EXTERNAL SCIENTIFIC REPORT

Transcription

1 Supporting Publications 2013:EN-466 Any enquiries related to this output should be addressed to Suggested citation: Madeleine Mattin, David Brodbelt, Claire Wylie, Marta Carbonell Antoñanzas, Laia Solano Gallego, Luis Espejo, Solenne Costard and Francisco Zagmutt; Data collection to characterise the impact of canine leishmaniosis and modelling of the role of animals in spreading Leishmania infantum within the European Union. Supporting Publications 2013:EN-466. [264 pp.]. Available online: European Food Safety Authority, 2013 EXTERNAL SCIENTIFIC REPORT Data collection to characterise the impact of canine leishmaniosis and modelling of the role of animals in spreading Leishmania infantum within the European Union Madeleine Mattin, David Brodbelt, Royal Veterinary College, UK Claire Wylie, Marta Carbonell Antoñanzas, Laia Solano Gallego, Universitat Autònoma de Barcelona, Spain Luis Espejo, Solenne Costard and Francisco Zagmutt, EpiX Analytics, USA ABSTRACT Canine leishmaniosis (CanL) is a major zoonosis transmitted by sand flies which can potentially cause severe fatal disease in humans and dogs. The disease is endemic in southern Europe and there are concerns about its introduction into the non-endemic areas of the European Union (EU). The objectives of this work were to investigate the efficacy of different control measures using a systematic review (SR), assess the role of dogs in the spread of Leishmania Infantum (L. infantum) in the EU and evaluate potential mitigation measures using simulation modelling, and evaluate the impact of the disease in endemic areas. The SR evaluated the effectiveness and safety of control measures for CanL using 23 clinical trials conducted on a total of 5861 dogs. Significant benefits were observed in the intervention groups for a number of vaccine types, deltamethrin collars, spot-ons, and prophylactic medication with domperidone. However, these results require confirmation in adequately powered well designed randomised controlled trials. The results of the stochastic simulation modelling suggest that there is a high risk of endemicity for disease-free EU regions with presence of a competent vector and where infected dogs are introduced. When evaluated individually and at a high level of implementation, the best mitigation measures in decreasing order of effectiveness were: repellent, vaccine, prophylactic medication and insecticide. Some combinations of mitigation measures showed a joint effect higher than the effect of individual measures. Test and exclusion of positive dogs moving into non-endemic areas was effective for low numbers of imports, but its benefits diminished as the number of imported dogs increased, given the low sensitivity of the diagnostic test used. The results of the impact assessment suggest clinical CanL is a common diagnosis in endemic parts of the EU. The estimated percentage of practice-attending dogs with a veterinary diagnosis of CanL by country was as follows: France 0.71%, Spain 3.71%, Portugal 2.92%, Italy 4.25% and Greece 7.80%. CanL can have a significant impact on the welfare of affected dogs and economic implications for owners, both for prevention and treatment. It is estimated that the use of preventative collars in these five countries cost dog owners 54 million in 2012 alone. The incidence of clinical human leishmaniosis is much lower than CanL, with reported data suggesting between 20 and 380 hospital admissions per country per year. Costs relating to hospital fees and time off work may amount to approximately 4.5million/year in Spain, Portugal, Italy and Greece. The current economic crisis is likely to have an adverse impact on the prevention and control of CanL. Royal Veterinary College, Universitat Autònoma de Barcelona and EpiX Analytics, 2013 KEY WORDS canine, leishmaniosis, Leishmania, prevention, epidemiology, systematic, impact

2 SUMMARY Canine leishmaniosis (CanL) is a major zoonosis which can potentially cause fatal disease in humans and dogs and is transmitted by sand flies. High rates of subclinical infection are common in endemic areas, and Leishmania Infantum (L. infantum) infection is spreading northward in Europe (Ready, 2010). The numbers of dogs travelling to southern Europe or imported from endemic areas, coupled with the existence and potential expansion of sand fly vector populations in Europe (Fischer et al., 2011) have raised serious concerns about the introduction of leishmaniosis into the non-endemic areas of Europe (Menn et al., 2010). The aim of this call was to carry out preparatory work in anticipation of a mandate on leishmaniosis from the European Commission. Systematic Review Canine leishmaniosis (CanL) is an important zoonotic disease; however the efficacy of classical control measures remains unclear. The objective of this review was to determine whether currently available preventative control interventions are efficacious at preventing natural L. infantum infection in dogs. Four electronic databases were searched; CAB Direct 2011, Web of Science 2011, U.S. National Library of Medicine 2011 (MEDLINE), Literatura Latino Americana e do Caribe em Ciências da Saúde (LILACS), along with eight sets of conference proceedings and the International Veterinary Information Service (IVIS) database. The reference lists of studies and review articles were checked. Dates were restricted from 1980 onwards and only publications in English, Italian, Portuguese, French or Spanish were included. Randomised controlled clinical trials (RCTs), non-randomised controlled clinical trials (OCTs), cohort studies and case-control studies that investigated prophylactic control measures for canine naturally-occurring L. infantum infection on parasite load, humoral or cellular immunity, infectivity, death, clinical disease and adverse effects were eligible for inclusion.two review authors independently assessed each study against the inclusion criteria, extracted data and assessed risk of bias for all included studies. The odds ratio (OR) and absolute risk reduction (ARR) for dichotomous outcomes and mean difference for continuous outcomes were calculated. The search yielded 937 articles, from which 84 full text articles for second stage screening. Twentythree eligible studies were included providing data on 5861 dogs, 12 studies on vaccinations (5 RCTs), 5 on repellent collars (2 RCTs), 4 on spot-on insecticides (2 RCTs) and 3 on prophylactic medication (all RCTs). All of the studies were considered to be at a high risk of bias, with the exception of two vaccination studies and one spot-on study which were considered to be at an unclear risk of bias. There are studies to support the use of control measures, in particular vaccines (200µg ALM protein, Leishmune, CaniLeish, LiESAp with MDP, and ALM with BCG), deltamethrin collars, 65% permethrin, 10% imidacloprid with 50% permethrin spot-ons and domperidone prophylactic medication as they tended to significantly reduce the proportion of dogs infected with L. infantum based on either parasitological or serological evidence. There were few studies reporting cellular immunity or the proportion of dead dogs or clinical illness following control intervention. Adverse effects were only reported in two vaccination studies and in one study on collars. There is some peer-reviewed evidence that control measures are effective in preventing CanL. However, current evidence is based on a small number of RCTs, all of which are either at high or unclear risk of bias. Well-designed and adequately powered clinical trials are needed to clearly establish efficacy of CanL control measures. Supporting publications 2013:EN-466 2

3 Simulation modelling The objectives of the simulation modelling were to estimate the probability of CanL endemicity in previously disease-free areas in the presence of a competent vector, following the introduction of infected dogs, and to evaluate the effectiveness of potential mitigation measures in reducing this potential endemicity. A review of existing simulation models on CanL was initially conducted to identify model(s) that could be used or adapted for the purpose of the project. Ten articles retained as potentially suitable were evaluated using a pre-defined approach, including modified Numeral Unit Spread Assessment Pedigree (NUSAP) data quality and assumption assessments for the three models deemed more appropriate. The structure and transmission dynamics of the model described by Dye (1996) were used to inform the development of parts of the simulation model used for the project, while its parameterisation was amended to reflect the context of the EU and incorporate data from the systematic review and impact assessment. A stochastic simulation model was developed to estimate the overall risk of CanL endemicity in a previously free area (P EndRegion ) in three steps. Firstly, the probability of introducing an infected dog into a non-endemic area was calculated for different pathways of introduction: dogs travelling to endemic areas with their owners (P Inf ), and dogs from endemic areas imported (e.g. commercial imports, adoptions, individual purchases) into non-endemic areas (P InfCA ). Then, following the introduction of an infected animal, CanL transmission in a previously disease-free area in the presence of a competent vector was simulated to estimate the probability of endemicity in an independent contact network of dogs (P End ). CanL transmission was simulated using individual-based stochastic continuous-time state transition modelling. Finally, endemicity in an area composed of one or more independent contact networks (P EndRegion ) was calculated using probability theory commonly used in import risk assessments. The impact of vaccination, prophylactic medication, repellent, insecticide, and diagnostic test and exclusion on the risk of CanL endemicity was assessed individually and in combination. In the absence of mitigation measures, following the introduction of infected dogs the probability of endemicity in at least one dog population in a previously free region with sand flies was high (92.2% to 100% when 10 to 100 dogs were imported from a endemic area). This probability was greatly influenced by seasonality (increased when travels of dogs and introductions occur when competent sand flies are active), the length of travel of susceptible dogs in endemic areas, and the number of dogs introduced in disease-free areas in the presence of competent vectors. When applied individually, mitigation measures reduced the risk of endemicity when used in a large percentage of the population. Testing and exclusion of positive dogs (i.e. dogs are tested before entering into or returning to a disease-free area, and positive dogs are not allowed to enter the area) had the highest impact on the overall probability of endemicity (68.7% mean reduction from scenario above) when a small number of dogs were imported, or returned from trips from endemic areas. However, this effect diminished as the number of imported dogs increased, because of the low sensitivity of the diagnostic tests used (52.6%, 95%CrI: 30.8%-74.0%). The most effective mitigation measure for P Inf and P End was the use of repellent, followed in decreasing order of effectiveness by: vaccination, prophylactic medication and insecticide use (99.6%, 88.5%, 75.6% and 55.1% mean reduction in P Inf from baseline scenario, and 85.9%, 60.6%, 22.4% and 0% mean reduction in P End from baseline scenario, respectively). Repellent was assumed to affect the biting rate of sand flies, and its effect was modelled as reducing a squared term used to calculate the vectorial capacity of sand flies as well as the number of infectious dogs from which infection can be acquired. The repellent s effectiveness might have been overestimated as the model does not consider that bites avoided by treated dogs may be redirected to other dogs or hosts. Conversely repellent effectiveness might have been underestimated due to the fact that repellents also have an insecticide effect, not represented in the model. The mechanism of action of vaccination and Supporting publications 2013:EN-466 3

4 prophylactic medication was modelled similarly, thus the difference in their impact on the probability of endemicity is due to the difference in their efficacy (mean: 86.4% vs. 74.1%, respectively, as derived from published RCTs). The modelled effect of insecticide decreased the density of sand flies, and thus the vectorial capacity. Some combinations of mitigation measures showed a joint effect higher than the effect of individual measures, even at lower levels of application. Notably, the use of vaccination (or prophylactic medication) and repellent reduced the probability of a dog returning infected from a travel to an endemic area by 94.9% (or 93.6%) on average when used on 80% of dogs, and by 67.9% (or 66.7%) on average when used on 40% of dogs. Impact assessment Leishmaniosis is a vector-borne zoonosis which can cause disease in companion animals, wildlife species and humans. The major route of transmission in both canine and human populations is via bites of phlebotomine sand flies. There is strong evidence that the dog is the main primary reservoir host of L. infantum. However, other domestic and non-domestic species may also be reservoir hosts. Clinical cases of CanL are common in endemic parts of the EU. Based on results of an online veterinary questionnaire, the estimated percentage of practice-attending dogs with a veterinary diagnosis of CanL by country was as follows: France 0.71%, Spain 3.71%, Portugal 2.92%, Italy 4.25% and Greece 7.80%. The economic burden of CanL includes the costs associated with prevention, treatment and mortality losses. The primary course of the CaniLeish vaccine was the most expensive control measure with an average cost of 150 for year one and thereafter for maintenance of immunity. Other preventative measures generally cost Treatment for a 30kg dog typically cost 200 if purchased from a veterinary clinic. Based on mean cost figures and data from a pharmaceutical company and the online veterinary questionnaire, it was estimated that dog owners in the five countries spent between million on vaccination, 35.9 million on testing and 44.8 million on treating CanL in Data from a pharmaceutical company were used to estimate that dog owners spent 54 million on deltamethrin collars in Mortality due to CanL can result in monetary losses due to the cost of euthanasia and replacement of a deceased pet, in addition to nonmonetary losses related to the impact on the human-animal bond. The indirect economic impacts on trade and tourism are likely to be minor, and based on available data were considered unreliable to quantify. The current economic crisis in the EU was likely to have a moderate to high impact on the use of prophylactic control measures and the diagnosis and treatment of dogs affected by CanL. Where data were available, the geographical distribution of clinical human leishmaniosis broadly followed that of CanL. Regional incidence of the disease in both species was highest in southeast France, Mediterranean Spain and the Italian regions of Sicily, Piemonte and Liguria. Although overall the incidence of clinically apparent human leishmaniosis was low, certain groups, such as infants and immunosuppressed people, were at a higher risk of disease and VL (visceral leishmaniosis) was reported to be fatal in up to 5% of human cases. Based on expert opinion, it was concluded that the impact on dog welfare and the likely prognosis of CanL depended on the clinical stage of the disease. Adverse effects associated with repellents and insecticides applied to the dog and immune stimulants were thought to have the lowest impact on dog welfare. Vector control was thought to be the most publicly acceptable control measure. Based on the level of web interest, temporal and regional public interest in leishmaniosis largely reflected the occurrence of human and canine cases in France, Spain and Italy. There was insufficient evidence that L. infantum has the potential to be used as a bioterrorism agent. The environmental impact of L. infantum and the control measures for CanL is likely to be minor, although the potential environmental impact of insecticide applied to the environment should be considered. Supporting publications 2013:EN-466 4

5 CONCLUSION There is some peer-reviewed evidence that control measures are effective in preventing CanL. However, current evidence is based on a small number of RCTs, all of which are either at high or unclear risk of bias. The results of the stochastic simulation modelling suggest that there is a high risk of endemicity for disease-free EU regions with presence of a competent vector and where infected dogs are introduced, but the prevalence in the dog meta-population in the region following disease establishment was not explicitly modelled, as disease was considered endemic when the disease remained in at least one dog contact network. When evaluated individually and at a high level of implementation, the best mitigation measures in decreasing order of effectiveness are: repellent, vaccine, prophylactic treatment and insecticide. Test and exclusion of positive dogs moving into nonendemic areas was effective for low numbers of imports, but its benefits diminished as the number of imported dogs increased, given the low sensitivity of the diagnostic test used. In endemic areas, CanL is a common diagnosis, which can have a significant impact on the welfare of affected dogs. Although less common than CanL, human leishmaniosis is a public health concern in endemic areas. CanL has a considerable economic impact in the EU, largely at the expense of dog owners. The current economic crisis may have an adverse impact on the prevention and control of CanL. Supporting publications 2013:EN-466 5

6 TABLE OF CONTENTS Abstract... 1 Summary... 2 Conclusion... 5 Table of contents... 6 Background as provided by EFSA Terms of reference as provided by EFSA Systematic Review Introduction and Objectives Description of the disease Description of the interventions Vaccination Collars Spot-ons Prophylactic medication Why is it important to do this review Objectives Materials and Methods Criteria for considering studies for this review Types of studies Types of subjects Types of interventions Types of outcome measures Primary outcomes Secondary outcomes Search methods for identification of studies Electronic searches Searching other resources Data collection and analysis Selection of studies Data extraction and management Assessment of risk of bias in included studies Measures of treatment effect Unit of analysis issues Dealing with missing data Assessment of heterogeneity Assessment of reporting bias Data synthesis Subgroup analysis and investigation of heterogeneity Sensitivity analysis Results Results of the search Description of studies Vaccination Description of included studies Risk of bias Effect of interventions Collars Description of included studies Risk of bias Effect of interventions Supporting publications 2013:EN-466 6

7 Spot-ons Description of included studies Risk of bias Effect of interventions Prophylactic medications Description of included studies Risk of bias Effect of interventions Overall summary Conclusions Summary of main results Overall completeness and applicability of evidence Quality of the evidence Potential biases in the review process Agreements and disagreements with other studies or reviews Recommendations Implications for practice Implications for research Assessment of the role of animals in the spread of L. infantum within the European Union Introduction and Objectives Review and evaluation of existing simulation models Introduction Materials and Methods Review Evaluation of existing simulation models Results Discussion Simulation modelling Materials and Methods Modelling the probability of endemicity in a disease free area following the introduction of an infected dog Vectorial capacity Evaluating mitigation measures Scenario analysis Sensitivity analysis Model main assumptions Results Sensitivity analysis Probability of one dog returning infected from a trip to an endemic area (P Inf ) Probability of endemicity within a contact network of dogs (P End ) Probability of endemicity in a region (P EndRegion ) Discussion Conclusions Impact assessment Introduction and Objectives Characteristics of the Studied Countries Materials and Methods Data collection Data analysis Results Epidemiological profile of Canine Leishmaniosis Materials and Methods Supporting publications 2013:EN-466 7

8 Data collection Data analysis Results Animal species affected by clinical disease Animal and human hosts Transmission Epidemiology in Humans Conclusions Local Epidemiology Materials and Methods Data collection Data analysis Results Benchmarking data (Panelvet, France) Veterinary questionnaire Shelter data Laboratory data National Reference Centre for Leishmania, Italy (C.Re.Na.L.) National Reference Centre for Leishmania, Greece (Hellenic Pasteur Institute) Published estimates of the prevalence of canine leishmaniosis in Europe Factors which may influence the future epidemiology of leishmaniosis Conclusions Economic Impact Materials and Methods Data collection Data analysis Results Direct economic losses Mortality Morbidity Indirect economic losses Trade and movement restrictions Impact on tourism Impact of the current economic crisis on the veterinary care of dogs affected by leishmaniosis Conclusions Human health impact Materials and Methods Data collection Data analysis Results Annual number of human cases Reported cases Hospital data Laboratory data (Hellenic Pasteur Institute, Greece) Typical and maximum severity Economic impact of human disease Cases attributable to animals and animal-human interactions Conclusions Societal Impact Materials and Methods Data collection Supporting publications 2013:EN-466 8

9 Data analysis Results Animal welfare Crisis generation potential Bioterrorism potential Conclusions Environmental impact Materials and Methods Data collection Data analysis Results Impact on of the disease on biodiversity Environmental impact of mortality due to leishmaniosis Environmental impact of control measures Conclusions Discussion Conclusions and Recommendations Appendix/Appendices A. Results of search strategies B. Reasons for exclusion and short description of the 64 studies excluded at the second screening phase of the SR C. Scoring matrix used to evaluate time limitation factors of published models identified through the literature search D. Pedigree matrix used for the data quality assessment of the models identified as most suitable for the simulation modelling component of the project: E. Pedigree matrix used for the assumption assessment of the models identified as most suitable for the simulation modelling component of the project: F. Model characteristics: disease and leishmania species modelled, and geographical area of relevance of the models: G. Model characteristics: type of models and populations of hosts and vectors represented: H. Model characteristics: Concept used to model transmission dynamics and model parameters:. 189 I. Time limitation scores of models identified as potentially suitable for the purpose of the project:190 J. Narratives of the ten published models as potentially suitable for the simulation modelling component of the project: K. Data quality assessment scores for parameters used in the human compartment of the simulation models: L. Data quality assessment scores for parameters used in the dog compartment of the simulation models: M. Data quality assessment scores for parameters used in the sand fly compartment of the simulation models: N. Assumption assessment scores: O. Assumption assessment results: for each assumption, score for each criteria and sum of scores 202 P. Structure of the simulation model Q. Derivation of parameters used in the simulation model R. Phylum criteria (Phylum, 2010) S. Location of Panelvet practices contributing data on canine leishmaniosis, France T. Results of online veterinary questionnaire U. Results of online questionnaire for dog shelters V. Laboratory Data, C.Re.Na.L, Italy (Vitale, 2013) W. Laboratory data, Hellenic Pasteur Institute, Greece (Dotsika, 2013) X. Published estimates of canine leishmaniosis infection and disease in France, Spain, Portugal, Italy and Greece Supporting publications 2013:EN-466 9

10 Y. Human leishmaniosis data Z. Maps of standardised morbidity ratios for hospitalised cases of human leishmaniosis in Spain and Italy, AA. Google Trends: regional web interest References Glossary and abbreviations Glossary Abbreviations Supporting publications 2013:EN

11 BACKGROUND AS PROVIDED BY EFSA Impact, modelling and control of canine leishmaniosis in the EU Canine leishmaniosis was listed in the 2011 EFSA Management Plan as a topic to be considered by the AHAW Panel. Leishmaniosis is a non-food-borne zoonosis with sand flies (Phlebotomus spp.) as the vector. Only two transmission cycles are known to be endemic in the European Union (EU): i) cutaneous leishmaniosis caused by L. tropica, which is usually regarded as an anthroponosis, although the parasite was isolated also from rodents and dogs and which occurs sporadically in Greece and ii) zoonotic, cutaneous and visceral leishmaniosis caused by L. infantum which occurs throughout the Mediterranean region with domestic dogs as main reservoir hosts. Recently high prevalences of L. infantum infection in apparently healthy wild carnivores were found in southern Spain, suggesting also the existence of a sylvatic cycle independent of dogs. Factors that may trigger changes in the distribution of the disease include climate and environment changes, changes in vector capacity and competence, importation or dispersal of vectors and reservoir hosts and travelling dogs with their owners. Currently, two questions need to be addressed that are: the impact of the disease and the role of animals in spreading L. infantum. Canine leishmaniosis (CanL) has become more apparent in northern latitudes where sand fly vectors are either absent or present in very low densities (Ready, 2010). For example, in the last decade, several publications have reported cases of L. infantum in Germany (Harms et al., 2003, Naucke and Schmitt, 2004, Menn et al., 2010, Mettler et al., 2005b). Most of these L. infantum infections might be explained by dog importation from, or travel to and back from, endemic regions, potentially followed by vertical transmission from bitch to pup or horizontal transmission by biting hounds. Canine leishmaniosis has a high prevalence of infection in endemic areas, involving as much as 63% 80% of the population (Berrahal et al., 1996, Solano-Gallego et al., 2001b) even if it is accompanied by a lower rate of apparent clinical disease. When present, clinical signs in dogs may manifest with variable degrees of severity. Substantial research published recently on the pathogenesis of CanL and immune responses during infection has contributed considerably to the current understanding of this complex zoonosis and its epidemiology (Quinnell et al., 2001, Solano-Gallego et al., 2001a, Oliva et al., 2006). These aspects were recently reviewed by a Scientific Report submitted to EFSA (EFSA, in press). Current studies have revealed new facts about the prevalence and spread of infection (e.g. data made available by EU-funded projects such as EDEN, EDENext, VBORNET, LeishRisk or public and veterinary surveillance programmes) and established novel concepts on its evolution and dynamics (Ferroglio et al., 2005, Leontides et al., 2002, Oliva et al., 2006, Solano-Gallego et al., 2001a). These new insights may help in modelling the spread of the disease and to assess its impact and evaluate efforts to prevent and control the disease and its spread into human populations. TERMS OF REFERENCE AS PROVIDED BY EFSA This contract/grant was awarded by EFSA to: Royal Veterinary College, UK Contractor/Beneficiary: Royal Veterinary College, UK; UAB Barcelona; EpiX analytics, USA Contract/grant title: Data collection to characterise the impact of canine leishmaniosis and modelling the role of animals in spreading Leishmania infantum within the European Union Contract/grant number: CT/EFSA/AHAW/2012/02 Supporting publications 2013:EN

12 1. Systematic Review INTRODUCTION AND OBJECTIVES 1.1. Description of the disease CanL has a wide spectrum of clinical manifestations, classically displaying a chronic disease course with periods of remission followed by relapse (Solano-Gallego et al., 2009). Clinical signs commonly include skin lesions such as non-pruritic exfoliative dermatitis ± alopecia, particularly affecting the pinnae and peri-ocular regions, or ulcerative, papular, nodular or mucocutaneous proliferative dermatitis. Additional clinical signs include generalised lymphadenomegaly, loss of body weight, and ocular lesions such as blepharitis and conjunctivitis. Affected dogs will also display a range of clinicopathological findings, with hyperproteinaemia, hyperglobulinaemia, hypoalbuminaemia, decreased albumin/globulin ratio, proteinuria, and non-regenerative anaemia occurring in more than 50% of sick dogs (Paradies et al., 2010, Baneth et al., 2008). The diagnosis of CanL is challenging as there are no methods capable of detecting infection in all infected animals (Solano-Gallego et al., 2009). Diagnosis should be performed based on physical and clinicopathological manifestations and by confirmation of infection, preferably using molecular and serological techniques. Examination of biopsy samples from the skin, lymph nodes or bone marrow commonly reveals the organism in infected dogs (Solano-Gallego et al., 2009). Real-time polymerase chain reactivity (PCR) is considered to be the most sensitive molecular technique with lymph node, bone marrow, skin or spleen tissue preferable over blood, buffy coat or urine due to improved sensitivity (Maia et al., 2009). Clinical disease is usually associated with high serum antibody levels (Pasa et al., 2005). Seropositivity is found in % of dogs with clinical signs and/or clinicopathological abnormalities consistent with CanL (Solano-Gallego et al., 2009). ELISA and immunofluorescent antibody tests (IFAT) have similar sensitivities and specificities and are considered effective in clinical illness diagnosis. The prognosis for treated dogs is reported to be good to excellent, although many dogs remain infected maintaining subclinical infection (Roura et al., 2012). The prognosis for dogs with renal disease or clinically ill untreated dogs is poor, with many animals dying of renal failure and immune-mediated diseases (Solano-Gallego et al., 2011) Description of the interventions Vaccination Many Leishmania antigens have been identified as potential vaccine candidates. However, very few have been tested in field trials and only three second-generation canine vaccines have been registered (Palatnik-de-Sousa, 2012). The FML-saponin vaccine (Leishmune ) was licensed in Brazil in 2003, where a second vaccine, Leish-Tec, a recombinant A2-antigen of Leishmania amastigotes adjuvanted by saponin, has also been licensed since In early 2011, CaniLeish, a formulation related to LiESAp, composed of 54-kDa excreted protein of L. infantum with MDP was the first vaccine for CanL licensed for use in Europe. Vaccination is designed to induce a strong protective cellular immune response against the specific antigens of CanL, with an effective immune response mounted approximately four weeks after the final vaccination (Palatnik-de-Sousa, 2012). Supporting publications 2013:EN

13 Collars Impact, modelling and control of canine leishmaniosis in the EU Vector control measures for CanL consist of products applied at the dog-level to prevent insect bites, including insecticides and repellents (Maroli et al., 2010). Deltamethrin-impregnated collars control sand fly feeding, and are commonly used for CanL prevention in Europe (Maroli et al., 2010). Deltamethrin is slowly absorbed into the skin of the animal where it completely covers the dog and retains its activity for approximately 5-6 months. Deltamethrin has a potent anti-feeding and insecticidal effect against sand flies (Killick-Kendrick et al., 1997). Protective collars should be applied for at least 1-2 weeks before travelling to endemic areas, or at the beginning of the vector season for dogs living in endemic regions, to ensure enough time for the deltamethrin to be absorbed and distributed over the dog s skin Spot-ons Directly applied insecticides are available as spot-on formulations for the protection of individual dogs and are also commonly used for CanL prevention in Europe (Maroli et al., 2010). Permethrin/imidacloprid spot-on formulations provide repellent activity against sand flies for up to three weeks, and the products can be reapplied every three weeks for as long as necessary (Miro et al., 2007a). Spot-on formulations should be applied at least two days before travelling to ensure animals are protected prior to exposure. Mass vector killing, carried out for some vector-borne diseases, is not considered an appropriate control measure for sand flies as they tend to inhabit focal areas rather than having a widespread distribution (Gálvez et al., 2010), and breeding sites remains unknown (Feliciangeli, 2004). Two further environmental control measures are used in Asian and Latin American countries for the prevention of human leishmaniosis (house spraying and bed nets) (Picado et al., 2012, Ostyn et al., 2008), but are not applicable control measures for the canine disease Prophylactic medication Two types of prophylactic mediation are available for CanL. Leisguard is a domperidone-based oral solution recently marketed for both prevention and treatment in several European countries (Gomez- Ochoa et al., 2009). Domperidone is a gastric prokinetic and anti-emetic drug that acts as a dopamine D2 receptor antagonist, resulting in the release of serotonin, which in turn stimulates prolactin production (Gomez-Ochoa et al., 2009). Prolactin has a central role in the immune reaction, although its mechanism of action is largely unknown. Allopurinol is widely used for the treatment of leishmaniosis in dogs with potential prophylactic efficacy (Noli and Auxilia, 2005). Allopurinol functions as an alternative substrate for hypoxanthine guanine phosphoribosyl transferase enzyme, allowing incorporation of allopurinol riboside into RNA, which leads to the inhibition of protein synthesis in the parasite (Denerolle and Bourdoiseau, 1999) Why is it important to do this review Though there are numerous papers evaluating individual control measures there are limited reports that objectively evaluate the different methods available to European member states. Two previously published systematic reviews either focused on control measures in Latin America (Romero and Boelaert, 2010) or only included vector control measures such as insecticides/repellents (Noli and Auxilia, 2005). There is, therefore, an identified gap in knowledge and justification for this review in order to place current evidence in the context of available interventions. Supporting publications 2013:EN

14 1.4. Objectives Impact, modelling and control of canine leishmaniosis in the EU The objective of this review was to determine whether currently available preventive control interventions are efficacious at preventing natural L. infantum infection in dogs. The review questions were what is the effect of different prophylactic control measures (i.e. vaccination, vector control measures, prophylactic medication, and other preventative control measures of L. infantum infection in dogs) for canine naturally-occurring L. infantum infection on: a) Canine L. infantum parasite load? b) Canine L. infantum humoral or cellular immunity? c) Canine L. infantum infectivity? d) Canine L. infantum clinical disease, confirmed by specific parasitological diagnostic methods? Supporting publications 2013:EN

15 MATERIALS AND METHODS 1.5. Criteria for considering studies for this review Types of studies Randomised controlled trials (RCTs), non-randomised clinical trials (OCTs), cohort studies and casecontrol studies that investigated prophylactic control measures for naturally-occurring L. infantum infection on parasite load, humoral (serology) or cellular immunity, infectivity, death, clinical disease and/or adverse effects in dogs were eligible for inclusion Types of subjects Studies that included dogs susceptible to naturally-occurring L. infantum infection but non-infected at the start of the study were eligible for inclusion. Serology must have been undertaken as a minimum technique to establish non-infection. Dogs with other infections were also included Types of interventions Any intervention/exposure, or combination of interventions/exposures, for the prevention of L. infantum infection were included if they were compared to placebo or control intervention Types of outcome measures Primary outcomes o Proportion of dogs infected with L. infantum based on serology or parasite detection Secondary outcomes 1. Proportion of dogs infected with L. infantum based on parasite detection 2. Proportion of dogs infected with L. infantum infection based on PCR detection 3. Quantitative PCR: measured as the mean value and standard deviation (SD) depending on the calculation performed by the paper 4. Proportion of dogs infected with L. infantum based on serological detection 5. Quantitative serology: measured as the geometric mean of the titers (based on IFA/DAT) or the arithmetic mean of the optical density/absorbance in nm (based on ELISA) and SD 6. Proportion of dogs infected with L. infantum based on cellular immunity tests 7. Size of the Delayed Type Hypersensitivity (DTH) reaction 8. Proportion of dogs that died infected with L. infantum 9. Proportion of clinically ill dogs infected with L. infantum 10. Adverse events related to the intervention Supporting publications 2013:EN

16 1.6. Search methods for identification of studies Electronic searches Four electronic databases were searched; 1. CAB Direct Web of Science 2011 (WOS) 3. U.S. National Library of Medicine 2011 (MEDLINE) 4. Literatura Latino Americana e do Caribe em Ciências da Saúde (LILACS) The following search strategy in CAB Direct was used: ab:(dog OR dogs OR canis OR canid OR canids OR canidae OR canine) OR title:(dog OR dogs OR canis OR canid OR canids OR canidae OR canine) OR subject:(dog OR canidae) AND title:(leishmania OR leishmaniosis OR leishmaniasis) OR ab:(leishmania OR leishmaniosis OR leishmaniasis) OR subject:(leishmania OR leishmaniosis) AND title:(control OR preventive OR preventative OR prophylactic OR prophylaxis OR prevention OR insecticide OR insecticides OR vaccine OR vaccines OR vaccination OR spot-on OR spot-ons OR collar OR collars OR immune stimulant OR immune stimulants) OR ab:(control OR preventive OR preventative OR prophylactic OR prophylaxis OR prevention OR insecticide OR insecticides OR vaccine OR vaccines OR vaccination OR spot-on OR spot-ons OR collar OR collars OR immune stimulant OR immune stimulants) OR subject:(insecticides OR "insect repellents" OR vaccination) The search strategies for WOS, MEDLINE and LILACS can be found in Appendix A. Dates were restricted from 1980 onwards and only publications in English, Italian, Portuguese, French or Spanish were included Searching other resources Eight sets of conference proceedings (1 st -4 th World Leishmaniasis Congress [WorldLeish], 1 st and 2 nd International Canine Leishmaniasis Forum, 1 st and 2 nd Italian Companion Animal Veterinary Association International Congress of Canine Leishmaniasis) and the International Veterinary Information Service (IVIS) database were also checked. The reference lists of studies and review articles were checked. Study authors were not contacted for additional information. No further grey literature was searched due to the limited time and resources available for the current review. Supporting publications 2013:EN

17 1.7. Data collection and analysis Selection of studies Two reviewers independently assessed each reference identified by the search to check its eligibility. There was good agreement between the review authors following discussion and it was therefore not necessary to consult a third review author to obtain consensus. Those references which appeared to meet the inclusion criteria were retrieved in full and further assessed independently by the same two reviewers. At this stage a third reviewer was consulted to resolve difference of opinion on whether two studies should be included Data extraction and management Data extraction was carried out by two reviewers independently using pre-specified forms, modified for each study design, regarding study characteristics, population characteristics, pre-trial diagnostics, intervention/comparator, outcomes and interim results with DistillerSR 1. No attempts were made to contact the study authors if data were incomplete. Data from studies published in duplicate were included only once. Where studies were available in full paper and abstract form, the full paper version was chosen. All discrepancies were resolved by discussion Assessment of risk of bias in included studies Two review authors independently assessed the risk of bias of the eligible studies. All discrepancies were resolved by discussion. The risk of bias assessment was based on the guidance in the Cochrane Handbook for Systematic Reviews of Interventions (Higgins and Green, 2008) and included assessment of the methods of randomisation and allocation concealment, level of blinding, presence of incomplete data, selective reporting, study duration (12 months or greater to allow investigation of a complete transmission season and for the potential subsequent development of infection), acknowledgement of financial support and other sources of bias. This was the only assessment tool used as only RCT and OCT study designs were identified. Each study was classified as a high, unclear or low risk of bias such that if one domain was not met the trial was considered a high risk of bias, and if one domain was unclear the trial was considered to have an unclear risk of bias. The reasons for each judgment from details provided in the study report were described Measures of treatment effect For categorical outcome measures the odds ratio (OR), using asymptotic normal calculation for the log of the OR, absolute risk reduction (ARR) and their corresponding 95% confidence interval (95% CI) were calculated for each measure. The baseline for all the ORs was the controls, i.e. OR>1 signified that the intervention had a lesser protective effect that control, with the exception of the cellular immunity tests for which immunity was protective. For continuous data the mean difference and the standard error (SE) were calculated for both the intervention and comparator groups Unit of analysis issues A single measurement for each outcome was collected and analysed from each dog Dealing with missing data Study authors were not contacted to obtain missing data due to time constraints. 1 DistillerSR, Evidence Partners, Ottawa, Canada Supporting publications 2013:EN

18 Assessment of heterogeneity Impact, modelling and control of canine leishmaniosis in the EU Assessment of the impact of statistical heterogeneity using the I 2 statistic, which describes the percentage of total variation across studies that is due to heterogeneity rather than sampling error (Higgins et al., 2011), was pre-specified in the protocol. Heterogeneity would be considered high when I 2 75%. Studies with statistical heterogeneity > 75% would not be pooled. For levels of I 2 less than 50% a fixed-effect model would be applied, for levels of I 2 between 50-75% a random-effects model would be applied. Where statistical heterogeneity was high (> 75%) or where clinical heterogeneity was identified it was planned to check the data or explore heterogeneity Assessment of reporting bias Production of a funnel plot to assess the review for publication bias if the number of clinical trials identified for inclusion in meta-analysis was more than 10 (Sterne et al., 2011) was pre-specified in the protocol Data synthesis Pooling of the results using meta-analysis was pre-specified in the protocol if the included studies were considered to be sufficiently comparable. It was pre-specified that results from RCT and non- RCT would not be pooled, and that meta-analysis would not be undertaken on poor-quality studies Subgroup analysis and investigation of heterogeneity Potential sources of clinical heterogeneity were considered to be: Dog characteristics: for which subgroup analyses were planned to compare age (dogs < 2 years old and those 2 years old) and breed (purebred v mixed breed), data description was planned for type (pet/stray), gender, neutered status and housing, and for which source (rehoming v private practice) and primary use were to be ignored. Intervention: for which heterogeneity of type was to be eliminated by examining each intervention and corresponding control measure individually and dose, number and route of administration were to be described. Diagnostic methods of infection prior to intervention: for which sensitivity analysis was planned to compare only the diagnoses made on the basis of serology, and tissue used and test sensitivity and specificity were to be described. Diagnostic methods of infection after intervention: for which method type, tissue used and test sensitivity and specificity were to be described Sensitivity analysis Sensitivity analysis was planned to compare only the diagnoses prior to intervention made on the basis of serology, only studies of suitable duration (at least one year of follow-up to allow for examination of the transmission season) and those considered to be high quality. Supporting publications 2013:EN

19 RESULTS 1.8. Results of the search The search resulted in 937 publications. After assessing for relevance titles and abstracts 84 potentially relevant studies were retrieved in full text (Appendix B). Twenty-three studies were included and 61 were excluded (Figure 1). Figure 1: Flow chart illustrating the results of the selection process of the studies modified from Moher et al. (2009). Supporting publications 2013:EN

20 Excluded studies Sixty-one studies were excluded for the following reasons (Appendix B). Six studies were not original scientific articles 17 studies were part of another study written up in full elsewhere One study did not include a domestic dog species One study did not assess leishmaniosis Two studies did not assess L. infantum infection Four studies were not on naturally occurring disease 14 studies were neither RCTs, OCTs, or observational analytic epidemiologic studies Five studies did not include a definitive diagnosis Six studies did not contain information on preventive control interventions Six studies did not provide results regarding the efficacy of the interventions to provide data for the outcomes of this review 1.9. Description of studies Descriptive data were extracted from the 23 included studies. The included study designs represented were either RCTs or OCTs, with no case-control or cohort studies identified. Each of the included studies was published in English. Overall, data on 5861 dogs were included in the review consisting of 12 studies on vaccinations (5 RCTs), 5 on collars (2 RCTs), 4 on spot-ons (2 RCTs) and 3 on prophylactic medication (all RCTs). Baseline characteristics of the enrolled dogs were reported to some extent for each of the studies. Clear presentation of statistical comparison of intervention groups with respect to signalment and baseline disease characteristics pre-treatment were rarely reported. Pooling of outcome data from multiple studies investigating similar treatment interventions was not possible owing to the heterogeneity of study designs (i.e. RCTs and OCTs) and reported outcome measures. The following results are presented by intervention type. Supporting publications 2013:EN

21 Vaccination Description of included studies A number of different vaccines were evaluated in the vaccination studies (Table 1, Table 2 and Table 3). Three OCTs in Brazil examined Leishmune vaccine; Lima et al. (2010) conducted a 12 month evaluation of the PCR status of pet dogs in comparison to untreated controls. Borja-Cabrera et al. (2008) carried out a 24 month study evaluating clinical illness in an uncharacterised population of dogs in comparison to controls treated with sterile saline. Nogueira et al. (2005b) performed a 12 month comparison to BCG in adult pet dogs examined by PCR, cytology and ELISA serology. Gradoni et al. (2005) examined MML and MPL vaccine using two different adjuvants, MPL-SE and Adjuprime. Each of the other studies examined unique vaccine types as described: da Silva et al. (2000) conducted a 24-month OCT of FML vaccine. Mohebali et al. (2004) examined ALM through a 16-month RCT. Oliva et al. (2012) carried out an OCT of CaniLeish. Lemesre et al. (2007) conducted a 24-month RCT of LiESAp and MDP. Mohebali et al. (1999) examined ALi and BCG vaccine and ALM and BCG through a 12 month RCT. Borja-Cabrera et al. (2002) conducted a 41- month long OCT of FML and QuilA. Dunan et al. (1989a) examined lyophilized protein through a 24-month long RCT. Finally, Genaro et al. (1996) conducted an 18-month RCT of Leishvacin. Supporting publications 2013:EN

22 Table 1: The signalment characteristics for studies on vaccination. Impact, modelling and control of canine leishmaniosis in the EU Study Publication Affiliation first author (da Silva et al., 2000) (Gradoni et al., 2005) (Mohebali et al., 2004) (Lima et al., 2010) (Oliva et al., 2012) (Borja- Cabrera et al., 2008) (Lemesre et al., 2007) (Nogueira et al., 2005b) (Mohebali et al., 1999) (Borja- Cabrera et al., 2002) (Dunan et al., 1989a) (Genaro et al., 1996) Full paper Universidade Federal do Rio de Janeiro Full paper Istituto Superiore di Sanità Full paper Tehran University of Medical Sciences Full paper Universidade Estadual Paulista Abstract University Federico II Naples Full paper Universidade Federal do Rio de Janeiro Full paper Full paper Full paper Institut de Recherche pour le Développement Universidade Estadual Paulista Tehran University of Medical Sciences Full paper Universidade Federal do Rio de Janeiro Full paper Abstract Pitié-Salpètrière School of Medicine Universidade de São Paulo Country Type Age Gender Neutered Housing Breed Brazil Pets only 4 months old Mixed Missing Not specified, however 'domestic dogs' so probably a combination of homes and outdoors 97.5% mongrels Italy Experimental dogs Missing Mixed Missing Kennels Beagle Iran Pets only >3 months Mixed Missing Not specified, however 'owned' Missing so probably combination of homes and outdoors Brazil Pets only Missing Mixed Missing Not specified, however 'owned' Missing so probably combination of homes and outdoors Italy and Missing Missing Missing Missing Kennels Missing Spain Brazil Missing Missing Missing Missing Missing Missing France Hunting dogs, some guard dogs, dogs Adults and puppies <6 Mixed Missing Outdoors Missing from breeding months facilities Brazil Pets only Adult Missing Missing Not specified, however owned Missing dogs, 'co-habitated in the same residence' Iran Pets only Missing Missing Missing Not specified, however 'domestic Missing dogs' so probably combination of homes and outdoors Brazil Pets only Missing Missing Missing Missing Missing France Pets only 8 months Missing Missing Missing Missing Brazil Missing >6 months Missing Missing Missing Missing Supporting publications 2013:EN by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender considered as an output adopted by the Authority. The European food Safety Authority reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

23 Table 2: The intervention characteristics for studies on vaccination. Impact, modelling and control of canine leishmaniosis in the EU Study Intervention Dose of treatment Route of administration (da Silva et al., 2000) (Gradoni et MML with al., 2005) a MPL (Gradoni et MML with al., 2005) b Adjuprime (Mohebali et al., 2004) (Lima et al., 2010) (Oliva et al., 2012) (Borja- Cabrera et al., 2008) (Lemesre et al., 2007) FML 1.5mg lyophilized FML antigen reconstituted in 1ml NaCl 0.9% sterile saline solution 45g/dose MML plus 50g/dose MPL -SE 45g/dose MML plus 1mg/dose Adjuprime Number of treatments Comparator Study duration in months Study design S/C 4 1ml sterile saline 24 OCT S/C 6 1ml/dose sterile saline 24 OCT S/C 6 0.1ml normal saline 24 OCT ALM 200µg protein I/D 1 Placebo group 16 RCT Leishmune Missing S/C 3 Untreated controls 12 OCT CaniLeish Missing Missing Missing Untreated controls 24 OCT Leishmune Missing S/C 4 1ml sterile saline 24 OCT LiESAp MDP with Missing S/C 3 Untreated controls 24 RCT ALM with (Nogueira et Leishmune Missing S/C 3 BCG (400µgram/dose) 12 OCT al., 2005b) (Mohebali et ALi with 1mg ALi protein and 400µg I/D 2 Normal saline 12 RCT al., 1999) a BCG BCG, autoclaved 1mg protein promastigotes of L. infantum (Mohebali et 1mg ALM protein and 400µg I/D 2 Normal saline 12 RCT al., 1999) b BCG BCG, promastigotes of L. major Supporting publications 2013:EN by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender considered as an output adopted by the Authority. The European food Safety Authority reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

24 (Borja- Cabrera et al., 2002) (Dunan et al., 1989a) (Genaro et al., 1996) FML QuilA Lyophilized protein Leishvacin with (a): S/C = Subcutaneous injection (b): I/D = Intra-dermal injection (c): OCT = Non-randomised controlled clinical trial (d): RCT = Randomised controlled clinical trial Impact, modelling and control of canine leishmaniosis in the EU Missing S/C 4 Saline 41 OCT Missing S/C µg sodium dodecyl sulphate and 20µg lactose. Reconstituted with 0.5ml saline and administered with 100µg MDP adjuvant First dose 600µg protein and 500µg BCG, 2nd/3rd/annual doses 600µg protein and 100µg BCG I/D 5 Magnesium hydroxide 10% in a solution of PBS buffer, ph 7.2, merthiolated 24 RCT 18 RCT Supporting publications 2013:EN by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender considered as an output adopted by the Authority. The European food Safety Authority reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

25 Diagnostic method Tissue Sensitivity Specificity Diagnostic method Tissue Sensitivity Specificity Methods of serology Antigens Antibodies Methods of parasite detection Case definition provided by the paper Table 3: The diagnostic methods for studies on vaccination. Impact, modelling and control of canine leishmaniosis in the EU Study Pre-trial diagnostics Post-trial diagnostics (da Silva et al., 2000) (Gradoni et al., 2005) a and b (Mohebali et al., 2004) Serology and PE Serology only Serology, PE and Leishmanin Skin Test in Serum and PE Serum only Serum, PE and skin Missing Missing Serology, PE, parasite detection and cellular immunity Missing Missing Serology, laboratory testing, parasite detection and cellular immunity Missing Missing Serology and PE Whole blood, bone marrow, lymph nodes, spleen, serum, liver and kidneys Whole blood, bone marrow, lymph nodes, serum Serum and PE Missing Missing ELISA only Missing Missing ELISA and IFA Missing Missing ELISA and DAT Supporting publications 2013:EN by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender considered as an output adopted by the Authority. The European food Safety Authority reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors. FML K39, L. infantum promastigotes of the WHO reference strain for L. infantum Promastigotes L. infantum Anti- Leishmania donovani Antileishmanial IgG, serum IgG antibodies against MML antigen Anti- Leishmania antibodies PCR and Clinical signs or deaths due to cytology kala-azar: loss of weight, cachexia, alopecia, onychogryphosis, apathy, anorexia, increase of popliteal lymph node size and ulcerative skin lesions. Considered symptomatic submitted to autopsy and to parasitological evaluation. Liver and spleen weighed after autopsy of symptomatic dogs to determine hepato-splenomegaly PCR, Clinical assessment by accurate cytology inspection for the presence of and seven signs attributable to culture Leishmania infection (dermatitis, skin ulcers, alopecia, ocular lesions, lymph adenopathy, onychogryphosis, weight loss) and by the evaluation of laboratory data (full blood count, total proteins and albumin/globulin ratio). Animals were scored for clinical and laboratory signs on a scale from 0 to 2/3, and the scores added up to give a clinical score NA Physically examined, signs including hair shedding, desquamative or ulcerative dermatitis, lymphadenopathy,

26 18% of dogs and cachexia (Lima et al., 2010) (Oliva et al., 2012) (Borja- Cabrera et al., 2008) (Lemesre et al., 2007) (Nogueira et al., 2005b) (Mohebali et al., 1999) a and b Serology, PE and parasite detection Serum, whole blood and bone marrow Missing Missing Parasite detection only Missing NA Missing Missing Serology, PE, laboratory testing and parasite detection Serology and PE Serology, PE and parasite detection Serology and PE Serology only Serum and PE Serum and bone marrow Serum and PE Serum only Whole blood and bone marrow Whole blood, bone marrow, lymph nodes and serum NA NA NA NA NA PCR only NA Missing Missing Missing Missing Missing PCR and culture Clinicopathological signs Missing Missing PE only PE only NA NA NA NA NA NA Alopecia, onychogryphosis, cachexia, anorexia, apathy, disseminated ulcers, skin lesions, keratitis, renal failure, loss of weight, lymph node enlargement or diarrhoea were recorded as visceral leishmaniasis Missing Missing Serology, PE and parasite detection Serology, PE and parasite detection Missing Missing Serology only Serum and bone marrow Whole blood, lymph nodes, skin and serum Serum only Missing Missing IFA only ELISA only Missing Missing ELISA, IFA and DAT Missing FML Missing Anti- Leishmania antibodies Anti-FML antibodies, anti-l. chagasi antibodies Anti- Leishmania antibodies PCR and culture PCR and cytology NA Appearance of classical clinical signs such as weight loss, cachexia, apathy, anorexia, uveitis, alopecia, exfoliative dermatitis, ulcerative skin lesions, mucosa paleness, onychogryphosis, presence of popliteal lymphoadenopathies and renal failure Clinical signs for visceral leishmaniasis, kala-azar symptoms NA Supporting publications 2013:EN by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender considered as an output adopted by the Authority. The European food Safety Authority reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

27 (Borja- Cabrera et al., 2002) (Dunan et al., 1989a) (Genaro et al., 1996) Serology and PE Serology and PE Serology only (a): PE = Physical examination Serum and PE Serum and PE Serum only Missing Missing Serology, PE, parasite detection and cellular immunity Missing Missing Serology, PE and parasite detection Missing Missing Serology, PE and parasite detection Whole blood, bone marrow, lymph nodes, spleen, serum, liver and kidneys Whole blood, bone marrow, lymph nodes and serum Serum and whole blood Missing Missing ELISA only Missing Missing IFA only Missing Missing ELISA and IFA FML Missing PCR and cytology Acetone-fixed promastigotes from a canine strain of L. infantum harvested from cultures in stationary phase Anti- Leishmania antibodies Cytology only Clinical signs or obits due to kala-azar. Liver and spleen weighed after autopsy of symptomatic dogs to determine hepato-splenomegaly Checked by their veterinarian. Physical data recorded. Physical signs of canine visceral leishmaniosis included wasting, enlargement of lymphoreticular organs, skin lesions and fever spikes Missing Missing Missing Clinical examination Supporting publications 2013:EN by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender considered as an output adopted by the Authority. The European food Safety Authority reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

28 Risk of bias None of the studies on vaccines was considered to be at an overall low risk of bias. Two studies were considered to be at an unclear risk of bias (Genaro et al., 1996, Mohebali et al., 2004). In general, studies had a high risk of bias for the generation of the randomisation sequence, allocation concealment and other sources of bias (Figure 2 and Figure 3). Figure 2: Assessment of methodological quality graph for vaccines: item presented as percentages across all 12 included studies. Leishmune (Lima et al., 2010) CaniLeish (Oliva et al., 2012) Leishmune (Borja-Cabrera et al., 2008) Leishmune (Nogueira et al., 2005) FML (da Silva et al., 2000) MML with MPL/Adjuprime (Gradoni et al., 2005) ALM (Mohebali et al., 2004) LiESAp with MDP (Lemesre et al., 2007) ALi/ALM with BCG (Mohebali et al., 1999) FML with QuilA (Borja-Cabrera et al., 2002) Lyophilized protein (Dunan et al., 1989) Leishvacin (Genaro et al., 1996) Randomisation Concealment Blinding Incomplete data Selective reporting Other bias Study duration Financial support Figure 3: Methodological quality summary for vaccines for each methodological quality item for each included study. Supporting publications 2013:EN

29 Allocation Adequacy of the method used to generate the allocation sequence Impact, modelling and control of canine leishmaniosis in the EU Of the 12 included studies, only two adequately reported the method used to generate the allocation sequence (Dunan et al., 1989a, Genaro et al., 1996). The remainder of the studies stated they performed randomisation but did not specify how this was generated, undertook OCTs and did not do so, or failed to mention this in their report. Allocation concealment None of the included studies adequately reported allocation concealment or did not undertake allocation concealment due to using an OCT design instead. Blinding Six studies reported adequate blinding (Gradoni et al., 2005, Mohebali et al., 2004, Borja-Cabrera et al., 2002, Dunan et al., 1989a, Genaro et al., 1996, Lemesre et al., 2007). One study did not report sufficient information for judgement (Oliva et al., 2012), while the others did not undertake adequate blinding. Incomplete outcome data Three studies reported if there were any subjects lost to follow up (Nogueira et al., 2005b, Lemesre et al., 2007, Mohebali et al., 1999). Three studies failed to report complete outcome data (Borja-Cabrera et al., 2008, Borja-Cabrera et al., 2002, Gradoni et al., 2005), and the other studies did not provide sufficient information to judge this. Selective reporting It was difficult to assess selective reporting as study protocols were not sought, therefore no studies were judged to be free from bias. One study (Dunan et al., 1989a) was judged to be at a high risk of bias by stating it would carry out DTH skin reaction tests but not carrying them out. Other sources of bias Dunan et al. (1999) was considered free from other sources of bias. Five studies failed to provide sufficient information to judge this (Nogueira et al., 2005b, Genaro et al., 1996, Gradoni et al., 2005, Lima et al., 2010, Mohebali et al., 2004), while the remainder were considered to have other sources of bias. da Silva et al. (2000) had a lack of baseline data on dog characteristics and did not allow an accurate evaluation of the infection rate because the vaccine product and the repeated leishmanin doses were likely to have interfered with the serological response with more than half of control subjects showing positive FML-ELISA tests. Oliva et al. (2012) had no explanation of the term 'naive dogs', and it was unclear whether there was baseline imbalance between the two sites used to recruit animals. Borja-Cabrera et al. (2008) recruited 600 healthy dogs from canine visceral leishmaniosis endemic towns showing previous negative results in Leishmania serology by IFAT. For ethical reasons, veterinarians were not able to keep an untreated and exposed control dog population. For comparison, 588 asymptomatic FML-seronegative dogs from another endemic area were included as the exposed untreated group. Lemesre et al. (2007) included healthy seronegative dogs in the study, even though they were PCR positive at bone marrow examination. Mohebali et al. (1999) had very small numbers in control group 1 (only 7 dogs), and the results are presented inconsistently between text and tables. Borja-Cabrera et al. (2002) distributed equivalent number of control and vaccinated Supporting publications 2013:EN

30 dogs in the prevalent and non-prevalent quarters of their study location, the probability of obtaining balanced groups through randomization of a necessarily small number of communities was unlikely to be high. Study duration Eleven studies were considered to be of adequate study duration (12 months or greater), while one study (Nogueira et al., 2005b) evaluated outcomes at 11 months post-intervention. Acknowledgment of financial support Seven studies were judged to be free from bias with respect financial support. Three studies reported commercial financial support (Gradoni et al., 2005, Borja-Cabrera et al., 2008, Nogueira et al., 2005b) but it was judged unclear whether this had biased the results, while the remaining two studies failed to report financial support (Oliva et al., 2012, Mohebali et al., 1999) Effect of interventions Primary outcome: Proportion of dogs infected with L. infantum based on serology or parasite detection Eleven studies reported data on this outcome; with two studies reporting data on two interventions (Table 4). The generated OR (Figure 4) and ARR (Figure 5) are presented. There was a statistically significant protective effect for vaccination for the overall proportion of dogs infected with L. infantum based on serology or parasite detection, in six of the studies (OR range , ARR range ), based on DAT serology results for ALM vaccine (Mohebali et al., 2004), ELISA serology results for ALM with BCG vaccine (Mohebali et al., 1999) b, a case definition of active infection, defined as no clinicopathological signs, positive bone marrow or lymph node PCR and culture for CaniLeish (Oliva et al., 2012), combined culture and bone marrow PCR for LiESAp with MDP vaccine (Lemesre et al., 2007), evidence of parasite DNA for Leishmune (Lima et al., 2010) and the lymph node PCR results for Leishmune (Nogueira et al., 2005b). There was a statistically significant non-protective effect for one study on lyophilized protein based on a case definition defined as a positive antibody test plus parasitological confirmation, or two or more sequential IFA tests showing a rise in antibody titres (Dunan et al., 1989a), and for the FML ELISA result by da Silva et al. (2000) due to a 100% positive reaction observed for the intervention group. The remaining five studies did not detect statistically significant differences between interventions based on ELISA serology for ALi with BCG vaccine (Mohebali et al., 1999) a, Leishvacin based on IFAT and ELISA serology (Genaro et al., 1996), both studies on MML with MPL/Adjuprime which defined infection as sub-patent, asymptomatic and symptomatic based on bone marrow PCR and/or IFAT serology (Gradoni et al., 2005) a and b and for the FML ELISA result reported by Borja-Cabrera et al which reported a 100% positive reaction in the intervention group. The study by Nogueira et al. (2005b) found the largest significant ARR, for every 100 dogs given Leishmune 54 cases of infection with L. infantum based on serology or parasite detection would be averted, followed by Lima et al. (2010) who found that 43 cases per 100 dogs would be averted also following Leishmune vaccination. Mohebali et al. (1999) b found for every 100 dogs given ALM with BCG vaccine 25 cases of infection with L. infantum based on serology or parasite detection would be averted, closely followed by Oliva et al. (2012) who found that 20 cases per 100 dogs would be averted following CaniLeish vaccination. Both Mohebali et al. (2004) and Lemesre et al. (2007) reported that 8 and 6 cases per 100 dogs would be averted due to vaccination with ALM vaccine and LiESAp with MDP respectively. Dunan et al. (1989a) found that 12 more cases per 100 dogs would Supporting publications 2013:EN

31 Number of positive Leishmania findings Number of negative Leishmania findings Number of positive Leishmania findings Number of negative Leishmania findings Odds Ratio and 95% CI ARR Risk and 95% CI Impact, modelling and control of canine leishmaniosis in the EU occur in the intervention group in dogs given lyophilized protein compared to the control solution, while da Silva et al. (2000) found that 32 more cases per 100 dogs would occur in the intervention group given FML vaccine. Table 4: Measures of effect for the proportion of dogs infected with L. infantum based on serology or parasite detection for vaccinations. Intervention Control Measures of effect Study FML (3.31, ) (-0.45, -0.2) (da Silva et al., 2000) MML with MPL (0.01, 4.24) 0.13 (-0.06, 0.29) (Gradoni et al., 2005) a MML with Adjuprime (0.01, 39.52) 0.01 (-0.21, 0.29) (Gradoni et al., 2005) b ALM (0.11, 0.73) 0.08 (0.02, 0.15) (Mohebali et al., 2004) Leishmune (0.002, 0.56) 0.43 (0.18, 0.63) (Lima et al., 2010) CaniLeish (0.10, 0.94) 0.20 (0.02, 0.36) (Oliva et al., 2012) LiESAp with MDP (0.01, 0.64) 0.06 (0.02, 0.11) (Lemesre et al., 2007) Leishmune (0.001, 0.38) 0.54 (0.29, 0.70) (Nogueira et al., 2005b) ALi with BCG (0.06, 2.00) 0.25 (-0.13, 0.57) (Mohebali et al., 1999) a ALM with BCG (0.13, 0.97) 0.25 (0.01, 0.43) (Mohebali et al., 1999) b FML with QuilA (0.02, 55.28) ( ) (Borja-Cabrera et al., 2002) Lyophilized protein (1.55, 6.62) (-0.19, -0.05) (Dunan et al., 1989a) Leishvacin (Genaro et al., 1996) (0.36, 4.99) (-0.02, 0.01) Supporting publications 2013:EN

32 Figure 4: Forest plot of the Odds ratio and 95% confidence interval for the proportion of dogs infected with L. infantum based on serology or parasite detection for vaccinations. The results for FML (da Silva et al. 2000), MML with MPL (Gradoni et al. 2005) and FML with QuilA (Borja-Cabrera et al. 2002) are not presented due to large confidence intervals and the visual masking of the other results. * = OR and 95% CI estimated based on Haldane correction. Figure 5: Forest plot of the Absolute risk reduction and 95% confidence interval for the proportion of dogs infected with L. infantum based on serology or parasite detection for vaccinations. * = OR and 95% CI estimated based on Haldane correction. Supporting publications 2013:EN

33 Number of positive Leishmania findings Number of negative Leishmania findings Number of positive Leishmania findings Number of negative Leishmania findings Odds Ratio and 95% CI ARR and 95% CI Impact, modelling and control of canine leishmaniosis in the EU Secondary outcomes Proportion of dogs infected with L. infantum based on parasite detection Seven studies reported data on this outcome; with one study reporting data on two interventions (Table 5). Data were replicated from the primary outcome for three of the studies as described above (Gradoni et al., 2005, Lima et al., 2010), with new data presented by five studies. There was a statistically significant protective effect for vaccination for the proportion of dogs infected with L. infantum based on parasite detection only, for two of the studies (OR range , ARR range ), based on bone marrow and/or lymph node PCR for CaniLeish (Oliva et al., 2012) and immunohistochemistry results for Leishmune (Nogueira et al., 2005b). There was a statistically significant non-protective effect for one study on lyophilized protein (Dunan et al., 1989a). There were non-significant results for PCR conducted on oligosymptomatic dogs for the FML vaccine (da Silva et al., 2000) and for the LiESAp and MDP PCR result by Lemesre et al. (2007). Table 5: Measures of effect for the proportion of dogs infected with L. infantum based only on parasite detection for vaccinations. Study Intervention Comparator Measures of effect FML (0.004, 2.93) 0.40 (-0.08, 0.77) (da Silva et al., 2000) MML with MPL (0.01, 4.24) 0.13 (-0.1, 0.38) (Gradoni et al., 2005) a MML with Adjuprime (0.01, 39.52) 0.01 (-0.21, 0.29) (Gradoni et al., 2005) b Leishmune (0.002, 0.56) 0.43 (0.18, 0.63) (Lima et al., 2010) CaniLeish (0.14, 0.95) 0.23 (0.02, 0.42) (Oliva et al., 2012) LiESAp with MDP (0.02, 1.03) 0.04 (0.04, 0.08) (Lemesre et al., 2007) Leishmune (0.003, 0.98) 0.24 (0.06, 0.39) (Nogueira et al., 2005b) Lyophilized protein (Dunan et al., 1989a) (1.33, 10.26) (-0.02, 0.13) (a): Data in italics are replicated from the primary outcome Supporting publications 2013:EN

34 Number of positive Leishmania findings Number of negative Leishmania findings Number of positive Leishmania findings Number of negative Leishmania findings Odds Ratio and 95% CI ARR and 95% CI Impact, modelling and control of canine leishmaniosis in the EU Proportion of dogs infected with L. infantum infection based on PCR detection Six studies reported data on this outcome; with one study reporting data on two interventions (Table 6). Data were replicated from the primary outcome for three of the studies (Gradoni et al., 2005, Lima et al., 2010) and from the overall parasite detection result for two studies (da Silva et al., 2000, Lemesre et al., 2007). New data were presented by two studies, A statistically significant protective effect for vaccination for the proportion of dogs infected with L. infantum based on PCR detection only based on combined bone marrow and lymph node PCR for CaniLeish (Oliva et al., 2012) and based on combined blood and lymph node PCR for Leishmune (Nogueira et al., 2005b). Table 6: Measures of effect for the proportion of dogs infected with L. infantum based only on PCR detection for vaccinations. Intervention Comparator Measures of effect Study FML (0.004, 2.93) 0.4 (-0.08, 0.77) (da Silva et al., 2000) MML with MPL (0.01, 4.24) 0.13 (-0.10, 0.38) (Gradoni et al., 2005) a MML with Adjuprime (0.01, 39.52) 0.01 (-0.21, 0.29) (Gradoni et al., 2005) b Leishmune (0.002, 0.56) 0.43 (0.18, 0.63) (Lima et al., 2010) CaniLeish (0.23, 0.84) 0.20 (0.04, 0.34) (Oliva et al., 2012) LiESAp with MDP (0.02, 1.03) 0.04 (0.04, 0.08) (Lemesre et al., 2007) Leishmune (Nogueira et al., 2005b) (0.001, 0.35) 0.33 (0.20, 0.45) (a): Data in italics are replicated from the primary outcome or the overall parasite detection outcome Quantitative PCR No studies measured this outcome. Supporting publications 2013:EN

35 Number of positive Leishmania findings Number of negative Leishmania findings Number of positive Leishmania findings Number of negative Leishmania findings Odds Ratio and 95% CI ARR and 95% CI Impact, modelling and control of canine leishmaniosis in the EU Proportion of dogs infected with L. infantum based on serological detection Nine studies reported data on this outcome; with two studies reporting data on two interventions (Table 7). Data were replicated from the primary outcome for five of the studies. It was possible to generate OR for each of the six new results and ARR for five due to equal proportions of positive and negative findings in both the intervention and comparator groups in one study. There were two statistically significant results, one study found a statistically significant protective effect for vaccination for the proportion of dogs infected with L. infantum based on serological detection only using ELISA only for ALM vaccine (Mohebali et al., 2004), while the other study found a nonprotective effect for lyophilized protein based only on IFAT serology (Dunan et al., 1989a). Nogueira et al. (2005b) reported new data on the FML ELISA results, while the other non-significant results were from IFAT testing for MML with MPL and with Adjuprime vaccines (Gradoni et al., 2005) and for LiESAp with MDP (Lemesre et al., 2007). Table 7: Measures of effect for the proportion of dogs infected with L. infantum based only on serological detection for vaccinations. Intervention Comparator Measures of effect Study FML (3.31, ) (-0.45, -0.20) (da Silva et al., 2000) MML with MPL (0.43, 9.80) (-0.45, 0.17) (Gradoni et al., 2005) a MML with Adjuprime (0.22, 5.86) (-0.37, 0.33) (Gradoni et al., 2005) b ALM (0.09, 0.70) 0.08 (0.02, 0.14) (Mohebali et al., 2004) LiESAp with MDP (0.20, 1.31) 0.04 (-0.02, 0.09) (Lemesre et al., 2007) Leishmune (0.43, 2.31) (Nogueira et al., 2005b) ALi with BCG (0.06, 2.00) 0.25 (-0.13, 0.57) (Mohebali et al., 1999) a ALM with BCG (0.13, 0.97) 0.25 (0.01, 0.43) (Mohebali et al., 1999) b FML with QuilA (0.02, 55.28) (-0.09, 0.09) (Borja-Cabrera et al., 2002) Lyophilized protein (1.56, 6.66) (-0.19, -0.05) (Dunan et al., 1989a) Leishvacin (Genaro et al., 1996) (0.36, 4.99) (-0.02, 0.01) (a): Data in italics are replicated from the primary outcome Supporting publications 2013:EN

36 Number of positive cellular immunity tests Number of negative cellular immunity tests Number of positive cellular immunity tests Number of negative cellular immunity tests Odds Ratio and 95% CI ARR and 95% CI Impact, modelling and control of canine leishmaniosis in the EU Quantitative serology Three studies reported data on this outcome. Two studies had a higher reading in the intervention group where the mean difference and SE was ± 0.22 (da Silva et al., 2000) and ± 0.13 (Borja-Cabrera et al., 2002) respectively. One study had a higher reading in the control group where the mean difference and SE was ± 0.14 (Lemesre et al., 2007). Proportion of dogs infected with L. infantum based on cellular immunity tests Two studies reported data on this outcome (Table 8). There was a statistically significant protective effect for vaccination for the proportion of dogs infected with L. infantum based on cellular immunity tests for both the FML vaccine (da Silva et al, 2000) and the FML with QuilA vaccine (Borja-Cabrera et al. 2002). Table 8: Measures of effect for the proportion of dogs infected with L. infantum based only on cellular immunity tests for vaccinations. Study Intervention Comparator Measures of effect FML (da Silva et al., 2000) FML with QuilA (Borja-Cabrera et al., 2002) (10.69, ) (-0.74, -0.47) (143.30, ) (-0.99, -0.85) Size of the Delayed Type Hypersensitivity reaction Two studies reported data on this outcome. Both studies had a higher reading in the intervention group where the mean difference and SE was ± 1.46 (da Silva et al., 2000) and ± 3.09 (Borja-Cabrera et al., 2002) respectively. Supporting publications 2013:EN

37 Number of positive Leishmania findings Number of negative Leishmania findings Number of positive Leishmania findings Number of negative Leishmania findings Odds Ratio and 95% CI ARR and 95% CI Proportion of dogs that died infected with L. infantum Impact, modelling and control of canine leishmaniosis in the EU Four studies reported data on this outcome; with one study reporting data on two interventions (Table 9). There was a statistically significant protective effect for vaccination for the proportion of dogs dying infected with L. infantum, for two studies on Leishmune and FML with QuilA vaccine presenting an OR < 1.00 (range ) and an ARR >0.00 (range ). The three studies on FML (da Silva et al. 2000), MML with MPL (Gradoni et al, 2005) a and MML with Adjuprime (Gradoni et al, 2005) b were statistically non-significant. Table 9: Measures of effect for the proportion of dogs that died infected with L. infantum for vaccinations. Intervention Comparator Measures of effect Study FML (0.01, 2.00) 0.07 (-0.01, 0.16) (da Silva et al., 2000) MML with MPL (0.02, 50.31) (-0.21, 0.22) (Gradoni et al., 2005) a MML with Adjuprime (0.03, 75.37) (-0.29, 0.21) (Gradoni et al., 2005) b Leishmune (0.002, 0.04) 0.27 ( ) (Borja-Cabrera et al., 2008) FML with QuilA (Borja-Cabrera et al., 2002) (0.01, 0.81) 0.17 (0.04, 0.32) Supporting publications 2013:EN

38 Number of positive Leishmania findings Number of negative Leishmania findings Number of positive Leishmania findings Number of negative Leishmania findings Odds Ratio and 95% CI ARR and 95% CI Proportion of clinically ill dogs infected with L. infantum Impact, modelling and control of canine leishmaniosis in the EU Six studies reported data on this outcome; with one study reporting data on two interventions (Table 10). There was a statistically significant protective effect for vaccination for the proportion of clinically ill dogs infected with L. infantum for two studies on FML with QuilA (Borja-Cabrera et al., 2002) and for Leishmune (Nogueira et al., 2005b). There was no significant difference for the remaining five studies. Table 10: Measures of effect for the proportion of clinically ill dogs infected with L. infantum for vaccinations. Intervention Comparator Measures of effect Study FML (0.11, 2.03) 0.05 (-0.06, 0.16) (da Silva et al., 2000) MML with MPL (0.38, ) -0,19 (-.0.43, 0.07) (Gradoni et al., 2005) a MML with Adjuprime (0.36, ) -0,19 (-0.49, 0.08) (Gradoni et al., 2005) b CaniLeish (0.07, 1.13) 0.15 (-0.01, 0.3) (Oliva et al., 2012) LiESAp with MDP (0.15, 2.70) 0.01 (-0.03, 0.05) (Lemesre et al., 2007) Leishmune (0.003, 0.8) 0.24 (0.08, 0.39) (Nogueira et al., 2005b) FML with QuilA (Borja-Cabrera et al., 2002) (0.01, 0.81) 0.17 (0.04, 0.32) Adverse effects related to the intervention Four studies reported data on this outcome. Two studies reported no adverse side effects (Gradoni et al., 2005) a and b ; one reported that local reactions occurred including ulcers in 64.51% of dogs, and papules in 4.0% of dogs (Mohebali et al., 2004), while the other reported signs of anorexia, non-lethal generalized anaphylactic reaction and hypersomnia (Dunan et al., 1989a). Supporting publications 2013:EN

39 Collars Description of included studies All five studies examined the effect of deltamethrin collars (Table 11, Table 12 and Table 13). Two RCTs were conducted in Italy and Iran of 24 and 12 months in duration respectively. Foglia Manzillo et al. (2006) focussing only on stray dogs and conducting cytological parasite detection and IFA serology, while Gavgani et al. (2002) examined all domestic dogs, killing all strays and only conducted DAT serology. The other three trials were OCTs: Aoun et al. (2008) was a Tunisian study of specific breeds of working dogs followed for 22 months and examined by PCR, culture and ELISA and IFA serology. Ferroglio et al. (2008) was a 24 month long Italian study of kennel and client-owned dogs of unknown age examined only by IFA serology. Maroli et al. (2001) was another 24 month long Italian study of adult and puppy pets examined only by IFA serology. Table 11: The signalment characteristics for studies on collars. Study Publication Affiliation first author (Aoun et Full paper Laboratoire de al., 2008) Recherche (Ferrogli o et al., 2008) (Foglia Manzillo et al., 2006) (Gavgani et al., 2002) (Maroli et al., 2001) Full paper Università di Torino Full paper Università di Napoli Full paper Tabriz University of Medical Sciences Full paper Istituto Superiore di Sanità Country Type Age Gender Neutered Housing Breed Tunisia Working dogs only months Italy Kennel and clientowned dogs Italy Stray dogs only months Iran All domestic dogs. All strays were killed Italy Pets only Adults and puppies Missing Missing Kennels German Shepherds, Rottweilers Malinois Missing Missing Missing Outdoors Missing Mixed Missing Kennels Mainly mongrels Missing Missing Missing Missing Missing Missing Missing Probably combination homes, outdoors of Missing and Supporting publications 2013:EN by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender considered as an output adopted by the Authority. The European food Safety Authority reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

40 Table 12: The intervention characteristics for studies on collars. Impact, modelling and control of canine leishmaniosis in the EU Study Intervention Dose of treatment Route of administration (Aoun et al., 2008) (Ferroglio et al., 2008) (Foglia Manzillo et al., 2006) (Gavgani et al., 2002) (Maroli et al., 2001) Number of treatments Comparator Study duration in months Untreated October) 2005 and 2006 controls Deltamethrin 760mg Collar Placed for 7 months, twice a year (April to Deltamethrin Missing Collar 1 Untreated controls Deltamethrin Large-size dogs 1g, small Collar Kept on from late May to mid-october, for Untreated and medium sized dogs two consecutive transmission seasons controls 0.76g Deltamethrin 40mg/g Collar 1 Untreated controls Deltamethrin 1g for large dogs, 0.76g Collar Administered in May 1998 and 1999, Untreated for small dogs, 1.52g for checked every fortnight between June and controls very large dogs October and refitted during this period (a): OCT = Non-randomised controlled clinical trial (b): RCT = Randomised controlled clinical trial Study design 22 OCT 24 OCT 24 RCT 12 RCT 24 OCT Supporting publications 2013:EN by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender considered as an output adopted by the Authority. The European food Safety Authority reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

41 Diagnostic method Tissue Sensitivity Specificity Diagnostic method Tissue Sensitivity Specificity Methods of serology Antigens Antibodies Methods of parasite detection Case definition provided by the paper Table 13: The diagnostic methods for studies on collars. Impact, modelling and control of canine leishmaniosis in the EU Pre-trial diagnostics Post-trial diagnostics Study (Aoun et al., 2008) (Ferroglio et al., 2008) (Foglia Manzillo et al., 2006) (Gavgani et al., 2002) (Maroli et al., 2001) Serology and PE Serology and PE Serology and PE Serology only Serology only (a): PE = Physical examination Serum and PE Serum and PE Serum and PE Serum only Serum only Missing Missing Serology, PE and parasite detection Missing Missing Serology only Missing Missing Serology, PE, laboratory testing and parasite detection Serology only Missing Missing Serology only Whole blood, serum and buffy coat Serum only Bone marrow, lymph nodes and serum Serum only Serum only 95 Missing ELISA and IFA Missing Missing IFA only Missing Missing IFA only DAT only Missing Missing IFA only MON1 Missing PCR and culture Missing Missing NA NA L. infantum zymodeme MON1 Promastigotes L. donovani Promastigotes WHO ref. strain of L. infantum Leishmania, FITCconjugated anti-dog IgG Cytology only Missing NA NA Missing NA NA Clinical signs of slimming, enlarged glands, hypergriffose and skin ulcers Inspection for the presence of seven signs attributable to Leishmania infection (lymph adenopathy, weight loss, ocular lesions, exfoliative dermatitis, alopecia, skin ulcers and onychogryphosis. Animals were scored for clinical and laboratory signs as asymptomatic (no signs), oligosymptomatic (1-3 signs) and symptomatic (>3 signs) Supporting publications 2013:EN by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender considered as an output adopted by the Authority. The European food Safety Authority reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

42 Aoun et al., 2008 Ferroglio et al., 2008 Foglia Manzillo et al., 2006 Gavgani et al., 2002 Maroli et al., 2001 Impact, modelling and control of canine leishmaniosis in the EU Risk of bias All the studies on collars were considered to be at an overall high risk of bias, particularly for allocation concealment, blinding, incomplete outcome data and other sources of bias (Figure 6 and Figure 7). Figure 6: Assessment of methodological quality graph for collars: item presented as percentages across all five included studies. Randomisation Concealment Blinding Incomplete data Selective reporting Other bias Study duration Financial support Figure 7: Methodological quality summary for collars for each methodological quality item for each included study. Supporting publications 2013:EN

43 Allocation Adequacy of the method used to generate the allocation sequence Of the five included studies, only one adequately reported the method used to generate the allocation sequence (Gavgani et al., 2002). Two studies failed to mention this in their report (Aoun et al., 2008, Foglia Manzillo et al., 2006) and two studies were OCTs. Allocation concealment None of the included studies adequately reported allocation concealment. Gavgani et al. (2002) was a matched-cluster RCT, two studies failed to mention this in their report (Aoun et al., 2008, Foglia Manzillo et al., 2006) while the other two studies were OCTs. Blinding One study reported adequate blinding (Ferroglio et al., 2008). The remaining four studies did not undertake adequate blinding. Incomplete outcome data Two studies reported if there were any subjects lost to follow up (Ferroglio et al., 2008, Gavgani et al., 2002), while the other three failed to report this. Selective reporting It was difficult to assess selective reporting as study protocols were not sought, therefore no studies were judged to be free from bias. Other sources of bias No studies were considered free from other sources of bias. Four studies failed to provide sufficient information to judge this (Foglia Manzillo et al., 2006, Gavgani et al., 2002, Aoun et al., 2008, Ferroglio et al., 2008), while Maroli et al. (2001) was considered to have bias due to study design recruiting animals from different sites. One area was selected as the intervention area, and four other towns designated as control sites. During two consecutive transmission seasons, collars were fitted to about 70% of the dogs owned by the population of the intervention site, estimated to be about 500. In the four control towns surveillance activities were continued. On the advice of local veterinarians, placebo collars were not fitted to the dogs in the control areas because their inability to control ticks and fleas may have given the impression they were of no use. Study duration All studies were considered to be of adequate study duration. Acknowledgment of financial support Two studies were judged to be free from bias with respect financial support (Gavgani et al., 2002, Maroli et al., 2001). One study reported commercial financial support (Foglia Manzillo et al., 2006) Supporting publications 2013:EN

44 Number of positive Leishmania findings Number of negative Leishmania findings Number of positive Leishmania findings Number of negative Leishmania findings Odds Ratio and 95% CI ARR and 95% CI Impact, modelling and control of canine leishmaniosis in the EU but it was judged unclear whether this had biased the results, two studies failed to report financial support (Aoun et al., 2008, Ferroglio et al., 2008) Effect of interventions Primary outcome: Proportion of dogs infected with L. infantum based on serology or parasite detection Five studies reported data on this outcome (Table 14), and OR (Figure 8) and ARR (Figure 9) were generated. There was a statistically significant protective effect for collars for the overall proportion of dogs infected with L. infantum based on serology or parasite detection for four of the five studies, with each study presenting an OR<1.00 (range ) and an ARR>0.00 (range ); based on IFAT serology only (Ferroglio et al., 2008, Foglia Manzillo et al., 2006, Maroli et al., 2001) and DAT serology (Gavgani et al., 2002). There was a non-significant result for testing on ELISA and IFAT (Aoun et al., 2008), The study by Foglia Manzillo et al. (2006) found the largest ARR, for every 100 dogs provided with deltamethrin collars 34 cases of infection with L. infantum based on serology or parasite detection would be averted, followed by Maroli et al. (2001) who found that 22 cases per 100 dogs would be averted following treatment. Aoun et al. (2008), Ferroglio et al. (2008) and Gavgani et al. (2002) reported that 16, 13 and 4 cases per 100 dogs would be averted due to deltamethrin collar placement respectively. Table 14: Measures of effect for the proportion of dogs infected with L. infantum based on serology or parasite detection for collars. Study Intervention Comparator Measures of effect Aoun et al., (0.003, 1.08) 0.16 (0.03, 0.3) Ferroglio et al., (0.04, 0.46) 0.13 (0.07, 0.2) (Foglia Manzillo et al., (0.09, 0.66) 0.34 (0.1, 0.53) 2006) (Gavgani et al., 2002) (0.22, 0.91) 0.04 (0.01, 0.07) (Maroli et al., 2001) (0.03, 0.31) 0.22 (0.13, 0.32) Supporting publications 2013:EN

45 Figure 8: Forest plot of the Odds ratio and 95% confidence interval for the proportion of dogs infected with L. infantum based on serology or parasite detection for collars. * = OR and 95% CI estimated based on Haldane correction. Figure 9: Forest plot of the Absolute risk reduction and 95% confidence interval for the proportion of dogs infected with L. infantum based on serology or parasite detection for collars. * = OR and 95% CI estimated based on Haldane correction. Supporting publications 2013:EN

46 Number of positive Leishmania findings Number of negative Leishmania findings Number of positive Leishmania findings Number of negative Leishmania findings Odds Ratio and 95% CI ARR and 95% CI Impact, modelling and control of canine leishmaniosis in the EU Secondary outcomes Proportion of dogs infected with L. infantum based on serological detection Five studies reported data on this outcome (Table 15). Data were replicated from the primary outcome for four of the studies, with new data presented by Gavgani et al. (2002). For each result there was a statistically significant protective effect for collars for the proportion of dogs infected with L. infantum based on serological detection only, with each study presenting an OR < 1.00 (range ) and an ARR >0.00 (range ). Table 15: Measures of effect for the proportion of dogs infected with L. infantum based only on serological detection for collars. Study Intervention Comparator Measures of effect Aoun et al., (0.003, 1.08) 0.16 (0.03, 0.3) Ferroglio et al., (0.04, 0.46) 0.13 (0.07, 0.2) 2008 (Foglia Manzillo et (0.09, 0.66) 0.13 (0.07, 0.2) al., 2006) (Gavgani et al., (0.002, 0.01) 0.85 (0.81, 0.88) 2002) (Maroli et al., 2001) (0.03, 0.31) 0.22 (0.13, 0.32) (a): Data in italics are replicated from the primary outcome Proportion of clinically ill dogs infected with L. infantum One study reported data on this outcome (Aoun et al., 2008), reporting no Leishmania positive findings in the intervention group of 42 dogs, compared to 3 positive findings out of 38 dogs in the comparison group, a non-significant reduction in risk (OR 0.21, 95% CI 0.01, 2.39, ARR 0.08, 95% CI 0.03, 0.21) Adverse effects related to the intervention Two studies reported data on this outcome. One study reported a local reaction in one dog (Aoun et al., 2008) and another reported no adverse effects (Maroli et al., 2001). Other secondary outcomes None of the studies on collars measured the proportion of dogs infected with L. infantum based on parasite detection, the proportion of dogs infected with L. infantum based on PCR detection, quantitative PCR, quantitative serology, the proportion of dogs infected with L. infantum based on cellular immunity tests, size of the delayed type hypersensitivity reaction or the proportion of dogs that died infected with L. infantum. Supporting publications 2013:EN

47 Spot-ons Description of included studies Studies on spot-ons investigated 65% permethrin or 10% imidacloprid with 50% permethrin solutions (Table 16, Table 17 and Table 18). Ferroglio et al. (2008) and Giffoni et al. (2002) conducted OCTs of 24 and 3 months respectively assessing the efficacy of permethrin spot-on. Insufficient signalment data were provided to compare the populations used which were sourced in Italy and Brazil respectively, and were analysed using IFA serological methods. Otranto et al. (2007) and Otranto et al. (2010) conducted RCTs of 12 months duration in Italy assessing the efficacy of a combination of imidacloprid and permethrin. Otranto et al. (2007) evaluated treatment every 7 and every 14 days and analysed PCR, cytology and rapid tests for serological determination, while Otranto et al. (2010) evaluated treatment every 21 days on average and Otranto et al. (2007)analysed PCR and IFA serology. Table 16: The signalment characteristics for studies on spot-ons. Study Publication Affiliation first author (Ferroglio et al., 2008) Full paper Università di Torino (Giffoni et al., 2002) Full paper Zoonosis Control Centre (Otranto et al., 2007) Full paper Università degli Studi di Bari (Otranto et al., 2010) Full paper Università degli Studi di Bari Country Type Age Gender Neutered Housing Breed Italy Combination of types Missing Missing Missing Outdoors Missing Brazil Other only Missing Missing Missing Missing Missing Italy Missing Over 7 weeks of age Missing Missing Kennels Missing Italy Stray dogs only months Missing Missing Kennels Missing Supporting publications 2013:EN by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender considered as an output adopted by the Authority. The European food Safety Authority reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

48 Table 17: The intervention characteristics for studies on spot-ons. Impact, modelling and control of canine leishmaniosis in the EU Study Intervention Dose of treatment Route of administration (Ferroglio et al., 2008) (Giffoni et al., 2002) (Otranto et al., 2007) a (Otranto et al., 2007) b (Otranto et al., 2010) Number of treatments 65% permethrin Missing Spot-on Every 30 days 65% permethrin Dogs weighing up to 3kg were treated with one - half of a tube (0.5ml); dogs weighing 3-15kg received the full contents of one tube (1.0ml); and dogs weighing more than 15 kg received two tubes (2.0ml) 10% imidacloprid and 50% permethrin 10% imidacloprid and 50% permethrin 10% imidacloprid and 50% permethrin (a): OCT = Non-randomised controlled clinical trial (b): RCT = Randomised controlled clinical trial Manufacturer's instructions, basis of the animal bodyweight ranges. Treated every 28 ± 2 days. Manufacturer's instructions, basis of the animal bodyweight ranges. Treated every 14 ± 2 days. Comparator Untreated controls Spot-on 3 Untreated controls Spot-on 7 Untreated controls Spot-on 14 Untreated controls Label instructions of manufacturer Spot-on Every 21 ± 2 days Untreated controls Study duration in months Study design 24 OCT 3 OCT 12 RCT 12 RCT 12 RCT Supporting publications 2013:EN by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender considered as an output adopted by the Authority. The European food Safety Authority reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

49 Table 18: The diagnostic methods for studies on spot-ons. Impact, modelling and control of canine leishmaniosis in the EU Study 65% permethrin (Ferroglio et al., 2008) 65% permethrin (Giffoni et al., 2002) 10% imidacloprid and 50% permethrin (Otranto et al., 2007) a 10% imidacloprid and 50% permethrin (Otranto et al., 2007) b 10% imidacloprid and 50% permethrin (Otranto et al., 2010) Diagnos tic method IFA only Pre-trial diagnostics Tissue Sensitivity Specificity Diagnostic method NA Missing Missing Serology only NA NA NA NA Serology only Rapid tests Rapid tests IFA only (a): PE = Physical examination Lymph nodes, skin and serum Lymph nodes, skin and serum Whole blood, bone marrow, skin and buffy coat Missing Missing Serology and parasite detection Missing Missing Serology and parasite detection Missing Missing Serology and parasite detection Post-trial diagnostics Tissue Sensitivity Specificity Methods of serology Serum only Serum only Lymph nodes, skin and serum Serum and PE Whole blood, bone marrow, skin, serum and buffy coat Antigen Antibodies Methods of parasite detection Missing Missing IFA only Missing Missing NA NA Missing Missing IFA only Missing Antibodies to canine visceral leishmaniasis Missing Missing Rapid tests Missing Missing Rapid tests Missing Missing IFA only L. infantum zymodeme MON1 NA NA NA PCR and cytology NA NA PCR and cytology Missing PCR only Case definition provided by the paper NA NA NA NA Supporting publications 2013:EN by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender considered as an output adopted by the Authority. The European food Safety Authority reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

50 Risk of bias None of the studies on spot-ons were considered to be at an overall low risk of bias. One study was considered to be at an unclear risk of bias (Otranto et al., 2010). In general, studies had a high risk of bias for the generation of the randomisation sequence, allocation concealment and other sources of bias (Figure 10 and Figure 11). Figure 10: Assessment of methodological quality graph for spot-ons: item presented as percentages across all four included studies. 10% imidacloprid 10% imidacloprid 65% permethrin 65% permethrin and 50% permethrin and 50% permethrin Ferroglio et al., 2008 Giffoni et al., 2002 Otranto et al., 2007 Otranto et al., 2010 Randomisation Concealment Blinding Incomplete data Selective reporting Other bias Study duration Financial support Figure 11: Methodological quality summary for spot-ons for each methodological quality item for each included study. Supporting publications 2013:EN

51 Allocation Adequacy of the method used to generate the allocation sequence Of the four included studies, only one adequately reported the method used to generate the allocation sequence (Otranto et al., 2010). Two studies were OCTs (Ferroglio et al., 2008, Giffoni et al., 2002) while Otranto et al. (2007) used a non-random method of sequence generation. Allocation concealment None of the included studies adequately reported allocation concealment. Two studies were OCTs (Ferroglio et al., 2008, Giffoni et al., 2002) while Otranto et al. (2007) did not conceal allocation. Otranto et al. (2010) did not report sufficient information to judge this. Blinding Three studies reported adequate blinding (Ferroglio et al., 2008, Otranto et al., 2010, Otranto et al., 2007). Giffoni et al. (2002) did not perform blinding. Incomplete outcome data One study reported if there were any subjects lost to follow-up (Ferroglio et al., 2008). Two studies did not report sufficient information to judge this (Otranto et al., 2010, Otranto et al., 2007) while Giffoni et al. (2002) had a large number of animals lost to follow-up. Selective reporting It was difficult to assess selective reporting as study protocols were not sought, therefore no studies were judged to be free from bias. Other sources of bias No studies were considered free from other sources of bias. Two studies failed to provide sufficient information to judge this (Ferroglio et al., 2008, Otranto et al., 2010), while the other two were considered to have other sources of bias. Giffoni et al. (2002) had a non-comparable baseline prevalence, dogs in one region were assigned to treatment with spot-on because of the high incidence of dogs positive for visceral leishmaniosis in this area. Dogs in another region were untreated controls. There was also low sensitivity of the test used to define infection IFAT and unclear intention to treat analysis. Otranto et al. (2007) recruited dogs from different sites. Dogs were housed in two kennels in different regions of the country where endemic CanL had been reported over the previous two years. One kennel (latitude North, longitude East) had about 1100 animals and an estimated CanL prevalence of 20%, while the other kennel (latitude North, longitude East) housed about 900 animals with an estimated CanL prevalence of 17%. Study duration Three studies were considered to be of an adequate study duration (Ferroglio et al., 2008, Otranto et al., 2010, Otranto et al., 2007), while Giffoni et al. (2002) evaluated outcomes at three months postintervention. Supporting publications 2013:EN

52 Number of positive Leishmania findings Number of negative Leishmania findings Number of positive Leishmania findings Number of negative Leishmania findings Odds Ratio and 95% CI ARR and 95% CI Impact, modelling and control of canine leishmaniosis in the EU Acknowledgment of financial support No studies were judged to be free from bias with respect financial support. Three studies reported commercial financial support (Giffoni et al., 2002, Otranto et al., 2007, Otranto et al., 2010) but it was judged unclear whether this had biased the results. One study failed to report financial support (Ferroglio et al., 2008) Effect of interventions Primary outcome: Proportion of dogs infected with L. infantum based on serology or parasite detection Four studies reported data on this outcome; with one study reporting data on two interventions (Table 19), and OR (Figure 12) and ARR (Figure 13) generated. There was a statistically significant protective effect for spot-ons for the overall proportion of dogs infected with L. infantum based on serology or parasite detection, for four of the studies (OR range ) and an ARR >0.00 (range ) including the IFAT serology for 65% permethrin (Ferroglio et al., 2008), IFAT serology for 10% imidacloprid and 50% permethrin (Otranto et al. 2010),and both doses of 10% imidacloprid and 50% permethrin based on positive immunochromatographic rapid tests or PCR/cytology (Otranto et al., 2007) a and b. There was no significant difference for the remaining study, however there was a trend towards a protective effect for 65% permethrin based on IFAT serology (Giffoni et al., 2002). The study by Otranto et al. (2010) found the largest ARR, for every 100 dogs given 10% imidacloprid and 50% permethrin 37 cases of infection with L. infantum based on serology or parasite detection would be averted. Ferroglio et al. (2008) found that 14 cases per 100 dogs would be averted due to 65% permethrin while Otranto et al. (2007) a and b found that 10 cases per 100 dogs would be averted due to 10% imidacloprid and 50% permethrin for both dosing regimens studied. Table 19: Measures of effect for the proportion of dogs infected with L. infantum based on serology or parasite detection for spot-ons. Study Intervention Comparator Measures of effect 65% permethrin (Ferroglio et al., 2008) 65% permethrin (Giffoni et al., 2002) 10% imidacloprid and 50% permethrin (Otranto et al., 2007) a 10% imidacloprid and 50% permethrin (Otranto et al., 2007) b 10% imidacloprid and 50% permethrin (Otranto et al., 2010) (0.04, 0.45) 0.14 (0.07, 0.20) (0.24, 1.47) 0.06 (-0.05, 0.16) (0.02, 0.40) 0.10 (0.05, 0.15) (0.006, 0.33) 0.10 (0.06, 0.15) (0.001, 0.20) 0.37 (0.24, 0.50) Supporting publications 2013:EN

53 Figure 12: Forest plot of the Odds ratio and 95% confidence interval for the proportion of dogs infected with L. infantum based on serology or parasite detection for spot-ons. * = OR and 95% CI estimated based on Haldane correction. Figure 13: Forest plot of the Absolute risk reduction and 95% confidence interval for the proportion of dogs infected with L. infantum based on serology or parasite detection for spot-ons. * = OR and 95% CI estimated based on Haldane correction. Supporting publications 2013:EN

54 Secondary outcomes Proportion of dogs infected with L. infantum based on serological detection Data were replicated from the primary outcome for one study with 8 positive findings in the intervention group of 74 dogs and 17 positive findings in the comparison group of 101 dogs (Giffoni et al., 2002), with no significant difference between groups based on only serological detection. Adverse effects related to the intervention Three studies reported no adverse effects (Otranto et al., 2007) a and b (Otranto et al., 2010). Other secondary outcomes None of the studies on spot-ons measured the proportion of dogs infected with L. infantum based on parasite detection, the proportion of dogs infected with L. infantum based on PCR detection, quantitative PCR, quantitative serology, the proportion of dogs infected with L. infantum based on cellular immunity tests, size of the delayed type hypersensitivity reaction, the proportion of dogs that died infected with L. infantum or the proportion of clinically ill dogs infected with L. infantum. Supporting publications 2013:EN

55 Prophylactic medications Description of included studies Two studies on prophylactic medications investigated domperidone liquid solution and one investigated allopurinol capsules (Table 20, Table 21 and Table 22). Saridomichelakis et al. (2005) conducted a 12-month long RCT assessing the efficacy of allopurinol in Central Greece on purebred and crossbreed pet dogs living outside, utilising PCR and cytology and ELISA and IFA serological methods for diagnosis. Two studies examined the efficacy of Domperidone, Gomez-Ochoa et al. (2012) and Llinas et al. (2011), conducting RCTs of 9 and 21 months respectively in Spain. No signalment data were provided by Llinas et al. (2011) and different doses of domperidone were used, 1ml/10kg/day and 0.5ml/10kg/day respectively. Gomez-Ochoa et al. (2012) evaluated DAT serology and cytology, while Llinas et al. (2011) only evaluated IFA serology. Table 20: The signalment characteristics for studies on prophylactic medications. Study Publication Affiliation first author (Saridomichelakis et al., 2005) Full paper Aristotle University of Thessaloniki (Gomez-Ochoa et al., 2012) Abstract Universidad de Zaragoza (Llinas et al., 2011) Abstract Hospital Veterinario Valenica Sur Country Type Age Gender Neutered Housing Breed Central Pets only Mixed Missing Outdoors Purebreds and Greece crossbreds Spain Missing Different Mixed Missing Kennels Different breeds ages Spain Missing Missing Missing Missing Missing Missing Supporting publications 2013:EN by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender considered as an output adopted by the Authority. The European food Safety Authority reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

56 Table 21: The intervention characteristics for studies on prophylactic medications. Study Intervention Dose of treatment Route of administration Number of treatments Comparator Study duration in months Identical-looking (Saridomichelakis et al., 2005) Allopurinol 20mg/kg/day Oral capsule Usually the first week of the month, from 7 months (Gomez-Ochoa et al., 2012) Domperidone 1ml/10kg/day Oral liquid 30 consecutive days one May-June, other September-October (Llinas et al., 2011) Domperidone 0.5mg/kg/day Oral liquid 30 consecutive days, on a 4-monthly basis during 21 months (a): RCT = Randomised controlled clinical trial Study design 12 RCT placebo tablets Untreated controls 9 RCT Untreated controls 21 RCT Supporting publications 2013:EN by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender considered as an output adopted by the Authority. The European food Safety Authority reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

57 Table 22: The diagnostic methods for studies on prophylactic medications. Pre-trial diagnostics Post-trial diagnostics Study Allopurinol (Saridomichelakis et al., 2005) Domperidone (Gomez-Ochoa et al., 2012) Domperidone (Llinas et al., 2011) Diagnostic method Serology, PE and parasite detection Serology and PE Serology only Tissue Sensitivity Specificity Diagnostic method Bone marrow, lymph nodes and serum Serum and PE Serum only Missing Missing Serology and parasite detection Missing Missing Serology, PE and parasite detection Missing Missing Serology and PE Tissue Sensitivity Specificity Methods of serology Whole blood, bone marrow, lymph nodes and serum Whole blood, bone marrow, lymph nodes and serum Serum and PE Missing Missing ELISA and IFA Missing Missing DAT only Missing Missing IFA only Antigens Antibodies Methods of parasite detection Missing Missing PCR and cytology Missing Missing Anti- Leishmania antibody Anti- Leishmania antibody titres Cytology only NA Case definition provided by the paper NA Clinical examinations, clinical signs compatible with CanL (peripheral lymphadenopathy and alopecia) Any clinical sign compatible with disease (a): PE = Physical examination Supporting publications 2013:EN by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender considered as an output adopted by the Authority. The European food Safety Authority reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

58 Risk of bias All of the studies on prophylactic medication were considered to be at a high risk of bias. In general, studies had a high risk of bias for acknowledgment of financial support (Figure 14 and Figure 15). Figure 14: Assessment of methodological quality graph for prophylactic medications: item presented as percentages across all three included studies. Allopurinol Domperidone Domperidone Saridomichelakis et al., 2005 Gomez-Ochoa et al., 2012 Llinas et al., 2011 Randomisation Concealment Blinding Incomplete data Selective reporting Other bias Study duration Financial support Figure 15: Methodological quality summary for prophylactic medications for each methodological quality item for each included study. Supporting publications 2013:EN

59 Allocation Adequacy of the method used to generate the allocation sequence Of the three included studies, only one adequately reported the method used to generate the allocation sequence (Saridomichelakis et al., 2005). The other two studies (Gomez-Ochoa et al., 2012, Llinas et al., 2011) reported themselves as RCTs but failed to mention this in their report. Allocation concealment All three studies failed to report sufficient information to judge this. Blinding Two studies (Gomez-Ochoa et al., 2012, Saridomichelakis et al., 2005) reported adequate blinding, while Llinas et al. (2011) measured clinical disease outcomes but did not perform blinding. Incomplete outcome data Saridomichelakis et al. (2005) had large losses-to follow up (40%) and was judged to be at a high risk of bias. Gomez-Ochoa et al. (2012) had no missing outcome data and was judged not to be at risk of bias, while Llinas et al. (2011) had insufficient reporting to judge this. Selective reporting It was difficult to assess selective reporting as study protocols were not sought, therefore no studies were judged to be free from bias Other sources of bias The studies failed to provide sufficient information to judge this. Study duration Two studies were considered to be of an adequate study duration (Llinas et al., 2011, Saridomichelakis et al., 2005), while one study (Gomez-Ochoa et al., 2012) evaluated outcomes at nine months postintervention. Acknowledgment of financial support Saridomichelakis et al. (2005) acknowledged financial support which was not judged to bias the results, while financial support was not acknowledged by Gomez-Ochoa et al. (2012) or Llinas et al. (2011). Supporting publications 2013:EN

60 Number of positive Leishmania findings Number of negative Leishmania findings Number of positive Leishmania findings Number of negative Leishmania findings Odds Ratio and 95% CI ARR and 95% CI Effect of interventions Impact, modelling and control of canine leishmaniosis in the EU Primary outcome: Proportion of dogs infected with L. infantum based on serology or parasite detection Three studies reported data on this outcome (Table 23), and OR (Figure 16) and ARR (Figure 17) were generated. There was a statistically significant protective effect for prophylactic medication for the overall proportion of dogs infected with L. infantum based on serology or parasite detection, for one study based on IFAT serology for domperidone (Llinas et al., 2011). Llinas et al. (2011) and Gomez-Ochoa et al. (2012) found that for every 100 dogs prophylactically medicated with domperidone 37 and 6 cases of infection with L. infantum based on serology or parasite detection would be averted, respectively. There was no significant difference for the other study, however there was a trend towards a non-protective effect for allopurinol based on bone marrow PCR (Saridomichelakis et al., 2005). Table 23: Measures of effect for the proportion of dogs infected with L. infantum based on serology or parasite detection for prophylactic medications. Study Intervention Comparator Measures of effect Allopurinol (Saridomichelakis et al., 2005) Domperidone (Gomez-Ochoa et al., 2012) Domperidone (Llinas et al., 2011) (0.38, 42.18) (-0.53, 0.17) (0.004, 1.11) 0.06 (0.01, 0.12) (0.05, 0.42) 0.37 (0.18, 0.52) Supporting publications 2013:EN

61 Figure 16: Forest plot of the Odds ratio and 95% confidence interval for the proportion of dogs infected with L. infantum based on serology or parasite detection for prophylactic medication. * = OR and 95% CI estimated based on Haldane correction. Figure 17: Forest plot of the Absolute risk reduction and 95% confidence interval for the proportion of dogs infected with L. infantum based on serology or parasite detection for prophylactic medication. * = OR and 95% CI estimated based on Haldane correction. Supporting publications 2013:EN

62 Number of positive Leishmania findings Number of negative Leishmania findings Number of positive Leishmania findings Number of negative Leishmania findings Odds Ratio and 95% CI ARR and 95% CI Impact, modelling and control of canine leishmaniosis in the EU Secondary outcomes Proportion of dogs infected with L. infantum based on parasite detection Data were replicated from the primary outcome for one study with 6 positive findings in the intervention group of 15 dogs and 1 positive findings in the comparison group of 7 dogs (Saridomichelakis et al., 2005) with no significant difference between groups based on only parasite detection. Proportion of dogs infected with L. infantum infection based on PCR detection Data were replicated from the primary outcome and overall parasite detection as above (Saridomichelakis et al., 2005), with no significant difference between groups based only on parasite detection. Proportion of dogs infected with L. infantum based on serological detection One study reported data on this outcome with 2 positive findings in the intervention group of 15 dogs and 0 positive findings in the comparison group of 7 dogs, a non-significant result (OR 2.78, 95% CI 0.12, 65.79, ARR -0.09, 95% CI -0.34, 0.26) (Saridomichelakis et al., 2005). Proportion of dogs that died infected with L. infantum One study reported data on this outcome with 0 positive findings in the intervention group of 120 dogs and 0 positive findings in the comparison group of 120 dogs, a non-significant result (OR 1.00, 95% CI 0.02, 50.81) (Saridomichelakis et al., 2005). Proportion of clinically ill dogs infected with L. infantum Two studies reported data on this outcome (Table 24), with a non-significant result from the Gomez- Ochoa et al. (2012) study on domperidone. The other study on domperidone by Llinas at al. (2011) reported a significant difference between groups based on clinical illness with a protective effect. Table 24: Measures of effect for the proportion of clinically ill dogs infected with L. infantum for prophylactic medications. Study Intervention Comparator Measures of effect Domperidone (Gomez-Ochoa et al., 2012) Domperidone (Llinas et al., 2011) (0.004, 1.11) 0.06 (0.01, 0.12) (0.05, 0.42) 0.37 (0.18, 0.52) Supporting publications 2013:EN

63 Adverse effects related to the intervention One study reported no adverse effects (Gomez-Ochoa et al., 2012). Impact, modelling and control of canine leishmaniosis in the EU Other secondary outcomes None of the studies on prophylactic medications measured quantitative PCR, quantitative serology, the proportion of dogs infected with L. infantum based on cellular immunity tests or the size of the delayed type hypersensitivity reaction. Supporting publications 2013:EN

64 Overall summary A summary figure of the ORs of the studies judged to be at the lowest risk of bias (the highest number of low risk/lowest number of high risk quality assessment criteria) for each intervention group is presented below (Figure 18). Two studies had an unclear risk of bias (Genaro et al., 1996, Otranto et al., 2010), while the other two were judged to be at a high risk of bias (Gavgani et al.,, 2002, Saridomichelakis et al., 2005). Figure 18: Summary forest plot of the Odds Ratio and 95% confidence interval for the proportion of dogs infected with L. infantum based on serology or parasite detection for the studies judged to be at the lowest risk of bias for each intervention group. * = OR and 95% CI estimated based on Haldane correction. Supporting publications 2013:EN

65 The summary figure of the ORs of the most efficacious intervention from each intervention group is presented below (Figure 19) (Llinas et al., 2011, Maroli et al., 2001, Nogueira et al., 2005b, Otranto et al., 2010). The risk of bias for the Otranto et al. (2010) paper was judged to be unclear while it was high for each of the other studies. Figure 19: Summary forest plot of the Odds Ratio and 95% confidence interval for the proportion of dogs infected with L. infantum based on serology or parasite detection for the studies judged to be the most efficacious for each intervention group. * = OR and 95% CI estimated based on Haldane correction. Supporting publications 2013:EN

66 CONCLUSIONS Summary of main results The aim of this review was to determine whether currently available preventative control interventions are efficacious at preventing natural L. infantum infection in dogs. In total, 23 trials were reviewed, consisting of 12 studies on vaccinations (5 RCTs), 5 on collars (2 RCTs), 4 on spot-ons (2 RCTs) and 3 on prophylactic medication (all RCTs). Substantial variations within the studies regarding baseline characteristics of the dogs involved, such as age, breed and gender, were observed. It was decided not to pool the results by performing meta-analysis because of the heterogeneity in the studies, particularly the study designs and study areas (investigating differing Leishmania infections with different incidence in different regions) and the potential sources of bias identified. The primary outcome was defined as the difference in proportion of dogs infected with L. infantum based on either serology or parasite detection. Six of the 13 studies regarding vaccination had a significant protective effect; two studies on Leishmune (Lima et al., 2010, Nogueira et al., 2005b), 200µg ALM protein (Mohebali et al., 2004), CaniLeish (Oliva et al., 2012), LiESAp with MDP (Lemesre et al., 2007) and ALM with BCG (Mohebali et al., 1999). Two studies had significant nonprotective effects, one on FML (da Silva et al., 2000) and one using lyophilized protein (Dunan et al., 1989a). Two studies [da Silva et al. (2000), Borja-Cabrera et al. (2002)] reported data on the FML ELISA results for the FML QuilA vaccine and FML vaccine respectively, resulting in a 100% positive reaction observed for the intervention group and no observation of a humoral response. There was statistically significant evidence from four of the five included studies that deltamethrin collars had a protective effect (Ferroglio et al., 2008, Foglia Manzillo et al., 2006, Gavgani et al., 2002, Maroli et al., 2001), and there was one non-significant result. Four of the five spot-ons studies had a significant protective effect, one using 65% permethrin (Ferroglio et al., 2008), the 10% imidacloprid and 50% permethrin evaluated by Otranto et al. (2010) and both the 28 ± 2 day and 14 ± 2 days treatments with 10% imidacloprid and 50% permethrin evaluated by Otranto et al. (2007b). Only one of the three studies on prophylactic medication showed a statistically significant effect, with a protective effect evident from domperidone (Llinas et al., 2011). More specific new evidence pertaining to overall parasite detection was obtained from a further five studies on vaccination; with two indicating a significant protective effect for CaniLeish (Oliva et al., 2012) and for Leishmune (Nogueira et al., 2005b), and one a significant non-protective effect for lyophilized protein (Dunan et al., 1989a). No data regarding overall parasite detection were provided from the studies on collars or spot-ons, and no new data on prophylactic medication were provided. Only studies on vaccination presented new specific data regarding PCR; with one study presenting a significant protective effect for CaniLeish (Oliva et al., 2012) and one a significant protective effect for Leishmune (Nogueira et al., 2005b). No data regarding PCR were provided from the studies on collars or spot-ons, and no new data on prophylactic medication were provided. There were no data for the review outcome on quantitative PCR from any of the included studies. New evidence pertaining only to serological diagnosis was obtained for a further six results for vaccination, with only two statistically significant results; one study found a protective effect for 200µg ALM protein (Mohebali et al., 2004) while the other study found a non-protective effect for lyophilized protein (Dunan et al., 1989a). New serological specific data from one study on collars revealed a significant protective effect (Gavgani et al., 2002). There were no new data on spot-ons, while there was a non-significant effect for allopurinol prophylactic medication. Only three studies on vaccination reported data on quantitative serology, with two studies reporting a higher reading in the intervention group for FML and the FML QuilA vaccines (da Silva et al., 2000, Borja-Cabrera et al., Supporting publications 2013:EN

67 2002) as these studies evaluated the humoral response to the vaccine components, and one study a higher reading in the control group (Lemesre et al., 2007) for the LiESAp with MDP vaccine. Only vaccination studies provided information on cellular immunity tests, specifically DTH, with two studies presenting a statistically significant protective effect for FML and FML with QuilA (da Silva et al., 2000, Borja-Cabrera et al., 2002) and the same two studies reporting data on the size of the DTH reaction, both showing a higher reading in the intervention group (da Silva et al., 2000, Borja-Cabrera et al., 2002). No studies reported data on other ex-vivo cellular immunity tests such as cytokine production in supernatants. Data regarding the proportion of dogs that died with L. infantum infection were provided by five studies on vaccination, with two statistically significant protective effects, for Leishmune and the FML with QuilA vaccine (Borja-Cabrera et al., 2008, Borja-Cabrera et al., 2002). There was a nonsignificant effect for allopurinol prophylactic medication. Data on the proportion of dogs with clinical illness were provided from seven studies on vaccination with two significant results indicating a protective effect for the Leishmune (Nogueira et al., 2005b) and FML QuilA vaccine (Borja-Cabrera et al., 2002). There was a non-significant effect for one study that reported this outcome for collars. Two studies reported this outcome for the prophylactic medication domperidone, with one reporting a non-significant effect (Gomez-Ochoa et al., 2012) and one reporting a significant protective effect (Llinas et al., 2011). Outcomes regarding adverse effects observed in the intervention group were recorded from a subset of studies in each intervention group. For vaccination, no adverse effects were noted for the MML with MPL or for the MML with MPL vaccines (Gradoni et al., 2005), while adverse effects were noted for the ALM and lyophilized protein vaccines (Mohebali et al., 2004, Dunan et al., 1989a). One study on deltamethrin collars reported local adverse reactions (Aoun et al., 2008), while the other reported no adverse effects (Maroli et al., 2001). Imidacloprid and permethrin spot-on did not cause any adverse effects in the studies which reported on this outcome (Otranto et al., 2007, Otranto et al., 2010), however no information was gathered on permethrin only spot-ons. Domperidone caused no side effects in the one study which reported data on this outcome (Gomez-Ochoa et al., 2012). No information was gathered on side effects related to allopurinol. The most efficacious interventions in each intervention group were each judged to be at a high risk of bias (Llinas et al., 2011, Maroli et al., 2001, Nogueira et al., 2005b, Otranto et al., 2010), with the exception of Otranto et al. (2010) which was judged to be at an unclear risk of bias. These studies displayed significant protective effects for Leishmune vaccination, deltamethrin collars, 10% imidacloprid and 50% permethrin spot-ons and domperidone prophylactic medication, and therefore warrant further investigation with future well-designed RCTs. Judging the studies on the basis of quality assessment, most reliance could be placed on the results of the studies by Genaro et al., 1996, Otranto et al., 2010 which both had an unclear risk of bias. Overall, therefore, the best evidence for control measures for CanL comes from the study by Otranto et al. (2010) which displayed a significant protective effect for 10% imidacloprid and 50% permethrin spot-on (OR 0.01, 95% CI 0.001, 0.20, ARR 0.37, 95% CI 0.24, 0.50) based on the proportion of dogs infected with L. infantum based on either serology or parasite detection. Supporting publications 2013:EN

68 1.11. Overall completeness and applicability of evidence Impact, modelling and control of canine leishmaniosis in the EU The interest in applying effective control measures for CanL is increasing, with ongoing research apparent in vaccination and prophylactic medications. Vaccination was most frequently considered in this review; however there is apparent recent increase in published evidence regarding the efficacy of prophylactic medications Quality of the evidence Evidence concerning all four groups of control measures reviewed is poor due to the risk of bias in existing studies. An a priori sample size was not calculated for most of the studies and the lack of statistical power limits the adequate evaluation of efficacy of control measures for CanL. Overall, the studies on control measures for CanL were judged to have a high risk of bias for the study design aspects, in particular randomisation, allocation concealment and blinding, and for the acknowledgement of financial support. There was an unclear risk of bias for selective reporting. While the study protocol allowed for the inclusion of case-control and cohort study designs none of these was identified in the final screening process. The inclusion of OCTs as well as RCTs will have affected the overall bias, as the scoring system was heavily biased towards gold-standard RCTs which provide the best level of evidence Potential biases in the review process Several aspects of the review process may have adversely affected the selection and assessment of the reviewed studies. CanL can present as a wide spectrum of clinical manifestations. Significant variability in the diagnostic processes between studies may have led to heterogeneity of the clinical status within the enrolled population, potentially influencing the response of dogs to the interventions. The range of study publication dates may have influenced the reliability of the diagnosis, as advancement of serological and parasitological methods of diagnoses have advanced since the earliest publication in 1989 (Dunan et al., 1989a) and the latest in 2012 (Gomez-Ochoa et al., 2012, Oliva et al., 2012). Various signalments of dog were represented in the studies which reported baseline characteristics and these differences may have been potential sources of variability in the treatment responses. Serological assessment of vaccine efficacy may have affected the results of this study. Some of the available vaccines (such as FML) elicit a humoral response that with available current methods cannot be differentiated between active infection and vaccination, whereas other vaccines (such as LiESAp) do not. The current review included studies published in abstract form, as well as full-papers, due to published evidence regarding newly available interventions, such as the first licensed European vaccine (CaniLeish ) and domperidone prophylactic medication. It is acknowledged that it was difficult to assess the risk of bias in studies published as abstract, as only limited data can be provided in this format. The assessment of risk of bias would be improved by standardised assessment of full papers for each intervention. More robust review of interventions that are to date only published in abstract form will hopefully be available in the near future when full-text papers are published. Due to time constraints it was not possible to request further information from publication authors, which may have improved the assessment of each of the criteria, nor was it possible to obtain further grey literature which may have identified further relevant studies. Furthermore, we did not seek the trial protocols in order to aid our assessment of selective reporting and we only included studies which could be readily translated into languages spoken by members of the group. These limitations should be considered for future reviews. Supporting publications 2013:EN

69 1.14. Agreements and disagreements with other studies or reviews Two previous systematic reviews regarding CanL have been published (Romero and Boelaert, 2010, Noli and Auxilia, 2005). Noli and Auxilia (2005) aimed to identify evidence of efficacy of molecules and/or therapeutic protocols to treat or prevent CanL. They identified three field trials investigating the preventive effect of repellent collars or spot-ons, all of which were also identified in this review (Gavgani et al., 2002, Giffoni et al., 2002, Maroli et al., 2001). They found good evidence for recommending deltamethrin collars (Gavgani et al., 2002, Maroli et al., 2001) and fair evidence for recommending spot-on permethrin (Giffoni et al., 2002), however they did not undertake metaanalysis and used their own risk of bias scoring system. Romero and Boelaert (2010) undertook a review of the effectiveness of novel visceral leishmaniasis control tools and strategies in Latin America. They identified 14 trials, of which only two were identified in this review (da Silva et al., 2000, Giffoni et al., 2002), as they considered culling an acceptable control measure, evaluated the effect of repellent interventions on sand fly populations and human incidence as outcome measures, which the present review did not. Romero and Boelaert (2010) did not undertake risk of bias assessment or meta-analysis. The present review offers a Cochrane standard risk of bias assessment, a more recent search, and focuses on all types of control measures used at the dog level for CanL. RECOMMENDATIONS Implications for practice There are studies to support the use of control measures to prevent L. infantum infection in both endemic and non-endemic areas, in particular vaccination with 200µg ALM protein, Leishmune, CaniLeish, LiESAp with MDP, and ALM with BCG, deltamethrin collars, 65% permethrin, 10% imidacloprid with 50% permethrin spot-ons and domperidone prophylactic medication. However, the risk of bias with the publications on these interventions needs to be considered. It is interesting to note an apparent lack of evidence of efficacy for the Brazilian commercial licensed Leish-Tec vaccine, a recombinant A2-antigen of Leishmania amastigotes adjuvanted by saponin vaccine Implications for research Well-designed, adequately powered RCTs are needed to determine whether using control measures for CanL confers prophylactic benefits. There is ongoing work to determine the efficacy of prophylactic medications and these results will provide further, valuable evidence. Supporting publications 2013:EN

70 2. Assessment of the role of animals in the spread of L. infantum within the European Union INTRODUCTION AND OBJECTIVES 2.1. Review and evaluation of existing simulation models Introduction The work conducted for the assessment of the role of dogs in the spread of L. infantum within the European Union (EU) consisted of two phases: a review and evaluation of existing simulation models, followed by the actual simulation modelling work. The objective of the initial review was to identify existing model(s) that could be used or adapted for the simulation modelling described in section 2.2. MATERIALS AND METHODS Review Although the review conducted was not a systematic review, it drew on a systematic approach. The models identified via the review were then evaluated using a set of pre-defined criteria further detailed below and in Appendix C. The search of existing simulation models 2 was performed in August 2012 in various databases, including CAB 3, ISI Web of Knowledge , and PubMed 5. Search terms included (leishmani* AND model*) AND (epidemiol* OR dynamic*) limited to titles or abstracts. A two-stage screening process was applied to exclude articles and models not relevant to the project. In addition, selected articles on other vector-borne disease modelling were also reviewed, either because they may be adapted to Leishmania, or to compare with the modelling approaches of the articles retrieved through the search described above Evaluation of existing simulation models Models identified as potentially suitable for the modelling component of the project were evaluated using the following pre-defined approach and set of criteria: A brief summary of key technical characteristics was provided; An evaluation of time limitation was performed using a scoring for: platform/software used, availability of model codes, ease of adaptation in EFSA-supported modelling platforms, and ease of use and adaptation (see Appendix C. ); Data needs were recorded; The models structures, technical characteristics, advantages and limitations were briefly described; 2 As described in the project specifications, the search conducted was not a formal systematic review of all possible simulation models Using EndNote Web ( 5 Supporting publications 2013:EN

71 RESULTS Impact, modelling and control of canine leishmaniosis in the EU Finally, the models identified as most suitable for the purpose of the project were subjected to a data quality assessment and an assumption assessment, both based on simplified versions of the NUSAP/Pedigree method (Boone et al., 2009, Boone et al., 2010). For the data quality assessment, each of the model parameters were scored a 0-4 scale (see Appendix D. for Pedigree matrix), and an overall model score was obtained by averaging the individual parameter scores. For the assumption assessment, the main models assumptions were scored on a 0-4 scale using the Pedigree matrix shown in Appendix E. The overall model score for the assumption assessment was obtained by summing the individual assumption scores. Figure 20 summarises the results of the search conducted on leishmaniosis modelling. Literature search: (leishmani* AND model*) AND (epidemiol* OR dynamic*) Databases: CAB, MEDLINE, ISI Limits: Titles or Abstracts Search results combined (n= 294) Articles screened on basis of title and abstract Included (n=32) Articles screened on basis of entire content 10 Articles with potentially useful model structures Excluded (n=262) 162 Molecular modelling 33 Risk factor 28 Spatial/temporal modelling 9 Control programmes 8 Impact of climate or climate change 6 Vaccine 5 Descriptive epidemiology 3 Non-Leishmania 4 Theoretical articles 2 Ecological modelling 2 Diagnostic test Excluded (n=22) 7 Ecological modelling 5Different modelling objectives 4 Descriptive epidemiology 2 Impact of climate or climate change 1 Control programmes 1 Economic modelling 1 Risk factor Figure 20: Summary of search strategy and findings Ten articles were finally retained as potentially suitable for the simulation modelling component of the project (Table 25). All 10 articles described deterministic, compartmental SIR-type of models. Two articles modelled CanL; the others simulated visceral leishmaniosis (3), cutaneous leishmaniosis (3), and American integumentary leishmaniosis (1) in humans and one did not specify the clinical Supporting publications 2013:EN

72 presentation or species modelled. In three articles, transmission dynamics were modelled using the concept of vectorial capacity, while the others (7) used contact rates. None of the models listed a platform supported by EFSA. None of the articles provided the model code, and some articles did not provide specific details of the model equations and parameters. The models simulating leishmaniosis in humans were assessed as more difficult to adapt for the purpose of the project, because they did not specifically consider the dog population in the transmission dynamics. Tables showing the key characteristics of the models and detailed time limitation scores are provided in Appendix F. to I. Table 25: List of articles identified as potentially relevant to the modelling work. ID Article 1 Hasibeder, G., C. Dye, and J. Carpenter Mathematical-modeling and theory for estimating the basic reproduction number of canine leishmaniasis. Parasitology. 105: Ben Salah, A., H. Smaoui, L. Mbarki, R. M. Anderson, and R. Ben Ismail Development of a mathematic model of the dynamics of the transmission of canine leishmaniasis. Archives de l'institut Pasteur de Tunis. 71(3-4): Gelvez Pinto L. N., J. C. Muskus, H. H. Andrade, and G. Munoz Mathematical simulation model for the study of the transmission dynamics of the leishmaniasis under a systemic approach, The 12th International Conference of the System Dynamics Society, eds. C. Monaghan and E. Wolstenholme (1994). 4 Dye, C The logic of visceral leishmaniasis control. American Journal of Tropical Medicine and Hygiene. 55(2): Palatnik-De-Sousa, C. B., L. M. Batista-De-Melo, G. P. Borja-Cabrera, M. Palatnik, and C. C. Lavor Improving methods for epidemiological control of canine visceral leishmaniasis based on a mathematical model. Impact on the incidence of the canine and human disease. Anais Da Academia Brasileira De Ciencias. 76(3): Bacaer, N. and S. Guernaoui The epidemic threshold of vector-borne diseases with seasonality: the case of cutaneous leishmaniasis in Chichaoua, Morocco. Journal of Mathematical Biology. 53(3): Rosales, J. C. and H. Yang Estimation of the basic reproducibility number for American tegumentary leishmaniasis in two sites in northeastern Salta Province, Argentina. Cadernos de SaudePublica 23(11): Chaves L. F., M. J. Hernandez, and S. Ramos Simulación de modelos matemáticos como herramienta para el estudio de los reservorios de la leishmaniasis cutánea americana. Divulgaciones Matemáticas. 16 (1): Elmojtaba, I. M., J. Y. T. Mugisha, and M. H. A. Hashim Mathematical analysis of the dynamics of visceral leishmaniasis in the Sudan. Applied Mathematics and Computation. 217(6): Agyingi, E. O., D. S. Ross, and K. Bathena A Model of the transmission dynamics of leishmaniasis. Journal of Biological Systems 19(2): The narratives for all 10 articles are available in Appendix J., below is a summary for the three models considered more suitable for the objectives of the project (Ben Salah et al., 1994, Dye, 1996, Rosales and Yang, 2007). Ben Salah et al. (1994) describe a deterministic SEI model restricted to dog and sand fly populations. The authors acknowledge data gaps and conducted a sensitivity analysis to assess the impact of uncertainty in model parameters on its outputs. When seasonality is introduced to the model, outputs match prevalence estimates observed in the field. Dye (1996) used a deterministic SEIR model to describe the transmission of L. infantum in dogs and humans in Brazil, using vectorial capacity rather than explicitly modelling the disease dynamics in the Supporting publications 2013:EN

73 sand fly population. The objective of the work was to evaluate the effectiveness of various control methods for canine and human visceral leishmaniosis. The model structure is appropriate and relevant to the project. The approach used to assess the relative efficacy of mitigation measures was incorrectly performed, and would not be applicable to the current project. The model is well specified and relatively easy to implement in its deterministic form. Rosales and Yang (2007) developed a deterministic SIR model to simulate the transmission of American integumentary leishmaniosis in humans in an endemic area of Argentina. Human, dog and vector populations were represented, and transmission was modelled using contact-based parameters. The objective of the model was to estimate the basic reproduction ratio and the force of infection during the epidemic period of the disease. Data quality and assumption assessments were performed on the three above mentioned models and the detailed results are presented in appendices K. to O. The overall mean data quality score of the model by Dye (2.8) was slightly higher than the ones of the models by Rosales and Yang (2.7), and Ben Salah et al. (2.4). Overall mean assumption scores were also very similar (2.7 for Ben Salah et al., and 2.6 for both Dye and Rosales and Yang). DISCUSSION The model review and evaluation yielded 10 models potentially relevant for the modelling of CanL in the EU. Although the approach used was not a systematic search, it followed pre-defined search term and evaluation criteria to allow for a fair identification and comparison of simulation models on leishmaniosis. Three models were retained for a second-step evaluation after an initial screening of the model scope and general approach, number of parameters and aspects likely to affect the timeframe of the project. The model by Dye (1996) was assessed as the most suitable for the purpose of the simulation modelling component of this project. Reasons for this included: appropriate structure, use of vectorial capacity rather than contact-based transmission parameters, good specification and relative ease of implementation in a deterministic form, and possibility to add seasonal variation and implement mitigation measures. The small difference of scores for the data quality and assumption assessment did not justify using these criteria to favour or discard any of the three models under scrutiny. The structure and transmission dynamics of the Dye (1996) were used to inform the development of the model presented in section 2.2. However, the model presented hereafter was largely expanded and adapted to reflect the context of the EU. Supporting publications 2013:EN

74 2.2. Simulation modelling INTRODUCTION AND OBJECTIVES The objectives of the simulation modelling were to: Estimate the probability of CanL endemicity in the presence of a competent vector in a previously CanL-free area, following the introduction of infected dogs; Evaluate potential mitigation measures. The overall risk of CanL endemicity in a previously free area was determined in three steps. Firstly, the probability of introducing an infected dog into a non-endemic area was calculated for two pathways of introduction: household travelling to endemic areas with their dogs, and dogs from endemic area imported into non-endemic areas (e.g. commercial imports, adoptions, individual purchases). Then, CanL transmission in a previously disease-free area in the presence of a competent vector was simulated to estimate the probability of endemicity in a hypothetical independent contact network of dogs. Finally, the overall risk of endemicity in an area composed of one or more independent contact networks (P EndRegion ) was calculated using basic probability theory. Figure 21 shows a conceptual representation of the scenarios evaluated. Supporting publications 2013:EN

75 Figure 21: Conceptual representation of the scenarios developed for the assessment of the role of dogs in the spread of L. infantum within the European Union. Grouped dogs represent an independent contact network of dogs 6, and dogs infected with L. infantum are highlighted in orange. Blue and orange arrows represent movements of non-infected and infected dogs, respectively, between endemic and non-endemic areas. Infected dogs can be introduced into non-endemic areas via commercial imports or adoptions or purchase of dogs from endemic areas, or household trips to endemic areas. Once introduced into disease-free areas, transmission of canine leishmaniosis may occur in the contact network of the infected dogs in the presence of competent sand fly vectors. The overall risk of introduction and endemicity is then a function of the number of infected dogs introduced in diseasefree areas and the probability that L. infantum will be transmitted and maintained in contact networks of dogs. A number of mitigation measures were implemented and their impact on the overall risk of CanL endemicity in a previously free area was assessed. The following mitigation measures were included alone or in combination: vaccination, prophylactic medication, repellent, insecticide, and diagnostic test and exclusion (i.e. refusal to introduce the dog in the disease-free area). The model was implemented in MS Excel, using the Monte Carlo 6.0 add-in (Palisade Corporation, Ithaca, NY). It was parameterised using data from the literature, publicly available databases (e.g. EUROSTAT) and expert opinion. Coordination with project team members 6 A contact network of dogs represents a group of dogs that interact and may be able to transmit CanL to each other throughout a competent vector (e.g. dogs in a neighbourhood, or a dog park). Supporting publications 2013:EN

76 working on the systematic review and the impact assessment ensured the use of the most relevant data available. The following CanL infection states in dogs were considered in the model. Susceptible: Dogs have not been infected with Leishmania promastigotes but can get infected. Latent (Infected): Dogs exposed to and infected with Leishmania amastigotes but not yet Infectious 7. Latent dogs may potentially progress to the Infectious state. Infectious (Clinical and Sub-Clinical): Infected dogs able to transmit Leishmania amastigotes to Susceptible dogs. Infectious Sub-Clinical: Infectious dogs not showing clinical signs of the disease or clinicopathological abnormalities consistent with CanL. Infectious Clinical: Infectious dogs with clinical signs of the disease and/or clinicopathological abnormalities consistent with CanL. Such dogs are more likely to be brought to a veterinarian, be diagnosed, and treated against leishmaniosis, and are thus modelled separately from Infectious Sub-Clinical dogs. Resistant: Dogs exposed to Leishmania promastigotes that developed an immune response (either via effective vaccination or treatment) preventing them from becoming Infectious. Dogs in this state remain non-infectious for the duration of the simulation. MATERIALS AND METHODS Sections to provide a shortened description of the different modules of the simulation model. More detailed descriptions of the model structure and parameterisation are provided in Appendix P Modelling the probability of endemicity in a disease free area following the introduction of an infected dog. Probability of introducing an infected dog into a non-endemic area Two main pathways were considered to lead to the introduction of infected dogs in disease-free areas: i) dogs from endemic areas that are commercially imported for sale in non-endemic areas, or dogs from endemic areas adopted or purchased by owners residing in non-endemic areas (P InfCA ) and ii) households from disease-free areas travelling to endemic areas with their dogs (P Inf ). These two pathways were modelled differently to accommodate for different probabilities of infection with L. infantum. For example, on average, dogs travelling to endemic areas could be exposed to sand flies for shorter periods than dogs residing in endemic areas. Probability of importing an infected dog into a non-endemic area (P InfCA ) Although dogs from endemic areas may have different levels of prevalence depending on housing, age, and the population of sand flies, modelling this heterogeneity would require prevalence estimates within different population strata such as outdoor, indoor, and kennel dogs. As such prevalence estimates are scarce or non-existent, it was assumed that any given dog from an endemic area was 7 Although this definition of latency might not be consistent with the terminology used in the parasitology and virology literature, it was chosen for consistency with transmission modelling terminology. Supporting publications 2013:EN

77 equally likely to be exported into non-endemic areas. Under this assumption, P InfCA was simply the true prevalence of CanL in endemic areas. The uncertainty of the P InfCA was estimated using a Bayesian latent class model from 3 cross-sectional studies reporting seropositivity to the IFAT test in endemic areas (Leontides et al., 2002, Keck and Dereuer, 2003, Baldelli et al., 2011). Probability of a dog returning infected from an endemic area (P Inf ) An Index Case module was developed to simulate the scenario where households from a CanL-free area travel with their dog(s) to an endemic area for a certain period of time during which the dog(s) can acquire L. infantum infection and travel back to the CanL-free area (Figure 22). This scenario returns the probability of a travelling dog becoming infected during a single household trip to a leishmaniosis-endemic area (P Inf ) Figure 22: Conceptual representation of the modelling framework (Index Case module) used to simulate probability of a dog returning infected from an endemic area, and potential mitigation measures The Index Case module was implemented using an individual-based stochastic continuous-time state transition modelling framework (see differential equation representation in Figure 23 and Appendix P. ). The module includes a winter period during which sand flies are absent and no transmission of CanL occurs. The infection state of each dog was evaluated daily for the duration of the trip, and the model record whether or not the travelling dog was infected at the end of the trip. This module was very similar to the Transmission module used described in the following section, but it is simpler as it assumes that the number of infectious and susceptible dogs and sand flies was constant for the duration of the trip, and only simulates CanL progression on the dog(s) travelling. Supporting publications 2013:EN

78 SAND FLIES Season Replacement rate (α) Vectorial Capacity (VC) Mortality (δ) Mortality (δ) Susceptible (S) Latent (L) ς Infectious Sub-Clinical (I) Mortality (δ) = -VC*(I + C)*(S/N) + αδn δs ρ = VC*(I + C)*(S/N) - (δ + ς)l = ςl - (δ + ρ)i Infectious Clinical (C) Mortality (δ) = ρi - δc N=S+L+I+C Figure 23: Conceptual design of the modelling module simulating the transmission of CanL in a contact network of dogs in the presence of sand flies where: σ is the rate of transition from Latent to Infectious Sub-Clinical, ρ is the rate of transition from Infectious Sub-Clinical to Infectious Clinical Probability of endemicity within a contact network of dogs (P End ) The Transmission module calculates the probability (P End ) of CanL endemicity in a previously free area, within an independent contact network of dogs and in the presence of sand flies, given that an infected dog was introduced (Figure 23). Endemicity in a previously free area was considered established when at least one dog (other than the index case) in the contact network was infected at the end of the simulation period. Therefore, endemicity in this modelling work does not provide information about the prevalence of CanL. The Transmission module was an individual-based stochastic continuous-time state transition modelling module similar to the Index Case module but with a different parameterisation. The infection state of each dog was simulated during a 3-year period from the day that the infected dog introduced into the contact network becomes Infectious. The module incorporates seasonality, with a winter period during which sand fly vectors are absent and no transmission of CanL can occur via vector. The model records the number of Latent and/or Infectious dogs at the end of the simulation, and endemicity was considered established when this number was larger than one. P Inf and P End are calculated in the Index Case and Transmission modules respectively using numerical integration methods. Therefore, a two-dimensional modelling approach (Cohen et al., 1996) was used to report the uncertainty in P Inf and P End, in the form of 95% predictive intervals (PI). By randomly sampling values from the uncertainty distributions in an outer loop and then simulating the variability in an inner loop, this approach allows for the reporting of P Inf and P End, conditional on the uncertainty distributions. This was critical to obtain the correct variance in the uncertainty distribution of the overall probability of endemicity in the region, the final output of the simulation model. Supporting publications 2013:EN

79 Although the capability to simulate more than one dog per trip was included in the Index Case and Transmission modules, for simplicity here P Inf and P End are reported considering only one dog travelling or introduced per simulation. Results including multiple dogs per simulation are also reported in selected scenarios Probability of endemicity within a region (P EndRegion ) A stochastic module was developed to derive the overall probability of endemicity in at least one independent contact network in a CanL previously free area with sand flies (P EndRegion ). Again, in this modelling work endemicity in a region was attained when at least one dog (other than the index case) in one contact network was infected with CanL, and endemicity was thus not referring to a certain level of CanL prevalence. It was assumed that a proportion of the total number of dogs introduced into disease-free areas (following commercial imports, adoptions, individual purchases from endemic areas, or travels to endemic areas) are infected with CanL, and a proportion of those infected may transmit the infection to other dogs in their contact network and generate endemicity in the previously free area after a certain period of time. This analysis assumes that all dogs from endemic areas are equally likely to be exported (i.e. randomly sampled), all dogs within a CanL-free area are equally likely to travel to endemic areas, imported or returning dogs are independent from each other (i.e. that any contact network has the same chance of receiving an infected dog imported from an endemic area), and that the number of dogs moved was relatively a small proportion of the total dog population in the exporting country Vectorial capacity Further details of the modelling approach are included in Appendix P., section Therefore, only the conceptual aspects of the modelling of spread of CanL are discussed here. In contact-based infectious disease models, one of the key parameters is, the rate of transmission per encounter with an infectious individual (often called effective contact rate). As this rate was assumed to be constant and therefore, exponentially distributed, an infectious individual I has the potential to transmit the disease to N others per time t (where N was the total population). However, only a fraction S/N of contacts with susceptible (S) individuals can result in infection. Therefore, the expected number of new infections in time t is ( *N)*(S/N)*I, which is often presented in its simplified mathematical form SI. That is, the incidence of infection is directly proportional to the effective contact rate, the number of infectious individuals, and the fraction of individuals in the population that are susceptible to the disease. However, vector borne diseases such as CanL are not typically based on direct contacts between infectious and susceptible individuals of the same species, but often have to pass through a vector; in this case, the sand fly. One option to model such diseases is to simulate the disease dynamics in both the vector and the host(s) populations and use a contact based principle such as the one described above to include directional effective contact rates from host to vector and vice versa. However, this requires several parameters such as the directional effective contact rates and the total number of vectors in the population, which are often unknown or difficult to estimate from empirical data. Another alternative was to simplify the disease dynamics in the vector by using the concept of vectorial capacity (VC). Vectorial capacity was defined as the number of secondary cases resulting from an infectious case in time t (usually, t was represented in days or weeks) in a susceptible Supporting publications 2013:EN

80 population, which is constant and independent of the prevalence in the sand fly population. The basic formula for VC was: where: m: Number of female sand flies per dog. : Number of female sand flies bites per day. µ: Daily mortality rate of female sand flies. VC= τ: Daily rate of transition from latent to infectious sand flies. The term represents the probability that any given sand fly survives a day; it was assumed to remain constant throughout the life of a sand fly and therefore, exponentially distributed. Thus represents the probability that a surviving sand fly was able to transfer the infection via bites, as 1/τ was the average latent period in sand flies and τ was the daily rate of transition from latent to infectious sand flies. Multiplying the density m by the squared bite rate 2 (as a sand fly has to bite both an Infectious and a Susceptible dog to transmit the disease) by returns the number of infectious bites per sand fly per day, and dividing it by the daily mortality rate µ then provides the number of daily new dog infections resulting from one infectious dog in a susceptible population (VC), which was equivalent to the parameter of the contact based modelling concept described earlier. Therefore, using this same construction the expected total new cases in time t was proportional to VC*I*S/N. For example, assuming a constant population of 50 animals N, one Infectious animal, and VC= 1 the expected number of new cases will be 1. This was directly equivalent to the SI formulation shown for a basic contact-based model, but has the advantage of not requiring directional contact rates or vector population sizes. For this work, vectorial capacity was favoured over modelling the dynamics of disease in the vector population for the following reasons: VC was consistent with the three models selected in the modelling review stage of this work, and is widely used in vector borne disease modelling. For example Dye (1996) explains that VC is suitable to describe the transmission rate of L. infantum among dogs because the transmission dynamics occur at a much faster time scale in sand flies than in dogs. He also points out that few sand flies live long enough to acquire infection. Even if the dynamics in the sand fly population were modelled, the numbers would be so large that the effect of variability would be almost irrelevant (because of the central limit theorem), whereas the uncertainty in the parameters affecting this dynamic would remain important. For example, the variability in the sand fly population could be modelled with Poisson (λ), where λ was the density of sand flies in the area. If lambda =10, the standard deviation σ of that distribution is 10 and therefore its coefficient of variation (CV)=0.32. In contrast, if lambda=10000, the CV is only 01. However, if lambda is uncertain and follows for example, a Gamma (10000,0.5), the variance (scale) of the Gamma distribution will directly affect that of the Poisson density regardless of its mean, so the uncertainty is by far the biggest driver in this example. Supporting publications 2013:EN

81 Evaluating mitigation measures A number of mitigation measures against CanL were evaluated as described below, in Figure 24 and Appendix P. Table 26 provides a summary of the most relevant estimates and sources used to model efficacy of each mitigation measures modelled. As discussed in the systematic review section, the quality of studies reporting efficacy of the different mitigation measures on the dog was highly variable and not comparable. Therefore, rather than using a pooled (meta-analytical) estimate of efficacy across studies, for most mitigation measures one or several studies were considered as individual estimates and therefore, were sampled from mixture distributions where the distributions from each study were sampled with equal weight. Table 26: Mean and 95% confidence interval (95%CI) or 95% credibility interval (95%CrI) of important parameters used in the simulation model. Parameter Mean 95%CI or Source 95%CrI Treatment efficacy 68.5% (a) Slappendel and Teske (1997), Manna et al. (2008) Repellent efficacy 95.4% Killick-Kendrick et al. (1997) Vaccine efficacy 86.2% Lemesre et al. (2007) Insecticide efficacy 58.8% Coleman et al. (2011) Prophylactic medication efficacy 73.3% (Llinas et al., 2011) Vectorial capacity (a) Expert opinion (Anonymous, 2013) Prevalence of Leishmania Infected dogs in an endemic area 10.7% (a) Leontides et al. (2002), Keck and Dereuer (2003), Baldelli et al. (2011) Proportion of Leishmania Infected dogs that are Infectious 40.6% (a) Gálvez et al. (2010), Miró et al. (2012), Papadopoulou et al. (2005), Aoun et al. (2009) Travelling days (a) EuroStat (2013e) Time to transition from Latent to Infectious Sub-Clinical (days) Oliva et al. (2006) Time to transition from Infectious Sub-Clinical to Infectious Clinical Oliva et al. (2006) (days) Test sensitivity in asymptomatic 52.6% 30.8%-74.0% (a) Mettler et al. (2005a) dogs (a) 95% credibility interval Supporting publications 2013:EN

82 Season Vectorial Capacity (VC) SANDFLIES Insecticide (φ) Repellent (γ) Replacement rate (α) Mortality (δ) Mortality (δ) Susceptible (S) Latent (L) ς Infectious Sub-Clinical (I) Mortality (δ) Vaccine (θ) Prophylactic medication (π) = -VC*(1-γ)*(I + C)*S/N + αδn δs = VC*(1-γ)*(I + C)*(S/N) - (δ + ς)l - (θ + π)l = ςl - (δ + ρ)i = ρi - δc - ηc ρ Resistant (R) Mortality (δ) Test & Treatment (η) = (θ + π)l + ηc δr N = S+L+I+C+R Infectious Clinical (C) Mortality (δ) Figure 24: Conceptual model of the effect of mitigation measures on the transmission of CanL in a contact network of dogs in the presence of sand flies. Mitigation measures are in blue. Vaccination: In epidemiological modelling, resistance is typically assumed to occur before exposure to the agent; that is, a vaccinated animal becomes Resistant without going through the Latent state. However, after several conversations with the project team and with panel members, to better represent the current understanding on CanL s dynamics it is assumed that dogs vaccinated against CanL become Resistant after being naturally exposed to leishmania infection (i.e. after becoming Latent). This prevents them from becoming Infectious during the simulation period and thus reduces both the prevalence and the incidence of CanL. Vaccination was implemented in both the Index case and the Transmission module, using two parameters: (1) vaccine use, which represents the proportion of dogs vaccinated (or the probability that travelling dogs from a household are vaccinated in the case of the Index case module), and (2) vaccine efficacy, corresponding to the proportion of vaccinated dogs that will become Resistant (Table 27). Prophylactic medication: Prophylactic medication was modelled in the same way as vaccination (Table 27). The drug used to model prophylactic medication against CanL in dogs was Domperidone (Llinas et al., 2011). Repellent: Repellents reduce the transmission of CanL by directly decreasing the sand fly biting rate, thus reducing the vectorial capacity of sand flies for CanL. As for vaccination, repellent is implemented using two parameters (use and efficacy), and repellent use status is modelled independently for each dog in the Transmission module while the Index Case module assumes the same repellent use status for all travelling dogs from a same household. Both modules assume that when a repellent is used on a Supporting publications 2013:EN

83 dog, it is applied on a regular basis during the entire period modelled. Repellent efficacy is defined as the capacity of the repellent used to prevent bites from sand flies and thus limit Leishmania transmission. When used on Susceptible dog(s), a repellent reduces the biting rate of sand flies (and therefore, the vectorial capacity) proportionally to its efficacy. When used on Infectious dog(s), a repellent reduces the transmission of Leishmania from these dogs to susceptible sand flies. This is represented in the modules by reducing the total number of Infectious dogs with repellent by a proportion equal to the repellent efficacy which is assumed to be equivalent to reducing the number of sand flies bites on all Infectious dogs with repellent proportionally to the repellent efficacy (Table 27). This models indirectly the fact that fewer sand flies get infected from Infectious dogs, which in turn reduce the incidence of Leishmania infection. Insecticide: The environmental application of insecticides was modelled by decreasing the density of sand flies, which in turn decreases their vectorial capacity. As with vaccination and repellent, insecticide is modelled using two parameters: use and efficacy. However, unlike vaccination and repellent, Insecticide use is the probability that insecticide is applied in the entire contact network environment. When used, insecticides result in a decrease of the density of sand flies proportional to its efficacy (Table 27). For this, it is assumed that insecticide is equally effective on infected and non-infected sand flies. Diagnostic test and exclusion: This mitigation measure assumes that dogs imported from endemic areas are tested to detect and exclude infected dogs. Therefore, the testing and exclusion of infected dogs is modelled for all the import pathways. As with other mitigation measures, use and efficacy are included. Use is simply the proportion of imported animals that are tested, whereas efficacy is the sensitivity (Se) of the diagnostic test (i.e. the conditional probability of testing positive given that the dog tested is infected). Thus, the number of undetected infected animals is proportional to the probability of infection in the imported dog (P InfCa and P Inf for the import and travelling pathways respectively), the percentage of use of testing, and the probability of false negative tests, 1-Se (Table 27). More details on the modelling of testing and exclusion for each pathway are provided in Appendix P. section 3.5. Ideally, the diagnostic test should be performed just before or as soon as possible after moving dogs into a non-endemic area, to prevent transmission of the disease from infected dogs. However, the sensitivity of the diagnostic test may be lower for early stages of the infection, e.g. for dogs which have been infected during a short trip to an endemic area (Mettler et al., 2005a). Supporting publications 2013:EN

84 Table 27: List of parameters used to model the transmission of canine leishmaniosis in a non-endemic area following the introduction of an infected dog with competent vector and the transmission of canine leishmaniosis to susceptible dogs travelling to endemic areas. Parameter Short explanation (a) Modelling Transmission parameters Time to transition from Latent to Distribution of the time from dog infection (from Infectious (σ) sand flies) to infectiousness (ability to transmit the infection to Time to transition from Infectious Sub-Clinical to Infectious Clinical (ρ) sand flies, days) Distribution of the time for dogs to go from the Infectious Sub-Clinical to Infectious Clinical state (days) Probability Distribution Distribution Parameter (b) Variability Weibull Shape:1.34 Scale: Variability Weibull Shape:4.3 Scale:233 Sources Oliva et al. (2006) Oliva et al. (2006) Vectorial capacity Number of female sand flies per dog (m) Number of female sand flies bites per day ( ) Transition rate from latent to infectious sand flies (τ) Daily mortality rate of female sand flies (µ) Number of female sand flies per dog Number of female sand fly bites per day Daily rate of transition from latent to infectious sand flies Daily survival probability of sand flies Characteristics of environment Proportion of Prevalence of Infectious Leishmania dogs in endemic areas Infected dogs that (%) are Infectious Prevalence of CanL in endemic areas Length of the winter season Continued.. Prevalence of infected dogs in endemic areas (%) Number of days when the environmental conditions are not favourable to the reproduction of sand flies (days) Uncertainty Gamma Shape:866 Scale: Uncertainty Normal µ: 2.03 σ: 0.29 Uncertainty Normal µ:6.5 σ: 0.53 Uncertainty Uncertainty Uncertainty Scenario analysis Empirical samples with equal weight Mixture of two equally weighted Betas Posterior Bayesian estimation of true prevalence 21d 42.5d Beta(alpha: 30,beta: 1048), Beta(alpha: 20,beta: 400) Fixed ICM: 150 TM: 90 Branco et al. (2013) Expert opinion (Anonymous, 2013) Expert opinion (Anonymous, 2013) Expert opinion (Anonymous, 2013) Gálvez et al. (2010), Miró et al. (2012), Papadopoulou et al. (2005), Aoun et al. (2009) - Leontides et al. (2002), Keck and Dereuer (2003), Baldelli et al. (2011) Oliva et al. (2006) Supporting publications 2013:EN

85 Parameter Short explanation (a) Modelling Characteristics of environment Day of the year Day of the year when the travelling travel to the endemic area was performed Travelling days Number of days per trip to endemic areas per year (days) Mitigation measures Repellent use Proportion of a certain dog (p ) population (e.g. region, country) that uses repellent Repellent efficacy Prophylactic medication use (π) (%) Proportional reduction of sand fly bites on dogs using repellent (%) Proportion of dogs that are treated (e.g. region or country) to prevent Leishmania infection (%) Impact, modelling and control of canine leishmaniosis in the EU Probability Distribution Distribution Parameter (b) Variability Uniform Min:0, Max:365 Variability Empirical ( histogram frequencies) Scenario analysis Uncertainty Scenario analysis 1-3d-20.7% 4-7d-38.1% 8-14d-6.3% 15-28d-11.1% 29-91d-3.6% d-0.2% Sources Assumes that travel can happen any day with equal chance EuroStat (2013e) Fixed NA Scenario analysis 1-Log- µ:-3.14 Normal (c) σ:0.33 Killick- Kendrick et al. (1997) Fixed NA Scenario analysis Prophylactic medication efficacy Diagnostic test Se Vaccine use (θ) Vaccine efficacy Insecticide efficacy Treatment efficacy (η) Continued.. Proportion of the treated dogs that won't become infected when exposed to infectious sand flies (%) Sensitivity of the test (Proportion of the Sub- Clinical infected dogs that are positive to the test) (%) Proportion of dogs in a certain population (e.g. region, country) that have been vaccinated (%) Proportion of the vaccinated dogs population that won't become infected when exposed to infectious sand flies (%) Proportional reduction in sand fly population as a consequence of the use of insecticide in an area (%) Proportion of the treated dogs that are cured and become Resistant (%) µ: -1.44, σ: Uncertainty 1-Log- Normal (c) 0.45, Truncated: max:1 Uncertainty Beta Alpha:10 Beta:9 Scenario analysis Uncertainty (Llinas et al., 2011) Mettler et al. (2005a) Fixed NA Scenario analysis 1-Log- µ:-2.43 Normal (c) σ: 1.04, Truncated max:1 Uncertainty 1-Normal µ:0.41, σ:0.079 Uncertainty Mixture of two equally weighted Betas Beta (alpha:12 beta:8) Beta(alpha: 33, beta: 10) Lemesre et al. (2007) Coleman et al. (2011) Slappendel and Teske (1997), Manna et al. (2008) Supporting publications 2013:EN

86 Parameter Short explanation (a) Modelling Probability Distribution Distribution Parameter (b) Dog s Life Mortality rate (δ) Dogs' life expectancy distribution (days) Variability Weibull Shape:42.47 Scale: Age Age distribution dogs Variability Log-Normal µ:1898 at any given time σ:1241, (days) Number of dogs Distribution of the Variability Zerotruncated lambda:0.74 travelling per number of dogs per household household Poisson Time to Time for a new Variability Pert Min:3 replacement replacement dog to be Max:3650 brought home/ Mode:120 introduced into the home after a dog is lost/dead (days) Probability of Proportion of Uncertainty Beta Alpha:52 replacement (α) households that Beta:53 replace a lost/dead dog (%) (a) Parameters expressed in days were transformed to weeks for the Transmission module. (b) ICM= Index Case module, TM=Transmission module (c) 1- Uncertainty distribution of the (log) relative risk. See appendix Q. Sources O Neill et al. (2012) Gálvez et al. (2010) Slater et al. (2008) McCutcheon and Fleming (2002) McCutcheon et al. (2002) Scenario analysis The individual and combined effect of mitigation measures on the main outcomes of the model was assessed (Table 28). In the baseline scenario (scenario 1a, 1b and 1c,Table 28), no mitigation measure was used, and P End was assessed for contact network sizes of 50 and 200 dogs, and for 1 and up to 5 infected dogs introduced in the non-endemic area. The individual effect of each mitigation measure (scenarios 2 to 21,Table 28) was evaluated in 20%-increments in levels of use (e.g. vaccination use in 20, 40, 60, 80 and 100% of the dog population). The combined effect of the mitigation measures was implemented for equally high (80%) and medium to low (40%) levels of use. The scenarios with combined mitigation measures assumed independency of use between them. The dual combinations of mitigation measures selected for evaluation were: vaccination and repellent, vaccination and insecticide, repellent and prophylactic medication, repellent and insecticide, and insecticide and prophylactic medication (scenario 22 to 31, Table 28). The vaccination and prophylactic medication were not combined as their effect was modelled using the same mechanism (Section 2.2.3). An extreme, best-case scenario where vaccination, repellent and insecticide were combined and implemented at 100% was also included (scenario 32, Table 28). Finally, the results of a survey administered to clinical veterinarians in endemic areas (see impact assessment section) were used to create a realistic scenario with 10% vaccine use, 50% repellent use and 5% prophylactic medication use. Variations in network size and number of dogs introduced are also reported for this scenario. Results are summarised with a mean and 95% predictive interval (95%PI). Effectiveness of the mitigation measures were evaluated using the proportional reduction of the mean P Inf, P End or P EndRegion. Supporting publications 2013:EN

87 Table 28: Proportion of population using mitigation measures, and contact network sizes used to evaluate changes in the probability of infection in one or multiple dogs after travelling to an endemic area (P Inf ) and probability of endemicity after introduction of one or multiple canine leishmaniosis infected dogs in a non-endemic area with competent vectors (P End )(a). Scenario ID Number of dogs (b) Network Size (n) Mitigation(s) affecting P Inf Mitigation (s) affecting P End Vacc Repel. Ins. Vacc. (c) Proph. (c) Repel. Ins. (c). (c) Proph. (c) (c) (c) (c) 1a b c a b (a) Insecticide was implemented as a proportion of networks using it. (b) Number of dogs travelling with the household. (c) Vacc.: vaccination; Proph.: prophylactic medication; Repel.: repellent; Ins.: insecticide Sensitivity analysis A two-stage sensitivity analysis was performed to identify the parameters that were most influential in the estimation of P Inf and P End. In a first stage, a high level of uncertainty was assigned to each parameter in order to identify the most influential parameters prior to a detailed data collection. For example, efficacies were modelled using a Beta (1,1) distribution, providing maximum uncertainty in the efficacy. This helped prioritising the data collection and parameterisation efforts of the model. In a second stage, the sensitivity analysis was repeated, but using parameters estimated from the data Supporting publications 2013:EN

88 collected (Table 27). The objective of this second stage was to identify the parameters influencing most the uncertainty in the model outputs. The sensitivity analysis assessed the conditional effect of input parameter quantiles on the output means. The mean of P Inf and P End was evaluated for given fixed percentiles of each input distribution, while keeping all other variables random. This approach can be more informative than other traditionally used sensitivity analyses, as it provides the actual effect on the output values rather than the effect on the output variance (i.e. in variance component sensitivity analyses) and doesn t assume a linear relationship between inputs and outputs (e.g. Spearman correlation and regression-based sensitivity analyses) Model main assumptions The following is a condensed list of the most relevant assumptions made in the model reported here: The density of vectors depends on environmental factors such as temperature and humidity which are highly variable within and between regions. Given the lack of information on such heterogeneous parameters, the density of sand flies within the contact network was assumed to be constant throughout the breeding season. Similarly, it was assumed that no sand flies were present during the winter season and therefore there was no transmission between dogs via sand flies as vectorial capacity was null. The vector randomly bites dogs in the contact network with no preference for any type of dog or infection state. Similarly, the feeding behaviour of infected and non-infected sand flies is the same. Sand flies stay in vicinity of their breeding sites (Sharma and Singh, 2008, Killick-Kendrick, 1999). When repellent was used on dogs, the model did not consider that bites avoided by treated dogs may have been redirected to other dogs or hosts. Conversely, repellents also have an insecticide effect that increases the mortality of sand flies by 30 to 67% (Vulpiani et al. (2011), not represented in the model. It was assumed that Infectious Clinical dogs remain Clinical until treated or dead, and that once treated they become Resistant for life with a probability given by the treatment efficacy. The mortality rate of dogs was the same regardless of their infection status or infection stage and all the dogs that died had the same probability of replacement. As transmission occurred only between dogs and sand flies within a contact network of dogs, it was assumed that no other reservoirs and/or host played a significant role in the transmission dynamics. Also, the contact networks of dogs were independent units with an independent vector population, and transmission between contact networks was only possible via movement of infected dogs. The model assumes that all the mitigations measures that have a time limited efficacy were repeatedly implemented in order to maintain their effect for the duration of the simulation. When used on a dog, repellent and prophylactic medication were assumed to be used repeatedly during the simulation period and according to their indication of use. Vaccination was assumed to generate immunity for the duration of the simulation (3 years). Supporting publications 2013:EN

89 RESULTS Sensitivity analysis Figure 25A and Figure 26A show the mean P Inf and P End, respectively, for the 2.5 and the 97.5 percentiles of the input parameters. For example, the mean P Inf for the lowest 2.5 and the highest 97.5 percentiles of day of the year travelling was 0% and 22.0% respectively, suggesting that the season when the travel occurred had a strong influence on the probability of infection during a trip to endemic areas. Figure 25B and Figure 26B show the changes on the mean P Inf and P End, respectively, for every 5 th percentile of each input parameter. Finally, Figure 25C and Figure 26C show Spearman s rank order correlation coefficient between each input parameter and P Inf and P End, respectively. The most influential parameter for P Inf was the day of the year travelling (i.e. the effect of seasonality on vectorial capacity), followed by the number of travelling days in an endemic area and the prevalence of Infectious dogs in the endemic area (Figure 25). Other influential parameters include the efficacy of the mitigation measures, the time to transition from Infectious Sub-Clinical to Infectious Clinical, and vectorial capacity parameters such as the number of female sand flies per dog and number of female sand flies bites per day. The most influential parameter for P End was the day of the year when the index case dog became Infectious (i.e. season when trip was made). Other influential parameters included the efficacy of mitigation measures (vaccine, repellent, and insecticide), vectorial capacity (number of female sand flies per dog and number of female sand flies bites per day) and infection transition parameters (time to transition from Infectious Sub-Clinical to Infectious Clinical, and time to transition from Latent to Infectious Sub clinical, Figure 26). Supporting publications 2013:EN

90 Figure 25: Sensitivity analysis of parameters affecting the probability of infection after a trip to CanL endemic areas (P Inf ): A) Conditional effect of 2.5 th and 97.5 th percentiles of input parameters on the mean P Inf, B) Changes in the P Inf for 10 th percent interval changes in input parameters, and C) Rank order correlation between parameters and P Inf. Supporting publications 2013:EN

91 Figure 26: Sensitivity analysis of the parameter affecting the probability of endemicity in a contact network following the introduction of an CanL infected dog (P End ): A) Conditional effect of input parameter 2.5 th and 97.5 th percentiles on the mean P End, B) Changes in the P End by 5 th percent interval changes in input parameter percentiles, and C) Rank order correlation between parameters and the P End. Supporting publications 2013:EN

92 Frequency Frequency Frequency Frequency Impact, modelling and control of canine leishmaniosis in the EU The mean vectorial capacity was 0.42 ( ) per day when no mitigation measures were implemented. There was a 0.4% probability that this vector capacity was greater than 1. The mean vectorial capacity was ( ) and ( ) when repellent was used alone or in combination with insecticide, respectively. The mean vectorial capacity when insecticide was applied alone to reduce the density of sand flies was 0.17 ( ) (Figure 27). A B Vectorial capacity Vectorial capacity C 6 5 D Vectorial capacity Vectorial capacity Figure 27: Distribution of the vectorial capacity of sand flies for CanL using A) no mitigation measure, B) repellent, C) insecticide, and D) repellent and insecticide Probability of one dog returning infected from a trip to an endemic area (P Inf ) When no mitigation measures were used (baseline scenario), the probability that one dog travelling to Leishmania-endemic areas became infected was 7.8% ( %), which was lower than 8.9% ( %) in households travelling with 1 to 5 dogs. This is likely because the most of the dog s owners (69%) have only one dog (Table 29). Supporting publications 2013:EN

93 The analysis of the separate effects of mitigation measures showed that when 100% of the dogs were vaccinated before travelling the P Inf was 0.9% (0-5.1%), an 88.5% reduction compared to not using vaccine. In contrast, low levels of vaccine use (i.e. 20% vaccine use) reduced P Inf to 5.8% ( %), which was a 25.6% reduction compared to not using vaccine (Figure 28A, Table 29). When repellent was used in 100% of dogs travelling, the probability of infection during the travel was almost zero (0.03%, 95%PI: 0-0.3%), representing a 99.6% reduction compared to not using repellent (Figure 28B, Table 29). When repellent was used in 20% of dogs travelling to an endemic area, P Inf decreased to 5.6% ( %), which was a reduction of 28.2% compared to not using repellent. Similarly when repellent was used in 80% of the dogs travelling, P Inf decreased by 80.8% compared to not using repellent. When prophylactic medication was used in 20% and 80% of dogs, the P Inf was 6.3% ( %) and 3.0% (0-8.5%), respectively, which represented a reduction of 19.2% to 61.5% compared to not using prophylactic medication (Figure 28C, Table 29. The P Inf when all the travelling dogs used prophylactic medication (i.e. 100%) was 1.9% (0 to 6.8%). The effect of insecticide applied on endemic areas showed to be less effective in reducing P Inf, as even when 100% of endemic areas were treated with insecticide, P Inf was reduced to 3.5% ( %), which represented 55.1% relative to not using insecticide (Figure 28D, Table 29). All combinations of mitigation measures that included a high usage (i.e. 80%) reduced P Inf to less than 2.5%. The most effective was the simultaneous use of repellent with vaccine, which reduced the P Inf to 0.4% (0-1.2%), representing a reduction of 94.9% compared to not using any mitigation measure and a reduction of 80.7% and 78.2% compared to using repellent or vaccine separately at 80%, respectively (Table 29). This suggests the joint effect of the two combined mitigations measures is higher than the effect of the measures applied individually in the population of dogs. At medium-low level of usage (i.e. 40%) the combinations of repellent and vaccination, and repellent and Prophylactic medication were the most effective combination of the evaluated mitigation measures, reducing P Inf to 2.5% ( %), and 2.6% (0.7-5,4%), which represented a reduction of 45.6% and 47.7% when compared to 40% of vaccine and prophylactic medication use separately, respectively (Table 29). When vaccination, repellent and insecticide were combined and implemented at 100%, the P Inf was 0.02% (0-0.1%), which represented a reduction of 99.7% compared to not using any mitigation measure. In contrast, the P Inf from the realistic scenario (use of vaccine at 10%, repellent at 50% and prophylactic medication at 5%) was 1.6% (0-5.2%) which was a reduction of 79.5% with respect to not using any mitigation measure when one dog was travelling with the household. When the network size increase to 200 dogs instead of 50, the P Inf was 6.5% ( %), which represented a reduction of 16.7% compared to the P Inf in a network size of 50 dogs (Table 29). Supporting publications 2013:EN

94 Figure 28: Probability of infection (mean and 95%PI) of a dog after a trip to a CanL endemic area (P inf ), by proportions of use of mitigation measures (blue lines) compared to no use of mitigation measures (red lines): A) Vaccine use, B) Repellent use, C) Prophylactic medication and D) Insecticide use. Supporting publications 2013:EN

95 Table 29: Probability of infection (P Inf ) of CanL (ranked from low to high mean P Inf ) for different scenarios of mitigation measures when used individually or in combination. Level of use of mitigation measures P Inf Vaccination Prophylactic Repellent Insecticide Mean (95% Predictive medication interval) % (0-0.1) % (0-0.3) % (0-1.2) % (0-1.4) % (0-1.9) % (0-5.1) % ( ) % (0-4.3) % (0-5.2) % (0-6.5) % (0-6.8) % (0-7.1) % ( ) % ( ) % ( ) % (0-8.5) % ( ) % ( ) % ( ) % ( ) % ( ) % ( ) % ( ) % ( ) % ( % ( ) % ( ) % ( ) % ( ) % ( ) % ( ) % ( ) (a) % ( ) % ( ) % ( %) (b) (a) Baseline scenario with a network size of 200 dogs (b) 1 to 5 dogs travelling with the households Probability of endemicity within a contact network of dogs (P End ) The P End following the introduction of an infected dog into a non-endemic area with presence of competent vectors was 72.3% ( %). Similarly, when 1 to 5 infected dogs per household were introduced into non-endemic areas the P End was 72.6% ( %). These findings suggest that leishmaniosis is very likely to become endemic within a few years following the introduction of an Supporting publications 2013:EN

96 Infectious dog. Nonetheless, endemicity was calculated as the probability that the basic reproductive ratio is greater than 1 (e.g. at least one new case results from the introduction of an Infectious dog into a fully Susceptible population), so no attempts were made to establish the long term prevalence in a newly endemic area. When 100% of repellent was used on dogs, the P End was 10.2% ( %, Figure 29), which represented a reduction of 85.9% compared to not using any mitigation measure. When vaccination, prophylactic medication or insecticide were used separately on 100% of dogs within the contact network, the P End were 28.5% ( %), 56.1% ( %) and 72.3% ( %), respectively (Figure 29, Table 30). These results showed that although repellent was the most effective mitigation measure, vaccination and prophylactic medication also reduced the probability of endemicity. Insecticide had no effect on the P End when used alone. When mitigation measures were used in combination, the best results decreasing the P End were observed when high levels of use were implemented. When repellent was implemented separately in 80% of the contact network, the P End decreased to 66.9% ( %), compared to not using any mitigation measure; however when repellent use was combined with vaccination, prophylactic medication or insecticide, P End decreased to 28.6% ( %), 38.0% ( %) and 63.9% ( %), respectively (Table 30). The combination of insecticide use with vaccination or prophylactic medication did not decrease the P End with respect to when vaccination or prophylactic medication was used separately. At a medium-low level of usage (e.g. 40%) the effect of the combination of mitigation measures on the P End was minimal reducing only 2 to 3 percentage units the P End compared to using each mitigation measure separately at the same usage level (Table 30). The lowest P End was observed when vaccination repellent and insecticide were implemented at 100% of usage the P End was 0.2% (0-1.5%). In contrast, the P End from the realistic scenario (use of vaccine at 10%, repellent at 50% and prophylactic medication at 5%) was 71.4% ( %) which was a reduction of 1.2% with respect to not using any mitigation measure (Table 30). Supporting publications 2013:EN

97 Probability of endemicity (P End ) Probability of endemicity (P End ) Probability of endemicity (P End ) Probability of endemicity (P End ) Impact, modelling and control of canine leishmaniosis in the EU 1 A 1 B Vaccine use Repellent use 1 C 1 D Prophylactic medication use Insecticide Use Figure 29: Changes in the probability of endemicity (P End ) following the introduction of a CanL infected dog into a non-endemic area with competent vector, by proportions of use of mitigation measures (blue lines) compared to no use of mitigation measures (red lines) A) vaccine, B) repellent, C) Prophylactic medication and D) Insecticide Supporting publications 2013:EN

98 Table 30: Probability of infection (P End ) with CanL (ranked from low to high mean P IEnf ) for different scenarios of mitigation measures when used individually or in combination. Level of use of mitigation measures P End Vaccination Prophylactic Repellent Insecticide Mean (95% Predictive medication interval) % (0-1.5) % ( ) % ( ) % ( ) % ( ) % ( ) % ( ) % ( ) % ( ) % ( ) % ( ) % ( ) % ( ) % ( ) % ( ) % ( ) % ( ) % ( ) % ( ) % ( ) % ( ) % ( ) % ( ) % ( ) % ( ) % ( ) % ( ) % ( ) % ( ) % ( ) % ( ) % ( ) (a) % ( ) % ( ) % ( ) (b) (a)with 1 to 5 index cases introduced to the contact network in the no-endemic area, (b)with a network size of 200 dogs The mean prevalence of infected dogs at the end of the 3 years of simulation was 27% (0-66.0%). The number of secondary cases at the end of the first week after the index case became infectious was 0.32 (0-3) with seasonality and 1.32 (0-5) without seasonality (i.e. sand flies are always active, Figure 30). This number of secondary cases after the first week of transmission in a non-endemic area is a similar concept to R0, but R0 refers to the number of infections during an infection generation while the number of secondary cases was calculated only for the first week of CanL transmission. Given the Supporting publications 2013:EN

99 Frequency Frequency Impact, modelling and control of canine leishmaniosis in the EU very long infectious period in dogs, up to several years, the practical interpretation of R0 is hindered as it would essentially cover the entire simulation period and result in a large number, while the number of secondary cases after the first week of CanL transmission is more meaningful. A Number of secondary cases Figure 30: Distribution of the number of secondary of canine leishmaniosis cases at the end of the first week after the introduction of an infectious dog in a non-endemic area with competent vectors when A) seasonality was implemented, and B) no seasonality was included. B Number of secondary cases Probability of endemicity in a region (P EndRegion ) For brevity, only the two most effective mitigation measures (repellent and vaccination use) were included in the scenario analysis of P EndRegion. Dog returning from an endemic area (travelling pathway) When no mitigation measure was implemented, both P Inf and P End were very high (means 7.8% and 72.3% respectively), which resulted in a high P EndRegion. Under this scenario, a P EndRegion of 100% was obtained for 170, 240 and 350 dogs returning from travel from an endemic area when 0%, 50% and 100% of the dogs were tested and excluded if positive, respectively. The effect of test and exclusion on reducing P EndRegion was important when a very small numbers of dogs (e.g. 10) were travelling to endemic areas, but quickly decreased and was almost null for large numbers of travelling dogs (e.g. 5,000). Repellent and vaccination in 80% of the dogs travelling decreased P EndRegion when used individually or in combination with a higher joint effect (Table 31). Supporting publications 2013:EN

100 Table 31. Effect of different levels of testing and exclusion and mitigation measures on the probability of CanL endemicity in at least one network of a non-endemic region, for different numbers of dogs travelling to endemic areas. Mitigation measure Test and exclusion level (%) None 0 None 50 None 100 Vaccination at 80% 0 Vaccination at 80% 50 Vaccination at 80% 100 Repellent at 80% 0 Repellent at 80% 50 Repellent at 80% 100 Repellent and Vaccination at 80% Repellent and Vaccination at 80% Repellent and Vaccination at 80% Number of dog travelling to endemic areas % 97.5% 100% 100% 100% ( ) ( ) ( ) ( ) ( ) 32.9% 93.8% 100% 100% 100% (0-95.2) (0-100) ( ) ( ) ( ) 20.1% 86.8% 100% 100% 100% (0-92.9) (0-100) ( ) ( ) ( ) 10.7% 56.6% 87.8% 92.8% 95.3% (0-71.8) (0-100) (0-100) (0-100) (0-100) 7.1% 47.7% 83.7% 91.0% 95.3% (0-66.0) (0-99.5) (0-100) (0-100) (0-100) 4.6% 35.4% 75.7% 86.5% 95.1% (0-58.3) (0-99.8) (0-100) (0-100) (0-100) 9.0% 53.7% 89.1% 94.9% 97.0% (0-74.3) (0-100) (0-100) (0-100) (0-100) 6.1% 45.1% 84.5% 92.5% 97.0% (0-72.4) (0-99.4) (0-100) (0-100) (0-100) 3.8% 32.3% 75.4% 87.6% 96.8% (0-68.6) (0-96.9) (0-100) (0-100) (0-100) 1.1% 9.4% 33.9% 49.5% 77.5% (0-21) (0-59.8) (0-94.2) (0-99.4) (0-100) 0.6% 7.0% 26.9% 42.7% 74.6% (0-16.4) (0-50.4) (0-88.4) (0-98.3) (0-100) 0.4% 4.4% (0-19.1% (0-31.7% (0-68.0% (0-0) 40.7) 76.0) 93.5) (0-100) Dog imported from endemic areas (import pathway) When no mitigation measure was implemented, the probability of CanL infection in imported dogs was given by the prevalence in the endemic areas. Under this scenario, a P EndRegion of 100% was obtained for 20, 30 and 80 dogs imported from endemic areas when 0%, 50% and 100% of the dogs were tested and excluded if positive, respectively. When repellent was used in 80% of the contact network in the non-endemic area, a P EndRegion of 100% was obtained when 270, 350 and 600 dogs imported from endemic areas when 0%, 50% and 100% of the imported dogs were tested and excluded if positive, respectively. Similarly, when 80% of the dogs in the contact network in the non-endemic area were vaccinated, a P EndRegion of 100% was obtained when 320, 450 and 710 dogs imported from endemic areas when 0%, 50% and 100% of the imported dogs were tested and excluded if positive, respectively (Table 32). When vaccination and repellent use was combined in the population of the non-endemic area, a P EndRegion of 100% was obtained when 770, 1,000 and 1,700 dogs imported from endemic areas when 0%, 50% and 100% of the imported dogs were tested and excluded if positive, respectively (Table 32). Supporting publications 2013:EN

101 Table 32. Effect of different levels of testing and exclusion and mitigation measures on the probability of CanL endemicity in at least one network of a non-endemic region, for different numbers of dogs imported from endemic areas (commercial imports, adoptions, individual purchases). Mitigation measure Test and exclusion level (%) None 0 None 50 None 100 Vaccination at 80% 0 Vaccination at 80% 50 Vaccination at 80% 100 Repellent at 80% 0 Repellent at 80% 50 Repellent at 80% 100 Repellent and Vaccination at 80% Repellent and Vaccination at 80% Repellent and Vaccination at 80% Number of dog imported or adopted from endemic areas % 100% 100% 100% 100% (0-100) ( ) ( ) ( ) ( ) 86.5% 100% 100% 100% 100% (0-100) ( ) ( ) ( ) ( ) 28.8% 89.0% 100% 100% 100% (0-97.5) (0-100) ( ) ( ) ( ) 45.3% 94.6% 99.7% 99.8% 100% (0-96.4) ( ) ( ) ( ) ( ) 36.2% 91.8% 99.6% 99.8% 100% (0-94.6) (0-100) ( ) ( ) ( ) 23.8% 84.5% 98.9% 99.8% 100% (0-89.7) (0-100) ( ) ( ) ( ) 51.2% 96.1% 99.8% 99.9% 99.9% (0-99.0) ( ) ( ) ( ) ( ) 41.7% 94.1% 99.7% 99.9% 100% (0-98.0) (0-100) ( ) ( ) ( ) 27.7% 88.2% 99.3% 99.9% 100% (0-95.1) (0-100) ( ) ( ) ( ) 26.7% 84.0% 98.8% 99.6% 100% (0-83.2) ( ) ( ) ( ) ( ) 20.7% 77.8% 97.9% 99.5% 100% (0-76.2) (0-100) ( ) ( ) ( ) 13.3% 65.5% 94.8% 98.5% 100% (0-64.2) (0-100) ( ) ( ) ( ) DISCUSSION The most influential parameters for P Inf were the day of year when the trip to an endemic area was performed, followed by the travelling length. These results are intuitive, as the first parameter is a proxy for the season when the trip was done, and the second has direct relationship with the risk of disease exposure to the sand fly during the trip. For this parameter, the travelling day was assumed to randomly occur at any time of the year, although travelling to endemic areas may be concentrated during certain periods of the year (winter or during summer school vacations). Evidently, if most of the trips occurred during winter when vectors were absent, P Inf would be close to 0. However if most of the trips occurred during the summer when the vector was present, P Inf is expected to be much larger. Similarly, longer trips linearly increase the P Inf because dogs were exposed to Leishmaniacarrying vectors for longer periods (when travels occurred during the vector breeding season). The distribution of the length of the trip was estimated from data obtained from European statistics (Eurostat, 2013f), and assumed that the trips accompanied by dogs followed the same distribution of duration as the trips reported. The most influential parameter for P End was the time of the year when the index case dog became Infectious: season was the key driver of the probability of endemicity in previously CanL-free areas. This parameter influences P End because transmission from the index case to dogs in the contact network was modelled only when the vector was present. If no vectors were present when the index case became Infectious, transmission was delayed until the next vector breeding season. Therefore, if Supporting publications 2013:EN

102 the index case showed clinical signs of leishmaniosis during winter and was successfully treated, or died, no transmission to the network was observed. It is also possible that this parameter was associated with the time of the year when the dog became infected in the endemic area and therefore influenced by the time of the year when travelling to an endemic area. The effect of the season is also evident in the increase in the number of secondary cases at the end of the first week after the introduction of an infectious dog. This value increased 4.2 times when compared the results with seasonality and without. Several parameters used to estimate P Inf, P End and P EndRegion were based on studies with different degrees of bias (See systematic review section). Parameters were estimated from the available information in these studies and therefore, assumptions were made to derive them. For example, the prevalence of CanL infectious dogs in endemic areas influenced P Inf and P EndRegion. This parameter was estimated using the prevalence of clinical cases in 2 regions in Spain (Gálvez et al., 2010, Miró et al., 2012) and in absence of data on prevalence of infectious animals, it was assumed that the prevalence of clinical cases was an approximation to the prevalence of Infectious dogs in endemic areas. Therefore it is likely that the prevalence of Infectious dogs was underestimated. Also, the uncertainties in the number of female sand fly bites per day, the daily rate of transition from latent to infectious sand flies were estimated using opinion from a limited number of experts. The influence of the transition parameters on P Inf and P End may have been underestimated in the sensitivity analysis, as only the effect of the parameter in individual dog transitions was measured, as opposed to the combined effect of all the simultaneous transitions in the model. Several key assumptions were made for this modelling work, one of the most important being that dogs became resistant after effective treatment. Although it is possible that some clinical dogs return to the clinical stage 6 months to 2 years after treatment (Solano-Gallego et al., 2009), (Oliva et al., 2010), it was assumed here that dogs with history of clinical infection would be monitored on a regular basis for seroconversion and clinical signs of leishmaniosis (Solano-Gallego et al., 2009), and that they would therefore be treated as needed within a short period of time. P End may have been underestimated if Resistant dogs can transition back to Infectious after treated or after losing vaccine immunity. The P End increased only slightly when 1 or 1 to 5 infected dogs entered the network (72.3% versus 78.7%), which may indicate that the number of infected dogs entering a non-endemic area has a small effect on P End. It was also assumed that dogs moved between networks within endemic and nonendemic areas independently (i.e. any contact network has the same chance of receiving an infected imported dog), which means that P EndRegion could have been overestimated if this assumption is unwarranted. Vectorial capacity represents the potential of sand flies to transmit leishmaniosis between dogs and it is independent of the prevalence of L. infantum infection in sand flies (Massad and Coutinho, 2012). As suggested by Dye (1996), vectorial capacity was used to elude modelling the transmission of L. infantum between sand flies, which occurs on a faster time scale than in dogs, and because few sand flies live long enough to acquire infection. Moreover, the transmission parameter of L. infantum in the sand fly population may be difficult to estimate. It was also assumed that the vectorial capacity is constant and independent of prevalence in the sand fly population. This may have overestimated the initial size of the outbreak because the vectorial capacity may take few weeks to reach stability at the start of the short breeding season in non-endemic areas. In contrast, by assuming a single introduction of an infected dog in a network instead of a continuous inflow of dogs, P End may have been underestimated. As P End is measured as the probability Supporting publications 2013:EN

103 of at least one new infected dog at the end of the simulated three year period, it is unlikely that the above assumption affected this parameter, nor the relative efficiency of the different mitigation measures. Based on a baseline scenario with no mitigations measures, both P inf and P end were very high (means 7.8% and 72.3% respectively), resulting in a probability of endemicity in at least one network in the region that becomes 100% only after a few infected dogs are imported (that is, in a Leishmania-free region with competent vectors). This number may seem high at first inspection but this is possible as dogs exhibit a prolonged infectious period which is often sub-clinical, allowing them to spread the disease for a very long time period. The baseline P Inf result agrees with field findings where susceptible animals were introduced in dog contact networks in endemic areas (Dye et al., 1993, Oliva et al., 2006). In these studies, a cohort of CanL free dogs was introduced into an endemic area in southern Italy (Oliva et al., 2006) and southern France (Dye et al., 1993), and dogs were physically examined and sampled every 1 to 3 months to evaluate the clinical signs of leishmaniosis and positivity to tests (PCR and IFAT). While in these studies dogs were exposed to sand flies during the whole five months of the breeding season, in our modelling study dogs travelling to endemic areas were exposed to sand flies only for the relatively short duration of the travel (when the travel overlapped with the breeding season). Nonetheless, once the simulated exposure time matched those of the cited studies, the simulation results were comparable. For example, Oliva et al. (2006) found an incidence of.036 cases/dog/week, which is comparable to the incidence when running a baseline scenario for 2.5 months with a full breeding season (Incidence rate ratio (IRR) 0.72, 95%CrI: 0.38,1.29). Likewise, Dye et al. (1993) found an incidence of cases/dog/week which is similar to the found when comparable conditions were simulated (IRR:1.1, 95%CrI:0.68, 1.79). In addition, the low sensitivity (52.6%, 95% CrI: %), of the diagnostic tests used in these studies (Mettler et al., 2005a) underestimated the true incidence of infected dogs. In the present study, we modelled the true infection state of the dogs. Unsurprisingly, the highest and lowest P Inf were observed in the scenarios when none and all the mitigation measures were implemented, respectively. Although, it is likely that none of these scenarios were realistic, they provide the range of the predicted P Inf for different usage levels and combinations of mitigation measures. Results show that P inf and P end can be greatly reduced by increasing the use of mitigation measures separately or in combination. The most effective mitigation measure was repellent (P Inf : 0.03% and P End : 10.2% when repellent is used in 100% of the dogs), followed in order of effectiveness by vaccine, prophylactic medication and insecticide. The main effect of repellent is to avoid the bites of sand flies which reduces the VC several folds. Moreover, repellents have a secondary insecticide effect, killing a proportion of the sand flies that were able to bite repellent users. The insecticidal effect of repellents was not modelled but it is likely to increase repellents efficacy, further reducing the P Inf and P End. This model did not include scenarios where sand flies would divert bites to other dogs without repellent, or feed on other species. Such diversion of bites may increase the P Inf in dogs that do not use repellent or increase the P End, if an alternative host is infected. Vaccination and prophylactic medication had similar effect in reducing the P Inf. Since both mitigation measures created resistance to leishmaniosis, their slight different effect on P Inf could be attributed to their different efficacies, as the vaccine efficacy modelled was 86.2% ( %) whereas prophylactic medication efficacy was 73.3% ( %). Insecticide showed to be the least effective mitigation measure in reducing the P Inf and P End. In contrast, in an earlier CanL modelling work, Dye (1996) concluded that insecticide was the most effective mitigation measure to reduce the prevalence of CanL when compared to vaccination, therapeutic treatment and killing infected dogs. However, Dye (1996) modelled the disease in an Supporting publications 2013:EN

104 endemic area, whereas in the present study P End was estimated following the introduction of an infectious dog in a non-endemic area. Dye (1996) also modelled insecticide as a percentage of change in the mortality rate of sand flies, whereas in our study vector density was decreased proportionally to the insecticide efficacy as estimated from a field study (Coleman et al., 2011). Although insecticide use was incorporated as a potential mitigation measure, some authors considered that mass killing of vectors using insecticide may not be appropriate to control the population of sand flies, because they tend to inhabit focal areas (Gálvez et al., 2010) and their breeding sites remains unknown (Feliciangeli, 2004). The systematic review did not identify any study that evaluated combinations of mitigation measures. In the current study the joint reduction of the combination of repellent and vaccine or repellent and prophylactic medication on P Inf and P End was higher compared to using these mitigation measures individually. However, this joint effect was only observed when both mitigation measures were used on 80% of dogs. Overall, the results of the impact assessment agree with the modelling results, as most of the areas with presence of competent vectors also had reported cases of CanL. However, in a few European NUTS3 areas with presence of sand flies no cases of CanL were reported by the veterinarians responding to the survey in the impact assessment, whereas the modelling results suggest that the probability of disease endemicity following introduction of infected dogs into a non-endemic area with a competent vector is high. However, the survey responses may not have included enough veterinary practices to be representative of the practices in the area and thus, may not be directly comparable to the modelling results. Furthermore, even if the probability of at least one dog to be infected in the department is high, the prevalence of CanL infection and thus the number of clinical cases may be low enough to stay undetected by the responding veterinary practices. Moreover, if the infection was recently introduced into a department, it may also be possible that veterinary clinics had not yet diagnosed cases of CanL, given the relatively slow progression of the disease and the potential for under or misdiagnosis of a disease that is not commonly seen by local practitioners. Finally, VBORNET maps reported only presence or absence of competent vectors, not abundance so it may also be possible that competent vectors were present, but in a low enough density to prevent further CanL spread in dogs. The P EndRegion for dogs returning to non-endemic areas can be reduced by implementing mitigation measures before, during and after the trip (repellent, vaccine, prophylactic medication, insecticide and test and exclusion). In contrast, the mitigation measures to reduce P EndRegion in dogs that are imported from endemic areas are limited to mitigation measures that reduce the probability that imported animals are infected (test and exclusion), and those that reduce the transmission of infection to susceptible dogs in the non-endemic area after imported (repellent, vaccine, prophylactic medication, insecticide). The results of the Import module showed that test and exclusion of positive dogs moving into non-endemic areas was effective for low numbers of imports, but its benefits diminished as the number of imported dogs increased. When interpreting the results it is important to consider that P EndRegion reports the probability of endemicity at any level: from one dog to 100% of the dogs in the contact network, within one to all contact networks. That is, no attempt was made to quantify the actual disease prevalence in the region, since data on the number of contact networks and how they connect are not available. So although P EndRegion numbers may seem relatively high, the prevalence in a region may still range from extremely low to extremely high. The probability of endemicity in at least one dog population in a previously CanL free area was high even under scenarios where all imported animals are tested and 80% of the animals in the importing Supporting publications 2013:EN

105 (non-endemic) were vaccinated and/or used repellent. The main drivers of these results are the high P end (probability of endemicity in one dog contact network within the region), the relatively high prevalence of CanL in endemic areas (P InfCA ) and the low sensitivity of the diagnostic test used to model the test and exclusion policy. As discussed, P End doesn t quantify the actual prevalence within a dog contact network; only whether at least one other dog is infected three years after an infectious dog is introduced in the network. The seroprevalence in the country of origin were sourced from several studies (Gálvez et al., 2010, Baldelli et al., 2011, Leontides et al., 2002, Keck and Dereuer, 2003) and ranged from 7.2 to 17.8%. Using these estimates assumes that the dogs imported were a random sample from the dog population in the country of origin. Although this assumption is common in risk assessment, it is possible that CanL prevalence in commercial dogs is lower than that of the general population in a country since commercial breeding facilities may regularly use mitigation measures such as vaccination or repellent, and/or may also house dogs indoors, further reducing the exposure to sand flies. Under these circumstances, the results of the import pathway scenarios may be more comparable to those of the traveling scenarios, where the risk of infection in dogs was highly reduced when using mitigation measures. Finally, an immunochromatographic-dipstick rapid test using a rk39 antigen with a sensitivity of 52.6% (95%CrI:30.8%-74.0%) was used to model the test and exclusion strategy (Mettler et al., 2005a). This diagnostic test and its sensitivity may not be representative of what is observed in field conditions. Therefore, the sensitivity of the test used for the test and exclusion of positive dogs is a parameter that may require refinement in future modelling work. If a test with a higher sensitivity was to be used to identify and exclude CanL infected dogs, this mitigation measure would be more effective at reducing P EndRegion. CONCLUSIONS The first objective of the simulation modelling component of this project was to estimate the probability of CanL endemicity following the introduction of infected dogs in previously disease-free areas. The stochastic model developed for the project estimated a high probability of endemicity for disease-free EU regions with competent sand flies where infected dogs are introduced (92.2% to 100% when 10 to 100 dogs were introduced). The second objective of the simulation modelling was to evaluate potential mitigation measures for CanL. The impact of the following mitigation measures was assessed individually and in combination: vaccination, prophylactic medication, repellent, insecticide, and diagnostic test and exclusion. When applied individually, mitigation measures reduced the risk of endemicity only when used in a high proportion of the population(s) ( 80%). Testing and exclusion of positive dogs had the highest impact on the overall probability of endemicity when relatively small numbers of dogs were imported from, or returned from trips to endemic areas. However, the benefits of this mitigation measure diminished as the numbers of imported dogs increased, due to the relatively low sensitivity of the diagnostic test used 52.6% (95% CrI:30.8%-74.0%). The most effective mitigation measure for P end and P inf was the use of repellent, followed in decreasing order of effectiveness by: vaccination, prophylactic medication and use of insecticide Some combinations of mitigation measures showed a joint effect higher than the effect of the same measures applied individually, for example repellent and vaccination or prophylactic medication, even when applied to a lower proportion of dogs. Supporting publications 2013:EN

106 The following main recommendations were formulated based on the results of the simulation modelling: - Mitigation measures on dogs in CanL free areas with the competent vectors were only effective when used in a large proportion of the animals within the contact network and therefore, would require a high level of compliance and may be costly to implement. Thus, efforts may be better directed at either dogs travelling to or from endemic areas, or commercial imports of dogs. - Dog owners should be advised to use control measures on their dogs during travels to CanL endemic areas to prevent L. infantum infection. Ideally, vaccination or prophylactic medication, slightly less effective but of lower cost should be complemented by the use of repellents during the full duration of the travel. Although not explicitly evaluated via modelling, the findings of this report suggest that dog owners from CanL endemic areas travelling with their dogs to non-endemic areas with the competent vector should also be advised to use repellent to reduce the chances that their dogs may spread the disease to the local sand fly population. Ideally, dogs known to be infected should not visit non-endemic areas, but this might be an unrealistic requirement given the open nature of travels within the European Union. - A test and exclusion policy could be a feasible alternative for commercial importations of dogs from endemic areas. Dogs should be tested with the best available diagnostic test before the date of importation, and positive animals should not be allowed to enter disease-free countries. The effectiveness of this strategy is tied to the sensitivity of the diagnostic test used. Therefore, when large numbers of dogs are imported and a test with a relatively low sensitivity is used, the benefits of this strategy are diminished. Because of this, additional mitigation measures aimed at reducing the risk of infection in dogs from endemic areas (repellent, vaccination, prophylactic medication) may also be requested for dogs to be imported into non-endemic areas. The test and exclusion policy may be difficult to implement and enforce for non-commercial movements of dogs across the EU (e.g. for travel between endemic and disease-free areas within the same countries or between countries), hence the suggestion to restrict this strategy to commercial movements of dogs. The conclusions and recommendations above are derived from a modelling abstraction of a very complex disease process. This abstraction required a series of simplified assumptions and the quantification of many important parameters, often from disparate sources of information. Therefore, the conclusions and recommendations should be considered in the context of the modelling work described here, and notably the limitations and data gaps described in the report. Supporting publications 2013:EN

107 3. Impact assessment INTRODUCTION AND OBJECTIVES The objective of the impact assessment was to evaluate the impact of canine leishmaniosis (CanL) in those areas where the disease is endemic, within the European Union (EU). The necessary data have been collected to characterise the impact of the disease, following the guidance of the manuals created by Phylum for OIE and the EC (Phylum, 2010) and Anses (Anses, 2012). The framework for the impact assessment is summarised in Figure 31. The economic, human health, societal and environmental impact of the disease and its control measures were explored. Epidemiological profile of canine leishmaniosis Characteristics of the studied countries Local epidemiology Figure 31: Schematic sequence of the impact assessment of canine leishmaniosis in the EU 3.1. Characteristics of the Studied Countries MATERIALS AND METHODS Data collection Published information and figures were sourced from databases and reports from national and international organisations to describe the key characteristics of the studied countries: France, Spain, Portugal, Italy and Greece. Where data were not readily available, these were sought through personal communications. Data on the distribution of phlebotomine sand fly vector species was provided by VBORNET (VBORNET, 2012). The French territories outside Metropolitan France (mainland and Corsica) were not considered Data analysis Impact assessment Disease Economic Human health Societal Environmental Control measures Economic Societal Environmental The economic status of the studied countries and the role and transportation of dogs within the EU were described in order to assess the impact of leishmaniosis in the context of the local socioeconomic environment. Human and canine population estimates and data relating to the economy, human health and veterinary sectors in each studied country were compiled into a table, so that comparisons could be made across the studied countries. Data on the distribution of phlebotomine sand fly species found in the studied countries were joined to shape-files downloaded from a database of global administrative areas (GADM), and presented as choropleth maps using ArcGIS 9 (ESRI, 2013). Supporting publications 2013:EN

108 RESULTS All of the studied countries: France, Spain, Portugal, Italy and Greece, are developed countries, situated within Western Europe. They are part of the EU, a political and economic union consisting of 27 member states. Although the studied countries have high OECD income levels, economic growth within the EU suffered as a result of the global financial crisis, resulting in a severe recession in most countries (The World Bank, 2013a). Domestic dogs are very popular in the EU, with 27% of households owning at least one dog (FEDIAF, 2010). A questionnaire reported that the most common reasons for dog acquisition in Europe were for companionship, guarding, hunting and working (assistance / herd dogs) (Batson, 2008). Furthermore, European legislation acknowledges the importance of companion animals to society and their contribution to quality of life. The European Convention for the Protection of Pets (ETS No. 125), an international treaty, lays down principles for responsible pet ownership and the welfare considerations relating to the control of stray populations (Council of Europe, 1987). In addition, all of the studied countries have national or regional laws relating to canine welfare and registration (Carodog, 2013). Further to the benefits of pet ownership to the individual, the industries and services associated with pet ownership contribute to the European economy. In 2010, the combined annual turnover from the pet food industry and pet related products and services (including veterinary care and insurance) was estimated to be 24 billion (FEDIAF, 2010). Companion animal ownership generates an estimated 550,000 jobs throughout Europe, including approximately 200,000 veterinary surgeons (FEDIAF, 2010). In 2009, the annual turnover of the EU veterinary service sector was estimated to be 13,000 million and 6,500 million of value added was generated (Eurostat, 2012). Dogs travel between EU member states for both non-commercial and commercial purposes (Europa, 2013f). To enable movement of up to 5 dogs for non-commercial reasons, each animal must be accompanied by a passport, which documents a valid rabies vaccination (Council of Europe, 2003). Individuals are identified by an electronic microchip or clearly readable tattoo (if applied before 3rd July 2011), which should correspond to details recorded in the passport (Europa, 2013f). When dogs are moved for commercial reasons, additional regulations as laid down in Council Directive 92/65/EEC, apply (Council of Europe, 1992). An authorised veterinary surgeon must conduct a clinical examination within 24 hours prior to travel, in order to verify that the animal is in good health and fit to travel. An Official Veterinarian of the member state of dispatch is also required to authorise a health certificate and to notify the movement through the Community Trade Control and Expert System (Europa, 2012). To prevent commercial movements of dogs being fraudulently claimed as non-commercial movements, the latter rules (Directive 92/65/EEC) are applicable when more than 5 dogs are moved at one time for non-commercial movements (Council of Europe, 2010). Although dogs can make an important socio-economic contribution, both owned and stray dogs can also pose a public health risk, including the transmission of zoonotic diseases. In addition, stray dog populations pose animal health and welfare issues (Tasker, 2007). Approaches to stray animal control vary in different countries, as there is no common European Community legislation detailing which specific control measures should be adopted (Voslarova and Passantino, 2012, Tasker, 2007). At least ten species of phlebotomine sand flies are found in at least one of the studied countries (VBORNET, 2012). The most widespread vector of L. infantum in the western Mediterranean basin is reported to be Phlebotomus perniciosus (P. perniciosus), although this species is absent in Greece (Figure 32a). P. ariasi, P. neglectus, and P. perfiliewi, are also reported to be proven competent vectors of L. infantum (Killick-Kendrick, 1999, Ready, 2010). P. tobbi (Rassi et al., 2012) and P. alexandri (Colacicco-Mayhugh et al., 2010, Azizi et al., 2006) are thought to be probable vectors of L. infantum and although the role of P. mascitti is uncertain, this species is frequently found in L. Supporting publications 2013:EN

109 infantum-endemic regions (Naucke et al., 2011). There is evidence that P. papatasi and P. sergenti are specific vectors for L. major and L. tropica respectively (Dostalova and Volf, 2012) and not involved in the transmission of L. infantum. P. similis is also a suspected vector of L. tropica, although the role of this species has yet to be proven (Christodoulou et al., 2012, Ozbel et al., 2011). Figure 32a. to Figure 32g. show the geographical distributions of the phlebotomine sand fly species in the studied countries (VBORNET, 2012). Maps of P. papatasi, P. sergenti and P. similis were excluded as they are not thought to be competent vectors of L. infantum. Figure 32a. P. perniciosus Figure 32b. P. ariasi Figure 32c. P. neglectus Figure 32: Geographical distribution of possible or proven sand fly vectors of L. infantum in France, Spain, Portugal, Italy and Greece (VBORNET, 2012) Continued on page 110 Supporting publications 2013:EN

110 Figure 32d. P. perfiliewi Figure 32e. P. tobbi Figure 32f. P. alexandri Figure 32g. P. mascitii Figure 32: Geographical distribution of possible or proven sand fly vectors of L. infantum in France, Spain, Portugal, Italy and Greece (VBORNET, 2012) Supporting publications 2013:EN

111 Table 33: Characteristics of studied countries: Geography, economy, human population & health Studied countries France Spain Portugal Italy Greece References Geography and human population Land area (km 2 ) 550, ,782 92, , ,957 (Europa, 2013e, Europa, 2013b, Europa, 2013d, Europa, 2013a, Europa, 2013c) Human population (2012) 63,409,191 46,196,276 10,541,840 60,820,696 11,290,067 (CIA, 2013) Human population density, (Eurostat, 2013c) (Inhabitants per km 2 ) Economy GDP per capita (current US$) 42, , , , ,629.8 (The World Bank, 2013b) GDP growth (Annual %) (The World Bank, 2013b) GDP growth (per qtr. %) 2012Q (Eurostat, 2013a) 2012Q Q Q Direct contribution of travel and tourism to GDP( billion)(%gdp) 77.6 (3.8%) 57.1 (5.4%) 9.4 (5.7%) 63.8 (4.1%) 12.0 (6.5%) (WTTC, 2013d, WTTC, 2013e, WTTC, 2013c, WTTC, 2013a, WTTC, 2013b) (9.7%) 160 (15.2%) (10.3%) 30.3 (16.4%) (15.9%) a (Eurostat, 2013a) a January 2013 Total contribution of travel and tourism to GDP( billion)(%gdp) Unemployment rates (%, seasonally adjusted, 03/2013) Human health Life expectancy from birth (years)(2012) (CIA, 2013) Total expenditure on health(%gdp)(2011) (World Health Organization, General government expenditure on health (% total expenditure on health) (2011) b) Supporting publications 2013:EN by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender considered as an output adopted by the Authority. The European food Safety Authority reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

112 Out-of-pocket expenditure as a % of private expenditure on health (2011) HIV prevalence, years (%)(2011) (World Health Organization, 2013a) Table 34: Characteristics of studied countries: canine population and veterinary sector statistics Canine population Canine population Studied countries France Spain Portugal Italy Greece 7,590,000 a (2010) 5,498,500 b (2008) References 1,500,000 (2011) a 6,355,309 (2013) c 196,081 (2011) a a(wahid, 2013) b(batson, 2008) c(ministero della Salute, 2013a) (WAHID, 2013) a estimated (2011) (2012) (2011) (canine population / land area) Canine population density (dogs per a km 2 ) (2010) Households owning 1 dog (%) (FEDIAF, 2010) Registration legislation Mandatory Registration & identification a Mandatory Registration & identification b Mandatory Registration & identification b Mandatory Registration & identification b Mandatory Registration & identification b a(ministère de l'agriculture de l'agroalimentaire et de la Forêt, 2013)b(Tasker, 2007) Veterinary sector Number of veterinarians 13,652 (2013) a 29,060 b 4,640 c 30,162 d - a(petit, 2013), b(instituto Nacional de Estadística, 2012), c(simões A., 2013), d(benini, Number of companion animal veterinarians Annual turn-over of the veterinary services sector ( million), 2009 Value added ( million) veterinary services sector, ,787 a (2013) 2013) 10,000 b 1,718 (2008) c 22,609 d 1,000 b a(petit, 2013),b(Batson, 2008), c(resende, 2008), d(benini, 2013) 2, (Eurostat, 2012) 1, Supporting publications 2013:EN by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender considered as an output adopted by the Authority. The European food Safety Authority reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

113 3.2. Epidemiological profile of Canine Leishmaniosis MATERIALS AND METHODS Data collection An initial literature search identified potential transmission routes of L. infantum and mammalian species which may be implicated in the spread of the parasite. The species reported to be affected by clinical disease were also identified. A questionnaire was sent to nine members of the LeishVet group to gain expert opinion on the epidemiology of L. infantum, to complement the information derived from published articles. The LeishVet group is a registered association, which consists of veterinary scientists from academic institutes with a specific interest in canine leishmaniosis (Solano-Gallego et al., 2009). Questions related to the type of host that different species are likely to be (Table 35), and the likely importance of transmission routes in the overall spread of leishmaniosis in endemic areas within the EU. Respondents were asked to indicate the certainty of their answers, based on available data (Table 36). Table 35: Definitions of animal hosts relating to the transmission of pathogens Host Definition(a) Accidental / incidental An infected host, that plays no role in the maintenance of the parasite Secondary reservoir A host that can transmit infection but cannot maintain parasite transmission in the absence of the primary host(s) Primary reservoir A host responsible for maintaining the parasite indefinitely in nature (a): Source: (Quinnell and Courtenay, 2009) Table 36: Qualitative categories for expressing certainty in relation to qualitative estimates Certainty category Interpretation(a) There are scarce or no data available; evidence is not provided in references but rather Low in unpublished reports or based on observations, or personal communication; authors report conclusions that vary considerably between them. Medium There are some but no complete data available; evidence is provided in small number of references; authors report conclusions that vary from one another. There are solid and complete data available; strong evidence is provided in multiple High references; authors report similar conclusions. (a): Adapted from: (EFSA, 2006) Data analysis The major epidemiological features of L. infantum were described with standard methods. The results of the literature search and questionnaire were used to create stacked bar charts and a radar chart, based on the criteria detailed in the Phylum manual (Phylum, 2010). The epidemiological characteristics of L. infantum were assessed based on the species affected, the persistence and transmission of the pathogen and the zoonotic potential. The result for each branch of the radar chart was the sum of the corresponding criteria divided by the number of criteria (ranging from 0 to 1). For details of the Phylum criteria, please consult Appendix R. Supporting publications 2013:EN

114 Percent of responses Impact, modelling and control of canine leishmaniosis in the EU RESULTS Animal species affected by clinical disease Infection of susceptible dogs with L. infantum can result in a wide spectrum of manifestations, which can potentially involve any organ, tissue or body fluid (Solano-Gallego et al., 2011). However, in endemic areas, there is a higher prevalence of subclinical infection than clinical disease in dogs (Moreno and Alvar, 2002, Miro et al., 2008, Baneth et al., 2008, Solano-Gallego et al., 2009). The severity of clinical disease in dogs can range from mild and self-limiting to very severe and potentially fatal (Solano-Gallego et al., 2009). See section for the common clinical signs and how the disease impacts on the welfare of dogs. Domestic cats can manifest both visceral and cutaneous forms of leishmaniosis (Poli et al., 2002, Marcos et al., 2009, Navarro et al., 2010, Ozon et al., 1998, Hervas et al., 1999), whereas horses have only been reported to have self-limiting cutaneous leishmaniosis (Solano-Gallego et al., 2003, Rolao et al., 2005, Koehler et al., 2002). Infection and clinical disease have also been reported in non-domestic canines (Fallah and Khanmohammadi, 2011, Tenorio Mda et al., 2011, Beck et al., 2008, Luppi et al., 2008) and lions (Dahroug et al., 2011, Libert et al., 2012) Animal and human hosts Infected dogs are often infectious and it is widely accepted that domestic dogs are the main reservoir hosts of L. infantum (Baneth et al., 2008, Quinnell and Courtenay, 2009, Alvar et al., 2004). Certain wildlife species are also potential reservoir hosts of L. infantum. For example, investigations in Fuenlabrada, Madrid, concluded that hares (Lepus granatensis) were a likely reservoir host of L. infantum and that infection in this species contributed to the outbreak of human cases (Molina et al., 2012, Aguado et al., 2013). LeishVet expert opinions on the likelihood of different species as the defined host types are presented in Figure % 90% 80% 70% 60% 50% 40% 30% 20% 10% 0% Dog Wild canines cats rabbits / hares humans rodent horses Don't know Accidental / incidental Secondary reservoir Primary reservoir Figure 33: Expert opinion on the role of different mammal species / orders as hosts of Leishmania infantum (responses from six experts) Transmission The most common route of transmission of the parasite among dogs is via the bite of the female phlebotomine sand fly (Killick-Kendrick, 1999), which inoculates the flagellated promastigote form of L. infantum into the skin of the dog during a blood meal. If a dog becomes infectious, amastigotes present in the upper dermis and/or the bloodstream may be ingested by sand flies. The parasites migrate to the sand flies saliva and then to another susceptible dog via a subsequent bite, completing the life cycle (Saridomichelakis, 2009). In Mediterranean countries, phlebotomine sand fly activity is Supporting publications 2013:EN

115 Percentage of responses Impact, modelling and control of canine leishmaniosis in the EU seasonal (Rossi et al., 2008) and the vector over-winters in the larval stage (Killick-Kendrick, 1999). Activity is nocturnal or crepuscular during the transmission period (Killick-Kendrick, 1999). Other routes of transmission in dog populations may also play a role in the epidemiology of the disease. It was proposed that the spread of L. infantum in fox hounds in the USA was the result of dogto-dog transmission (Duprey et al., 2006, Gaskin et al., 2002). Case reports of natural Leishmania infection transmitted via venereal (Naucke and Lorentz, 2012, Silva et al., 2009) and vertical (Boggiatto et al., 2011) routes in dogs, provide further evidence of alternative transmission routes. Furthermore, blood transfusions have been shown to successfully transfer L. infantum infection between dogs (Owens et al., 2001). There was less evidence that non-sand fly vectors are important in the spread of CanL in dog populations, although naturally infected fleas (Coutinho and Linardi, 2007) and Rhipicephalus sanguineus ticks (Dantas-Torres, 2006a) were experimentally infectious to hamsters. A consensus of expert opinion provided further evidence that transmission of L. infantum by sand flies was of major importance in the overall spread of CanL in dog populations (LeishVet group, Figure 34). Transmission via blood transfusion was thought to have minor to moderate importance and all other routes were considered to be of minor or no / negligible importance. 100% 90% 80% 70% 60% 50% 40% 30% 20% 10% 0% Sand fly vectors Blood transfusion Venereal Vertical Direct Non-sand fly contact vectors between dogs Don't know None / negligible Minor Moderate Major Figure 34: LeishVet Expert opinion of the importance of transmission routes in dogs in the overall spread of leishmaniosis in endemic areas of the European Union (responses from six experts) Epidemiology in Humans Zoonotic leishmaniosis is an important public health concern in many countries (Ready, 2010, Quinnell and Courtenay, 2009). The most severe form of the disease in humans, visceral leishmaniosis (VL), is frequently fatal if left untreated (World Health Organization, Ready, 2010, Desjeux, 2004). Common clinical signs and clinicopathological abnormalities include organomegaly, lymphadenopathy, pyrexia, anaemia, leukopaenia and thrombocytopaenia (Fernandez-Guerrero et al., 2004, Maltezou et al., 2000, Lita et al., 2002, Papadopoulou et al., 2005). The cutaneous form of the disease is reported to be more benign and can manifest as single (Murray et al., Bongiorno et al., 2009) or multiple lesions (Bongiorno et al., 2009). Humans can be infected with L. infantum through the bite of an infected sand fly (Killick-Kendrick, 1999), the transmission route that all experts considered to be of major importance in the spread of human leishmaniosis in endemic parts of the EU (LeishVet group, Figure 35). However, non-vector routes have also been reported in human populations including vertical transmission (Meinecke et al., 1999, Boehme et al., 2006), venereal transmission (Symmers, 1960), transmission through organ transplant (Antinori et al., 2008), via blood transfusions (Cummins et al., 1995, Dey and Singh, 2006) and through sharing contaminated needles (Cruz et al., 2002, Morillas-Marquez et al., 2002). LeishVet expert opinion on the importance of these routes is presented in Figure 35. Supporting publications 2013:EN

116 Percent of responses Impact, modelling and control of canine leishmaniosis in the EU 100% 90% 80% 70% 60% 50% 40% 30% 20% 10% 0% Don't know No/negligible Minor Moderate Major Figure 35: Expert opinion of the importance of transmission routes in humans in the overall spread of leishmaniosis in endemic areas of the European Union (responses from six experts) CONCLUSIONS In summary, leishmaniosis is a vector-borne zoonosis which can cause disease in companion animals, wildlife species and humans. Haematophagous phlebotomine sand flies transmit the parasite between infectious and susceptible hosts, although transmission can occur by other non-sand fly routes. The majority of the literature and consensus of Leishvet experts would suggest that the dog is the primary reservoir host, although other domestic and non-domestic species may be reservoir hosts. Figure 36 indicates the relative impact of the disease in relation to its effect on animals and people based on the criteria detailed in the Phylum manual (Phylum, 2010). For details on the criteria, see Appendix R. Humans Affected species Persistence Transmission Figure 36: Epidemiological profile of L. infantum based on the Phylum criteria (Phylum, 2010). Maximum impact for a given parameter is represented by a score of 1 (minimum 0) on the respective axis. Supporting publications 2013:EN

117 3.3. Local Epidemiology MATERIALS AND METHODS Data collection The surveillance of CanL was discussed in a focus group with LeishVet members to direct primary data collection (LeishVet, 2012). National reference laboratories in each of the studied countries were contacted via to enquire about the possibility of sharing data on CanL. Data were also requested from a French benchmarking company (Panelvet, 2012), which extracts clinical and financial data from the practice management systems (computer based accounting and clinical record systems) of primary care veterinary clinics in France. The National Observatory of Leishmaniosis in Portugal (ONleish), ministries of agriculture, universities, and public health and private laboratories were also approached. An online questionnaire for veterinary practitioners, which included questions on the annual number of confirmed CanL cases and the estimated weekly canine consultation caseload, was translated into the relevant languages. Member organisations of the FVE (Federation of Veterinarians of Europe) and companion animal veterinary associations of the studied countries were contacted and sent links to the questionnaire to their members via or by publishing the link on their website. Web searches were used to identify dog rescue centres and shelters in France, Spain, Portugal, Italy and Greece. An online questionnaire, aiming to estimate the prevalence of CanL seen in non-owned dogs, was sent via to these organisations. The questionnaire was translated into the local languages. Respondents were asked to estimate the number of dogs which had stayed at the establishment in the last twelve months and whether they routinely tested for L. infantum. They were asked how many of these dogs tested positive for Leishmania and how many were sick as a result of CanL. The same questions were asked in relation to the current situation at the shelter (i.e. the current total number of dogs, the number testing positive and the number sick due to CanL) in case annual data were not available or less reliable. Published articles relating to the frequency and distribution of CanL in the studied countries were reviewed. The region, study period, sample size, sampling methods, diagnostic tests used and prevalence or incidence estimates were recorded for each article. Grey literature was also searched for relevant articles on the prevalence and incidence of leishmaniosis in the relevant countries. Factors which may influence the future epidemiology of leishmaniosis were also considered when undertaking the literature search. These include reported changes to the vector or pathogen, movement of pets and the potential impact of climate change Data analysis The proportion of practice-attending dogs treated for or euthanased due to CanL in France was estimated using invoice data from a veterinary bench-marking company (Panelvet, 2012). Cases were defined as dogs that received treatment for CanL (allopurinol, meglumine antimoniate, miltefosine) or dogs which were tested or treated (but not vaccinated) and euthanased within 60 days of the last test or treatment prescribed. Data provided by Panelvet were analysed using Microsoft Excel to generate descriptive statistics and exported to ArcGIS 9 (ESRI, 2013) to create maps showing the spatial distribution of CanL and administration of the CaniLeish vaccination. Standard error maps of these estimates were also created. Data collected from the online veterinary questionnaire were cleaned and analysed using Microsoft Excel. Prevalence (period prevalence, the proportion of practice-attending dogs with a confirmed Supporting publications 2013:EN

118 veterinary diagnosis of CanL within a 12 month period) was calculated at the NUTS 1, 2 and 3 levels using the following equation: The typical number of dogs seen per week was recorded and multiplied by the average number of working weeks per year to derive the annual number of consultations involving a dog. The number of working weeks per year was calculated by subtracting the average collectively agreed paid annual leave (weeks) for each country (Cabrita and Ortigão, 2011) from the number of weeks per year (52). To account for individual animals attending clinics more than once per year, this figure was divided by reported mean number of times a dog attends a clinic per year (Mercader, 2013, Panelvet, 2012). The mean and median NUTS 2 level prevalence estimates were calculated using standard methods. Incidence risk (cumulative incidence) was calculated by dividing the number of new cases by the estimated denominator population at risk during Questionnaires with missing numerator or denominator data were excluded from the prevalence and incidence estimates. The prevalence and incidence estimates were presented as choropleth maps in ArcGIS 9 (ESRI, 2013), using shape-files downloaded from a database of global administrative areas (GADM). In some instances, NUTS 3 regions were combined to become compatible for joining to shape-files. Maps showing the number of responses per NUTS 3 area and the standard errors of the prevalence and incidence estimates were also created. Data obtained from the questionnaire for dogs shelters were used to calculate within-shelter disease and infection prevalence using Microsoft Excel as defined for the veterinary questionnaire data. Data were aggregated to produce regional estimates and associated standard errors and presented as choropleth maps in ArcGIS 9 (ESRI, 2013). Supporting publications 2013:EN

119 RESULTS Benchmarking data (Panelvet, France) Ninety-seven clinics located in 52 French departments contributed data relating to approximately 180,000 dogs each year between August 2010 and August 2012 (See Appendix S. for distribution maps of contributing clinics). The mean numbers of dogs per clinic were 1,860 and 1,854 for each 12 month period respectively. The mean number of vets per clinic was 2.2 full time equivalents. A mean of 844 dogs were estimated to be under the care of each full time veterinary practitioner annually. The mean number of times dogs attended the clinics was reported to be 3.9 visits per year (Richard, 2012). The percentage of dogs treated for or euthanased due to CanL was 0.052% (95% CI: %) and 0.046% (95% CI: %) between August 2010 August 2011 and August August 2012 respectively. Eighteen (34.6%) departments had practices contributing data with at least one case of CanL and the apparent prevalence at the department-level ranged from 0% to 0.83% (Figure 37). During the first and second 12 months of the study period, 415 (0.23%) and 735 (0.41%) dogs were tested for Leishmania infection. The most commonly used tests were the SNAP tests produced by Virbac (n = 131, 11.1%) and Idexx (n = 334, 28.3%). One hundred and twenty nine tests (10.9%) were specified as being performed at the time of vaccination. The type of test used was not specified for 48% of invoices for Leishmania tests. A total of 565 (0.31%) dogs were vaccinated against Leishmania, following the introduction of the CaniLeish vaccine in August Twenty-seven departments (51.9% of departments represented by Panelvet practices) vaccinated at least one dog, and the proportion of dogs receiving the vaccine per department ranged from 0% to 9.53%. The pattern of vaccination administration generally reflected the disease prevalence (Figure 38). Supporting publications 2013:EN

120 Legend NUTS 3 (department) level percentage of practice-attending dogs treated for or euthanased due to CanL Figure 37a: 2011 Legend Standard error (SE) of NUTS 3-level percentage of practice-attending dogs treated for or euthanased due to CanL Figure 37c: SE 2011 Figure 37b: 2012 Figure 37d: SE 2012 Figure 37: Percentage of practice-attending dogs treated for or euthanased due to CanL in France 2011 (Figure 37a) and 2012 (Figure 37b) and standard error maps of these estimates (Figure 37c and d). Legend NUTS 3 (department) level percentage of practice-attending dogs vaccinated against CanL Legend NUTS 3 (department) level standard error of percentage of practice-attending dogs vaccinated against CanL Figure 38a Figure 38b SE Figure 38: Percentage of practice-attending dogs vaccinated for canine leishmaniosis in France, 2012 (Figure 38a) and a standard error map of these estimates (Figure 38b). Supporting publications 2013:EN

121 Veterinary questionnaire Impact, modelling and control of canine leishmaniosis in the EU One thousand, two hundred and thirty-one questionnaires were completed by veterinary practitioners, including 625 from France, 369 from Spain, 57 from Portugal, 67 from Italy and 113 from Greece. Thirteen questionnaires were not analysed as they were completed by vets working outside the studied countries or not currently in employment. The number of responses within countries varied by area, especially in France where higher responses were received from southern departments (Figure 39). The national (NUTS 1) prevalence estimates of veterinary-diagnosed CanL (percentage of practiceattending dogs with a confirmed veterinary diagnosis of CanL) ranged from 0.71% in France to 7.8% in Greece (Table 37). This represented a median of 1 case of leishmaniosis seen by the responding vets over the previous year in France and a median of 25 cases seen by the vets in Greece. Vets in Spain, Portugal and Italy saw a median of 8 cases per year; of which 5-6 were incident cases diagnosed within the last 12 months. Generally, prevalence and incidence estimates were highest in the Mediterranean regions of the studied countries (Figure 40 and Figure 42). Standard errors reflect that these estimates were generally more precise in parts of France and Spain compared to the other countries (Figure 41 & Figure 43). Differences between the national (NUTS 1) and average regional (NUTS 2) prevalence estimates were particularly marked for France (Table 37 and Table 38). See Appendix T. for individual NUTS 2 estimates and standardised morbidity ratios maps (NUTS 3). Table 37: Estimated frequency of confirmed veterinary diagnosis of CanL in dogs attending veterinary clinics over a 12 month period ( ) in France, Spain, Portugal, Italy and Greece. Cases per vet Median (range) Incident All cases Dogs per vet Median (range) Dogs per Consults per week year Dogs seen per year NUTS 1 Prevalence % (a) (95% CI) NUTS 1 Incidence per 1,000 (95% CI) Country France 1 (0-150) 1 (0-190) 50 (0-600) 2, ( ) (0-9400) ( ) ( ) Spain 6 (0-220) 8 (0-220) 30 (0-209) 1, (0-9798) (0 3266) ( ) ( ) Portugal 5.5 (0-80) 8 (0-80) 20 (1-100) ( ) ( ) ( ) ( ) Italy 5 (0-100) 8 (0-250) 25 (3-150) ( ) ( ) ( ) ( ) Greece 20 (0-200) 25(1-350) 30 (2-250) 1, ( ) ( ) ( ) ( ) (a) Prevalence = percentage of practice-attending dogs with a confirmed veterinary diagnosis of CanL Table 38: Regional estimates of the percentage of practice-attending dogs diagnosed with CanL. Regional (NUTS 2) prevalence estimates (%) Mean Median Range Country France Spain Portugal Italy Greece Supporting publications 2013:EN

122 Number of responses per NUTS 3 area Figure 39: Number of responses to a veterinary questionnaire which aimed to estimate the percentage of practice-attending dogs diagnosed with CanL in France, Spain, Portugal, Italy and Greece. Figure 40: Veterinary estimates of the percentage of practice-attending dogs diagnosed with CanL (pre-existing and incident) in France, Spain, Portugal, Italy & Greece over 12 months ( ). Figure 41: Standard error of the estimated percentage of practice-attending dogs diagnosed with CanL. Supporting publications 2013:EN

123 Figure 42: Estimated incidence of veterinary diagnosed CanL in dogs attending veterinary clinics (number of new CanL cases diagnosed per 100,000 practice-attending dogs over a 12 month period) Figure 43: Standard error of incidence of canine leishmaniosis in dogs attending veterinary clinics (standard error of number of new CanL cases diagnosed per 100,000 practice-attending dogs over a 12 month period) The most common forms of CanL control measures used in practice attending dogs were repellents and insecticides applied to the dog (Figure 44). In addition to the control measures specified in the questionnaire, some vets recommended keeping dogs indoors at night, using environmental insecticides and using drugs typically used to treat CanL as prophylactic medication. Allopurinol was the most frequently prescribed treatment for CanL, although meglumine antimoniate and miltefosine were used relatively frequently in some countries (Figure 45). In addition, some vets used treatments not specified in the questionnaire (antibiotics, corticosteroids, levamisole, complementary therapies, milbemycin, immunotherapy, neutraceuticals, ACE inhibitors, anthelmintics and autovaccination). The proportions of respondents using treatments not specified in the questionnaire were: 7.5% (France), 9.2% (Spain), 10.5% (Portugal), 17.9% (Italy) and 6.3% (Greece). Supporting publications 2013:EN

124 France Spain Portugal Italy Greece France Spain Portugal Italy Greece France Spain Portugal Italy Greece France Spain Portugal Italy Greece France Spain Portugal Italy Greece Responses France Spain Portugal Italy Greece France Spain Portugal Italy Greece France Spain Portugal Italy Greece Responses Impact, modelling and control of canine leishmaniosis in the EU 100% 90% 80% 70% 60% 50% 40% 30% 20% 10% 0% No reponse 0% 0<-20% 20-40% 40-60% 60-80% >80% Vaccination Repellents / insecticides (applied to the dog) Immune stimulants (domperidone) Figure 44: Veterinary estimates of the proportions of all dogs attending their clinic that receive control measures for leishmaniosis. The Y axis represents the percentage of respondents per country indicating that level of measure use. 100% 90% 80% 70% 60% 50% 40% 30% 20% 10% 0% No response Never Rarely Sometimes Often Always Allopurinol Meglumine antimoniate Miltefosine Domperidone Amphotericin B Figure 45: Frequency of treatments used for canine leishmaniosis in veterinary clinics in France, Spain, Portugal, Italy & Greece. The Y axis represents the percentage of respondents per country indicating that level of treatment use. Supporting publications 2013:EN

125 Responses Shelter data Impact, modelling and control of canine leishmaniosis in the EU Overall, 121 completed questionnaires were received from 517 dogs shelters between 19 th January and 1 st May 2013, giving an overall response rate of 23.4% (range between countries: %). The total number of dogs inhabiting the shelters at the time of questionnaire completion was 9,539, with a median of 52 dogs per shelter (range dogs). The number of dogs admitted to the shelters annually was estimated at 20,472, with a median of 133 dogs entering each shelter per year (range 5 3,000). Virtually all of the shelters rescued dogs from within the local area, 19.0% also took dogs in from other parts of the country and 8.3% rescued dogs that originated from other countries. Approximately one third of shelters routinely tested all dogs entering the premises for Leishmania infection, although the proportions of shelters routinely screening dogs varied between countries (Figure 46). Among populations undergoing routine screening, the proportion of dogs testing positive that showed clinical signs varied between shelters, countries and the time frame considered. For the population of dogs currently inhabiting the shelters, the level of clinical disease among dogs testing positive for L. infantum ranged from %, whilst for the population inhabiting the shelters over the last 12 months the range was %. See Appendix U. for details on the prevalence of L. infantum infection and disease in dogs routinely screened for Leishmania at shelters, and the proportions of dogs testing positive that showed clinical signs. 100% 80% 60% 40% 20% 0% France Spain Portugal Italy Greece All countries No response Don't know No Yes Figure 46: Proportion of 121 dog shelters in France, Spain, Portugal, Italy and Greece routinely screening dogs for L. infantum infection Table 39 and Table 40 show the apparent prevalence of clinical leishmaniosis in dogs inhabiting shelters over the two time periods, irrespective of whether they routinely test for Leishmania. Apparent prevalence (the percentage of dogs inhabiting the shelters that were sick due to CanL) at the time of questionnaire completion ranged from 0.26% in France to 5.18% in Greece. During the 12 month period prior to questionnaire completion, apparent prevalence ranged from 0.12% in France to 6.07% in Italy. Figure 47 shows local (NUTS 3) prevalence estimates of clinical CanL in the 12 month period prior to questionnaire completion and Figure 48 shows the standard errors of these estimates. These prevalence estimates were considered less likely to be influenced by seasonal effects than the point prevalence (current) estimates at the time of questionnaire completion. Furthermore, differences in euthaniasia policies (Carodog, 2013) and the considered ease of re-homing dogs with CanL (Figure 49) differed somewhat between the studied countries, which may have had a greater impact on point prevalence than period prevalence of CanL. Supporting publications 2013:EN

126 Table 39: Apparent point prevalence of clinical leishmaniosis in dogs inhabiting shelters in France, Spain, Portugal, Italy and Greece (prevalence at the time of questionnaire completion, 2013) Country Sick dogs (n) Number of dogs Prevalence % (95% CI) Responses to question (%) France ( ) 41 (95.3) Spain ( ) 26 (100.0) Portugal ( ) 11 (100.0) Italy ( ) 15 (93.8) Greece ( ) 23 (92.0) Total ( ) 116 (95.9) Table 40: Apparent prevalence of clinical leishmaniosis in dogs inhabiting shelters in France, Spain, Portugal, Italy and Greece (during the 12 month period prior to questionnaire completion, ) Country Sick dogs (n) Number of dogs Prevalence % (95% CI) Responses to question (%) France ( ) 14 (32.6) Spain ( ) 26 (100.0) Portugal ( ) 11 (100.0) Italy ( ) 14 (87.5) Greece ( ) 21 (84.0) Total ( ) 85 (70.2) Supporting publications 2013:EN

127 Responses Impact, modelling and control of canine leishmaniosis in the EU Figure 47: NUTS 3-level apparent prevalence of clinical leishmaniosis in dogs inhabiting animal shelters in France, Spain, Portugal, Italy and Greece, within a 12 month period ( ). Figure 48: Standard errors of apparent prevalence of clinical leishmaniosis in dogs inhabiting animal shelters in France, Spain, Portugal, Italy and Greece, within a 12 months period ( ). 100% 80% 60% 40% 20% 0% France Spain Portugal Italy Greece All countries No response Don't know No Yes Figure 49: Shelter response to question: Is it difficult to rehome a dog with leishmaniosis? The Y axis represents the percentage of respondents per country providing each answer. Supporting publications 2013:EN

128 Laboratory data Impact, modelling and control of canine leishmaniosis in the EU Regional canine data were received from the National Reference Centre for Leishmania, Italy (C.Re.Na.L.) and the Hellenic Pasteur Institute, Greece. Other reference laboratories did not respond to data requests or did not volunteer to share unpublished data National Reference Centre for Leishmania, Italy (C.Re.Na.L.) The Leishmania National Reference Centre in Palermo (C.Re.Na.L.) provides epidemiological surveillance from all Italian regions. Seroprevalence data are provided by the Istituti Zooprofilattici Sperimentali laboratory network (IZS), which use the indirect immunofluorescent-antibody test (IFAT) for diagnosis (Vitale, 2013). The cut-off value for seropositivity was established in every institute and is related to the prevalence-status of the region. Duplicate samples from the same animal were removed whenever possible (Vitale, 2013). During a five year period ( ), 351,957 samples were analysed and the seroprevalence and 95% confidence intervals were calculated for each region (See Appendix V. ). In many regions, the majority of samples were submitted in order to diagnose sick animals seen by vets but some regions carried out screening of healthy dogs in accordance with regional surveillance plans (Emilia- Romagna, Campania, Lazio and Sicilia) (Vitale, 2013), which may have resulted in comparably lower estimates than if only sick dogs were sampled. In 2009, the median regional seroprevalence was 19.7% (range %). The seroprevalence of L. infantum in the Italian regions in 2009 are presented in Figure 50. Figure 50a Figure 50b Figure 50: Regional seroprevalence of Leishmania infantum diagnosed at Istituti Zooprofilattici Sperimentali laboratories, Italy, 2009 (Figure 50a) and standard errors of these results (Figure 50b) National Reference Centre for Leishmania, Greece (Hellenic Pasteur Institute) The Hellenic Pasteur Institute receives human and canine samples from the Great Athens area (Dotsika, 2013). All of the samples were taken from suspect clinical cases and multiple samples from the same individuals were counted once. The same IFAT technique has been used for over 25 years and the cut-off titre for sero-positivity was 1:100 in dogs. In the last 5 years, private diagnostic veterinary laboratories have been established across Greece, so not all cases of CanL will be referred to the Hellenic Pasteur Institute. This is also likely to explain in part the observed decrease in the number of samples submitted for analysis in recent years (Figure 51). Supporting publications 2013:EN

129 The median annual percentage of samples testing positive was 23.4% (range %). See Appendix W. for further seroprevalence figures Number of samples % positive Figure 51: Annual number of canine samples tested and percentage testing positive for L. infantum at the Hellenic Pasteur Institute, Greece, The left Y axis refers to the number of samples and the right Y axis refers to the percentage of samples testing positive Published estimates of the prevalence of canine leishmaniosis in Europe Based on seroprevalence data from Italy, France, Spain and Portugal, it was estimated that 2.5 million dogs (16.7%) in western Mediterranean countries are infected with L. infantum (Moreno and Alvar, 2002). However, the prevalence of infection may be greater than this figure, as higher proportions of dogs test positive by PCR than serology in endemic areas. (Quinnell and Courtenay, 2009, Morales- Yuste et al., 2012, Baneth et al., 2008). For example, a cross-sectional study of 100 dogs living in a shelter in Mallorca (Spain), estimated a disease prevalence of 13% and a seroprevalence of 26%, yet PCR detected Leishmania DNA in at least one tissue in 63% of dogs (Solano-Gallego et al., 2001a). Furthermore, the sensitivity of PCR depended on which tissue was sampled (Solano-Gallego et al., 2001a). In addition to the influence of the type of diagnostic technique on apparent prevalence, the sensitivity and specificity of individual tests vary depending on the clinical severity, the function which dogs perform and the epidemiological characteristics of the area under study (Morales-Yuste et al., 2012). Sampling techniques and cut-off values also influence prevalence estimates (Morales-Yuste et al., 2012). Depending on the study objective, the time of year may also affect the results of epidemiological studies (Fernández-Bellon et al., 2008). The ranges of published regional prevalence estimates of clinical disease and infection due to L. infantum are presented for each country in Table 41. These estimates reflect that generally the apparent prevalence of infection is higher than that of disease and that the infection prevalence estimate will be influenced by diagnostic technique adopted (serology versus PCR). For further information on regional prevalence estimates from individual studies, including details of location and sampling, see Appendix X. Supporting publications 2013:EN

130 Table 41: Published CanL disease and infection prevalence estimates in France, Spain, Portugal, Italy and Greece Country Disease prevalence (%) Seroprevalence (%) PCR positive (%) References Min Max Min Max Min Max France (a) (Aoun et al., 2009, Lachaud et al., 2002) Spain (b) 67 (Alonso et al., 2010, Ballart et al., 2013, Miró et al., 2012, Tabar et al., 2008, Chitimia et al., 2011, Solano-Gallego et al., 2001a) (Cortes et al., 2012, Portugal (a) n/a Schallig et al., 2013, Cardoso et al., 2004a) Italy n/a (Morosetti et al., 2009, Rossi et al., 2008) (Leontides et al., 2002, Greece n/a (c) 61.9(d) Papadopoulou et al., 2005, Athanasiou et al., 2012) (a): Signs compatible with canine leishmaniosis; not all confirmed (b): Clinically healthy seronegative dogs (c): Symptomatic dogs only (d): Asymptomatic dogs only A risk map based on a systematic review of published literature on the distribution L. infantum (Trotz- Williams and Trees, 2003) showed that the majority of prevalence studied were conducted in the southern parts of the studied countries. The force of infection was calculated to be highest in the Mediterranean islands and southern Italy. A risk map for France, Spain, Portugal and Italy, based on environmental predictors, predicted prevalence to be highest in southwest France, southern Italy, Corsica and the Italian islands (Franco et al., 2011).. The results of three national surveys of CanL in veterinary clinics across France have shown the disease to be endemic along the Mediterranean coast and in the southeast part of the country (Bourdeau et al., 2011). The results of the second national survey, based on questionnaires completed by 992 veterinary clinics between 2002 and 2004, estimated the national prevalence of CanL to be 0.41% (Bourdeau et al., 2004). The survey was repeated in 2010 and an extension of cases into west and northwest France was observed (Bourdeau et al., 2011). In 2009, 3,974 dogs attending veterinary clinics across Portugal participated in a national survey (Cortes et al., 2012). Clinical signs consistent with CanL were apparent in 10.46% of this population. A direct agglutination test was used to detect infection and the true prevalence of CanL infection in Portugal was estimated to be 6.31% (apparent prevalence was 5.86%). The true prevalence of Leishmania infection per district varied from 0.94% to 17.40% (Cortes et al., 2012). A recent study of clinically normal dogs attending veterinary clinics in mainland Greece reported that 19.4% of 2,620 dogs were infected with Leishmania based on series testing with IFAT and an ELISA test. The regional seroprevalence estimates ranged from 2.1% in Florina to 30.1% in Attiki (Athanasiou et al., 2012). Supporting publications 2013:EN

131 Factors which may influence the future epidemiology of leishmaniosis In recent years, there has been evidence of the northward spread of leishmaniosis in Italy, which has been attributed to changes in vector distributions; possibly a result of climate change (Maroli et al., 2008, Baldelli et al., 2011). Climate change may increase the density of sand fly populations by shortening the lifecycle or extending the breeding season of the vector. Furthermore, higher temperatures may also allow the establishment of vectors in new locations (Maroli et al., 2008). These concerns have also been identified in Portugal and Spain (Casimiro et al., 2006, Martin-Sanchez et al., 2009). In addition to changes in the vector, changes to the Leishmania parasite itself may impact on the severity or frequency of disease. Meglumine antimoniate treatment failure rates in human VL patients have increased in southern Europe in recent years (Gradoni et al., 2003), possibly as a consquence of widespread use of this drug in dogs (Gramiccia et al., 1992). Furthermore, a genetic hybrid between two Leishmania species has been reported in Portugal, raising issues of the potential for the transfer of drug resistance and changes in virulence as a result of genetic exchanges (Ravel et al., 2006). Concerns over drug resistant prompted the development of an EC project Leishnatdrug-R ; the main objective of which was to design molecular tools to monitor the emergence and spread of drug resistance in leishmaniosis (LeishRisk, 2013). The general health status of a human population may impact on the occurrence of human cases of leishmaniosis (Gradoni et al., 2003). HIV is a risk factor for leishmaniosis (Alvar et al., 2008), and although the prevalence of HIV in Greece (0.2% in 2011) is lower than the other studied countries (World Health Organization, 2013a), there has been a rapid increase in the number of HIV case reports among injecting drug users in recent years (EMCDDA/ECDC, 2008, Kentikelenis et al., 2011). Although causation cannot be proved, there was a temporal association between a reduction of provision of prevention services and this outbreak. Historically, increases in HIV prevalence have coincided with economic crises (EMCDDA/ECDC, 2008), which may also have a possible impact on the occurrence of leishmaniosis. Increased travel and movement of pets and humans from endemic areas may also contribute to the changing epidemiology of the disease within and between countries (Menn et al., 2010, Teske et al., 2002, El Hajj et al., 2004, Dujardin et al., 2008, Leishrisk, 2007). CONCLUSIONS Based on the results of the veterinary questionnaire and published literature, there is a high prevalence of CanL in the Mediterranean coastal areas of the studied countries. Veterinary diagnosed cases were uncommon in central and northern France and parts of northwest Spain, although there is potential for spread within these countries as has been observed in Italy. Data from Greece suggests a high prevalence of the disease throughout the country. Data from shelters show a similar level of disease in non-owned dogs and laboratory data from Greece and Italy support high regional seroprevalence of L. infantum in suspect cases. Supporting publications 2013:EN

132 3.4. Economic Impact MATERIALS AND METHODS Data collection Financial data were requested from a veterinary bench-marking company (Panelvet, 2012), in order to estimate the cost of diagnosis and treatment and prophylactic vaccination for CanL. Three veterinary surgeons working in Leishmania-endemic parts of France were asked to comment on these figures and provide the typical cost of these services at their clinics. Comparable data were collected from two veterinary clinics in the autonomous community of Madrid, Spain. These were compared to prices on online veterinary pharmacies. Indirect losses, as a result of the impact of leishmaniosis on trade or tourism, were evaluated by searching secondary data sources and seeking expert opinion in the questionnaire sent to LeishVet members (as described in section 3.2). The number of dogs moved for commercial purposes was requested from the Community Trade Control and Expert System and chief veterinary officers in each studied country. The pharmaceutical companies manufacturing the CaniLeish vaccine, Scalibor deltamethrin collars and Advantix spot-on insect repellent were contacted and data on the numbers of products sold to each studied country were requested. Data from the online veterinary questionnaire (see section 3.3.4) were used to estimate the proportions of dogs using control measures in each country in order to estimate the possible costs related to use of these products. In addition to the economic losses as a result of the disease and control measures, the impact of the current economic crisis on veterinary services was explored. Both the LeishVet questionnaire and the online questionnaire for veterinary practitioners contained questions on the impact of the economic crisis on the prevention, treatment and euthanasia of dogs with CanL Data analysis The direct economic losses as a result of CanL were reported based on morbidity and mortality estimates derived from the primary data. Panelvet data was used to calculate the mean costs of vaccinating, testing or treating a dog for leishmaniosis by dividing the total value generated by each of these services by the number of transactions recorded. These values were compared to estimates derived from other sources. Only costs relating to canine disease were considered in this section. In order to estimate the total amount dog owners in the 5 EU countries spent on the primary CanLeish vaccination course in 2012, the proportion of dogs vaccinated was estimated from the online veterinary questionnaire data (see section 3.3.4). National canine population estimates (see Table 34) were then used to derive the estimated number of dogs vaccinated in clinics, which was subsequently multiplied by the mean cost of vaccination. In order to calculate the total amount dog owners spend on testing and treating CanL in the studied countries, the estimated number of CanL cases and the mean costs of testing and treating CanL were used. The estimated numbers of CanL cases per region (NUTS 2) were summed to calculate the approximate number of veterinary-diagnosed CanL cases per country. The regional number of CanL cases was calculated by multiplying the regional prevalence estimates from the online veterinary questionnaire (see Appendix T. ) by each estimated regional canine population. Published estimates of the regional canine population were not available, so it was assumed that the regional canine population would be proportional to the reported regional human population data. National canine Supporting publications 2013:EN

133 population estimates (see Table 34) were multiplied by the proportion of the human population inhabiting each region (Eurostat, 2013d). The mean regional CanL prevalence was used in regions where no veterinarians responded to the questionnaire. Overall country prevalence estimates (as opposed to regional level estimates) as presented in section were not used as there was regional variation in response rate within certain countries (particularly France), which appeared to have skewed some estimates. It was assumed that the CanL prevalence estimates from the veterinary questionnaire (the percentage of practice-attending dogs with a confirmed veterinary diagnosis of CanL) would reflect the prevalence of CanL among the general dog population. It was not possible to estimate the proportion of dogs that attended veterinary clinics in the studied countries. To calculate the cost of testing for L. infantum, it was assumed that for each dog diagnosed with CanL, three would be tested for L. infantum infection. Based on serological laboratory data, this was likely to be a conservative estimate ( % of samples tested positive at national reference laboratories in Italy and Greece respectively, see section 3.3.6). To calculate the approximate overall costs of treatment and mortality losses, it was assumed that the estimated 10% of CanL cases that died or were euthanased were not also treated, whilst the remaining 90% of dogs diagnosed with CanL at veterinary clinics would be treated. RESULTS The economic losses due to CanL can include direct losses as a result of mortality (monetary value and human-animal bond) and morbidity (cost of diagnosis and treatment) and indirect costs resulting from the impact on tourism, trade or animal movements Direct economic losses The economic impact of leishmaniosis reflects the value of dogs to society. The direct economic losses due to morbidity or mortality in companion animals can include both monetary and non-monetary components (Rushton, 2011). Companion animals can have a positive impact on their owners mental and physiologic health status (Friedmann and Son, 2009) and whilst the value of the animal-human bond is difficult to quantify, it has a non-monetary, metaphysical value. Additionally, companion animals can be defined as goods according to the civil codes of many EU countries (Carodog, 2013) and animals that are bought and sold will have an economic price and therefore have monetary value (Rushton, 2011). Additionally, the price an owner is willing to pay to keep a pet can be used as a proxy for their value (Shaw, 2012). Furthermore, working animals, such as hunting or guarding dogs, may perform an economic role which has a monetary value (Rushton, 2011). For example, in France alone there were estimated to be approximately 1,500 guide dogs, 6,000 to 7,000 service dogs, and 50,000 to 100,000 dogs used for sport (Grandjean, 2013). Therefore morbidity and mortality in dogs as a result of disease can be considered to result in direct economic loss, although the exact value would be difficult to establish Mortality There are few data on mortality estimates of dogs with clinical CanL. Oliveira et al. (2010) reported that veterinary clinicians in Portugal provided the following responses when questioned on how frequently euthanasia was elected instead of treatment for CanL: occasionally (58% of respondents), frequently (23%), never (14%) or always (5%). In the same survey, 43.1% and 34.3% of vets reported that less than 10% and between 10 and 25% of treated cases were euthanased respectively. One hundred and sixty-one (91.0%) of the leishmaniosis cases identified from the Panelvet data were treated, whereas sixteen (9.0%) cases were euthanased within 60 days of the last test or treatment. These figures suggest approximately 10 25% of diagnosed dogs were likely to be euthanased as a result of CanL. However, LeishVet experts reported that the likelihood of euthanasia depends on the severity of disease (see section 3.6.3). The replacement cost of dogs are highly variable; typically Supporting publications 2013:EN

134 around 100 for a crossbred and for a purebred dog in the some of the studied countries, although the cost of purebred animals can reach several thousand euros (Segundamano, 2013) Morbidity Cost of control measures In addition to mortality, the economic burden of disease includes the costs of prevention and the costs of treating affected animals (Rushton, 2011). Over 21 million dogs are estimated to live in the studied countries (see section 3.1). The number of dogs susceptible to and infected by Leishmania living in and travelling to the studied countries is considerable; even if the most conservative prevalence estimates, as described in section 3.3, are adopted. The cost of the primary vaccination course for CanL exceeds the cost of the other preventative measures, as summarised in Table 42. Table 42: Cost of control measures of canine leishmaniosis in France and Spain, 2013 (euros) France CaniLeish vaccination Primary course Per injection / booster Scalibor deltamethrin collars Advantix Pipettes (4) Small Large 1-4kg >25kg Leishguard 60ml Panelvet(a) Private clinics(b) Online pharmacy(c) Spain Private clinic(d) Private clinic(e) Online pharmacy(f) Online pharmacy(g) (a): Clinics contributing data to Panelvet benchmarking company (Panelvet, 2012) (b): Personal communication (Vergne, 2013): Data from 3 veterinarians working in the south east of France (c): Online pharmacy (Zubial.fr, 2013) (d): Personal communication (Hernaez Martinez, 2013) (e): Personal communication (Clinica Veterinaria Butragueno, 2013) (f): Online pharmacy (MedicAnimal.com, 2013) (g): Online pharmacy (Farmacia del Sol, 2013) A primary vaccination course (three injections) of the CaniLeish vaccination was typically between and Veterinary clinics in France purchase the vaccines from the wholesaler at a cost of (excluding VAT) and the mark-up to clients can reach up to 100% (Richard, 2012). If the proportions of dogs vaccinated for Leishmania, as estimated by the online veterinary questionnaire (see section 3.3.4) are representative of the situation in the general canine populations in the studied countries, an estimated 756,526 dogs may have been vaccinated in Assuming a mean cost of per primary vaccination course, a very rough estimate of 113 million may have been spent by dog owners in France, Spain, Portugal, Italy and Greece in 2012 (Table 43). However, dogs that have received the primary course will only need one booster injection in subsequent years (NOAH, 2013b), at a lower cost than the primary course. Furthermore, as the vaccine has only recently been Supporting publications 2013:EN

135 introduced, acceptance and up-take of this control measure may change over time. The pharmaceutical company that produces CaniLeish were unable to provide annual sales figures. However, the company was able to confirm that over 1 million doses had been sold by February 2013 (Schultheiss, 2013). Assuming 1 million doses were given as part of a primary course (3 doses per dog) at a cost of 150; owners would have spent approximately 50 million on vaccination, just under half of the figure estimated from the online veterinary questionnaire data. It is recommended that dogs are screened for Leishmania infection using a blood test prior to CaniLeish vaccination (NOAH, 2013b); an additional cost associated with this control measure (see below for estimates). Although the number of tests for Leishmania performed by Panelvet practices increased after the vaccine was introduced, the number of dogs vaccinated exceeded the number that were tested in 12 (44.4%) of the departments using the vaccine, indicating that not all of the veterinary practitioners complied with the vaccine manufacturer recommendations. One hundred and twenty nine Leishmania tests (10.9% of tests) were specified as being performed at the time of vaccination. Table 43: Estimated national costs of vaccinating, testing and treating canine leishmaniosis in France, Spain, Portugal, Italy and Greece and losses due to mortality (all costs are in euros) Canine population (n) (a) Vaccination (b) (%) Cost CanL cases (n) Cost of testing Cost of treating Mortality monetary losses Country France 7,590, % 15,675,000 27,204 1,958,712 2,448, ,043 Spain 5,498, % 34,342, ,823 13,523,256 16,904,059 1,878,229 Portugal 1,500, % 20,427,632 37,049 2,667,528 33,34, ,489 Italy 6,355, % 40,589, ,003 16,488,216 20,610,258 2,290,029 Greece 196, % 2,444,204 16,973 1,222,080 1,527,587 1,697,313 Total 21,139, ,478, ,052 35,859,792 44,824,693 4,980,521 (a): See Table 33 for references (b): CanLeish primary course of vaccination (three injections) An estimated 2.7 million deltamethrin impregnated Scalibor collars were sold in the studied countries in 2012 (Table 44) (Lussot, 2013). The Scalibor collar is sold as an antiparasitic for dogs against sand flies, ticks and mosquitoes but is primarily used in the studied countries for the prevention of CanL and is strongly marketed for that reason (Merck Animal Health, 2013). The cost of the collars varies depending on the seller and the collar size (Table 42) but assuming a mean cost of 20, this equates to the dogs owners in the studied countries spending 54 million on Scalibor collars in Table 44: Estimated sales of deltamethrin Scalibor collars in France, Spain, Portugal, Italy and Greece, 2012 (Lussot, 2013) Collars sold in 2012 Percentage of total collars sold Country France 500, Spain 1,000, Portugal 200, Italy 800, Greece 200, Total 2,700, Supporting publications 2013:EN

136 Cost of testing and treating a dog for leishmaniosis Impact, modelling and control of canine leishmaniosis in the EU Estimates of the cost of testing for Leishmania in veterinary clinics were: (in French PanelVet practices, based on testing 1,150 dogs at a total cost of 35,947.10) (Panelvet, 2012), 15( Speed Leish test from Virbac in clinics in France) (Vergne, 2013), and (for tests done at private clinics in Madrid, Spain) (Hernaez Martinez, 2013, Clinica Veterinaria Butragueno, 2013). A higher fee of was charged to send samples to an external laboratory in Spain (Clinica Veterinaria Butragueno, 2013). The Hellenic Pasteur institute reported that pet owners are required to pay for the IFAT tests performed at the laboratory, which costs (Dotsika, 2013). Assuming that for each dog diagnosed with CanL, three dogs would be tested at an average cost of 24 per test, it was estimated that dog owners in France, Spain, Portugal, Italy and Greece spent approximately 35.9 million on testing for L. infantum infection in 2012 (Table 43). In addition to performing tests to diagnose sick dogs, healthy dogs are screened for L. infantum infection as part of government surveillance programmes in parts of Italy (Vitale, 2013). Testing for Leishmania as part of these programmes is undertaken at public veterinary clinics (Sistema sanitario nazionale) and is free of charge to dog owners. Istituto Zooprofilattico Sperimentale laboratories, which process the samples, are funded by the Ministero della Salute (Baldi, 2013). If samples are taken at private veterinary clinics in order to diagnose clinical CanL cases, owners pay for the cost of diagnostic tests (Baldi, 2013). In addition to the costs of diagnostic tests, there may be additional fees to the owner including the cost of the veterinary consultation and sampling fees. The severity, duration and clinical features of CanL can vary between individuals (Solano-Gallego et al., 2009), which may result in different treatment plans and therefore costs for different cases. However, the mean cost of treating a dog with leishmaniosis was in clinics contributing benchmarking data in France (based on treating 161 dogs at a total cost of 8,295) (Panelvet, 2012). Generally, pet owners purchase medication for their animals directly from veterinary clinics in France, although on-line pharmacies are an alternative option. If a human drug is prescribed to an animal, this is dispensed by a human pharmacy (Desmas, 2012). Veterinary clinicians in France (Vergne, 2013) estimated that the typical cost of treating a dog for CanL at their clinics was 27 for a 5kg dog, 80 for a 15kg dog and 160 for a 30kg dog. Medication for companion animals can be obtained from both veterinary clinics and pharmacies in Spain. A veterinary clinician working in Spain charged clients an estimated for one month s treatment for a 30kg dog (15 boxes of Meglumine antimoniate). The cost of obtaining comparable treatment from a local pharmacy was estimated to be approximately a quarter of this price. Assuming that 90% of CanL cases receive treatment at a cost of , it was estimated that dog owners in the studied countries would spend approximately 44.8 million on treating CanL in 2012 (Table 43) Indirect economic losses Trade and movement restrictions As detailed in section 3.1, the legislation regarding the movement of pets focuses largely on rabies control (Europa, 2013f). There are no specific requirements relating to CanL, although the Italian Ministry of Health website recommends preventative measures for CanL when taking dogs to endemic regions (Ministero della Salute, 2013b). Dogs moved for commercial reasons must undergo a clinical examination by an authorised veterinarian to verify that the animal is in good health and fit to travel (Council of Europe, 1992). It is possible that dogs with clinical leishmaniosis intending to travel may not satisfy these requirements. In 2012, 3,591 dogs were notified as being imported to France from other EU member states and 918 dogs were exported from France to other member states for commercial purposes (Angot, 2013). Data were not available for other countries. Supporting publications 2013:EN

137 France Spain Portugal Italy Greece France Spain Portugal Italy Greece France Spain Portugal Italy Greece Responses Responses Impact on tourism Impact, modelling and control of canine leishmaniosis in the EU There is a strong tourist industry in all of the studied countries. The direct contribution of travel and tourism as a percentage of GDP ranged from % in the countries (see Table 33). LeishVet experts generally thought that leishmaniosis had a minor impact on tourism, although the responses ranged from no impact to high impact Impact of the current economic crisis on the veterinary care of dogs affected by leishmaniosis Generally, both LeishVet experts and veterinary clinicians responding to the online survey thought that the economic crisis was most likely to have a moderate to high impact on the use of prophylactic control measures and the diagnosis and treatment of dogs affected by leishmaniosis (Figure 52 and Figure 53). The responses relating to the impact on euthanasia decisions were more diverse. Generally, the impact of the economic crisis was considered to be higher in Spain, Greece and Portugal than France and Italy, which is consistent with the magnitude of economic hardship of the respective countries as reported in Table 33. In the free text answer, one LeishVet expert and several veterinary clinicians specified that the impact of the economic crisis on the use of the Leishmania vaccination was likely to be higher than the impact on the use of other prophylactic control measures (repellents, collars etc.). 100% 80% 60% 40% 20% 0% Prophylaxis Diagnosis and treatment Euthanasia decisions Veterinary service (relating to canine leishmaniosis) Don't know No impact Minor Moderate High Extreme Figure 52: Expert opinion of the likely impact of the current economic crisis on the veterinary care of dogs affected by leishmaniosis (responses from 6 experts). 100% 90% 80% 70% 60% 50% 40% 30% 20% 10% 0% No answer Don't know None Mild Moderate High Extreme Prophylaxis Diagnosis and treatment Euthanasia Figure 53: Veterinary opinion of the likely impact of the current economic crisis on the veterinary care of dogs affected by leishmaniosis in France, Spain, Portugal, Italy and Greece. Supporting publications 2013:EN

138 CONCLUSIONS In summary, approximate costs of control measures for a 30kg dog were as follows: primary vaccination course (three injections) of the CaniLeish vaccination was typically between and , Scalibor collars cost approximately 20 to 30, Advantix cost 30 to 40 and domperidone (Leishguard ) cost approximately 30 (a 30kg dog requires 2 units). Costs of diagnostic tests ranged from per test and treatment for a 30kg dog was approximately 160 to 270 if purchased from veterinary clinics. Mean cost figures and data from industry and the online veterinary questionnaire were used to estimate that dog owners spent between million on CanL vaccination, 35.9 million on testing and 44.8 million on treating CanL in Based on the industry data, owners spent an additional 54 million on repellent collars alone across the studied countries in Mortality as a result of CanL can result in monetary losses due to the cost of euthanasia and replacing a deceased pet, in addition to loss of the non-monetary value of the human-animal bond. Although the cost to dog owners may be high, veterinary clinics and pharmaceutical companies may derive financial benefit from the prevention and treatment of CanL. The indirect economic impacts on trade and tourism were considered likely to be minor. The current economic crisis in the EU was likely to have a moderate to high impact on the use of prophylactic control measures and the diagnosis and treatment of dogs affected by leishmaniosis. Supporting publications 2013:EN

139 3.5. Human health impact MATERIALS AND METHODS Data collection Data on the annual number of notified or hospitalised human leishmaniosis cases were requested from the national ministries of health, statistical authorities and reference centres for human leishmaniosis in each country. Summary data were sourced from online databases and reports. Journal articles and grey literature were searched for information on the typical severity of the illness Data analysis Where available, the incidence risk of leishmaniosis or the hospitalisation rates were calculated based on the annual number of cases or hospitalisation episodes per region respectively. Population estimates derived from the national institute of statistics (INE, 2013b, Istat, 2013, INE, 2013a, EL.STAT, 2010) were used as denominators. Standardised morbidity ratios (SMRs), which are the ratios between the observed number of cases and the expected number of cases assuming a homogenous spatial distribution were calculated using Microsoft Excel when regional data were available. The data were then used to create maps of incidence and SMR using ArcGIS 9 (ESRI, 2013). In addition, descriptive statistics such as mortality rates, median hospital duration and cost per case were calculated. Only cases with a primary diagnosis of visceral leishmaniosis, cutaneous leishmaniosis (urban) and leishmaniosis (unspecified) were included. Patients with leishmaniosis recorded as a secondary diagnosis and leishmaniosis not caused by L. infantum infection were not included in the analysis. The results were used to create a radar chart summarising the zoonotic profile of L. infantum, based on the criteria detailed in the Phylum manual (Phylum, 2010). RESULTS Annual number of human cases Reported cases Notification of visceral leishmaniosis (VL) and cutaneous leishmaniosis (CL) is compulsory in Greece, Italy and 12/17 autonomous communities in Spain. VL, but not CL, is notifiable in Portugal and neither form of the disease is notifiable in France (Dujardin et al., 2008). A World Health Organisation-led project estimated that under-reporting rates of VL and CL in the studied countries were fold and fold respectively (Alvar et al., 2012). The Centralized Information System for Infectious Diseases The Centralized information system for infectious diseases (CISID) holds data collected using standardised reporting forms from different sources. Data on the annual number of reported leishmaniosis cases and incidence per 100,000 inhabitants from 2000 to 2010 are summarised in Table 45 (CISID, 2013). The median incidence ranged from 0.11 cases per 100,000 inhabitants in Portugal to 0.42 cases per 100,000 inhabitants in Greece. The case definition used for this data was requested via but no response was received. Supporting publications 2013:EN

140 Table 45: Annual number of reported human leishmaniosis cases and incidence per 100,000 inhabitants ( ) (CISID, 2013) Country Annual number of cases Incidence per 100,000 inhabitants Median Range Median Range France Spain Portugal Italy Greece Recent estimates Further to the published data above, recent numbers of notified cases were obtained through personal communications or from online publications. France: the Centre National de Référence des Leishmanioses (CNRL) collects data on the number of autochthonous (indigenous) and imported cases of leishmaniosis diagnosed in France (CNRL, 2013). In 2010, 169 cases of human leishmaniosis were declared in metropolitan France, 11 of which were thought to be autochthonous (indigenous) cases (CNRL, 2013). In 2009, autochthonous cases were exclusively located in departments in southeast France (Dedet, 2009). Spain: the number of leishmaniosis cases reported by the National Epidemiological Surveillance Network was 271 in 2011 and 244 in 2012 (provisional data) (Rodriguez Valin, 2013). These estimates are much higher than the median annual number of cases in Table 45. Portugal: the annual numbers of human leishmaniosis cases, as reported by the surveillance system of national notifiable diseases, were 11, 11, 8 and 11 cases ( ) (Sousa Pinto, 2013), in line with the estimate reported in Table 45. Italy: National epidemiological bulletins, published by the Italian Ministry of Health, summarise the numbers of notified cases of visceral (VL) and cutaneous (CL) leishmaniosis (Ministero della Salute, 2013c). The median (range) of reported VL and CL cases were 128 (22 185) and 36 (8 73) respectively. See Appendix Y. for annual figures. Greece: The Hellenic Center for Disease Control and Prevention, Ministry of Health, produce monthly reports on the number of cases, location and age/sex distributions of notifiable diseases (HCDCP, 2013). Forty-one cases were reported in 2012, similar to the median annual number of cases in Table 45. See Appendix Y. for further details Hospital data Data were received from the Ministries of Health in Italy and Spain (Ministero della Salute, 2012, Ministerio de Sanidad, 2013a). The hospital discharges registers includes all discharges produced in general public hospitals in Spain (Ministerio de Sanidad, 2013b) and 99.7% of public and 95.5% of private hospitals in Italy (Ministero della Salute, 2011). Summary data on leishmaniosis cases in public hospitals in mainland Portugal (Jorge Nogueira and Valente Rosa, 2013b, Jorge Nogueira and Supporting publications 2013:EN

141 Valente Rosa, 2013a, Jorge Nogueira and André Giria, 2012, Jorge Nogueira and André Giria, 2010, Jorge Silva et al., 2008a, Jorge Silva et al., 2008b) and all private and public hospitals in Greece (EL.STAT, 2013) were available from published reports and an online database. No hospital data were available from France. All countries that provided data used The International Classification of Diseases codes (World Health Organization, 2013c). The annual incidence risk was highest in Greece (1.35 cases per 100,000 inhabitants during 2008, (95% CI: )), although data were only available for one year (Table 46). Incidence risk was generally higher in Spain (range cases per 100,000 inhabitants) than Portugal (range cases per 100,000 inhabitants). Incidence risk in Italy could not be calculated as it was not possible to identify whether episodes of hospitalisation were re-admissions; some patients may have been counted more than once. Twenty-six per cent of all the hospital admissions in Spain were readmissions. The hospitalisation rates were similar for Italy (range hospitalisations per 100,000 inhabitants) and Spain (range hospitalisations per 100,000 inhabitants). Table 46: Annual number and incidence of hospital admissions and patients with a primary diagnosis of leishmaniosis in Spain, Portugal, Italy and Greece. Year Annual number of hospital admissions with leishmaniosis as a primary diagnosis Spain Portugal Italy Greece Annual number of patients with leishmaniosis as a primary diagnosis Spain Portugal Italy Greece Annual incidence risk of hospitalisation for leishmaniosis per 100,000 inhabitants (95% CI) Spain ( ) 0.32 ( ) 0.48 ( ) Portugal 0.39 ( ) 0.24 ( ) 0.29 ( ) 0.16 ( ) Greece 1.35 ( ) Annual hospitalisation rate for leishmaniosis per 100,000 inhabitants (95% CI) Spain ( ) 0.41 ( ) 0.61 ( ) Italy ( ) 0.52 ( ) 0.46 ( ) The majority (78.2%) of leishmaniosis cases hospitalised in Spain were coded as the visceral form, 16.7% were unspecified forms of the disease, 0.9% were urban cutaneous leishmaniosis and 4.1% were not thought to be caused by L. infantum, so were excluded from the analysis. Approximately two-thirds (64.4%) of hospitalisations with a primary diagnosis of leishmaniosis in Italy were coded as visceral leishmaniosis, 21.1% were unspecified forms, 10.6% were cutaneous and 3.9% were not thought to be caused by L. infantum. Supporting publications 2013:EN

142 The incidence risk of hospitalised human leishmaniosis cases was generally highest in the south and east and lowest in the north and west of Spain (Figure 54). In 2011, the incidence in Madrid increased, reflecting an outbreak reported the Fuenlebrada region (Aguado et al., 2013). Legend Annual incidence of hospitalised leishmaniosis cases per 100,000 inhabitants Legend Standard error of annual incidence of hospitalised Leishmaniosis cases per 100,000 inhabitants inhabitants Figure 54a:2009 Figure 54d:2009 Figure 54b:2010 Figure 54e:2010 Figure 54c:2011 Figure 54f:2011 Figure 54: Annual incidence of hospitalised patients with a primary diagnosis of leishmaniosis per 100,000 inhabitants in 2009 (Figure 54a), 2010 (Figure 54b) and 2011 (Figure 54c) and standard error of annual incidence in 2009 (Figure 54d), 2010 (Figure 54e) and 2011 (Figure 54f). Supporting publications 2013:EN

143 Figure 55 shows that the highest rates of hospital admissions for leishmaniosis were generally in provinces in Sicily, Liguria and Sardinia. See Appendix Z. for maps showing the SMRs of the regional incidence of hospitalised human leishmaniosis cases in Spain and Italy. Legend: Annual rate of hospitalisations for leishmaniosis per 100,000 inhabitants Legend Standard error of hospitalisation rate Figure 55a: 2009 Figure 55d: 2009 Figure 55b 2010 Figure 55e 2010 Figure 55c 2011 Figure 55f 2011 Figure 55: Province level incidence rates of hospital admissions with a primary diagnosis of leishmaniosis, 2009 (Figure 55a), 2010 (Figure 55b) and 2011 (Figure 55c) and standard error of incidence rates of hospital admissions in 2009 (Figure 55d), 2010 (Figure 55e) and 2011 (Figure 55f), Italy. Supporting publications 2013:EN

144 Laboratory data (Hellenic Pasteur Institute, Greece) Impact, modelling and control of canine leishmaniosis in the EU The Hellenic Pasteur Institute receives samples from suspect human cases in hospitals in the Greater Athens area, Greece. The annual numbers of samples received from humans were much lower than the annual number of samples from dogs (see section ), although the numbers submitted from both species has declined in recent years. The cut-off for sero-positivity in humans was 1:400 (IFAT) (Dotsika, 2013). The median percentage testing positive each year was 4.6% (range %) (Figure 56). See Appendix W. for further seroprevalence figures. No other human laboratory data were available from the other studied countries Number of samples % positive Figure 56: Annual number of human samples tested and percentage positive for Leishmania infantum at the Hellenic Pasteur Institute, , Greece. The left Y axis refers to the number of samples and the right Y axis the % positive Typical and maximum severity A review by Michel et al. (2011) concluded that most humans infected with L. infantum have asymptomatic infection. Risk factors for clinical disease include age (younger people) (Maltezou et al., 2000), HIV infection (Desjeux and Alvar, 2003) and other immunosuppressive states (Antinori et al., 2008). The prevalence of HIV among hospital patients with a primary diagnosis of leishmaniosis in Spain (data described in section ) was 21.33% and co-infection with HIV was coded for 7.0% of hospital admissions with a primary diagnosis of leishmaniosis in Italy. When compared with the prevalence of HIV among the general population aged (0.4% for both countries (World Health Organization, 2013a)), there appears to be an association between being diagnosed with leishmaniosis and HIV, as reported in previous work. A bimodal age distributions was evident in the hospital data from Spain and Italy, with peaks in infancy and middle age (data described in section ). The median age of hospital admission due to leishmaniosis in Spain was 38 years (range 0 to 95 years). Cutaneous leishmaniosis commonly presents as single or multiple erythematous plaques or papules (Aguado et al., 2013). Visceral leishmaniosis is more severe and common clinical signs and clinicopathological abnormalities include organomegaly, lymphadenopathy, pyrexia, anaemia, leukopaenia and thrombocytopaenia (Fernandez-Guerrero et al., 2004, Maltezou et al., 2000, Lita et al., 2002, Papadopoulou et al., 2005). The cure rate of leishmaniosis with treatment is reported to exceed 95% in France, Italy, Portugal and Greece; Spain reported a fatality rate for visceral leishmaniosis of 5% (Alvar et al., 2012). Mortality rates of 4.1% and 2.7% in hospitalised leishmaniosis cases in Portugal and Spain respectively were calculated from the data described in section All of the fatal leishmaniosis cases in Spain were diagnosed with co-morbidities, including two patients with concurrent HIV. Supporting publications 2013:EN

145 Economic impact of human disease Impact, modelling and control of canine leishmaniosis in the EU The median estimated cost per admission for leishmaniosis in Spain was 4, (range 2, ,071.83). The mean cost of leishmaniosis cases to public hospitals in Spain was estimated to be 1,234,818 annually (based on data described in section ). Assuming equivalent costs per hospitalisation and admission rates in Portugal, Greece and Italy; hospital fees from admissions with a primary diagnosis of leishmaniosis could cost 4,085,298 in the four countries that provided hospital data. The median number of days a patient with leishmaniosis spent in hospital were as follows: 9 days (range days) in Spain, 12 days (range days) in Italy, 9 days in Greece and 8 (2010) and 9 (2011) days in Portugal (based on data described in section ). In addition to hospitalisation costs, an indirect economic cost results from loss of earnings of patients due to time off work. Days off work due to leishmaniosis could equate to labour costs of 427,940 annually in Spain, Portugal, Italy and Greece. This estimate was based on the typical length of hospitalisation, the estimated mean number of admissions and the labour cost per day (Eurostat, 2013b) for each country. Employment rates were based on the proportion of patients aged 16 to 60 year in Spain and recent unemployment rates in each country (Eurostat, 2013a). No hospital data were available from France Cases attributable to animals and animal-human interactions As described in section 3.1, people often live in close proximity to dogs, which are considered to be the main reservoir of infection. LeishVet experts considered transmission of the parasite to people via sand flies was of most importance. However, other routes, especially needle sharing in intravenous drug users were also considered to play a less important role in the transmission of leishmaniosis (see section 3.2.6). It was not possible to estimate the frequency of leishmaniosis cases attributable to people sharing close proximity with animals, although infections via sand flies (most likely infected from non-human animals) are likely to be responsible for most human cases. CONCLUSIONS In summary, the incidence of clinical human leishmaniosis in the studied countries was relatively low and broadly follows a similar geographical pattern to CanL. However, certain groups were at a higher risk of disease and VL was reported to be fatal in % of cases. The economic impact due to human disease is not insignificant; costs relating to hospital fees and time off work may amount to approximately 4.5million/year in Spain, Portugal, Italy and Greece. Figure 57 represents the zoonotic profile of L. infantum based on the Phylum manual (Phylum, 2010). See Appendix R. for details. Human-to-human transmission Severity in humans Diagnosis in humans Animal-to-human transmission Prevention and treatment Figure 57: Zoonotic profile of Leishmania infantum based on the Phylum criteria (Phylum, 2010). Supporting publications 2013:EN

146 3.6. Societal Impact MATERIALS AND METHODS Data collection The impact on animal welfare was assessed by searching journal articles for the prevalence of clinical signs in affected dogs and potential adverse effects of therapy for CanL. Data sheets for preventative measures and treatments for CanL were consulted to identify known adverse effects. The LeishVet questionnaire included questions on the animal welfare impact of CanL and the control measures for the disease. RVC colleagues working in the animal welfare field were consulted during the construction of the questionnaire. The crisis generation potential and the impact of the media were assessed by including relevant questions in the LeishVet questionnaire. Search terms related to leishmaniosis and other selected zoonotic diseases were entered into Google Trends (Google Trends), so that the relative frequency of web searches (used as a proxy for public interest and concern) for these diseases could be compared over time and geographically in each of the studied countries Data analysis The results of the questionnaire and literature search were used to create stacked bar charts and a radar chart, based on the criteria detailed in the Phylum manual (Phylum, 2010). See Appendix R. for details. Graphs generated by Google (Google Trends) compared the temporal web interest in leishmaniosis with other selected zoonotic diseases, based on web searches of relevant terms. Major news headlines relating to web search terms were also indicated on the graphs. The geographical differences in web interest, based on the frequency of search term usage in different locations, was represented by circles (frequency of web searches in towns and cities) or by choropleth maps (frequency of web searches in a region). RESULTS Animal welfare Infection with L. infantum can result in a wide spectrum of clinical signs (Solano-Gallego et al., 2011), so the impact of CanL on the welfare of a dog will depend on how the disease manifests itself in that individual. Skin lesions are the most common manifestation of CanL (Solano-Gallego et al., 2011) and include papular dermatitis (Ordeix et al., 2005), alopecia and desquamation, ulcerative dermatosis, nodular disease, pustular dermatitis (Ferrer et al., 1988) and onychogryphosis (Ciaramella et al., 1997). Lymphadenomegaly, splenomegaly, lethargy and weight loss are also common clinical findings (Ciaramella et al., 1997, Slappendel, 1988). Glomerulonephritis and proteinuria are frequently reported in CanL (Costa et al., 2003, Zatelli et al., 2003) and although azotaemia is a relatively uncommon finding, severe renal failure can develop and is a major cause of mortality in affected dogs (Solano-Gallego et al., 2011, Ciaramella et al., 1997). A range of ocular and periocular manifestations (Ciaramella et al., 1997, Peña et al., 2000), polyarthritis (Santos et al., 2006), bone lesions (Agut et al., 2003) and epistaxis have also been reported (Petanides et al., 2008). The LeishVet group have developed a staging system for the disease. The four clinical stages, ranked by severity, are based on the serological status, clinical signs and laboratory findings of affected dogs (Solano-Gallego et al., 2009). The LeishVet experts agreed that stage I (mild) disease had no to minor impact on the welfare of affected dogs (Figure 58), and was most likely to result in complete recovery Supporting publications 2013:EN

147 Stage I Stage II Stage III Stage IV Stage I Stage II Stage III Stage IV Stage I Stage II Stage III Stage IV Stage I Stage II Stage III Stage IV Percentage of responses Stage I Stage II Stage III Stage IV Stage I Stage II Stage III Stage IV Stage I Stage II Stage III Stage IV Stage I Stage II Stage III Stage IV Percentage of responses Impact, modelling and control of canine leishmaniosis in the EU (Figure 59). Stage IV (very severe) disease generally had a high to extreme impact on animal welfare, although 60% of experts thought that this stage had a moderate impact on pain. Complete recovery was thought to be rare or never occur and death and euthanasia were considered common outcomes in dogs with stage IV disease. Stages II and III of disease had intermediate scores. It was not possible to estimate the frequency at which each clinical stage of disease occurred in the canine population. 100% 90% 80% 70% 60% 50% 40% 30% 20% 10% 0% No impact Minor Moderate High Extreme Demeanour Pain Physical / neurological Essential functions Figure 58: Expert opinion of the animal welfare impact of clinical stages (I-IV) of canine leishmaniosis (responses from six experts). 100% 90% 80% 70% 60% 50% 40% 30% 20% 10% 0% Always Often Sometimes Rarely Never occurs Complete recovery Relapses Long term / permanent disease Death or euthanasia due to leishmaniosis Figure 59: Expert opinion on the likely outcome of clinical stages (I-IV) of canine leishmaniosis (responses from six experts). Adverse effects of the CaniLeish vaccine include local reactions, which may resolve spontaneously in 2 15 days and hyperthermia, apathy and digestive disorders which are common and may last 1 6 days. Anaphylaxis is reported to be uncommon (NOAH, 2013b). Insecticides or repellents can be applied to dogs as spot-ons, sprays or collars (Podaliri Vulpiani et al., 2011). The datasheets of licensed insecticide / repellent products report that skin sensitivity, lethargy, behaviour changes, gastro-intestinal symptoms and neurological signs rarely occur following application of imidacloprid and permethrin (NOAH, 2013a) or deltamethrin (NOAH, 2013c). Central effects have been reported to Supporting publications 2013:EN

148 rarely occur following administration of domperidone (Gomez-Ochoa et al., 2009), an immune stimulant which has been used as a prophylactic medication and treatment for CanL. The side effects of other treatments for CanL are variable. Allopurinol can promote hyperxanthinuria and xanthinecontaining urolith formation (Osborne et al., 2009), whereas a potential side effect of meglumine antimoniate and amphotericin B treatment is nephrotoxicity (Baneth and Shaw, 2002). In clinical trials, adverse effects associated with miltefosine were vomiting (16% of dogs treated) and diarrhoea (12% of dogs treated), which lasted 1-2 days in most cases (Virbac, 2013) LeishVet experts generally reported that the control measures with the lowest adverse impact on animal welfare were repellents and insecticides and immune stimulants (Table 47). Culling was reported to have the highest adverse effect on the welfare of dogs, although only 50% of experts responded to this question. Most experts thought that the public were likely to completely or mostly accept vector control measures, whereas culling pet dogs was thought to be the least publically acceptable control measure (Figure 60). LeishVet experts considered that the most common adverse effects of each treatment had similar impacts on dog welfare, although adverse effects of allopurinol were reported to occur less frequently but have a longer duration than the other treatments. Table 47: Expert opinion of the animal welfare impact of the control measures and treatments of canine leishmaniosis in relation to the expected frequency, intensity and duration of the compromise to welfare (responses from 6 experts). Scores relate to minimum 1 and maximum 5 (see footnote) Median (range) Frequency (a) Control / preventative measure Number of responses Median (range) Intensity (b) Number of responses Median (range) Duration (c) Number of responses Vaccination 3 (1-3) 4 2 (2-3) 4 3 (2-3) 4 Repellents / insecticides applied to the 2 (1-2) 6 2 (1-2) (2-4) 5 dog Immune stimulants 2 (1-2) 5 1 (1-3) (1-3) 4 Culling 5 (4-5) 3 5 (3-5) 3 5 (1-5) 3 Movement restrictions 3 (3) (2-3) (2-5) 2 Treatment Allopurinol 2 (2-3) (2-3) 6 4 (3-4) 5 Meglumine antimonite 3 (3-4) 6 3 (2.5-4) 6 3 (3-4) 6 Miltefosine 3 (2-3) 5 3 (2-3) 5 3 (2-3.5) 5 Amphoterin B 3 (3) 1 3 (3) 1 3 (3) 1 Domperidone 2 (2) 1 2 (2) 1 3 (3) 1 (a): 1 = very rarely occurs, 2 = rarely occurs, 3 = sometimes, 4 = often, 5 = always occurs (b): 1 = very minor impact, 2 = minor, 3 = moderate, 4 = high, 5 = extreme impact (c): 1 = seconds / minutes, 2 = hours, 3 = days, 4 = months, 5 = years / permanent Supporting publications 2013:EN

149 Responses Responses Impact, modelling and control of canine leishmaniosis in the EU 100% 90% 80% 70% 60% 50% 40% 30% 20% 10% 0% Don't know Completely acceptable Mostly acceptable Some acceptance Mostly unacceptable Completely unacceptable Figure 60: Expert opinion on the likely public acceptance of control measures of canine leishmaniosis (responses from six experts). The Y-axis represents percentage of responses Crisis generation potential The crisis potential of a disease relates to the likelihood that it would create a major public concern (Phylum, 2010). This depends on the zoonotic potential, the seriousness and the acceptable level of risk, the existence of public awareness campaigns and amplifying effect of the media (Anses, 2012). LeishVet experts generally reported that leishmaniosis had more impact via outbreak reports than public awareness campaigns. Expert opinion on the overall level of public concern was variable; the most frequent responses were there was moderate to high levels of concern in endemic areas (Figure 61). 100% 80% 60% 40% 20% 0% Media coverage (outbreak Media coverage reports) (public awareness Level campaigns) of public concern Don't know None / no impact Minor Moderate High Extreme impact / concern Figure 61: Expert opinion on the media coverage and level of public concern relating to canine leishmaniosis in endemic parts of the EU (responses from six experts) Figure 62 and Figure 63 show the Google trend graphs and maps which represent the relative temporal and regional web search volumes for leishmaniosis and other zoonotic diseases. The temporal web search volumes for leishmaniosis (Figure 62) are comparable to the search volumes for other zoonoses within a country but are not comparable between countries; the height of the lines are relative to the peak web search for any search term, not exclusively searches for leishmaniosis (100 on the y-axis). Due to the limitations on searches imposed by Google trends, it was not possible to explore the relative web search volumes between the studied countries. Supporting publications 2013:EN

150 Figure 62a Impact, modelling and control of canine leishmaniosis in the EU Figure 62b Rabies Leishmaniosis Canine leishmaniosis Babesiosis News headlines Figure 62c Figure 62d Figure 62: The number 100 on the y-axis represents the relative peak web search interest Figure 63: Larger, darker circles represent relatively higher web interest in leishmaniosis in towns and cities Figure 62: Web search interest and news headlines relating to zoonoses, France (Fig.62a), Spain (Fig. 62b), Portugal (Fig. 62c) &Italy (Fig. 62d) Figure 63a Figure 63b Figure 63c Figure 63d Figure 63e Figure 63: Local web search volume for leishmaniosis in France (Figure 63a), Spain (Figure 63b), Portugal (Figure 63d) and Italy (Figure 63e), 2007 to 2013, and web search volume in Spain in 2012 only (Figure 63c). Supporting publications 2013:EN by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender considered as an output adopted by the Authority. The European food Safety Authority reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

151 Generally, web searches for leishmaniosis were more frequent than web searches for babesiosis but less frequent than for rabies in the studied countries (Figure 62). In 2011 and 2012, there were peaks in web interest for leishmaniosis in Portugal and Spain respectively (Figure 62c and b). Peaks in Italy appeared to correspond to news headlines and followed a seasonal pattern (Figure 62d). Regional web interest in leishmaniosis over a six year period largely reflected the distribution of cases in humans and dogs in Spain, Italy and France (Figure 63). In 2012, the regional interest in Spain was highest in Madrid (Figure 63c). This was likely to have reflected increased public interest relating to an outbreak of human cases in Fuenlebrada, Madrid, during this period (Aguado et al., 2013, Molina et al., 2012). The spike in web searches volume in Spain (Figure 62b) also corresponded to the time period of the outbreak. The frequency of web searches using the terms dog and rabies were generally more uniform across regions when compared with the searches for leishmaniosis in Spain, France and Italy. Web interest in all three search terms was concentrated in coastal regions of Portugal (see Appendix AA. for distribution maps for the results for all search terms). Search volumes in Greece were not sufficient to create graphs and maps Bioterrorism potential Leishmania infantum is not listed in the OIE/CFSPH classification of pathogens or the USDA list of selected agents and toxins (Phylum, 2010). There was almost complete consensus among LeishVet experts that there was no potential to use the pathogen as a bioterrorism agent (one expert thought that there was low potential). CONCLUSIONS Based on the expert opinion of six LeishVet members, the impact on dog welfare and the likely prognosis of CanL depends on the clinical stage of the disease, although in very severe cases the disease was thought to have an extreme impact on welfare. Adverse effects are reported for vaccination, repellents and prophylactic medications for CanL. LeishVet experts generally considered that repellents and insecticides applied to the dog and prophylactic medication had the lowest impact on dog welfare and vector control was thought to be most publicly acceptable control measure. Based on the level of web interest, temporal and regional public interest in leishmaniosis largely reflected the occurrence of human and canine cases in Spain, Italy and France. There was insufficient evidence that L. infantum has the potential to be used as a bioterrorism agent. Figure 64 represents the societal impact of CanL based on the criteria detailed in the Phylum manual (Phylum, 2010). For further details on the Phylum criteria, please consult Appendix R. Figure 64: Societal impact of canine leishmaniosis based on the Phylum criteria (Phylum, 2010). Maximum impact for a given parameter is represented by a score 0-1 (minimum 0) on the respective axis. Supporting publications 2013:EN

152 3.7. Environmental impact MATERIALS AND METHODS Data collection The impact of leishmaniosis on biodiversity was assessed by searching the literature for case reports and epidemiological studies relating to clinical disease in wildlife species. The National Veterinary Service in Portugal and researchers in wild animal health were contacted regarding surveillance of leishmaniosis in wildlife. The ecological impact of control measures was also considered through literature and web searches. The European Environmental Agency was contacted regarding the potential environmental impact of pet carcass disposal Data analysis A descriptive analysis of the environmental impact of leishmaniosis and the control measures for the disease was performed. RESULTS Impact on of the disease on biodiversity Certain species of wild carnivores have been reported to suffer with clinical leishmaniosis (section 3.2.3), although infection is usually subclinical in these species (Quinnell and Courtenay, 2009). However, the reported low prevalence of clinical disease in wild animals may be due to sampling bias towards healthy animals (Quinnell and Courtenay, 2009). Leishmaniosis is not part of the active surveillance plans for wildlife in Spain (Gortázar Schmidt, 2012), France (Moinet, 2012) or Portugal (Afonso, 2012). However, the national wildlife disease surveillance program in Spain includes passive surveillance and researchers at the Instituto de Investigación en Recursos Cinegéticos (IREC) conduct opportunistic research on agents including Leishmania (Gortázar Schmidt, 2012). In France, the SAGIR network analyses data from post mortems of dead wild birds and mammals performed at the local departmental veterinary laboratories. On exceptional occasions, analyses for leishmaniosis have been done on foxes, wolves and badgers (Moinet, 2012) Environmental impact of mortality due to leishmaniosis The European Environmental Agency did not have any information relating to the environmental impact of cremating pet dogs. Based on benchmarking data (section 3.3.3) and LeishVet expert opinion (section 3.6.3), mortality due to leishmaniosis is not likely to exceed 1% of the total animal population (the cut-off for significant mortality in the Phylum manual (Phylum, 2010)) Environmental impact of control measures The application of environmental insecticides as a leishmaniosis control measure could have environmental impacts as a result of pollution and effects on non-vector insect species (Podaliri Vulpiani et al., 2011). Unintentional exposure to high levels of pyrethroids can also have adverse effects on human health (Bradberry et al., 2005, Bouwman and Kylin, 2009). Infants living in areas with malaria control measures were found to have intakes of insecticide which exceeded recommended acceptable levels (Bouwman and Kylin, 2009). No evidence was found that repellents applied to the dog had an environmental impact. Supporting publications 2013:EN

153 CONCLUSIONS In summary, although there have been case reports of clinical disease in wild animals, there is no evidence that L. infantum causes morbidity or mortality levels which may threaten entire populations of wild animals. Some control measures (environmental insecticides) may have environmental or human health considerations. Supporting publications 2013:EN

154 DISCUSSION Leishmaniosis is a vector-borne zoonosis which can cause disease in companion animals, wildlife species and humans. The major route of transmission in both canine and human populations is via bites of phlebotomine sand flies, although transmission can occur by other non-sand fly routes. In addition, sharing contaminated needles was considered to be of importance in the spread of human leishmaniosis in intravenous drug users. There is strong evidence that the dog is the main primary reservoir host of L. infantum, although other domestic and non-domestic species may also be reservoir or incidental hosts. Where data were available, the geographical distribution of clinical human leishmaniosis generally followed that of CanL. In France, the few reported autochthonous human cases were concentrated in the southeast parts of the country in agreement with data collected on the distribution of the canine disease. Several veterinarians based in northern France who responded to the online questionnaire, reported that the few CanL cases they had diagnosed had travelled to endemic regions prior to developing disease. The regions of Sicily, Piemonte and Liguria were found to have consistently higher levels of human and canine clinical leishmaniosis than other parts of Italy; whereas the northeast had lower levels of disease in both species. The frequency of canine and human cases was highest in Mediterranean regions and lowest in the north and west of Spain. The prevalence of CanL appeared to be highest in parts of central Portugal and lowest in the north, consistent with the results of a recent national survey (Cortes et al., 2012), whereas in Greece, CanL appeared to be relatively prevalent and more homogenously spatially distributed. Data from a veterinary benchmarking company in France (Panelvet, 2012), were used to estimate a CanL disease prevalence of approximately 0.05% (the percentage of practice-attending dogs treated for or presumably euthanased due to CanL). This was a much lower estimate than the prevalence estimate observed in a national survey (0.41%) (Bourdeau et al., 2004) and those calculated from the online veterinary questionnaire. The Panelvet estimate was based on invoice data relating to CanL treatments or euthanasia following testing for leishmaniosis. This case-finding method would not have detected CanL cases prescribed treatments other than allopurinol, meglumine antimoniate and miltefosine or therapies sold and dispensed that were not entered onto the clinics practice management systems (computer based accounting systems). In the online questionnaire, 7.5% of vets in France indicated that they used alternative or additional treatments to the main treatments specified in the questionnaire. Moreover, dogs receiving treatment purchased from human or online pharmacies would not have been included as cases. Hence, these estimates were likely to be a conservative estimate at best. Based on veterinary estimates from the online veterinary questionnaire, the estimated national prevalence of clinical CanL (percentage of practice-attending dogs with a confirmed veterinary diagnosis of CanL) by country was as follows: France 0.71%, Spain 3.71%, Portugal 2.92%, Italy 4.25% and Greece 7.80%. This represented a median of 1 CanL case seen by each vet in France per year, 8 cases seen per vet in Spain, Portugal and Italy and 25 cases seen per vet in Greece per year. These estimates were based on veterinary confirmed cases of leishmaniosis seen at clinics. Due to different clinical approaches, specific criteria for the case definition (e.g. specific clinical signs, serology results exceeding a certain titre) were not specified in the questionnaire. The local prevalence estimates would have depended on factors including the clinical acumen of the vets, the proportion of suspect cases which underwent investigations and the characteristics of the diagnostic techniques adopted. In order to calculate the number of dogs seen per year, the estimated canine caseload was divided by three (assuming that a dog attends a clinic three times per year on average). The latter figure was based on data from a Spanish company that holds economic data (Veterinary Management Studies) and a French benchmarking company (Panelvet) (Mercader, 2013, Panelvet, 2012). It was assumed that the mean number of times a dog attended a clinic per year would be similar for all the Supporting publications 2013:EN

155 studied countries. If the mean number of annual visits per dog was under or overestimated for a country, this would have resulted in an under or overestimation of prevalence. Additionally, given some evidence of regional variation in response rate within certain countries, e.g. in France where more responses were seen in endemic areas, the regional prevalences were reported and the mean and median regional prevalence were suggested to be more representative estimates of the national disease load for each country. Finally, the generalisability of these estimates to the general dog population will depend on the proportion of all dogs which attended veterinary clinics in each country, though it is likely that owned dogs with clinical disease of sufficient severity were likely to present to a veterinarian for treatment, and as such these estimates presented are likely to be reflective of clinically relevant disease. Additional estimates reported from shelter populations and laboratory diagnoses support a similar prevalence of disease in un-owned or relinquished dogs and a high seroprevalence among suspect CanL cases respectively. The primary course of the CaniLeish vaccine was the most expensive control measure, with an average cost of approximately for initiation of control (and thereafter approximately euros per year). Deltamethrin impregnated Scalibor collars and Advantix spot-on treatments appeared to be the most common prophylaxes and cost approximately 25 and 30 respectively whilst domperidone (Leishguard ) cost approximately 30 (for a 30kg dog). Average costs of diagnostic tests were estimated at 24 per test. Average treatment for a 30kg dog was approximately 200 if purchased from a veterinary clinic. These figures equate to dog owners in the studied countries spending an estimated 54 million on Scalibor collars and 118 million on CaniLeish vaccination in 2012 (although based on industry data the latter is likely to be an overestimate). Assuming 90% of diagnosed cases were treated, at a conservative estimate of 100 per case, the cost of treatment for CanL in the studied countries could amount to approximately 44.8 million. Mortality as a result of leishmaniosis can result in monetary losses due to the cost of euthanasia and replacing a deceased pet, in addition to non-monetary losses related to compromise to the human-animal bond. The indirect economic impacts on trade and tourism were assessed as being minor. The current economic crisis in the EU was considered likely to have a moderate to high impact on the use of prophylactic control measures and the diagnosis and treatment of dogs affected by leishmaniosis and may result in an increase in the prevalence in owned dogs across the studied countries. The incidence of clinical human leishmaniosis cases in the studied countries was much lower than the incidence of clinical CanL. The incidence of clinical human leishmaniosis was highest in Greece and lowest in France and Portugal; with Italy and Spain having intermediate risk; a similar pattern to the national estimates of CanL incidence. Certain groups, such as infants and immunosuppressed people (especially HIV positive people), were at a higher risk of leishmaniosis and VL was reported to be fatal in 2.7-5% of cases. The economic impact due to human disease is not insignificant; costs relating to hospital fees and time off work were estimated to amount to 4.5million/year in Spain, Portugal, Italy and Greece. Based on expert opinion, it was concluded that the impact on dog welfare and the likely prognosis of CanL depended on the clinical stage of the disease, although in very severe cases the disease can have an extreme impact on welfare. Experts thought that the adverse effects associated with repellents and insecticides applied to the dog and immune stimulants had the lowest impact on dog welfare and vector control was thought to be most publicly acceptable control measure. These conclusions were based on the responses to a questionnaire completed by six LeishVet members. The certainties of these conclusions are low due to the small number of responses and the limitations respondents may have when answering some questions. Some questions were open to interpretation, such as the duration of the welfare impact of culling, whereas others may have been inherently difficult to answer due to the heterogeneity of impact at the level of the individual dog. Supporting publications 2013:EN

156 Based on the level of web interest, temporal and regional public interest in leishmaniosis largely reflected the occurrence of clinical human and canine cases in France, Spain and Italy. There was insufficient evidence that L. infantum has the potential to be used as a bioterrorism agent. The environmental impact of L. infantum and the control measures for CanL was likely to be minor. However, clinical disease in wildlife species has been reported and the potential environmental impact of insecticide applied to the environment should be considered. Supporting publications 2013:EN

157 CONCLUSIONS AND RECOMMENDATIONS Canine leishmaniosis (CanL) is an important zoonosis affecting animals and people worldwide. Though the quality of published evidence is only moderate, systematic appraisal of available data and reports suggests there are a number of efficacious preventative control measures available to prevent infection and transmission of leishmaniosis in dogs, including repellents, vaccination and prophylactic medication. Despite the availability of these preventative methods, the modelling work performed for this report suggests that the potential for spread via dogs to non-endemic European countries where a competent sand fly vector exists is high. Mitigation measures to prevent the introduction of canine leishmaniosis into a non-endemic region require implementation in a high proportion of dogs within the population at risk. Based on this simulation modelling, the methods that appear most effective to prevent introduction of disease in a non-endemic area are repellent, followed in decreasing order of effectiveness by vaccination, prophylactic medication and the use of insecticide. This study has identified a significant burden of disease in dogs within a number of Mediterranean European member states. Leishmaniosis results in human hospital admissions within affected countries; it has significant financial consequences for owners of dogs and for human health services in endemic regions and it represents an increasingly important potential zoonotic threat. Though effective mitigation measures exist, their success requires implementation in a high proportion of the primary canine host population to be effective, hence a substantial commitment by member states is required to prevent the introduction of the disease into non-endemic regions where sand flies are present. The systematic evaluation of scientific evidence for efficacy of currently available preventative control interventions for natural L. infantum infection in dogs identified a number of appropriate prophylactic modalities. The systematic review (SR) identified 23 eligible studies, providing data on 5861 dogs and featured 12 studies on vaccinations, 5 on repellent collars, 4 on spot-on insecticides and 3 on prophylactic medication. Substantial variations within the studies regarding baseline characteristics of the dogs involved, such as age, breed and gender, were observed. All of the studies were considered to be at a high risk of bias, with the exception of two vaccination studies and one spot-on study which were considered to be at an unclear risk of bias. Nonetheless, there are studies to support the use of control measures, in particular vaccines (200µg ALM protein, Leishmune, CaniLeish, LiESAp with MDP, and ALM with BCG), deltamethrin collars, 65% permethrin, 10% imidacloprid with 50% permethrin spot-ons and domperidone prophylactic medication as they all tended to significantly reduce the proportion of dogs infected with L. infantum (relative to a control group) based on either parasitological or serological evidence. In the context of evidence for prevention of infection, the magnitudes of effects reported were broadly comparable across interventions, with vaccinations tending to have the greater protective effects. There was a significant protective effect for the overall proportion of dogs infected with L. infantum based on serology or parasite detection, in six of the vaccination studies (OR range ), for ALM (Mohebali et al., 2004), BCG (Mohebali et al., 1999) b, CaniLeish (Oliva et al., 2012), LiESAp with MDP (Lemesre et al., 2007), and Leishmune (Lima et al., 2010, Nogueira et al., 2005b). The attributable risk reductions ranged from 0.06 to 0.54 suggesting between 6 and 54% of cases could be prevented with vaccination. For deltamethrin collars, there was a protective effect for the overall proportion of dogs infected with L. infantum for four of the five studies (OR range , Ferroglio et al., 2008; Foglia Manzillo et al., 2006; Maroli et al., 2001; Gavgani et al., 2002). The attributable risk reductions ranged from 0.04 to For spot-ons four studies reporting reduced odds of disease when treated with ORs reported of between 0.01 and 0.14 and the attributable risk reductions ranged from 0.10 to 0.37 suggesting between 10 and 37% of cases could be prevented with spot-on use (Ferroglio et al., 2008, Otranto et al. 2010, Otranto et al., 2007a and b). There was a Supporting publications 2013:EN

158 statistically significant protective effect for one study based on domperidone used as prophylactic medication (Llinas et al., 2011). Llinas et al. (2011) and Gomez-Ochoa et al. (2012) found that for every 100 dogs prophylactically medicated with domperidone 37 and 6 cases of infection with L. infantum would be averted, respectively (though the latter was not statistically significant). There are studies to support the use of the studied control measures to prevent L. infantum infection in both endemic and non-endemic areas. However, the risk of bias with the publications on these interventions needs to be considered. This systematic review highlighted that well-designed, adequately powered RCTs are needed to determine whether using control measures for CanL confers prophylactic benefits. There is ongoing work to determine the efficacy of prophylactic medications and these results will provide further, valuable evidence. The simulation modelling component of this project estimated the probability of CanL disease persistence (called endemicity throughout the report) following the introduction of infected dogs in previously CanL-free areas with competent vectors. The stochastic model developed for the project estimated a high probability of CanL endemicity in at least one dog population within CanL-free EU regions with competent sand flies where infected dogs are introduced (92.2% to 100% when 10 to 100 dogs were introduced) and no mitigation measures are implemented. The impact of the following mitigation measures was assessed: vaccination, prophylactic medication, repellent, insecticide, and diagnostic test and exclusion. Although the systematic review did not identify studies evaluating combinations of mitigation measures, their combined effect was evaluated in the modelling section. When mitigation measures were used in combination, their effectiveness in reducing the risk of CanL infection or transmission was higher than when used individually; however this effect was only observed when mitigation measures were used in a high proportion of the population (e.g. 80%) of dogs. Testing and exclusion of positive dogs had the highest impact on the overall probability of endemicity when relatively small numbers of dogs were imported from, or returned from trips to endemic areas. However, the benefits of testing and exclusion diminished as the numbers of imported dogs increased, due to the relatively low sensitivity of the diagnostic test used 52.6% [95%CrI: 30.8%- 74.0%, (Mettler et al., 2005a)]. The use of repellent was the most effective mitigation measure on the probability of endemicity following the introduction of infected dogs in a CanL-free area, and on the probability of a dog returning infected after travelling to an endemic area, followed in decreasing order of effectiveness by: vaccination, prophylactic medication and use of insecticide. Some combinations of mitigation measures such as repellent and vaccination or prophylactic medication showed a joint effect higher than the effect of individual measures, even when applied to a lower proportion of dogs in the population. Overall, the results of the impact assessment agree with the modelling results, as most of the areas with presence of competent vectors also had reported cases of CanL. However, in a few European NUTS3 areas with presence of sand flies no cases of CanL were reported by the veterinarians responding to the survey in the impact assessment, whereas the modelling results suggest that the probability of disease endemicity following introduction of infected dogs into a non-endemic area with a competent vector is high. However, the survey responses may not have included enough veterinary practices be representative of the practices in the area and thus, may not be directly comparable to the modelling results. Furthermore, even if the probability of at least one dog to be infected in the department is high, the prevalence of CanL infection and thus the number of clinical cases may be low enough to stay undetected by the responding veterinary practitioners. Moreover, if the infection was recently introduced into a department, it may also be possible that veterinary clinics had not yet diagnosed cases of CanL, given the relatively slow progression of the disease and the potential for under or misdiagnosis of a disease that is not commonly seen by local practitioners. Finally, VBORNET maps reported only presence or absence of competent vectors, not abundance so it may Supporting publications 2013:EN

159 also be possible that competent vectors were present, but in a low enough density to prevent further CanL spread in dogs. Based on this simulation modelling work, a number of recommendations were formulated. Mitigation measures on dogs in CanL-free areas with competent vectors were only effective when used in a large proportion of the dogs, and would thus require a high level of compliance and may be costly to implement. Thus, efforts may be better directed at either dogs travelling to or from endemic areas, or commercial imports of dogs from endemic areas. Dog owners should be advised to use control measures on their dogs during travels to CanL endemic areas to prevent L. infantum infection: vaccination or prophylactic medication ideally, complemented by the use of repellent during the full duration of the travel. Although not explicitly evaluated via modelling, the findings of this report suggest that dog owners from CanL endemic areas travelling with their dogs to non-endemic areas with competent vector(s) should also be advised to use repellent to reduce the chances that their dogs may spread the infection to the local sand fly population. Ideally, dogs known to be infected should not visit non-endemic areas, but this might be an unrealistic requirement given the open nature of travels within the European Union. A test and exclusion policy could be a feasible alternative for commercial importations of dogs from endemic areas. Dogs should be tested with the best available diagnostic test before the date of importation, and positive animals should not be allowed to enter CanL-free countries. The effectiveness of this strategy is tied to the sensitivity of the diagnostic test used. Therefore, when large numbers of dogs are imported and a test with a relatively low sensitivity is used, the benefits of this strategy are diminished. Because of this, additional mitigation measures aimed at reducing the risk of infection in dogs from endemic areas (repellent, vaccination, prophylactic medication) may also be requested for dogs to be imported into non-endemic areas. The test and exclusion policy may be difficult to implement and enforce for non-commercial movements of dogs across the EU (e.g. for travel between endemic and CanL-free areas within the same countries or between countries), hence the suggestion to restrict this strategy to commercial movements of dogs. These conclusions and recommendations are derived from a modelling abstraction of a very complex disease process. This abstraction required a series of simplified assumptions and the quantification of many important parameters, often from disparate sources of information. Therefore, the conclusions and recommendations should be considered in the context of the modelling work described here, and notably in light of the limitations and data gaps described in the report. The impact assessment evaluated the impact of CanL in those areas where the disease is endemic, in particular in France, Spain, Portugal, Italy and Greece. Based on expert opinion and published literature, the dog was widely accepted to be the main primary reservoir host of L. infantum (Baneth et al., 2008, Quinnell and Courtenay, 2009, Alvar et al., 2004). The major route of transmission of the parasite in both canine and human populations was considered to be via bites of phlebotomine sand flies. Although many infections remain subclinical, leishmaniosis has been reported to cause disease in companion animals, wildlife species and humans (Poli et al., 2002, Marcos et al., 2009, Navarro et al., 2010, Ozon et al., 1998, Hervas et al., 1999, Solano-Gallego et al., 2003, Rolao et al., 2005, Koehler et al., 2002, Fallah and Khanmohammadi, 2011, Tenorio Mda et al., 2011, Beck et al., 2008, Luppi et al., 2008, Dahroug et al., 2011, Libert et al., 2012, Ready, 2010, Quinnell and Courtenay, 2009, Solano- Gallego et al., 2011). Routinely collected companion animal epidemiological data were not available and based on results of an online veterinary questionnaire, clinical cases of CanL were estimated to be common in endemic parts of the EU with greatest estimates in Italy and Greece, and least disease reported in France of the five studied countries. The economic burden of CanL was mostly at the expense of dog owners and included the costs associated with prevention, treatment and mortality. The CaniLeish vaccine was the most expensive Supporting publications 2013:EN

160 control measure, with an average cost of 150 for the primary course and each subsequent year for maintenance of immunity. Other preventative measures such as repellent collars and spot-on, typically cost Treatment costs were variable, but standard treatment for a 30kg dog with clinical leishmaniosis cost approximately 200 if purchased from a veterinary clinic. Based on mean cost figures and data from pharmaceutical companies and the online veterinary questionnaire, it was estimated that dog owners in the five countries spent between 50 million and million on CanL vaccination, approximately 54 million on deltamethrin collars, 35.9 million on testing and 44.8 million on treating CanL in The indirect economic impacts on trade and tourism were likely to be minor. LeishVet experts and veterinary clinicians responding to the online questionnaire generally thought that the current economic crisis in the EU was likely to have a moderate to high impact on the use of prophylactic control measures and the diagnosis and treatment of dogs affected by CanL. This impact may consequently increase the prevalence of CanL in the future. Where data were available, the geographical distribution of clinical human leishmaniosis cases broadly followed that of clinical CanL. Regional incidence of the disease in both humans and dogs was highest in southeast France and Spain and the Italian regions of Sicily, Piemonte and Liguria. Although overall the incidence of clinical human leishmaniosis was low, approximately people required hospitalisation each year in each country and certain groups, such as infants and immunosuppressed people, were at a higher risk of disease. The evaluation of the adverse impact of CanL and its control measures on dog welfare was largely based on the opinions of six experts. The perceived welfare impact depended on the clinical stage of the disease. Although it was not possible to reliably estimate the frequency at which different stages of CanL occurred in affected dogs, many clinical cases are likely to experience welfare compromise. Generally, adverse effects associated with repellents and insecticides applied to the dog and immune stimulants were considered to have the lowest impact on dog welfare. Experts thought that vector control was likely to be the most publicly acceptable control measure. Based on the level of web interest, temporal and regional public interest in leishmaniosis largely reflected the occurrence of clinical human and canine cases in France, Spain and Italy. There was little evidence that L. infantum and the control measures for CanL had major environmental consequences, although the potential impact of environmental insecticide use should be considered. In summary, though routinely collected and robust epidemiological data were rarely immediately available, particularly for the evaluation of canine disease, the data estimates reported represent at best a good approximation. Despite this uncertainty in the exact frequency and cost of the disease, canine leishmaniosis represents a disease with significant animal and human health impacts. It affects the welfare of dogs that are clinically infected with the condition and has economic implications for owners of affected dogs and for human health care. On systematic evaluation of available evidence, canine leishmaniosis appears a preventable disease with a number of effective preventative modalities available. The spread to non-endemic regions, based on simulation modelling, appears likely and this should be a major concern to EU member states. Mitigation strategies to prevent this spread are available and appear effective, though they require a concerted effort by non-endemic countries if they are to be successful in averting the imposition of endemicity. The disease appears relatively prevalent in dogs in Mediterranean countries and there are a number of impacts for infected dogs and for their owners. However, data sources to allow ongoing surveillance of the current disease burden in companion animals are very limited and there is a major need for the routine collection of epidemiological data from companion animals in member states to be able to measure and rapidly respond to changes in disease frequency. Finally, the zoonotic impact is significant, though most reported cases relate to high risk groups including young children and immune-suppressed individuals. Supporting publications 2013:EN

161 Acknowledgements The authors are grateful to Dr. Gioia Capelli, Prof. Maria Grazia Pennisi, Prof. Gaetano Oliva, Prof. Gad Baneth, Prof. Ozbel, Prof. Domenico Otranto and Dr. Filipe Dantas-Torres for providing references used in this study. The authors would like to thank Dr. Huybert Groenendaal for his valuable input to the modelling work presented here. The authors would also like to thank Prof. Maria Grazia Pennisi, Prof. Patrick Bourdeau, Prof. Guadalupe Miró Corrales, Prof. Luís Cardoso, Prof Alexander Koutinas, Prof. Gaetano Oliva and Prof. Gad Baneth for their expert advice and opinion. The authors are grateful to Nancy De Briyne for sending the online veterinary questionnaire to FVE member associations and the following people who translated the online questionnaires for veterinary clinicians and dog shelters: Aris Polyviou, Maria Ferrara, Carmela Caponi, Roula Tsourouti, Alex Mattin, Ana Pascual and João Sucena Afonso. The authors would also like to thank everyone who contributed data to the impact assessment, including Eleni Dotsika, Fabrizio Vitale, Robert Armstrong, David Lussot and Dominique Grandjean; and the questionnaire respondents working in dog shelters and veterinary clinics. Finally, the authors would like to thank EFSA scientific staff who provided input and guidance during this project, including Dr. Sofie Dhollander, Dr. José Cortinas Abrahantes, and Ms Elisa Aiassa, and also EFSA AHAW panel members Dr. Ana Afonso, Dr. Aline de Koeijer and Dr. Edith Authié. Supporting publications 2013:EN

162 APPENDIX/APPENDICES A. RESULTS OF SEARCH STRATEGIES Search 1. U.S. National Library of Medicine 2011 (MEDLINE) Add to builder Query Items found #69 Add Search ((#68) AND #39) AND # #68 Add Search ((((((((((((((((((#40) OR #41) OR #42) OR #43) OR #44) OR #45) OR #46) OR #47) OR #48) OR #49) OR #50) OR #51) OR #52) OR #53) OR #54) OR #55) OR #56) OR #66) OR #67 #67 Add Search vaccination[mesh Terms] #66 Add Search insect repellents[mesh Terms] 1664 #56 Add Search insecticides[mesh Terms] #55 Add Search immune stimulant*[title/abstract] 178 #54 Add Search collars[title/abstract] 838 #53 Add Search collar[title/abstract] 5506 #52 Add Search spot-ons[title/abstract] 3 #51 Add Search spot-on[title/abstract] 955 #50 Add Search vaccination[title/abstract] #49 Add Search vaccines[title/abstract] #48 Add Search vaccine[title/abstract] #47 Add Search insecticides[title/abstract] #46 Add Search insecticide[title/abstract] #45 Add Search prevention[title/abstract] #44 Add Search prophylaxis[title/abstract] #43 Add Search prophylactic[title/abstract] #42 Add Search preventative[title/abstract] 6364 #41 Add Search preventive[title/abstract] #40 Add Search control[title/abstract] #39 Add Search (((#35) OR #36) OR #37) OR # #38 Add Search leishmaniasis[mesh Terms] #37 Add Search leishmaniasis[title/abstract] #36 Add Search leishmaniosis[title/abstract] 371 #35 Add Search leishmania[title/abstract] #34 Add Search ((((((((#25) OR #26) OR #27) OR #28) OR #29) OR #30) OR #31) OR #32) OR # #33 Add Search canidae[mesh Terms] #32 Add Search dogs[mesh Terms] #31 Add Search canidae[title/abstract] 211 #30 Add Search canine[title/abstract] #29 Add Search canids[title/abstract] 446 #28 Add Search canid[title/abstract] 250 #27 Add Search canis[title/abstract] 6196 #26 Add Search dogs[title/abstract] #25 Add Search dog[title/abstract] Supporting publications 2013:EN

163 2. Web of Science 2011 (WOS) # #3 AND #2 AND #1 Databases=SCI-EXPANDED, SSCI, A&HCI, CPCI-S, CPCI-SSH Timespan=All Years Lemmatization=On # 4,042,497 TS=(control OR preventive OR preventative OR prophylactic OR prophylaxis OR 3 prevention OR insecticide OR insecticides OR vaccine OR vaccines OR vaccination OR spot-on OR spot-ons OR collar OR collars OR immune stimulant OR immune stimulants) Databases=SCI-EXPANDED, SSCI, A&HCI, CPCI-S, CPCI-SSH Timespan=All Years Lemmatization=On # 2 27,496 TS=(leishmania OR leishmaniosis OR leishmaniasis) Databases=SCI-EXPANDED, SSCI, A&HCI, CPCI-S, CPCI-SSH Timespan=All Years Lemmatization=On # 1 272, 414 TS=(dog OR dogs OR canis OR canid OR canids OR canidae OR canine) Databases=SCI-EXPANDED, SSCI, A&HCI, CPCI-S, CPCI-SSH Timespan=All Years Lemmatization=On 3. CAB Direct 2011 ab:(dog OR dogs OR canis OR canid OR canids OR canidae OR canine) OR title:(dog OR dogs OR canis OR canid OR canids OR canidae OR canine) OR subject:(dog OR canidae) (n=179,216) AND title:(leishmania OR leishmaniosis OR leishmaniasis) OR ab:(leishmania OR leishmaniosis OR leishmaniasis) OR subject:(leishmania OR leishmaniosis) (n=25,345) AND title:(control OR preventive OR preventative OR prophylactic OR prophylaxis OR prevention OR insecticide OR insecticides OR vaccine OR vaccines OR vaccination OR spot-on OR spot-ons OR collar OR collars OR immune stimulant OR immune stimulants) OR ab:(control OR preventive OR preventative OR prophylactic OR prophylaxis OR prevention OR insecticide OR insecticides OR vaccine OR vaccines OR vaccination OR spot-on OR spot-ons OR collar OR collars OR immune stimulant OR immune stimulants) OR subject:(insecticides OR "insect repellents" OR vaccination (n=161,546) Total = 177 Supporting publications 2013:EN

164 4. Literatura Latino Americana e do Caribe em Ciências da Saúde (LILACS) Abstract and title: (dog OR dogs OR canis OR canid OR canids OR canidae OR canine) (n=4321) AND (leishmania OR leishmaniosis OR leishmaniasis) (n=3088) AND (control OR preventive OR preventative OR prophylactic OR prophylaxis OR prevention OR insecticide OR insecticides OR vaccine OR vaccines OR vaccination OR spot-on OR spot-ons OR collar OR collars OR immune stimulant OR immune stimulants ) (n=54,983) Total =114 Supporting publications 2013:EN

165 Reference Original scientific article Part of a study written up in full elsewhere Canine population Leishmania infection Leishmania infantum/chagasi infection, Naturally-occurring infection A randomised or non-randomised controlled clinical trial, or observational analytic epidemiologic study Definitive diagnosis established using a specific diagnostic method Preventive control intervention applied to the canine population Results presented to provide evidence of the efficacy Short description of study Impact, modelling and control of canine leishmaniosis in the EU B. REASONS FOR EXCLUSION AND SHORT DESCRIPTION OF THE 64 STUDIES EXCLUDED AT THE SECOND SCREENING PHASE OF THE SR (Albarracin et al., 2011) Yes No Yes Yes Yes Yes No Yes No Unclear Serum samples of 74 canines were tested. Of all samples tested, only 13 resulted positives, with a prevalence of 17.56%. The variables type of housing, breed, age, sex, and origin, did not show any association with the disease transmission. Supporting publications 2013:EN by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender considered as an output adopted by the Authority. The European food Safety Authority reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

166 (Almeida et al., 2012) Yes No Yes Yes Yes Yes No Yes Yes Unclear A transversal study was carried out in endemic areas from Cuiabá, State of Mato Grosso, to assess data on seroprevalence and risk factors associated to canine infection. Four hundred and thirty (430) dogs were randomly evaluated through indirect fluorescence antibody test (IFAT) considering variables related to the animals, the environment and the knowledge by owners on CVL aspects and control. (Andrade al., 2008) et Yes No Yes Yes Yes Yes No Yes No Unclear Descriptive study of the canine population of urban area Aracatuba, Sao Paulo, Brazil, from (Aoun et al., 2009) Yes No Yes Yes Yes Yes No Yes Yes Unclear To evaluate the real prevalence of asymptomatic carriage in dogs by means of real time quantitative PCR (qpcr) and serology. Included prospectively 140 military dogs wearing deltamethrineimpregnated collars. (Badaro et al., 2002) Yes No Yes Yes Unclear Yes Yes No Yes No 7-year trial was conducted in Jacobina (Brazil); all dogs were fitted with deltamethrin-impregnated collars, which were changed every 4 months during the year. Supporting publications 2013:EN by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender considered as an output adopted by the Authority. The European food Safety Authority reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

167 (Borja- Cabrera et al., 2005) Yes Yes Yes Yes Yes Yes Yes Yes Yes Unclear A group of 550 FML-seronegative and asymptomatic dogs from Brazilian canine visceral leishmaniasis endemic areas was vaccinated with 3 sc doses of Leishmune and one annual booster. (Cabrera et al., 2003) Yes No Yes Yes Yes Yes Yes Yes No Unclear In order to evaluate factors related to the increase of the risk for Leishmania (Leishmania) chagasi infection in dogs we have screened 365 dogs by anti-leishmania immunofluorescent antibody test (IFAT) and captured sand flies in the domestic and peridomestic environment. (Camargo- Neves et al., 2005) Yes Yes Yes Yes Unclear Yes No Yes Yes Unclear The study was conducted in Andradina. A canine cohort study was established between October 2002 and April Deltamethrin impregnated collars 4% Scalibor were employed in all seronegative dogs living in the city, and the seropositive were euthanized. Supporting publications 2013:EN by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender considered as an output adopted by the Authority. The European food Safety Authority reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

168 (Camargo- Neves et al., 2009) Yes No Yes Yes Unclear Yes No Yes Yes Unclear The study was conducted from in Andradina (Brazil). Collars were employed in all seronegative dogs and the seropositive dogs were euthanized. (Camargo- Neves et al., 2010) Yes Yes Yes Yes Unclear Yes No Yes Yes Unclear The cohort study was done from 2002 to 2005 in Andradina small town of the state of SP, Brazil, which has an estimated population of 55,161 inhabitants and 15,600 dogs. Serologic evaluations with the IFAT were done at six monthly intervals. ScaliborR collars were fitted on all seronegative dogs and all seropositive dogs were euthanized. The collars were replaced in Apr and Oct/2003 and Apr/2004. (Collin et al., 2007) Yes No Yes No No No No No No Unclear Focused on the role of Lutzomyia longipalpis salivary proteins as potential vaccine candidates in dogs against L. infantum infection. Developed a model reproducing natural exposure of dogs to sand fly saliva. Supporting publications 2013:EN by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender considered as an output adopted by the Authority. The European food Safety Authority reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

169 (Cortes et al., 2012) Yes No Yes Yes Yes Yes No Yes Yes Unclear This study presents the results of a canine epidemiological survey in a mediterranean region where human and canine leishmaniasis (CanL) are endemic - Portugal. The main goal was to identify risk factors, which can be relevant for Leishmania infection control. (Courtenay et al., 2009) Yes No Yes Yes Yes Yes Yes Unclear Yes Unclear Nine pairs of villages matched by the pre-intervention human aged-adjusted prevalence of positive DTH reactions to Montenegro skin test were randomised within pairs to receive community-wide insecticide or placebo pour-on treatment of dogs. (da Silva et al., 2012) Yes No Yes Yes Yes Yes Yes Yes No Unclear The objective of this study was to assess the association between peridomestic socioeconomic and environmental factors and the presence of dogs seropositive for Leishmania chagasi in the City of Teresina, Brazil. Methods: This case-control study was based on the results of a routine seroepidemiological survey among domestic dogs carried out in Supporting publications 2013:EN by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender considered as an output adopted by the Authority. The European food Safety Authority reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

170 (Dalla Villa et al., 2008) (Dantas- Torres, 2006b) Yes Yes Yes Yes Yes Yes No Yes No Unclear The Istituto Zooprofilattico Sperimentale dell'abruzzo e del Molise 'G. Caporale' (IZS A&M) manages two kennels in Teramo and uses standard management procedures that include regular behavioural and physical examinations. All results recorded in a computer database were analysed to identify production goals and improve welfare, in line with a population medicine model. Prevalence and incidence of different pathologies were recorded and analysed to improve veterinary management and organisation. No No Yes Yes Yes Yes No Yes Yes Unclear Review of the potential of Leishmune vaccine for the prevention and control of canine visceral leishmaniosis and its potential as a transmission-blocking vaccine. (Davies et al., 2002) Yes Yes Yes Yes Yes Yes Yes Yes Yes Unclear Study of the impact of insecticide impregnated dog collars on the incidence of zoonotic visceral leishmaniasis in dogs and children. (Diouani et al., 2008) Yes No Yes Yes Yes Yes No Yes No Unclear A follow-up study of 917 dogs was undertaken between 1994 and 1995 in the focus of visceral leishmaniasis in northern Tunisia. Supporting publications 2013:EN by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender considered as an output adopted by the Authority. The European food Safety Authority reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

171 (Dunan et al., 1989b) Yes Yes Yes Yes Yes Yes Unclear Yes Yes Unclear The assay, conducted in an endemic zone of canine leishmaniasis (Provence - Marseilles' region) interested a population of 557 dogs. 392 of them serologically and clinically indemn, were surveyed for 2 years (surveyed, serological and parasitological controls). (Fallah et al., 1998) Yes Yes Yes Yes Yes No Yes Unclear Yes Unclear In this study, 16 eligible dogs were randomly divided to 4 groups. Group 1: were received Autoclaved leishmania infantum vaccine with BCG. Group 2: were received Autoclaved leishmania major vaccine with BCG. Group 3: were received BCG alone and Group 4: were received normal saline. (Fallah et al., 2000) Yes Yes Yes Yes Yes Yes Yes Yes Yes Unclear Preparation and evaluation of Leishmania vaccines in Iran for the control of kala-azar. 152 domestic dogs from one of the village of Menshkin-shahr district were tested by with 3 serological methods. Supporting publications 2013:EN by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender considered as an output adopted by the Authority. The European food Safety Authority reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

172 (Freyburger et al., 2012) No No Yes Yes Yes Yes No No Yes Unclear This article describes the development of vaccines to control vector-borne leishmaniasis and babesiosis in dogs in France. (Gambino al., 1997) et Yes No Yes Yes Yes Yes No Yes No Unclear Study of the clinical manifestations of 178 pet dogs, 43 asymptomatics, 49 with symptoms of active leishmaniasis and 86 oligosymptomatics in Sicily. (Garcez et al., 2008) Yes No Yes Yes Yes Yes Yes Unclear Yes No The efficacy of topical deltamethrin for protecting dogs against Lutzomyia longipalpis and risk factors for human kala-azar in a mineral prospection area were respectively investigated in Salvaterra and Juruti municipalities, Pará State. (Goto 2011) et al., Yes No No Yes Yes No No No Yes Unclear Description of a multiantigen vaccine candidate comprised of antigens known to be protective in animal models, including dogs, and to be recognized by humans immune to visceral leishmaniasis. Supporting publications 2013:EN by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender considered as an output adopted by the Authority. The European food Safety Authority reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

173 (Gradoni et al., 2012) Yes No Yes Yes Unclear Unclear Yes Unclear Yes Unclear Results of a pilot xenodiagnosis study of vaccination with LiESP/QA-21 (CaniLeish ). CaniLeish reduced the intensity of infection in Phlebotomus perniciosus fed on Leishmania infantum infected dogs. (Gradoni, 2006) No No Yes Yes Yes Yes No Yes Yes Unclear Review of vaccines against leishmaniasis. (Ikeda-Garcia et al., 2009) Yes No Yes Yes Yes Yes No Unclear Yes Unclear Evaluation of the four IgG subclasses in symptomatic dogs with Leishmania sp. and in dogs vaccinated against visceral leishmaniasis. Three groups of dogs from both sexes, several breeds and ages were used. The first group was composed of 45 clinically healthy animals, coming from non endemic areas for visceral leishmaniasis (control), the second was composed of 45 symptomatic dogs with visceral leishmaniasis and the third was composed of 37 clinically healthy dogs which had been vaccinated against canine visceral leishmaniasis. Brazil Supporting publications 2013:EN by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender considered as an output adopted by the Authority. The European food Safety Authority reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

174 (Juliao et al., 2007) Yes No Yes Yes Yes Yes No Yes No Unclear The efficacy of topical deltamethrin for protecting dogs against Lutzomyia longipalpis and risk factors for human kala-azar in a mineral prospection area were respectively investigated in Salvaterra and Juruti municipalities, Pará State. (Maroli et al., 1999) Yes Yes Yes Yes Yes Yes Yes Yes Yes Unclear A report is given on the rationale and study design of a field trial to evaluate the efficacy on the incidence and prevalence of canine leishmaniasis (Leishmania infection; CanL) of deltamethrinimpregnated collars (Scalibor ProtectorBands, Hoechst Roussel Vet) in a natural dog population. (Maroli et al., 2002) Yes Yes Yes Yes Yes Yes Yes Yes Yes Unclear Two field studies were carried out to evaluate the efficacy of Scalibor protector bands on the transmission of the canine leishmaniasis in a highly endemic focus of southern Italy with two cohorts of dogs. Supporting publications 2013:EN by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender considered as an output adopted by the Authority. The European food Safety Authority reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

175 (Menz et al., 2009a) Yes Yes Yes Yes Yes Yes Unclear Unclear Yes Unclear Study of the incidence of human and canine visceral leishmaniasis in endemic areas of Brazil following vaccination of dogs with Leishmune. (Menz et al., 2009b) Yes Yes Yes Yes Yes Yes Unclear Unclear Yes Unclear Study of the decrease of the incidence in human and canine visceral leishmaniasis after dog vaccination with Leishmune in Brazilian endemic areas. (Mohebali al., 1998) et Yes No Yes Yes Yes No Yes Yes Yes Unclear Sixteen dogs were randomly divided into four groups. 1) Autoclaved Leishmania infantum vaccine with BCG, 2) autoclaved l. Major vaccine with BCG, 3) BCG, 4) normal saline. (Mohebali, 2003) Yes Yes Yes Yes Yes Yes Yes Yes Yes Unclear In this study, 406 dogs with owners and with no anti-leishmania antibodies from Parikhan village in Meshkin-Shahr district, Iran were physically examined. 315 healthy seronegative dogs with no response to leishmanin were selected and randomly injected by either alum-precipitated autoclaved Leishmania major (ALM) vaccine (200 µg) of Leishmania protein precipitated in roughly 620 pg of aluminium hydroxide mixed with BCG (roughly each dog received 2 million CFUs) or injected with PBS as a control. Supporting publications 2013:EN by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender considered as an output adopted by the Authority. The European food Safety Authority reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

176 (Morais et al., 2008) Blood samples were taken from the dogs 4 months post injection and were skin tested with leishmanin (LST). The injected dogs were followed-up for 16 months. Results: The examination of the site of injection showed that the vaccine is well tolerated and no severe reaction was observed. The LST conversion rate was significantly higher in the vaccinated group compared to the placebo group. The efficacy evaluation performed at 16-month post vaccination indicated that administration of this vaccine plus BCG had 54.6% protective fraction effects against canine visceral leishmaniasis. Yes No Yes Yes Yes Yes Yes Yes No Unclear The aim of this work is to evaluate the success of actions in the control of visceral leishmaniasis in the North West Health District of Belo Horizonte. (Moreira et al., 2003) Yes No Yes Yes Yes Yes No Yes No Unclear Study of a cohort of dogs in an urban area in Brazil to determine whether incidence varied with age, breed, and environmental characteristics. The mean follow-up was 1.5 years, and the crude incidence rate was 11.8 cases/100 dog-years (95% confidence interval [CI] = ). (Netto et al., 2005) Yes No Yes Yes Yes Yes Yes Unclear Yes No To test the efficacy of the trifusion antigens as a preventive vaccine or immunotherapy against L.chagasi we performed two studies: a double blind randomized controlled by placebo and an open randomized 4 arms. Supporting publications 2013:EN by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender considered as an output adopted by the Authority. The European food Safety Authority reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

177 (Nogueira al., 2005a) et (Nunes et al., 1991) Yes Yes Yes Yes Yes Yes Yes Unclear Yes Unclear The potential Leishmune vaccine effect on the interruption of the transmission of the disease,was assayed by monitoring, in untreated (n = 40) and vaccinated dogs (n = 32) of a Brazilian epidemic area: the kala-azar clinical signs, the FMLseropositivity and the Leishmania parasite evidence by immunohistochemistry of skin and PCR for Leishmanial DNA of lymph node and blood samples. On month 11 after vaccination, untreated controls showed: 25% of symptomatic cases, 50% of FML-seropositivity, 56.7% of lymph node PCR, 15.7% of blood PCR and 25% of immunohistochemical positive reactions. Yes No Yes Yes Yes Yes No Yes No Unclear A serological survey for canine visceral (VL) and American cutaneous leishmaniasis (ACL) was carried out during , to assess the effects of the prophylactic measures adopted in areas where there was a risk of transmission of the diseases in Rio de Janeiro. (Ogunkolade et al., 1988) Yes No Yes Yes Yes No No Yes Yes Unclear Immunization of dogs with a Leishmania infantum-derived vaccine. A partially-purified extract of Leishmania infantum was administered to healthy dogs. Post-immunization sera were found to neutralize the infectivity of L. infantum and to abate the development of L. major. Supporting publications 2013:EN by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender considered as an output adopted by the Authority. The European food Safety Authority reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

178 (Oliveira-Lima et al., 2002a) Yes No Yes No No No No No No Unclear Deltamethrin impregnated polyvinyl chloride dogs collars were tested to assess if they were as effective in protecting dogs. Minas Gerais, Brazil. (Oliveira-Lima et al., 2002b) Yes No Yes Yes Yes Yes No No Yes No Preliminary results of a field trial to evaluate deltamethrinimpregnated collars for the control of canine leishmaniasis in northeast Brazil. (Palatnik-de- Sousa et al., 2009) Yes No Yes Yes Yes Yes Unclear Yes Unclear No Demonstration that vaccination with Leishmune did not interfere with the serological control campaign (110,000 dogs). Only 1.3% of positivity (76 among 5860) was detected among Leishmune uninfected vaccinees. Analysis of the possible additive effect of Leishmune vaccination over dog culling, on the decrease of the incidence of CVL and VL in two Brazilian endemic areas, from 2004 to Supporting publications 2013:EN by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender considered as an output adopted by the Authority. The European food Safety Authority reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

179 (Palatnik-de- Sousa, 2001) Yes Yes Yes Yes Yes Yes Yes Unclear Yes Unclear Studies on the FML vaccine: a second-generation candidate for vaccination against murine and canine visceral leishmaniasis. (Papierok al., 2005) et Yes Yes Yes Yes Yes Yes Yes Unclear Yes Unclear Protection against natural visceral leishmaniasis in dogs immunized using a new vaccinal candidate: field trial. (Parra et al., 2007) Yes No Yes Yes Yes Yes No No Yes No A group of 600 healthy and asymptomatic dogs from Brazilian canine visceral leishmaniasis endemic areas was vaccinated with three sc doses of Leishmune which is the industrialized formulation of the FML-saponin, recently licensed for commercialization in Brazil, which previously showed 76-80% vaccine efficacy against canine visceral leishmaniasis. Supporting publications 2013:EN by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender considered as an output adopted by the Authority. The European food Safety Authority reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

180 (Podaliri Vulpiani et al., 2009) Yes No Yes Yes Yes Yes No No Yes Unclear This retrospective study refers to a 4-year period, from January 2004 to December 2007, considering a group of dogs kennelled by the Istituto Zooprofilattico Speimentale dell Abruzzo e del Molise G. Caporale in a long term shelter, located in Teramo, Abruzzo, Italy where CanL is endemic. (Puccini, 1999) Yes No Yes Yes Yes Yes Yes Yes Yes No A study of the efficacy of deltamethrin-impregnated dog collars (Scalibor ProtectorBands) for the control of canine leishmaniasis is described. 120 farm dogs were serologically negative (IFAT) for leishmaniasis in May-June 1998 when the collars were put on 61 dogs; 59 were kept as untreated controls. At an examination in November 1998, one uncollared dog showed adenopathy, variation of its seroprotein pattern and a positive serology (1:80). (Quinnell al., 1997) et Yes No Yes Yes Yes Yes No Yes No Unclear Estimation of the incidence rate, serological conversion rate and basic case reproduction number (R0) of Leishmania infantum from a cohort study of 126 domestic dogs exposed to natural infection rates over 2 years on Marajo Island, Para State, Brazil. Supporting publications 2013:EN by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender considered as an output adopted by the Authority. The European food Safety Authority reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

181 (Reithinger et al., 2004) Yes No Yes Yes Yes Yes Unclear Unclear Yes No In a zoonotic visceral leishmaniasis (ZVL)-endemic area in Brazil, deltamethrin-impregnated collars (DMC) were fitted to 136 dogs for 5 months and significantly reduced the odds of increasing their anti-leishmania antibody titer during this period by 50% (95% confidence interval 29-87%, P=0.01), as compared with a population of 97 uncollared dogs with pre-intervention prevalence within the same town. (Ribeiro et al., 2005) Yes No Yes Yes Yes Yes No Yes Yes No Seventeen dogs from the same shelter in an endemic visceral leishmamasis region were observed for 24 months. (Ribeiro, 2009) No No Yes Yes Yes Yes No No No Unclear Treatment of canine visceral leishmaniasis in Brazil. Supporting publications 2013:EN by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender considered as an output adopted by the Authority. The European food Safety Authority reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

182 (Ristow and Perez Junior, 2009) Yes No Yes Yes Unclear Yes No No Yes Unclear Study of seroconversion of dogs after vaccination against canine leishmaniasis. The purpose of this work was to document the lack of seroconvertion of animals vaccinated with a vaccine based in fucose mannose ligand using current kits approved in Brazil. (Sabate et al., 2012) Yes No Yes Yes Yes Yes Yes Yes Yes No 93 healthy and seronegative dogs (DAT <1/400) were treated with Leisguard (Domperidone) (Sagols et al., 2012) Yes No Yes Yes No No No Yes Yes Unclear Twelve Beagle dogs were included in the study. The dogs were divided into two groups: a control group (n = 2) and a vaccinated group (n = 10) which was given three injections of LiESP/QA-21 (CaniLeish ; Virbac, France), a vaccine composed of purified excreted/secreted proteins (ESP) from Leishmania infantum. Supporting publications 2013:EN by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender considered as an output adopted by the Authority. The European food Safety Authority reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

183 (Senra et al., 1985) (Silva 2005) et al., Yes No Yes Yes Yes Yes No Yes No Unclear Study of the general aspects of the control, serological survey in dogs and treatment of the human case, from an outlying district of Santarém, whose diagnosis was confirmed serologically at the Instituto Evandro Chagas (IEC), started the research work to verify the dimensions of Focus leishmaniasis in Santarem. The survey conducted by the IEC entolomológico showed high density of Lutzomyia longipalpis in various districts, with infection rate of 7.1%. The serological survey of canine population, a total of 4593 samples showed 1,486 (32.3%) were positive. Yes No Yes Yes Yes Yes No Yes No Unclear An in-depth study of dogs as a reservoir for Leishmania chagasi in the peri-urban environment through clinical and serological follow-up using the immunofluorescence and Western blot techniques. (Vieira and Coelho, 1998) No No Yes Yes Yes Yes No No No Unclear Review of visceral leishmaniasis in general. (Yeadon, 2011) No No Yes Yes No Yes No No No Unclear Review of the proposed changes to the Pet Travel Scheme (PETS). Supporting publications 2013:EN by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender considered as an output adopted by the Authority. The European food Safety Authority reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

184 C. SCORING MATRIX USED TO EVALUATE TIME LIMITATION FACTORS OF PUBLISHED MODELS IDENTIFIED THROUGH THE LITERATURE SEARCH Supporting publications 2013:EN by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender considered as an output adopted by the Authority. The European food Safety Authority reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

185 D. PEDIGREE MATRIX USED FOR THE DATA QUALITY ASSESSMENT 8 OF THE MODELS IDENTIFIED AS MOST SUITABLE FOR THE SIMULATION MODELLING COMPONENT OF THE PROJECT: 8 The data quality assessment were not performed on the data used in the published articles, but rather on data required for application of the model to the context of the European Union. Supporting publications 2013:EN by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender considered as an output adopted by the Authority. The European food Safety Authority reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

186 E. PEDIGREE MATRIX USED FOR THE ASSUMPTION ASSESSMENT OF THE MODELS IDENTIFIED AS MOST SUITABLE FOR THE SIMULATION MODELLING COMPONENT OF THE PROJECT: Supporting publications 2013:EN by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender considered as an output adopted by the Authority. The European food Safety Authority reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

187 F. MODEL CHARACTERISTICS: DISEASE AND LEISHMANIA SPECIES MODELLED, AND GEOGRAPHICAL AREA OF RELEVANCE OF THE MODELS: Supporting publications 2013:EN by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender considered as an output adopted by the Authority. The European food Safety Authority reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

188 G. MODEL CHARACTERISTICS: TYPE OF MODELS AND POPULATIONS OF HOSTS AND VECTORS REPRESENTED: Supporting publications 2013:EN by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender considered as an output adopted by the Authority. The European food Safety Authority reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

189 H. MODEL CHARACTERISTICS: CONCEPT USED TO MODEL TRANSMISSION DYNAMICS AND MODEL PARAMETERS: Supporting publications 2013:EN by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender considered as an output adopted by the Authority. The European food Safety Authority reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

190 I. TIME LIMITATION SCORES OF MODELS IDENTIFIED AS POTENTIALLY SUITABLE FOR THE PURPOSE OF THE PROJECT: Supporting publications 2013:EN by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender considered as an output adopted by the Authority. The European food Safety Authority reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

191 J. NARRATIVES OF THE TEN PUBLISHED MODELS AS POTENTIALLY SUITABLE FOR THE SIMULATION MODELLING COMPONENT OF THE PROJECT: a. Article ID 1 (Hasibeder et al., 1992) i. Description: This article describes a deterministic SEIR (susceptibleexposed-infectious-recovered) model for the spread of leishmaniosis in dogs (Figure 65). The authors assumed two modelling scenarios; type A dogs that become infected and infectious for life with a high death rate because of the infection, and type B dogs that are able to develop an immunological response and susceptible after cure. The model does not include other reservoirs or humans populations. b. Figure 65: Flow diagram of the compartmental model of canine leishmaniosis (reproduced 9 from Hasibeder et al. (1992)) i. Advantages and limitations: This model introduces heterogeneity in the vector bite rate, depending on the dog activity (indoor/outdoor or pet dog/working dog). Dye (1996) cites this model as the theoretical foundation for the model he later reported. c. Article ID 2 (Ben Salah et al., 1994) 9 Permission to reproduce images from published articles will be requested if model diagrams are included in the final project report. Supporting publications 2013:EN

192 i. Description: This article written in French describes a deterministic SEI (susceptible-exposed infectious) model, and is restricted to dogs and sand flies populations (Figure 66). Figure 66: Flow diagram of canine leishmaniosis (reproduced from Ben Salah et al. (1994)). Advantages and limitations: Authors provided a sensitivity analysis of the model parameters, as they acknowledged that some parameter values were not supported with field data. The model overestimates the prevalence of leishmaniosis; therefore they introduced a seasonality component of the vector population to successfully match the prevalence observed in the field. c. Article ID 3 (Gelvez Pinto, 1994) i. Description: The authors modelled the transmission of leishmaniosis in four subsystems: humans, animals, vectors, environment (altitude, latitude, temperature, and rainfall level), and control measures (vector control). ii. Advantages and limitations: Authors indicated that the model included a set of 120 differential equations, but they are not explicit. The diagram of the model structure provided in the article is small and unreadable. The model was parameterised using literature and expert opinions. d. Article ID 4 (Dye, 1996) i. Description: The author used a deterministic SEIR model to describe the transmission of L. infantum in dogs (Figure 67). The objective of the article was to evaluate the effectiveness of various methods to control canine and human visceral leishmaniosis. The model includes both dog and human populations, whereas the disease dynamics of the sand flies is modelled indirectly. Supporting publications 2013:EN

193 Figure 67: Flow diagram of the compartmental model of zoonotic visceral leishmaniosis. Solid lines indicate natural transitions between states; unlabelled inputs and outputs represent births and death rates, respectively. Dashed lines mark additional pathways made possible by specified methods of control (reproduced from Dye (1996)). ii. Advantages and limitations: The model is centred on the dynamics of L. infantum in the dog population. Rather than explicitly modelling diseases states in the sand fly population, the author used vectorial capacity to combine parameters associated with the vector to evaluate the effect of different mitigation measures. Multiple control practices were explicitly incorporated to evaluate their effect on decreasing L. infantum incidence in both humans and dogs. The model is well specified and relatively easy to implement in its deterministic form. As with any deterministic ODE model, it may not be appropriate to model disease eradication in small populations. Some model parameters were obtained from studies in Brazil and may not be relevant to the European reality. Also, several parameters were not explicitly reported, as the study focused only on the relative change in L. infantum incidence based on different disease mitigation measures. The sensitivity analysis used to discuss the relative efficacy of mitigation measures was incorrectly performed, as the author only evaluated the effect that percentual changes in the mitigation measure parameters could have on disease incidence/prevalence. Since some mitigation measures may have much more percentual impact than others, the findings of the study are questionable. Nonetheless, the model structure is appropriate and potentially relevant to this study. Supporting publications 2013:EN

194 e. Article ID 5 (Palatnik-de-Sousa et al., 2004) i. Description: The authors used the Dye (1996) model to evaluate the effect of a higher rate of dog removal (compared to the removal rate used by Dye (1996)) on the dynamics of the leishmaniosis in dogs in Brazil. ii. Advantages and limitations: An attempt was made to derive the model parameters from empirical data from Brazil, but several estimates were dubious or incorrectly estimated. f. Article ID 6 (Bacaer and Guernaoui, 2006) i. Description: This is a deterministic SIR (susceptible-infected-recovered) model, which therefore considers that infected individuals can become immune (or removed from the disease dynamics) after infection. The model includes seasonality in the vector population, and a latent period in humans. ii. Advantages and limitations: The model parameter values were estimated from epidemiological data from a province in Morocco. The model emphasis is on cutaneous leishmaniosis in humans and includes humans and vectors populations, but does not include vertebrate reservoirs. g. Article ID 7 (Rosales and Yang, 2007) i. Description: The authors used a deterministic SIR model to evaluate the transmission of American integumentary leishmaniosis in humans in an endemic area of Argentina (Figure 68). The model includes human, dog, and vector populations. The authors used the model to estimate the basic reproduction rate and the force of infection during the epidemic period of the disease. Supporting publications 2013:EN

195 Figure 68: Compartmental model describing the transmission of American integumentary leishmaniosis. Model includes two vertebrate hosts (humans and dogs) and an invertebrate host (vector) (reproduced from Rosales and Yang, 2007). ii. Advantages and limitations: This model is similar to the one presented by Dye (1996), but used contact-based transmission parameters (similar to the β parameter of the Kermack-McKendrick SIR model) rather than the vectorial capacity used in Dye (1996). Model parameters were estimated epidemiological leishmaniosis data from an endemic area of Argentina. h. Article ID 8 (Chaves, 2008) i. Description: Authors present a deterministic SIS (susceptible-infected-susceptible) model with incidental hosts (humans), two different reservoir hosts (dogs and donkeys), and the vector population for the transmission of American cutaneous leishmaniosis in Venezuela. The main objective of the model was to explain a leishmaniosis outbreak in a town in Venezuelan and therefore, the model parameter were estimated from this outbreak. Two model formulations were developed, treating the two reservoirs species as separated populations or as a group. Although the main model is deterministic, they also included uncertainty using a Normal distribution for the demographic and environmental parameters, as the authors argued that the deterministic model would likely predict eradication given the small population of the town. This article is written in Spanish. Supporting publications 2013:EN

196 ii. Advantages and limitations: Only the results of the main deterministic model were discussed and they concluded that both species of reservoirs (dogs and donkeys) were likely involved in this outbreak. This model inference is limited, as it only simulates the transmission of American cutaneous leishmaniosis between humans in a defined small population setting, but provides sparse details on the animal reservoirs. i. Article ID 9 (Elmojtaba et al., 2010) i. Description: The authors presented a SIR model for the transmission of visceral leishmaniosis between humans, reservoir animals and the vector in Sudan (Figure 69); unlike other models, this model assumes that humans can become either immune or a reservoir. Figure 69: Compartmental model (reproduced from Elmojtaba et al, 2010). ii. Advantages and limitations: The model simulates the transmission of Leishmania between humans, and the effect of treatment to reduce the cases of human reservoirs. The authors also explored the control of vectors to reduce the biting rate, and concluded that treatment and vector control are needed to eradicate the disease. j. Article ID 10. (Agyingi et al., 2011) i. Description: This is a deterministic SIS model; therefore infected individuals can become susceptible after infection. This model includes three types of populations; human (incidental host), animals (reservoirs), and sand flies (vector). Authors simulated different scenarios with low and high biting rates of the vector and recovery rate of the reservoir. Some scenarios lead to disease-free equilibrium (i.e. disease eradication). Supporting publications 2013:EN

197 ii. Advantages and limitations: The model was designed for cutaneous leishmaniosis in humans. However authors indicated that the model can be used for other presentations of the disease in humans. Although the model includes a generic animal reservoir, it does not specifically include a dog population. Supporting publications 2013:EN

198 K. DATA QUALITY ASSESSMENT SCORES FOR PARAMETERS USED IN THE HUMAN COMPARTMENT OF THE SIMULATION MODELS: The table lists the main parameters of the three models being assessed, indicates in which models the parameters are used, and provide NUSAP/Pedigree data quality scores for each parameter. Supporting publications 2013:EN by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender considered as an output adopted by the Authority. The European food Safety Authority reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

199 L. DATA QUALITY ASSESSMENT SCORES FOR PARAMETERS USED IN THE DOG COMPARTMENT OF THE SIMULATION MODELS: The table lists the main parameters of the three models being assessed, indicates in which models the parameters are used, and provide NUSAP/Pedigree data quality scores for each parameter. Supporting publications 2013:EN by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender considered as an output adopted by the Authority. The European food Safety Authority reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

200 M. DATA QUALITY ASSESSMENT SCORES FOR PARAMETERS USED IN THE SAND FLY COMPARTMENT OF THE SIMULATION MODELS: The table lists the main parameters of the three models being assessed, indicates in which models the parameters are used, and provide NUSAP/Pedigree data quality scores for each parameter. Supporting publications 2013:EN by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender considered as an output adopted by the Authority. The European food Safety Authority reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

201 N. ASSUMPTION ASSESSMENT SCORES: The table lists the main assumptions of the three models being assessed, indicates in which models the assumption is explicit (E) or implicit (I), and provides NUSAP/Pedigree scores for each assumption. Supporting publications 2013:EN by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender considered as an output adopted by the Authority. The European food Safety Authority reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

202 O. ASSUMPTION ASSESSMENT RESULTS: FOR EACH ASSUMPTION, SCORE FOR EACH CRITERIA AND SUM OF SCORES Supporting publications 2013:EN by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender considered as an output adopted by the Authority. The European food Safety Authority reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

203 P. STRUCTURE OF THE SIMULATION MODEL This appendix describes various technical aspects of the implementation of the simulation model used to calculate: Probability of a dog returning infected from an endemic area (P Inf ) - Index Case module Probability of endemicity within a contact network of dogs (P End ) - Transmission module Probability of endemicity within a region (P EndRegion ) - Import module 1. Modelling canine leishmaniosis transmission between Dogs 1.1. Differential equations Transmission of canine leishmaniosis was simulated in the Index Case and Transmission modules to calculate P Inf and P End, using a stochastic continuous-time state transition modelling framework, which was individual-based for the dog populations while it used vectorial capacity (VC) to generalize the potential of the sand fly population to transmit the disease to dogs under the assumption of independence from the prevalence of infection in sand flies. The complex stochastic model can be more easily understood using the differential equation representation given below. = -VC*(I + C)*S/N + αδn δs = VC*(I + C)*S/N - (δ + σ)l = σl - (δ + ρ)i = ρi - δc Where: S: Susceptible population L: Latent (infected) population I: Infectious Sub-Clinical population C: Infectious Clinical population N: total population (S+L+I+C) σ : rate of transition from Latent to Infectious Sub-Clinical ρ: rate of transition from Infectious Sub-Clinical to Infectious Clinical δ: Mortality rate α: proportion of dogs that died that are replaced And where VC, the vectorial capacity, is given by: where: VC = Supporting publications 2013:EN

204 m: Number of female sand flies per dog. : Number of female sand flies bites per day. µ: Daily mortality rate of female sand flies. τ: Daily rate of transition from latent to infectious sand flies. Impact, modelling and control of canine leishmaniosis in the EU In the absence of mitigation measures, in a contact network of N dogs, Susceptible (S) dogs can become infected with L. infantum, and become Latent (L) before becoming Infectious Sub-Clinical (I). Infectious Sub-Clinical dogs that develop signs of the disease become Infectious Clinical (C). Susceptible dogs become infected at rate VC*(I + C)*S, and can die at rate δ*s. Latent dogs can become Infectious Sub-Clinical at rate σ*l, or die at a rate δ*l. Infectious Sub-Clinical dogs become Infectious Clinical at a rate ρ*i, and can die at a rate δ*i. Infectious Clinical dogs can die at a rate δ*c. After dogs die, their owner may replace them with new dogs at rate αδn. The representation above is a simplification of a stochastic model where the disease progression of each dog was modelled individually. Therefore, the rates above were actually derived from a variety of probability distributions with non-constant instantaneous survival rates. The transition of individual dogs between states on day d was simulated with Bernoulli trials, where the probabilities of transition (success in the trial) were calculated as described in sections to Probabilities of transition between states Probability of transition from Susceptible to Latent Susceptible dogs may become infected when an infectious sand fly bites them. As proposed by Dye (1996) assuming homogeneous mixing (all sand flies are equally likely to bite any given dog) the expected number of new dog infections on day d E[L d ] is proportional to VC (e.g. the expected number of new infections per Infectious dog per day d)multiplied by the number of Infectious dogs (I + C) in the network on day d and by S d /N d, the proportion of the total population N that is Susceptible and day d. Therefore, assuming a constant risk of infection during the exposure period and excluding the elements not relevant to the vectorial capacity, number of new infections on day d is proportional to: L d Poisson(VC*(I d + C d )*S d /N d ) However, if any of the terms above are large enough, L d can be greater than the total number of Susceptible individuals S. For example, if VC=2, I=30 and N=50 and all other 20 dogs are Susceptible, L d would be 24 which is greater than S. To avoid this, a Binomial approximation to this Poisson rate was used, where L d ~ Binomial(S d,e[l d ]). As each dog is modelled individually, the probability of infection for any given Susceptible dog is then P s L, d =VC*(I d + C d)*1/n Where the term S/N is replaced with 1/N. This approximation worked well for all the range of parameters and populations tested in the model. An alternative version of the model, where the number of newly Infectious individuals was simulated using a Poisson distribution and then allocated to individual dogs, was implemented and yielded equal results. If large populations and VC are modelled, an extra term can be added to restrict P s E[L d ] >1, i.e. L, d to 1 when Supporting publications 2013:EN

205 Probability of transition from Latent to Infectious Sub-Clinical The probability of transition from Latent to Infectious P L P L l,d = 1-exp(-(f(d)/ 1-F(d))), l on day d is given by: Where (f(d)/ 1-F(d) is the instant failure rate ( failure refers to the transition from Latent to Infectious Sub-Clinical): f(d) is the probability density at time d and F(d) is the cumulative function of the duration of the latent period up to time d Probability of transition from Infectious Sub-Clinical to Infectious Clinical The probability of transition from Infectious Sub-Clinical to Infectious Clinical P I calculated in a similar way to P L l,d : P I C, d = 1-exp(-(f(d)/ 1-F(d))), C on day d is Where f(d) is the probability density function at time d, and F(d) is the cumulative density function up to time d of the duration of the Infectious Sub-Clinical period Vectorial capacity The VC provides an estimate of the number of new secondary cases infected from an infectious case per time step. It has been extensively used to model vector borne diseases (e.g. Dye, 1996, Riveiro et al. 2010). VC was modelled as follows: where: m = Number of female sand flies per dog. VC = = Number of female sand flies bites per day. µ =Daily mortality rate of female sand flies. τ = Daily rate of transition from latent to infectious sand flies. The term represents the probability that any given sand fly survives one day; it is assumed to remain constant throughout the life of a sand fly and therefore, exponentially distributed. The term represents the probability that a surviving sand fly is able to transfer the infection via bites, as 1/τ is the average latent period in sand flies and τ is the daily rate of transition from latent to infectious sand flies. Multiplying the density m by the squared bite rate 2 (to represent the fact that a sand fly has to bite both an Infectious and a Susceptible dog to transmit the disease) by returns the number of infectious bites per sand fly per day, and dividing it by the daily mortality rate µ then provides the number of daily new dog infections resulting from one Infectious dog in a Susceptible population. Sand flies breeding season starts when daily average temperatures are higher than 14 C (Hartemink et al., 2011), therefore no sand fly activity is observed during the winter months. The Transmission module includes two seasons, winter and non-winter. During the winter season there are no active sand flies and thus no transmission of canine leishmaniosis. The duration of winter is modelled using a fixed value (Appendix Q. ), shorter in the Index Case module than in the Transmission module to represent differences in climates between endemic and non-endemic areas. Supporting publications 2013:EN

206 2. Modelling the overall probability of endemicity into a region A stochastic import risk assessment module was implemented to calculate the overall probability of at least one contact network becoming endemic within a previously CanL free region (P EndRegion ). In the Import module, the number of infected dogs introduced into the disease-free region (N Inf ) when no testing and exclusion strategy was considered was modelled via two pathways: N InfTravel ~Binomial(N pop, P Inf ) N InfImport ~Binomial(N pop - N InfTravel, P InfCA ) where N pop is the total number of dogs imported from endemic areas, P Inf is the probability that a dog returning from an endemic area is infected, and P InfCA is the probability that a dog imported from a CanL endemic area is infected. Assuming independency between the dogs imported and that dogs are equally likely to enter any contact network within the non-endemic region, P EndRegion was then calculated for dogs returning from endemic areas as: And for dogs imported from endemic areas as: 1-(1-P End ) NInfTravel 1-(1-P End ) NInfImport where P End is the probability of endemicity in any contact network as calculated in the Transmission module for both pathways. The term (1-P End ) NInf represents the probability that none of the N Inf imported dogs produce infection within their recipient network, so 1-(1-P End ) NInf provides the probability of endemicity (following the introduction of an infected dog) in at least one contact network in the region. This calculation is very simplified, as it does not take into account clustering and/or overlapping of networks, nor does it consider the actual number of contact networks within a region, which was not available at the time this analysis was performed. 3. Modelling mitigation measures The stochastic model of transmission of canine leishmaniosis was amended to include control measures, as shown in the amended differential equation representation below. = -VC*(1-γ)*(I + C)*S/N + αδn δs = VC* (1-γ)*(I + C)*S/N - (δ + σ)l - (θ + π)l = σl - (δ + ρ)i = ρi - δc - ηc = (θ + π)l + ηc δr With VC given by: Supporting publications 2013:EN

207 VC = And where the following terms in green represent control measures: π: prophylactic medication proportion θ: vaccination proportion η: curative treatment proportion γ: repellent proportion υ = insecticide proportion In addition to the transmission dynamics previously presented in section 1.1, the amended model accounts for the effects of control measures described hereafter. Resistant (R) dogs include dogs that have developed an immune response which prevents them from becoming Infectious, as a result of effective vaccination or treatment (curative or preventative). Latent dogs can become Resistant following either prophylactic medication (at rate π L) or vaccination (at rate θ*l). Infectious Clinical dogs become Resistant after treated at a rate η*c. Infectious Clinical (C) dogs that are not diagnosed and treated (or in which the treatment is not effective) remain infectious. Resistant dogs can die at a rate δ*r. As with the basic representation of transmission of canine leishmaniosis, the representation above is a simplification. The transition between infection states was modelled individually for each dog. For each of the control measures the proportion shown above was modelled using two parameters, use and efficacy, derived from a variety of probability distributions (Appendix Q. ). The transition of individual dogs between states on day d was simulated with Bernoulli trials, where the probabilities of transition (success in the trial) were calculated as described in sections to Probability of transition from Latent to Resistant The probability that vaccinated dogs become Resistant on day d is given by: P L R, d = Vaccine use * Vaccine efficacy The probability that dogs treated preventatively against canine leishmaniosis transition to the Resistant state on day d is given by: P L R, d = Prophylactic medication use * Prophylactic medication efficacy 3.2. Probability of transition from Clinical to Resistant Infectious dogs with clinical signs of canine leishmaniosis that are successfully treated transition to the Resistant state. The model assumes that only a proportion of dogs are diagnosed but as soon as it happens, dogs are treated. Treatment use is thus defined as the proportion of dogs diagnosed with leishmaniosis and subsequently treated. Treatment efficacy is defined as the probability that a treated dog becomes Resistant. Assuming independence between these two parameters, the probability that clinical dogs diagnosed with canine leishmaniosis and treated transition to the Resistant state on day d is given by: P C R, d = Treatment use * Treatment efficacy Supporting publications 2013:EN

208 3.3. Effect of repellent on vectorial capacity and transition from Latent to Infectious Whether repellent is used on a dog or not is simulated using a Bernoulli (p ) distribution where p is the proportion of the dog population on which repellent is used. When used on Susceptible dog(s), repellent reduces the biting rate of sand flies (and therefore, the vectorial capacity) proportionally to its efficacy: = *(1-Repellent efficacy) When used on Infectious dog(s), repellent reduces the transmission of Leishmania from these dogs to Susceptible sand flies. This is represented in the models by reducing the total number of Infectious dogs with repellent by a proportion equal to the repellent efficacy: (I + C) = (I + C)*(1- Repellent efficacy) 3.4. Effect of insecticide on vectorial capacity Whether insecticide is applied in the area of interaction of the contact network of dogs being modelled or not is simulated using a Bernoulli (p ) distribution where p is the probability that insecticide is applied. When used, insecticides result in a decrease of the density of sand flies proportional to its efficacy: m = m*(1-insecticide efficacy) 3.5. Testing and exclusion on import module This mitigation measure assumes that a rapid screening test is used on animals and that positive dogs are excluded i.e. not allowed for introduction/movement into the non-endemic areas. The N pop dogs to be imported are tested with certain proportion of use (P test ) and test sensitivity (Se). P test is the proportion of dogs tested and is assumed to be the same for all CanL infection states. Se is the probability of a dog testing positive given that the dog is infected. The probability of any infected not being detected P D- is then: P D- =(1-P test ) * P CanL + P test * P CanL *(1-Se) where (1-P test ) * P CanL is the joint probability of any given dog to be untested and infected, and P test * P CanL *(1-Se) is the joint probability of a dog being tested, infected, and test negative. P CanL was P InfCA and P Inf for the import and travelling pathways, respectively. Therefore, the total number of infected dogs entering the non-endemic area N Inf was Binomial(N pop, P D- ), which was calculated separately for both pathways. 4. Modelling dog populations The modules that simulate transmission of canine leishmaniosis represent small population of dogs, which include travelling dogs, index case dogs, and contact network dogs. These dog populations are modelled over a certain period of time: i) from the day that the index case became Infectious in the Transmission module and the day of arrival in the endemic area in the Index Case module, and ii) for 3 years in the Transmission module or for the duration of travel in the Index Case module. Supporting publications 2013:EN

209 4.1. Contact network size A contact network of dogs represents a group of dogs that interact and may be able to transmit canine leishmaniosis to each other via sand flies, because they share a common geographical area and often have direct (e.g. walking or playing at the same time) or indirect (e.g. walking in the same parks at different times) contacts. The contact network size is modelled in the Index Case and Transmission modules as a fixed value (50 dogs) that represents a high level of contact between dogs in the network (Westgarth et al., 2009); however that can be changed for a further scenario analysis Day of the year travelling The day of the year travelling is the day when the travel to endemic area occurs, and is used to determine if the travel occurred during the season where sand flies are inactive (winter) or active (non-winter period) Travelling days The number of days travelling to an endemic area represents the period of exposure (in days) of travelling dogs to canine leishmaniosis infection in an endemic area Prevalence of canine leishmaniosis in endemic areas The prevalence of CanL is used in the Index Case module to estimate the number of infected dogs in the contact network of the endemic area where Susceptible dogs travel to. The variability and uncertainty of the number of Infectious dogs is modelled using a Binomial distribution with n number of dogs in the contact network (50) and the probability of Infected given by a posterior Bayesian estimation of true prevalence Age In the model, dogs in a contact network have different initial ages, and during the modelling period some may die and be replaced. The variability of the age (in days) of each i th dog in the network at day 0 is modelled using a LogNormal distribution (mean, SD) Daily mortality rate The variability of the daily mortality rate of the i th dog during the x th day of age (P idead ) is modelled using a Bernoulli distribution with a probability P idead coming from a Weibull distribution with shape = α and scale = β where: 4.7. Replacements After a certain period of time following the death of a dog (time to replacement, given by a Pert distribution), a replacement dog is introduced into the contact network with a probability of replacement modelled using a Beta distribution. The model assumes that all replacement dogs are Susceptible. The age of the Susceptible replacement dogs is given by the same distribution used for the age of the dogs in the initial contact network at day 0. As more than two dog replacements within three years are unlikely given the mortality and age of dogs, up to 2 replacements are allowed in the model Number of dogs travelling per household Supporting publications 2013:EN

210 P Inf and P End were calculated considering only one dog per trip, and one Infectious dog introduced into a contact network in a non-endemic area. However, both the Index Case and Transmission modules can simulate more than one dog per trip or index cases. Selected scenarios were run to assess the impact on outputs of multiple dogs per travel or introduction into a network of dogs. In these scenarios, the variability on the number of dogs travelling per household was modelled using a zerotruncated Poisson distribution with a parameter lambda. Supporting publications 2013:EN

211 Q. DERIVATION OF PARAMETERS USED IN THE SIMULATION MODEL 1. General approach and inference The parameters used in the modelling section included vector and dog population parameters, parameters affecting the disease dynamics at the individual dog level, and efficacy of different mitigation measures, among others. Therefore, a variety of sources of data were consulted to parameterize the model. When available, peer-reviewed evidence was preferred to derive parameters. In absence of relevant published literature, expert opinion was used (see Impact Assessment section). Whenever possible, the published articles selected were informed by the results of the systematic review portion of this project. However, as the systematic review and the modelling work were carried in parallel, not all articles used for the model parameterization match those from the systematic review. In presence of uncertainty, a maximally noncommittal approach to the parameter estimation was taken. In other words, when candidate uncertainty distributions were available to model a parameter, the one with the largest uncertainty (e.g. greatest variance) was chosen. Moreover, in cases with no empirical evidence to support a parameter, a scenario analysis approach was favoured. Both classical (frequentist) and Bayesian statistical methods were used to derive parameters modelling population heterogeneity, variability and randomness, and parameter uncertainty. The choice of inference was based on convenience and parsimony (e.g. analytical solutions preferred over simulation ones), and were also based on the amount of data available. In general, when little data was available to derive a parameter, Bayesian methods with vaguely informative priors were preferred over frequentist methods, whereas when sufficient data was available, the simplest method (regardless of inference type) was selected. For example, when estimating the uncertainty in a proportion, if the number of successes s out of n trials was reported, the Bayesian conjugate prior Beta(S+1, n-s+1) was used as its analytical form could be directly presented in the model. Likewise, when the results from RCTs of mitigation measures were reported, efficacy was calculated using the standard formula:, where the uncertainty in the RR was modelled using the classical asymptotic normal approximation of the ln RR: Therefore, the probability density function of RR in the antilog was, which is equal to, as parameterized in software. All mitigation measures were assumed to be at least as efficacious as placebo and therefore, the uncertainty distributions of efficacy were truncated at 1, yielding a minimum efficacy of 0. When no analytical solutions were possible, Bayesian posteriors using MCMC were estimated elsewhere (using the WinBUGS software) and resampled in the model. An example of such case was the estimation of true prevalence of infected dogs in endemic areas using latent class analysis. Finally, when conflictive data sources were present but potentially relevant, a mixture distribution was used to sample (with equal weight) from the uncertainty distributions from each data source. Supporting publications 2013:EN

212 The parameters included in the model are listed in Table 27 in the main document. 2. Vectorial capacity The uncertainty in the mean number of sand flies per dog m was estimated from a study in a central municipality of Portugal (Branco et al., 2013). In this study, 264 CDC miniature traps were used in 17 parishes to capture sand flies. The number of traps per parish and the mean number of Phlebotomus perniciosus captured per trap per night were used to parameterize a Gamma (Scale, Shape) conjugate using a vague 1/ prior, where the shape parameter was the total number of sand flies captured in all traps and days and the scale parameter was the inverse of the total number of trap/days. In absence of dog specific data, it was then assumed that the traps used in the study were equally attractive to sand flies than dogs and therefore the mean number of sand flies per trap was similar to the mean number of sand flies per dog. The uncertainty in the mean number of bites per female sand fly per day was elicited from experts. Four experts were asked to provide an estimation of the minimum, maximum and most likely value of the total number of bites (blood meals) taken by an average female sand fly. One expert didn t provide an estimation of the most likely value, so was excluded from further analysis. As the expert opinions largely agreed (i.e. little heterogeneity in their response), a fixed effect meta-analysis using the inverse of the variance (Higgins & Green,2008) to weight the three expert opinions was used to estimate the mean bites and SE bites of the mean total bites per fly. The uncertainty in the mean total bites was then Normal( bites, SE bite ). Finally, Normal( bites, SE bite ) was divided by the mean sand fly lifespan (in days or weeks) to convert it to, the daily (or weekly) bite rate. The resulting distribution of was used directly in the calculation of Vectorial Capacity. The uncertainty of τ, the daily rate of transition from latent to infectious sand flies was estimated using the opinion of two experts using the same methodology as described for above.finally the uncertainty in the mean lifespan of sand flies was estimated from the opinion of two experts. However, the experts provided diverging estimates of the lifespan suggesting that they were providing estimates for different species. Therefore, the inverse of the mean daily (or weekly) lifespan estimates from each expert were sampled with equal weight from a categorical distribution to estimate the uncertainty in, the daily (or weekly) mortality rate of female sand flies. 3. Parameters for the disease dynamics in the dog population 3.1 Transition from Latent to Infectious Sub-Clinical The variability in the length of the latent period (from exposure to Infectious) was estimated using data from Oliva et al. (2006). This was a three-year longitudinal cohort study where non-infected dogs were introduced to a CanL endemic area in Italy. Dogs were tested on a regular basis (every 1 to 3 months) throughout the study. To estimate the variability in the length of the Latent period, it was assumed the dogs became Latent when they first tested positive to nested-pcr in bone marrow aspirate samples. Similarly, it was assumed that the time the dogs became Infectious was the time when dogs were positive to all tests (nested-pcr in bone marrow aspirate samples, IFAT in serum samples with a cutoff of 1:160 dilution and microscopy and culture of bone marrow aspirate samples). The information above was tabulated, and a parametric Weibull survival analysis was performed on these data to estimate the shape and scale parameter of a Weibull distribution of the latent period (Table 27). Supporting publications 2013:EN

213 3.2 Transition from Infectious Sub-Clinical to Infectious Clinical The variability in the length of the period between Infectious Sub-Clinical to Infectious Clinical was estimated from Oliva et al. (2006), using a parametric survival analysis, as described for the latent period. 3.3 Treatment Efficacy The uncertainty of the treatment efficacy was estimated from two clinical studies in dogs using Meglumine antimoniate alone or combined with Allopurinol. These drugs are considered the first line option for the treatment for canine leishmaniosis (Solano-Gallego et al., 2009, Oliva et al., 2010). Treatment also reduces the parasite load and the transmission of the parasite to sand flies (Oliva et al., 2010). Slappendel and Teske (1997) found that 32(s) out of 42 (n) dogs with clinical leishmaniosis recovered from the clinical stage. Similarly, Manna et al. (2008) found that 11(s) out of 18(n) patients recovered from the clinical stage. A Beta(s+1, n-s+1) distribution was used to estimate the uncertainty in the efficacy from each study, and the resulting distributions were sampled with equal weights. 4. Other environmental characteristics 4.1 Prevalence of canine leishmaniosis in endemic areas The prevalence of Infectious dogs in contact networks within endemic areas was estimated using the prevalence of clinical cases from two studies from Spain. Gálvez et al. (2010) surveyed a population of 1,076 (n) and found 29 (s) dogs that tested positive and had clinical signs compatible with leishmaniosis. Miró et al. (2012) found 19 (s) seropositive dogs with clinical signs of leishmaniosis out of a population of 418 (n) surveyed dogs. Beta(s+1, n-s+1) distributions were used to represent the uncertainty in the prevalence of clinical cases from each study. The distributions were then combined in a mixture distribution and sampled with equal weights. It was assumed that the prevalence of clinical cases of leishmaniosis is an approximation to the prevalence of Infectious dogs. 4.2 Length of the winter season (seasonality) Oliva et al. (2006) described that sand fly activity in an endemic area in Italy can be normally observed from the end of May to mid-october; therefore the duration of the non-winter period was assumed to be 5 months in endemic areas and a shorter 3 months in non-endemic areas that would presumably have lower temperatures and therefore, a shorter sand fly season. 4.3 Travelling days The mean number of outbound trips from European countries and length of the trips was obtained from 293,456,842 records from 2009 to 2011 from the European tourism information database (Eurostat, 2013). This information was recorded into 6 categories of trip lengths (frequency): from 1 to 3 nights (20.7%), 4 to 7 nights (38.1%), 8 to 14 nights (26.3%), 15 to 28 nights (11.1%), 29 to 91 nights (3.6%), and 92 to 365 nights (0.2%). Business trips were excluded from this analysis, as it was assumed that in most cases business travellers would not be accompanied by dogs. Given the very large number of records, no inference was used for this parameter and thus, the travel lengths were assumed to be representative and directly sampled using a continuous empirical distribution. Supporting publications 2013:EN

214 5. Parameters for dog populations 5.1 Age Impact, modelling and control of canine leishmaniosis in the EU In a cross-sectional serological study in Spain Gálvez et al. (2010) reported a mean age of 1898 and standard deviation of 1241 days in the study population surveyed (These parameters were used directly in a Lognormal distribution to model the age distribution of dogs. This approximation was taken as age was not a highly sensitive parameter in the model, so it was assumed that this age distribution was sufficient to represent the age distribution of the network of dogs in a non-endemic area, and that of dogs imported from non-endemic areas. 5.2 Daily mortality rate A Weibull distribution was fitted to reported data on longevity of UK breeds of dogs (O Neill et al., 2012) using MLE methods (Table 27). 5.3 Replacements A survey of 103 pet owners in Canada (McCutcheon and Fleming (2002) found that the time to replacement varied from 4 days to 10 years with a median of 4 months. These values were incorporated in a Pert distribution to model the variability in the time to replacement. The same study reported that 51(s) out 103(n) pet owners replaced their lost pet with another pet. These results were used to parameterize a Beta(s+1, n-s+1) distribution which was used to model α, the probability of replacement. 5.4 Number of dogs travelling per household Slater et al. (2008) reported that 73% of the dog owners in central Italy had 1 dog, 19% 2 dogs, 4% 3 dogs, 3% 4 dogs and 1% 5 dogs or more. Similar values were also reported by Murray et al. (2010) who surveyed pet owners in the UK. A zero-truncated Poisson distribution was fitted to these data using MLE methods (Table 27). It was assumed that the distribution of number of dogs travelling with their owners is the same as the distribution of number of dogs per household. 6. Modelling Mitigation measures The uncertainty in the efficacy of mitigation measures followed the general form of, and was explained in detail in section 1of this appendix. Therefore, the sections below focus on the data sources used to derive efficacy Vaccination The relative risk of leishmaniosis of vaccinated dogs versus control dogs was estimated from data from a vaccine trial that tested a precursor of a commercially used vaccine (Lemesre et al., 2007). Supporting publications 2013:EN

215 6.2. Repellent The uncertainty in repellent efficacy was derived from a study that measured the number of sand fly bites on dogs using collars impregnated with Deltamethrin vs. controls (Killick-Kendrick et al., 1997) Insecticide The overall efficacy of insecticides in reducing the density of sand flies was estimated using a study that evaluated 5 different insecticides (Lambda-Cyhalothrin, Pyrethrins, Carbaryl, Deltamethrin and Malathion) and 8 different application protocols in a military base in Iraq (Coleman et al., 2011).The density of sand flies was evaluated using sand fly light traps. Results indicated a mean reduction of 58.8% in the density of sand flies when comparing treated and untreated areas, and reported the mean and SE number of sand flies for both treated and control groups. The ratio of the mean number of sand flies trapped in the treated group over those in the control group provides a measure of the relative reduction of the treatment, and therefore, subtracting one from this ratio provides the efficacy of the treatment. The uncertainty in this ratio was modelled using a Normal approximation to the ratio of two normal distributions. As no other sources of data for insecticide were available from Europe, using this data assumes that a similar efficacy can be achieved if insecticides are applied in endemic or non-endemic areas of Europe Prophylactic medication A clinical trial reporting leishmaniosis in Domperidone treated dogs versus control dogs (Llinas et al., 2011) was used to calculate efficacy using the same methods as described for other efficacies Diagnostic test sensitivity Mettler et al. (2005a) found that 9 and 10 out of 17 asymptomatic dogs were detected using Dipstick test and ELISA test, respectively. The uncertainty of the diagnostic tests sensitivity was modelled using a Beta distribution with parameters alpha (number of test positives plus 1) and beta (number of true positives minus number of test positives plus 1). Supporting publications 2013:EN

216 R. PHYLUM CRITERIA (PHYLUM, 2010) Table 48: Global Epidemiology assessment criteria Affected species Types of susceptible species Impact, modelling and control of canine leishmaniosis in the EU Criteria definition Answer Score Overall results Livestock No 0/1 Wildlife Yes 1/1 Pets Yes 1/1 Persistence Animals Possible infection of the next generations (chronic carriers, healthy carriers) Yes 1/1 Environment Survival of the pathogen in the environment (soil, water etc.) No 0/2 Wildlife Wildlife is a reservoir of the pathogen Yes 2/2 Vectors Biological cycles or persistence in vectors No 0/2 Transmission Speed of spread Modalities of transmission Epidemiology in humans Transmission to humans Transmission from humans Air-borne diseases, or diseases transmitted by flying vector species Yes 2/2 Direct close contact (e.g. breeding, bites etc.) Yes 1/1 Proximity or indirect contact (e.g. fomites, equipment) Yes 1/1 Soil agent No 0/1 Water- or feed-borne disease No 0/2 Vector-borne Yes 3/3 Air-borne No 0/4 Human form of the disease Yes 1/1 Possible (common) human-to-animal transmission No 0/1 Possible (common) interhuman transmission Yes 1/1 2/3 3/7 7/14 2/3 Table 49: Zoonotic impact assessment criteria Severity in humans Criteria definition Answer Score Zoonoses Possible human form of the disease Yes 1/1 Severity of human form (if untreated) Invariable fatal (score 4) Potentially fatal in complicated cases (3) Symptomatic expression without mortality (2) Asymptomatic or mild disease in humans (1) 3/4 Overall results 4/5 Diagnosis Clinical diagnosis No clinical diagnosis elements (score 3) Clinical suspicion only (2) Supporting publications 2013:EN /3 3/6

217 Laboratory diagnosis Pathognomonic signs (1) Impact, modelling and control of canine leishmaniosis in the EU No laboratory test or confirmation technique (3) Difficult or unreliable confirmation technique (2) Good diagnostic test (1) Prevention and treatment Vaccine No vaccine (score 3) Intermediate efficacy or availability (2) Effective vaccine (1) Medical treatment No effective treatment (3) Intermediate efficacy or availability (2) Total recovery without relapse (1) Transmissibility Animal to human transmission Interhuman transmission High risk (e.g. flying human-adapted vector) (4) Significant risk (e.g. hazardous widely consumed food, non-flying human adapted vector) (3) Professional or occupational risk hazard (2) Only accidental human cases (1) High risk (e.g. flying human-adapted vector) (4) Easy transmission by proximity or direct contact (3) Transmission by direct close contact (2) Only vertical or accidental transmission, in high risk persons or particularly invasive contact (1) Epidemiological dead-end in humans (0) 1/3 3/3 4/6 1/3 4/4 1/4 4/4 1/4 Table 50: Societal impact assessment criteria Animal welfare Nature of animal discomfort Duration of welfare problem Crisis generation potential Perception of threat to humans Acceptability of threat Perception of fraud Criteria definition Answer Score Alteration in general state / demeanour Significant pain Disablement (physical / neurological) Chronic effect on essential functions Usually fatal Cases commonly involve permanent alterations or relapses Possible human cases Occupational exposure (transmission to humans by direct contact with animals) Indirect exposure (vector-borne or food-borne) Commonly fatal in human cases Animal-to-human transmission Interhuman transmission Food-borne zoonoses Exposure to potentially contaminated food products Vector borne zoonoses Human population exposed to vectors Economic operators may profit from contingencies Yes Yes Yes Yes No 4/5 Possible 1/2 Yes No Yes No Yes Yes No No 2/4 1/1 1/1 0/1 0/1 Yes 1/1 Yes 1/1 Yes 1/1 Overall results 5/7 10/14 Supporting publications 2013:EN

218 Authorities preparedness Amplifier effect of the media Up-to-date public awareness programmes No 1/1 Recent (<3 years) occurrence of the disease reported in the media Important disease-related media concern Bioterrorism potential OIE / CFS Category A classification of Category B pathogens Category C USDA list of Select Agents and Toxins Local list Included in official local list of potential bioterrorism agents Yes Yes No No No Listed pathogen No 0/1 1/1 1/1 0/4 No 0/1 0/6 Table 51: Environmental impact assessment criteria Pollution: effluents and residues Disposal of dead animals Biodiversity Plant biodiversity Wildlife disease Criteria definition Answer Score Mortality exceeds 1% of total population Protocol for disposal of dead animals Contamination of manure Disease effects animal dependent functions (pollination, fertilisation) Disease liable to cause massive mortality and threaten wildlife populations Presence of wildlife surveillance Official surveillance system for the disease in wildlife No No No No No Yes No Overall results 1/3 1/3 1/10 Supporting publications 2013:EN

219 S. LOCATION OF PANELVET PRACTICES CONTRIBUTING DATA ON CANINE LEISHMANIOSIS, FRANCE Figure 70: Location of 97 practices contributing data relating to canine leishmaniosis in France (Panelvet, 2012) Figure 71: Number of Panelvet practices per department, France 2012 (Panelvet, 2012) Supporting publications 2013:EN

220 T. RESULTS OF ONLINE VETERINARY QUESTIONNAIRE Table 52: Number of responses to veterinary questionnaires and estimates of regional prevalence of canine leishmaniosis in dogs attending veterinary clinics in France, Spain, Portugal, Italy and Greece, France Île-de-France Champagne-Ardenne Picardie Haute-Normandie Centre Basse-Normandie Bourgogne Nord-Pas-de-Calais Lorraine Alsace Franche-Comté Prevalence (%) Spain Prevalence (%) Andalucía 2.27 Aragón 2.97 Cantabria 0.31 Castilla-La Mancha 3.04 Castilla y León 0.9 Cataluña 6.15 Ceuta y Melilla - Comunidad de Madrid 3.32 Comunidad Foral de Navarra 1.45 Portugal Prevalence (%) Algarve Alentejo Norte Centro Italy Prevalence (%) Abruzzo Apulia Basilicata Calabria Campania Emilia-Romagna Friuli-Venezia Giulia Lazio Liguria Lombardia Responses (n) Responses (n) Responses (n) Responses (n) Pays de la Loire Bretagne Poitou-Charentes Aquitaine Midi-Pyrénées Limousin Rhône-Alpes Auvergne Languedoc-Roussillon Provence-Alpes-Côte d'azur Corse Comunidad Valenciana Extremadura Galicia Islas Baleares Islas Canarias La Rioja País Vasco Principado de Asturias Región de Murcia Lisboa Região Autónoma dos Açores Região Autónoma da Madeira Marche Molise Piemonte Sardegna Sicily Toscana Bolzano-Bozen Umbria Valle d'aosta Veneto Prevalence (%) Prevalence (%) Prevalence (%) Prevalence (%) Responses (n) Responses (n) Responses (n) Responses (n) Supporting publications 2013:EN

221 Greece Prevalence (%) Anatoliki Makedonia kai Thraki 7.43 Attiki 7.41 Ayion Oros - Dytiki Ellada Dytiki Makedonia 6.58 Ionioi Nisoi Ipeiros 3.80 Responses (n) Kentriki Makedonia Kriti Notio Aigaio Peloponnisos Stereá Elláda Thessalia Voreio Aigaio Prevalence (%) Responses (n) Fig. 72a Fig. 72b Fig. 72c Fig. 72d Fig. 72e Figure 72: Standardised morbidity ratio (SMR) of incidence estimates of canine leishmaniosis in dogs attending practices in France (Figure 72a), Spain (Figure 72b), Portugal (Figure 72c), Italy (Figure 72d) and Greece (Figure 72e). SMR relates to within-country estimates and is a ratio of the observed cases to the expected cases assuming a homogenous spatial distribution. Supporting publications 2013:EN

222 U. RESULTS OF ONLINE QUESTIONNAIRE FOR DOG SHELTERS Table 53: Prevalence of disease and infection due to L. infantum in dogs routinely screened for Leishmania at shelters in endemic countries within the EU Dogs inhabiting shelters routinely screening for L. infantum at time of questionnaire completion (a) Country Number with clinical disease (%) Number testing positive (%) Number of dogs tested (n) Proportion of infected dogs with clinical disease (%) France 0 (0) 0 (0) 50 - Spain 56 (5.5) 126 (12.3) Italy 57 (4.0) 110 (7.8) Greece 35 (5.8) 87 (14.5) Total 148 (4.8) 323 (10.4) Dogs inhabiting shelters routinely screening for L. infantum over the last 12 months Country Number with clinical disease (%) Number testing positive (%) Number of dogs tested (n) Proportion of infected dogs with clinical disease (%) France 0 (0) 0 (0) Spain 89 (3.5) 125 (4.9) Italy 257 (19.1) 268 (19.9) Greece 55 (3.4) 119 (7.3) Total 401 (6.5) 512 (8.4) (a): Only shelters which had entered values for the total number of dogs living in the shelter, the number testing positive and the number of sick dogs due to leishmaniosis were included in these calculations. Questionnaires with missing values for any of the relevant questions were excluded from this part of the analysis Supporting publications 2013:EN

223 % positive % positive V. LABORATORY DATA, C.RE.NA.L, ITALY (VITALE, 2013) Impact, modelling and control of canine leishmaniosis in the EU Table 54: Annual number of samples tested and positive for Leishmania infantum at Istituti Zooprofilattici Sperimentali laboratories, Italy, Total positive samples Total number of samples Seroprevalence Year ,480 58,951 23% ,124 70,481 20% ,850 61,411 23% ,027 73,989 16% ,509 87,125 17% 60% 50% 40% 30% 20% 10% 0% Figure 73: Regional seroprevalence of Leishmania infantum diagnosed at Istituti Zooprofilattici Sperimentali laboratories, Italy, Samples not received from Abruzzo or Molise. 60% 50% 40% 30% 20% 10% 0% Figure 74: Regional seroprevalence of Leishmania infantum diagnosed at Istituti Zooprofilattici Sperimentali laboratories, Italy, Samples not received from Abruzzo or Molise. Supporting publications 2013:EN

224 % Positive % Positive % Positivite Impact, modelling and control of canine leishmaniosis in the EU Figure 75: Regional seroprevalence of Leishmania infantum diagnosed at Istituti Zooprofilattici Sperimentali laboratories, Italy, Samples not received from Calabria Figure 76: Regional seroprevalence of Leishmania infantum diagnosed at Istituti Zooprofilattici Sperimentali laboratories, Italy, Figure 77: Regional seroprevalence of Leishmania infantum diagnosed at Istituti Zooprofilattici Sperimentali laboratories, Italy, Supporting publications 2013:EN

225 W. LABORATORY DATA, HELLENIC PASTEUR INSTITUTE, GREECE (DOTSIKA, 2013) Table 55: Annual number of samples tested and positive for Leishmania infantum at the Hellenic Pasteur Institute, Greece, Year Number of samples Dogs (a) % positive Humans (b) Number of % positive samples (a): Dog and Human serum samples from suspected cases referred for diagnosis in the Hellenic Pasteur Institute, Greece (personal communication Dr. A. Mentis, Director of Diagnostics Department) and in the Center of Athens Veterinary Institutions (personal communication Dr S. Boutsini, Director of Institute of Infectious and Parasitic Diseases) (b): Serum samples were screened for L. infantum specific antibodies using the indirect immunofluorescence test (IFAT) Proportion of hospitalised children with suspect leishmaniosis testing positive with an IFAT test and more recently with an rk39 immunochromatographic test. Figure 78: Proportion of suspect human leishmaniosis cases admitted in the Agia Sophia, Athens Childrens Hospital, Greece, testing positive for L. infantum. Personal communication Dr M. Giannaki, Director of Department of Microbiology Supporting publications 2013:EN

226 X. PUBLISHED ESTIMATES OF CANINE LEISHMANIOSIS INFECTION AND DISEASE IN FRANCE, SPAIN, PORTUGAL, ITALY AND GREECE. Table 56: Published estimates of canine leishmaniosis infection and disease in France, Spain, Portugal, Italy and Greece. France Cévennes 2 vegetation zones: Holm and White Oak (HWO) & Holm Oak (HO) Cévennes Var, Corsica, Bouchesdu-Rhône Data collection period Sampling / study population 1999 Dogs attending veterinary clinics February- April 1997 Hunting dogs not treated for leishmaniosis 2007 / 2008 Male military dogs wearing deltamethrinimpregnated collars All regions 2004 Dogs attending veterinary clinics Spain Data collection period Sampling / study population Sample size 259 dogs tested (134 HWO zone, 125 in HO zone) Compatible clinical signs or disease Prevalence / incidence estimate Serology (test and cut-off) 17.2 % (HWO) 9.6 % (HO) (IFAT, cut-off 1/80) % 30% (IFAT, cut-off 1:40 and counter immunoelectrophoresis, cut-off =1 line) % 0.71% (ELISA, cut-off 0.23 OD) 14% (Western Blot) 992 clinics completed questionnair e Sample size 80% PCR PCR Reference (Keck and Dereuer, 2003) (Lachaud et al., 2002) 41.4% PCR (Aoun et al., 2009) 0.41% (Bourdeau et al., 2004) Compatible clinical signs or disease Prevalence / incidence estimate Serology (test and cut-off) PCR Reference Supporting publications 2013:EN by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender considered as an output adopted by the Authority. The European food Safety Authority reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

227 Priorat 1985 to 1994 spring and early summer Impact, modelling and control of canine leishmaniosis in the EU Annual rabies vaccination campaign % (based on data from ). Annual incidence 5.7% (Dot-ELISA, Cut-off 1:800) Supporting publications 2013:EN (Fisa et al., 1999) Mallorca Published 2001 Dog shelter % 26% (ELISA) 67% (PCR) (Solano-Gallego et al., 2001a) Alicante June 1999 Annual rabies vaccination campaign 807 0% 22% (ELISA) (Alonso et al., 2010) Northern Greater Madrid June October 1992 Annual rabies vaccination campaign % (IFAT, cut-off 1:80) (Amela et al., 1995) Northwest Spain Published 2006 Dogs attending veterinary clinics (IFAT, cut-off 1:100) (Amusategui et al., 2004) Madrid region Biannually (November / April Dog shelters % 7.8% (range 0.9% April % April 1997) (IFAT, cut-off 1:100) (Miro et al., 2007b) Madrid region late May/ early July 2006 & 2007 Annual rabies vaccination campaign Alpujarras April June 2006 Dogs attending organised events (e.g. compulsory annual rabies vaccination) Alicante and Valencia, Southeastern Spain May July and October December 2008 March May 1991 Murcia City's Municipal Zoonosis Control Centre % (IFAT, cut-off 1:80) (Gálvez et al., 2010) % (IFAT, cut-off 1:160) (Martin- Sanchez et al., 2009) 212 7% (ELISA) 67% (Chitimia et al., 2011) Alpujarras Nearly all of canine census Northern Spain 2011 Stray dogs 418 Cantabria 2%, Asturias 4.7%, Orense 35.6% and Biscay 0% (IFAT, cut-off 1:100) In and around Barcelona May 2005 and December 2006 Dogs admitted to the Veterinary Teaching Hospital of Barcelona % (IFAT, cut-off 1:160) (Acedo Sanchez et al., 1996) (Miró et al., 2012) 69 clinically healthy dogs & 84 dogs with clinical signs 21.7% (clinically healthy) 35.7% (sick) (PCR) (Tabar et al., 2009) by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender considered as an output adopted by the Authority. The European food Safety Authority reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

228 In and around Barcelona May 2005 and July 2007 Clinically healthy seronegative dogs donating blood to the Animal Blood Bank Faculty of Veterinary Medicine, Autonomous University of Barcelona Lleida October 2009 Owned dogs (mostly hunting dogs living in kennels) Data collection period Sampling / study population Portugal National January 2009 Dogs attending veterinary clinics Évora Annual rabies vaccination campaign compatible with vectorborne diseases % (PCR) (Tabar et al., 2008) % Tested positive with 2 methods (IFAT, cut off 1:80, ELISA, cut off 24 U, Western blot, Immunochromatographic dipstick test) Sample size Compatible clinical signs or disease Prevalence / incidence estimate Serology (test and cut-off) % Apparent prevalence: 5.86%. True prevalence. 6.31%. True prevalence per district varied from 0.94% to 17.40% (Direct Agglutination Test (cut off titre 400)) 3,614 (1990) 0.8% (1990) 2% (1999) 3.9% (1990) 9.4% (1999) PCR (Ballart et al., 2013) Reference (Cortes et al., 2012) (Schallig et al., 2013) Supporting publications 2013:EN by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender considered as an output adopted by the Authority. The European food Safety Authority reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

229 2010 3,563 (1999) 1,485 (2010) Northeast Portugal Randomly sampled domestic dogs 0.9% (2010) 5.6% (2010) (Direct agglutination test) % (positive for at least 3 tests) (Direct agglutination test and ELISA using five different defined antigens) (Sousa et al., 2011) Urban areas of Lisbon December 2002 December 2003 Owners volunteering dogs and dog shelter 277 (owned) 97 (stray) 19.2% Owned (18.4%) Stray (21.6%) (IFAT, cut-off = 1:64, lymph node aspirates for culture) (Cortes et al., 2007) Peso da Régua, Alto Douro April 1999 Annual rabies vaccination campaign % 20.4% (fast agglutination screening test (FAST) and direct agglutination test (DAT), cut-off titre 400) (Cardoso et al., 2004b) Alijó, Alto Douro April June 2000 Annual rabies vaccination campaign % 18.7%, 95% CI: Northwest (2.5%) Intermediate (11.4%) Southern (49.9%) (DAT, cut-off titre 400) (Cardoso et al., 2004a) Data collection period Sampling / study population Sample size Compatible clinical signs or disease Supporting publications 2013:EN Prevalence / incidence estimate Serology (test and cut-off) PCR Reference Italy NW of Bologna July 1987-July 1988 Clinically healthy owned dogs 802 0% (IFAT, cut-off 1:40) (Baldelli et al., 1992) Gargona promonotory, November1989 Dogs living on farms % (Brandonisio et by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender considered as an output adopted by the Authority. The European food Safety Authority reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

230 Foggia, Apulia May 1990 (IFAT, cut-off 1:40) al., 1992) Campania Published 2002 Asymptomatic dogs selected by veterinarians working in the study area % (IFAT, cut-off 1:40) (Cringoli et al., 2002) Piedmont Jan March Perugia August 2005 Februrary 2007 Northern Continental Italy Olevano Romano, Rome January 2003 and May 2005 February to April 2004 Apulia May 1998 (Foggia) and April/May 2003 Asymptomatic dogs % Turin, 5.8% Ivrea, 3.9% Casale, 0.4% Aosta (IFAT, cut-off 1:160) Local canine registry offices were used to randomly select owned dogs >2 years old Asymptomatic dogs living near Autochthonous cases referred by local practitioners % (IFAT, cut-off 1:80) Owned dogs % of seropositive dogs Farm / kennelled dogs not treated with enctoparasiticides Western Liguria Group A: Clinical suspicion of infection. Group B: Any dog eligible Supporting publications 2013:EN Po valley (2.6%) Pre-Alpine sites (1.8%) (IFAT, cut-off 1:160) 137 farm dogs (Foggia) 303 (163 farm, 140 kennel dogs) Bari & Taranto A: n = B: n = (Subgroup I:, n = 2333; Subgroup II: 33.3% (IFAT, cut-off 1:80) Foggia: Seroprevalence = 8.75% Inicidence 4.25 / 100 dog-years Bari & Taranto: serprevalence = 18.8% Incidence: 9.52 / 100 dogyears (IFAT, cut-off 1:80) Group A (Oct 1990 Sept 1994)= 26% Group B: subgroup I = 30.3%, subgroup 2 = 22.1% (IFAT, cut-off = 1:160 or 1:40 if (Ferroglio et al., 2005) (Maresca et al., 2009) (Maroli et al., 2008) (Rossi et al., 2008) (Paradies et al., 2006) (Zaffaroni et al., 1999) by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender considered as an output adopted by the Authority. The European food Safety Authority reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

231 Teramo, Abruzzo January 2004 December 2007 Bari and Ginosa January February 2005 Impact, modelling and control of canine leishmaniosis in the EU Bologna Dog shelter 2007: : : 245 Bolzano-South Tyrol May-June 2008 Kennelled dogs (housed at night) Mount Versuvius area, Mass serological Naples Province screening for L. infantum Greece Karditsa, Trikala and Fthiotida Ioannina, Arta, Thesprotia and Preveza Data collection period March and April 1999 n = 2123) clinical symptoms) Dog shelter 10.2% (2004 before intervention introducing) (IFAT, cut-off depending on clinical signs and albumin/globulins ratio) Kennelled dogs % (32.2% Bari, 26.1% Ginosa) (Immunochromatographic dipstick test) 2007: 12.6% 2008: 5.6% 2009:4.9% (IFAT, cut-off 1:160) Sampling / study population Asymptomatic hunting dogs Randomly selected asymptomatic dogs and clinically suspect dogs seen at small animal practices (Podaliri Vulpiani et al., 2009) (Otranto et al., 2007) (Baldelli et al., 2011) 40 0% 0% (IFAT, cut-off 1:40) (Morosetti et al., 2009) % (IFAT, cut-off 1:40) (Maroli et al., 2001) Sample size Compatible clinical signs or disease Prevalence / incidence estimate Serology (test and cut-off) 73 Prevalence 14.3% and incidence 17.6% (IFAT, cut-off 1:50) 1550 Asymptoma tic (1200) Symptomati c (350) 24% asymptomatic 45.5% symptomatic dogs were seropositive (IFAT, cut-off 1:60) PCR Prevalence 61.9% and incidence 47.1% (PCR) Reference (Leontides et al., 2002) (Papadopoulou et al., 2005) Supporting publications 2013:EN by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender considered as an output adopted by the Authority. The European food Safety Authority reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

232 Greater Athens Asymptomatic dogs volunteered by owners Greek mainland End of winter Clinically normal dogs admitted to private practices for routine examinations % (IFAT, cut-off 1:100) 2, % (positive to both tests) Ionnina (21.0%) Florina (2.1%) Serres (20.5%) Evros (4.3%) Evritania (17.0%) Evia (4.3%) Attiki (30.1%) IFAT (cut off 1:200) and ELISA (Sideris et al., 1999) (Athanasiou et al., 2012) Supporting publications 2013:EN by the author(s) in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender considered as an output adopted by the Authority. The European food Safety Authority reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document, without prejudice to the rights of the authors.

233 Y. HUMAN LEISHMANIOSIS DATA Table 57: Annual number of reported cases of visceral and cutaneous leishmaniosis in Italy, (Ministero della Salute, 2013c) Form of leishmaniosis Number of cases of leishmaniosis reported per year Visceral Cutaneous Table 58: Monthly cases of human leishmaniosis reported to the Hellenic Center for Disease Control and Prevention, Greece (monthly cases in 2012 and median and range for each month ) (HCDCP, 2013) Month Cases of human leishmaniosis reported per month 2012 (n) median (range) January (1 16) February (2 7) March (2 6) April (2 7) May 2 4 (3 8) June (2 9) July (2 7) August (1 10) September 3 4 (2 6) October 10 3 (1 8) November 1 4 (0 6) December (1 10) Supporting publications 2013:EN

234 Table 59: Geographical distribution of human leishmaniosis cases reported to the Hellenic Center for Disease Control and Prevention, Greece, 2012 (HCDCP, 2013) Cases of human leishmaniosis (n) Region Eastern Macedonia and Thrace 0 Central Macedonia 2 Western Macedonia 0 Epirus 1 Thessalia 7 Ionian islands 3 Western Greece 3 Sterea Greece 2 Attica 15 Peloponnese 5 Northern Aegean 1 Southern Aegean 0 Crete 1 Unknown 1 Total 41 Supporting publications 2013:EN

235 Z. MAPS OF STANDARDISED MORBIDITY RATIOS FOR HOSPITALISED CASES OF HUMAN LEISHMANIOSIS IN SPAIN AND ITALY, Figure 79a Figure 79b Code Region 1 Andalucía 2 Aragón 3 Cantabria 4 Castilla la Mancha 5 Castilla y León 6 Cataluña 7 Ceuta and Melilla 8 Comunidad de Madrid 9 Comunidad Foral de Navarra 10 Comunidad Valenciana 11 Extremadura 12 Galicia 13 Islas Baleares 14 Isles Canarias 15 La Rioja 16 País Vasco 17 Principado de Asturias 18 Región de Murcia Figure 79c Figure 79: Standardised morbidity ratio for NUTS 2 level incidence risk of hospital admissions with a primary diagnosis of leishmaniosis, 2009 (Figure 79a), 2010 (Figure 79b) and 2011 (Figure 79c), Spain. Supporting publications 2013:EN

236 Figure 80a Region code Name of region (capital) 1 Piemonte 2 Valle d'aosta 3 Lombardia Trentino-Alto 4 Adige 5 Veneto 6 Friuli-Venezia Giulia 7 Liguria 8 Emilia-Romagna 9 Toscana 10 Umbria 11 Marche 12 Lazio 13 Abruzzo 14 Molise 15 Campania 16 Puglia 17 Basilicata 18 Calabria 19 Sicilia 20 Sardegna Figure 80b Figure 80c Figure 80: Standardised morbidity ratios for province level incidence rates of hospital admissions with a primary diagnosis of leishmaniosis, 2009 (Figure 80a), 2010 (Figure 80b) and 2011 (Figure 80c), Italy. Supporting publications 2013:EN

237 AA. GOOGLE TRENDS: REGIONAL WEB INTEREST Figure 81a. Figure 81b. Figure 81c. Figure 81: Regional search volume, France : Chien ( Figure 81a.), La rage (rabies) (Figure 81b.) and Leishmaniose (Figure 81c.) BB. Figure 82a. a. Figure 82b. Figure 82c. Figure 82: Web search volume in towns and cities, France : Chien (dog)(figure 82a.), La rage (rabies)(figure 82b.) and Leishmaniose (Figure 82c.) Figure 83a. Figure 83b. Figure 83c. Figure 83: Web search volume in towns and cities, Spain : Perros (Figure 83a.), Rabbia (rabies)(figure 83b.) and Leishmaniasis (Figure 83c.) Supporting publications 2013:EN

238 Figure 84a. Figure 84b. Figure 84c. Figure 84: Regional search volume, Portugal : cão (dog)(figure 84a.), raiva (rabies) (Figure 84b.) and Leishmaniose (Figure 84c.) Figure 85a. Figure 85b. Figure 85c. Figure 85: Web search volume in towns and cities, Portugal : cão (dog)(figure 85a.), raiva (rabies)(figure 85b.) and Leishmaniose (Figure 85c.) Figure 86a. Figure 86b. Figure 86c. Figure 86: Web search volume in towns and cities, Italy : cane (dog)(figure 86a.), rabbia (rabies)(figure 86b.) and Leishmaniosi (Figure 86c.) Supporting publications 2013:EN