A large new study analyzing nearly 200,000 matched patient pairs has delivered a counterintuitive finding that may uproot common assumptions about the safety of melatonin: adults over 50 prescribed melatonin went on to develop dementia and Parkinson’s disease at significantly higher rates than those prescribed benzodiazepines (BZDs) or zolpidem.
Published June 24 in Scientific Reports by a Taiwanese research team (Chang, Su, Yang, Chen, Wu, and Gau), the study used a target trial emulation design on the TriNetX Global Collaborative Network, pulling electronic health records from 2015 through 2024. After rigorous propensity-score matching, the analysis compared 189,858 pairs of new melatonin users versus new BZD users, and 190,889 pairs of melatonin users versus zolpidem users.
The headline numbers are striking. Compared with BZD users, melatonin users had more than double the risk of developing all-cause dementia (HR 2.09, 95% CI 2.01-2.16) and Alzheimer’s disease (HR 2.11, 95% CI 1.96-2.26), a 63% higher risk of Parkinson’s disease (HR 1.63, 95% CI 1.51-1.76), and more than double the risk of vascular dementia (HR 2.39, 95% CI 2.21-2.60). The pattern held when melatonin was compared against zolpidem, though the hazard ratios were somewhat lower: HR 1.74 for all-cause dementia, 1.54 for Alzheimer’s, 1.44 for Parkinson’s, and 1.99 for vascular dementia. All findings were statistically robust with narrow confidence intervals, reflecting the enormous sample size.
Median follow-up ranged from 1.88 to 3.70 years depending on the comparison and outcome.
The critical caveat
These results might seem alarming at first glance, but the authors themselves issue a forceful warning against causal interpretation. “These findings should NOT be interpreted as causal,” they write. The most likely explanation, they argue, is confounding by indication and residual confounding.
The core problem is straightforward: patients who are prescribed melatonin are systematically different from those prescribed benzodiazepines or zolpidem, and standard matching may not fully capture those differences. Melatonin is often prescribed as a “milder” sleep aid for patients who are already perceived as vulnerable – older adults with early cognitive complaints, people with complex comorbidities, or those whose doctors are already worried about neurodegenerative risk. In other words, the higher dementia and Parkinson’s rates in melatonin users may simply reflect that these patients were already on a trajectory toward neurodegenerative disease before they ever started the supplement. The melatonin is correlated with that risk, not causing it.
This phenomenon, known as confounding by indication or protopathic bias, is a well-known challenge in pharmacoepidemiology. It is especially difficult to resolve when the condition being studied (early-stage dementia or Parkinson’s) develops insidiously over years and may already be present in prodromal form at the time of treatment initiation.
Challenging the neuroprotection narrative
The finding is notable because it directly challenges a widely held assumption. Melatonin is available over the counter in many countries and is often perceived by both the public and clinicians as not just safe but potentially neuroprotective. Some preclinical research and small human studies have suggested melatonin may reduce oxidative stress, modulate neuroinflammation, and even slow cognitive decline. These results, however, point in the opposite direction, though the authors caution that the study cannot test neuroprotection claims – it only compares melatonin users against users of other sleep medications in a real-world clinical setting.
What this means clinically
For clinicians, the immediate takeaway is not that melatonin is dangerous, but that its widespread use in older adults should not be assumed risk-free or protective. The study underscores that all sleep aids carry trade-offs, and that prescribing decisions in older populations should be made with careful attention to the patient’s full clinical picture, including early signs of neurodegeneration that may be subtle and easily missed.
Benzodiazepines and zolpidem carry their own well-documented risks – falls, fractures, dependence, and cognitive impairment in their own right – which this study was not designed to compare on a level playing field. The study’s head-to-head numbers should not be read as evidence that BZDs or zolpidem are “safer” for the brain, only that the observed risk profile for melatonin was unexpectedly worse in this particular analysis.
Limitations
Beyond the central issue of confounding by indication, the study has several other important limitations. As a retrospective database analysis using target trial emulation, it cannot establish causation. Follow-up was relatively short at under four years median, which may be insufficient to capture long-term neurodegenerative trajectories. Medication adherence was assumed from prescription records but not verified. Dosing information and over-the-counter melatonin use – which may differ in important ways from prescription use – were not captured. The database also lacks detailed information on lifestyle factors, family history, and other unmeasured confounders that could influence dementia risk.
The authors note that future prospective studies with longer follow-up, detailed cognitive assessments, and better control for confounding are needed to clarify whether the observed associations represent a real causal effect, residual bias, or some combination of both.
For now, the study serves as a sobering reminder that “safe” is not the same as “studied,” and that even a widely used over-the-counter supplement deserves the same pharmacovigilance scrutiny as any prescription drug – especially in vulnerable older adults.
Source: Chang HC, Su YJ, Yang SC, Chen CC, Wu MC, Gau SY. Comparative risk of dementia and Parkinson’s disease associated with melatonin, benzodiazepines, and zolpidem in adults with sleep disorders: a target trial emulation study. Scientific Reports. 2026. DOI: 10.1038/s41598-026-58529-4. PMID: 42342736.