Large Study Finds No Evidence Cannabis Use Accelerates Cognitive Decline or Raises Dementia Risk in Older Adults
By Dr. Benjamin Caplan, MD | Board-Certified Family Physician, CMO at CED Clinic | Evidence Watch
A large observational and genetic study drawing from UK Biobank and the Million Veteran Program found no significant association between cannabis use and accelerated cognitive decline or increased dementia risk in older adults. While these null findings are broadly reassuring, the crude way cannabis exposure was measured in both cohorts means clinicians should not yet treat this as definitive proof of long-term cognitive safety, particularly for heavy or prolonged users.
Large Study Finds No Evidence Cannabis Use Accelerates Cognitive Decline or Raises Dementia Risk in Older Adults
Researchers caution that imprecise exposure measurement, older-cohort survivor biases, and the absence of dose, duration, and product-type data limit the degree of confidence that can be drawn from these otherwise reassuring findings across two of the largest population cohorts studied to date.
#72
Strong Clinical Relevance
Directly addresses a common clinical question about cannabis and cognitive aging in a rapidly growing patient population, though exposure measurement gaps temper confidence.
Dementia Risk
Older Adults
Mendelian Randomisation
Observational Epidemiology
Cannabis use among adults over 50 has risen sharply in the past decade, driven by expanding legalization and growing interest in cannabis for age-related conditions including chronic pain, insomnia, and anxiety. Despite this surge, rigorous evidence on whether cannabis use affects long-term cognitive trajectories or dementia risk in older populations has been remarkably thin. Patients and clinicians alike need credible data to weigh the potential benefits of cannabis use against the theoretical risk of hastening cognitive decline during the very years when neurodegeneration accelerates naturally.
The question of whether cannabis use contributes to cognitive aging has lingered for years without adequate large-scale investigation. This study draws on two of the largest available population cohorts to address the gap. In the UK Biobank (UKB), researchers examined cross-sectional and longitudinal cognitive performance across five domains in nearly 80,000 participants, comparing those who reported ever using cannabis to those who did not. In the Million Veteran Program (MVP), they modeled incident all-cause dementia among more than 222,000 veterans, using a clinical diagnosis of cannabis use disorder (CanUD) as the exposure variable. A Mendelian randomisation (MR) analysis was layered on top to probe whether any observed associations might reflect causal relationships rather than confounding.
The headline finding is one of absence: cannabis use was not significantly associated with faster cognitive decline in UKB, and CanUD was not significantly associated with incident dementia in MVP (hazard ratio 1.11, 95% confidence interval 0.97 to 1.26, p=0.12). Cannabis users actually scored slightly higher on baseline cognitive tests, though this almost certainly reflects healthy-user selection bias rather than a genuine cognitive benefit. MR analyses found no evidence of causal effects in either direction. The authors are forthright about the study’s central limitation: cannabis exposure was measured only as binary lifetime use (UKB) or clinical disorder diagnosis (MVP), with no data on dose, frequency, duration, product type, or route of administration. Coupled with survivor bias inherent in older cohorts and the possibility that the confidence interval for the dementia hazard ratio does not fully exclude a modest excess risk, these null findings should be read as cautiously reassuring rather than definitively exonerating.
I appreciate the scale and methodological ambition of this study. It brings Mendelian randomisation to a question that has been plagued by confounding for decades, and the null findings across two independent cohorts carry real weight. That said, I want to be clear about what “no significant association” does and does not mean. The confidence interval around the dementia hazard ratio still stretches up to 1.26, which would represent a clinically meaningful risk if it turned out to be real. And the exposure definitions used here are so coarse that a person who tried cannabis once at age 20 is grouped with someone who has used it daily for 40 years. That matters enormously for the clinical question patients are actually asking.
In my practice, I share findings like these with patients who worry that their cannabis use may be contributing to cognitive problems, and I find them genuinely reassuring at a population level. But I do not use them to dismiss concerns about heavy or escalating use in someone already showing subtle cognitive changes. I still counsel older patients to use the lowest effective dose, favor products with known cannabinoid profiles, and to remain vigilant about changes in memory or executive function. The absence of evidence of harm is not the same as evidence of absence, and until we have studies that capture dose and duration properly, clinical prudence remains the appropriate posture.
This study sits at an important inflection point in the research arc. Prior work on cannabis and cognition has been dominated by small cross-sectional studies, often in younger populations and frequently complicated by polydrug use. The few longitudinal studies that exist in older adults have been underpowered and inconsistent. By combining large population cohorts with MR analyses, this paper raises the methodological bar substantially. However, it also makes clear how much further the field needs to go. The two cohorts capture fundamentally different exposure strata: any lifetime use in UKB and clinically diagnosed disorder in MVP. The middle ground, where most older cannabis users actually live (regular but not disordered use), remains essentially unstudied at this scale.
From a pharmacological standpoint, clinicians should note that cannabis interacts with the endocannabinoid system in ways that are both neuroprotective and potentially neurotoxic depending on dose, ratio of THC to CBD, and individual vulnerability. High-THC products may carry different risk profiles than balanced or CBD-dominant preparations, but this study cannot distinguish between them. Safety considerations for older patients also include the well-documented acute effects on attention and psychomotor performance, which can compound fall risk and driving hazards independent
Have thoughts on this? Share it:
