Impact of converging sociocultural and substance-related trends on US autism rates: combined geospatiotemporal and causal inferential analysis

Abstract

AbstractWhilst cannabis is known to be toxic to brain development, it is unknown if it is driving rising US autism rates (ASMR). A longitudinal epidemiological study was conducted using national autism census data from the US Department of Education Individuals with Disabilities Act (IDEA) 1991–2011 and nationally representative drug exposure (cigarettes, alcohol, analgesic, and cocaine abuse, and cannabis use monthly, daily, and in pregnancy) datasets from National Survey of Drug Use and Health and US Census (income and ethnicity) and CDC Wonder population and birth data. Analysis was conducted in R. 266,950 were autistic of a population of 40,119,464 8-year-olds in 1994–2011. At national level after adjustment, daily cannabis use was significantly related to ASMR (β estimate = 4.37 (95%C.I. 4.06, 4.68), P < 2.2 × 10–16) as was first pregnancy trimester cannabis exposure (β estimate = 0.12 (0.08, 0.16), P = 1.7 × 10–12). At state level following adjustment for cannabis, cannabigerol (from β estimate = – 13.77 (– 19.41, 8.13), P = 1.8 × 10–6) and Δ9-tetrahydrocannabinol (from β estimate = 1.96 (0.88–3.04), P = 4 × 10–4) were significant. Geospatial state-level modelling showed exponential relationship between ASMR and Δ9-tetrahydrocannabinol and cannabigerol exposure. Exponential coefficients for the relationship between modelled ASMR and Δ9-tetrahydrocannabinol and cannabigerol exposure were 7.053 (6.39–7.71) and 185.334 (167.88–202.79; both P < 2.0 × 10–7). E-values are an instrument related to the evidence for causality in observational studies. High E-values were noted. Dichotomized legal status was linked with elevated ASMR. Data show cannabis use is associated with ASMR, is powerful enough to affect overall trends, and persists after controlling for other major covariates. Cannabinoids are exponentially associated with ASMR. The cannabis–autism relationship satisfies criteria of causal inference.