Prenatal pet keeping and caregiver-reported attention deficit hyperactivity disorder through preadolescence in a United States birth cohort

WHEALS recruited pregnant women with due dates from September 2003 through December 2007, and who were seeing a practitioner at 1 of 5 clinics in the Henry Ford Health System to establish a birth cohort [11, 34]. All women were in their second trimester or later, were aged 21–49 years, and were living in a predefined geographic area in western Wayne County that included the western portion of the city of Detroit as well as the suburban areas immediately surrounding the city. Mothers were interviewed in the clinic prenatally and postpartum interviewer-administered questionnaires were completed at child age 1, 6, 12, and 24 months. Children and their parent/guardian were invited to return for a clinic visit at child age 2 years and again at child age 10–12 years for assessment of child health. All participants provided written, informed consent and study protocols were approved by the Henry Ford Health System Institutional Review Board; at the age 10–12 year visit, children provided written, informed assent.

At the age 10–12 year visit, the caregiver (95% the mother) reported if the child had ever been diagnosed with ADHD. The caregiver also reported if the child had ever been diagnosed with autism spectrum disorder (ASD), Asperger’s syndrome, or sensory processing disorder (SPD); Asperger’s syndrome was asked separately since children may have been diagnosed prior to the Diagnostic and Statistical Manual of Mental Disorders, Fifth Edition (DSM-5). A report of “suspect” diagnosis was also considered positive. Children were considered neurotypical (NT) if they did not have a caregiver-reported ASD, Asperger’s, ADHD, or SPD diagnosis. Children without an ADHD diagnosis but who had a caregiver-reported ASD, Asperger’s or SPD diagnosis were excluded from the analysis, given the small sample size with these conditions.

A subset of 325 WHEALS children had their medical record abstracted for additional health information, including ADHD diagnosis. Caregiver report of ADHD diagnosis was validated within this subset using the kappa (κ) statistic to evaluate agreement. Landis and Koch [35] criteria were used to evaluate strength of agreement. There was near perfect agreement between caregiver-reported ADHD and the medical record (κ = 0.84, 95% CI 0.78, 0.91). Given that only a subset of children had medical record review and the high level of agreement between caregiver report and the medical record, we utilized caregiver-reported ADHD in statistical analysis.

During the prenatal maternal interview, the mother was asked about presence of pets in the home, the number of hours they spent indoors, and the type of pet (e.g., dog or cat). Some studies have attributed health benefits primarily to dog ownership [2, 36, 37]; thus, we defined prenatal dog keeping and prenatal cat keeping seperately, as report of having a dog or cat, respectively, indoors at least 1 h per day.

The mother self-reported race, date of birth, marital status, household income, education, parity, smoking history and exposure to environmental tobacco smoke at the prenatal interview. Maternal prenatal care records were abstracted to obtain antibiotic and antifungal use [38]. Height and weight were also abstracted and body mass index (BMI) at first prenatal care visit (mean gestational age at measure 9.1 ± 4.9 weeks; 83% taken during the first trimester) was calculated. Although first measured BMI in pregnancy and self-reported prepregnancy BMI have been shown to be highly correlated in other studies [39], the first measured BMI during pregnancy represents both prepregnancy body size and early pregnancy-related weight gain. Delivery records for WHEALS women were abstracted to obtain delivery mode, birth weight, and gestational age at delivery. Low birth weight was defined as a birthweight < 2500 g and preterm delivery as gestational age at delivery < 37 weeks.

Differences in basic characteristics by ADHD vs. NT development were first compared using standard parametric statistical methods, including chi-square tests and analysis of variance. Prenatal pet exposure (dogs and cats, separately) was then associated with ADHD using logistic regression to obtain odds ratios (OR) and 95% CIs. Given the growing and consistent evidence that male fetuses are more vulnerable than female fetuses to exposures during gestation [40], sex-specific effects of pets on ADHD was a priori hypothesized and formally tested using interaction terms in logistic regression models. When the interaction term was statistically significant, sex-specific models were fit to aid in interpretation.

Not all WHEALS children/caregiver pairs completed a 10–12 year questionnaire (a comparison of those with and without 10–12 year questionnaire data is presented in Table 1). Inverse probability weights (IPW) [41‐43] were used to account for loss to follow-up by age 10–12 years, defined by completion of the 10–12-year questionnaire. Weights were calculated as the inverse of the “treatment” received. In other words, if p = probability of follow-up, then w = 1/p for children with follow-up, and w = 1/(1-p) for children without follow-up. Briefly, IPWs up weight the importance of children who were followed through age 10–12 years but were not very likely to have been followed based on baseline characteristics (e.g., children of mothers who smoked during pregnancy). Similarly, it down weights the importance of children who were followed through age 10–12 years but were very likely to have been followed (e.g., children of mothers with a high level of education). The following factors were included in the calculation of the IPWs: maternal race, insurance, household income, maternal education, firstborn child, maternal smoking during pregnancy, prenatal environmental tobacco smoke exposure, urban residence, marital status, mode of delivery, prenatal alcohol use, child sex, prenatal dogs, prenatal cats, maternal history of allergies and asthma, maternal age at delivery, gestational age at delivery, birthweight, and completion of the 2-year clinic visit. Balance in these covariates was assessed using standardized differences before and after weighting, with imbalance defined as absolute value > 0.20. In addition to weighting the models to account for loss to follow-up, models were additionally adjusted for potentially confounding covariates. Confounders were selected based on having a significant association with either prenatal pet keeping or ADHD. These adjustment covariates included maternal race, household income, maternal smoking during pregnancy, prenatal antifungal use, child sex, maternal BMI first measured in pregnancy, gestational age at delivery, and birthweight.

If complete case analysis were used in the adjusted models, approximately 25% of the data would be excluded. To better handle this missingness, multiple imputation was used; all adjustment covariates as well as exposures (prenatal dogs/cats) and outcome (ADHD) were used to calculate 25 total imputed datasets. Each imputed dataset was modeled using logistic regression (with IPWs normalized to sum to the actual sample size as well as confounders included); final effect estimates were obtained by pooling results. All analyses were performed using SAS version 9.4 (SAS Institute Inc., Cary, NC) and R version 3.4.1.