Direct, indirect and total effects of 13-valent pneumococcal conjugate vaccination on invasive pneumococcal disease in children in Navarra, Spain, 2001 to 2014: cohort and case–control study

M Guevara 1 2 3 , A Barricarte 1 2 3 , L Torroba 2 4 , M Herranz 2 5 , A Gil-Setas 2 4 , F Gil 6 , E Bernaola 2 5 , C Ezpeleta 2 4 , J Castilla 1 2 3 , Working Group for Surveillance of the Pneumococcal Invasive Disease in Navarra 7 1. Instituto de Salud Pública de Navarra, Pamplona, Spain 2. Navarra Institute for Health Research (IdiSNA), Pamplona, Spain 3. CIBER Epidemiología y Salud Pública (CIBERESP), Spain 4. Department of Clinical Microbiology, Complejo Hospitalario de Navarra, Pamplona, Spain 5. Department of Pediatrics, Complejo Hospitalario de Navarra, Pamplona, Spain 6. Department of Pediatrics, Hospital García Orcoyen, Estella, Spain 7. Members of the group are listed at the end of the article


Introduction
The 7-valent pneumococcal conjugate vaccine (PCV7) has proved highly effective in preventing invasive pneumococcal disease (IPD) caused by the serotypes included in its formulation [1,2]. However, its impact has varied across countries due to factors that may include differences in serotype distribution, vaccination coverage and characteristics of vaccination programmes [3,4]. New, higher valency pneumococcal conjugate vaccines (PCVs) containing 10 (PCV10) and 13 (PCV13) serotypes were licensed on the basis of non-inferiority of immunogenicity compared with PCV7 [5]; thus, post-licensure studies are required to assess their effects under real-life conditions.
To date (June 2015), there are few studies published on the direct effect of PCV13 against IPD, and all except one have used the indirect cohort method and have hence been limited to evaluating it against vaccine serotypes rather than total IPD [6][7][8]. PCVs can also reduce IPD incidence among unvaccinated individuals as a result of reduced transmission. This indirect or 'herd' effect has been studied in unvaccinated age groups [9][10][11][12], but not in children targeted for vaccination. The total effect accounts for both the direct and indirect effects on vaccinated individuals [13,14].
In Navarra, Spain, PCVs became available for private purchase in June 2001 (PCV7), November 2009 (PCV10) and June 2010 (PCV13), and are publicly funded only for children with selected IPD risk factors, including cardiovascular, respiratory, neurological, renal or hepatic disease, diabetes, cancer, immunosuppression, HIV infection, haemoglobinopathy, and cerebrospinal fluid leak [15]. The Spanish Association of Paediatrics recommends PCV for all children younger than 5 years [16], and coverage has increased progressively through  a B vs A in the cohort study is the direct effect in the baseline period.
b The direct effect in the case-control study is estimated by the odds ratio of vaccination vs non-vaccination in cases compared with controls. the private market, reaching 78% in children up to 23 months of age in 2013 [17]. Most of the vaccinated children have received a complete 3 + 1 schedule, with doses at 2, 4 and 6 months plus a booster dose at 12-15 months. Since 2010, PCV13 has been the predominant PCV in use. After the change from PCV7 to PCV13, IPD incidence from all serotypes decreased by 69%, from 60.7 to 18.7 cases/100,000 inhabitants, in children younger than 5 years [17], in line with what has been observed in other countries [10,11,[18][19][20][21].
The aim of this study was to estimate the effect of PCV13 on IPD incidence in vaccinated (direct and total effects) and unvaccinated (indirect effect) children younger than 5 years.

Study design
A population-based cohort study with follow-up during 2001-2014, and a nested case-control study from July 2010 to December 2014 were conducted in Navarra, a region with ca 640,000 inhabitants, including ca 34,700 aged less than 5 years, in 2014 [22]. The Navarra Ethical Committee for Medical Research approved the study protocol.

Sources of information and variables
The Navarra Health Service provides healthcare, free at point of service, to 97% of the inhabitants of the region. Clinical records have been computerised since 2000 and include reports from primary care, hospital admissions, the regional vaccination register, and laboratory test results.
Vaccination history was obtained from the regional vaccination register [23], which includes all doses received by children, including those acquired in the private market. Vaccine doses were counted starting 15 days after their administration and the 14 days after receiving the first dose were not considered for the analysis. Cases of IPD were identified through the active laboratory-based surveillance. IPD was defined as isolation, PCR or antigen detection of Streptococcus pneumoniae from a normally sterile body site. Pneumococcal isolates were serotyped at the national reference laboratory (Instituto de Salud Carlos III, Madrid) by the Quellung reaction or dot-blot assay, and were classified as PCV7 serotypes (4, 6B, 9V, 14, 18C, 19F, 23F), additional PCV13 serotypes (1, 3, 5, 6A, 7F, 19A), or non-PCV13 serotypes.

Cohort study to evaluate the indirect and total vaccine effects
The cohort included children covered by the Navarra Health Service from birth to their fifth birthday, the end of the follow-up on 31 December 2014 or date of death, whichever occurred first. Cox regression was performed to obtain hazard ratios (HR) with their 95% confidence intervals (CI). Age in days was used as the underlying time scale, with entry time defined as age at 1 January 2001 or 75 days of age if it was later, and exit time as 59 months of age, age at IPD diagnosis or death, or at 31 December 2014, whichever occurred first. Calendar periods and PCV status were defined as time-dependent variables. Person-years (PY) at risk were used as the denominators of the IPD incidence rates. Since a culture was taken from all suspected cases and PCR and antigen detection were progressively introduced as complementary tests, culturenegative cases (n = 13) were excluded to maintain comparability between the periods.
To evaluate the indirect and total effects of the PCVs with respect to the pre-vaccine situation, we considered four periods according to PCV use and coverage in children younger than 5 years: the baseline period (2001-2004) during which PCV7 use was low; the In another cohort analysis we evaluated the specific effect of the change from PCV7 to PCV13, considering two periods: 2005-2009, or the period of PCV7 use; and 2011-2014, or the period of PCV13 use. Three exclusive categories of vaccination status were defined in the following order: at least one dose of PCV13, at least one dose of PCV7 or PCV10, and no dose of PCV.
The year 2010 was excluded from this analysis because this was a transition year with appreciable use of PCV7, PCV10 and PCV13.
The incidence of IPD in unvaccinated children during the baseline period was used as the reference to estimate the indirect effect by comparison with the IPD incidence in unvaccinated children in each PCV period, and to estimate the total effect by comparison with the incidence in vaccinated children in each PCV period ( Figure) [13,14]. Where zero cases were observed in one group, the p value was obtained by the two-tailed mid-p exact test.

Case-control study to evaluate the direct vaccine effect
A case-control study, nested within the cohort, included as case patients all children born since June 2008 (as they might have received at least one dose of PCV13) and who were diagnosed with IPD by culture, PCR or antigen detection between July 2010 and December 2014. For each case, eight controls were selected from children with no previous IPD, individually matched by paediatric practice, district of residence and date of birth (± 2 months). Of all the children who met these eligibility criteria, the eight with dates of birth closest to that of the case were selected. Previous inclusion of a twin was an exclusion criterion.
Healthcare computerised databases were used to obtain the sex, date of birth, paediatrician, district of residence, premature birth (< 37 weeks' gestation), low birth weight (< 2,500 g), major chronic illness (defined as cardiovascular, respiratory, neurological, renal or hepatic disease, diabetes, immunosuppression or cancer), primary care visits in the previous 12 months, other children younger than 5 years in the household, and parental income level (< EUR 18,000 and ≥ EUR 18,000/year).
The reference date for cases was the date of symptom onset, and for controls, the date on which their age exactly matched the age in days of their corresponding case at the time of symptom onset. Different categorisations of PCV status were used to analyse the effect of either PCV13 including mixed schedules or PCV13-only schedules, with or without distinction of the number of doses received. Vaccination with PCV7 or PCV10 without PCV13, and non-PCV vaccination were assigned to two separate categories.
In different analyses we evaluated the effect of receiving PCV13 on the risk of IPD due to all serotypes, to PCV13 serotypes, to additional PCV13 serotypes and to non-PCV13 serotypes, using non-PCV vaccination as the reference. A sensitivity analysis was performed excluding children with any medical condition. Adjusted matched odds ratios (OR), with their 95% CI, were calculated using conditional logistic regression. We assessed for confounding by including additional variables one by one in the model. Covariates were removed if they did not change the OR by at least 15%. Vaccine effects were calculated as (1 -HR) x 100 or (1 -OR) x 100. Two-tailed p values < 0.05 were considered to be statistically significant.    In a similar analysis limited to IPD cases from PCV13 serotypes, the total effect was 90% (HR: 0.10; 95% CI: 0.03-0.32) ( Table 3).

Evaluation of the direct vaccine effect
Between July 2010 and December 2014, 34 cases of IPD were included in the case-control study. The median age was 18.9 months (range 3.4-57.7 months). Clinical presentations were bacteraemia (17 cases), pneumonia (14 cases) and meningitis (three cases). In 30 cases the serotype was available: 10 cases were caused by PCV13 serotypes and 20 cases by non-PCV13 serotypes. There were two vaccine failures due to serotype 3 in immunocompetent children who had received four doses of PCV13. Additionally, there was one case due to serotype 19A (a PCV13-only serotype) in a child who had received three doses of PCV10 and a booster dose of PCV13. All but one of the seven cases in unvaccinated children were due to PCV13 serotypes, while most of the cases in vaccinated children (18 of 21) were caused by non-PCV13 serotypes.
The 34 cases and 272 matched controls presented similar sociodemographic characteristics and underlying medical conditions, with the exception of a higher proportion of males in cases (71% vs 51%, p = 0.037) (  (Table 5).

Discussion
In a cohort of children younger than 5 years followed up during the 14 years in which PCV7 was introduced and its subsequent replacement by PCV13, we observed large reductions in the incidence of IPD, both in vaccinated children (total effect, 76%) and in those not vaccinated (indirect effect, 78%). The effect against PCV-serotype cases was earlier and more pronounced in vaccinated than in unvaccinated children, but these differences disappeared when we evaluated the effect against IPD due to all serotypes.
The replacement of PCV7 by PCV13 was followed by a reduction of 90% in the incidence of IPD due to PCV13 serotypes in children who had received PCV13.
Other studies have also reported a high effectiveness of PCV13 against IPD due to vaccine serotypes: 86% in Quebec, Canada [8], and 75% in the UK [7], in the first 3 and 3.5 years after PCV13 introduction, respectively. In our case-control analysis, as in the study from Quebec, the effectiveness estimate for non-PCV13 serotypes was negative, although with a wide CI including the null effect.
In our cohort analysis, the incidence of IPD due to non-PCV13 serotypes among children who had received PCV13 increased compared with the incidence in unvaccinated children in the pre-PCV period. This finding suggests some vaccine-induced replacement, a phenomenon well-documented for PCV7 [3,[24][25][26][27], which may be beginning to occur with PCV13 [12,[19][20][21]28]. Nevertheless, the initial incidence of non-PCV13 serotype IPD was low, and its increase has been much smaller than the reduction in vaccine serotype incidence, resulting in a considerable net population benefit of vaccination. The replacement effect may be a potential source of bias to be corrected for in indirect cohort studies [6,7].
The two vaccine failures observed in children completely vaccinated with PCV13 were due to serotype 3, for which different studies suggest lower effectiveness [6,7,11,29].
To the best of our knowledge, ours is the first study to estimate the indirect effect of PCV13 against IPD in children younger than 5 years, and the results are consistent with those of other studies that have described reductions in IPD incidence in unvaccinated age groups after the change to PCV13 [10,11,[17][18][19]. A study in Boston in the United States between July 2010 and June 2012 observed an indirect effect against nasopharyngeal colonisation in unimmunised children as vaccine uptake reached 75% [30], however, we observed an indirect effect against PCV7 serotypes starting in the period 2005-2007, when coverage reached only 48%. The strong indirect effect of PCV13, added to serotype replacement in vaccinated children, leads to apparently paradoxical results, such as the low or absent direct effect of the vaccine against all-serotype IPD. In this situation the total effect is the measure that best reflects the benefit in the vaccinated population.
This study has certain limitations. Although the study size was small, it was enough to sustain the statistically significant findings presented in the results but not for more disaggregated analysis. Some of the estimates' CIs were wide and should be interpreted with caution.
In the comparisons between periods in the cohort study, we cannot rule out the possibility that some of the changes detected could have been due to temporal fluctuations in specific serotype incidence unrelated to vaccine use, giving an over-or underestimation of the indirect and total vaccine effects. However, the high effectiveness of PCV13 against vaccine serotypes was confirmed in the case-control study limited to the last period, in an analysis not affected by temporal fluctuations. The cohort analysis took into account age, sex, period and PCV status, but was not adjusted for other variables. Although some residual confounding could be possible, the results were consistent with those of the case-control analysis in which we did adjust for other variables. The fact that the same study has found a protective effect of PCVs for vaccine serotypes but not for non-vaccine serotypes also argues against important residual confounding. PCV7 was already available in the baseline period (2001)(2002)(2003)(2004), with an average coverage of 11%; accordingly, some indirect effect cannot be ruled out. Nonetheless, the incidence of IPD in children younger than 5 years was still very high (75 per 100,000 PY), indicating little impact of vaccination. Few cases were not serotyped, and they were excluded from some analyses; however, sensitivity analyses were performed in the cohort study assigning these cases alternately to each serotype group (data not shown), and the main results were hardly affected. This study also has a number of strengths. Populationbased surveillance was active and consistent throughout the follow-up period. The case-control design achieved good comparability by individual matching, and was also adjusted for relevant covariables. The intermediate levels of vaccine coverage allowed a sufficient number of vaccinated and unvaccinated individuals to evaluate the direct and indirect effects.
In conclusion, PCV13 was highly effective in preventing vaccine-serotype IPD. With vaccine coverage around 76% in children, PCV benefits have been substantial and similar in vaccinated and unvaccinated children through strong total and indirect effects. Signs of possible serotype replacement in vaccinated children highlight the importance of ongoing surveillance and development of new pneumococcal vaccines. Joint assessment of vaccine effects at the individual and population level helps to better understand the complex dynamics of changes in the epidemiology of IPD that follow changes in the pneumococcal vaccination programme. The important vaccine benefit at the population level supports the recommendation for universal PCV vaccination in children.