Study design and setting

To examine the association of maternal 25-hydroxyvitamin D (25[OH]D) levels and self-reported asthma in young women and viral bronchiolitis severity in their infants, we conducted a cross-sectional study of 340 mother-infant dyads enrolled in the TCRI. The methods regarding the establishment of the TCRI have been reported elsewhere. In brief, the TCRI is a prospective study of 670 term, non–low-birthweight infants (<1 year of age at study entry) enrolled with their biological mothers at the time of an infant acute viral respiratory infection. A convenience sample of mother-infant dyads were recruited in the inpatient, emergency department, and clinic settings during viral respiratory seasons September through May 2004-2008 at a single academic institution.

The primary goals of TCRI are to investigate the association of viral respiratory infection during infancy, maternal atopy, and development of early childhood asthma and atopy. This analysis includes the subset of women from whom a blood sample was obtained at study entry (n = 340) and their infants. The study was approved by the Vanderbilt University Institutional Review Board. Women provided informed consent for themselves and their infants.

At enrollment trained research nurses administered an in-person structured questionnaire during the acute infant visit. The questionnaire included questions regarding demographics, the details of the acute illness, previous medical history of the mother and infant, the infant feeding method, the detailed family history of atopic diseases, maternal responses to the International Study of Asthma and Allergies in Children (ISAAC) questionnaire, and smoking exposure.

To ascertain the women’s vitamin D status at study entry, we measured levels of 25(OH)D, the primary circulating form of the hormone, using whole-blood specimens on blood spot cards. The 25(OH)D levels were determined by high-performance liquid chromatography tandem mass spectrometry performed by ZRT Laboratories (Beaverton, OR).

Self-reported maternal asthma status, ever asthma, was determined using the validated ISAAC questionnaire and defined as a positive response to the question, “Have you ever had asthma,” asked as part of the ISAAC questionnaire, and/or to the question. “Were you diagnosed with asthma as a child?” In addition, we determined whether women who reported a history of asthma had prevalent symptoms by their response to the question, “Have you had wheezing or whistling in the chest in the past 12 months?”

Infant viral lower respiratory tract illness (LRTI) severity was determined using both the physician discharge diagnosis and postdischarge chart review. Infants were considered to have a viral lower respiratory tract illness if they had a physician diagnosis of bronchiolitis or wheezing or if on chart review the infant had signs and symptoms consistent with bronchiolitis or viral lower respiratory tract infection including grunting, nasal flaring and/or chest wall retractions and findings of diffuse wheezing, rales, or rhonchi, and the absence of alternative diagnoses. LRTI severity was determined using the ordinal bronchiolitis score, which ranges from 0 to 12, with 12 being the most severe. The bronchiolitis score includes respiratory rate, the presence and extent of flaring or retractions, room air oxygen saturation, and the presence and extent of wheezing.

We included variables that could confound the relationship between maternal 25(OH)D level and maternal asthma and/or infant bronchiolitis severity, including the following: self-reported maternal race/ethnicity, maternal age, maternal smoking, environmental tobacco smoke exposure, infant birthweight, infant sex, infant age at enrollment, infant feeding method, infant insurance status, year of enrollment (September/October through May 2004-2005 through September through May 2007-2008), season of enrollment (fall, winter, or spring), and season of infant birth.

In this study’s cross-sectional analyses, the variables of interest included maternal 25(OH)D level and its relationship to the maternal asthma prevalence and the severity of LRTI in her infant. Power and sample size determinations were conducted for the analyses of the original cohort and not this substudy. In descriptive analyses, the demographics of the population were determined by maternal 25(OH)D level, which was categorized using the historical categorization as sufficient (>30 ng/mL), insufficient (20–30 ng/mL), or deficient (<20 ng/mL) based on levels important for bone health.

Characteristics of mother-child dyads were compared by whether blood was available for vitamin D determination. In univariate analyses, χ ² tests were used for categorical variables, and the Wilcoxon rank sum or the Kruskal-Wallis test were used as appropriate for continuous variables. Analyses were stratified by maternal race because race is a major determinate of 25(OH)D levels. Median 25(OH)D levels were compared in women with and without a history of asthma and by whether women with asthma had prevalent symptoms. In analyses stratified by maternal race/ethnicity, we evaluated the association of maternal 25(OH)D levels and maternal asthma using multivariable logistic regression adjusting for factors including maternal age, cigarette smoking, infant insurance, and season of enrollment.

To prevent overfitting of logistic regression models, propensity score adjustment was also performed for maternal 25(OH)D levels. The propensity score analysis has gained increasing application because it properly adjusts for many confounding factors simultaneously and preserves analytical power by combining factors into 1 score. For maternal 25(OH)D level outcomes, we used the proportional odds logistic regression model to derive the propensity score using the covariates outlined above for the multivariable logistic regression model. Restricted cubic splines were used to allow for nonlinear adjustment between propensity score and outcome. Propensity score adjustment was used as a sensitivity analyses in the subset of white women and as the primary analysis in the subset of African American women because of the small sample size.

We investigated the association between maternal 25(OH)D level and viral LRTI severity in the subset of the cohort with both maternal blood available for 25(OH)D level determination and an infant with a viral LRTI. In analyses stratified by maternal race, the univariate relationship of maternal 25(OH)D level and infant LRTI severity score were assessed using Spearman correlation coefficient (rho). Multivariable analyses used proportional odds logistic regression for the outcome of respiratory illness severity. Among white women and their infants, multivariable regression analyses controlled for the following variables: maternal age, infant birthweight, infant age at enrollment, year (2004-2005 to 2007-2008) of enrollment, season of enrollment, season of birth, infant insurance, history of any breast-feeding, and other children in the home. Propensity score adjustment was used as a sensitivity analyses in the subset of white women and as the primary multivariable regression analysis in the subset of African American women because of the small sample size.