Effectiveness of 10-valent pneumococcal conjugate vaccine estimated with three parallel study designs among vaccine-eligible children in Finland

EFFECTIVENESS OF 10-VALENT PNEUMOCOCCAL CONJUGATE VACCINE ESTIMATED WITH THREE PARALLEL STUDY DESIGNS AMONG VACCINE-ELIGIBLE CHILDREN IN FINLAND Hanna Rinta-Kokko1, Kari Auranen2, Maija Toropainen3, J. Pekka Nuorti3,4, Hanna Nohynek3, Lotta Siira3, Arto A. Palmu5 1 Department of Public Health Solutions, Finnish Institute for Health and Welfare, Helsinki, Finland 2 Department of Mathematics and Statistics and Department of Clinical Medicine, University of Turku, Finland 3 Department of Health Security, Finnish Institute for Health and Welfare, Helsinki, Finland 4 Health Sciences Unit, Faculty of Social Sciences, University of Tampere, Finland 5 Department of Public Health Solutions, Finnish Institute for Health and Welfare, Tampere, Finland Corresponding author Hanna Rinta-Kokko, MSc Finnish Institute for Health and Welfare Mannerheimintie 166, P.O. Box 30 00271 Helsinki, Finland [email protected] tel. +358 295248141 This is the accepted manuscript of the article, which has been published in Vaccine. 2020, 38(6), 1559-1564. https://doi.org/10.1016/j.vaccine.2019.11.049 ABSTRACT Background Ten-valent pneumococcal conjugate vaccine (PCV10) was introduced into the Finnish National Vaccination Programme (NVP) in 9/2010. We estimated the individual-level effectiveness (VE) of PCV10 in children during eight years of vaccine implementation. Methods Data on invasive pneumococcal disease (IPD) were collected from national, population-based surveillance, and vaccination status from the vaccination register. Vaccine-eligible children were followed from six months of age until end of 2018 (born 6/2010 or later, ages 6‒102 months). VE was estimated with three parallel approaches: full cohort, nested case-control and indirect cohort designs adjusting with age-group, sex and calendar year. Results VE against PCV10 serotype IPD was estimated at 93% (95% credible interval 87‒97%), 98% (90‒100%) and 100% (98‒100%), and against PCV10-related serotypes at 46% (-13‒72%), 51% (-24‒79%), and 78% (-7‒ 97%), with the full cohort, nested case-control, and indirect cohort designs, respectively. The estimated VE against non-PCV10-related (neither PCV10 nor PCV10-related) serotypes was negative but included zero effectiveness (full cohort VE -67%, -711‒52%; nested case-control VE -77%, -929‒59%). VE against all IPD was estimated with these two methods at 54% (24‒71%) and 61% (26‒79%). Over time, VE against PCV10 IPD remained stable but VE against all IPD decreased. Discussion All designs provided estimates that were concordant with each other, but those with the full cohort design were usually the most precise. PCV10 offered sustained high VE against PCV10-serotype IPD on vaccinated children during the first decade after introduction into the programme. 1 INTRODUCTION Pneumococcal conjugate vaccines (PCVs) have been successful in providing direct protection for vaccinated children as well as herd protection for all members of the population against pneumococcal diseases caused by the serotypes included in the vaccines. Replacement in carriage and disease by serotypes not included in the vaccines has partially reduced the overall impact of conjugate vaccines, although less in child populations compared with adults [1]. The population-level impact of a pneumococcal vaccination programme can be assessed with a before-after design [2]. To assess the impact, time series data from a long enough pre-introduction period are needed to reliably estimate the baseline incidence of IPD and to control secular trends in serotype-specific rates. During wide-scale use, the direct protective effect of the vaccine on the vaccinated (vaccine effectiveness, VE) can be estimated by comparing the incidence of disease in the vaccinated and unvaccinated parts of the population in parallel [3]. Both the vaccinated and unvaccinated experience the indirect effects of the vaccination programme but unvaccinated individuals are used as the comparator group. Although information on vaccination status is often available for specific subpopulations, this is rare for the entire population. Several parallel study designs, e.g. the indirect cohort, screening, and case-control designs have been used for VE estimation in certain subpopulations [4-10]. VE of 13-valent pneumococcal conjugate vaccine has also been estimated for the whole child population of New South Wales and Western Australia [11]. The indirect cohort and case-control designs have been compared in VE estimation of PCVs e.g. in the US [5, 12] and Brazil [13, 14]. In Finland, surveillance of IPD dates back to 1995, with comprehensive linkage of cases to the national population register and serotyping data since 2004, long before the implementation of universal infant vaccination with ten-valent pneumococcal conjugate vaccine (PCV10) in 2010. The PCV10 vaccination programme has shown very high impact in reducing the incidence of all IPD in vaccine-eligible children, driven by a marked decrease in IPD by PCV10 serotypes together with slow replacement in children by IPD caused by non-PCV10 serotypes [15, 16]. A national vaccination register covering all infant vaccinations has been in place since 2009 [17], allowing the estimation of VE within entire cohorts of vaccine-eligible children. The aim of this work was to estimate VE of PCV10 in children 6‒102 months of age eight years after the programme implementation, and to compare different estimation methods and their congruence in the Finnish setting characterised by high vaccine uptake. 2 METHODS 2.1 Vaccination programme PCV10 was introduced after a public tendering process into the Finnish national vaccination programme (NVP) without catch-up vaccinations in September 2010. All children born on or after June 1, 2010, have been eligible for vaccination with a 2+1 schedule at 3, 5 and 12 months of age. Vaccination coverage rose quickly above 90% and the uptake of at least one dose of PCV10 was estimated at 95.5% in the birth cohort of 2015 [18]. Prior to PCV10 vaccinations in the FinIP trial [19] and NVP, 7-valent pneumococcal conjugate was recommended for certain risk groups, but pneumococcal vaccines were not routinely used and the vaccine uptake was minimal (<2%). Influenza vaccine was introduced into the NVP in the beginning of the influenza season 2007/2008 for children aged 6 to 36 months. The uptake of influenza vaccine in this age group has been estimated to vary between 15 and 40% [17]. 2.2 Data sources and case definition The study cohort was defined based on the Finnish Population Information System as all vaccine-eligible children (born between 6/2010 and 6/2018 and followed from 1/2011 through 12/2018, from 6 months up to 102 months of age, Figure 1). The vaccination status of each individual child in the cohort was retrieved from the National Vaccination Register. The child was defined as vaccinated if at least one dose of PCV10 was registered. Based on this definition, vaccination coverage was 91.7% in the cohort. The individual follow-up started at the age of six months to ensure that the majority (95.3%) of the vaccinated study subjects had received 2 doses of PCV10 before their first birthday. Some 0.3% of children had received the 13-valent conjugate vaccine; in the analysis they were considered as vaccinated with PCV10. Children whose vaccinations were incompletely covered by the National Vaccination Register (altogether 14%, 71882 out of 508734 vaccine-eligible children) were excluded from the analysis. Reasons for incomplete vaccination information were related to either the child’s health care center not providing complete data, or to the child itself having moved to another municipality or abroad [17]. Cases of invasive pneumococcal disease (IPD) were identified from the National Infectious Diseases Register (NIDR), a population-based electronic laboratory surveillance system maintained by the Finnish Institute for Health and Welfare (THL). The IPD surveillance in Finland and the THL laboratory methods have been described earlier [20, 15]. There have been no major changes in care seeking practices or detection methods during the study period. The cases, defined as isolations of Streptococcus pneumoniae from blood or cerebrospinal fluid, were categorised according to the causative serotype into three mutually exclusive groups: PCV10 serotypes (1, 4, 5, 6B, 7F, 9V, 14, 18C, 19F, 23F), PCV10-related serotypes that belong to the same serogroups as the PCV10 serotypes, and non-PCV10-related serotypes. All register-based information was linked by using the unique national personal identity code. 2.3 Vaccine effectiveness VE was defined as one minus the relative rate of IPD in vaccinated compared with unvaccinated in the study cohort (Figure 1). VE was estimated against the three serotype groups, and separately against the PCV10-related serotype 19A and against all IPD. Five vaccinated cases without serotype data were included only in the analysis of VE against all IPD. Three study designs were used to estimate VE. In the full cohort design all children in the study cohort were included and followed until IPD, death or end of study period. VE was estimated with Poisson regression adjusted by age group (6‒23, 24‒47, 48‒102 months), sex and calendar year. In the nested case-control design, five controls for each IPD case were selected from the case's risk set, matching with age, sex and calendar year. Case’s risk set included all children in the study cohort who were at risk of IPD at the time of case occurrence. Conditional logistic regression was used to estimate VE. In the indirect cohort design, VE was estimated as the odds of vaccination in IPD cases of the target serotype or serotype group compared to the odds of vaccination in non-PCV10-related cases. VE was estimated with logistic regression adjusted by age group, sex and calendar year. To assess any time trends in serotype-specific incidence rates or VE, the study period was split into two parts: years 2010‒2014 (early period) and 2015‒2018 (latter period). Statistical inferences were performed within the Bayesian framework. Uninformative prior distributions were used: normal distribution with mean 0 and variance 106 for predictor parameters and Gamma distribution with mean 1 and intensity 105 for precision. All analyses were carried out with R (version 3.4.4) and the INLA library. Results are presented as point estimates (posterior mean) and 95% posterior probability (credible) intervals (CI). 2.4 Ethical approval The authorisation for research use of surveillance data was granted by THL, the regulatory agency with jurisdiction over national health registry data (THL/1090/6.02.00/2013). The THL Institutional Review Board approved the study. 3 RESULTS Table 1 presents VE against IPD in vaccine-eligible children estimated with three parallel study designs. There were altogether 150 IPD cases: 15 caused by PCV10, 78 by PCV10-related, and 52 by non-PCV10related serotypes. Serotypes 14 (7 cases) and 19F (4 cases) were the most common PCV10 serotypes. Serotype 19A was the most common PCV10-related serotype with 60 cases, followed by 23B, 23A and 6A with 6, 4 and 3 cases, respectively. All three 6A cases had been vaccinated. Serotypes 3 (21 cases), 22F (10 cases), and 15B/C (6 cases) were the most common non-PCV10-related serotypes. VE against PCV10-serotype IPD was estimated at 93% (95% credible interval 87‒97%), 98% (90‒100%) and 100% (98‒100%) with the full cohort, nested case-control and indirect cohort designs, respectively. VE against PCV10-related IPD was estimated respectively at 46% (-13‒72%), 51% (-24‒79%) and 78% (-7‒97%). The majority of PCV10-related IPD cases were caused by serotype 19A. Therefore, the estimates of VE against 19A IPD under the three designs were close to that of the whole PCV10-related IPD group (50%, - 19‒78%; 45%, -54‒79%; 66%, -76‒95%). VE against non-PCV10-related IPD was estimated at -67% (-711‒52%) and -77% (-929‒59%) with the full cohort and nested case-control designs, respectively. VE against all IPD was estimated at 54% (24‒71%) with the full cohort and 61% (26‒79%) with the nested case-control design. Of note, VE against non-PCV10related IPD or all IPD could not be estimated with the indirect cohort design since IPD cases of non-PCV10related serotypes were used as controls in the analysis. When the follow-up was split into two sub-periods, some trends in VE were discerned (Table 2). The estimates of VE against PCV10 IPD remained the same as for the entire study period, but the credible intervals were wider due to smaller case counts, especially in 2015‒2018. Towards the end of the latter period (2015‒2018), vaccine-type disease had practically been eliminated as there was only one case of serotype 19F in 2017 in an unvaccinated child and no cases caused by PCV10 serotypes in 2018. The incidence of 19A disease increased from 0 to 8.7/100 000 person-years in unvaccinated and from 2 to 3.8/100 000 person-years in vaccinated children from the early (2010–2014) to the latter (2015–2018) period. Based on the full cohort design, VE against 19A was 59% (4–81%) during the latter period. The point estimate of VE against non-PCV10-related IPD decreased from -32% to -105% with the full cohort design in 2015‒2018. However, the credible intervals were wide and included zero effectiveness (-1860‒ 63% in 2015‒2018). The incidence of non-PCV10-related IPD was low throughout the study period. However, vaccinated children generally had higher non-PCV10-related IPD incidence than unvaccinated children. VE against all IPD decreased from the early period (70%; 35‒85%) to the latter period (39%; -24– 67%), as estimated with the full cohort design. 17. Baum U, Sundman J, Jääskeläinen S, Nohynek H, Puumalainen T, Jokinen J. Establishing and maintaining the national vaccination register in Finland. Euro Surveill. 2017;22(17):pii=30520. DOI: http://dx.doi.org/10.2807/1560-7917.ES.2017.22.17.30520 18. Finnish Institute for Health and Welfare (THL), Finland. The Vaccination Register. Helsinki: THL; [Accessed 13 Aug 2019]. Available from: https://thl.fi/en/web/vaccination/vaccinationcoverage/national-vaccination-register 19. Palmu AA, Jokinen J, Borys D, Nieminen H, Ruokokoski E, Siira L et al. Effectiveness of the tenvalent pneumococcal Haemophilus Influenzae protein D conjugate vaccine (PHiD-CV10) against invasive pneumococcal disease: A cluster randomised trial. Lancet 2013;381(9862):214-22. 20. Siira L, Kaijalainen T, Lambertsen L, Nahm MH, Toropainen M, Virolainen A. From Quellung to multiplex PCR, and back when needed, in pneumococcal serotyping. J Clin Microbiol. 2012;50(8):2727-31. 21. Palmu AA, Toropainen M, Kaijalainen T, Siira L, Lahdenkari M, Nieminen H, et al. Direct and indirect effectiveness of the 10-valent pneumococcal conjugate vaccine against carriage in a cluster randomized trial. Pediatr Infect Dis 2017;36:1193-200 22. Nuorti JP, Rinta-Kokko H, Toropainen M, Siira L, Virtanen MJ, Nohynek H, et al. Invasive pneumococcal disease in adults after infant 10-valent pneumococcal conjugate vaccination in Finland. Submitted for publication. 23. Toropainen M, Nyholm O, Siira L, Jalava J, Rinta-Kokko H, Nuorti JP, et al. Genomic epidemiology and antimicrobial resistance in invasive serotype 19A pneumococci before and after introduction of 10-valent pneumococcal conjugate vaccine (PCV10) in Finland. In: 11th International symposium on pneumococci and pneumococcal diseases, Melbourne, Australia; 2018. 24. Andrews N, Waight P, George R, Slack M, Miller E. Impact and effectiveness of 23-valent pneumococcal polysaccharide vaccine against invasive pneumococcal disease in the elderly in England and Wales. Vaccine 2012(30):6802-8. 25. Andrews N, Waight P, Borrow R, Ladhani S, George R, Slack M, Miller E. Using the indirect cohort design to estimate the effectiveness of the seven valent pneumococcal conjugate vaccine in England and Wales. PLoS ONE 2011;6(12):e28435. doi:10.1371/journal.pone.0028435 Table and Figure legends Table 1. Vaccine effectiveness against IPD in PCV10 vaccinated children, estimated with three parallel study designs (full cohort; nested case-control; indirect cohort) Finland, September 2010 – December 2018. Footnote: a PCV10 serotypes in the data: 6B, 7F, 14, 19F, 23F; b PCV10-related serotypes in the data: 6A, 6C, 6D, 7C, 9N, 19A, 23A, 23B; c non-PCV10-related serotypes in the data: 3, 11A, 12F, 15A, 15B/C, 22F, 24F, 33F, 34, 35F, 38. Table 2. Vaccine effectiveness against IPD in PCV10 vaccinated children, estimated with three parallel study designs (full cohort; nested case-control; indirect cohort). Finland, September 2010 – December 2014 and January 2015 – December 2018. Figure 1. IPD cases and their vaccination status in the study cohort of PCV10 eligible children in Finland. The dashed vertical line represents the split of the study period into two sub-periods in which VE was estimated separately (early period 2010–2014; latter period 2015–2018). [PLEASE USE COLOR IN PRINT] Table 1. Vaccine effectiveness against IPD in PCV10 vaccinated children, estimated with three parallel study designs (full cohort; nested case-control; indirect cohort) Finland, September 2010 – December 2018. Design/Serotype group Unvaccinated Vaccinated Vaccine effectiveness (%) Estimate 95% crebile interval 2010-2018 Full cohort design Cases Follow-up years Cases Follow-up years Adjusted with age group, sex and calendar year PCV10 serotypes a 8 114440 7 1580743 93.4 87.2-96.6 PCV10-related serotypes b 9 114440 69 1580743 46.2 -13.2-71.9 19A 7 114440 53 1580743 50.1 -19.2-78.2 Non-PCV10-related serotypes c 2 114440 50 1580743 -66.7 -711.4-52.1 All IPD 19 114440 131 1580743 54.1 23.7-71.1 Nested case-control design Cases Controls Cases Controls Matched with age, sex and calendar year PCV10 serotypes 8 9 7 66 98.0 89.9-99.7 PCV10-related serotypes 9 29 69 361 50.5 -23.6-79.2 19A 7 24 53 276 45.2 -53.8-79.2 Non-PCV10-related serotypes 2 15 50 245 -76.6 -928.8-58.9 All IPD 19 54 131 696 61.1 25.6-79.3 Indirect cohort design Cases Cases Adjusted with age group, sex and calendar year PCV10 serotypes 8 7 99.9 98.4-100 PCV10-related serotypes 9 69 78.3 -6.8-96.7 19A 7 53 65.7 -75.7-94.9 Non-PCV10-related serotypes 2 50 Reference - All IPD 19 131 - - Footnote: a PCV10 serotypes in the data: 6B, 7F, 14, 19F, 23F; b PCV10-related serotypes in the data: 6A, 6C, 6D, 7C, 9N, 19A, 23A, 23B; c non-PCV10-related serotypes in the data: 3, 11A, 12F, 15A, 15B/C, 22F, 24F, 33F, 34, 35F, 38. Table 2. Vaccine effectiveness against IPD in PCV10 vaccinated children, estimated with three parallel study designs (full cohort; nested case-control; indirect cohort). Finland, September 2010 – December 2014 and January 2015 – December 2018. Period/Design/ Unvaccinated Vaccinated Vaccine effectiveness (%) Serotype group Estimate 95% crebile interval 2010-2014 Full cohort design Cases Follow-up years Cases Follow-up years Adjusted with age group and sex PCV10 serotypes 6 33678 5 400907 93.0 77.3-97.9 PCV10-related serotypes 2 33678 11 400907 54.1 -143.6-88.1 19A 0 33678 8 400907 NA NA Non-PCV10-related serotypes 1 33678 16 400907 -32.1 -1203.7-76.7 All IPD 9 33678 32 400907 70.4 35.2-85.3 Nested case-control design Cases Controls Cases Controls Matched with age, sex and calendar year PCV10 serotypes 6 7 5 48 98.3 88.5-99.8 PCV10-related serotypes 2 5 11 60 67.6 -154.5-95.2 19A 0 3 8 37 NA NA Non-PCV10-related serotypes 1 6 16 79 -30.3 -1666.3-84.1 All IPD 9 18 32 187 82.3 48.4-94.0 Indirect cohort design Cases Cases Adjusted with age group and sex PCV10 serotypes 6 5 99.2 88.1-100 PCV10-related serotypes 2 11 88.1 -255.4-99.7 19A 0 8 NA NA Non-PCV10-related serotypes 1 16 Reference - All IPD 9 32 - - 2015-2018 Full cohort design Cases Follow-up years Cases Follow-up years Adjusted with age group and sex PCV10 serotypes 2 80761 2 1179836 93.6 54.1-99.1 PCV10-related serotypes 7 80761 58 1179836 43.3 -36.5-72.2 19A 7 80761 45 1179836 59.4 4.3-80.7 Non-PCV10-related serotypes 1 80761 34 1179836 -104.8 -1860.1-62.6 All IPD 10 80761 99 1179836 38.5 -23.5-66.6 Nested case-control design Cases Controls Cases Controls Matched with age, sex and calendar year PCV10 serotypes 2 2 2 18 96.9 51.7-99.9 PCV10-related serotypes 7 24 58 301 45.2 -53.9-79.2 19A 7 21 45 239 56.8 -25.1-84.3 Non-PCV10-related serotypes 1 9 34 166 -123.4 -2576.7-68.3 All IPD 10 36 99 509 40.3 -39.4-73.1 Indirect cohort design Cases Cases Adjusted with age group and sex PCV10 serotypes 2 2 100.0 NA PCV10-related serotypes 7 58 76.5 -70.3-98.1 19A 7 45 80.2 -39.8-98.4 Non-PCV10-related serotypes 1 34 Reference - All IPD 10 99 - - Figure 1