We developed a decision-analytic model to compare the costs and outcomes associated with a national immunization program of PCV13, as is currently implemented throughout Canada, compared with PCV10, which was previously used in some provinces following usage of PCV7 and before transitioning to PCV13. Overall, widespread use of PCV13 in Canada was expected to have a substantial public-health impact and be cost-saving due to its demonstrated immunogenicity and broader serotype coverage when compared with PCV10.
Our analysis approach was similar to other Canadian-specific analyses in the literature, in terms of the decision-analytic model structure, diseases considered, and approaches to input data [13
]. Talbird et al. performed an all-Canada analysis that compared the use of PCV10 to PCV7. Thus, it was difficult compare results of our analysis to that analysis. Although we saw many similarities in the input data (i.e., incidence, mortality, costs, sequelae utilities), Talbird and colleagues drew their serotype coverage data from a different surveillance-coordinating center and estimated serotype coverage as an average over a 10-year period. Conversely, we considered current serotype coverage as reported by the National Centre for Streptococcus. In addition, Talbird et al. considered PCV10 direct effects on NTHi in IPD and AOM, in which the effect of PCV10 on NTHi-caused AOM was substantial (30%–38% of AOM due to Streptococcus pneumonia
and 38%–57% of non-Streptococcus pneumonia
disease due to NTHi) [31
]. Talbird et al. also considered a utility decrement due to having AOM, making the approach in our analysis more conservative [55
Chuck et al. [14
] compared the use of PCV10 to PCV13 in Alberta. Their results differed slightly in that they found PCV10 to be cost-effective when considering coverage of NTHi in AOM. Although our analyses were similar around assumptions on utilities, incidence of disease sequelae, and costs, Chuck et al. assumed a greater rate of mortality. In addition, Chuck and colleagues considered an incidence combined with serotype coverage in Alberta as estimated from the National Centre for Streptococcus. An in-depth comparison of the incidence and serotype coverage data values found that the incidence along with serotype coverage assumed in their analysis was much lower than assumed in our analysis. A comparison of these data values with Talbird et al. showed our data values to be more in line with Talbird et al. [31
]. It is unclear whether the incidence along with serotype coverage was much lower in Alberta than for the rest of Canada. As a result, the value of PCV13 could be overestimated. All in all, to account for these differences, we performed multiple scenario analyses around our assumptions on both incidence and serotype coverage. It will be important for decision makers to examine these analyses closely and determine which analysis most closely resembles their situation.
The disease epidemiology has changed dramatically since the introduction of PCV7 in 2003 [1
]. Specifically, there has been a fair amount of serotype replacement. Figure shows the trends of PCV7-covered serotypes over time, as measured by a number of surveillance studies. As seen in Figure , while PCV7-covered serotypes have decreased approximately 6-fold, PCV7 non-covered serotypes have increased 2- to 4-fold in various Canadian territories and provinces. From the Institut National de Sante Publique du Quebec, we also observed that in the early years of PCV7, PCV7 covered approximately 50% of pneumococcal disease (Figure ) [18
]. Over the years and because of the implementation of PCV7, PCV7-covered serotypes have been dramatically reduced. Examining current serotype coverage, we observed that PCV13 was estimated to cover approximately 50% of pneumococcal disease. In addition, data from Laboratoire De La Sante Publique de Quebec for 2009 showed 19A as the most prevalent serotype across all ages, covering nearly 18% of all serotype isolates in Quebec and representing almost 68% of vaccine-preventable IPD covered by PCV 13. Data from the National Centre for Streptococcus for 2009–2010 confirmed 19A to be the most prevalent among 0- to 4-year-olds [23
]. With PCV13 being the only vaccine to cover 19A, significant reductions in disease might be expected. Sensitivity analysis using the various serotype coverage estimates showed PCV13 to remain cost-saving when compared with PCV10.
Figure 4 Change in PCV7 and non-PCV7 serotype coverage over time. (a) Change in PCV7 Serotype Coverage Over Time. PCV7=7-valent pneumococcal polysaccharide-protein conjugate vaccine. Solid line represents coverage of PCV7 serotypes for Canada (more ...)
PCV10 has been shown to have a lower immunogenic response than PCV7 among the seven common serotypes, as demonstrated by lower geometric mean antibody concentrations and by percentage of infants achieving serum immunoglobulin G seropositivity
]. In our base-case analysis, we did not consider the lower immunogenic effect of PCV10. However, in sensitivity analysis we examined the impact of reduced effectiveness and showed that PCV13 continued to dominate PCV10. PCV7 consistently demonstrated a statistically significant reduction in nasopharyngeal carriage [45
], which is a critical component to creating an indirect effect. The data on statistically significant reduction in nasopharyngeal carriage is now emerging [48
]. PCV10 has not demonstrated this reduction [10
]. As a result, we assumed no indirect effects to PCV10 in the base case. It is possible that PCV10 could exhibit indirect effects. Thus, we examined the presence and absence of indirect effects for PCV10 and PCV13, respectively, in sensitivity analyses. In all analyses, PCV13 was found to remain cost-saving compared with PCV10.
In cost-effectiveness analyses of pneumococcal conjugate vaccines, assumptions around AOM incidence and efficacy played an important role in the direct medical costs accrued. Although the costs associated with a single AOM event were low relative to hospitalized disease, due to the high incidence of AOM these costs were a substantial contributor to the disease burden calculated within our model. In addition, because the etiology of AOM is typically all-cause, a critical component to the cost-effectiveness of pneumococcal vaccines was the impact of vaccine effectiveness on NTHi. Recent economic analyses have shown that PCV10 reduces costs and is cost-effective when effectiveness on NTHi is considered [14
]. Chuck et al. [14
] assumed that PCV10 would have a 3% increase in direct effects on all AOM due to NTHi. However, Talbird et al. presented an etiology-weighted estimate of 22.9% reduction in all-cause AOM relative to 6.7% for PCV7. Subsequent analyses also used this approach [58
]. Details of this approach have been discussed elsewhere [55
In this analysis, we considered both direct effects and a combination of direct and indirect effects of the vaccines. However, evidence of indirect effects for PCV10 and PCV13 are limited in the literature. Indirect effects for PCV7 have been firmly established in the Canada, the US, Australia, and the UK [2
]. Because PCV13 shares the same carrier protein as PCV7, we assumed PCV13 would incur similar indirect effects as PCV7. In addition, a recent analysis performed by the CDC demonstrated a decline in PCV13-type IPD in adults in the US [49
]. However, we acknowledge that a recent presentation by De Wals and colleagues [64
] noted that the existence of indirect effects in the presence of PCV7 for all-cause PNE could not be identified in Quebec. An indirect effect for IPD in adults was observed, but the authors also observed “strong and rapid” serotype replacement. Thus, a targeted vaccination program in adults may provide better protection to this group [64
]. We examined of the use of PCV10 and PCV13 in the presence and absence of indirect effects and found PCV13 to be cost-saving in both scenarios. We acknowledge that we did not explicitly examine the impact of serotype replacement following the use of higher-valent conjugate vaccines. As these data become available, it will be important to consider this impact.
A limitation of this model is the uncertainty around vaccine prices due to the confidential nature of tender negotiations. When including indirect effects, our model was not sensitive to PCV13 price variation; PCV13 remained a cost-saving option even if PCV13 was twice the price of PCV10. In addition, PCV13 remained cost-saving regardless of the reduction in cost per dose for PCV10 when the cost per dose of PCV13 was held constant. This was due to the substantial burden of disease caused by PCV13-specific serotypes compared with PCV10-covered serotypes. This difference in coverage represents more than 27% of all serotypes in Quebec [18
]. Due to the confidential nature of negotiated tenders, it will be important for decision-makers to perform these analyses under their own pricing scenarios so that they can better understand the economic value of PCV13 compared with PCV10.