Objectives: Our main objective was to attempt to replicate the findings of Cheslack-Postava and colleagues. A secondary objective was to examine whether maternal age, sex, and birth year—using a cutoff of 1998, the year Canada instituted a mandatory folic acid fortification program—act as effect modifiers of the IPI-autism association.
Methods: Records related to Manitoba births between 1988 and 2005 were extracted from population-based administrative datasets at the Manitoba Centre for Health Policy. First- and second-born siblings from the same mother were identified, applying the following exclusion criteria: multiple gestation, pregnancy loss between the first and second sibling, and autism in the first-born. Three case definitions were used, representing increasing probability of true-positive case status: 1) At least one autism code in the Education, Health, or Children’s Special Services datasets (the latter contains information on children identified for a Canadian autism surveillance program through a provincial agency that coordinates the provision of services for children with special needs in Manitoba); 2) Two or more codes for autism in one or more of the preceding datasets; and 3) a record in the Children’s Special Services dataset. Logistic regression models were fit to estimate the association between the IPI and autism while controlling for other pregnancy-related variables and sociodemographic factors.
Results: A total of 41,066 second-born siblings met the inclusion criteria. Using ≥36 months as the reference, the adjusted odds ratios from the models examining main effects increased from Case Group One (n=472) to Case Group Three (n=142) across all IPI categories, ranging from 1.23 (0.91-1.65) to 1.70 (0.95-3.05) for IPIs <12 months; 1.09 (0.83-1.43) to 1.56 (0.91-2.66) for IPIs of 12-23 months; and 1.06 (0.78-1.44) to 1.30 (0.71-2.39) for IPIs of 24-35 months. The interactions between IPI and maternal age, sex, and birth year were not significant, but the IPI odds ratios derived from those models attained significance within certain strata (e.g. maternal age 30-34 years), particularly for Case Group Three.
Conclusions: While the findings from the main-effects models were not significant, the point estimates and lower confidence limits for the shorter IPIs suggest an association, albeit a weaker one than observed in the California study. The difference in the proportions of our case and comparison groups with an IPI <12 months (2.1%-3.2%) was substantially smaller than that reported by Cheslack-Postava and colleagues (14.4%). Thus, it seems unlikely that the smaller number of cases in our study can explain the discrepant findings; rather, they may be due to differences in the two studies’ case groups or to variations in characteristics of the underlying populations that may influence the IPI-autism association, such as ethnic background. Further analyses are planned to explore the potential impact of these differences on the findings.
See more of: Epidemiology
See more of: Prevalence, Risk factors & Intervention