Study Sample
We used data collected from mothers and their infants participating in the HOME Study, an ongoing prospective birth cohort in the Cincinnati metropolitan area designed to examine low-level environmental toxicant exposure and the efficacy of injury and lead hazard controls in the home. From March of 2003 to January 2006, women were identified from seven prenatal clinics associated with three hospitals. Eligibility criteria for the study included: <19 weeks gestation; > 18 years old; living in a house built before 1978; living in Brown, Butler, Clermont, Hamilton, or Warren counties, intention to continue prenatal care and deliver at collaborating obstetric practices, negative HIV status; and not receiving seizure, thyroid, or chemotherapy/radiation medications. We mailed letters to 5,512 women > 18 years of age who were living in a home built before 1978 to see if they were eligible and interested in participating in our study. Of the 1,263 eligible women, 468 enrolled in our study. Our analyses were restricted to singleton infants.
Tobacco Smoke Measurements
Self-Reported Exposure
Women self-reported secondhand and active tobacco smoke exposures for the periods between conception and 20 weeks (measured at 20 week home visit) and 20 weeks and birth (measured at a 4 week postpartum visit). Trained interviewers asked the women the average number of cigarettes they smoked per day, the number of smokers living in the home, and the number of cigarettes smoked per day by other people in the home for each time period. We classified the woman's exposure status during each time period as unexposed, exposed to secondhand tobacco smoke, and exposed to active tobacco smoke.
Serum and Meconium Biomarkers of Exposure
Women provided serum samples at 16 weeks gestation, 26 weeks gestation, and within 24 hours of birth. All samples were stored at -20°C until they were transported to the Centers for Disease Control and Prevention (CDC) laboratory for analysis, where they were stored at < -20°C. Serum samples were analyzed for cotinine, a biomarker of nicotine exposure, using high performance liquid chromatography-tandem mass spectrometry (HPLC-MS/MS) [18, 19]. The limit of detection (LOD) for this assay was 0.015 ng/mL with a coefficient of variation (CV) ranging from 3-4% at high concentrations (1 ng/mL) to 10% at low concentrations (0.1 ng/mL).
Meconium specimens were collected from infants during their hospital stay by placing cellulose fiber inserts inside infant diapers. After it was soiled, the diaper and insert were initially stored in a labeled plastic bag in a designated hospital refrigerator until they were collected by study staff, usually within 24 hours. Meconium stools were collected throughout the hospital stay, until the first milk stool appeared. Study staff pooled meconium samples from diapers into polyethylene containers using a spatula. Pooled meconium samples were stored at -20°C until transported to CDC laboratories where they were stored at < -20°C. Study staff took care not to collect fibers from the diaper or cellulose insert.
For this analysis approximately 0.5 g of meconium were digested at room temperature in 3 mL of 5 N potassium hydroxide containing the internal standards (trideuterated nicotine, cotinine and hydroxycotinine; 1.25 ng of each per sample except for nicotine which was 5 ng). The digested material was extracted with methylene chloride and ethanol, back-extracted into hydrochloric acid, neutralized, buffered, and applied to a CleanScreen DAU column which was processed essentially as described by Ostrea et al [4]. HPLC-MS/MS was used to quantify the concentrations of the NIC, COT, and 3HC relative to the deuterated internal standards. Samples were analyzed using multiple reaction monitoring and concentrations were calculated from the ratios of native and labeled ions in the samples compared to a 10-point calibration curve incorporating the three analytes. Transition ions monitored in this work were 192.9/80.0 and 192.9/133.9 for hydroxycotinine, 177/98.1 and 177/80.1 for cotinine, and 162.9/130, 162.9/117 for nicotine. Corresponding single transitions were monitored for each internal standard. The LODs for these three compounds were 0.946 ng/g for NIC, 0.070 ng/g for COT, and 0.092 ng/g for 3HC.
Each analytical run included two blank samples and a low and high concentration quality control (QC) sample. The low and high QC materials were prepared from two separate, pooled samples of meconium that had been pre-tested and confirmed to have different levels of all three metabolites present. Each pool was prepared by taking 25 g of its respective meconium sample and diluting with 600 mL 5 N KOH. This was allowed to stir for 2 hr to dissolve the solid and assure even mixing. Each pool was dispensed with continuous mixing in 3 mL aliquots into 200 sequentially numbered 16 × 125 mm screw cap tubes. After all the tubes were prepared, they were stored at -70°C until use. The low-concentration QC samples had a relative standard deviation (RSD) of 25% (NIC), 8% (COT), and 14% (3HC), while the high-concentration QC samples had a CV of 15% (NIC), 5% (COT), and 6% (3HC). Accuracy evaluations were conducted at target concentrations of 0.5, 2.5 and 7.5 ng/g (2, 10 and 30 ng/g for nicotine). All samples had an analytical bias of less than ± 10%.
Infant Birth Weight and Covariates
Infant birth weight (in grams) was abstracted from hospital medical records and was analyzed as a continuous variable. Maternal age, race, education, and marital status were gathered at the first prenatal care visit. Maternal depression was assessed using the Beck Depression Inventory-II (BDI-II) which was administered during a home visit at approximately 20 weeks [20]. Parity was obtained from maternal medical records. Maternal weight (in kg) was collected at the initial clinic visit at 16 weeks gestation.
Statistical Analysis
We conducted our statistical analysis in two stages. First we examined the relationship between the various self-report and biomarkers of prenatal tobacco smoke exposure. Second, we examined and compared the association between meconium and serum biomarkers of tobacco smoke exposure and infant birth weight.
We started our statistical analysis by comparing women and infants with complete meconium, self-report, serum, and birth weight data to women with missing data. We corrected for the right-skewed distribution of serum cotinine and meconium tobacco smoke metabolite concentrations using the log10-transformation in analyses using continuous variables. Serum and meconium cotinine values <LOD were randomly imputed from the left tail of the log10-normal distribution [21].
Relationship between Self-Report and Biomarkers of Tobacco Smoke Exposure
We compared methods of classifying tobacco smoke exposures among mother-infant pairs with at least two prenatal serum cotinine measurements and a valid meconium measurement available. We created several measures of cumulative prenatal tobacco smoke exposure using either self-reported tobacco smoke exposure or prenatal serum cotinine concentrations.
First, we created a summary variable of self-reported prenatal tobacco smoke exposure over the course of pregnancy. This five category variable reflected the level and duration of exposure: unexposed, SHS exposure in one period, SHS exposure in both periods, active exposure in one period (the other period was unexposed or SHS exposure), and active exposure in both periods.
Next, we averaged the available serum cotinine measurements to create a continuous summary measure women's prenatal tobacco smoke exposure between 16 weeks gestation and birth. From this, we created categories of unexposed (< LOD), secondhand exposure (LOD to 3 ng/mL), and active exposure (> 3 ng/mL). The threshold of 3 ng/mL for active smoking was chosen based on results from the 1999-2004 National Health and Nutrition Examination Survey which compared self-reported smoking status and serum cotinine levels among a representative sample of the US population [22].
Among women with all three serum cotinine and meconium measurements, we further quantified cumulative exposure to prenatal tobacco smoke by creating a summary variable that described the number of prenatal serum measurements that a woman was exposed to secondhand or active tobacco smoke. This seven category, ordinal variable counted the number of measurements that a woman had serum cotinine concentrations indicative of secondhand (zero, one, two, or three) or active (one, two, or three) tobacco smoke exposure. Women in any of the secondhand categories could not have had active exposure at another time point, while women in the active categories could have had another serum measurement consistent with secondhand or no exposure.
Finally, we investigated whether the timing of the serum measurements influenced the association between the number of serum cotinine measurements indicative of secondhand or active tobacco smoke exposure and meconium NIC concentrations. We did this using a seven category variable that summarized the number and timing of serum cotinine concentrations consistent with secondhand and active tobacco smoke exposure at 16 weeks and birth. We limited this analysis to NIC concentrations because NIC was detected most frequently and the results with NIC concentrations were similar to those using COT and 3HC concentrations.
We calculated the geometric mean (GM) and corresponding 95% confidence interval (CI) of meconium tobacco smoke metabolite concentrations according to self-reported and serum cotinine concentration categories described above. We also calculated the proportion of infants with detectable meconium tobacco smoke metabolite concentrations according to self-reported or serum categories.
Association between Biomarkers of Prenatal Tobacco Smoke Exposure and Infant Birth Weight
The second stage of our analysis examined the association between biomarkers of tobacco smoke exposure and infant birth weight. We chose to examine birth weight because there is a well-established inverse relationship between serum cotinine concentrations and infant birth weight [23–28].
We compared the results for the different biomarkers of tobacco smoke exposure since many cohorts only have the resources to collect one exposure measurement. We estimated and compared the associations between continuous log10-transformed prenatal serum cotinine and meconium tobacco smoke metabolite concentrations and infant birth weight using linear regression. Coefficients from these analyses represent the mean change in infant birth outcome for a 10-fold increase in tobacco smoke biomarker concentration. In addition, we examined the association between categorical serum and meconium tobacco smoke metabolite concentrations and infant birth weight. Serum cotinine concentrations were categorized using the thresholds described above. Several different meconium tobacco smoke metabolite concentrations were used to discriminate secondhand from active tobacco smoke exposure based on sensitivity and specificity analyses.
In all of the analyses examining the association between prenatal tobacco smoke exposure and infant birth weight, we adjusted for confounders identified using a directed acyclic graph (DAG) [29]. DAGs are a better method to assess the role of confounding variables compared to change in estimate and significance testing approaches [30]. Based on our DAG, all models included maternal age, maternal education, maternal race, marital status, depression (), and maternal weight (kg). We did not adjust for gestational age since it was an intermediary on the causal pathway between prenatal tobacco smoke exposure and infant birth outcomes.
Ethical Considerations
The Institutional Review Boards (IRB) of the University of North Carolina-Chapel Hill, Cincinnati Children's Hospital and Medical Center, and CDC approved this study. The Cincinnati Children's Hospital and Medical Center IRB was involved in the oversight of this study. All mothers provided written informed consent for themselves and their children prior to enrollment in the study.