Dengue epidemiology in selected endemic countries: factors influencing expansion factors as estimates of underreporting*

Transcription

1 Tropical Medicine and International Health doi:10./tmi.1298 volume 20 no 7 pp july 201 Review Dengue epidemiology in selected endemic countries: factors influencing expansion factors as estimates of underreporting* Nguyen T. Toan 1, Stefania Rossi 2, Gabriella Prisco 3, Nicola Nante and Simonetta Viviani 1 Clinical Research Unit, Pasteur Institute, Ho Chi Minh City, Vietnam 2 Department of Molecular and Developmental Medicine, University of Siena, Siena, Italy 3 Postgraduate School of Public Health, University of Siena, Siena, Italy Department of Molecular and Developmental Medicine, Postgraduate School of Public Health, University of Siena, Siena, Italy Abstract objective Dengue fever is globally considered underestimated. This study provides expansion factors (EFs) for dengue endemic selected countries and highlights critical issues in the use of EFs. methods We identified dengue epidemiological cohort studies from 2000 to July 2013 through a literature search using PubMed, Web of Science and Lilacs (Latin American and Caribbean Health Sciences Database), predefined keywords and inclusion/exclusion criteria, and included Brazil, Colombia, Nicaragua, Peru, Puerto Rico, Venezuela, Bangladesh, Cambodia, India, Indonesia, Philippines, Singapore, Sri Lanka, Thailand and Vietnam. Dengue national and local passive surveillance data were derived from WHO regional websites, PAHO, SEARO and WPRO. EFs were calculated as CI cohort studies/ci passive data for both national and local levels. results Cohort studies differed in case definition, laboratory test used and surveillance methods. The information on SEARO, PAHO and WPRO websites differed in terms of dengue epidemiological variables, population denominators and completeness. The highest incidence was reported by PAHO countries followed by WPRO and SEARO countries. EFs may vary for the different variables and denominators used for calculation. EFs were the highest in SEARO countries and lowest in PAHO countries. A trend for lower local EFs was observed. conclusions The use of EFs for quantifying dengue underreporting may be problematic due to lack of uniformity in reporting dengue both active and passive surveillance data. Quality dengue surveillance data are urgently needed for a better estimate of dengue disease burden and to measure the impact of preventive intervention. keywords dengue, epidemiology, endemic countries, underreporting, expansion factors Introduction Dengue fever (DF) is caused by infection with the dengue virus (DENV), a R virus that occurs as four recognised serotypes: DENV1, DENV2, DENV3 or DENV, which belong to the Flavivirus genus of the virus family Flaviviridae. These viruses are transmitted in humans by mosquitoes (primarily Aedes aegypti). Infection with a DENV can result in a range of symptoms, from subclinical disease to debilitating but transient dengue fever (DF) to lifethreatening dengue hemorrhagic fever (DHF) or dengue shock syndrome (DSS) [1, 2]. *Full free access from The most severe forms of dengue disease DHF and DSS are life threatening, and children with DENV infection are particularly at risk of progressing to severe dengue (DHF/DSS) [2, 3]. There is no specific treatment for dengue. Supportive care includes control of fever and pain with antipyretics/analgesics and adequate fluid intake. In countries with case management programmes in place, the case fatality rate can be reduced to close to zero []. Routine laboratory diagnosis of dengue infections is based on one or more of the following: the detection of DENVspecific antibodies (by plaque reduction neutralisation test, ELISA IgM or IgG); virus isolation; detection of viral R by quantitative reverse transcription polymerase chain reaction (qrtpcr); or detection of viral nonstructural protein (NS1) antigen by enzyme John Wiley & Sons Ltd

2 linked immunosorbent assay (ELISA) [1 3]. The geographical spread of both the mosquito vectors and the viruses has led to a global resurgence of epidemic DF and emergence of DHF in the past 2 years, with the development of hyperendemicity in many urban and periurban centres of the tropics. Dengue is hyperendemic in Asia, the Pacific area and Central and South America (including the Caribbean) [ 7]. However, dengue is considered globally underreported: with 2. billion people are at risk of a dengue infection, at least 0 million people infected, > cases of DHF and approximately deaths annually are likely attributable to dengue (available at Accessed 23 July 201) [3, ]. Routine surveillance performed through notification of dengue cases by healthcare providers and enhanced surveillance, a form of surveillance usually in response to an epidemic alert [2], are in place in most dengue endemic countries to detect dengue fever. In the last decade, enhanced surveillance has been used more generally to increase dengue awareness [8]. However, data reported through this system are considered to underestimate the true DF incidence [3, ]. DF incidence estimates can also be derived from ad hoc epidemiological studies (active surveillance) that over the past 30 years have been conducted in some of the dengue highendemic regions [9 ]. These studies provide the best estimate of the epidemiology of dengue and show how the sensitivity of a surveillance system can be improved by conducting active surveillance [12, 13]. WHO headquarters (HQ) and WHO regional offices [1, 2, 1, 1] issued a number of guidelines and recommendations for reporting dengue fever that are important to consider as dengue is endemic in three WHO regions: WHO SouthEast Asia Regional Office (SEARO), WHO Western Pacific Regional Office (WPRO) and the Pan American Health Organization (PAHO). Although recent estimates for the African region show that dengue is endemic in many parts of the continent [], Africa was not included in this review for scarcity of systematic data collection. The lack of reliable and precise data on dengue incidence renders problematic the true estimate of the disease burden, that is the health, social and economic impact for the population affected by dengue [2, 3]. Recent studies [1 18] gave quantitative estimates of the degree of underreporting (expansion factor, EF) in the Americas and SouthEast Asia, and in 2013, a report with EFs for the Americas and SouthEast Asia was published as the result of a collaboration between Malaysian Universities and the Ministry of Health of Malaysia [19, 20]. Officially reported dengue episodes need to be adjusted for underreported to obtain a more accurate estimate of the number of apparent dengue episodes occurring in a specific population. These adjustment factors, sometimes called expansion factors (EFs), are ideally based on empirical data (e.g. cohort or capture recapture studies), but have been also obtained from expert opinion [21]. The EF is calculated as the ratio of the best estimate of the number of cases and reported number of cases in a specific population in one year, where an EF=1 indicates perfect reporting of dengue and EF>1 represents the degree of underreporting [19]. The aim of this study was twofold: first, to compare dengue epidemiological data reported through the national and local routine surveillance systems and data from ad hoc epidemiological studies performed in dengue endemic selected countries from PAHO, SEARO and WPRO regions and, second, to identify critical factors able to influence the calculation of the EFs as an estimate of the extent of underreporting. Methods Search strategy and selection criteria We performed a literature review using PubMed, Web of Science and Lilacs (Latin American and Caribbean Health Sciences Database) using the research categories Dengue diseases, Active and Passive Surveillance and Selected endemic countries (Brazil, Colombia, Nicaragua, Peru, Puerto Rico, Venezuela, Bangladesh, Cambodia, India, Indonesia, Philippines, Singapore, Sri Lanka, Thailand, Vietnam) with the terms dengue, burden, incidence, cohort, outpatient and active surveillance. The search period was January 2000 to July 2013, and we considered publications in English, French, Spanish and Portuguese. We also reviewed bibliographies of pertinent articles and searched for relevant unpublished studies in the grey literature databases: Systems for Information in Grey Literature (SIGLE), EThOS Electronic theses online service, Inside Conferences and Dissertation Abstracts. We included studies designed as cohort prospective (community based or school based) or designed as capture recapture (community based or school based) and studies having active data of dengue incidence referred from 2000 to We excluded data derived from estimate or modelling studies, studies not reporting case definition or laboratory confirmation, studies recruiting suspected or confirmed cases, entomological surveillance data or health economic data, studies related to clinical findings or viral molecular characterisations, and data not referred to selected endemic countries (Thailand, Vietnam, Singapore, Philippines, India, Sri Lanka, Bangladesh, Indonesia, Venezuela, Nicaragua, Puerto Rico, 201 John Wiley & Sons Ltd 81

3 Colombia, Brazil, Cambodia, Peru). Two investigators (SR and GP) independently reviewed the title and the abstracts of the articles identified using the search strategies described and retained those articles that met the inclusion/exclusion criteria. Disagreements were resolved through discussions with a third investigator (SV). Statelevel passive data reported to WHO Regional Offices, PAHO, SEARO and WPRO, were accessed at the respective websites (available at en. Accessed 2 July 2013) for the same time period. As our aim was to calculate EFs also for the local level, to retrieve local dengue routine surveillance data for each selected country, a number of national organisations website were consulted: MoH/MoHp, National Statistical Office/Dept of Statistics, Bureau of epidemiology, Lardidev and DengueNet (available at dengue/denguenet/en/. Accessed 10 February 201). Expansion factor calculation EFs were calculated as CI cohort studies /CI passive data for both national and local level and for each year. In addition, we calculated lefs to explore factors that can potentially influence EFs such as laboratory confirmation of cases, incidence type [cumulative or incidence density (ID)], cases from urban or rural area, age group of cases and collection of cases during the dengue season or annually. Epidemic years To identify the epidemic years, we calculated the mean of the annual number of dengue cases in the studied period for each given country and then defined up to three epidemic years for a given country as those corresponding to the years which reported the highest peaks of dengue cases above the mean. Results Literature search of cohort surveillance studies (1 January July 2013) All cohort studies identified during the study period and included in this analysis according to the predefined inclusion/exclusion criteria were cohort prospective studies (Table 1) [8, 22 3]. Secondary publications of the main study were also included [37 7]. The studies show some degree of variability with respect to (i) incidence measures; (ii) age of subjects enrolled in the cohorts; (iii) active surveillance strategies; (iv) specific area of the country where incidence of dengue is measured, that is rural, urban, or rural and urban combined; (v) laboratory assays for case confirmation; and (vi) annual or seasonal dengue cases reporting (Table 1). Some studies calculated cumulative incidence (CI) or provided information to calculate it, while others reported the ID with different measure units: personyear, personseason and persondays. Moreover, different denominators of CI were used: cohort size at the end of surveillance season [3], all subjects active in the cohort for at least half of the study year, whether or not an annual sample was taken at the end of the year [33], total population under study, with paired annual samples available [22], cohort of schoolchildren at the end of the study, who gave a complete set of three consecutive samples (one baseline and two followups) [2], total number of people who began the study [28]. With the purpose of obtaining uniformity of incidence measures, for each study, we calculated the CI expressed as the number of confirmed symptomatic cases/population cohort size. For one study in Thailand (Ratchaburi) [32], the original ID reported in the article was used, as no information was provided for CI calculation. Most cohorts were composed of children and/or adolescents, only one included infants and toddlers (2 1 months) [28]. The majority of studies adopted a communitybased or schoolbased surveillance strategy, while enhanced communitybased surveillance was used only in two studies in Nicaragua and Puerto Rico [8, 33]. A study in Indonesia was based on sentinel active community surveillance [9], that is the study of disease rates in a specific cohort, geographic area or population subgroup [2]. In Peru, both schoolbased and communitybased surveillance were used in the same time period for the same population [31], whereas schoolbased and communitybased surveillance were used in Vietnam for the same cohort, but during different seasons [3], i.e. during, respectively, school term (September May) and school vacation (JuneAugust, peak dengue season). The studies employed different strategies in identifying and sampling febrile cases. In schoolbased studies, the identification of the febrile cases corresponded to the days of school absenteeism. For example, in Nicaragua, the resource constraints permitted the study nurse to visit the absent child s house only on the third day of absence [22], while in Thailand, in Ratchaburi, active surveillance for school absence was conducted daily [32]. Otherwise, communitybased studies appeared to be uniform with respect John Wiley & Sons Ltd

4 Table 1 Summary of active dengue surveillance: prospective cohort and capture recapture studies PAHO countries Colombia Nicaragua Active surveillance for febrile detection in prospective cohort (Study population, age) School based ( 19 years old) Dengue season cases urban area: Students from primary and secondary schools in Medellin (three different sites: districts of San Javier, Poblado and Laureles); cases were identified by school absenteeism due to a febrile episode of less than 7 days of duration A study nurse visited each school every days. If the schoolchild met the inclusion criteria, a study physician performed a complete medical examination A clinically suspected dengue episode with virological or serological confirmation. Specific IgM antibodies detected by DEN IgM capture ELISA. Detection of dengue virus R by RT PCR School based ( 1 years old) Annual cases urban area: Daily monitoring school absenteeism Monitoring fever at school by study nurse at third day Selfapproach by subject when they have got fever Children with fever and denguelike symptoms were identified as suspected cases of dengue. Acute (first day of identification) and convalescent phase serum samples were collected. Reverse transcriptase polymerase chain reaction (RTPCR) and virus isolation were performed on acute phase samples Communitybased enhanced surveillance (2 9 years old) Annual cases* urban area: Household visit by study team, using GPS device Febrile children approach health centre/ hospital Dengue detection and classification into four categories (A, B, C, D) Confirmed case by IgM/IgG MACELISA (ME), VI, RTPCR, HI Seroconversion test: IgM/IgG ME Confirmed case by IgM/IgG MACELISA (ME), VI, RTPCR, HI Seroconversion test: IgM/IgG ME 9,7,9, *Note: Annual incidence and expansion factors were calculated in relation to the annual dengue season, beginning in July of each year. Cumulative incidence CI* (Incidence density ID) May December 2010: CI =.9/10 3 (1/2379) June December 20: CI =.9/10 3 (9/180) May 2001 May 2002: CI Symptomatic= 8./10 3 (/7) CI Total = 8/10 3 (/7) May 2002 May 2003: CI Symptomatic =8.3/10 3 (/719) CI Total 7/10 3 (1/719) August 200 June 200: CI symptomatic =./10 3 (17/3713) CI total = 8/10 3 (319/3713) August 200 June 200: CI symptomatic = 17./10 3 (/398) CI total = 1/10 3 (09/389) August 200 June 2007: CI Symptomatic = 3./10 3 (13/33) CI total = 8/10 3 (207/33) August 2007 June 2008: CI Symptomatic = 17./10 3 (/37) CI total = 73/10 3 (28/37) 201 John Wiley & Sons Ltd 83

5 Table 1 (Continued) PAHO countries Peru Puerto Rico Active surveillance for febrile detection in prospective cohort (Study population, age) Schoolbased surveillance ( to 17yearold children and their adult family members, all age) Annual cases urban area: Monitoring school absenteeism in school term and weekly home visit (from one to three times per week) in school holiday Confirmed case by virus isolation, RTPCR, IgM/IgG MACELISA (ME). Seroconversion: PRNT Communitybased surveillance (All age) Annual cases urban area: Doortodoor febrile illness surveillance programme. Household visit three times per week to detect febrile illness Hospitalisation in suspected dengue case Two blood samples (acute and convalescent) Active dengue cases were identified by viral isolation, IgM serology (elevated IgM antibody titres > 1:00) in the acute sample, convalescent sample or both), or a fourfold rise in IgG antibody titres between acute and convalescent samples (ELISA) Communitybased, enhanced surveillance (All age) Annual cases urban area: Active case finding by support healthcare provider s identification and reporting of symptomatic dengue cases among residents Confirmed case by RTPCR, IgM MACELISA (ME). Seroconversion by MN or quantitative IgG ELISA in case PCR (+) or IgM seroconversion A prospective followup study in a cohort of schoolchildren ( 13 years old), who gave a complete set of three consecutive samples (one baseline and two followups) Dengue Season cases urban area: We assumed that the cohort was representative of the population at risk for contracting dengue, which state surveillance data showed to be 13 yearold children and adolescents (LARDIDEV, unpublished observations) Consecutive samples were assayed by PRNT Cumulative incidence CI* (Incidence density ID) January 2000 August 200: CI = 2.1/10 3 (annul mean data) April 200 December 200: School based ( 17 years) CI = 9.7/10 3 (/3) ID = 12.9/10 3 personyears Community based ( 17 years) CI = 1.3/10 3 (2/137) ID = 23./10 3 person years Community (0 98 years) CI =./10 3 (/80) ID = 17.1/10 3 personyears June 200 May 200: CI = 7.7/10 3 (1/20 12) May November 2002: CI = 2.8/10 3 (183/710) May November 2003: CI = 1.9/10 3 (120/710) John Wiley & Sons Ltd

6 Table 1 (Continued) PAHO countries Venezuela Active surveillance for febrile detection in prospective cohort (Study population, age) Community based ( 9 years old) In this study, active surveillance was incorporated as a part of the prospective study design. The participants houses were visited three times a week A biannual blood sample was taken for each study participants, to establish the prevalence and month incidence of dengue infection Dengue season cases urban area: Confirmed case by IgM/IgG MACELISA (ME), PRNT, RTPCR Cumulative incidence CI* (Incidence density ID) September December 2007: CI = 18.7/10 3 (7/209) SEARO countries Indonesia Thailand Active surveillance for febrile detection in prospective cohort (Study population, age) Sentinel active community surveillance (18 years old) Annual urban area: Factory personal office notifies the event (failed to show up for work) Physician evaluation Mobile team to contact 2 h when absenteeism Confirmed case by IgM/IgG MAC ELISA (ME), RTPCR, VI; Seroconversion by HI and PRNT Prospective, multicentre, active fever surveillance, cohort study (community, schools, health centres and/or private health clinics, depending on each study site s setting). Children 2 1 years old Annual cases rural and urban areas: Laboratoryconfirmed dengue: NS1 positive. Probable dengue: IgM positive and/ or fourfold rise in IgG Prospective, multicentre, active fever surveillance, cohort study (community, schools, health centres and/or private health clinics, depending on each study site s setting. Children 2 1 years old Annual cases rural and urban areas: Laboratoryconfirmed dengue: NS1 positive Probable dengue: IgM positive and/ or fourfold rise in IgG Cumulative incidence CI* (Incidence density ID) : ID Symptomatic = 18/10 3 personyears ID Asymptomatic = /10 3 personyears CI = 37.9/10 3 (90/237) June 2010 July 20: CI = 3.9/10 3 (1/) June 2010 July 20: CI = 20.1/10 3 (/299) 201 John Wiley & Sons Ltd 8

7 Table 1 (Continued) SEARO countries Thailand Kamphaeng Phet Thailand Ratchaburi WPRO countries Cambodia, Kampong Cham Active surveillance for febrile detection in prospective cohort (Study population, age) Schoolbased surveillance from Jun to Nov ( 1 years old) Dengue season cases urban area: School absenteeism report and then home visit by village health worker (four persons in mobile team) Confirmed case by IgM/IgG MAC ELISA (ME), RTPCR, VI, HI Schoolbased surveillance (3 1 years old) Annual cases rural area: Selfreporting febrile episode at home and school absenteeism report. Project field coordinator visit household, two times per week in vacation Confirmed case by IgM/IgG MAC ELISA (ME), RTPCR, VI Active surveillance for febrile detection in prospective cohort (Study population, age) Active communitybased surveillance (0 19 years old) Mainly during the rainy season rural and urban areas combined: Fever was detected by notification of trained mother or weekly visit of village team. Village team inform to investigation team Investigation teams visit household and check, take blood samples (acute and convalescent) if dengue is suspected Confirmed case by IgM/IgG, and then, RTPCR was used when IgM (+) and VI was used when RTPCR (+) Cumulative incidence CI* (Incidence density ID) 2000 June November: CI = Total 21./10 3 (37/1713) CI = Sympt 7./10 3 (13/1737) CI = Inapp. 1/10 3 (2/1713) 200: CI = Total 9.3/10 3 (120/2023) CI = Sympt 1.3/10 3 (33/2023) CI = Inapp. 0/10 3 (81/2023) 200: CI = Total 2./10 3 (10/2021) CI = Sympt 13./10 3 (27/2021) CI = Inapp. 38.1/10 3 (77/2021) 200: CI = Total 98./10 3 (201/2039) CI = Sympt.1/10 3 (90/2,039) CI = Inapp. 0./10 3 (103/2039) 2007: CI = Total 3.8/10 3 (128/2007) CI = Sympt 19./10 3 (39/2007) CI = Inapp. 2./10 3 (8/2007) 200 ID = 17.7/10 3 personyears 2007: ID = 3.8/10 3 personyears 2008: ID = 7./10 3 personyears 2009: ID = 32.9/10 3 personyears Cumulative incidence CI* (Incidence density ID) 200 May November: CI = 13./10 3 (89/7) ID = 13./10 3 personseason 2007 June December: CI = 2./10 3 (30/10 08) ID = 7.8/10 3 personseason 2008 April December: CI = 1.2/10 3 (7/773) ID = 17./103 personseason John Wiley & Sons Ltd

8 Table 1 (Continued) WPRO countries Philippines Vietnam Long Xuyen Active surveillance for febrile detection in prospective cohort (Study population, age) Prospective communitybased study (infants 2 1 months) Annual cases semiurban area: During the rainy season (June November 2007), mothers were encouraged to bring their infants to the San Pablo City Health Office for the evaluation of outpatient febrile illnesses. Acuteand convalescentphase (day 1) blood samples were obtained from study infants with febrile illnesses that did not have an obvious source at time of presentation. Clinical data and blood samples were collected, and surveillance was performed for symptomatic and inapparent DENV infections. Confirmed case by IgM/IgG ELISA, RTPCR Prospective, multicentre, active fever surveillance, cohort study (community, schools, health centres and/or private health clinics, depending on each study site s setting). (Children 2 1 year) Annual cases rural and urban areas: Laboratoryconfirmed dengue: NS1 positive Probable dengue: IgM positive and/or fourfold rise in IgG Schoolbased and communitybased surveillance, (2 1 years old) Annual cases Urban areas: In school term (September May): Daily absenteeism report at school Household visit by collaborators Hospital or mobile team s examination Two blood samples (acute and convalescent) In school vacation (June August, peak dengue season): Same procedure but household visit three times per week instead of daily absenteeism report Confirmed case by IgM/IgG MacElisa (ME), VI. Retrospective samples were tested by NS1, realtime RTPCR. Seroconversion by IgM/IgG MACELISA (ME) Cumulative incidence CI* (Incidence density ID) January 2007 January 2008: CI = 9.0/10 3 (0/1) 2007: ID = Total: 9/10 3 personys ID Inapparent = 103/10 3 personys ID Apparent = 1/10 3 personys ID Hospitalised ID Outpatients = 1/10 3 personys June 2010 July 20: CI = : 20.2/10 3 (/297) Case definition A: serological or virological confirmation: (i) positive dengue virus isolation test result or (ii) IgM antidengue antibodies detected in acute or convalescent serum or (iii) an increase in antidengue IgG titre of at least fourfold between acute and convalescent sera 200: CI = 33.3/10 3 (73/219) 200: CI = 1.7/10 3 (1/3239) 200: CI = 2./10 3 (77/31) 2007: CI = 3./10 3 (109/3081) Case definition B: virological confirmation: positive test result by either virus isolation or NS1 antigen or qrtpcr assays 200: CI = 2./ : CI =.7/ : CI = 2.7/ : CI = 37.0/ John Wiley & Sons Ltd 87

9 Table 1 (Continued) WPRO countries Active surveillance for febrile detection in prospective cohort (Study population, age) Prospective, multicentre, active fever surveillance, cohort study (community, schools, health centres and/or private health clinics, depending on each study site s setting). Children 2 1 years Annual cases rural and urban areas: Laboratoryconfirmed dengue: NS1 positive Probable dengue: IgM positive and/or fourfold rise in IgG Cumulative incidence CI* (Incidence density ID) June 2010 July 20: CI = 20./10 3 (3/1) *Calculated by author when data available from article: CI = confirmed symptomatic cases/cohort size (otherwise, CI Total = asymptomatic and/or inapparent + symptomatic cases/cohort size). ID were taken from the article, if published, whenever data for CI calculation were not available. Local passive surveillance data available from article or reference of the article. Specific study design features: Venezuela (A prospective followup study in a cohort of schoolchildren, 13 years old) and Philippines (A prospective nested case control study of DENV infections during infancy). to febrile case surveillance. Most of the cohort studies were conducted in urban areas, with only few exceptions in rural [32, ] and rural urban combined [2, 2] areas. In almost all studies, the primary endpoint was fever with laboratory confirmation for dengue. Laboratory confirmation was reached in the vast majority of studies by reverse trascriptase polymerase chain reaction (RTPCR) with additional ELISA for IgM and IgG and plaque reduction neutralisation test (PRNT). One study [2] used the detection of nonstructural protein (NS1) by ELISA [2], and another performed in Venezuela was designed as a prospective followup study with seroconversion to PRNT as primary endpoint [2]. In Appendix 1, dengue serotype distribution detected by active surveillance studies is reported. Almost all four serotypes circulated in the same year in the study locations, and the proportion of isolates by serotype is consistently reported. In some studies, data were collected in the dengue epidemiological season [2 28, 30, 3, 3, 7] as in Thailand, where active case surveillance of study participants was limited to the dengue season from June to November each year [27, 3]. In other studies, data were collected throughout the calendar year [8, 22, 2, 2, 29, 31 3] (Table 1). For the purpose of calculating EFs, we considered data collected during the dengue season as representative of the calendar year. Incident cases outside the dengue season comprise no more than % of all cases of the year [33, 3]. We identified six previous cohort studies one in the Philippines [2], four in the Americas (Colombia [30], Peru [31], Venezuela [2, 28]) and one in multisite locations in five Asian countries (Indonesia, Philippines, Thailand, Malaysia and Vietnam) [2]. In the period considered, we did not find any publication reporting of active surveillance data for India, Sri Lanka, Bangladesh and Singapore. For Brazil, we found a cohort study [8] that did not meet our selection criteria. Another multisite active surveillance study has completed recruitment in Brazil in Results were not published at the time of this review (available at NCT Accessed 10 September 2013). Dengue cases on PAHO, SEARO and WPRO websites Dengue cases reported by selected endemic countries to the respective WHO Regional Offices during the study period are published in PAHO, SEARO and WPRO websites (available at en/. Accessed 2 July 2013) and presented in Table 2. On the 2th of July 2013, PAHO data were updated at 13th of July 2013, SEARO at 31st of December 2012 and WPRO at 31st of December 20. PAHO and WPRO countries reported incidence data on denominator 10 that we transformed in 10 3 as more convenient for comparison. As SEARO countries reported only absolute numbers, for each country and each year, we calculated the incidence by 10 3 using absolute number of dengue cases as in SEARO website and respective country population as reported in (Accessed 31 July 2013) ( ) and SEARO country population ( ). SEARO selected countries did not report laboratoryconfirmed data in the period considered; PAHO countries did since 2003, and no WPRO countries reported laboratoryconfirmed data except Cambodia in 2009 (Table 2) John Wiley & Sons Ltd

10 Table 2 Reported cumulative incidence of dengue fever and dengue haemorrhagic fever in selected countries by national surveillance transformed in 10 3 (all age, except Cambodia 0 1 years) PAHO Countries Brazil 1. () 2. (). () 2 () 0.7 (0.2) 1.2 () 2 () 3 (1.1).3 () 2.8 (1.3).2 (2.).0 (0.0) 3.0 (0.).0 (0.) Colombia 0. () 2.7 () 2.1 () 2. (0.3) 1. (0.3) 1. (0.) 1.8 (0.) 2.1 (0.) 1.2 (0.1) 2.2 (1.2).9 (3.3) 1. (0.) 2.2 (0.2) 2.9 (1.0) Nicaragua 1. () 0. () 0. () (0.) 0.2 (0.1) 0.3 () 0.3 () 0.3 (0.2) (0.3) 3.3 (0.) 1.2 (0.2) 2.3 (0.2).9 (1.1) 1.7 (0.2) Peru 0.2 () 0.9 () 0.3 () 0.1 (0.0) 0. () 0.2 (0.0) 0.1 (0.0) 0.1 (0.1) 0. (0.1) 0.3 (0.2) 0. (0.) 1 (0.3) 1 (0.) 0.3 (0.2) Puerto Rico 0. () 1.3 () 0.7 () 0.9 (0.3) 0.8 (0.2) 1. () 0.8 () 2.8 (0.8) 0.9 (0.2) 1.7 (0.). (2.) 1. (0.) 3. (1.) 1.9 (0.9) Venezuela 0.9 () 3. () 1. () 1.1 () 1.2 () 1.7 () 1. (0.2) 2.9 (1.8) 1.7 (0.) 2.3 (0.2). () 1.1 () 0. () 0. () SEARO Countries 2000* 2001* Bangladesh 0.0 () 0.0 () 0.0 () 0.0 () 0.0 () 0.0 () 0.0 () 0.0 () 0.0 () 0.0 () 0.0 () 0.0 () 0.0 () India 0.0 () 0.0 () 0.0 () 0.0 () 0.0 () 0.0 () 0.0 () 0.0 () 0.0 () 0.0 () 0.0 () 0.0 () 0.0 () Indonesia 0.1 () 0.2 () 0.2 () 0.2 () 0. () 0. () 0. () 0.7 () 0.7 () 0.7 () 0. () 0.2 () 0.3 () Sri Lanka 0.2 () 0.2 () 0. () 0.2 () 0.8 () 0.3 () 0. () 0. () 0.3 () 1.7 () 1. () 1.3 () 2.1 () Thailand 0.3 () 2.3 () 1.8 () 1.0 () 0. () 0.7 () 0. () 1.0 () 1. () 0. () 1.8 () 1.0 () 1.2 () WPRO Countries Cambodia 0.2 () 0.8 () 0.9 () 0.9 () 0.8 () 0.7 () 1.2 () 2.8 () 0.7 () 0.8 (0.0) 0.8 () 1.2 () Philippines 1.1 () 0.3 () 0.2 () 0. () 0.3 () 0. () 0. () 0. () 0. () 0. () 1. () 1.3 () Singapore 0.2 () 0.7 () 1.7 () 1. () 2.7 ().0 () 0.9 () 2. () 1. () 0.9 () 1.1 () 1.0 () Vietnam 0.3 () 0. () 0. () 0. () 1.0 () 0.7 () 0.8 () 1.2 () 1.1 () 1.2 () 1. () 0.8 () ( ) Laboratoryconfirmed incidence. *Incidence calculated using the number of dengue cases reported by SEARO and the population of Puerto Rico, Thailand, Ratchaburi and Cambodia = Hospitalised Cases. Incidence calculated using the number of dengue cases reported by SEARO and the population reported by countries statistics WHO. By regional analysis epidemiologic update on the dengue situation in the Western Pacific region, 20 (Arima). 201 John Wiley & Sons Ltd 89

11 Brazil Colombia Nicaragua Perù Puerto Rico Venezuela PAHO ( ) Passive cumulative incidence (CI/10 3 population) SEARO ( ) Bangladesh/ India Indonesia Sri Lanka Thailand WPRO (200020) Cambodia Philippines Singapore Vietnam Figure 1 Dengue Incidence in selected endemic countries by WHO Region John Wiley & Sons Ltd

12 Dengue cases were collected from different health institutions. For some countries such as Thailand, Cambodia, Philippines, Nicaragua and Puerto Rico [12], dengue cases were collected exclusively from hospital surveillance, whereas for the majority of countries, dengue cases were collected from both outpatient clinics and hospitals. All PAHO countries reported serotypes distribution by year. WPRO countries report serotype distribution since 200, whereas the information on serotype distribution is not available for SEARO countries (Table A1, Appendix 1). Overall, we observed that all four dengue serotypes circulated in the same year in each country. While active surveillance ad hoc studies consistently reported the proportion of isolated serotypes, only Vietnam reported the proportion of isolates by serotype in and 20 as part of the passive surveillance system. The highest incidence of dengue was observed in PAHO s countries consistently across the period under study with three peaks identified in 2001, 2007 and Venezuela reported the highest incidence and Brazil the lowest. Within the WPRO s countries, Vietnam and Singapore reported the highest incidence with peaks in 200 and Within SEARO s countries, Thailand and Sri Lanka were the highest reporting countries with an appreciable peak in 2010 and 2012 (Figure 1). During the study period, Indonesia reported incidence between 0.1 and 0.7/10 3. Bangladesh and India reported so few dengue cases in relation to population size that the incidence is close to zero (Table 2). Dengue cases reported by regional or provincial level to national routine surveillance As the purpose of this study was also to estimate the local EFs by comparing active data with passive data PAHO SEARO WPRO Colombia Nicaragua Peru Puerto Rico 200 Venezuela Indonesia Thailand K. P Thailand R Thailand 2010 Cambodia Philippines Vietnam nef = CIactive data/ci passive National data) Expansion factor (EF) 82 Epidemic CI Active : Cumulative Incidence = [(n symptomatic not hospitalized laboratory confirmed cases/cohort size)/10 3 ] CI Passive : Cumulative Incidence = [(n symptomatic notified clinical cases/tiol population, all age)/10 3 ] PAHO WPRO SEARO Colombia/Medellin 20 1 Nicaragua/Managua 2001 Nicaragua/Managua Nicaragua/Managua Nicaragua/Managua Nicaragua/Managua Nicaragua/Managua Peru/Dept Loreto (Iquitos) 200 Peru/Dept Loreto (Iquitos) 200 Peru/Dept Loreto (Iquitos) 200 Venezuela/Aragua (Maracay) Venezuela/Aragua (Maracay) Venezuela/Aragua (Maracay) Thai/Ratchaburi Muang District 200 Thai/Ratchaburi Muang District 2007 Thai/Ratchaburi Muang District 2008 Thai/Ratchaburi Muang District 2009 Figure 2 Comparison of Dengue Incidence from active and passive surveillance : nef s vs. lefs Epidemic year Active Schoolbased Surveillance Active Communitybased Enhanced Surveillance Active Communitybased Surveillance Active School and Communitybased Surveillance Cambodia/Kampong Cham 200 Cambodia/Kampong Cham 2007 Cambodia/Kampong Cham 2008 Vietnam/Guang (LongXyyen) 200 Vietnam/Guang (LongXyyen) 200 Vietnam/Guang (LongXyyen) 200 Vietnam/Guang (LongXyyen) lef = CIactive/CI passive Local data) CI Active : Cumulative Incidence = [(n symptomatic not hospitalized laboratory confirmed cases/cohort size)/10 3 ] CI Passive : Cumulative Incidence = [(n symptomatic notified clinical cases/local population, all age)/10 3 ] 201 John Wiley & Sons Ltd 81

13 obtained from both national and local level, we identified passive surveillance data referring to the most representative geographical location of the active surveillance sites in the websites mentioned in the methods and to articles [22, 33 3, 7] where passive data were reported (Table A2, Appendix 2 & Table A3, Appendix 3). Dengue cases or incidence were transformed in CI/10 3 population as mentioned above. Expansion Factors (EFs): comparison between active and passive incidence Figure 2 shows EFs by national (nefs) or local (lefs) level, year and WHO regions. EFs were calculated for those countries and regions/province for which active surveillance data were available from cohort study as well as passive surveillance data for both national and local level. The highest national expansion factors found were nef = 12 for Indonesia in 2001 (Figure 2) followed by Thailand (Ratchaburi) in 2009 with nef = 82, Thailand (Kamphaeng Phet) in 200 with nef = 73, Indonesia in 2010 with nef = 0, Nicaragua in 200 and 2007 with nef = 8 and 9, respectively, and Indonesia in 200 with nef =. The lowest nefs are observed for Colombia in 2010 and 20 with nef = 1 and 3, respectively, Puerto Rico in 200 with nef =, Venezuela in 2007 with nef =, and Cambodia in 200 and Thailand in 2010 with nef =. Overall, the lowest nefs were found for WPRO countries, the highest in SEARO countries. There was a trend for lower lefs than nefs in all countries with the exception of Venezuela (Aragua) in 2002 and 2003, and Colombia (Medellin) in 20 (Figure 2). In locations where both active schoolbased surveillance and active communitybased surveillance were in place at the same time, lef was lowest (Figure 2). Among the factors that can potentially influence EFs, we did not find a substantial difference in lefs if cases who were reported through passive surveillance are laboratoryconfirmed or only clinical diagnosis (Figure A1 in Appendix ), and if incidence measures were differently expressed (Figure A2 Appendix ). lefs are substantially different if cases compared are from rural or urban area (Figure A3 in Appendix ), if their age or age group is different (Figure A in Appendix ) and if they are collected during the dengue season or annually (Figure A in Appendix ). The serological study (active study) performed in Maracay, Venezuela, in 2002 and 2003 showed the highest lefs (Figure 2 and Figure A in Appendix ) [2]. We did not observe any trend in EFs according to the epidemic year (Figure 2). Discussion We found high levels of dengue underreporting in selected endemic countries as estimated by EFs. First, nefs were substantially higher in SEARO countries than PAHO and WPRO countries with an outlier value for Indonesia in Second, lefs were generally lower than nefs with few exceptions (lefs were higher than nefs in Medellin, Colombia in 20 and Maracay, Venezuela in 2003 and 2007) (Figure 2). When an ad hoc active surveillance study is conducted in a given area, the passive reporting system of that area may benefit with more cases reported through the routine surveillance system. Another explanation can be due to the site selection bias in cohort studies. As they are time and resource consuming, they tend to be performed in areas where high DENV transmission is expected to occur. Therefore, on average, incidence of dengue in areas with cohort studies is higher than the incidence at national level. Third, routine surveillance data as published in SE ARO, PAHO and WPRO are expressed differently in terms of numerator, denominator, source of surveillance, serotype reporting and method for laboratory confirmation of dengue cases. Similarly, active surveillance studies used different methods for case detection, case definition, laboratory methods, seasonality definition and serotype reporting. Fourth, we found that several factors influenced lefs estimates, such as the area of surveillance (urban or rural), the age group and the period of data collection (seasonal vs. annual). Our study has some limitations. For Brazil, India and Bangladesh, EFs were not calculated. For Brazil, we did not find any active surveillance data meeting study inclusion criteria, although routine surveillance data were found in the PAHO website for the study period. For India and Bangladesh, also no active data were found for the study period, and very few dengue cases were reported on the SEARO website, although dengue is considered to be endemic in both countries.[2] The exclusion of such a large population exposed to dengue has limited the findings of our study. Another limitation is that we did not further elaborate on circulation of the predominant serotype, as only qualitative data were available from passive surveillance data with the exception of Vietnam. We propose a systematic analysis of EFs by comparing cohort studies with national and local routine data. Because dengue transmission is usually higher in areas where cohort studies are conducted than in the wider geographical area, the nefs may have been overestimated. No clear trend in EFs was identified during epidemic years. A possible explanation is that during epidemic John Wiley & Sons Ltd

14 years, there might be more awareness of dengue and some nondengue febrile illnesses may be reported as dengue, which would increase the total cases reported. However, particularly in regions with lower quality or less access to health care, there is substantial healthsystem congestion during outbreaks, which may reduce the number of reports. EFs for dengue have been reported in other recent studies as an estimate of underreporting based on the annual average for 10 years, [13] with the intent to provide stable values over the period considered. However, it is important to consider that dengue transmission has a marked geographical and temporal heterogeneity [3,, 9, 0] and that the rate of reporting and health quality systems [18] or health systems accessibility [1] are correlated. We showed that not only nef and lef differ (Figure 2), but that EFs are influenced by the yearly epidemiological fluctuation of dengue and by the different parameters used for EFs calculation such as the denominator, the surveillance methods used, the age group and the area where surveillance is performed (Figure 2, Appendix ). All or some of these factors should be considered when estimating EFs as it is important to obtain as much as possible unbiased values. On the other hand, it may be complex to deal with the extreme variability and uncertainty of some estimates. The use of extensive sensitivity analysis should be considered as a possible approach to minimise variation. We found that the existing passive surveillance system is not optimal in providing reliable dengue data. Although we cannot expect that true dengue incidence data are provided through the surveillance system, it would be desirable to have some standardisation (i.e. case definition, disease severity DSS/DHS, serotype) among the WHO regional offices on the data requested to the individual countries. Some monitoring and active solicitations should be in place for overseeing the qualitative and quantitative aspects of reporting. If surveillance is not improved at both country and WHO regional office level, the use of the data generated will not be even helpful for monitoring trends in dengue transmission. Accurate estimates of dengue disease burden are important because they allow informed policy decisions, increase dengue awareness and help define funding and research priorities from different institutions (governments, donors, NGOs, corporations). Cohort studies are probably the most efficient and most popular in providing true dengue disease incidence data. As they are complex studies to implement, and resource and time consuming, the method to be used (case definition area and population selection, laboratory assay, time, duration) must be carefully considered. Seroepidemiological studies can also provide information on dengue disease incidence, although we identified only one [23]. They should be explored further. EFs were developed to measure the economic cost of dengue in countries. The use of this metric at the national or multiple regional levels should be cautious, particularly when comparing cohort studies conducted in highendemic areas. The use of timeaggregated data [17, 18] to calculate EFs is probably acceptable to estimate the cost of dengue disease. As we have shown, EFs estimate varies by year due to a number of factors that make their use questionable in estimating the extent of underreporting at national level. lefs may be more reliable in estimating underreporting at the local or regional level. Reliable dengue incidence data are urgently needed to evaluate interventions that aim to prevent and control dengue epidemics, including new innovative approaches to control mosquitoes [2] and widespread use of dengue vaccines [3, ]. Acknowledgements This work originated by the thesis dissertation of NTT to the MSc course in Vaccinology and Pharmaceutical Clinical Development (Novartis Vaccines Academy and University of Siena A.A ). The authors thank Dr. Sue Ann Costa Clemens, M.D., PhD. Prof. Pediatric Infectious Diseases Director Novartis Vaccines Academy, Dr. Audino Podda Audino, M.D., PhD Head of Clinical Development & Regulatory Affairs at Novartis Vaccines Institute for Global Health, Siena, Italy, and Prof. Emanuele Montomoli BSc, MSc President Master Technical Scientific Committee, Master in Vaccinology and Pharmaceutical Clinical Development, Novartis Academia & University of Siena, Italy. We are grateful to the librarians Cristina Costantini and Roberto Faleri of the Medical Faculty Library, University of Siena, Italy, for their help in the search strategy. References 1. World Health Organization (WHO). Dengue Haemorrhagic Fever: Diagnosis, Treatment, Prevention and Control (2nd edn). WHO:Geneve, World Health Organization (WHO). Dengue: Guidelines for Diagnosis, Treatment, Prevention and Control: New Edition. WHO: Geneva, Simmons CP, Farrar JJ, Nguyen vv, Wills B. Dengue. N Engl J Med 2012: 3: Mayurasakorn S, Suttipun N. The impact of a program for strengthening dengue hemorrhagic fever case management on the clinical outcome of dengue hemorrhagic fever 201 John Wiley & Sons Ltd 83

15 patients. Southeast Asian J Trop Med Public Health 2010: 81: World Health Organization (WHO). Global Strategy for Dengue Prevention and Control WHO: Geneva, Bhatt S, Gething PW, Brady OJ et al. The global distribution and burden of dengue. Nature 2013: 9: World Health Organization (WHO). Global Alert and Response. Impact of Dengue. disease/dengue/impact/en/. 21 August Ramos MM, Arguello DF, Luxemburger C et al. Epidemiological and clinical observations on patients with dengue in Puerto Rico: results from the first year of enhanced surveillance June 200May 200. Am J Trop Med Hyg 2008: 79: Sangkawibha N, Rojanasuphot S, Ahandrik S et al. Risk factors in dengue shock syndrome: a prospective epidemiologic study in Rayong, Thailand. I. The 1980 outbreak. Am J Epidemiol 198;120: Burke DS, Nisalak A, Johnson DE, Scott RM. A prospective study of dengue infections in Bangkok. Am J Trop Med Hyg 1988: 38: Thein S, Aung MM, Shwe TN et al. Risk factors in dengue shock syndrome. Am J Trop Med Hyg 1997: : Beatty ME, Stone A, Fitzsimons DW et al. Best practices in dengue surveillance: a report from the AsiaPacific and Americas Dengue Prevention Boards. PLoS Negl Trop Dis 2010: : e German RR, Lee LM, Horan JM, Milstein RL, Pertowski CA, Waller MN, Updated guidelines for evaluating public health surveillance systems: recommendations from the Guidelines Working Group. MMWR Recommendations and reports: morbidity and mortality weekly report Recommendations and reports. Centers for Disease Control , 1 3; quiz CE WHO Regional Office for SouthEast Asia. Comprehensive Guidelines for Prevention and Control of Dengue and Dengue Haemorrhagic Fever, Revised and expanded edition. WHO Regional Office for SouthEast Asia: New Delhi, India, WHO Pan American Health Organization. Dengue and Dengue Haemorrhagic Fever in the Americas: Guidelines for Prevention and Control. Washington, DC: WHO Pan American Health Organization, Shepard DS, Coudeville L, Halasa YA, Zambrano B, Dayan GH. Economic impact of dengue illness in the Americas. Am J Trop Med Hyg 20a: 8: Undurraga EA, Halasa YA, Shepard DS. Use of expansion factors to estimate the burden of dengue in Southeast Asia: a systematic analysis. PLoS Negl Trop Dis 2013: 7: e Undurraga EA, Halasa Y, Shepard DS. Expansion factors: a key step in estimating dengue burden and cost in Southeast Asia. Am J Trop Med Hyg 20: 8: Shepard DS, Undurraga EA, Lees RS, Halasa Y, Lum LC, Ng CW. Use of multiple data sources to estimate the economic cost of dengue illness in Malaysia. Am J Trop Med Hyg 2012: 87: Shepard D, Lees R, Ng CW, Undurraga EA, Halasa YA & Lum LC. Burden of dengue in Malaysia: Report from Collaboration between Universities and the Ministry of Health of Malaysia: Brandeis University, Schneider Institutes for Health Policy, Heller School for Social Policy and Management, 20b. 21. Shepard DS, Undurraga EA, BetancourtCravioto M et al. Approaches to refining estimates of global burden and economics of dengue. PLoS Negl Trop Dis 201: 8: e Balmaseda A, Hammond SN, Tellez Y et al. High seroprevalence of antibodies against dengue virus in a prospective study of schoolchildren in Managua, Nicaragua. Trop Med Int Health 200: : Balmaseda A, Standish K, Mercado JC et al. Trends in patterns of dengue transmission over years in a pediatric cohort study in Nicaragua. J Infect Dis 2010: 201: Capeding MR, Chua MN, Hadinegoro SR et al. Dengue and other common causes of acute febrile illness in Asia: an active surveillance study in children. PLoS Negl Trop Dis 2013: 7: e Capeding RZ, Brion JD, Caponpon MM et al. The incidence, characteristics, and presentation of dengue virus infections during infancy. Am J Trop Med Hyg 2010: 82: Comach G, Blair PJ, Sierra G et al. Dengue virus infections in a cohort of schoolchildren from Maracay, Venezuela: a 2year prospective study. Vector Borne Zoonotic Dis 2009: 9: Endy TP, Chunsuttiwat S, Nisalak A et al. Epidemiology of inapparent and symptomatic acute dengue virus infection: a prospective study of primary school children in Kamphaeng Phet, Thailand. Am J Epidemiol 2002a: 1: Espino C. Active surveillance and incidence rate of dengue infection in a cohort of high risk population in Maracay, Venezuela. University of South Florida, Scholar Commons Citation. Graduate Theses and Dissertations. [23 September 2013]. 29. Porter KR, Beckett CG, Kosasih H et al. Epidemiology of dengue and dengue hemorrhagic fever in a cohort of adults living in Bandung, West Java, Indonesia. Am J Trop Med Hyg 200: 72: Restrepo BN, Piedrahita LD, Agudelo IY, ParraHenao G, Osorio JE. Frequency and clinical features of dengue infection in a schoolchildren cohort from Medellin, Colombia. J Trop Med 2012: 2012: Rocha C, Morrison AC, Forshey BM et al. Comparison of two active surveillance programs for the detection of clinical dengue cases in Iquitos, Peru. Am J Trop Med Hyg 2009: 80: SabchareonA,SirivichayakulC,LimkittikulKet al.dengue infectioninchildreninratchaburi,thailand:acohortstudy.i. Epidemiologyofsymptomaticacutedengueinfectioninchildren, PLoSNeglTropDis2012::e John Wiley & Sons Ltd