Mendelian randomization is what separates this from another pooled correlation, and it is also where the coverage runs out. The design asks whether genetic variants associated with an exposure also track with the outcome [12], which addresses the objection the authors raise about their own observational inputs: people who use drugs differ from people who do not in ways that independently move stroke risk [16]. Here the genetic evidence supported links between cocaine use disorder and both brain hemorrhage and cardioembolic stroke, in which a clot travels to the brain [13], and between cannabis use disorder and stroke overall, with large-artery stroke standing out [14]. Amphetamines got no such test, because suitable data were not available [15]. Study author Megan Ritson called the work the most comprehensive analysis ever conducted on recreational drug use and stroke, and compelling evidence that cocaine, amphetamines and cannabis are causal risk factors [10].
That leaves the causal argument oddly distributed. The substance with the steepest under-55 association, amphetamines at a 174 percent increase [7], which is a risk ratio near 2.74 [4], rests entirely on observational pooling. The substance with the shallowest under-55 association, cannabis at 14 percent [9], is one of the two carrying genetic support [5].
Note also what the genetic arm indexes: use disorder, a diagnosed pattern, rather than use [13][14]. The pooled percentage increases are about use [4][5][6]. So the claim from Eric Harshfield that it is the drugs themselves raising risk, not just other lifestyle factors among users [11], is best supported at the dependent end of the exposure range, which is not where most people reporting cannabis use sit.
The age stratification rewards a second look. Cannabis in under-55s is about 38 percent of its all-age value, 14 divided by 37 [1]; amphetamines rises to roughly 1.43 times its all-age figure, 174 over 122 [2]; cocaine barely moves, 97 against 96 [3]. The mechanisms the paper lists run on different clocks. Long-term cocaine use accelerating plaque buildup takes years, while blood pressure surges, vessel narrowing and spasm, and amphetamine-associated vessel inflammation are fast [17][19]. An age pattern in which the acute mechanisms dominate the young subgroup is consistent with that reading, though the analysis as reported does not test it, and does not say how many of the 32 studies fed each age-stratified estimate [8].
The 100 million is the pooled population of the contributing studies [2][1], not a count of users and not a count of strokes, and neither of those counts appears in the reporting available, nor do confidence intervals [7]. There is also no baseline. Without absolute stroke incidence in under-55s, a 14 percent relative increase cannot be converted into extra strokes per thousand people [6]. The authors are explicit that the results do not mean every recreational drug user will have a stroke [21].
The thing this does not tell you is whether asking changes outcomes. Nothing in the design tests whether a drug history taken from a young patient with a stroke syndrome alters management, and the implication the authors themselves draw is public health measures to reduce substance abuse [11]. Even so, associations near a doubling for cocaine and near a tripling for amphetamines in the under-55 group [8][7] are large enough that exposure belongs in the differential rather than in the social history. The cannabis figure I would hold more loosely, for the two reasons above: the smallest age-specific effect, and a genetic test run on a heavier exposure than most users have.
