Is cannabis exposure associated with childhood leukemia rates across US states?

Cannabis use was independently associated with pediatric ALL rates in multiple regression models adjusted for other substances, income, and ethnicity. A dose-response relationship was observed across cannabis use quintiles (Chi.Sq.=613.79, p=3.04x10^-70). Cannabis-legal states had higher ALL rates (2.395 vs 2.127/100,000, p=5.05x10^-10). Spatiotemporal models showed associations with specific cannabinoids including THC and cannabigerol.

Ecological geospatial analysis combining state-level ALL rates from CDC/NCI SEER database with drug use data from the National Survey of Drug Use and Health (74.1% response rate), cannabinoid concentration data from the DEA, and census data. Analyzed using robust regression, spatiotemporal models, inverse probability weighting, and causal inference methods including e-Values.

Pediatric ALL rates in the US have risen 93.5% between 1975 and 2016 for unclear reasons. If cannabis exposure contributes to this rise, it would be a significant public health concern as cannabis use and legalization expand.

This study makes a provocative causal claim based on ecological data. While the statistical associations are striking, ecological studies correlating state-level exposures with state-level disease rates cannot establish that individual cannabis users' children develop leukemia. The findings warrant further investigation with individual-level data.

Ecological design (state-level correlations cannot prove individual-level causation). Many potential confounders at the state level. Cannabis use data is self-reported. The authors make strong causal claims that exceed what ecological data can support. Multiple comparisons increase risk of spurious findings.