Vaccine 27 (2009) Contents lists available at ScienceDirect. Vaccine. journal homepage:

Transcription

1 Vaccine 27 (2009) Contents lists available at ScienceDirect Vaccine journal homepage: Review Cost-effectiveness of pneumococcal polysaccharide vaccination in adults: A systematic review of conclusions and assumptions Isla Ogilvie a, Antoine El Khoury b,c, Yadong Cui b, Erik Dasbach b, John D. Grabenstein b, Mireille Goetghebeur a, a BioMedCom Consultants inc., 1405 TransCanada Highway, Suite 310, Montreal, Quebec, Canada H9P 2V9 b Merck & Co., PA 19486, USA c University of Arkansas for Medical Sciences, Little Rock, AR, USA article info abstract Article history: Received 4 December 2008 Received in revised form 12 May 2009 Accepted 21 May 2009 Available online 9 June 2009 Keywords: Pneumococcal disease Pneumococcal polysaccharide vaccine Cost-effectiveness Systematic review Streptococcus pneumoniae infections in adults are associated with substantial morbidity, mortality, and costs. A literature review was conducted to identify strengths and limitations of the cost-effectiveness of pneumococcal polysaccharide vaccine studies. A comparative analysis of the impact of model parameters on cost-effectiveness ratios was complemented by systematic assessment of the studies. We identified 11 economic evaluations of pneumococcal polysaccharide vaccine (PPV-23) in adults. In general, all 11 studies found that vaccination with PPV-23 is a cost-effective, and in some cases a cost-saving strategy for the prevention of invasive pneumococcal disease (IPD). The systematic assessment indicated that the results of the cost-effectiveness studies of PPV-23 are influenced by the values applied to vaccine efficacy, IPD incidence and case-fatality Elsevier Ltd. All rights reserved. Contents 1. Introduction Methods Literature review Comparative analysis of economic evaluations Assessment of relevance and strengths of the economic evaluations Results Studies included in analysis Comparative analysis of economic evaluations Impact of indication Impact of the pneumococcal conjugate vaccine (PCV-7) Impact of age and comorbidities Impact of setting Impact of perspective and costs considered Impact of study funding source Impact of other critical parameters Relevance and strengths of the economic evaluations Vaccine-efficacy assumptions Epidemiological assumptions for IPD and pneumococcal pneumonia Revaccination assumptions Discussion Conclusions Acknowledgments Appendix A. Supplementary data References Corresponding author. address: mireille goetghebeur@biomedcom.org (M. Goetghebeur) X/$ see front matter 2009 Elsevier Ltd. All rights reserved. doi: /j.vaccine

2 4892 I. Ogilvie et al. / Vaccine 27 (2009) Introduction Pneumococcal disease (PD) due to S. pneumoniae is a major cause of morbidity and mortality in adults, leading to substantial clinical and economic burden worldwide. S. pneumoniae infections include mucosal non-invasive infections (such as pneumonia mainly community-acquired pneumonia (CAP), otitis media, and other upper respiratory tract infections), and invasive pneumococcal disease (IPD) including bacteremia, septicemia and bacterial meningitis. [1] S. pneumoniae is the underlying pathogen in percent of cases of CAP [2 4] which, as the most common manifestation of pneumococcal infection among adults, constitutes a major clinical and public-health problem. IPD rates are highest among children under one year of age, and among adults over 65 years of age. [5,6] The US-based Active Bacterial Core Surveillance program reported an incidence for IPD of 43.3 per 100,000 children under 1 year of age and 39.6 per 100,000 for the over 65 year old population in [7] The incidence of CAP is also high among the elderly [2 4,8]. A US study reported an incidence of hospitalized CAP of 4.9 per 1000 person-years for year olds and 15.3 cases per 1000 person-years for over 75 year olds [8]. IPD rates reported from the US, [9,10] are higher than those reported in Europe, with those from the UK, Sweden, France and Belgium being low compared with those reported in other European countries [11,12]. Case-fatality rates due to IPD are significant among the elderly population ( 65 years of age): in 2007, rates in the US were 7.6 fatalities per 100,000 [7,13]. While the elderly only represented one-third of the IPD cases, they accounted for 48.9 percent of all deaths due to IPD [7]. Multivariate analyses of risk in patients with IPD shows that in addition to older age ( 65 years) (OR: 2.5 [95%CI: ] P =.004), underlying chronic disease (OR: 2.0 [95%CI: ] P =.025), immunosuppression (OR: 1.9 [95%CI: ] P =.035), and severity of disease (OR: 16.1 [95%CI: ] P <.001) are significantly associated with case-fatality for invasive serotypes; similar results were reported for other serotypes [14]. IPD poses significant demands on healthcare resources because it generally leads to hospitalization; one Canadian study estimated that the economic burden for all PD was 2004 US$500,000 per 100,000 persons of the Canadian population [15]. IPD represented 17 percent of these costs and non-invasive disease 83 percent; hospitalized pneumococcal pneumonia constituted the largest fraction (78%) of the costs among adults over 65 years of age [15]. Increasing antibiotic resistance of S. pneumoniae, in the 1960s and 70s, led to the development of the current 23-valent pneumococcal polysaccharide vaccine (PPV-23) [16]. PPV-23 is indicated primarily for the immunization of adults, and children with chronic diseases [17]. Pneumococcal conjugate vaccines were subsequently developed, which showed efficacy in infants [17]. Currently new vaccines, both conjugate and common protein vaccines against S. pneumoniae, are in development but are not yet licensed for use in adults. The evidence for PPV-23 effectiveness in the elderly and atrisk groups has been interpreted variously [18 25]. The most recent Cochrane Review concludes that PPV-23 is protective against IPD in adults, with limited evidence for efficacy in adults with chronic illness, while evidence of a protective effect against all-cause pneumonia is inconclusive [26]. Many economic analyses of PPV-23 have been performed in different countries and with different methodologies. Previous reviews of the cost-effectiveness of PPV-23 reported that vaccination was cost-effective [27,28]. However, to our knowledge, no systematic review of the literature on the cost-effectiveness of PPV-23 has been done. The objective of this study was to conduct a structured and systematic review of the literature on the economic evaluations of pneumococcal polysaccharide vaccination in adults. 2. Methods 2.1. Literature review An extensive literature search was conducted to identify articles ( ) pertaining to the cost-effectiveness of the PPV-23 vaccine against PD in the adult population. Sources included PubMed, EMBASE, Cochrane reviews, AHRQ Evidence, and the Center for Reviews and Dissemination databases. The search was limited to human subjects and to English language articles. The following search terms were used: cost; economic; cost-effectiveness; cost utility; cost benefit; and pneumococcal polysaccharide vaccine. Bibliographies of retrieved articles were screened as were related articles from Medline. All regions of the world were explored Comparative analysis of economic evaluations The impact of setting, design and model parameters on costeffectiveness ratios was explored in a comparative analysis of the economic studies identified. The following data was systematically extracted and evaluated: type of economic analysis; time horizon; perspective; setting; study cohort; base-year currency; costs (including vaccination costs; hospitalization costs; medication costs; physician visits, and indirect costs); discount rate; data sources for incidence and mortality of PD; vaccine-efficacy assumptions; cost-effectiveness ratios, sensitivity analysis results; and funding sources. All costs were standardized to 2007 US$ using the consumer price index and exchange rates for Assessment of relevance and strengths of the economic evaluations The relevance and strengths of each study were assessed using a semi-quantitative instrument designed to explore economic evaluations; part of the Evidence and Value: Impact on DEcision- Making (EVIDEM) framework [29]. This instrument Relevance and Validity-Economic Evaluation was derived from a synthesis of scientific standards and international guidelines for economic evaluations and instruments assessing their quality [30 38]. It allows analysis and reporting of the relevance and strengths of an economic evaluation in a structured fashion, prompting systematic consideration of 11 dimensions including: target population, intervention and setting, comparator, perspective and costs, outcome measures, parameter estimates and sources, time horizon, discount rate, event pathway/model, sensitivity analysis, and conclusions. Due to the subjective nature of these types of analyses [39], transparent reporting for each dimension was combined with a three-step deliberative process to reach a consensus on potential limitations in relevance and strength of studies. Firstly, an evaluator analyzed each study and provided comments for each dimension. All evaluations were then reviewed by a second investigator and comments were validated by an independent expert. Although the EVIDEM instrument can be used to assign a score for each study we focused our analysis on the relevance, strength and limitations of the assumptions and conclusions of the studies. 3. Results 3.1. Studies included in analysis A literature search was conducted to identify cost-effectiveness studies pertaining to the use of PPV-23 for the prevention of PD in adults. While the initial search recovered 31 papers, eighteen were rejected because cost or effectiveness data for PPV-23 vaccination was not included. Of thirteen cost-effectiveness studies remaining, two were excluded. One study assumed that patients received

3 I. Ogilvie et al. / Vaccine 27 (2009) Fig. 1. Literature search and study inclusion. pneumococcal vaccine with another intervention [40], the second only reported per-capita annual costs of hospitalization with CAP [41]. Eleven studies comparing PPV-23 vaccination with no vaccination were included in this analysis (Fig. 1) [9 12,42 48] Comparative analysis of economic evaluations Cost-effectiveness ratios from the studies were compared by disease; age group; country; and study perspective (societal or healthcare) (Table 1). The majority reported that vaccination was cost-effective (less than $50,000 per life-year gained [LYG] or per quality-adjusted life-year [QALY] gained) [49] for the prevention of IPD in adults. Comparative analysis across studies revealed the impact of several factors on cost-effectiveness ratios Impact of indication Ten of the studies evaluated the cost-effectiveness of PPV-23 for the prevention of IPD only in the base-case analysis. Most of the studies used incidence rates for overall IPD including meningitis, bacteremic septicemia, and bacteremic pneumonia (Table 1). Costeffectiveness ratios ranged from $9810 to $26,160 per LYG and from $9.08 (cost-saving) to $53,955 per QALY (Table 1). A single Italian study considered meningitis and bacteremic pneumonia only, reporting a higher cost-effectiveness ratio of $34,375 per LYG [42]. One study assumed that PPV-23 is effective against all PD including (bacteremic and non-bacteremic) pneumonia and meningitis, and two others provided secondary analyses including nonbacteremic pneumonia or all pneumococcal pneumonia (Table 1) [12,43,47]. For these studies cost-effectiveness ranged from $221 per LYG (cost-saving) to $30.4 per LYG and from cost-saving to $3379 per QALY [43,47] Impact of the pneumococcal conjugate vaccine (PCV-7) The introduction of the 7-valent pneumococcal conjugate vaccine (PCV-7) has been associated with a marked decrease in disease incidence caused by PCV-7 serotypes in the elderly population as well as among infants [50]. Only one study used incidence data from the post-pcv-7 era ( ) [44]. This US study reported that the cost-effectiveness of vaccinating over 65 year olds with PPV-23 was $ per QALY [44] Impact of age and comorbidities Most of the studies investigated the cost-effectiveness of PPV- 23 vaccination against IPD in elderly populations ( 65 years old). In general, vaccination was cost-effective for those over 65 years old [10 12,42,44,46 48], and cost-saving in one US study (Fig. 2) [10]. Base-case values ranged from $9810 to $34,375 per LYG [42,46], and from $9.08 (cost-saving) [10] to $53,955 per QALY [12]. For those over 85 years old, incremental cost-effectiveness ratio (ICER) estimates increased by twofold in two studies ($36, ,396 per QALY) [11,12] compared to the base-case, but remained similar to base-case in a third study (cost-saving: $19.6 per QALY) [10]. One US study addressed the vaccination of year olds and examined the effect of race on cost-effectiveness. Values ranged from $5 per QALY (cost-saving) for the black population to $5691 per QALY for non-blacks, due to higher risk of IPD among African-Americans [9]. A second US study reported an ICER for vaccination at aged 50 years of $35, per QALY; when followed by revaccination at age 65 years, the ICER was $25, per QALY [44]. Three studies included patients with comorbidities (chronic heart, lung, liver or renal disease, immuno-deficiencies, or diabetes mellitus), who are known to be at a higher risk of PD [51]. Vaccination of all high-risk elderly with PPV-23 in a UK study cost $20,042 per LYG, whereas vaccination of all over 65 year olds including those at high risk produced a lower ICER of $17,984 [48]. In contrast, in a US study, vaccination of all high-risk patients aged years old was cost-saving, while vaccinating the general immunocompetent population, (excluding high-risk patients) produced an ICER of $4492 per QALY [9]. This difference may be due to the younger population of the US study having fewer comorbidities; comorbidities assessed by these two studies were similar [9,48] Impact of setting Two related European multinational studies compared costeffectiveness from country to country [11,12]. Other studies were from Belgium, [43] Italy, [42] the Netherlands, [46] the UK [45,47,48] and the US. [9,10] Cost-effectiveness for the over 65 year old population varied from country to country but was below $50,000 per LYG or per QALY in most cases. Differences in base-case costeffectiveness between countries were due to differences in the economic and epidemiological variables for each country, in particular incidence and mortality data for IPD, and hospitalization costs [11,12] Impact of perspective and costs considered The majority of studies used a healthcare perspective, excluding indirect costs (Table 1). Only one study used a societal perspective including lost productivity, and estimated a base-case ICER of $10,999 per LYG for the vaccination of over 65 year olds [47]. From a healthcare-payer perspective, the base-case ICER increased to $12,871 per LYG indicating that lost productivity had a limited impact on cost-effectiveness [47] Impact of study funding source The source of study funding did not appear to affect the ICERs among the 11 studies examined.

4 Table 1 Overview of economic evaluations of vaccinating adults ( 50 years of age) with the 23-valent pneumococcal polysaccharide vaccine included in the review. Country/Study Design/Funding source Costs included Incidence/mortality data sources Europe Evers, 2007 [11] (Belgium, France, Spain, Scotland, Sweden, Denmark, England/Wales, Germany, Italy, the Netherlands) Ament, 2000 [12] (Belgium, France, Scotland, Spain and Sweden) Belgium De Graeve, 2000 [43] Italy Merito, 2007 [42] The Netherlands Postma, 2001 [46] UK Parsons, 2006 [45] Lifetime cohort model 10 European countries Healthcare perspective 65 years IPD Funding: unrestricted grant from Aventis-Pasteur MSD Lifetime cohort model Healthcare perspective 65 years IPD and Pneumococcal pneumonia Funding: grant from Aventis-Pasteur MSD Lifetime decision tree analysis; Healthcare perspective years; 65 years All pneumococcal disease Funding: none stated 5 year cohort model Healthcare perspective 65 years Meningitis and bacteremic pneumonia Funding: none stated 5-year cohort model Healthcare perspective 65 years IPD Funding: none stated Retrospective estimate of costs Healthcare perspective. All ages; IPD Funding: none stated 1999 Euros for country specific direct medical costs for vaccination (vaccine price; administration) and hospital care (average length of stay; physicians; hospital overhead; medications) Outpatient care and future costs excluded Discount rate 3% 1995 Euros for country specific direct medical costs for vaccination (vaccine price; administration) and hospital care (hospital overhead; physicians; average length of stay; medications) Outpatient care and future costs excluded Discount rate 3% 1995 Euros for direct medical costs hospitalization (average length of stay; hotel/nursing costs; clinical intervention costs; antibacterials) and vaccination costs (vaccine price; adverse events; administration) Discount rate 5% 1995 Euros for total costs to healthcare sector of vaccination campaign (vaccine doses; vaccination and administration costs) initial outpatient care; and hospitalization (average length of stay; room and board; nursing care; radiology; surgery; physicians; medications; laboratory tests; intensive care) Discount rate 3% 1995 Euros for vaccination costs (vaccine price; adverse events; administration) and hospitalization costs (average length of stay; intensive care; medication; excluded physician services) Discount rate 4% 2002 UK pounds for vaccination costs (limited to vaccine purchase) and hospitalization costs includes one outpatient visit (procedures/medicines) Discount rate 3.5% IPD: reports from clinical microbiology laboratories - isolates from sterile site; if not available then from literature Mortality rates: from local reports/literature IPD: reports from clinical microbiology laboratories - isolates from sterile site Pneumococcal pneumonia: 40% of hospitalized cases of all-cause pneumonia (ICD-9-CM codes) Mortality rates: from local reports/literature Pneumococcal pnuemonia: from literature Fedson, 1993 [64] Meningitis: from literature Mortality rates: from literature Bacteremic pneumonia: Estimated from Hospital Information system data for pneumococcal pneumonia (ICD9 codes); estimated % of bacteremic cases from literature Mortality due to bacteremic pneumonia from literature Meningitis: regional surveillance system including mortality: 3 6% of all IPD IPD: ICD9 codes for pneumococcal-related hospitalizations and mortality rates from national registration of hospitalizations IPD: Analyzed medical records for 206 IPD patients admitted to a single hospital including mortality Vaccine effectiveness Shapiro: [53] for 65 year old vaccine effectiveness is 75% in the 1st year declining to 33% in the 6th year Serotypes covered 88% Shapiro: [53] for 65 year old vaccine effectiveness is 75% in the 1st year declining to 33% in the 6th year Serotypes covered 88% 60% for those <65 years; 55% for those 65 years; Based on various studies Serotypes covered 95% Duration of protection: 5 years: Shapiro [53] Shapiro: [53] First 3 years: years: 80% years: 67% 85 years: 46% Year years: 71% years: 53% 85 years: 22% Serotypes covered 96% 64% overall Waning immunity; age taken into account Serotypes covered 80% Duration of protection 5 years; Shapiro [53] Serotypes covered 96% Duration of protection: 6 years; based on data from Butler, 1993 [55] Outcomes Base-case: cost-effectiveness ratios D 9239 to D 23,657 per QALY (1999) Base-case: cost-effectiveness ratios from D 11,000 to D 33,000 per QALY Assuming a common incidence (50 cases/100,000) and mortality rate (20 40%), the cost-effectiveness was <D 12,000 per QALY in all 5 countries (1995) Vaccinating between the ages of 18 and 64 years results in 2 LYG compared to no vaccination, and an additional cost of D 11,800 per LYG Vaccinating elderly people ( 65 years of age) leads to >9 LYG and a saving of D 1250 (1995) Baseline net costs per event averted: D 34,681 (95%CI: D 28,699 to D 42,929) Baseline net costs per LYG: D 23,361 (95%CI: D 16,419 to D 38,297) (2001) Baseline net costs were between D 6000 and D 16,000 per LYG (1995) Cost per LYG: ,609 (2002) 4894 I. Ogilvie et al. / Vaccine 27 (2009)

5 Mangtani, 2005 [47] 10 year cohort model Societal perspective 65 years IPD and non-bacteremic pneumonia Funding: Wellcome Foundation grant 1998/99 UK pounds for vaccination program (vaccine; adverse events; administration); hospitalization (average length of stay; intensive care; medication); indirect costs (production losses to society) Discount rate 6% IPD: Laboratory reports for England and Wales Non-bacteremic pneumonia: From literature Mortality: from literature 50% Duration of protection: 10 years; based on data from Mangtani, 2003 [54] Assuming 50% efficacy against mortality and morbidity from IPD lasting 10 years, with a booster at 5 years, estimated cost per LYG was 5507 (1998/99) Melegaro, 2004 [48] Lifetime cohort model Healthcare perspective 65 years IPD Funding: Medical research council (UK) and European Union grants US Smith, 2008 [44] Lifetime Markov model examined 8 vaccination strategies for PPV-23 including revaccination: Healthcare perspective 50 years IPD Funding: Unrestricted grant from Merck & Co via St Margeret s Foundation Sisk, 2003 [9] Sisk, 1997 [10] Lifetime Markov model: Healthcare perspective years IPD Funding: Unrestricted grant from Pasteur Merieux MSD Lifetime Markov model Healthcare perspective 65 years IPD Funding: Grant from CDC 2000 UK pounds for primary care and hospitalization costs (medication; average nursing costs; costs of bed; diagnosis and treatment costs) and vaccination costs (vaccine price; delivery) Discount rate 3% 2003 US$ for vaccination (vaccine price; administration) and IPD treatment costs Discount rate 3% 1995 US$ for vaccination (vaccine price; administration; adverse events) hospitalization (Medicare payments) Discount rate 3% 1993 US$ for vaccination (vaccine price; administration; adverse events) hospitalization (Medicare payments) Discount rate 3% IPD: from CDSC/RSIL database and Hospital Episode Statistics database Mortality rates: hospital discharge codes IPD: Post-PCV-7 Incidence and case-fatality, CDC active bacterial core (ABC) surveillance data ( ) Incidence of disability due to IPD: used Meningitis incidence from ABC as a proxy IPD: Incidence and mortality CDC active surveillance programs IPD: From regional surveillance programs Mortality rates: from literature From RCTs and case/control studies 65% low risk group 20% high-risk group Duration of protection: [53,65] high risk: 5 years low risk: 6.5 years Vaccine efficacy/effectiveness against IPD was based on a Delphi panel Duration of protection estimated as 7 10 years depending on age group and immuno-competency by the Delphi panel Declining protection over time; based on 6-year age-specific data from Shapiro, 1991 [53] Serotypes covered 88% Declining protection over time; based on 6-year age-specific data from Shapiro, 1991 [53] Serotypes covered 88% Base-case cost-effectiveness: Vaccinating all high-risk elderly gives a cost per LYG of 9477 Base-case cost-effectiveness: Vaccinating all 65+ years old, with or without high-risk conditions, gives a lower cost per LYG of 8504 (2000) Base-case cost-effectiveness; Vaccinating all patients at age 65 gives a cost per QALY of US$3341 Vaccination at 50 years of age and 65 years of age gives a cost per QALY of US$23,120 Repeated vaccination at ages 50-, 60-, 70-, and 80 years of age cost US$66,818 per QALY (2003) Excluding future medical costs, vaccination saved US$4 per QALY for blacks and cost US$4351 per QALY for non-blacks. Including future medical costs, cost per QALY was US$6459 for blacks and US $12,374 for non-blacks (1995) Base-case cost per QALY was cost-saving for all age groups Bact: bacteremia; CAP: community-acquired pneumonia; CDC: Centre for Disease Control; CDSC/RSIL: Communicable Disease Surveillance Centre/Respiratory and Systemic Infection Laboratory; Pneu: Pneumococcal pneumonia; ICD-9-CM: International classification of diseases-ninth edition Clinical modification; ICD-10: International classification of diseases-tenth edition; IPD: invasive pneumococcal disease; LYG: life year gained; N/A: not applicable; PPV: pneumococcal polysaccharide vaccine; QALY: quality adjusted life year; RCT: randomized controlled trial. I. Ogilvie et al. / Vaccine 27 (2009)

6 4896 I. Ogilvie et al. / Vaccine 27 (2009) Fig. 2. Cost-effectiveness of vaccinating adults ( 65 years of age unless otherwise indicated) against IPD with the 23-valent pneumococcal polysaccharide vaccine Costeffectiveness per quality adjusted life year (QALY) Cost-effectiveness per life year gained (LYG): Max: maximum; Min: minimum; ICER: incremental cost-effectiveness ratio Impact of other critical parameters Sensitivity analysis was performed to test key parameters in ten studies. The exception was a retrospective economic analysis which contained few details [45]. Vaccine efficacy was a critical parameter in most studies accounting for the largest variation in cost-effectiveness values in 4 studies [10 12,46] Incidence rates of IPD and pneumococcal pneumonia and case-fatality rates were also a major source of variability in cost-effectiveness results [9 12,42,43,46 48,52]. Other factors affecting cost-effectiveness ratios included vaccine cost, [9,10,52] and vaccination strategy. In two studies, vaccination was assumed to occur in the same physician visit as influenza vaccination. Scenario analyses in which PPV-23 was administered alone increased ICERs up to two-fold [11,12]. Discount rate, hospital admissions, length of stay in hospital, and adverse events generally had less effect on ICERs. The minor impact of discounting may be due to the short study perspective (i.e., 5 10 years model duration) and possibly to the relatively short duration of vaccine protection assumed by all the models Relevance and strengths of the economic evaluations Studies were assessed on 11 dimensions for their relevance to decision-making and strength of design. Details of assessments for each study are reported in the Appendix and a summary provided in Table 2. Most studies used relevant patient populations, outcomes and perspectives. All studies compared PPV-23 vaccination to no vaccination. One study contained little detail and was therefore difficult to assess [45]. All the studies were affected by difficulties estimating vaccine efficacy, IPD incidence and case-fatality rates (Table 2) Vaccine-efficacy assumptions Limitations in the efficacy and effectiveness data for PPV-23 vaccination against IPD affected the strength and relevance of most of the studies. The sources used for vaccine effectiveness are presented in Table 3. Three studies used efficacy data derived from the analysis of many sources [43,47,48]. The remaining studies [9 12,42 44,46] used data from a single large case control study by Shapiro et al., 1991 [53]. While this study provided information about age-related decline in vaccine effectiveness and vaccination duration (Table 3), it is a single observational study and as such may be biased by uncontrolled confounders [53]. Efficacy estimates for vaccine-related types of IPD ranged between 22 and 75 percent. For those studies including high-risk individuals, efficacy estimates were around 20 percent [9,44,48]. Two of the reviewed studies assumed vaccine efficacy against pneumococcal pneumonia to be the same as for IPD, [12,43] and a

7 I. Ogilvie et al. / Vaccine 27 (2009) Table 2 Summary of limitations regarding relevance and strengths of economic evaluations included in the review (for details see Appendix). Study Evers, 2007 [11] (update of Ament, 2000) Ament, 2000 [12] De Greave, 2000 [43] Merito, 2007 [42] Postma, 2001 [46] Parsons, 2006 [45] Mangtani, 2005 [47] Melegaro, 2004 [48] Smith, 2008 [44] Summary of limitations regarding relevance and strength Validity and relevance of clinical studies used in the economic evaluation are limited due to the lack of definitive studies of vaccine efficacy This is a study looking at cost-effectiveness in 10 European countries; a weakness of comparative study is that country specific sources of clinical and cost data differ in validity making comparison difficult between countries. IPD incidence and mortality data for Italy was not available and therefore was estimated based on the other countries; this is questionable as there is a large variation in incidence data between countries Average length of stay (ALOS) in hospital and mortality rates for IPD for Spain, Scotland, The Netherlands, and Belgium were based on those for pneumonia due to a lack of other data It was assumed that vaccination with PPV-23 would take place during the same physician visit as the influenza vaccine; this is questionable Validity and relevance of clinical studies used in the economic evaluation are limited due to the lack of definitive studies of vaccine efficacy The secondary analysis of pneumococcal pneumonia is not in line with effectiveness data and is therefore of questionable validity This is a study looking at cost-effectiveness in five European countries; a weakness of comparative studies is that country specific sources of clinical and cost data differ in validity making comparison difficult between countries Average length of stay (ALOS) in hospital and mortality rates for IPD for Spain, Scotland and Belgium were based on those for pneumonia due to a lack of other data It was assumed that vaccination with PPV-23 would take place during the same physician visit as the influenza vaccine; this is questionable This economic evaluation is problematic due to the extrapolation of the efficacy data for IPD to cover all pneumococcal disease affecting the validity of the conclusions Validity and relevance of clinical studies used in the economic evaluation are limited due to the lack of definitive studies of vaccine efficacy Country specific incidence and mortality data were lacking as was frequency of hospitalization data for either pneumococcal pneumonia or meningitis; data from the international literature for these parameters varies widely Validity and relevance of clinical studies used in the economic evaluation are limited due to the lack of definitive studies of vaccine efficacy A minor limitation is the estimation of incidence and mortality data for bacteremic pneumonia; this was based on hospitalization rates for pneumonia in the Lazio region adjusted by the percentage of bacteremic cases among community-acquired pneumonia of any aetiology Some information was missing for the SA (e.g. ranges) and validity could not be assessed Validity and relevance of clinical studies used in the economic evaluation are limited due to the lack of definitive studies of vaccine efficacy The costing for hospitalization lacked transparency; and the SA were not comprehensive This study is an economic analysis of a cohort of confirmed IPD cases in a single hospital; while the clinical data is direct data, cost data was taken from national averages This study has very limited reporting and is therefore difficult to assess for validity and relevance No sensitivity analysis was done except for vaccine efficacy for which the source of data is unknown The extrapolation of this small study to the UK population is problematic Validity and relevance of clinical studies used in the economic evaluation are limited due to the lack of definitive studies of vaccine efficacy Efficacy/effectiveness data against IPD was taken from a systematic review of available efficacy data by the same authors and is conservative The secondary analysis of pneumococcal pneumonia is not in line with effectiveness data and is therefore of questionable validity The time horizon chosen is questionable (10-year duration based on 1 or 2-vaccinations) and was taken from a single US indirect cohort study in which vaccination efficacy was not shown to decline with time since vaccination; however the population (all ages) used in that analysis is not reflective of the target population; in addition repeat vaccinations are not usually recommended Validity and relevance of clinical studies used in the economic evaluation are limited due to the lack of definitive studies of vaccine efficacy Vaccine efficacy values for the elderly (over 65 years of age) were taken from a meta-analysis of RCTs and case control studies done by the same authors; these values are in line with other publications but do not take into account waning immunity with age in the over 65 population Differences between elderly with or without underlying risk conditions (in terms of incidence rate, responsiveness to the vaccine, duration of protection and background mortality rate) were taken into account The high-risk group in this study comprises patients with immunocompromising conditions as well as conditions that increase the risk of IPD Validity and relevance of clinical studies used in the economic evaluation are limited due to the lack of definitive studies of vaccine efficacy A Delphi panel was used to derive vaccine efficacy/effectiveness based on age appropriate efficacy/effectiveness data for IPD from a single large case control study. Duration of protection was estimated to be between 7- and 10-years depending on age group and immunocompetency by the Delphi panel; this is longer than most other studies Hypothetical no-vaccination IPD incidence, case-fatality and disability rates were derived using vaccination level estimates and vaccine effectiveness levels set by the Delphi panel; this is problematic given the uncertainty surrounding effectiveness estimates in the Shapiro study, incidence rates were post-pcv-7 Incidence of disability due to IPD was based on meningitis incidence data as a proxy; this is questionable Incidence of IPD; age related IPD costs (discharged alive and died) and discount rate do not appear to be covered in the SA although this may be due to poor reporting Revaccination strategies were explored in the model although revaccination is not routinely recommended

8 4898 I. Ogilvie et al. / Vaccine 27 (2009) Table 2 (Continued ) Study Sisk, 1997 [10] Sisk, 2003 [9] Summary of limitations regarding relevance and strength Validity and relevance of clinical studies used in the economic evaluation are limited due to the lack of definitive studies of vaccine efficacy A regression model was used to derive age appropriate efficacy/effectiveness data against IPD based on a single large case control study from the US. This model has a lifetime horizon; after 6 years (duration of protection)) both vaccinated and unvaccinated cohorts face the same risk of death Validity and relevance of clinical studies used in the economic evaluation are limited due to the lack of definitive studies of vaccine efficacy A regression model was used to derive age appropriate efficacy/effectiveness data against IPD based on a single large case control study from the US. This model has a lifetime horizon; after 6 years (duration of protection) both vaccinated and unvaccinated cohorts face the same risk of death third assumed vaccine efficacy to be 37.5 percent against both nonbacteremic pneumonia and IPD [54]. In most studies duration of protection of the vaccine was based on Shapiro et al. (Table 4) which showed that vaccine efficacy decreased with time since vaccination [53]. Most studies assumed duration of protection between 5 and 6.5 years. One study [44] employed a Delphi panel to estimate PPV-23 effectiveness using data from the Shapiro study, [53] and assumed a duration of protection between 7 and 10 years depending on patient age. [44] While this duration of protection is outside the time period studied by Shapiro, it may be consistent with their findings as vaccine efficacy reported in the Shapiro study did not drop to zero after 5 years. [53] Two studies, [45,47] used data from an observational study by Butler et al., [55] that recorded no decline in vaccine efficacy over time since vaccination; they assigned durations of protection of 10 years [45] and 6 years [47] respectively Epidemiological assumptions for IPD and pneumococcal pneumonia Data sources for incidence of IPD and case-fatality due to IPD also affected relevance and study strength. Input values and sources used for IPD and pneumonia incidence are given in Table 4. Most studies considered overall IPD incidence rates. Some studies used IPD incidence data from public-health surveillance programs such as that from the US Centers for Disease Control & Prevention (CDC) [9,10,48]. Other studies used ICD-9 codes for pneumococcal-related hospitalizations, [46] local clinical microbiology laboratory data; [11,12,45,47] and data from published sources (Table 4) [11,43]. There was substantial variation in the input values used for incidence rates of IPD; from 20.5 to 50 cases per 100,000 population for year olds, and from 49.2 to 110 cases per 100,000 for over 85 year olds. This variation in input values accounts for much of the variation in cost-effectiveness ratios between and within studies. In the study by Evers et al. [11] incidence rates of IPD for over 65 year olds varied from 29.5 to 63.9 cases per 100,000 between the 10 countries studied in part due to differences in data sources from different countries. Variation may also be due to differences in the frequency and quality of blood cultures between countries. When sensitivity analyses were performed using an incidence of 50 cases per 100,000 and a mortality rate of 30 percent for all 10 countries differences between countries were markedly reduced. ICERs were between $4862 and $17,389 per QALY compared with $14,099 $36,101 per QALY for the base-case epidemiological values [11]. Only one study was conducted after the introduction of PCV-7. Smith et al. calculated a hypothetical no vaccination with PPV-23 incidence rate for IPD, based on CDC incidence data for IPD from 2003 to 2004 adjusted by vaccination effectiveness and uptake [44]. Incidence rates ranged from 26.2 per 100,000 for year olds to 73.2 per 100,000 for over 85 year olds, values that fall within the range used by the other studies. Three studies included pneumococcal pneumonia or nonbacteremic pneumonia; two studies obtained incidence data from published literature [43,47], while the third estimated the incidence of pneumococcal pneumonia to be equal to 40 percent of hospital discharges for all-cause pneumonia (ICD-9-CM codes) [12]. There was significant variation in input values for pneumococcal pneumonia; between 70 and 640 cases per 100,000 for year olds and between 530 and 2630 cases per 100,000 for over 85 year olds. Case-fatality rates included in studies were obtained from public-health surveillance, [9,10] hospital-registration data, [42,46,48] and international literature [11,12,42,43,47]. Casefatality rates ranged from 4.65 percent to 30 percent for the year old age group [10,44]. Due to a lack of case-fatality data for IPD in certain countries, two related multicountry studies based case-fatality rates for IPD on hospital-mortality data (ICD-9 codes: all-cause pneumonia) [11,12]. Smith et al. calculated case-fatality rates based on CDC mortality data for IPD from 2003 to 2004 adjusted by vaccination-effectiveness and uptake [44] Revaccination assumptions Revaccination with PPV-23 is currently recommended in many European countries, especially for high-risk populations [56]. It is not recommended routinely in the US, except for patients at high risk, and those aged 65 or older who have not been vaccinated within the last 5 years [57]. Revaccination was included in two UK cost-effectiveness studies [47,48], and in the two multicounty European studies [11,12]. Smith et al. examined the cost-effectiveness of a range of revaccination schedules including: no revaccination, vaccination in line with US recommendations, and revaccination every 10 years [44]. 4. Discussion This review identified eleven economic evaluations comparing PPV-23 vaccination with no vaccination. The majority of studies reported that vaccination of over 65 year olds with PPV-23 was a cost-effective (less than US$50,000 per LYG or per QALY), and sometimes cost-saving, strategy for the prevention of IPD. Studies that specifically included pneumococcal pneumonia reported lower ICERs, compared to those limited to IPD. Structured assessment revealed that the relevance and strength of the studies was limited by data available on vaccine effectiveness, and also by data sources for incidence and case-fatality rates of IPD and pneumonia. A 2001 review of economic evaluations for PPV-23, [27] reported a cost-effectiveness range of (1999 US$) $ ,390 per QALY gained for universal vaccination and from $2642 to $10,646 for the vaccination of over 65 year olds [27]. A review of five costeffectiveness studies from Europe and North America also reported that vaccination of the elderly with PPV-23 was cost-effective [28]. This study updates previous reviews and provides a systematic review of the cost-effectiveness of PPV-23, based on a structured analysis of the relevance and strengths of the studies included. In this analysis cost-effectiveness ratios for the vaccination of over 65 year olds ranged from $9810 to $34,375 per LYG, [42,46] and from $9.08 (cost-saving) [10] to $53,955 per QALY. These ICERs fall into the same range as those for other vaccination programs for elderly populations, such as vaccination of those over 65 years of age against influenza ($15,535 per LYG in 2002 in Japan; $980 per QALY in 2000 in Japan) [58,59]. Most of the cost-effectiveness studies used a target population over 65 years of age [60]. This population is at elevated risk

9 Table 3 Effectiveness studies used in the economic evaluations included in the review. Study Population/setting Protective efficacy (%: 95% CI) Studies using data (ref #) Efficacy Duration of protection Shapiro, 1991 [53] Case control study; N = 1054 patients with IPD confirmed by isolation of vaccine serotype S. pneumoniae: 1054 controls were matched by age, underlying illness, and site of hospitalization; Median age: 67.6 years; age range years The aggregate protective efficacy of the vaccine against infections caused by the serotypes included in the vaccine was calculated as a percentage: 1 minus the odds ratio of having been vaccinated times 100 Aggregate protective efficacy of PPV-23 against serotypes in the vaccine in immunocompetent patients by age group and time since vaccination Age (years) Time since vaccination: Protective efficacy (%: 95% CI) <3 years 3 5 years >5 years <55 93 (82 97) 89 (74 96) 85 (62 94) (70 95) 82 (57 93) 75 (38 90) (51 92) 71 (30 88) 58 ( 2 to83) (20 87) 53 ( 15 to 81) 32 ( 67 to 72) ( 31 to 78) 22 ( 90 to 68) 13 ( 174to54) Aggregate protective efficacy of PPV-23 against vaccine serotypes Population Protective efficacy (%: 95% CI) All 56% (42% 67%) Immunocompetent 61% (47% 72%) Immunocompromised/high 21% ( 55 to 60%) risk * [9 12,42 44,46] [9 12,42 44,66] Study Population/setting Efficacy (%: 95% CI) Duration of protection (%: 95% CI) Data used by Butler, 1993 [55] US Indirect cohort study; N = 2837 persons with IPD confirmed by isolation of S. pneumoniae; Median age 57 years for vaccinated; 50 years for unvaccinated) Vaccine efficacy estimated by comparing distribution of pneumococcal serotypes that caused infection in unvaccinated and vaccinated subjects Efficacy for immunocompetent patients stratified by age Efficacy Duration of protection Vaccine efficacy and time since [43] [45,47] vaccination Age (years) Protective efficacy (%: 95% CI) Interval (years) Protective efficacy (%: 95% CI) Overall 60% (30 77%) <2 51% (22% 69%) % (30% 87%) % (28% 70%) 75 78% (54% 89%) % (24% 89%) 9 80% (16% 95%) I. Ogilvie et al. / Vaccine 27 (2009)

10 4900 I. Ogilvie et al. / Vaccine 27 (2009) Table 3 (Continued ) Data used by Study Population/setting Efficacy (%: 95% CI) Duration of protection (%: 95% CI) Efficacy Duration of protection Efficacy against IPD All: 81% n/a [43] Sims, 1988 [67] US Case control study; N = 1054 patients with IPD confirmed by isolation of vaccine serotype S. pneumoniae: 1054 controls were matched by age, underlying illness, and site of hospitalization; Median age: 67.6 years; age range years Efficacy against IPD All: 70% (37 86%) n/a [43] Farr, 1995 [68] US Multicenter, case control study; 244 controls and 122 patients hospitalized for IPD; aged 55 years; matched for admission date, hospital records, and underlying diseases n/a [48] Efficacy against IPD ( 65 years old) All: 65% ( 49 to 92%) High risk : 20% ( 188 to 78%) Meta-analysis of efficacy studies for PPV-23 against pneumonia and IPD in the elderly ( 65 years old) Melegaro, 2004 [66] International n/a [47] Efficacy of 50% against IPD used in study Efficacy against IPD ( 65 years old) RCTs: 30% 47% Observational studies: 47% 81% Efficacy against non-bacteremic pneumonia and IPD: 37.5% Systematic review of randomised clinical trials and observational studies on the efficacy of PPV-23 against IPD Mangtani, 2003 [54] International CI: confidence interval; IPD: invasive pneumococcal disease; PPV-23 pneumococcal polysaccharide vaccine -23 serotypes. * Immunocompromised/high-risk patients were defined as having the following conditions: asplenia, dysgammaglobinemia, renal transplantation, nephrotic syndrome, hematological cancers, metastatic cancers, and systemic lupus erythromatosus. Matched analysis of vaccinated and unvaccinated patients; all ages. High-risk patients were defined as those with COPD, bronchogenic carcinoma and other chronic conditions. compared to younger adults, and is routinely vaccinated in many countries [56]. Overall, vaccination of over 65 year olds was a costeffective strategy. In addition, two studies presented evidence that vaccination of year olds for the prevention of IPD is also cost-effective [9,44]. However, cost-effectiveness was reduced in the very old (over 85 years) in two studies that assumed reduced vaccine efficacy in the very elderly [11,12]. Many elderly people have comorbidities, such as diabetes mellitus, that raise their risk for contracting IPD [57]. Mixed results were reported on the impact of comorbidities on the cost-effectiveness of vaccination [9,48]. Only lost productivity was examined as an indirect cost, and appeared to have limited impact, [47] congruent with the retired status of modeled populations. Although ICERs varied from country to country (e.g., from $12,934 to $33,120 per QALY for 10 Western European countries), [11] reflecting differences in the epidemiological and economic input variables for each country, vaccination was cost-effective for all countries. [11,12] Among the 11 studies examined, the source of study funding did not appear to affect the ICERs, although such potential influence has been reported for economic evaluations of vaccines [61]. Analysis of economic evaluations included in this review, revealed that strengths and limitations were related principally to data sources used to model the impact of vaccination on target populations. While PPV-23 is known to offer protective efficacy against both IPD and pneumococcal pneumonia the precise level of efficacy is unclear, especially in the elderly and high-risk groups. This gap in available evidence affected the relevance and strength of all studies examined. Several meta-analyses of RCTs for PPV-23-vaccine efficacy in adults have been published, but due to varying selection criteria, the results from these meta-analyses differ [18 25]. This has lead to controversy about the effectiveness and value of the PPV-23 vaccine, especially for elderly and high-risk populations. A recent Cochrane Review concluded that there was evidence of protection against IPD in populations of healthy adults in highincome countries, with a protective efficacy of 80 percent (95%CI: 59 90%) from an analysis of RCTs [26]. Observational studies provided evidence of a protective efficacy for IPD in immunocompetent adults (52% [95% CI 37 61%]), including those over 55 years of age [26]. Evidence of a protective effect for the vaccine against all-cause pneumonia in healthy populations from low-income countries was reported from the analysis of RCTs, however this protective effect was inconclusive, and there was no evidence of a protective effect in high income countries; substantial heterogeneity applied to both of these sub-analyses [26]. New data on the effectiveness of PPV-23 in populations receiving the vaccine is required to validate future economic evaluations. Another important issue affecting the relevance and strength of the economic evaluations was the variability of IPD incidence rates reported in different countries, which explains in part the large variability in ICERs. Lower IPD rates are reported in the UK, Sweden, France and Belgium, compared with other European countries [11,12], and are highest in the US [9,10]. This may be explained by variation in the frequency and quality of blood cultures; diagnosis of the infection based on blood culture is not sensitive [1] and ascertainment of the aetiology of pneumonia is neither specific nor sensitive [1]. Alternatively there may be a higher level of surveillance or a difference in the uptake of vaccination in the US. Sensitivity analyses by Evers et al., 2007 included standardization of incidence rates across countries at 50 cases per 100,000; when compared with the base-case cost-effectiveness ratios this reduced the differences between countries markedly and in most cases, improved the cost-effectiveness [11]. Results of this study should be considered in light of its limitations. Comparison of cost-effectiveness ratios is always a challenge given different settings, currency and outcomes. Although ICERs have been standardized to a single currency and year to facilitate