|Home | About | Journals | Submit | Contact Us | Français|
The cost-effectiveness of 13-valent pneumococcal conjugate vaccine (PCV13) compared with 23-valent pneumococcal polysaccharide vaccine (PPSV23) among US adults is unclear.
To estimate the cost-effectiveness of PCV13 vaccination strategies in adults to assist vaccination policy decision-making.
A Markov state-transition model, lifetime time horizon, societal perspective. Simulations were performed in hypothetical cohorts of US 50-year-olds. Vaccination strategies and effectiveness estimates were developed by a Delphi expert panel; indirect (herd immunity) effects resulting from childhood PCV13 vaccination were extrapolated based on observed PCV7 effects. Data sources for model parameters included CDC Active Bacterial Core surveillance, National Hospital Discharge Survey and Nationwide Inpatient Sample data, and the National Health Interview Survey.
Pneumococcal disease cases prevented and incremental costs per quality-adjusted life year (QALY) gained.
In the base case scenario, PCV13 given as a substitute for PPSV23 in current recommendations (i.e., vaccination at 65 years and at younger ages if comorbidities are present) cost $28,900/QALY gained compared with no vaccination and was more cost-effective than the currently recommended PPSV23 strategy. Routine PCV13 at ages 50 and 65 years cost $45,100/QALY compared with PCV13 substituted in current recommendations. Adding PPSV23 at age 75 to PCV13 at ages 50 and 65 years gained 0.00002 QALYs, costing $496,000/QALY gained. Results were robust in sensitivity analyses and alternative scenarios, except when low PCV13 effectiveness against nonbacteremic pneumococcal pneumonia was assumed or when greater childhood vaccination indirect effects were modeled. In these cases, PPSV23 as currently recommended was favored.
Overall, PCV13 vaccination was favored compared to PPSV23, but the analysis is sensitive to assumptions about PCV13 effectiveness against NPP and the magnitude of potential indirect effects from childhood PCV13 on pneumococcal serotype distribution.
The 23-valent pneumococcal polysaccharide vaccine (PPSV23) has been recommended for prevention of invasive pneumococcal disease (IPD) in adults since 19831. Most studies show that PPSV23 provides some protection against IPD, but studies have reached contradictory conclusions for its prevention of nonbacteremic pneumococcal pneumonia (NPP)1-2, which causes several hundred thousand illnesses annually in the US3. Large randomized controlled trials of PPSV23 conducted in developed countries have not found evidence of efficacy against NPP among community-dwelling older adults or among younger adults with chronic illness1, 4-5. Routine childhood vaccination with the 7-valent pneumococcal conjugate vaccine (PCV7) has dramatically decreased both IPD and NPP in children through both direct and indirect (herd immunity) vaccine effects6-7 and reduced adult pneumococcal disease through indirect effects6-9. The introduction of a pediatric conjugate vaccine containing six additional serotypes (PCV13) is expected to further reduce pneumococcal disease in children and adults10-11.
Prior analyses suggest that adult pneumococcal conjugate vaccination could prevent more disease than PPSV23, due to its potential effectiveness against both NPP and IPD12. Although PCV7 has been shown to prevent NPP in children13, PCV13 effectiveness in preventing NPP in adults is currently unknown and the subject of an ongoing clinical trial14. In addition, routine childhood vaccination with PCV13 will likely result in further indirect effects in adults10, perhaps limiting the potential benefits of adult vaccination. As PCV13 is currently under FDA licensure review for use among adults 50 years and older15, decisions about vaccination policy must weigh tradeoffs between the possibility of decreased NPP vs. fewer serotypes covered by PCV13, on a background of childhood vaccination-related changes in pneumococcal epidemiology and suboptimal adult vaccination uptake16. To address these issues, we used decision modeling techniques to examine the effectiveness and cost-effectiveness of pneumococcal vaccination strategies among adult cohorts 50 years of age and older.
Using a Markov state-transition model (Figures 1 and and2),2), we examined six pneumococcal vaccination strategies developed by a Delphi expert panel process (see below): 1) no vaccination, 2) the present US Advisory Committee on Immunization Practices (ACIP) adult recommendations (vaccinate all persons with PPSV23 at age 65; those who received PPSV23 before age 65 for a comorbid condition are recommended to receive another dose at age 65 or later if at least 5 years have passed since the previous dose)1, 3) substituting PCV13 for PPSV23 in current ACIP recommendations, 4) PCV13 at age 50 and PPSV23 at age 65, 5) PCV13 at ages 50 and 65, and 6) PCV13 at ages 50 and 65, then PPSV23 at age 75. Strategies were compared using identical hypothetical cohorts of 50 year old US adults, with cohorts followed as they aged. We used a lifetime time horizon, a societal perspective, and a 3% discount rate for costs and benefits, converting costs to 2006 US dollars17. Quality of life was modeled using health state utility weights, with 0 equaling death and 1 denoting perfect health; quality adjusted life years (QALY) are the product of the health state utility and the length of time in that state.
The Markov model followed cohorts yearly until death. Pneumococcal disease was assumed to occur as in the Box, where the risk of infection is modified the likelihood and effectiveness of vaccination and by projected herd immunity effects from childhood PCV13. In the model, these factors varied by age and presence or absence of comorbidity. In the base case scenario, PPSV23 was assumed to have no effect on NPP4, while PCV13 was assumed to protect against NPP; these assumptions were modified in sensitivity analyses. Individuals could be in health states where pneumococcal disease risk was average or high and, within high-risk health states, persons with immunocompromising conditions were considered separately from those with other comorbid conditions. Persons at average risk could transition to the high-risk group due to development of a comorbid condition. We assumed no transitions from high risk to average risk. Patients could become disabled or die from pneumococcal infection, or recover.
We used 2006 National Health Interview Survey (NHIS) data to segment cohorts into comorbid illness groups to model differential vaccine effectiveness and pneumococcal disease rates based on age and comorbidity, using Centers for Disease Control and Prevention (CDC) definitions for immunocompromising and other comorbid conditions18. Asthma and cigarette smoking were recently added as pneumococcal vaccination indications1. In the model, asthma was included as a comorbid condition; smoking was not, due to difficulties in capturing reliable smoking information in CDC data. We also used Framingham Study19-22 and SEER23 data to model age-related risk of comorbid conditions in the oldest age groups. Patients with both immunocompromising and other comorbid conditions were included in the immunocompromised group; persons with HIV were also assumed to be included in the immunocompromised group. Transitions from other comorbid conditions to immunocompromised were based on SEER age-specific cancer incidence rates23.
CDC Active Bacterial Core surveillance (ABCs) data from 2007-2008 were used to model IPD rates and age-specific likelihood of illness from vaccine-contained serotypes (Table 1). We used previously published methods12 to derive hypothetical “no vaccine” IPD rates. Deaths were modeled using US mortality tables24. For hospitalized NPP rates, we extrapolated from all-cause pneumonia hospitalization rates (Table 2), estimating that pneumococcal pneumonia accounts for 30% of all-cause pneumonia hospitalizations3, 12, 25. We used National Hospital Discharge Survey data26 for our base case analysis, with Nationwide Inpatient Sample data6 used in sensitivity analyses. We assumed that the relative likelihood of hospitalized NPP among age- and comorbidity-specific groups was similar to that observed for IPD, and used rate ratios to model this assumption. Age- and comorbidity-specific rate ratios for IPD, from ABCs data, were applied to NPP rates to calculate age- and comorbidity-specific hospitalized NPP rates (Table 3). We also assumed that the serotype distributions were similar for hospitalized NPP and IPD. Due to uncertainty regarding the age- and comorbidity-specific frequency and serotype distribution of outpatient NPP and its lower cost27-28, we limited our analysis to hospitalized NPP.
As described previously29, an expert panel, using the modified Delphi technique30, estimated PPSV23 effectiveness against IPD (Table 4). A second Delphi panel, comprised of ACIP Pneumococcal Vaccines Working Group members, estimated PCV13 effectiveness against vaccine serotypes causing IPD and NPP (Table 5) and selected the vaccination strategies to be modeled. PCV13 relative effectiveness against NPP was estimated to be 18% lower than for IPD (range 10-60%) in healthy 50 year olds, 25% lower (range 10-60%) in healthy 65 year olds, and 30% lower (10-60%) in persons with immunocompromising and/or other comorbid conditions; these calculations’ results are shown in Table 6. Disability after IPD was modeled using meningitis as a proxy, understanding that not all meningitis is disabling but other IPD syndromes can be. For disability after NPP31, we used 50% of the IPD value and varied it from 0-100% in sensitivity analyses. Vaccination adverse event data and quality of life utilities were literature-based32-34. Vaccine35, administration36, and average adverse event costs ($0.03 per dose34) came from published sources. IPD and NPP hospitalization costs were obtained from 2006 Healthcare Cost and Utilization Project data37.
Potential indirect effects from childhood PCV13 were modeled in two steps. First, changes in adult IPD rates were extrapolated from indirect effects seen post-PVC79. We assumed that age-specific decreases in IPD would occur in commonly carried serotypes (3, 6A, 7F, and 19A) added to PCV13, as was observed for colonizing PCV7 serotypes9, and increases would occur in non-PCV13 serotype IPD. No herd effects were assumed for serotypes 1 and 5, which are considered uncommon colonizers11, 38. Next, age-specific changes in vaccine serotype distributions were modeled, based on observed changes after PCV7 introduction8, leading to a diminished relative likelihood of PCV13 serotype IPD in the first 1-5 years after PCV13 licensure. Longer-term effects were modeled based on observed age-related serotype distribution changes. Potential indirect PCV13 effects for NPP were modeled using point estimates for decreases observed after PCV7 introduction6.
In the base case scenario, we assumed 60.1% adherence to age-based vaccination recommendations and 33.9% adherence to comorbidity-based vaccination recommendations, based on observed PPSV23 uptake16, 39. In strategies modeling current vaccination recommendations, all persons with a comorbid condition diagnosed before age 50 were vaccinated at age 50. Persons aged 50 to 64 developing a comorbid condition were vaccinated that year, then revaccinated either at age 65 or five years after the first vaccination, whichever came last.
Table 7 depicts model parameter values. Parameters were varied individually in 1-way sensitivity analyses and varied simultaneously in probabilistic sensitivity analyses, where random draws from distributions were performed and the effectiveness or cost-effectiveness of each strategy calculated 3000 times. Distributions were chosen based on parameter characteristics and level of certainty. Parameters whose distributions were least certain (e.g., utility weights) were assigned uniform distributions, vaccine effectiveness estimates were assigned triangular distributions, and parameters whose distributions were most certain (i.e., derived from clinical trial or epidemiologic data) were assigned distributions based on data characteristics and ability to account for distribution skewness. In separate analyses, we varied the relative likelihood of IPD or NPP due to vaccine serotypes from 90-110% of the baseline values and considered greater herd immunity effects.
With no vaccination, the lifetime risk from age 50 onward for hospitalized NPP was 9.3%, for IPD was 0.86%, and for death due to pneumococcal disease was 1.8%. Thus, among the cohort of approximately 4.3 million US 50-year-olds in 200640, the model estimated 396,087 NPP hospitalizations, 36,576 IPD cases and 75,647 deaths due to pneumococcal disease over their lifetime. Table 8, top, summarizes the public health impact of different vaccination strategies. PPSV23 strategies prevented more IPD than strategies using only PCV13, while strategies using 2 scheduled PCV13 doses prevented more NPP.
In sensitivity analyses, results were most sensitive to variation in vaccine effectiveness estimates and magnitude of herd effects from childhood PCV13 vaccination on adult NPP. Using Nationwide Inpatient Sample data for all-cause pneumonia rates instead of the National Hospital Discharge Survey minimally affected results.
Due to the unknown effectiveness of adult PCV13 against NPP, we also performed a worst-case NPP scenario (Table 8, bottom), where PCV13 effectiveness against NPP was set at the low range of estimates (Table 6) and base case IPD effectiveness estimates for both vaccines were used. Despite worst-case assumptions, more total pneumococcal disease cases and deaths were prevented by strategies containing PCV13 compared to the current PPSV23 recommendation strategy.
Incremental cost-effectiveness analysis results under base case assumptions are shown in Table 9 (top). In accordance with guidelines17, we present results as incremental cost-effectiveness ratios, ordering strategies by cost and eliminating strategies that are strictly dominated (more costly and less effective) or extended dominated (having higher incremental cost-effectiveness ratios than more effective strategies)41.
With no vaccination, the total per person cost of IPD and hospitalized NPP from age 50 onward was $1047. Compared to no vaccination, current PPSV23 recommendations (vaccination at age 65 and at younger ages if comorbidities are present) cost $34,600/QALY gained. However, this strategy is eliminated due to extended dominance, since its cost-effectiveness ratio is greater than that of the more effective PCV13 substituted in current recommendations strategy, which cost $28,900/QALY gained. PCV13 given routinely at ages 50 and 65 years cost $45,100/QALY compared with PCV13 substituted in current recommendations. PCV13 at ages 50 and 65 followed by PPSV23 at age 75 gained 0.00002 more QALYs per person and cost $496,000/QALY gained. There are no absolute criteria for cost-effectiveness but, in general, interventions costing <$20,000/QALY gained are felt to have strong evidence for adoption, interventions costing $20,000-$100,000/QALY have moderate evidence, and those costing >$100,000/QALY have weaker evidence for adoption42-43.
When PCV13 effectiveness against NPP is set at low range estimates (Table 9, bottom), current PPSV23 recommendations were favored, costing $34,600/QALY gained. PCV13 substituted in current recommendations cost $131,000/QALY gained; other strategies had prohibitive cost-effectiveness ratios. If, in this scenario, PPSV23 effectiveness against IPD is also at the low estimates, the current PPSV23 recommendation strategy cost $60,200/QALY and PCV13 substituted in current recommendations cost $93,000/QALY gained.
The analysis was also sensitive to vaccine costs. PCV13 substituted into current recommendations cost >$100,000/QALY when PCV13 vaccination cost was >$237 (base case estimate=$128). The analysis was not sensitive to individual variation of other parameter values (Table 10).
A probabilistic sensitivity analysis of base case results is shown as a cost-effectiveness acceptability curve (Figure 3), showing the proportion of cost-effectiveness calculations that would be considered acceptable from a societal standpoint for various willingness-to-pay thresholds. This analysis, which includes the likelihood of worst-case scenario for effectiveness against NPP, PCV13 given routinely at ages 50 and 65 would be favored if willingness-to-pay (or acceptability) thresholds were >$50,000/QALY. Changing vaccine effectiveness distributions from triangular to uniform or beta distributions left results essentially unchanged.
In a series of separate sensitivity analyses on the base case model, we relaxed modeling assumptions to test model robustness. If PCV13 had no direct effects on adult nonbacteremic pneumonia, the current PPSV23 policy cost $34,600/QALY and all PCV13 strategies were strictly dominated (i.e., more costly and less effective). If greater herd immunity effects on pneumococcal disease rates (leading to fewer IPD and NPP cases) were modeled, current recommendations using PCV13 cost $38,400/QALY and PCV13 at ages 50 and 65 cost $51,000/QALY gained. Modeling greater decreases in disease likelihood from a vaccine serotype (declining with age from a 16% relative likelihood of disease from PCV13 serotypes compared to nonvaccine serotypes at age 50, rather than from 24% in the base case) led to the current recommendations using PCV13 strategy being dominated, current recommendations using PPSV23 costing $38,900/QALY gained, and PCV13 at ages 50 and 65 costing $106,000/QALY. In the base case analysis, we assumed that 30% of hospitalized all-cause pneumonia was NPP. If NPP accounted for 20% pneumonia hospitalizations, current PPSV23 recommendations cost $34,400/QALY, PCV13 in current recommendations cost $59,100/QALY, and PCV13 at ages 50 and 65 cost $74,200/QALY.
We also examined scenarios where PPSV23 was effective against NPP. PCV13 at 50 and 65 would continue to be favored (at a $100,000/QALY criterion) if PPSV23 effectiveness against NPP was <44% at age 50 and <38% at age 65. At higher effectiveness levels, the currently recommended PPSV23 strategy was favored.
Our analysis favors vaccinating adults with PCV13 instead of PPSV23, and suggests that PCV13 given either as a substitute for PPSV23 in current recommendations or routinely at ages 50 and 65 might reduce pneumococcal disease burden in an economically reasonable fashion. A two-dose PCV13 strategy at 50 and 65, although having a higher cost-effectiveness ratio, addresses the complexity of risk-based recommendations and is consistent with moves away from comorbidity-based strategies, exemplified by recent changes in influenza vaccination recommendations44.
However, results favoring PCV13 were sensitive to assumptions regarding PCV13 effectiveness against NPP. If effectiveness against NPP is low, then current PPSV23 recommendations are favored, but if, in addition, PPSV23 effectiveness against IPD is at the experts’ low range estimate, PCV13 substituted into current recommendations again becomes economically reasonable. This analysis highlights the tradeoffs that policy makers must consider when choosing among adult pneumococcal vaccination strategies. PPSV23 covers 23 serotypes and could prevent more IPD than PCV13 but, based on US and Western European studies, has no consistent effect on NPP4-5. Based on the experience with PCV7 program6-7, PCV13 is likely to prevent NPP. Since NPP is much more common than IPD, PCV13, despite its narrower serotype coverage, should prevent more pneumococcal disease than PPSV23. However, when modeling the possibility of PPSV23 preventing NPP, PPSV23 could be favored, but only if its effectiveness against NPP is higher than would be expected based on trial results4-5. At present, PCV13 effectiveness against pneumonia in adults is unknown. However, PCV7 was consistently effective against pediatric pneumonia in randomized clinical trials13, 46-48.
Results were also sensitive to the magnitude of indirect effects from childhood PCV13. Herd immunity from the added 6 serotypes will likely reduce disease rates in adults. Indirect effects of childhood PCV7 on carriage and IPD rates among adults have been documented8-9, 49. Decreases in all-cause pneumonia hospitalizations in adults 50 years and older were observed following PCV7 introduction, although these reductions were not statistically significant in one study6, but were significant in another7. However, given documented decreases in IPD rates due to indirect effects, we modeled decreases in adult NPP rates based on published point estimates, which could bias the analysis against PCV13. In any case, larger than expected indirect effects from childhood PCV13 would reduce the value of adult PCV13 strategies.
Several limitations should be considered in interpreting our findings. The analysis is limited by the lack of data on PCV13 effectiveness. A randomized trial of PCV13 among 85,000 adults ≥65 years of age is currently underway in the Netherlands; data collection will end 12/2011 but results will not be available until 201314. The magnitude of herd effects and changes in serotype distribution resulting from childhood PCV13 are also unknown. We modeled these based on data from the PCV7 experience. We excluded outpatient pneumonia due to difficulties in accurate estimation of disease rates and characteristics, and the relatively minor role of outpatient costs; again possibly biasing against PCV13. Other studies estimated that outpatient NPP accounted for ≤5% of adult pneumococcal disease costs28, 50. Our model is based on 50-year-old cohorts followed over their lifetime; this analysis does not consider cohorts of differing ages or other issues germane to ensuring population immunity, such as catch up vaccination. Finally, as infrastructure and resources to support adult vaccination are limited, achieving sufficient coverage with PCV13 is uncertain39, 51.
In conclusion, our analysis suggests that PCV13 might prevent more pneumococcal disease compared to the current PPSV23 vaccination recommendations, while remaining economically reasonable. However, these conclusions are sensitive to assumptions regarding PCV13 effectiveness against NPP, PPSV23 effectiveness against IPD and NPP, and herd immunity effects. Model estimates of adult PCV13 impact would be strengthened by evidence of PCV13 effectiveness against NPP from ongoing clinical trials and availability of data on the indirect effects of childhood PCV13 on adult pneumococcal disease rates.
Baseline probability of infection (pre-childhood PCV13) × Projected relative likelihood of infection (indirect effect, post- vs. pre-PCV13) × Relative likelihood of infection from a vaccine serotype (pre-PCV13) × Projected change in infection serotype likelihood (indirect effect) × Probability of being vaccinated × Probability of vaccine effectiveness
We thank the Working Group on Pneumococcal Vaccines of the Advisory Committee on Immunization Practices for providing their expert opinion of pneumococcal vaccine efficacy estimates. We also thank Matthew R Moore, MD, MPH and Ruth Link-Gelles for data from the Active Bacterial Core Surveillance System at CDC.
This work was supported by the NIAID (R01 AI076256)
This study was supported by the National Institute of Allergy and Infectious Diseases, Grant No. R01AI076256
Richard Zimmerman and Mary Patricia Nowalk have a research grant from Merck, Inc., on HPV vaccine
Disclaimer: The findings and conclusions in this report are those of the authors and do not necessarily represent the views of the Centers for Disease Control and Prevention.