All study participants were children <18 years of age who had IPD. Pneumococcal isolates from normally sterile clinical samples, such as blood, cerebrospinal fluid, pleural effusions, and joint fluid, were collected and examined.

Medical institutions that had a microbiology laboratory and a pediatric department with hospitalization facilities were permitted to participate actively in surveillance. An estimated 25% of all general hospitals in Japan participated. Participating hospitals were distributed nearly uniformly throughout Japan (Figure 1). These hospitals took part in the surveillance project after the laboratory director or hospital director granted permission in writing. A questionnaire collecting information for every case-patient was completed anonymously in accordance with the ethical guidelines for conducting epidemiologic studies in Japan.

Pneumococcal isolates were collected nationwide during 3 periods. The first surveillance period was April 2010–March 2011, when voluntary vaccination with PCV7 was estimated to be <10% (vol-PCV7: 2010). The second period was April 2011–March 2012, when the estimated vaccination rate was 50%–60% because of the Urgent Promotion of Vaccination incentive (post-PCV7: 2011). The third period was April 2012–March 2013, when the vaccination rate was 80%–90% just before introduction of PCV13 (pre-PCV13: 2012). We collected a total of 602 isolates from 341 general hospitals: 300 in 2010, 146 in 2011, and 156 in 2012.

Serotypes of all isolates were determined by the capsular quellung reaction using antiserum purchased from the Statens Serum Institute (Copenhagen, Denmark). Alterations in 3 penicillin-binding protein (PBP) genes mediating β-lactam resistance in S. pneumoniae—pbp1a (PBP1A), pbp2x (PBP2X), and pbp2b (PBP2B)—were identified by real-time PCR methods that we developed and reported previously (21). The PCR system detected the presence of amino acid substitution(s) in conserved amino acid motif(s), such as serine-threonine-methionine-lysine, in each PBP. The genes mef (A) and erm (B), which confer resistance to macrolide antimicrobial drugs, also were identified by real-time PCR (21).

Genotype (g) based on molecular analysis is represented here as penicillin-susceptible S. pneumoniae (gPSSP) possessing 3 normal pbp genes; penicillin-intermediate S. pneumoniae (gPISP), further classified as gPISP (pbp2x), gPISP (pbp1a+pbp2x), or gPISP (pbp2x+pbp2b); or penicillin-resistant S. pneumoniae (gPRSP), possessing all 3 abnormal pbp genes (22). Relationships between susceptibilities to β-lactam agents among phenotypes of S. pneumoniae and resistance genotype were described previously (21). Genotypes involving macrolide resistance are represented here as the following: macrolide-susceptible S. pneumoniae not possessing any genes (MLS); intermediate macrolide resistance to 14- or 15-membered macrolides mediated by the mef(A) gene (MLR-mef[A]); high macrolide resistance to all macrolides mediated by the erm(B) gene (MLR-erm[B]); and high macrolide resistance possessing both genes (MLR-mef[A]+erm[B]).

For each strain transferred to our laboratory, we promptly identified capsular type by quellung reaction, and resistance genotype was determined by real-time PCR. Results were returned immediately to medical staff at the referring hospital. Medical personnel caring for the patients considered this surveillance and results reporting system very helpful, and it was continued for 3 years.

Multilocus sequence typing for the strains collected was performed according to previously described methods with slight modifications (23). Primers listed on the website of the Centers for Disease Control and Prevention (http://www.cdc.gov/ncidod/biotech/strep/alt-MLST-primers.htm) were used. Multilocus sequence typing and eBURST analyses were performed according to published methods (http://spneumoniae.mlst.net).

We assessed statistical significance of differences in serotype distribution between the 3 periods, age groups, and specific infectious diseases. We used χ² tests or the Fisher exact test, using Ekuseru-Toukei 2012 software for statistics (Social Survey Research Information Co., Ltd., Tokyo, Japan).

Estimated rates of vaccination with PCV7 were <10% in 2010 (voluntary-PCV7) but rose to 50%–60% with funding in 2011 (post-PCV7) and to 80%–90% with enhanced implementation of PCV7 just before the transition from PCV7 to PCV13 in 2012 (pre-PCV13). S. pneumoniae isolates from patients with IPD, collected every year, decreased by half in 2011 and 2012 from those in 2010. In particular, vaccine serotypes decreased significantly among patients <4 years of age (p<0.001) but not among those >5 years of age (p = 0.733) (Table 1). For all patients, coverage rate by PCV7 decreased rapidly from 73.3% in 2010 to 54.8% in 2011 to 14.7% in 2012, and prevalence proportion of nonvaccine serotypes increased for 3 years (p<0.001).

We compared year-to-year changes in prevalence of vaccine and nonvaccine serotypes of S. pneumoniae in patients who had meningitis; sepsis and bacteremia; pneumonia; and other invasive infections, including cellulitis, arthritis, endocarditis, and empyema during each of the 3 years studied (Table 2). Pneumonia cases were included only when S. pneumoniae was isolated from blood cultures. Vaccine serotype strains decreased significantly in patients with meningitis (p = 0.006), sepsis and bacteremia (p<0.001), and pneumonia (p<0.001).

We also determined correlations between changes of serotypes and resistance genotypes in all 602 isolates for each period (Figure 2). These resistance genotypes were classified according to real-time PCR results concerning 3 PBP genes: pbp1a, pbp2x, and pbp2b.

Overall, gPRSP and gPISP (pbp1a+pbp2x), respectively, declined from 54.7% and 14.3% in 2010, to 47.3% and 8.2% in 2011, to 26.3% and 5.1% in 2012. Among them, vaccine serotypes 6B, 14, 19F, and 23F, including gPRSP and gPISP (pbp1a+pbp2x), which showed high frequency in 2010, decreased significantly during the 2 subsequent periods (p<0.001). Of serotypes contained in PCV13, serotype 6A, showing cross-protective immunity with 6B, also decreased markedly in 2012, but prevalence of serotype 19A increased. Serotypes 1, 3, and 7F included a very small number of isolates.

Coverage rate of PCV13 decreased rapidly from 89.0% in 2010 to 72.6% in 2011 to 39.1% in 2012 (p<0.001). Nonvaccine serotypes increased, especially serotypes 24 and 33F, identified as gPSSP; 15B, 15C; and 22F, gPISP (pbp2x); and 15A and 35B, gPRSP.

Main clonalities within major nonvaccine serotypes were identified as sequence type (ST) 3111 and ST2331 in serotype 19A, ST63 in serotype 15A, ST199 in serotypes 15B and 15C, ST433 in serotype 22F, ST338 of clonal complex (CC) 156 in serotype 23A, ST5496 of CC2572 in serotype 24, ST717 in serotype 33F, and ST558 in serotype 35B. STs in gPRSP among nonvaccine serotypes were as follows: major strains of serotype 15A belonged to ST63; serotype 15C belonged to ST83 of CC81; serotype 35B belonged to ST558; and serotype 16F belonged to ST8351 of CC3117.

Macrolide resistance was classified into 4 groups: MLS, MLR-mef(A), MLR-erm(B), and MLR-mef(A)+erm(B). Comparing results between 2010 and 2012, proportions of resistance types of MLR-mef(A), MLR-erm(B), and MLR-mef(A)+erm(B) in vaccine serotypes decreased, whereas that of MLR-erm(B) in nonvaccine serotypes increased significantly (p<0.001) (Figure 3).

Relationships between serotypes and macrolide resistance genes changed from year to year (Figure 4). In contrast to decreases in vaccine serotype strains possessing erm(B) or mef(A) genes, nonvaccine serotypes 15A, 15B, 15C, and 24, which possess the erm(B) gene mediating high macrolide resistance, increased. Almost all strains of serotype 19A possessed both genes erm(B) and mef(A).

As for increases and decreases in the proportion of each serotype isolated from patients <5 years of age from 2010 (vol-PCV7) to 2012 (pre-PCV13), vaccine serotype strains 6B, 14, 19F, and 23F, and 6A (which shows cross-protective immunity with 6B, occurring frequently among IPD cases) decreased markedly (p<0.001 for 4 vaccine serotypes, p = 0.005 for 6A). Serotype 19A, included in PCV13, increased (p<0.001). Other serotypes included in PCV7 or PCV13, except for 6A and 19A, changed minimally. Overall, proportions of nonvaccine serotypes not covered by PCV13 increased, particularly serotypes 15A, 15B, 15C, and 24 (each p<0.001), and 22F (p = 0.004).

Only 23 patients had infections caused by strains of vaccine serotypes (14.7%), 21 of whom had not received PCV7 (Table 3). Of the remaining 2 patients, a 1-year-old child infected with a strain of serotype 6B had received 2 doses of vaccine. The other patient, 5 years of age, was infected with a strain of serotype 19F; vaccination history was unknown.