Assessing the efficiency of catch-up campaigns for the introduction of pneumococcal conjugate vaccine: a modelling study based on data from PCV10 introduction in Kilifi, Kenya

With the aid of Gavi, the Vaccine Alliance (Gavi for short), many low-income countries, in particular across Africa, have introduced pneumococcal conjugate vaccines (PCVs) into their infant immunisation programmes. However, there remain Gavi countries, particularly in south Asia and northern Africa and some with large infant populations, who are yet to follow [1]. Country policy makers, along with global stakeholders, have a high interest in achieving optimal health impact from PCVs as quickly as possible; however, approaches for achieving maximum and rapid impact have to be weighed against relative cost. In situations where vaccine supply is constrained, as was the case several years ago for PCVs, issues of efficiency and equity in vaccine use are also a consideration [2]. The World Health Organisation (WHO) recommends that catch-up campaigns can be used as part of the introduction of PCVs to accelerate the build-up of herd protection and hence PCV impact [3]. However, it is unclear if such catch-up campaigns are an efficient way to use PCVs or if the gains from this approach are less than the relative increase in the number of doses required.

The value of a catch-up campaign is assessed by quantifying the trade-off between the costs of vaccinating additional age groups and the benefit of additional direct and indirect protection. However, there are few observational data on the impact of PCV campaigns, particularly from low- and middle-income countries (LMICs), to quantify the optimal approach of catch-up campaigns. One of the few well-studied examples of a PCV introduction catch-up campaign in an LMIC occurred in Kilifi, Kenya. The 10-valent pneumococcal non-typeable Haemophilus influenzae protein D-conjugate vaccine (PCV10) was introduced into the Kenyan routine childhood vaccination programme in early January 2011 using the WHO Expanded Programme on Immunisation (EPI) schedule of three infant doses at 6, 10 and 14 weeks. Additionally, in Kilifi County, at the introduction of the cohort programme, a three-dose catch-up campaign was offered to all infants younger than 12 months of age and a two-dose catch-up to children 12–59 months of age.

We fitted a transmission dynamic model of pneumococcal carriage (a precondition for disease and the source of person-to-person community transmission) and disease to detailed pre- and post-PCV introduction data from Kilifi. This allowed us to quantify the marginal impact of the PCV catch-up campaign on carriage and disease in Kilifi over the hypothetical impact of a routine cohort vaccination programme alone in that setting. Using this framework, we aimed to estimate the dose efficiency of alternative catch-up campaigns in relation to PCV cohort introduction alone.

Our model was able to reproduce Kilifi’s pre- and post-vaccination epidemiology of near VT elimination across age groups and serotype replacement with NVTs (Fig. 1). The introduction of PCV10 together with a catch-up campaign in children under 5 years old was predicted to prevent 220 (172–270) cases of IPD in Kilifi within the first 10 years after the start of the vaccination programme. Once the full direct and indirect effects (herd protection and serotype replacement) of the programme are established, the vaccination programme is predicted to avert 23 (17–28) cases of IPD annually (Fig. 2a); the majority of those among children younger than 5 years old (Fig. 1 and Additional file 1: Figure S1).

The catch-up campaign among children up to 5 years of age was estimated to accelerate direct and indirect effects of PCV. By doing so the Kilifi programme was estimated to prevent an additional 65 (48–84) cases of IPD (Fig. 2b) over 10 years in the overall population, if compared to a cohort introduction without a catch-up campaign. A catch-up programme confined to children younger than 2 years or younger than 1 year of age was estimated to prevent an additional 34 (26–43) or 18 (14–22) IPD cases, respectively, if compared to cohort introduction alone. The majority of cases averted by the catch-up campaigns would have occurred within the first 6–8 years after the start of vaccination (Fig. 2a). The age distribution of cases averted through catch-up campaigns was similar to that averted through routine immunisation (Additional file 1: Figure S1).

Within the first 10 years of the PCV infant programme in Kilifi, about 205,000 doses of vaccine were predicted to be used as part of the routine immunisation schedule. The under 5 catch-up campaign required 17,000 additional doses of vaccine (Fig. 3a). We estimated that, in the 10 years following introduction of PCV10 in Kilifi, routine vaccination without any catch-up campaigns would use 1321 (1058–1698) doses of PCV for each case of IPD averted. As herd protection gradually develops, this programme gains efficiency in the first years after introduction; that is, the annual number of cases prevented increases while the number of vaccinated individuals remains similar (Fig. 2a). By the 10th year after the start of cohort vaccination without a catch-up campaign, we estimated that routine vaccination uses 910 (732–1184) doses of PCV per IPD case averted.

Our results were not sensitive to variations in vaccine coverage (Additional file 1: Figure S4 and Table S1) or competition, the duration of protection, the relative protection of infants by PCV as compared to toddlers and the vaccine efficacy against carriage and IPD (Additional file 1: Figure S3).

In many high-income countries PCVs have been introduced with the help of catch-up campaigns to accelerate the direct and indirect protection that is offered to the community [27–29]. We used extensive data from the KHDSS, a well-studied mix of rural and urban Kenyan communities representing a typical low-income setting, to estimate the incremental effects that different catch-up campaigns are likely to have over routine vaccination and, therefore, whether PCV catch-up campaigns are an efficient use of PCV supply. We found that rapidly increasing the protection in the community via catch-up not only reduces cases of IPD by direct protection of older children but also reduces the burden of IPD in the whole childhood population by developing herd protection more rapidly. Any of the three catch-up programs considered in the analysis were estimated to use fewer vaccine doses to prevent a case of IPD than cohort introduction during the first 10 years; the catch-up schedules were more efficient than the routine cohort vaccination programme alone even after full herd effects were in place, in the 10th year of the programme. While the additional catch-up doses given to 1 year olds were estimated to provide the largest marginal efficiency, we find that cohort introduction alongside a catch-up campaign in children under 5 years old was the most efficient introduction strategy overall.

Data on the observed impact of PCV catch-up campaigns are sparse and mostly circumstantial. Catch-up campaigns of different sizes have been used for introducing PCVs into countries including the UK [30], USA [31], Israel [32], Brazil [33] and Kilifi, Kenya [11]. However, a head-to-head comparison with cohort introductions that would allow an evaluation of the additional impact of the catch-up is challenging because of the dissimilarity of the underlying population and other factors including vaccine coverage, intensity of pneumococcal transmission, differences in demographic structure and population mixing, serotype distribution and prevalence of epidemiological risk factors such as HIV infection.

As well as extending direct protection to older children who are also at high risk of pneumococcal disease, catch-up campaigns also rapidly increase the proportion of individuals in the transmitting population who are protected against VT acquisition and hence onward transmission. This indirect effect is non-linear, preventing a high number of infections for each increment in vaccine coverage when that coverage is low but suffering from a saturation effect for higher coverage levels. As a result, predictions of the optimal extent of catch-up campaigns need to account for these non-linear effects; i.e. they need to incorporate transmission dynamics.

Most of our posterior estimates that had an informative prior were similar to that prior, showing that in most instances the model is able to match the data well using the pre-specified parameter space. The notable exception was the vaccine efficacy against carriage in toddlers. While the model was unable to replicate the observed steep decrease in VT prevalence following vaccination using the mean prior estimate of 36% efficacy, the posterior suggests a mean efficacy of 55% which has been observed in other sites [23] and falls well into the range of the prior estimate.

We have restricted our analyses to catch-up campaigns that targeted age groups under 5 years of age, as those were deemed feasible both from a programmatic and a supply point of view. However, including older children may well be efficient, in particular in settings where older children contribute substantially to the transmission of pneumococci. Also, we have not considered programmatic issues associated with implementation of catch-up campaigns. Due to the immense additional burden on available staff, catch-up campaigns can disrupt routine immunisation services. Furthermore, we have studied the most efficient use of PCV10 supply but not the cost-effectiveness or affordability of catch-up campaigns for PCV10 introduction. One of the major differences in a cost-effectiveness analysis is that it takes into account the higher delivery costs of vaccine through a supplementary immunisation activity. Assuming that doses delivered as part of a PCV10 catch-up campaign were up to 75% more expensive than doses delivered through the routine EPI schedule, however, did not qualitatively change our findings on the superior efficiency of catch-up programmes.

We did not account for population growth in our model, which may impact the transmission dynamics in the post-vaccination era and hence on our findings. However, modelling work predicting the impact of PCV10 in Kilifi from pre-vaccination data has shown that accounting for population growth in Kilifi is unlikely to qualitatively change the prediction but only slightly reduces the long-term impact of vaccination on IPD [34]. As the impact of a catch-up campaign is mostly visible within a few years after vaccination, it is likely largely unaffected by long-term changes in demographics. Hence, accounting for population growth is likely to further favour the use of catch-up campaigns for introduction of PCV. Other models have taken into account more of the diversity of pneumococcal serotypes by either modelling them individually or by using finer groupings [19, 34–36]. Despite the considerable heterogeneity of serotypes in regard to their ecology within both our VT and NVT group, our model captures the post-vaccination dynamics well. The impact of catch-up campaigns largely concerns the acceleration of long-term impact of the programme, and hence nuances in the dynamics of specific serotypes are unlikely to qualitatively change our findings. We have not taken into account potential underreporting of IPD episodes. However, extra attention has been paid to that during the study period; the same procedures were followed at Kilifi hospital so that ascertainment has not changed, and hence any such bias in our estimates for relative impact of PCV introduction strategies should be minimal. Further, our primary objective was to estimate the relative impact of PCV10 introduction with and without catch-up campaigns. Hence, we chose not to incorporate factors that influence the absolute impact of pneumococcal vaccination but that should make little difference to the impact of catch-up campaigns relative to cohort introduction, such as impact on otitis media, pneumonia or mortality, secular trends, antibiotic consumption, comorbidities or cost. We fit age-group- and serotype-group-specific pathogenicity and assumed that those would stay unchanged during the study period. While vaccination may have led to a disproportionate emergence of some serotypes and hence changes in serotype-group pathogenicity, Nurhonen and Auranen have shown that the assumption of proportional expansion and hence unchanged pathogenicity generally holds [37], and our model fits well to both pre- and post-vaccination IPD and carriage without the need for allowing a change in pathogenicity.

The generalisability of our results beyond KHDSS is dependent on a number of factors. We show in a sensitivity analysis the robustness of our findings to vaccine coverage, vaccine efficacy against the carriage and IPD, the ratio of toddler to infant protection, the duration of vaccine-induced protection and the between-serotype group competition (see Additional file 1: Figure S3 and Figure S4). However, other factors that could not be systematically assessed in this analysis include transmission intensity and serotype distribution. In settings with higher transmission intensity, birth cohort PCV introduction likely takes longer to establish full herd effects. As a result, catch-up campaigns that accelerate the build-up of herd protection have the potential to prevent more IPD cases and hence be an even more efficient strategy for PCV use. Furthermore, we did not account for potential cross-reactivity of PCV10 against serotypes 6A and 19A, which has been reported previously [38, 39]. However, both 6A and 19A carriage prevalence increased in the post-vaccination era in Kilifi [11].

We assumed that two PCV10 doses in infancy, given as part of the routine EPI schedule, are similarly efficacious at preventing VT carriage and disease as a single catch-up dose in toddlers and young children. Fitting to the data from Kilifi, our model did not reject this hypothesis. While our results are robust to factors including these differences in relative protection in infants and toddlers and variable vaccine coverage (proportion of protected infants and toddlers), the number of doses that are administered to establish protection could have a larger impact. Twelve months after the introduction of PCV10 in Kilifi, 76% of infants eligible for three doses of PCV10 aged less than 1 year had received at least two doses of PCV10, and 62% of children 1–4 years old had received at least one dose. We have chosen the dosing of catch-up campaigns to align with what was rolled out in Kilifi; however, other dosing regimens have been used [40], notably in south Africa with two doses in infancy followed by a booster dose at 9 months of age [41]. In our analysis we assume for simplicity that all children receive the exact number of doses that in this analysis was deemed sufficient to induce protection. Drop-out rates in Kilifi are relatively low; e.g. more than 97% of infants who received one dose go on to receive a second dose before 1 year of age, but including drop-outs in the analysis would further decrease the efficiency of the cohort introduction in comparison to the catch-up campaigns. To define protection in our model, we used two doses in infancy and one dose for catch-up campaigns as a protective schedule but assumed that vaccinated children would eventually receive three doses as part of the routine schedule or alternatively two doses or one dose if part of the catch-up campaign in <1 year old or older children, respectively. Assuming instead that the children protected through routine immunisation and catch-up campaigns had received two doses and one dose, respectively, did not qualitatively change the results.

Pneumococcal conjugate vaccines (PCVs) are among the most expensive vaccines currently available and make up more than 30% of the annual budget of Gavi. Proposed ways to use PCVs more efficiently include a potential reduction in the number of infant doses if herd effects have been established [42] or a dilution of the current formulation. We show here that catch-up campaigns present an important, readily available tool which can increase the efficiency of the PCV’s impact on disease at introduction. For countries yet to introduce, or potentially also for countries with lagging coverage, strategies that include catch-up campaigns warrant serious consideration.