Reevaluating the Association Between Autism and Lithium Exposure During Pregnancy—Reply

In Reply In response to our article, Barbanti Zancheta et al asked why we did not include some known risk factors for autism as confounders in our models. We left out gestational age as it is likely a mediator in the causal pathway that would induce bias if included, while there is no apparent reason to think that paternal age or child age at diagnosis affects maternal residential lithium exposure levels. In terms of the study design and effect estimates we chose, we would like to point out that a nested case-control design is a well established and cost-efficient epidemiologic approach to estimate exposure effects (odds ratios) for rare outcomes. We extracted autism spectrum disorder (ASD) diagnoses—primarily made by child psychiatrists following International Statistical Classification of Diseases and Related Health Problems, Tenth Revision (ICD-10) criteria—from the Danish Psychiatric Registry. As our study focused on the prenatal period, a critical window for neurodevelopment, our lithium exposure assessment was specific for the pregnancy period only and did not include the periods before and after pregnancy. Thus, we targeted an exposure period relevant to our hypothesis that is quite different from studies measuring chemical elements in the blood of children already affected by ASD. Although lithium was measured at waterworks that supply only half of the Danish population, these were selected to represent all regions of Denmark. Thus, our model has nationwide coverage for the Danish population. A study is considered ecological if it relies on group-level exposure and group-level outcome data. This is different from the individual-level case-control study we conducted using individual-level records for child diagnosis and individual-level exposure measures based on geocoded addresses during the mother’s pregnancy.