Low Cord Blood Pneumococcal Immunoglobulin G (IgG) Antibodies Predict Early Onset Acute Otitis Media in Infancy

Transcription

1 American Journal of Epidemiology Copyright O 1997 by The Johra Hopkins University School of Hygiene and Public Health AD rights reserved Vol. 145, No. 11 Printed in USA. Low Cord Blood Pneumococcal Immunoglobulin G (IgG) Antibodies Predict Early Onset Acute Otitis Media in Infancy Juan C. Salazar, Kathleen A. Daly, 2 ' 5 G. Scott Giebink, Bruce R. Lindgren, 45 Carol L. Liebeler, 1-5 Mary Meland, 13 and Chap T. Le 4-5 Low maternally derived serum immunoglobulin G (IgG) antibodies to Streptococcus pneumoniae capsular polysaccharides (PS) combined with the inability of infants to produce anti-ps antibody may explain onset of otitis media in the first 6 months of life. To explore this relation, cord blood samples were assayed for anti-ps IgG antibodies from 414 of 592 infants enrolled in a study of early onset otitis media between 1991 and Infants' ears were examined at health supervision and illness visits for the first 6 months of life in a large Minneapolis-St. Paul, Minnesota, health maintenance organization. Antibodies to seven common pneumococcal serotypes (3, 4, 6B, 14,18C, 19F, and 23F) were measured by enzyme-linked immunoabsorbent assay (ELISA). Cox's regression analysis revealed that among infants with a sibling otitis media history, those with low concentrations of type 14 or 19F anti-ps cord blood antibody had earlier otitis media onset than those with higher cord blood antibody concentrations (relative risks (RR) (95% confidence intervals (Cl)) = 1.77 ( ) and 1.89 ( ), respectively). Day care attendance also increased risk (RR = 1.56, 95% Cl ). Breastfeeding, parental smoking, and low anti-ps antibody to pneumococcal serotypes 3, 4, 6B, 18C, and 23F did not significantly affect the risk of early otitis media. Am J Epidemiol 1997; 145: antibodies; infant, newborn, diseases; otitis media; pneumococcal infections; Streptococcus pneumoniae Otitis media is the most common disease in children younger than age 5 years, and it is responsible for over 30 percent of all pediatric health care visits in the United States (1). The majority of children (61-93 percent) have at least one otitis media episode before their fifth birthday, and percent develop recurrent otitis media or chronic otitis media with effusion (OME) (2, 3). A recent study (4) estimated that each year in the United States million children are diagnosed with otitis media at an estimated annual cost of $5 billion. Received for publication June 28, 1996, and accepted for publication February 12, Abbreviations: EUSA, enzyme-linked immunoabsorbent assay; HMO, health maintenance organization; IgA, immunoglobulin A; IgG, immunoglobulin G; OME, otitis media wfth effusion; PS, pneumococcal capsular potysaccharide. 1 University of Minnesota School of Medicine, Department of Pediatrics, Minneapolis, MN. 2 University of Minnesota School of Medicine, Department of Otolaryngology, Minneapolis, MN. 3 HeatthPartners, Division of Pediatric Medicine, Minneapolis, MN. 4 University of Minnesota, School of Public Health, Division of Biostatistics, Minneapolis, MN. 5 University of Minnesota, Otitis Media Research Center, Minneapolis, MN. 8 Fundacion Valle del UN, Call, Colombia. Infants who experience otitis media during the first 6 months of life are at higher risk for both recurrent otitis media and chronic OME (2, 3, 5). Delaying the first otitis media episode beyond age 6 months would be expected to reduce the risk of recurrent otitis media and chronic OME. Other factors that increase the risk of these conditions include male gender, day care attendance, passive cigarette smoke exposure, and sibling history of recurrent otitis media (5-8). Breastfeeding has been shown to protect infants from acute and recurrent otitis media (2). However, little is known about risk factors for early infant otitis media. Pneumococci cause percent of all acute otitis media (9). A total of 90 different pneumococcal capsular polysaccharide (PS) serotypes are known (10); however, in the United States, only 10 types account for 79 percent of pneumococcal otitis media (type 14, 15 percent; type 19F, 15 percent; type 23F, 13 percent; type 6B, 10 percent; type 3, 9 percent; type 19A, 7 percent; type 6A, 4 percent; type 9V, 3 percent; type 18C, 2 percent; and type 4, 1 percent) (11). Hence, types 14 and 19F together cause 30 percent of pneumococcal acute otitis media and percent of all acute otitis media disease. Type-specific serum antibodies directed against PS have been found to protect against pneumococcal dis- 1048

2 Low Cord Blood Pneumococcal Antibodies and Otitis Media 1049 ease (12, 13). In four studies (14-18), low serum anti-ps immunoglobulin G (IgG) antibodies correlated with higher susceptibility to recurrent otitis media. Ten Swedish children with recurrent otitis media had lower cord blood anti-ps concentrations to types 6A and 19F compared with children who did not have recurrent otitis media (18). Pneumococci were the predominant bacteria isolated from middle ear fluid in children with low cord blood anti-ps antibody, and pneumococcal acute otitis media did not occur in children with recurrent otitis media whose antibody concentrations were high (18). Because cord blood anti-ps antibody is derived principally from maternal anti-ps IgG antibody (19, 20), low maternal anti-ps antibody appears to be a risk determinant of recurrent otitis media and, specifically, pneumococcal otitis media in the offspring. These studies were not designed to calculate the relative risk of developing otitis media as a function of cord blood antibody concentration, and they had small sample sizes. Moreover, these studies did not control for possible confounding exposures such as sibling otitis media history, day care attendance, passive smoke exposure, or breastfeeding. The present study was designed to identify risk factors for otitis media onset in the first 6 months of life, and to correlate cord blood anti-ps antibody concentrations with time to otitis media onset while controlling for potentially confounding exposures such as sibling otitis media history, day care attendance, passive smoking exposure, and breastfeeding. Early otitis media onset was defined as otitis media that occurred before age 6 months. MATERIALS AND METHODS Population The study population comprised infants who were born to two separate but similar cohorts of eligible women followed at a large health maintenance organization (HMO) in Minneapolis-St. Paul, Minnesota. The first group of women were enrolled during the third trimester of pregnancy from participants in the Diana Project, a National Institutes of Healthsponsored study designed to measure the effects of maternal nutritional and life-style exposures before and during pregnancy on infant birth weight. Women aged years were eligible to enroll if they planned a pregnancy within 2 years. Women were excluded if they had conditions that could interfere with gestation or infant birth weight (e.g., diabetes mellitus or cardiovascular disease), voluntary interruption of pregnancy in the past 12 months, or history of infertility. Recruitment letters were sent to a second group of women aged years who received obstetric care at the same HMO. Women were not eligible if they were in their third trimester of pregnancy or had left the HMO. Only women from both groups whose infants were born during the study enrollment period and would receive care at the HMO staff-model clinics were eligible for the study. Data collection Women completed data forms during pregnancy and monthly from birth to 6 months of age. These were designed to measure sociodemographic variables (maternal and paternal age, combined household income, ethnicity/race, and maternal education), environmental variables (home passive smoking exposure from either parent, and day care attendance), family history variables (parental and sibling history of recurrent ear infections, defined as &3 episodes of otitis media in a 12-month period, or history of tympanostomy tubes, chronic otorrhea, or persistent middle ear fluid), nutritional variables (breastfeeding alone, breastfeeding with bottle, or bottle alone). Birth weight and gestational age were abstracted from the medical record. Cord blood samples were collected at birth and sent to the hospital laboratory, where they were separated and the sera stored at 20 C until transfer to the University of Minnesota for analysis. Infants were examined by study pediatricians, pediatric nurse practitioners, and family practitioners (n = 75) at their regularly scheduled health maintenance examinations (ages 2 weeks, and 2, 4, and 6 months), and when they were ill. At each examination, the physician or nurse-practitioner completed an examination form, which queried about illness symptoms (fever, irritability, and altered sleep pattern), tympanic membrane appearance (color, mobility, translucency, presence of middle ear fluid, and otorrhea), and middle ear diagnosis. Ninety percent of examinations during the first 6 months were conducted by study physicians. If patients were seen outside the HMO or if a form was not completed at the examination, medical records were abstracted by the research nurse. The primary outcome measure was age at the first otitis media episode. Patients were considered to have early otitis media if they were diagnosed with acute otitis media or OME by age 6 months. To assess the consistency of the physicians' specific observations and diagnosis, recorded signs and symptoms were compared with the middle ear diagnosis using an a priori diagnostic algorithm in which OME was defined as absent fever and irritability, normal sleep pattern, and presence of middle ear fluid. Middle ear fluid was considered present if the ear examination revealed at least one of three findings: fluid seen (bubbles, air/ fluid level, or otorrhea), tympanic membrane bulging,

3 1050 Salazaretal. or with decreased or absent mobility. Acute otitis media was defined as presence of middle ear fluid, otorrhea, or a tympanic membrane that was red, yellow, or bulging. Interobserver reliability was measured by calculating percent agreement and kappa statistics between a subset of the study pediatricians, a validated research nurse otoscopist, and one investigator (GSG). Laboratory methods Serum IgG antibodies against seven common pneumococcal serotypes (3, 4, 6B, 14, 18C, 19F, and 23F) were measured by enzyme-linked immunoabsorbent assay (ELISA) after absorbing sera with pneumococcal cell wall polysaccharide (C-Ps), as we have previously described (21). Note, however, that the antigencoating buffer contains 0.05 percent Tween 20 and not 0.5 percent as stated in Salazar et al. (21). Serum antibody concentrations were calibrated against the Federal Drug Administration (FDA) lot 89-SF reference, and individual anti-ps IgG antibody values in lot 89-SF were determined in our laboratory by a corollary response method (22). This method is based on equivalent absorbances between a reference ELISA and the anti-ps ELISA performed in parallel under identical assay conditions. Using a logarithmic regression of the reciprocal dilution versus absorbance, the interpolated dilutions at 0.3 absorbance units for both assays represent equivalent amounts of antibodies. The known concentration (yu.g per ml) of the reference preparation at this dilution is used to assign anti-ps antibody concentrations to the reference, lot 89-SF, in the anti-ps ELISA. We used a reference ELISA in which human Reference Preparation for Serum Protein (RPSP-Lot 4, College of American Pathologists, Northfield, Illinois) immunoglobulin is captured by goat anti-human polyvalent immunoglobulin (Sigma #1-8758, St. Louis, Missouri). The RPSP-Lot 4 has been assigned mg IgG per ml, 1.11 mg IgM per ml, and 2.30 mg immunoglobulin A (IgA) per ml. In the anti-ps ELISA, NUNC Maxisorb microplates (Catalogue #439454, Gibbco Scientific, Coon Rapids, Minnesota) were coated separately with each serotype-specific PS for direct binding of serum antibodies. For six of ten PS types, our anti-ps IgG antibody assignments for lot 89-SF were not significantly different from the values reported by Quataert et al. (22) (table 1). Our assignments for types 9V, 14, 18C, and 19F anti-ps IgG, however, differed significantly from those reported by Quataert et al. (22). The FDA reports that lot 89-SF is lyophilized lot 89-S and is identical in antibody composition (C. Frasch, Food and Drug Administration, personal communication, 1994). Differences in ELISA methodology between our laboratory TABLE 1. Lot 89-SF reference serum values, determined by corollary response methodology in the Investigators' laboratory (means of 36 replicate assays), used to assign anti-capsular polysaccharide (PS) immunoglobulin G (IgG) antibody concentrations In cord blood serum compared with lot 89-S assigned values (means of four replicate assays) reported by Quataert et al. (22) PS antigen B 9V 14 18C 19F 23F 89-SF assigned value (standard deviation) University ol Minnesota 7.20(1.69) 2.39(1.07) 4.62(1.34) 6.46(1.36) 19.18(5.34) 13.10(4.50)* (4.64)** 7.40 (2.51)*** 8.79 (3.26)**** 10.16(2.46) Difference between mean reference values: * 0.001, *** p , ** * p Quataert etal (0.85) 2.36 (0.30) 4.07 (0.47) 5.75 (0.36) (2.77) 6.90(1.85) (3.65) 4.46 (0.94) (2.67) 8.14(1.07).0.010, * p< and the Quataert laboratory most likely explain the variation in lot 89-SF and 89-S antibody assignments. Until other laboratories report anti-ps IgG antibody assignments for lot 89-SF using similar corollary response methodology, conclusions regarding assignment accuracy are premature. Pending such time that consensus is reached, interested persons may transform antibody results to either assignment with the use of the mean values shown in table 1. Antibody assays were performed after all infants were 6 months of age; therefore, middle ear diagnoses were made without knowledge of antibody concentrations. During the antibody assay phase, principal investigators, study pediatricians, study participants, and the research nurse were masked to antibody results; laboratory technologists were masked to clinical outcomes. Statistical analysis Summary statistics were calculated for all subjects and for specified subgroups. These statistics included means and medians for continuous demographic variables, and proportions for categorical variables. Association between categorical variables was analyzed by chi-square tests. Because antibody concentration distributions were skewed, they were log transformed. Relative risks for developing otitis media by age 6 months as a function of log-antibody concentration for each of the seven pneumococcal types was analyzed by Cox's proportional hazards regression (23). Antibody concentration was modeled as a categorical variable by dividing subjects into quartiles based on antibody concentration. The following covariates were

4 Low Cord Blood Pneumococcal Antibodies and Otitis Media 1051 also included in the model: sibling otitis media history (yes/no), entry into day care by age 6 months (yes/no), feeding type (breast/breast + bottle/bottle) and exposure to parental cigarette smoking at 2 weeks or 6 months (yes/no). An interaction term between significant covariates and age at first otitis media was used to assess the proportional hazards assumption of no change in risk as a function of time. The product limit (or Kaplan-Meier) method was used to estimate the rate of the first otitis media episode over time (23). Subgroups defined by the concentration of cord blood anti-ps IgG antibody concentration were compared by the log-rank test (23). Infants were first stratified into three groups: those with antibody concentration in the lowest quartile, those with antibody concentration in the middle two quartiles, and those with antibody concentration in the highest quartile. Infants were then stratified into two groups: low antibody (lowest quartile) and high antibody (highest three quartiles). RESULTS Among Diana Project participants, 339 women were eligible for our study, and 264 women participated. Of 2,262 recruitment letters sent to the second group of women who received care at the same HMO, 647 women responded, 395 were eligible, and 368 consented to participate. Of the 632 women who agreed to participate, 11 withdrew during pregnancy, and 10 enrolled twice (two different pregnancies). The 611 individual participants had 627 infants in the course of the study (10 sets of siblings and six sets of twins). Two infants were not eligible because of cleft palate or microcephaly, and 32 infants were voluntarily withdrawn from the study before they completed 6 months of followup. Of the 605 neonates with some follow-up, cord blood samples were collected on 425 infants. Nearly all of the study subjects who contributed cord blood samples were white (97 percent), and infants were almost equally distributed between males (49 percent) and females (51 percent). Only 12 subjects (3 percent) belonged to a household with an income less than $20,000 per year, whereas 130 (30 percent) had household incomes greater than $60,000. Birth weight was between 1.5 and 5.1 kg (mean 3.3 kg). Mothers' mean and median ages were both 31 years. At age 2 weeks and 6 months, 80 and 39 percent of the infants, respectively, were fully or partially breast-fed. At 6 months, 48 percent attended out-ofhome day care. Only 12 percent of infants at age 2 weeks and 13 percent at 6 months had a parent who smoked. Sibling otitis media history, as previously defined, was documented in 26 percent of subjects. Demographic characteristics were not significantly different for enrolled subjects whose cord blood was not collected. Internal consistency of the pediatricians' otitis media diagnoses correlated well with the a priori middle ear algorithm. For instance, 95 percent of infants aged <2 months and 98 percent of infants aged 2-6 months who were diagnosed to have acute otitis media had a bulging, red tympanic membrane. Similarly, 77 percent of infants aged <2 months and 90 percent of infants aged 2-6 months diagnosed with OME had an opaque tympanic membrane with decreased mobility. Kappa statistics for agreement between a sample of 20 HMO pediatricians, the research nurse otoscopist, and validated otoscopist were all between 0.54 and Statistical analyses were performed based on pediatrician-generated middle ear diagnoses. Although parents in the participating HMO chose a primary physician to provide continuity of care for their child, that individual was not always available when the child was acutely ill (e.g., with otitis media). Thus, participants in this study were seen by a number of different examiners (mean = 2.11, standard deviation (SD) 1.1, range 1-7) during the 6 months of follow-up. Among the 10 sets of sibling pairs enrolled in the study, the mean number of examiners was 6.7 (SD 2.8, range 3-12) for the older sibling during 2 years of follow-up, and 2.1 (SD 2.1, range 1-7) for the younger siblings during 6 months of follow-up. Cord blood anti-ps IgG antibody quartiles are shown in table 2. The 25th percentile breakpoint was 0.24 ng/ml for type 14 anti-ps IgG antibody and 0.47 yxg/ml for type 19F. Geometric mean titers were highest for type 6B antibody (1.67 /xg/ml) and lowest for type 4 (0.51 /u-g/ml). Type 14 and 19F antibody concentrations were skewed. For type 14, 41.6 percent of subjects had antibody concentrations <0.50 /xg/ml (range ); 27 percent of subjects had type 19F antibody concentrations <0.50 p,g/ml (range ). After transformation, all log-antibody results were normally distributed; therefore, statistical analyses were performed on transformed values. Correlation coefficients among anti-ps IgG antibodies varied from 0.70 (p < 0.001) between types 6B and 18C to 0.24 {p < 0.001) between types 3 and 14. Correlation coefficients were 0.63 (p < 0.001) between types 6B and 19F, and 0.28 (p < 0.001) between types 14 and 19F. Univariate analysis revealed a significant association between sibling otitis media history and early otitis media (j? , p < 0.01). Infants who attended day care by age 6 months were more likely to be diagnosed with otitis media than infants who did not attend day care (jf = 3.88, p < 0.05). Type of infant feeding and exposure to parental smoking were

5 1052 Salazaretal. TABLE 2. Cord blood anti-capsular polysaccharide (PS) Immunoglobulin G (IgG) (Mfl/niL.) quartile distributions for seven pneumococcal serotypes, geometric mean titere (GMT) and 95% confidence intervals (CI): Early Otitis Media Study, PS antigen Lowest quartile Middle two quartlles Highest quartde GMT 95% CI 3 4 6B 14 18C 19F 23F O ^ O not significantly associated with early onset otitis media. Kaplan-Meier analysis revealed that infants with type 14 (log-rank x 2 = 5.0, p = 0.08) and 19F (logrank x 2 = 5.4, p = 0.07) anti-ps IgG antibody concentrations in the lowest quartiles tended to have an earlier onset of otitis media than infants with concentrations in the middle two and the top quartiles. Kaplan-Meier analyses were repeated after stratifying subjects into two groups: infants with antibody in the lowest quartile, and infants with antibody in the highest three quartiles. Infants with low type 14 antibody had their first otitis media episode significantly earlier than infants with higher antibody (p = 0.03), a difference that persisted until at least age 6 months (figure 1). Similarly, infants with low type 19F antibody were more likely to develop early otitis media than those with higher antibody concentrations (p = 0.02; data not shown). Among infants with a positive sibling history, 78 percent of infants who had low type 14 anti-ps IgG had otitis media by age 6 months compared with 56 percent of infants who had higher antibody concentrations (log-rank x 2 ~ 6.5, p = 0.01) (figure 2). Time to otitis media onset was also significantly shorter among infants with low concentrations of type 19F antibody compared with those with higher concentrations (logrank x 2 ~ 6.0, p = 0.01). Similar trends were observed among infants with a negative sibling otitis media history, although differences in survival distributions functions were not statistically significant for type 14 (log-rank x 2 = 1-6, p = 0.21) or for type 19F (log-rank )? = 2.2, p = 0.14). By means of Cox's regression model, it was found that infants with type 14 antibody in the lowest quartile had an increased risk of early otitis media (RR = 1.64, 95 percent CI ) (table 3). Infants with low 19F antibody had a similarly increased risk of early otitis media (RR = 1.65, 95 percent CI ). Positive sibling otitis media history and entry into day care by age 6 months also increased the risk of early otitis media. Risk of early otitis media was not 1.00-PTrrr S0.70 o :o.eo! 0.50 s 0.40 S Age in months FIGURE 1. Kaplan-Meier survival distribution function for age at the first otitis media (OM) episode by cord blood type 14 anticapsular polysaccharide (PS) immunoglobulin G (lg(3) antibody concentration In infants whose antibody concentration was in the lowest quartile (s0.24 jig/ml, broken line) or higher three quartiles (>0.24 jtg/ml, solid line) antibody concentrations: Early Otitis Media Study, The difference between groups was significant (p = 0.03). significantly related to parental smoking, lack of breastfeeding, or low anti-ps IgG antibodies to types 3, 4, 6B, 18C, and 23F pneumococci. All covariates in Cox's proportional hazards regression model met the proportional hazards assumption of no change in relative risk of early otitis media between birth and age 6 months, except sibling otitis media history (Wald x 2 = P < 0.01). Therefore, adjusted relative risks for significant variables were calculated separately after stratifying subjects by sib-

6 Low Cord Blood Pneumococcal Antibodies and Otitis Media 1053 o Age in months FIGURE 2. Kaplan-Meier survival distribution function for age at the first otitis media (OM) episode by cord blood type 14 anticapsular polysaccharide (PS) immunoglobulin G (IgG) antibody concentration in infants wtth a positive sibling OM history and whose antibody concentration was in the lowest quartile (=50.21 ng/ml, broken line) or higher three quartiles (>0.21 Mg/mL, solid line) antibody concentrations: Early Otitis Media Study, The difference between groups was significant (p = 0.01). ling otitis media history. Among infants with a positive sibling otitis media history and low type 14 cord blood antibody, the relative risk of early otitis media was 1.77 (95 percent CI ) (table 4). Similarly, for infants with low type 19F antibody, the relative risk of early otitis media was 1.89 (95 percent CI ). For infants with low concentrations of both type 14 and 19F antibody, the relative risk of early otitis media was 3.10 (95 percent CI ). Relative risks (95 percent CIs) for infants with a negative sibling otitis media history are also shown in table 4. There were no interactions between any of the variables remaining in the model for each strata. Tobacco smoke exposure and breastfeeding history did not increase or decrease the relative risk of early otitis media and therefore were excluded from the regression model in both strata. DISCUSSION Onset of otitis media during the first few months of life is a known risk factor for recurrent otitis media and chronic OME (2, 3). Results from this study demonstrate that low cord blood type 14 and 19F anti-ps IgG antibodies are independent predictors of early otitis media onset. These results are consistent with the observation that types 14 and 19F pneumococci are the most prevalent bacterial causes of acute otitis media in the United States (11). Thus, low maternal concentrations of types 14 and 19F IgG antibodies result in low neonatal antibody levels and early onset otitis media, which leads to recurrent and chronic disease. The relation between antibody concentration and early onset otitis media remained significant even after controlling for other known otitis media risk factors: day care attendance, parent cigarette smoke exposure, breastfeeding, and sibling otitis media history. Cord blood IgG anti-ps antibodies to types 3, 4, 6B, 18C, and 23F were not associated with increased risk of early otitis media, probably because these types are less frequent causes of otitis media than types 14 and 19F. Because infants do not produce natural anti-ps antibodies until the latter half of the first year and well into the second year of life, neonates with low anti-ps antibody at birth are likely to be at greater risk of pneumococcal otitis media than neonates with higher cord blood antibody concentrations. Higher neonatal anti-ps antibody concentrations against types 14 and 19F pneumococci would be expected to delay the onset of otitis media. Neonatal antibody concentrations reflect maternal antibody production and placental transport during gestation. IgG is actively transported during intrauterine life (24), and IgGl subclass antibodies are transferred to a greater extent than IgG2, IgG3, and IgG4 antibodies (19). Correlation between neonatal and maternal serum concentrations of IgGl subclass antibody against types 1, 6A, 14, 19F, and 23F PS is strong (19, 25), but type 7F anti-ps IgG neonatal to maternal antibody correlation is weak (26). These differences are thought to result from a differential transfer of antibodies depending on their affinity to placental receptors (26). O'Dempsey et al. (27), however, have reported that cord blood/maternal blood anti-ps antibody ratios of individual mother/child pairs tended to hold their rank order between PS types. Investigation of the large differences in antibody transfer between mother/child pairs revealed that cord/maternal antibody ratios were significantly negatively correlated with maternal antibody levels, and positively correlated with gestational age at delivery. Ratios were not correlated with maternal age at delivery, parity, or ethnic group.

7 1054 SaJazaretal. TABLE 3. Adjusted* relattve risks for earty otitis media (OM) using Cox's proportional hazards regression analysis: Earty Otitis Media Study, Covartates Significant Low type 14 cord blood antibody} Low type 19F cord Wood antibody! Sibling otitis media history (yes/no) Day care at 6 months (yes/no) Relattve risk %Clt P value < <0.01 <0.01 Not significant Breastfeeding history (breast/both/bottle) Parents smoke (yea/no at 6 months) Low type 3 cord blood antibody}: Low type 4 cord blood antibody} Low type 6B cord Wood antibody}: Low type 18C cord Wood antibody} Low type 23F cord blood antibody}: * Relative risks were adjusted for all listed variables. t Cl, confidence interval. } Lowest quartile vs. highest three quartiles. TABLE 4. Adjusted* relative risks for early otitis media (OM) using multivariate Cox's proportional hazards regression analysis stratified by sibling otitis media history: Early Otitis Media Study, Covartataa Positive sibling OM history Low type 14 antibody} Low type 19F antibody} Day care at 6 months (yes/no) Negative sibling OM history Low type 14 antibody}: Low type 19F antibody} Day care at 6 months (yes/no) Relative risk * Relative risks were adjusted for all listed variables. t Cl, confidence interval. t Lowest quartile vs. highest three quartiles. 95%Clt Anti-PS antibody titers reach a nadir between ages 6 and 18 months because of the disappearance of maternal antibody and the slow maturation of the infant's anti-ps antibody production, which is unrelated to nasopharyngeal pneumococcal carriage during the first 2 years of life (28). O'Dempsey et al. (27) reported that anti-ps antibody levels in infants of mothers immunized with pneumococcal PS vaccine during the last trimester of pregnancy declined rapidly. Estimated antibody half-lives were days, and vaccinated and control infants had similar antibody levels by age 3 months. Their study, however, was confounded by the presence of malaria parasitemia in 24 percent of the vaccinated women; malaria infection significantly reduced the maternal anti-ps antibody response and reduced the cord blood/maternal blood antibody ratio. Low cord blood anti-ps antibody concentrations to common pneumococcal types may be due to several factors. Lack of prior maternal colonization with type 14 and 19F pneumococci or cross-reacting antigens, or low immunogenicity of these two PS antigens might explain low cord blood type-specific antibodies. Although there are no data that describe anti-ps antibody concentrations in women by age, pneumococcal antibody prevalence has been reported (29) to increase with age in healthy adult males. Among young men, only 15 percent had anti-ps antibody to eight common pneumococcal serotypes, compared with 33 percent of middle-aged working men and 34 percent of elderly men (29). Age differences were statistically significant for types 1, 14, and 23F pneumococci. IgG antibody production deficiency has been reported in both children and adults (30, 31), and low cord blood anti-ps IgG antibody could be explained by a genetically determined maternal antibody production disorder. This notion is supported by the interaction between sibling otitis media history, low antibody concentration to types 14 and 19F PS, and early otitis media in our study. Previous studies have suggested that antibody response to polysaccharide antigens is subject to genetic control (32). Specifically, individuals with the hereditary immunoglobulin allotype G2m(n) do not mount an adequate IgG2 response to PS. IgG2-deficient women would be unable to mount an antibody response to pneumococcal PS, and they would therefore transfer less IgG anti-ps antibody to their infants (33). These individuals would also be low responders when immunized with pneumococcal PS. Although maternal IgG2 deficiency could explain some of the differences between infants whose cord blood contained low or undetectable antibody and those with higher antibody concentrations, it seems

8 Low Cord Blood Pneumococcal Antibodies and Otitis Media 1055 unlikely that all women whose infants had low cord blood type 14 and 19F IgG antibody concentrations (25 percent of our population sample) had an IgG production deficiency. IgA deficiency is more common than IgG deficiency (24). Because fetal serum from IgA-deficient women has significantly higher IgGl concentrations but lower IgG2 and IgG4 concentrations than fetal serum from women with normal IgA (34), it is possible that children born to IgA-deficient women are at higher risk of early infection. This relation needs further study, especially in children with low cord blood anti-ps antibody and a positive sibling history. We did not obtain maternal serum samples in this study. Susceptibility to early otitis media could also be the result of higher type 14 and 19F anti-ps antibody concentrations required for protection compared with other type-specific antibodies. There are currently insufficient data to judge protective type-specific antibody concentrations in children or adults. Further defining relations between cord blood types 14 and 19F anti-ps IgG antibodies and age at otitis media onset may lead to strategies that reduce the incidence of early onset otitis media, subsequent recurrent otitis media, and chronic OME. Because it is unlikely that infant PS-protein conjugate vaccines will induce protective anti-ps IgG antibodies during the first 2-4 months of life (35), passive infant protection through maternal immunization could provide sufficient transplacental IgG antibody to forestall early onset otitis media. Maternal immunization is likely to be more effective in this regard if the vaccine induces IgGl antibodies which are transferred at higher concentrations than the other subclasses. ACKNOWLEDGMENTS This work was supported by grant nos. PO1-DCOO133 and R01-DC01242 from the National Institute on Deafness and Other Communication Disorders, the Minnesota Medical Foundation, and the Deafness Research Foundation. Dr. Salazar also received support from the American Legion and Auxiliary Heart Research Fund. REFERENCES 1. Teele DW, Klein JO, Rosner B, et al. Middle ear disease and the practice of pediatrics: burden during the first five years of life. JAMA 1983;249: Teele DW, Klein JO, Rosner B, et al. Epidemiology of otitis media during the first seven years of life in children in greater Boston: a prospective cohort study. J Infect Dis 1989;160: Marchant CD, Shurin PA, Turczyk VA, et al. Course and outcome of otitis media in early infancy: a prospective study. J Pediatr 1984;104: Gates GA. Cost-effectiveness considerations in otitis media treatment. Otolaryngol Head Neck Surg 1996;14: Harsten G, Prellner K, Heldrup J, et al. Recurrent acute otitis media. A prospective study of children during the first three years of life. Acta Otolaryngol 1989;107:ll Daly K, Giebink GS, Le CT, et al. Determining risk for chronic otitis media with effusion. Pediatr Infect Dis J 1988; 7: Daly KA. Epidemiology of otitis media. Otolaryngol Clin North Am 1991;24: Iversen M, Birch L, Lundqvist GR, et al. Middle ear effusion in children and the indoor environment: an epidemiological study. Arch Environ Health 1985;40: Del Beccaro MA, Mendelman PM, Inglis AF et al. Bacteriology of acute otitis media. A new perspective. J Pediatr 1992; 120: Henrichsen J. Six newly recognized types of 5. pneumoniae. J Clin Microbiol 1995;33: Butler JC, Breiman RF, Lipman HB, et al. Serotype distribution of Streptococcus pneumoniae infections among preschool children in the United States, : implications for development of a conjugate vaccine. J Infect Dis 1995;171: Giebink GS, Koskela M, Vella PP, et al. Pneumococcal capsular polysaccharide-meningococcal outer membrane protein complex conjugate vaccines: immunogenicity and efficacy in experimental pneumococcal otitis media. J Infect Dis 1993; 167: Shapiro ED, Berg AT, Austrian R, et al. The protective efficacy of polyvalent pneumococcal polysaccharide vaccine. N Engl J Med I991;325:1453-6O. 14. Freijd A, Hammarstrom L, Persson MA, et al. Plasma antipneumococcal antibody activity of the IgG class and subclasses in otitis prone children. Clin Exp Immunol 1984;56: Kalm O, Prellner K, Freijd A, et al. Antibody activity before and after pneumococcal vaccination of otitis prone and nonotitis prone children. Acta Otolaryngol 1986; 101: Prellner K, Kalm O, Pedersen FK. Pneumococcal antibodies and complement during and after periods of recurrent otitis. Int J Pediatr Otorhinolaryngol 1984;7: Kalm O, Prellner K, Pedersen FK. Pneumococcal antibodies in families with recurrent otitis media. Int Arch Allergy Appl Immunol 1984;75: Prellner K, Kalm O, Harsten G, et al. Pneumococcal serum antibody concentrations during the first three years of life: a study of otitis-prone and non-otitis-prone children. Int J Pediatr Otorhinolaryngol 1989;17: Gasparoni A, Avanzini A, Ravagni Probizer F, et al. IgG subclasses compared in maternal and cord serum and breast milk. Arch Dis Child 1992;67: Hazlewood M, Nusrat R, Kumararatne DS, et al. The acquisition of anti-pneumococcal capsular polysaccharide, Haemophilus influenzae type b and tetanus toxoid antibodies, with age in the UK. Clin Exp Immunol 1993;93: Salazar JC, Londoflo LJ, Liebeler C, et al. Pneumococcal seroepidemiologic study of Colombian and Minnesota children. Pediatr Infect Dis J 1995; 14: Quataert SA, Kirch CS, Wiedl LJQ, et al. Assignment of weight-based antibody units to a human antipneumococcal standard reference serum, lot 89-S. Clin Diagn Lab Immunol 1995,2: Lee ET. Statistical methods for survival data analysis. Belmont, CA: Lifetime Learning Publications, Oxelius VA, Svenningsen NW. IgG subclass concentrations in preterm neonates. Acta Paediatr Scand 1984;73: Anderson P, Porcelli S, Pichichero M. Natural maternal and cord serum antibodies to pneumococcal serotypes 6A, 14, 19F and 23F polysaccharides. Pediatr Infect Dis J 1992; 11:677-9.

9 1056 Salazaretal. 26. Chudwin DS, Wara DW, Schiffman G, et al. Maternal-fetal transfer of pneumococcal capsular polysaccharide antibodies. Am J Dis Child 1985;139: O'Dempsey TJD, McArdle T, Ceesay SJ, et al. Immunization with a pneumococcal capsular polysaccharide vaccine during pregnancy. Vaccine 1996;14: Gray BM, Converse GM m, Huhta N, et al. Epidemiologic studies of Streptococcus pneumoniae in infants: antibody response to nasopharyngeal carriage of types 3, 19, and 23. J Infect Dis 1981;144: Mustier DM, Groover JE, Rowland JM, et al. Antibody to capsular polysaccharides of Streptococcus pneumoniae: prevalence, persistence, and response to revaccination. Clin Infect Dis 1993;17: Oxelius VA, Laurell AB, Lindquist B, et al. IgG subclasses in selective IgA deficiency: importance of IgG2-IgA deficiency. N Engl J Med 1981 ;304: Umetsu DT, Ambrosino DM, Quinti I, et al. Recurrent sinopulmonary infection and impaired antibody response to bacterial capsular polysaccharide antigen in children with selective IgG subclass deficiency. N Engl J Med 1985;313: Ambrosino DM, Schiffman G, Gotschlich EC, et al. Correlation between G2m(n) immunoglobulin allotype and human antibody response and susceptibility to polysaccharide encapsulated bacteria. J Clin Invest 1985;75: Rynnel-Dagoo B, Freijd A, Hammarstrom L, et al. Pneumococcal antibodies of different immunoglobulin subclasses in normal and IgG deficient individuals of various ages. Acta Otolaryngol 1986;101: Thorn H, Carter PE, Duffty P, et al. Comparison of IgG subclasses in foetal serum, maternal serum at delivery and milk in IgA-deficient and control women. Acta Paediatr 1994; 83: Giebink GS, Liebeler CL, Daly KA, et al. Pneumococcal conjugate vaccine immunogenicity in infants and antibody persistence to age two years. In: Lim D, Bluestone C, Casselbrant M, et al, eds. Recent advances in otitis media. Proceedings of the 6th International Symposium. Hamilton, Ontario: BC Decker, Inc, 1996:495-8.