Removing One Study Reduced Cannabis Crash-Estimate Heterogeneity Without Producing Significance
A sensitivity analysis reduced inconsistency between the older-driver collision studies after one study was removed, but the pooled association remained nonsignificant. Greater consistency did not establish a precise or causal effect.
The initial collision results varied substantially
The review reported considerable heterogeneity among collision-involvement estimates. One included study showed a significant inverse trend, unlike the general direction of the other studies. The investigators examined how excluding it affected the synthesis.
The exclusion changed more than one property
The reported heterogeneity fell to 38%, and the summary odds ratio increased numerically. It nevertheless remained statistically nonsignificant. These are separate findings: reduced inconsistency is not equivalent to crossing a significance threshold, and a numerical increase does not by itself establish a meaningful increase in risk.
Sensitivity is diagnostic rather than a license to discard
An exclusion analysis shows how much a particular study influences the result. It does not prove that the excluded work was invalid or identify the correct biological effect. Methodological differences, selection and chance remain relevant. A complete account retains the main analysis and the sensitivity result, explaining what changed and what did not. This review’s evidence concerns selected observational comparisons in older drivers, not an experimental dose-response curve or an individualized impairment test.
Related analysis: the study’s main findings and limitations.
Research cited
Chinna-Meyyappan and colleagues. Risk of Motor Vehicle Collisions and Culpability among Older Drivers Using Cannabis: A Meta-Analysis. Brain Sciences (2023). DOI: 10.3390/brainsci13030421.
