Impact of 13-Valent Pneumococcal Conjugate Vaccine Used in Children on Invasive Pneumococcal Disease in Children and Adults in the United States: Analysis of Multisite, Population-based Surveillance

Background

Streptococcus pneumoniae, or pneumococcus, is a major cause of morbidity and mortality globally. In 2000, a 7-valent pneumococcal conjugate vaccine (PCV7, Prevnar®, Wyeth) was introduced into the routine infant immunization program in the U.S., using a schedule of doses at 2, 4, 6, and 12–15 months of age.¹ Rates of invasive pneumococcal disease (IPD) caused by PCV7 serotypes declined dramatically among children. Because PCV7 also prevented transmission of PCV7 serotypes, rates of IPD among unvaccinated groups also declined.² PCV7 was also linked to reductions in otitis media visits³ and pneumonia hospitalizations.⁴ During subsequent years, serotype replacement resulted in increases in non-PCV7-type IPD that were modest relative to reductions in PCV7-type IPD.⁵ Despite these reductions, pneumococcus caused about 4 million episodes of disease in the U.S. resulting in $7·7 billion in direct and indirect costs in 2004.⁶

In 2010, a 13-valent conjugate vaccine (PCV13, Prevnar-13®, Pfizer) replaced PCV7.^7,8 PCV13 included serotypes causing replacement disease in the U.S. and was licensed without a randomized clinical trial. Post-licensure evaluation was, thus, the first opportunity to evaluate the effects of PCV13 on prevention of IPD. Our objectives were to evaluate the population-level impact of PCV13 on incidence of IPD among all ages and to assess whether PCV13 introduction was associated with serotype replacement.

Methods

We used a long-standing surveillance system to compare rates of IPD before and after PCV13 introduction. We identified IPD cases through Active Bacterial Core surveillance (ABCs), an active population- and laboratory-based surveillance system that is part of CDC’s Emerging Infections Program. The methodology used by ABCs is described in full at http://www.cdc.gov/abcs/index.html. We included cases identified from July 1, 2004 through June 30, 2013 in ten continuously participating ABCs sites: selected counties in California, Colorado, Georgia, Maryland, New York, Oregon, and Tennessee, and the states of Connecticut, Minnesota, and New Mexico. The total population under surveillance was nearly 30 million.

We defined IPD cases as isolation of S. pneumoniae from normally sterile sites (e.g., blood, cerebrospinal fluid). Laboratory audits ensured completeness of reporting. PCR for diagnosis of IPD is not uniformly available in the U.S. and such cases are not captured by ABCs. Medical records were reviewed to obtain demographic and clinical information. Isolates were serotyped by Quellung at CDC’s Streptococcus Laboratory or the Minnesota Department of Health Laboratory. For our analysis, we assigned serotypes to the following categories: 1) PCV7-types (4, 6B, 9V, 14, 18C, 19F, 23F, and 6A), 2) PCV13/nonPCV7-types (serotypes 19A, 7F, 5, 3, and 1, which are included in PCV13 but are not affected by PCV7), 3) PPV11 types (serotypes included in 23-valent pneumococcal polysaccharide vaccine [PPV23, PNEUMOVAX 23®, Merck] but not in PCV13: 2, 8, 9N, 10A, 11A, 12F, 15B, 17F, 20, 22F, and 33F) and 4) non-PCV13-types (types not included in PCV13). Note that categories 3 and 4 overlap. Although serotype 6A is included in PCV13 and not in PCV7, we treated it as a PCV7 serotype because of documented cross-reactivity and disease reduction associated with the 6B antigen in PCV7⁹. Antimicrobial susceptibility testing against penicillin, erythromycin, clindamycin, trimethoprim-sulfamethoxazole, tetracycline, chloramphenicol, levofloxacin, and vancomycin was performed using broth microdilution and isolates were classified as susceptible, intermediate, or resistant according to published guidelines¹⁰. Meningitis breakpoints for penicillin were used for meningitis cases; non-meningitis breakpoints were used for all other cases. For all antibiotics, we combined intermediate and fully resistant strains into a “nonsusceptible” category. Any isolate nonsusceptible to three or more classes was considered “multiply nonsusceptible”.

We calculated case-fatality ratios (CFR) as the proportion of cases with fatal outcomes among those with known outcomes (>99% of all cases). Comorbid conditions were collected as per the ABCs protocol¹¹ and classified according to recommendations of the Advisory Committee on Immunization Practices (ACIP).^12,13

ABCs case reporting and isolate collection were considered to be surveillance activities and were exempt from CDC institutional review. The protocol was also assessed for review at each site and, when necessary, institutional review board approval was obtained. Informed consent was not required.

We estimated vaccination coverage using immunization information systems (IIS) which are confidential, population-based systems that consolidate data from vaccine providers. As a proxy for coverage in ABCs areas, we used IIS sentinel sites located in Michigan, Minnesota, North Dakota, New York City, Oregon, and Wisconsin that collectively include approximately 2·0 million children aged <5 years. We used SAS® (version 9.3, SAS Institute, Inc., Cary, NC) and Excel® 2010 (Microsoft Corp., Redmond, Washington) to calculate unweighted intra-site mean PCV13 coverage based on 2011 post-Censal population estimates and vaccination records from IIS sentinel sites (queried 2 February 2013). PCV13 primary series coverage estimates included doses of PCV13 administered before 12 months of age to children born during 1 July 2010 through 1 July 2011. Post-primary PCV13 booster dose coverage estimates include doses of PCV13 administered before 19 months of age to children born during 1 July 2010 through 1 December 2010. PCV13 supplemental dose estimates include PCV13 doses administered to children aged 14–59 months born 1 July 2007 through 1 May 2009 who previously completed a routine or catch-up schedule recommended by the ACIP.¹⁴¹¹

Results

During July 2004 through June 2013, we identified 33,688 IPD cases; 89% had serotyping results. The prevalence of at least one underlying condition (apart from age) that is an indication for PCV13 or PPV23^12,16 increased slightly among children and adults with IPD (Table 1) after the time of PCV13 introduction. The proportions of cases resulting in hospitalization were also marginally higher in the latter period in both groups while case-fatality rates did not change. Finally, the proportions of cases caused by specific clinical syndromes changed modestly in each age group.

During July 2010 through June 2012, mean coverage with ≥3 PCV13 doses administered before 12 months of age was 76% (range 67%–90%) among age-eligible children; 65% (range 55%–83%) of all age-eligible children received ≥3 doses of PCV13 administered before 12 months and a booster dose of PCV13 during ages 12 through 18 months. Among children 14–59 months who had received a complete schedule of PCV7, a mean of 63% (range 40%–88%) received the ACIP-recommended supplemental dose of PCV13 during July 2010 through June 2012.

Discussion

Our analysis demonstrates substantial and rapid reductions in IPD within three years of the introduction of PCV13 in the U.S. The serotypes most affected were those most common before PCV13 introduction, particularly serotypes 19A and 7F, and the age groups that experienced the earliest reductions in PCV13/nonPCV7-type IPD were the groups targeted for vaccination: children less than five years old. These reductions became evident rapidly—within six months of PCV13 introduction—possibly because PCV13 was introduced into the routine infant immunization schedule as a simple replacement for PCV7, a vaccine with over 80 percent coverage by 2008, along with substantial use of the supplemental dose of PCV13 among toddlers and other children <5 years old.¹⁸

We found a reduction in IPD in adults associated with PCV13 introduction in children. In all adult age groups, PCV13/nonPCV7-type IPD (especially serotypes 19A and 7F) declined by 58–72% (comparable to that observed early after PCV7 introduction), leading to overall reductions of IPD of 12–32%. These findings are consistent with the hypothesis that PCV13 prevents nasopharyngeal colonization with serotypes 19A and 7F among children, and therefore, transmission of these types between children and adults.¹⁹ Similar to the experience with PCV7, the reductions we observed among adults became evident very soon—by January 2011 in 18–49 year-olds—after vaccine introduction for children. For these reductions to be attributed to PCV13 use among adults, PCV13 would have to be licensed, recommended, and implemented by early 2011. PCV13 was not licensed for adults until December 2011 and ACIP refrained from recommending its use for adults until October 2012 and, even then, only for adults with immunocompromising conditions.¹² Therefore, we believe that PCV13 use among adults cannot explain our findings. In September 2014, ACIP recommended use of PCV13 for all adults 65 years of age and older but with the caveat that this recommendation should be revisited in 2018,¹³ primarily because of the large indirect effects demonstrated here. Continued monitoring of disease among adults will assist in determining whether this recommendation should be continued.

Comparing the period after PCV13 introduction to the period before, we observed modest increases in case-fatality ratios and in the proportion of cases with underlying conditions. Among pediatric cases, we observed a trend toward reduced prevalence of pneumonia²⁰ compared to other syndromes. Among adults, the proportion of cases resulting in bacteremia declined after PCV13 introduction compared to before. Potential explanations for these changes include differential effects on individual serotypes that may predispose to certain syndromes and differential effectiveness among children with and without underlying conditions.

We were unable to assess a reduction in IPD caused by serotype 3 (included in PCV13 but not PCV7). One recent publication suggests a reduction in serotype 3 IPD cases seen in 8 pediatric hospitals²¹ while data from a large, national surveillance program in England and Wales suggest no evidence of effectiveness of PCV13 against serotype 3.^19,20 A recent randomized controlled trial comparing immunogenicity and efficacy of PCV13 and PCV7 against nasopharyngeal colonization suggests no effect of PCV13 on nasopharyngeal colonization with serotype 3.¹⁷ Definitive evidence of effectiveness of PCV13 against serotype 3 colonization and IPD requires more study.

Serotype replacement has been documented since PCV7 introduction.²² A recent review of serotype replacement disease following pneumococcal conjugate vaccine introduction from multiple surveillance programs around the world indicated that serotype replacement would not be expected within two years of PCV13 introduction²¹.²² Indeed, we observed some evidence of serotype replacement but only among adults 50–64 years of age and only during the third year after PCV13 introduction (Table 2). Early evidence of serotype replacement may be emerging in Europe.²³ Importantly, we did observe reductions in antibiotic nonsusceptible IPD that are largely attributable to reductions in IPD caused by serotype 19A, the serotype associated with increased antibiotic nonsusceptibility before PCV13 introduction²⁴.²⁴ Vaccination is an important tool in combatting antimicrobial resistance.

To quantify the effect of PCV13 introduction, we used a potential outcomes modeling approach^25,26 with advantages over “before-after” comparisons.² First, rates of serotypes 19A and 7F were increasing in ABCs areas before PCV13 introduction. Therefore, estimates of PCV13 impact would depend, in part, on the arbitrary selection of baseline incidence. A high baseline incidence rate would lead to an overestimate of effect while a low baseline rate would underestimate the effect. Second, our method takes into consideration all data points during the period of observation, not just those chosen as the “before-after” comparison points. Finally, this method leads directly to an estimate of cases prevented, an estimate previously derived indirectly.²

Our analysis does have certain limitations. For children who had already received a full series of PCV7, a single supplemental dose of PCV13 was recommended²⁷.²⁷ We are unable to assess the relative contributions of the full 4-dose series vs. the supplemental dose. Our model assumed that the incidence of IPD caused by PCV13/nonPCV7 serotypes would have continued increasing after PCV13 introduction. While this assumption is reasonable for the first 2–3 years after introduction, experience with epidemics of serotype 1 suggests that some population-level immunity is achieved within a few years after the epidemic starts. Thus, we cannot expect that rates of PCV13/nonPCV7-type IPD would have continued to increase indefinitely. Additionally, as with any national estimates from limited geographical surveillance, it is possible the ABCs areas are not fully representative of the entire country. However, given the high variability in ABCs areas with respect to geography, socio-economic status, and urbanicity, this is unlikely to change our conclusions substantially.

In summary, PCV13 has already demonstrated dramatic reductions in IPD among children and, through herd protection, adults. In the first three years alone, an estimated 30,000 cases of invasive disease and 3,000 deaths have been prevented. The continued success of the pediatric PCV13 program will be critical for policy-making related to the recently adopted age-based recommendations for PCV13 in adults.

Supplementary Material