Cost effectiveness of child pneumococcal conjugate vaccination in middle-income countries

Abstract

Policy-makers require information on the potential benefits of and economic case for pneumococcal conjugate vaccination in middle-income countries. We built decision analysis models to evaluate a three-dose infant series of the 7-, 10- or 13-valent pneumococcal conjugate vaccines in 77 middle-income countries compared with no vaccination, accounting for direct protection of vaccinated children as well as herd protection and serotype replacement in unvaccinated children and adults. Over 10 years, pneumococcal vaccination would prevent at least 11.0 million cases and 314 000 deaths in children under-5, one-third of the pneumonia and invasive disease cases and deaths that would occur in this age group without vaccination. Herd protection would prevent 3.1 million cases and 163 000 deaths in older children and adults. A total of 11.1 million discounted disability-adjusted life-years (DALY) would be averted. At a dose cost of $10 for lower- middle-income and $20 for upper-middle-income countries, the net pooled (for all countries together) discounted vaccination cost would be $18.1 billion ($1600 per DALY averted). Vaccination would be cost effective for 72 countries with the 7-valent vaccine and for all countries with the 10- or 13-valent vaccines. The economic case for vaccination is compelling for middle-income countries.

1 Introduction

Streptococcus pneumoniae is a leading cause of childhood morbidity and mortality worldwide and a frequent cause of serious infections in older children and adults. Pneumococcal conjugate vaccination is an effective means of prevention,¹ and introduction of vaccination in low-income countries has been catalysed by innovative financing mechanisms. However, adoption of pneumococcal conjugate vaccine (PCV) by middle-income countries involves unique challenges. These countries are unlikely to receive as much financial assistance as low-income countries and may perceive the cost of the vaccine as high relative to that of other vaccines.

A comprehensive evaluation of the potential health benefits and economic case for PCV in middle-income countries has not yet been performed. This information is crucial to assist national and international policy-makers, including donors and international health organisations, with decisions regarding vaccine introduction. In this study, the projected health benefits, costs and cost effectiveness of PCV in middle-income countries were evaluated, taking into account both direct protection of vaccinated children and indirect effects of vaccination on unvaccinated children and adults.

2 Methods

2.1 Analytic overview

The cost effectiveness of PCV was assessed in the 42 countries classified by the World Bank as upper-middle-income and the 35 lower-middle-income countries that are ineligible for Global Alliance for Vaccines and Immunization (GAVI) assistance (Box 1). A three-part decision tree model was built. The first part addressed outcomes due to direct vaccine protection of vaccinated children under the age of 5 years (Figure 1A) . The second and third parts considered outcomes due to indirect effects of vaccination, consisting of herd protection (Figure 1B) and serotype replacement (the increase in disease caused by non-vaccine pneumococcal serotypes as they fill the ecological niche left when vaccination eradicates vaccine serotypes) (Figure 1C) in unvaccinated children and adults. Health benefits and costs of the first 10 years of a vaccination programme were modelled over a lifetime horizon. Experts in pneumococcal disease epidemiology and vaccinology provided advice on model structure and inputs.

Country-specific data were used when possible, but for several parameters (detailed below and in the Supplementary Technical Appendix), information is known only from certain middle- or high-income countries, necessitating extrapolation to other countries. As a result, although analyses were performed by country, pooled results for the middle-income countries are presented together. Given the lack of data for national analyses for many middle-income countries, we aim to provide a broader insight into the cost effectiveness of vaccination for countries in this income stratum based upon the best information available.

2.2 Vaccination programme

The vaccination programme evaluated in the model consisted of three doses administered at 6, 10 and 14 weeks of age, with no catch-up vaccination for older children. The base-case analysis focused on the 7-valent pneumococcal conjugate vaccine (PCV7) because it has the largest evidence base. The newer 10-valent (PCV10) and 13-valent (PCV13) vaccines were also examined (Supplementary Technical Appendix, Table A4). It was assumed that only a single formulation would be adopted by a given country and that vaccination rates in a country would be the same as for three doses of diphtheria–tetanus–pertussis vaccine (DTP-3).²

2.3 Disease-related probabilities

2.3.1 Children under-5

For vaccinated and unvaccinated children under the age of 5 years, pneumococcal pneumonia, pneumococcal meningitis and or non-pneumonia, non-meningitis invasive pneumococcal disease (NPNM IPD) (such as bacteraemia or osteomyelitis) were assessed (Figure 1A). All syndromes could resolve with no sequelae or could result in death. Meningitis could also lead to chronic sequelae consisting of seizure disorder, motor deficit, mental retardation or deafness.

Selected disease-related probabilities for children under-5 are presented in Table 1 . Estimates of incidence and case fatality rates (CFR) for pneumococcal meningitis, pneumonia and NPNM IPD were from the recently completed WHO Haemophilus influenzae type b and Streptococcus pneumoniae Global Disease Burden (Hib/Pneumococcus GDB) Study,³ which provides the most comprehensive and systematic appraisal to date of pneumococcal disease incidence and mortality in children aged 1 month to <5 years. The study does not include otitis media and we were unable to locate other sources on this syndrome's incidence in middle-income countries; as a result, it was not included in the analyses.

Probabilities of chronic meningitis sequelae were derived from analyses of meningitis outcomes in children in high-income countries and The Gambia,^4,5 the only settings for which data were available. Methods used to extrapolate probabilities to each middle-income country are detailed in the Supplementary Technical Appendix. The resulting probabilities ranged from a median of 0.13 for motor deficit to 0.29 for hearing loss.

2.3.2 Older children and adults

Because data were limited on the pneumococcal disease burden in older children and adults, clinical pneumonia (rather than pneumococcal pneumonia) and pneumococcal sepsis (rather than all pneumococcal NPNM disease) were evaluated in these age groups. Pneumococcal meningitis (Figure 1B) was also evaluated. Only deafness was included as a chronic complication owing to lack of data on other sequelae. [See the Supplementary Technical Appendix for complete references, methods and assumptions.]

Estimates of serotype coverage for older children and adults were obtained from a meta-analysis of pneumococcal isolates causing IPD,⁶ a regional analysis for Latin America and the Caribbean⁷ and national surveillance data from South Africa (A. von Gottberg, personal communication, 2008). Selected disease-related probabilities for older children and adults are presented in Table 2 .

2.4 Direct effects of vaccination (vaccine efficacy)

An estimate of vaccine efficacy against IPD (Table 1) was obtained from a recent meta-analysis by Klugman et al.¹ In the Hib/Pneumococcus GDB Study, the proportion of pneumonia cases and deaths due to pneumococcus was estimated from vaccine trial data using a vaccine probe approach, with efficacy against IPD used as a proxy for efficacy against pneumonia (this proxy was chosen because the lack of diagnostic tools for accurately identifying pneumococcal pneumonia cases prevents determination of the true efficacy against pneumococcal pneumonia).³ For internal consistency within our model, an analogous approach was therefore used, employing the efficacy against IPD as the efficacy against pneumococcal pneumonia.

Given that the vaccine efficacy of PCV10 and PCV13 is not yet known, we attempted to derive estimates of efficacy for the two formulations from immunogenicity data, drawing upon expert guidance for our methods. The resulting values were so close to the efficacy for PCV7 that it was assumed that PCV10 and PCV13 had the same efficacy against IPD and pneumonia as PCV7. However, PCV10 and PCV13 were projected to prevent more disease owing to their greater serotype coverage. It was conservatively assumed that direct vaccine protection would last only through 5 years of age.

2.5 Herd protection

Herd protection was assumed to benefit unvaccinated subjects of all ages and to occur for all pneumococcal disease syndromes (Tables 1 and 2). The incidence of pneumococcal meningitis and sepsis decreased based on the vaccine serotype coverage in a given region as well as estimates of herd protection for vaccine-type IPD. The incidence of clinical pneumonia decreased by the estimated herd protection effect for clinical pneumonia. A linear decrease in vaccine-type disease, based on the average annual herd protection effect following PCV7 introduction in the USA, was modelled. A steady state was assumed to occur after the first 6 years of vaccination for IPD and the first 5 years for pneumonia.

Estimates of herd protection were derived from US IPD surveillance data and an analysis of clinical pneumonia admissions in the USA.^8–10 Because the US experience may not be directly analogous to other countries, it was conservatively assumed that one-half of the US herd effect would occur in middle-income countries. No herd protection was assumed in countries with <50% vaccine coverage (Equatorial Guinea and Gabon).

2.6 Serotype replacement

The same replacement effect observed in IPD in the USA following 6 years of vaccine implementation was assumed (Tables 1 and 2).^9,11 The effects of serotype replacement were modelled as an increase in non-vaccine-type (NVT) IPD following the introduction of PCV7 (Figure 1C) using the following assumptions: a constant annual burden of vaccine-type disease would be averted following vaccine introduction; no catch-up vaccination campaign would be undertaken for older children; after a 2-year lag (due to the lack of a catch-up campaign), NVT disease would increase over the subsequent 6 years; after 6 years, serotype replacement would reach a steady state and the annual burden of NVT disease would remain constant; and at <50% vaccine coverage, no serotype replacement would occur, consistent with herd protection assumptions. [See the Supplementary Technical Appendix for complete references and methods.]

2.7 Costs

Analyses were conducted from the societal perspective, which included costs borne by governments and families. These included direct medical costs (medical facility and personnel costs as well as costs of diagnostic tests, procedures and medications), non-medical costs (patient or caregiver time lost from work and transportation costs) and vaccination costs (costs of vaccine doses and administration). Costs were expressed in 2005 US$ to be consistent with many of the key cost data sources.^7,12–14 Selected costs are summarised in Table 3 for children under-5 and in Supplementary Table 1 for older children and adults.

The vaccine dose cost used for lower- and upper-middle-income countries was $10 and $20, respectively, consistent with the cost of €11.50 (approximately $15) negotiated by Brazil, an upper-middle-income country, for PCV10.¹⁵ Dose costs were assumed to be the same for PCV7, PCV10 and PCV13 because costs negotiated for middle-income countries were not necessarily expected to differ substantially for PCV10 and PCV13 compared with PCV7. Indeed, although the price for PCV13 is higher than that for PCV7 in the USA, the price for PCV10 or PCV13 for middle-income countries will likely be lower than for any PCV formulation in high-income countries, as suggested by the cost established by Brazil. The cost of vaccine administration was assumed to be $5 per dose for lower- and upper-middle-income countries.

To estimate facility and personnel costs, unit costs from the WHO Choosing Interventions that are Cost-Effective (CHOICE) initiative were used.¹³ The data sources for direct medical and non-medical costs were published and unpublished country- and regional-level studies (L.F. Dans and G.V. Gregorio, unpublished report, 2001).^{7,12,14,16–18} Medication unit costs for older children and adults were guided by the International drug price indicator guide.¹⁹

2.8 Cost effectiveness

For calculations of cost effectiveness, disability-adjusted life-years (DALY) and disease and vaccination costs were discounted at 3% per year.²⁰ The cost-effectiveness ratio (CER) was calculated as [(vaccination costs – averted disease costs)/(DALYs averted)]. For each vaccine formulation, the comparator was no vaccination programme. Because the same dose cost and vaccine efficacy against vaccine-type disease was assumed for PCV7, PCV10 and PCV13, differences in cost effectiveness were due to differences in serotype coverage of the three formulations.

DALYs were estimated using standard methods and assumptions,²¹ including age-weighting, and employed estimates of life expectancy from standard life tables.² Standard disability weights were applied for acute disease episodes and chronic meningitis sequelae.²²

2.9 Sensitivity analyses

To identify key drivers of cost effectiveness, sensitivity analyses were conducted by varying values of key model parameters over plausible ranges. Key parameters included disease incidence, CFR, serotype coverage, vaccine efficacy and level of herd protection. [Refer to the Supplementary Technical Appendix for ranges used for each parameter.]

2.10 Software

Analyses were performed with TreeAge Pro Suite (release 2009) (TreeAge Software, Inc., Williamstown, MA, USA) and Microsoft Excel^® (Microsoft Corp., Redmond, WA, USA). [Refer to the Supplementary Technical Appendix for auditing methods.]

3 Results

3.1 Cases and deaths averted

3.1.1 Children under-5

It was projected that without vaccination, children under the age of 5 years in middle-income countries would experience 33.1 million cases of pneumococcal disease from 2010 to 2019, resulting in 961 000 deaths. Accounting for direct protection, herd protection and serotype replacement effects, infant vaccination with PCV7 could potentially prevent 11.0 million cases (33%) (Table 4 ). Among the averted cases, 10.5 million (95%) would be due to pneumococcal pneumonia, with 497 000 (4%) due to NPNM IPD and 85 000 (1%) due to meningitis. Vaccination with PCV7 would prevent 314 000 deaths (33%), with 249 000 (79%) due to pneumonia, 32 000 (10%) due to NPNM IPD and 33 000 (10%) due to meningitis. Adoption of PCV10 or PCV13 could avert 15.1 million (46%) or 16.1 million (48%) cases and 434 000 (45%) or 463 000 (48%) deaths, respectively.

Among children under-5, direct protection would account for 99% of averted cases and deaths (Table 4), with only 1% averted due to herd protection. Serotype replacement associated with PCV7 would result in 190 000 cases and 6000 deaths due to non-vaccine serotypes, increases of 0.6% for each relative to the total projected cases and deaths (vaccine- and non-vaccine-type) that would occur without vaccination.

3.1.2 Older children and adults

Among older children and adults, it was projected that without vaccination 440 million cases of clinical pneumonia, resulting in 23.6 million deaths, and 19.0 million cases of pneumococcal sepsis and meningitis, resulting in 2.6 million deaths, would occur from 2010 to 2019 (Table 4). Accounting for herd protection and serotype replacement effects, PCV7 would avert 3.1 million total cases (0.7%) and 163 000 total deaths (0.6%). Adoption of PCV10 or PCV13 would prevent 5.3 million (1%) or 6.9 million (1%) total cases and 301 000 (1%) or 406 000 (2%) total deaths, respectively.

Serotype replacement associated with PCV7 would result in 5.2 million cases and 287 000 deaths due to non-vaccine serotypes, increases of 1% for each relative to the total projected cases and deaths (vaccine- and non-vaccine-type) without vaccination (Table 4). For older children and adults, the number of cases and deaths due to non-vaccine serotypes resulting from serotype replacement would be substantial relative to the number of vaccine serotype cases and deaths averted by herd protection. In fact, for meningitis and sepsis, vaccination with PCV7 would result in more new NVT cases and deaths due to serotype replacement than averted vaccine-type cases and deaths due to herd protection (Table 4). The same would be true for PCV10 and meningitis cases and deaths.

For all age groups combined, including children under-5, 74% of averted cases and 65% of averted deaths would be in lower-middle-income countries (Supplementary Table 2).

3.2 Vaccination costs, disease costs and disability-adjusted life-years averted

Infant pneumococcal vaccination in all middle-income countries from 2010 to 2019 at the same coverage rate as DTP-3 would require 1.2 billion doses at a total programme cost of $20.5 billion, discounted at 3% per year (Table 5 ). From 2010 to 2019, infant vaccination with PCV7 would avert a total of $2.4 billion in discounted disease costs. The net cost of a vaccination programme (vaccination costs – disease costs) with PCV7 would therefore be $18.1 billion. Disease costs that would be prevented with PCV10 or PCV13 are shown in Table 5. Lower-middle-income countries would account for approximately one-half of disease costs averted.

From 2010 to 2019, vaccination with PCV7 could avert a total of 11.1 million discounted DALYs in all middle-income countries (Table 5). Adoption of PCV10 or PCV13 would avert 16.4 million or 18.5 million discounted DALYs, respectively. The majority of averted DALYs (84%) would be due to prevention of pneumonia cases and deaths (data not shown). Although only approximately one-half of the disease costs averted would be in lower-middle-income countries, these countries would account for approximately two-thirds of the DALYs averted.

3.3 Cost effectiveness

If projected costs and health outcomes were pooled for all middle-income countries, the incremental cost per DALY averted of vaccination with PCV7 compared with no vaccination would be $1600 (Table 5). The CER for the pooled lower-middle-income countries would be $1500/DALY averted, whilst for the pooled upper middle-income countries it would be $1900/DALY averted. The median CERs for the lower- and upper-middle-income countries, respectively, would be $1100 and $5400 per DALY averted. Results for PCV10 and PCV13 are shown in Table 5.

Using the WHO criteria of three times the gross domestic product (GDP) per capita as the threshold for cost-effective interventions,²³ vaccination with PCV7 would be cost effective for 72 countries, with Barbados, Belarus, Montenegro, Serbia and Seychelles as the exceptions. Vaccination with PCV7 would be highly cost effective (have a CER lower than GDP per capita) for 53 countries (Figure 2A) . Adoption of either PCV10 or PCV13 would be cost effective for all countries and highly cost effective for 68 or 71 countries, respectively (Figure 2B and 2C). In general, a lower GDP per capita would correlate tightly with a lower CER (i.e. better cost effectiveness) for vaccination, whilst the correlation would be less tight at higher GDP per capita.

3.4 Sensitivity analyses

The projected pooled cost effectiveness of PCV7 for all middle-income countries was robust over a range of estimates for various model parameters. Cost effectiveness was most sensitive to vaccine dose cost, vaccine serotype coverage and pneumococcal disease incidence (Figure 3 ). At the highest vaccine dose cost ($15 for lower-middle-income countries and $30 for upper-middle-income countries), vaccination remained cost effective for 66 countries and highly cost effective for 36 countries. Varying CFRs and herd protection had a moderate impact on cost effectiveness. Increasing the vaccination coverage rate to 100% of the DTP-3 rate also had a moderate effect due to incorporation of herd effects for countries that have DTP-3 rates <50% and thus had no herd effects included in the base-case analysis. Medical and non-medical disease costs had only a small effect.

3.5 Additional analyses

It was estimated that China, a lower-middle-income country, accounted for 47% of DALYs and 56% of costs averted due to PCV7 in lower-middle-income countries and contributed 31% of the DALYs and 30% of costs averted for all middle-income countries. Because it dominated the results, an analysis was performed excluding China from the pooled countries. When China was excluded from cost-effectiveness calculations, the CER for PCV7 for the pooled lower-middle-income countries decreased from $1500 to $1100 per DALY averted, whilst the CER for all middle-income countries decreased from $1600 to $1500 per DALY averted.

A regional comparison of the disease burden that would be averted by the three vaccine formulations was also performed. Owing to better serotype coverage, PCV10 would provide considerable additional benefit in some regions compared with PCV7, whereas the additional coverage by PCV13 relative to PCV10 would lead to only modest additional benefit in any region (Figure 4 ). The increase in DALYs averted by PCV10 relative to PCV7 would be highest in Africa, Asia and Latin America.

4 Discussion

In this first comprehensive analysis of pneumococcal conjugate vaccination in middle-income countries, vaccination with PCV7 was found to be cost effective in all but five countries, and with PCV10 or PCV13 it would be cost effective in all countries. Over the first 10 years, accounting for direct effects and herd protection and serotype replacement, infant vaccination with PCV7 would avert at least 11.0 million (33%) pneumococcal pneumonia and IPD cases and prevent 314 000 deaths in children under-5. Although direct vaccine protection would account for the majority of health benefits in young children, herd protection would avert an additional 3.1 million cases and 163 000 deaths due to all-cause pneumonia, pneumococcal sepsis or pneumococcal meningitis in older children and adults, contributing 44% of DALYs averted for all groups. Approximately two-thirds of DALYs averted would be in lower-middle-income countries, consistent with their larger populations and greater disease burden.

PCV10 and PCV13 could yield substantial additional benefit in Africa, Asia and Latin America given the substantial additional serotype coverage provided by these formulations compared with PCV7 as well as the large populations and correspondingly large disease burden in middle-income countries in these regions. The current findings support adoption of PCV10 and PCV13 in these regions in particular.

The possible detrimental effects of serotype replacement were included in the analyses. Although serotype replacement might be lower in middle-income countries owing to lower rates of vaccination than in the USA, the same magnitude of replacement observed in the USA was assumed. Under this assumption, serotype replacement was projected to have a substantial impact in older children and adults. This finding resulted from (1) our conservative approach of assuming one-half the herd protection effect observed in the USA, (2) the much smaller magnitude of herd protection versus direct protection effects in averting disease and (3) the predominance of NVT disease in older children and adults at baseline (Table 2). Even with US-level serotype replacement, however, vaccination would remain cost effective. A better understanding of other factors that may affect both serotype replacement and herd protection, such as choice of vaccination schedule and adult carriage of non-vaccine serotypes, is needed to project with greater certainty the implications of these effects for middle-income countries.

Vaccine dose cost was a key driver of cost effectiveness. Increasing the cost within plausible ranges (from $10 to $15 for lower-middle-income countries and from $20 to $30 for upper-middle-income countries) resulted in a CER above the threshold for cost effectiveness (three times GDP per capita) for 11 of the 77 countries. Herd protection was a moderately important driver of cost effectiveness, underscoring the need to improve knowledge of this effect in middle-income countries.

The Sabin Vaccine Institute's analysis of pneumococcal conjugate vaccination in Latin American and Caribbean countries is among the only other published economic evaluations of these vaccines in middle-income countries.⁷ For a $20 per dose cost, the study projected a CER of $1619/DALY averted, similar to the estimate of $1600/DALY averted calculated for the pooled middle-income countries in the current analysis. The CER derived here for middle-income countries, based on direct and herd protection, falls between the CER previously estimated for low-income countries ($91/DALY averted)²⁴ and for the USA ($7500 per life-year saved).²⁵

The current study had multiple limitations. Because of the paucity of country-specific data for several model parameters, such as serotype coverage and many disease costs, findings for lower- or upper-middle-income countries combined are reported. Information is particularly lacking on herd protection and serotype replacement in middle-income countries. We addressed these indirect effects in our analysis because of their suspected importance, drawing upon expert opinion to extrapolate from US data, but follow-up evaluations are needed to establish whether the assumptions from the US experience will hold in the middle-income countries. Data are also lacking regarding pneumococcal disease incidence in older children and adults, necessitating extrapolation from all-cause disease estimates in these age groups. In addition, the analysis did not give vaccination credit for preventing otitis media owing to the lack of information on the syndrome's incidence and complications, as well as the expected impact of vaccination, in middle-income countries. This was a conservative choice, however, since the estimates of health benefits and cost effectiveness associated with vaccination would likely be even greater with otitis media taken into account.

A new understanding of the worldwide burden of pneumococcal disease in young children, together with availability of vaccine formulations better suited to the serotype distributions in Asia, Africa and Latin America, make initiation of pneumococcal vaccination in middle-income countries more desirable than ever. The current study indicates that the economic case for vaccine adoption is compelling for middle-income countries and robust over a range of epidemiological and pricing scenarios. However, further information is needed on budget impact, affordability and sustainability of pneumococcal vaccine introduction. Efforts to determine empirically the impact of vaccination in middle-income countries will be important for assessing the accuracy of the estimates used as the basis for this analysis.

Authors' contributions

MMN, TAL, OL, MDK and AS designed the study; MMN, AT, OL, MDK and AS collected the data; MMN, AT, OL, MDK, LBR and AS analysed the data; all authors interpreted the data; MMN, AT, OL, MDK and AS wrote the manuscript and all authors critically revised the manuscript. All authors read and approved the final manuscript. MMN, AT and AS are guarantors of the paper.

Funding

This work was funded by the GAVI Alliance through collaborative arrangements among GAVI's PneumoADIP at Johns Hopkins Bloomberg School of Public Health (Baltimore, MD, USA), University of Medicine & Dentistry of New Jersey (UMDNJ)–New Jersey Medical School (Newark, NJ, USA), and Harvard Medical School (Boston, MA, USA). The study sponsors had no role in the study design; collection, analysis and interpretation of data; writing of the manuscript; and decision to submit the paper for publication.

Conflicts of interest

LBR's husband owns stock in Pfizer, which acquired Wyeth, one of the manufacturers of pneumococcal conjugated vaccine, in 2009. AS has received a grant from Wyeth/Pfizer to her institution for a cost-of-illness study among South African children with acute lower respiratory infection. All other authors declare no conflicts of interest.

Ethical approval

Not required.

Acknowledgements

The authors are grateful to Steve Black, Cynthia Whitney and Shabir Madhi for thoughtful contributions as expert panellists. They also thank David Goldblatt for valuable input.

References