Pediatric Pneumococcal Serotypes in 4 European Countries

TOC Summary: Non–heptavalent pneumococcal conjugate vaccine serotypes have increased in Spain, France, Belgium, and England and Wales.

ing of the dynamics of the additional serotypes is needed to help decision making on future vaccine strategies. This study describes and compares temporal trends of PCV7 serotypes and NVTs among children in 4 European countries, taking into account the levels of PCV7 use. We focused on the emergence of serotypes 1, 7F, and 19A because they were responsible for most of the NVT increase.

Methods
This population-based study is based on surveillance data collected prospectively by the national reference laboratories (NRLs) of Spain, Belgium, France, and England and Wales during July 1996-June 2006. (In the United Kingdom, Scotland and Northern Ireland were not included in this study because they use separate surveillance systems.) IPD isolates were referred by laboratories throughout each country to the NRL. IPD surveillance among children was enhanced in England and Wales, France, and Belgium starting in 1996, 2002, and 2005, respectively, by encouraging microbiologists to systematically refer pneumococcal isolates to the NRL for typing.

Defi nitions and Inclusion Criteria
We included all IPD cases, defi ned by isolation of S. pneumoniae from a normally sterile fl uid in children <15 years of age and referred to the NRL of their country. One isolate per disease episode was used in the analysis. Serotypes targeted by PCV7 were grouped as vaccine types (PCV7 types). All other serotypes were considered NVTs. Meningitis was defi ned as isolation of S. pneumoniae in cerebrospinal fl uid (CSF).
An epidemiologic year was July through June. A prevaccine period was defi ned as the 3 epidemiologic years during July 1999-June 2002. The postmarketing period was 2005-2006.

Microbiologic Testing
Serotype and antimicrobial susceptibility were determined by each NRL as described (9,(15)(16)(17). In Spain, all strains of serogroups 6 and 19 were subjected to PCR serotype identifi cation (18). In Belgium during 1996-2004 (before PCV7 marketing), 36% of isolates had the serogroup but not the serotype determined (17% for serogroup 19). However, all serogroup 19 isolates from children <2 years of age were typed, and during the postmarketing period, the serotype was determined for all isolates received. Isolates with missing serotype were assumed to follow the same serotype distribution as isolates from the same serogroup, by year and patient age group.
Isolates were considered susceptible, intermediate, or resistant to antimicrobial drugs according to Clinical and Laboratory Standards Institute criteria (penicillin intermediate and resistant, MIC 0.12-1.0 mg/L and MIC >1 mg/L, respectively; erythromycin resistant, MIC >0.5 mg/L) (19). England and Wales data on antimicrobial drug resistance were not available for this study.

Vaccine Coverage and Macrolide Use
Because studies estimating vaccine coverage used different methods among countries, we used vaccine doses to estimate a proxy of vaccine coverage. Data on monthly vaccine doses sold or distributed were provided by the PCV7 manufacturer (Wyeth, Brussels, Belgium; Madrid, Spain; Maidenhead, UK; and Paris, France) and by the Health Protection Agency for doses distributed by the UK Department of Health. Assuming that all doses were administered to children <2 years of age at an average of 3 doses per child (allowing for missed doses and catch-up schedules), we calculated the proportion of children <2 years of age who should have received an average of 3 PCV7 doses. We also calculated the number of vaccine doses distributed per 1,000 children <5 years of age for comparison with serotype-specifi c incidences in children <5 years of age.
We collected data on use of antimicrobial drugs from the European Surveillance of Antimicrobial Consumption (20). We compared these data with the serotype-specifi c incidence of antimicrobial drug-resistant isolates.

Data Analysis
For all incidence calculations, we adjusted numbers of cases to the rate of underreporting to the NRL to estimate total numbers of cases and control for surveillance enhancement over time. Underreporting rates were calculated by country, year, and age group by dividing the respective number of cases with an isolate typed at the NRL by the total number of laboratory-confi rmed IPD cases estimated in the country. Total numbers of IPD cases were estimated by different methods: in Belgium and France, through periodic capture-recapture studies and correction for laboratory coverage (7,21); in England and Wales, by reconciliation of 2 large datasets (22); in Spain, by calculation of hospital underreporting rates as a proxy (23). Age-specifi c incidence rates were computed by dividing adjusted numbers of cases by the respective midperiod population of each country and age group, by using population fi gures from Eurostat for Spain and from the national institutes for statistics of Belgium, France, and England and Wales. For serotype-specifi c incidence calculations, cases with missing serotype data were accounted for by multiplying the overall incidence by the yearly proportions of serotypespecifi c disease in each age group (12).
Trends in incidence over time were tested for linear model by the t test, except for France because of missing data points; this model showed an overall better fi t for the 3 studied serotypes. Correlation between vaccine doses per 1,000 children <5 years of age and serotype-specifi c incidence over years was tested by the Pearson correlation test, allowing for a 6-month lag between vaccine use and incidence. We also compared the annual average of serotype-specifi c incidence of the prevaccine period with the postmarketing period and computed incidence rate ratios (IRRs) and their exact 95% confi dence intervals; p values were calculated by the Fisher exact test. We considered p values <0.05 signifi cant. All statistical analyses were calculated by using STATA version 10.1 (StataCorp, College Station, TX, USA).  (Table 1). IPD incidence per 100,000 children <15 years of age increased during the 10-year period in Spain, Belgium, and England and Wales from 14.0 to 18.5, 20.1 to 28.2, and 6.1 to 10.6, respectively, but remained fairly stable in France, ≈7.0.

PCV7 Coverage
Isolates from CSF represented 15% of invasive isolates in patients <15 years in Spain, 11% in Belgium, 16% in England and Wales, and 32% in France. Because blood isolates were underrepresented in France, NRL data were adjusted to the CSF/blood distribution reported by national epidemiologic surveillance, by year and age group (7). Only adjusted data are presented here.

PCV7-Type and NVT IPD Cases
In children <5 years of age, incidence of PCV7-type disease started to decrease shortly after PCV7 introduction in Spain, Belgium, and France; the decrease was inversely related to increasing vaccine sales ( Figure 1). Between the prevaccine period and the last study year (2005-2006), PCV7-type IPD signifi cantly declined by 58%, 22%, and 52%, respectively, in these 3 countries (Table 2). In England and Wales, where vaccine sales were negligible, PCV7-type IPD increased by 25%, but the proportion of IPD caused by PCV7-type decreased slightly, from 75% to 70% (p = 0.004). In older children, PCV7-type IPD showed no clear trend, except in Belgium, where it signifi cantly decreased ( Figure 2; Table 3).
In contrast, incidences of NVT signifi cantly increased in all 4 countries during the 10-year period (p<0.001), Serotype 1 disease increased signifi cantly in each country during the 10-year period in both age groups ( Figure 3). In children <5 years of age, the increase in serotype 1 began before PCV7 sales began. However, the largest increases occurred from 1999-2002 to 2005-2006, when incidence increased by 2-4-fold in both age groups in all countries (Tables 2, 3). Increases in the incidence of serotype 1 did not correlate signifi cantly with PCV7 sales, except in France (Pearson r = 0.903, p = 0.036). In the 5-14year age group, incidences were lower, but in 2005-2006, serotype 1 constituted in average 50% of IPD in that age group compared with 13% in children <5 years of age. All serotype 1 isolates were susceptible to penicillin. Incidence and proportion of erythromycin-resistant serotype 1 was low but increased in Belgium in 2004-2006.

Serotype 7F
Most (74%) serotype 7F cases occurred among children <5 years of age. In this group, IPD increased substantially in each country, mostly during 2004-2006, and correlated signifi cantly with PCV7 sales, except in France (Pearson r = 0.901, p = 0.037 in Spain; r = 0.988, p = 0.002 in Belgium; r = 0.965, p = 0.008 in England and Wales; and r = 0.746, p = 0.148 in France) (Figure 4). From the prevaccine period to 2005-2006, incidence increased most in Spain and least in England and Wales (Table 2). In children 5-14 years of age, incidence rates also increased ( Table 3), but numbers of cases were small. All serotype 7F isolates were susceptible to penicillin, and only 6/315 isolates in children <5 years of age were erythromycin resistant.

Serotype 19A
Serotype 19A disease affected predominantly children <5 years of age (94% of cases), for whom incidence rates more than doubled (IRR range 2.4-6.8) over the period ( Figure 5). Incidence had already increased before PCV7 sales started in Belgium, Spain, and England and Wales, but the trend was signifi cant only in Belgium (during 1996-2004). After PCV7 marketing, increases correlated significantly with vaccine sales in Spain and Belgium (Pearson r = 0.929 and 0.884, p = 0.022 and 0.047, respectively). From the prevaccine period to 2005-2006, 19A incidence signifi cantly increased in Spain, France, and Belgium (Table 2); in England and Wales, the 27% increase was not signifi cant. In children 5-14 years of age, numbers of cases were too small to identify any signifi cant change.
Serotype 19A isolates showed high and increasing levels of antimicrobial drug resistance in Belgium, Spain, and particularly in France (data unavailable from England and Wales). In the <5-year age group, the prevalence of nonsusceptible strains ranged 0%-50% in 1996-1997 and increased in 2005-2006 to 21%, 48%, and 86% for penicillin and 67%, 61%, and 77% for macrolides in Belgium, Spain, and France, respectively. Full penicillin resistance was rare (0%-6% of isolates). Spain and Belgium shared similar patterns: incidence of penicillin-susceptible 19A increased more than resistant strains; incidence of erythromycin-resistant strains increased more than erythromycin-susceptible strains ( Figure 6); in 1998-2006, the use of penicillin in ambulatory setting, calculated in defi ned daily doses per 1,000 inhabitants per day, increased slightly, and macrolide use declined by 31%-38%. In France, where nonsusceptible isolates predominated, use of penicillin and macrolides initially was much higher than that in the other countries but declined (-17% and -37%, respectively) until 2006. In England and Wales where use of antimicrobial drugs was initially lower, macrolide use became similar to Belgium and Spain from 2004 onward, but data on resistance were not available.

Evolution of Meningitis
Numbers of meningitis cases caused by individual serotypes were small, especially for serotype 1. In children <5 years of age, the incidence of meningitis from serotypes 19A and 7F combined also increased signifi cantly in Spain, Emerging Infectious Diseases • www.cdc.gov/eid • Vol. 16

Discussion
This study compared the dynamics of NVT disease in 4 countries before universal PCV7 vaccination. In Spain, Belgium and France, serotype 1, 7F and 19A incidence increased considerably under rising (though moderate) vaccine coverage. In England and Wales, where PCV7 use was negligible, serotype 1 disease increased substantially, 7F disease rose less than in the other countries, and 19A disease increased nonsignifi cantly. The proportion of IPD caused by serotype 19A remained stable in England and Wales, and the incidence of meningitis caused by 19A and 7F did not change signifi cantly, suggesting that an increase in case detection caused part of the changes in 19A and 7F incidence. Indeed, a study showed that the 21% increase in IPD incidence in southwest England during 1996-2005 was no longer observed after adjustment for annual bloodculturing rates (24).
In Spain, the marked increase in NVT occurred concomitantly with PCV7 use, which led several studies to conclude that vaccine-induced replacement of serotypes had largely contributed to this increase (6,11). On the basis of our study fi ndings, we suggest that vaccine use contributed to the increase in serotypes 7F and 19A. Both serotypes increased markedly under increasing PCV7 coverage and remained stable or increased less in the countries not using PCV7. Increases in 19A and 7F incidences correlated signifi cantly with increasing vaccine sales in Spain and Belgium and were more pronounced in children <5 years of age, at whom PCV7 was aimed, than in older children. In Belgium and Spain, where both serotypes had already increased before PCV7 introduction, its slope escalated after PCV7 use. Although these data were observed under moderate PCV7 coverage, recent data from Belgium, England and Wales, and France indicate additional increases in the adjusted incidences of serotypes 7F and 19A under universal vaccination (2007-08) and high vaccine coverage (25)(26)(27).
However, we also suggest that vaccine-induced serotype replacement alone cannot explain the increase in NVT. First, serotype 1 rose well before PCV7 marketing in Belgium and Spain, affected predominantly older age groups, and increased in England and Wales in the absence of vaccine use. Second, serotype 19A increased in Belgium  and Spain before PCV7 use. Similarly, serotype 7F or 19A disease also increased in countries not using PCV7 (28)(29)(30). Third, some EU countries with widespread PCV7 use did not experience similar rises (31,32).
Other factors most likely contributed to the increases. Cyclical trends of serotype 1 were described in Scandinavian countries before any PCV7 use (3,33,34). A wave of serotype 1 (and possibly 7F) may have occurred in these 4 neighboring countries. Conversely, the high use of antimicrobial drugs, especially macrolides, allegedly favored the increase of nonsusceptible serotype 19A (14,35). A modeling study suggested that use of antimicrobial drugs played a larger role than did PCV7 use in the increase of resistant 19A in the United States (36). In 3 countries in our study where 19A incidence (and resistant strains) increased, use of antimicrobial drugs was higher than in England and Wales where 19A stayed stable. However, incidence of penicillin-and erythromycin-susceptible 19A strains also increased in Spain and Belgium. Macrolide use decreased 37%-41% during 1998-2006 in the 3 countries, whereas 19A incidence increased 80%-253%. Similarly, serotype 19A incidence increased in England and Wales after uni-versal PCV7 vaccination despite stable macrolide use (26). The role of antimicrobial drugs is thus diffi cult to delineate and suggests a synergistic effect of antimicrobial drugs and PCV7. Other factors for replacement have been raised (high prevalence of NVT carriage and low vaccine coverage), but they confl ict with current knowledge (37): 7F is a rarely carried serotype, and 19A and 7F increased further under higher PCV7 coverage. These confl icting views suggest that factors leading to replacement disease are still not fully understood; its cause is probably multifactorial and population dependant.
Our study has several limitations. First, enhancement of pediatric IPD surveillance and possible changes in blood culture practices could not be completely controlled by our methods of adjusting for underreporting. This limitation certainly applied to England and Wales, where reconciliation of 2 datasets could not totally adjust for the increase in case reporting and blood culturing, which most likely contributed to the increased incidence in nearly all serotypes studied, probably leading to overestimation of the NVT increase in England and Wales. Increase in blood culturing over time in the other countries is not suggested by the sharp   decrease in PCV7-type IPD, the similar trends in serotypespecifi c meningitis incidence (based on CSF isolation), and data on blood cultures in hospitals in Belgium (+13%  from 1999-2002 to 2005-2006 while NVT IPD increased  210%). Second, missing serotype data (more frequent in the prevaccine period) may have led to imprecision in serotype distributions; however, similar age-specifi c PCV7type and NVT distributions and trends were observed in other studies in Belgium, France, Spain, and England (5)(6)(7)(8)24), PCV7 serotype coverage did not vary with the geographic origin of pneumococcal strains in France (7), and the age and sample distribution of children for whom serotype information was available did not differ from that of other children in the Belgium dataset. Finally, estima-tion of vaccine coverage assumed that all PCV7 doses were administered at an average schedule of 3 doses for children <2 years of age. This method may overestimate PCV7 coverage because a proportion of children are likely to be incompletely vaccinated given the high cost of PCV7 paid for by parents (38), but it may also underestimate coverage because many children received fewer doses in catch-up vaccination. However, coverage values were close to those estimated by population surveys (6,7,11,38).
Such an ecologic study cannot determine which rise in disease incidence is attributable to vaccine, secular trends, or use of antimicrobial drugs, and other possible factors may have contributed. However, the strength of this study is in the comparison of epidemiologic changes in 4 countries showing variations in serotype dynamics, vaccine use, and antimicrobial drug use.
The increase in incidence of serotypes 1, 7F, and 19A has partly countered the positive impact of PCV7 on overall IPD incidence in the fi rst 2-5 years of nonuniversal vaccine use in Belgium, France, and Spain. The new 10-valent (1 and 7F) or 13-valent (1, 7F, and 19A) conjugate vaccines include these serotypes. However, a better understanding of serotype dynamics and contribution of vaccine and antimicrobial drug use is essential to guide decisions on the implementation of new vaccines and to assess their impact. Multicountry studies are useful for comparing serotype dynamics among population groups that have different levels of vaccine and antimicrobial drug use, but analyses should account for underreporting and prevaccine trends. this study was conducted and now works for the Belgian Health Care Agency and as a freelance epidemiologist. Her research interests include the epidemiology of vaccine-preventable diseases and the assessment of vaccination impact.