Impact of Converging Sociocultural and Cannabinoid-Related Trends on US Autism Rates Dataset: Combined Geospatiotemporal and Causal Inferential Analysis

Published: 18 June 2020| Version 1 | DOI: 10.17632/p7myt3fbzs.1
Albert Reece


Background: Whilst cannabis is known to be toxic to brain function and brain development in many respects it is not known if its increasing availability is associated with the rising US autism rates, whether this contribution is sufficient to effect overall trends and if its effects persist after controlling for other major covariates. Methods: Longitudinal epidemiological study 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. Geotemporospatial and causal inference analysis conducted in R. Results: 266,950 autistic of a population of 40,119,464 eight year olds 1994-2011. At the national level after adjustment daily cannabis use was significantly related (β-estimate=4.37 (95%C.I. 4.06-4.68), P<2.2x10-16) as was cannabis exposure in the first trimester of pregnancy (β-estimate=0.12 (0.08-0.16), P=1.7x10-12). At the state level following adjustment cannabis use was significant (from β-estimate=8.41 (3.08-13.74), P=0.002); after adjustment for varying cannabis exposure by ethnicity and other covariates (from β-estimate=10.88 (5.97-15.79), P=1.4x10-5). Cannabigerol (from β-estimate=-13.77 (-19.41—8.13), P = 1.8x10-6) and Δ9-tetrahydrocannabinol (from β-estimate=1.96 (0.88-3.04), P=4x10-4) were also significant. Geospatial state-level modelling showed an exponential relationship between ASMR and both Δ9-tetrahydrocannabinol and cannabigerol exposure; effect size calculations reflected this exponentiation. Exponential coefficients for the relationship between modelled ASMR and THC- and cannabigerol- exposure were 7.053 (6.39-7.71) and 185.334 (167.88-202.79; both P<2.0x10-7). In inverse probability-weighted robust generalized linear models ethnic cannabis exposure (from β-estimate=3.64 (2.94-4.34), P=5.9x10-13) and cannabis independently (β-estimate=1.08 (0.63-1.54), P=2.9x10-5) were significant. High eValues in geospatial models indicated that uncontrolled confounding did not explain these findings. Therefore the demonstrated relationship satified the criteria of causal inference. Dichotomized legal status was geospatiotemporally linked with elevated ASMR. Conclusions: Data show cannabis use is associated with ASMR, is powerful enough to affect overall trends, and persists after controlling for other major drug, socioeconomic, and ethnic-related covariates. Selected cannabinoids are exponentially associated with ASMR. The cannabis-autism relationship satisfies criteria of causal inference.



Edith Cowan University, University of Western Australia


Epidemiology, Pediatrics, Cannabinoid, Autism, Autism Spectrum Disorder, Pediatric Neurology, Cannabis Abuse