Risk Factors for Invasive Pneumococcal Disease in Children in the Era of Conjugate Vaccine Use
OBJECTIVE: We conducted a case-control study to evaluate risk factors for invasive pneumococcal disease (IPD) among children who were aged 3 to 59 months in the era of pneumococcal conjugate vaccine (PCV7).
METHODS: IPD cases were identified through routine surveillance during 2001–2004. We matched a median of 3 control subjects to each case patient by age and zip code. We calculated odds ratios for potential risk factors for vaccine-type and non–vaccine-type IPD by using multivariable conditional logistic regression.
RESULTS: We enrolled 782 case patients (45% vaccine-type IPD) and 2512 matched control subjects. Among children who received any PCV7, children were at increased risk for vaccine-type IPD when they had underlying illnesses, were male, or had no health care coverage. Vaccination with PCV7 did not influence the risk for non–vaccine-type IPD. Presence of underlying illnesses increased the risk for non–vaccine-type IPD, particularly among children who were not exposed to household smoking. Non–vaccine-type case patients were more likely than control subjects to attend group child care, be male, live in low-income households, or have asthma; case patients were less likely than control subjects to live in households with other children.
CONCLUSIONS: Vaccination with PCV7 has reduced the risk for vaccine-type IPD that is associated with race and group child care attendance. Because these factors are still associated with non–vaccine-type IPD risk, additional reductions in disparities should be expected with new, higher valency conjugate vaccines.
WHAT'S KNOWN ON THIS SUBJECT:
Before 7-valent conjugate vaccine introduction in the United States, children of certain racial and ethnic groups, those with chronic medical conditions, and children who attended group child care were known to be at increased risk for pneumococcal disease.
WHAT THIS STUDY ADDS:
Vaccination with 7-valent pneumococcal conjugate vaccine eliminated disparities in risk of invasive pneumococcal disease associated with race and group childcare attendance.
Before 7-valent conjugate vaccine (PCV7) introduction in the United States in 2000, the highest rates of invasive pneumococcal disease (IPD) were reported among children who were younger than 2 years.1,2 Children of certain racial and ethnic groups,1,3,–,5 those with chronic medical conditions,1,6 and children who attend group child care7,8 are known to be at increased risk for pneumococcal disease.
The Advisory Committee for Immunization Practices (ACIP) recommended PCV7 for all children who were younger than 2 years and children who were aged 2 to 4 years in high-risk groups, including children with certain underlying illnesses6; those who attend group child care; or black, American Indian, and Alaska Native children. In 2007, ACIP updated the recommendations to include all children who are aged 2 to 59 months.9 A dose of 23-valent polysaccharide vaccine after the PCV7 series is recommended for children who are aged 2 to 4 years and have certain underlying illnesses.
PCV7 introduction in the United States led to dramatic reductions in IPD burden.10,–,12 Postlicensure studies have demonstrated high effectiveness against IPD in the general population of children.13 Because a new conjugate vaccine targeting 13 serotypes was recently licensed in the United States, an assessment of risk factors for IPD is needed to evaluate which children continue to develop IPD despite receiving PCV7. We evaluated risk factors for IPD among children who were aged 3 to 59 months as part of a matched case-control study that was designed to evaluate effectiveness of PCV7 after its widespread uptake in the United States.13
The study was conducted in sites that participate in Active Bacterial Core surveillance (ABCs), an active population-based and laboratory-based system. Children who were aged 3 to 59 months and residing in select counties in California, Colorado, Georgia, Minnesota (entire state beginning 2002), New York, Oregon, and Tennessee and the entire state of Connecticut were eligible for participation. According to 2002 postcensus population estimates,14 the surveillance areas included ∼1.7 million children who were younger than 5 years (698 960 children who were younger than 2 years and 991 499 children who were aged 2 to 4 years).
IPD cases, defined as isolation of Pneumococcus from a normally sterile site (eg, blood, cerebrospinal fluid, pleural fluid) of a surveillance area resident, were identified through routine ABCs procedures.15 Active contact with all participating microbiology laboratories and regular laboratory audits ensured completeness of reporting. Site personnel reviewed medical records to obtain limited clinical information. Isolates were serotyped at reference laboratories (Centers for Disease Control and Prevention [CDC] and Minnesota Department of Health).
Children who were aged 3 to 23 months and had IPD onset from January 1, 2001, through June 30, 2003, and children who were aged 24 to 59 months and had IPD onset from January 1, 2001, through May 31, 2004, and met ABCs case definition, for whom a pneumococcal isolate was available for serotyping, and whose parents or guardians gave informed consent to participate were eligible for enrollment as case patients. The enrollment period for children who were aged 24 to 59 months was extended to increase a sample size for the evaluation of catch-up schedules. Children with previous IPD episodes were excluded from the study.
Selection of Control Subjects
Children who were aged 3 to 59 months and living in a surveillance area on the culture date of the matching case child were eligible for enrollment as control subjects. For each enrolled case child, we obtained from the birth certificate registry a list of 15 (20 in Colorado) children who were born within 14 days of the case child's birthday and whose mother's residence at the time of birth was within the same zip code as the case child's residence. We followed a standard protocol and had to exhaust 3 telephone numbers or address search methods and 15 telephone call attempts to a child's parent or guardian made at different times and on different days before excluding a potential control subject. We sought to enroll 3 control subjects per case patient but contacted parents or guardians of several potential control children simultaneously; however, children who were further removed by birth date from a matched case child's birthday could not be used in the analysis until potential control subjects who were born closest to the case child's birthday were either enrolled or approached for enrollment but excluded. Children were not eligible to be control subjects if they had IPD, they had a sibling who was enrolled in the study, or their parent or guardian declined to participate.
We conducted telephone interviews of parents or guardians of case and control children by using a standardized questionnaire that addressed the child's demographics (eg, race, gender), medical history (eg, birth weight, underlying illnesses, recent infections and antimicrobial use), household characteristics (eg, household size and income, parent's educational level), exposure to tobacco smoke, group child care attendance, breastfeeding history, vaccination history (when written records were available), and contact information for child's health care providers. For exposures that may change with time (eg, tobacco smoke, breastfeeding, child care attendance), the questions focused on the month before illness for case patients and their matched control subjects. Information on underlying illnesses (eg, asthma, cystic fibrosis, sickle cell disease, heart disease, chronic lung conditions, immunosuppressing conditions), recent infections, recent antimicrobial use, and the childhood vaccines received was collected from health care providers from whom children reportedly received routine medical care and immunizations. In Tennessee, Georgia, and Oregon, vaccination registries were used to verify vaccination histories. The study protocol was approved by institutional review boards at the CDC and in each of the ABCs sites.
Data were aggregated at the CDC and analyzed by using SAS statistical software.16 We used Mantel-Hansel χ2 to compare characteristics of children who were enrolled with those who were not enrolled, on the basis of information collected through routine surveillance. We used multivariable conditional logistic regression models to calculate odds ratios (ORs) for potential risk factors for IPD caused by the serotypes that are included in PCV7 (PCV7 types: 4, 6B, 9V, 14, 18C, 19F, and 23F) and IPD that is caused by all other serotypes (non-PCV7 types). We included interaction terms in the multivariable model to evaluate risk factors separately among vaccinated (those who received ≥1 PCV7 dose) and unvaccinated (those who received no PCV7). We combined children who received any PCV7 into a vaccinated group, because our previously published data13 showed high effectiveness of ≥1 PCV7 dose in healthy children. We assessed potential confounders, 2-way interactions, and co-linearity in a multivariable model. Vaccination doses that were given ≤14 days before case culture date were excluded. For all analyses, P < .05 was considered statistically significant.
During the study period, from a total of 1267 IPD case patients who were identified among children who were aged 3 to 59 months at participating sites, 1121 (88%) were study-eligible cases. We were not able to enroll 339 (30%) case patients (Fig 1A). From the remaining 782 (70%) case patients who were included in the analysis, 353 (45%) had PCV7-type IPD and 429 (55%) were caused by non-PCV7 types. Most common PCV7 types were 14 (n = 76 [10%]) and 19F (n = 70 [9%]), and the most common non-PCV7 serotypes were 19A (n = 100 [13%]) and 6A (n = 39 [5%]; Supplemental Table 5). Compared with case children who were not enrolled in the study, enrolled case patients were more likely to be white (54% vs 41%; P < .001) and less likely to have died from IPD (0.8% vs 2.7%; P = .006).
From the 8018 children who were identified as potential control subjects, 5506 (69%) could not be enrolled (Fig 1B) and the remaining 2512 (31%) were included in the analysis. The median number of control subjects was 3 per case patient (range: 1–10); 692 (88%) case patients had ≥3 control subjects enrolled, and the remaining 90 (12%) case patients had 1 or 2 control subjects enrolled.
Comparing Characteristics of Case Patients and Control Subjects
Control subjects differed from case children on a number of factors, including race, gender, breastfeeding history, household exposure to smoking, group child care attendance, and socioeconomic factors (Table 1). After controlling for the receipt of ≥1 PCV7 dose, child's race (black or other versus white), any household exposure to smoking, household crowding, low household income, education level of a caregiver (high school or less), and not having private health insurance were associated with increased risk for IPD. Among children who attended group child care, after controlling for the receipt of ≥1 PCV7 dose, risk for IPD was highest during the first 6 months of attendance and for those who attended for >10 hours per week. Among children with a history of breastfeeding, risk for IPD was lowest for children who were currently breastfed or those who had stopped breastfeeding within 1 month of case culture date (Table 1).
Fifty percent of case patients and 67% of control subjects (P < .001) received ≥1 PCV7 dose (Table 2); a higher proportion of control subjects than case patients received ≥3 doses of PCV7 or other childhood recommended vaccines. A significantly higher proportion of case patients (11%) than control subjects (4%) reported presence of underlying illnesses (Table 2). Underlying illnesses that most commonly were identified in our study population were systemic steroid use, congenital heart disease, immune system disorder or HIV/AIDS, and any type of cancer. After controlling for the receipt of ≥1 PCV7 dose, cancer, immune system disorder or HIV/AIDS, and nephrotic syndrome or renal failure had the strongest association with IPD (P < .01 for each).
Multivariable Analysis of IPD Risk Factors
Among children who received ≥1 PCV7 dose, case children with PCV7-type IPD were more likely than control subjects to report underlying illnesses, to be male, or to have no health care coverage and were less likely than control subjects to be up-to-date for vaccination against Haemophilus influenzae type b (Hib; Table 3). Group child care attendance, race, household income, or recent history of antibiotic use did not influence the risk for PCV7-type IPD among vaccinated children. Among unvaccinated children (received no PCV7), children with PCV7-type IPD were more likely than control children to have underlying illnesses, attend group child care, be male, live in low-income households, have no health care coverage, or report recent antibiotic use and were less likely than control subjects to be white or to be up-to-date for other childhood vaccines.
We also assessed risk factors for IPD that is caused by non-PCV7 serotypes (Table 4). The presence of underlying illnesses influenced the risk for non–PCV7-type IPD, and the risk was different for children who were exposed to tobacco smoke and for those with no household exposure to tobacco smoke. In households with no exposure to smoking, children with underlying illnesses were 3.3 times more likely (95% confidence interval [CI]: 2.2–5.1) to have non–PCV7-type IPD compared with children with no underlying illnesses. In households with smokers, there was no association between the presence of underlying illnesses and non–PCV7-type IPD (OR: 1.0 [95% CI: 0.5–2.3]). Smoking in the household was positively associated with non–PCV7-type IPD among children with no underlying illnesses (OR: 1.5 [95% CI: 1.1–2.0]), with 9% (95% CI: 2%–16%) of cases in this group potentially attributable to household smoking, after adjustment for all other significant factors. Children who attended group child care were at increased risk for non–PCV7-type IPD; however, the increased risk that was associated with group child care attendance was higher among black children (OR: 4.2 [95% CI: 2.4–7.5]) compared with white children (OR: 1.7 [95% CI: 1.3–2.4]; Table 4). Case patients with non–PCV7-type IPD were more likely than control subjects to live in low-income households (<$30 000), to be male, to have a history of asthma, to have been of low birth weight (<2500 g), or to have had tympanostomy tubes placed 30 days before case IPD onset. Case patients with non–PCV7-type IPD were less likely than control subjects to be up-to-date on vaccination against Hib or to live in households with children younger than 18 years (Table 4). Receipt of ≥1 PCV7 dose did not influence the risk for non–PCV7-type IPD.
Clinicians often rely on identification of risk factors to guide formulation of a differential diagnosis and patient care decisions. The primary findings of our study demonstrate that vaccination with PCV7 has changed the importance of some traditional factors that would raise suspicion for IPD. Perhaps most important, vaccination eliminated disparities in risk for PCV7-type IPD associated with race and group child care attendance, factors that in the past have been among the strongest predictors of patients who are at risk.1,3,–,5,7,8 After PCV7 introduction, IPD incidence declined substantially among black children in the United States, reducing disparities in disease incidence between black and white populations, findings confirmed by our study.17,18 Likewise, our results show that among children who received ≥1 PCV7 dose, group child care attendance did not influence the risk for PCV7-type IPD.
Dramatic reductions in PCV7-type IPD incidence after vaccine introduction and small increases in non-PCV7 types have led to the shift in the distribution of pneumococcal serotypes that cause disease, with non-PCV7 types now causing the majority of IPD cases.19,20 Our study found no increased risk for non–PCV7-type disease among vaccine recipients. This suggests that serotype replacement observed in populations after PCV7 use is occurring because of increased transmission of non-PCV7 types, not because vaccination with PCV7 directly increases individual risk for non–PCV7-type disease. Among the case patients with non–PCV7-type IPD, group child care attendance and race remained associated with increased risk for disease. These findings support efforts in several US states to mandate PCV7 for children who attend group child care.21 The remaining risk that is associated with these factors suggests that additional reduction in disparities by using higher valency conjugate vaccines should be possible; a formulation that covers 10 serotypes is licensed in a small number of countries, and a 13-valent formulation was licensed in the U.S. in February 2010.
Although children who live in low-income households may be more vulnerable to many health problems than other children, our data indicate that markers of poverty were associated with IPD independent of race and other factors. Among children who were vaccinated with ≥1 PCV7 dose, those who lacked health care coverage were at higher risk for PCV7-type IPD, suggesting that missed vaccine doses or delays in seeking care for upper respiratory infections or acute otitis media might be more frequent in children without health care coverage than among the insured. Low household income in our study was independently associated with the increased risk for non–PCV7-type disease. In addition, being fully vaccinated against Hib disease was protective against both PCV7-type and non–PCV7-type disease, most likely reflecting benefits of better access to routine health care. Analysis of PCV7 uptake in the United States revealed that children who did not receive the age-appropriate PCV7 doses were more likely than those who did to belong to minority race groups, live in low-income households, and receive vaccinations from public rather than private providers.22,23 These disparities in PCV7 uptake were related to multiple shortages that were reported during this study period,24,25 as well as to differences in individual state vaccine financing policies.26 Although ACIP recommendations for PCV7 use do not target uninsured children or those in low-income households, government programs such as Vaccines for Children should help to ensure that the vaccine is available for this vulnerable group.
Chronic medical conditions continue to be an important risk factor for IPD. Chronic conditions were a particularly strong explanation for the occurrence of PCV7-type IPD among children who had received ≥1 PCV7 dose. Clinical efficacy data for PCV7 are lacking for these high-risk groups with the exception of children with HIV infection.27 In our previous postlicensure study of vaccine effectiveness that used this same data set, PCV7 prevented IPD among children with underlying illnesses, although the effectiveness was significantly lower (86%) compared with the effectiveness measured among healthy children (94%).13 Another study of effectiveness of PCV7 among children with sickle cell disease produced a similar point estimate of effectiveness (81%–84%).28 Although the results of immunogenicity studies show adequate immune response among children with sickle cell disease29,–,31 and with HIV,32,33 the same concentrations of serotype-specific antibodies may not be as good of a correlate of protection for children with certain immunocompromising conditions compared with healthy children.
Associations between breastfeeding and decreased risk for IPD8 and between secondhand smoke exposure and increased risk for IPD7,34 have been reported in previous studies. In our study, both factors were important on univariate analysis, but breastfeeding was not significant in the multivariable models after controlling for other factors. Whether breastfeeding no longer provides a significantly protective effect in the PCV7 era, the effects of breastfeeding are small when compared with large herd effects that are associated with PCV7,35 or we had trouble accurately assessing or defining breastfeeding behavior. We did find an association between household smoke exposure and risk for non–PCV7-type IPD, in particular among children without chronic medical conditions. Because smoking status of caregivers was self-reported, the respondents' concern about providing a “correct” answer may have limited our ability to measure this exposure accurately. Nonetheless, our findings demonstrate that a proportion of cases can potentially be attributed to household exposure to smoking and provide yet another reason for health care providers to speak to parents of young children about limiting exposure to secondhand smoke in households.
This study is subject to the following limitations. First, a case-control design can be susceptible to both bias and confounding. We tried to avoid potential bias at the design stage through age and zip code matching of control subjects to case patients to control for age-specific differences in IPD risk and to control loosely for potential socioeconomic factors. In addition, we obtained information about and controlled for multiple known risk factors for IPD. Next, dramatic changes in pneumococcal serotype distribution19,20 may limit our ability to generalize some of our study findings to the current period; however, the most common non-PCV7 serotypes that are included in 10-valent and 13-valent PCV formulations accounted for 8.6% and 51.0% of non–PCV7-type IPD, respectively, among case patients who were included in our study and 9.8% and 64% of non–PCV7-type IPD, respectively, among ABCs cases in 2006–2007. Our findings on risk factors for non–PCV7-type IPD therefore remain relevant to the current time period. Last, information on household exposures was collected through parent or guardian interview, which may introduce recall and misclassification bias. We tried to limit recall bias by restricting the enrollment for those who were interviewed within 120 days after the case culture date. Misclassification of households with smokers as nonsmoking households, especially among children who had underlying conditions, may have limited our ability to find a significant association. The strength of our study was its large sample size, which allowed us to evaluate multiple potential risk factors among children with PCV7-type and non–PCV7-type IPD and those who were vaccinated with PCV7.
Vaccination with PCV7 is an effective tool for reducing excess risk for IPD among traditional risk groups such as black children and children who attend group child care. As new conjugate vaccines with broader serotype coverage become available, additional reduction in disparities in risk for IPD should be expected. Presence of underlying illnesses remains a significant risk factor for IPD, a finding that demonstrates the need to remain suspicious of IPD among children with high-risk medical conditions, even among those who have been vaccinated. In addition, many children with underlying medical conditions, even those who have already received PCV7, may benefit from the additional serotype coverage that the new, expanded-valent conjugate vaccines may provide and should be candidates for catch-up vaccination as those vaccines become available. Government vaccine financing policies that can ensure that these new vaccines are accessible for children from the most vulnerable socioeconomic groups can also help further reduce risk for IPD.
Funding for the study was provided by the CDC's Antimicrobial Resistance Working Group, the CDC's Emerging Infections Program, and the National Vaccine Program Office.
We thank the following people for contributions to data collection, analysis, laboratory work, or study direction: James Howgate, Laura Rainer, Kathryn E. Arnold, and Wendy Baughman, Georgia Emerging Infections Program; Catherine Muehlenbein, Maria Thomas, and Eugene D. Shapiro, Yale University; Nancy Barrett, Connecticut Emerging Infections Program; Pam Daily, Susan Brooks, and Emma Lucas, California Emerging Infections Program; Mary Glode, Elizabeth Esterl, Molly Byrne, and Hazel Senz, Children's Hospital, Aurora, CO; Mathew Finke, Colorado Department of Public Health and Environment; Pamala Gahr, Catherine Lexau, Billi Juni, Anita Glennen, Bonnie Koziol, Gertrud Kupferschmidt, Kerry MacInnes, and Darla Tuil, Minnesota Department of Health; Bridget Anderson, Shelley Zansky, Christine Long, and Dina Hoefer, New York Emerging Infections Program; Karen Stefonek, Linda Duke, and Michelle Barber, Oregon Department of Human Services; Brenda Barnes, Belinda Redd, Jane Conners, Patty Sackett, Jan Roulstone, Terri McMinn, and Selma Archer, Tennessee Emerging Infections Program; Chris Van Beneden, Tami Skoff, Carolyn Wright, Sandra McCoy, Richard Facklam, and Bernard Beall, Centers for Disease Control and Prevention; and M. Leticia McElmeel, Letitia Fulcher, Christa Trippy, and Sharon Crawford, University of Texas Health Sciences Center at San Antonio.
- Accepted March 31, 2010.
- Address correspondence to Tamara Pilishvili, MPH, CDC Mail stop C-23, 1600 Clifton Rd NE, Atlanta, GA 30333. E-mail:
FINANCIAL DISCLOSURE: Dr Schaffner has received a consulting fee from Wyeth and is a member of the Safety Evaluation Committee for experimental vaccine trials for Merck; Dr Bennett has served on Adult Vaccine Advisory Boards for Wyeth and Merck; the other authors have no financial relationships relevant to this article to disclose.
- PCV7 =
- 7-valent pneumococcal conjugate vaccine •
- IPD =
- invasive pneumococcal disease •
- ACIP =
- Advisory Committee on Immunization Practices •
- ABCs =
- Active Bacterial Core surveillance •
- CDC =
- Centers for Disease Control and Prevention •
- OR =
- odds ratio •
- Hib =
- Haemophilus influenzae type b •
- CI =
- confidence interval
- Rudolph KM,
- Parkinson AJ,
- Reasonover AL,
- Bulkow LR,
- Parks DJ,
- Butler JC
- Pastor P,
- Medley F,
- Murphy TV
- Levine O,
- Farley M,
- Harrison L,
- et al
- 9.↵Centers for Disease Control and Prevention. Updated recommendation from the Advisory Committee on Immunization Practices (ACIP) for use of 7-valent pneumococcal conjugate vaccine (PCV7) in children aged 24–59 months who are not completely vaccinated. MMWR Morb Mortal Wkly Rep. 2008;57(13):343–344
- Black S,
- Shinefield H,
- Baxter R,
- et al
- Kaplan SL,
- Mason EO Jr.,
- Wald ER,
- et al
- 14.↵US Census Bureau. Population estimates. Available at: www.census.gov/popest/datasets.html. Accessed April 27, 2010
- 16.↵SAS Institute Inc. SAS Version 9.1 [computer program]. Cary, NC: SAS Institute Inc; 2002–2003
- Moore MR,
- Gertz RE,
- Woodbury RL,
- et al
- 21.↵Immunization Action Coalition. Vaccination information for healthcare professionals: pneumococcal conjugate vaccine mandates for children in daycare. Available at: www.immunize.org/laws/pneuconj.asp. Accessed June 1, 2009
- Smith PJ,
- Nuorti JP,
- Singleton JA,
- Zhao Z,
- Wolter KM
- 24.↵Centers for Disease Control and Prevention. Notice to readers: decreased availability of pneumococcal conjugate vaccine. MMWR Morb Mortal Wkly Rep. 2001. Available at: www.cdc.gov/mmwr/preview/mmwrhtml/mm5036a3.htm. Accessed April 27, 2010
- 25.↵Centers for Disease Control and Prevention. Notice to readers: limited supply of pneumococcal conjugate vaccine—suspension of recommendation for fourth dose. MMWR Morb Mortal Wkly Rep. 2004. Available at: www.cdc.gov/mmwr/preview/mmwrhtml/mm5305a6.htm. Accessed April 27, 2010
- Adamkiewicz TV,
- Silk BJ,
- Howgate J,
- et al
- O'Brien KL,
- Swift AJ,
- Winkelstein JA,
- et al
- Halasa NB,
- Shankar SM,
- Talbot TR,
- et al
- Copyright © 2010 by the American Academy of Pediatrics