The study sample of 222 children is a subset of a cohort of 276 children born between July 1994 and January 1995 who were enrolled in a dust control intervention trial at 6 months of age . In order to be enrolled in the trial, children (5–7 months old at baseline) had to reside in Rochester, New York and families must have had no plans to relocate for at least 3 months . The 222 children included in this sub-analysis had complete information on asthma diagnosis and other covariables of interest. Parents/guardians were asked whether their child had been diagnosed with asthma by a physician in the past 12 months; children with a response at either 48, 60, or 72 months of age to this question were included in this study sample. Parents or guardians provided written consent and the study protocol was approved by the Institutional Review Board at the University of Rochester Medical Center, Rochester, New York.
The outcome of interest in this study was parental/guardian report of a physician diagnosis of asthma in the preceding 12 months. This information was collected from a self-report questionnaire at the child’s 48-, 60-, and 72-month follow-up visits that asked, “In the past 12 months did the doctor say he/she [the child] had asthma?” Children were considered to have physician-diagnosed asthma if their parent/guardian reported a diagnosis at any one of these three time points.
Blood lead measurement
Venous blood samples were collected when children were 6, 12, 18, 24, 36, 48, 60, and 72 months of age [20, 21]. All blood samples were analyzed for lead by electrothermal atomic absorption spectrometry (ETAAS) at the New York State Department of Health (NYSDOH) Wadsworth Center’s Trace Elements laboratory, a reference laboratory for blood lead. Blood lead was measured using a Perkin Elmer Model 400ZL atomic absorption spectrometry equipped with a transversely heated graphite atomizer (THGA) and longitudinal Zeeman background correction (PerkinElmer Life and Analytical Sciences, Shelton, CT). The THGA instrument was calibrated daily before each run with aqueous lead standards traceable to the National Institute of Standards and Technology (NIST, Gaithersburg, MD). Blood samples and QC materials were diluted 1 + 9 with reagents and 12 μL samples injected into the atomizer. Further details on this method, validation, and re-validation have been previously published [22, 23]. The limit of detection was estimated at ~ 1.0 μg/dL, with a limit of quantification at ~3.0 μg/dL. The standard deviation of repeatability of measurements ranged from 0.1 to 0.3 μg/dL for blood lead concentrations less than 10 μg/dL, and varied by less than 2% for blood lead measurements above 20 μg/dL . Three concentrations of NYSDOH (Albany, NY) blood-based reference materials (including one < 10 μg/dL) were analyzed before, during, and after each analytical run as part of the laboratory’s internal quality assurance program .
Average and peak blood lead concentrations were calculated from serial blood lead measures, similar to previous analyses in this cohort [21, 24]. Lifetime average blood lead level was computed by dividing the total area under each child’s age-by-blood-lead curve by 42 (42 = 48–6 months), and infancy average blood lead concentration was computed by dividing the total area under the child’s age-by-blood-lead curve by 18 (18 = 24–6 months). Peak infancy concentration is the child’s highest blood lead level measured from 6 to 24 months and peak 48-month concentration is the child’s highest blood lead level measured from 6 to 48 months. Infancy blood lead measures were included in the analysis due to the rapid development of the immune system during this timeframe and, thus, may serve as a critical window for exposure to lead [8, 25]. Concurrent blood lead level is the measurement taken on the day of the 48-month visit. As in previous analyses, missing age-specific blood lead levels were imputed using conditional means regression, utilizing the values of non-missing blood lead levels . The percentage of missing and imputed blood lead values were 2%, 7%, 8%, 14%, 16%, 19%, 14% and 4% at 6, 12, 18, 24, 36, 48, 60, and 72 months, respectively. In general, blood lead concentrations were moderately to strongly correlated, ranging from r = 0.21 (p < 0.05) between 6 and 24 months and r = 0.89 (p < 0.05) at 60 and 72 months, supporting the use of imputation.
At each visit, parents/guardians completed a questionnaire concerning the child’s health history, parental health, socioeconomic status, demographic information, and the home environment. Covariates of interest included data collected from the child’s birth record such as sex, birthweight (grams), and maternal parity. Maternal race (white/non-white) and whether the child was ever breastfed were collected at the child’s 6-month visit. The average adjusted income variable used in the analysis incorporated household income as well other sources of income (e.g., child support or government assistance) reported at the 54- through 72-month visits. This variable was calculated by averaging household income reported at 54, 60, 66, and 72 months, adding any average other income reported from that same time period, and subtracting out the average rent paid during that same period. Data collected at the child’s 72-month follow-up included maternal education (less than high school, high school/general education development (GED), more than high school), total number of cigarettes smoked per day in the household, and any child daycare attendance in the past 12 months.
Statistical models included blood lead concentration as a continuous variable or as a three-level ordinal classification variable with categories ≤ 5.0 μg/dL, 5.0–9.9 μg/dL, and ≥ 10.0 μg/dL. These groupings were chosen based on recent Centers for Disease Control and Prevention (CDC) reference levels, clinical utility, and as having significance for informing public health policy [26, 27]. For instance, a goal of Healthy People 2020 was to eliminate blood lead levels greater than 10 μg/dL, and presently, the CDC considers 5 μg/dL as an action level for the introduction of monitoring of the child and remediation in the home . Candidate confounders were identified from the existing literature [14, 17, 18, 28,29,30,31,32,33]. To identify the minimally sufficient set of adjustment variables, a directed acyclic graph (DAG) was constructed using DAGitty software (version 3.0) . The minimally sufficient set of variables identified from the DAG were child sex, birthweight, daycare attendance, maternal race, education, parity, ever having breastfed, average adjusted income, and total number of cigarettes smoked per day in the household.
The risk ratio for asthma and respective 95% confidence interval (CI) were estimated using log-binomial regression for each parameterization of lead. In total, 10 primary models were fit: five unadjusted, and five adjusted, examining (1) concurrent 48-month blood lead levels, (2) peak 48-month blood lead levels, (3) lifetime average at 48-month blood lead levels, (4) peak 24-month blood lead levels, and (5) infancy average at 24-month blood lead levels. As a secondary analysis, continuous parametrizations of age-specific lead measurements at 6, 12, 18, 24, 36, 48, 60, and 72 months of age were examined. Risk ratios and p values for linear trend were considered statistically significant if p < 0.05. All data management and analyses were performed using the SAS software system (SAS Institute Inc., Cary, NY, USA; version 9.4).