New genetic analysis complicates, not confirms, sleep apnea's link to dementia risk

A two-sample Mendelian randomization study reports that genetically predicted obstructive sleep apnea is tied to lower, not higher, odds of a general dementia diagnosis, a finding that runs against decades of observational research and that its own authors say needs replication.

A Mendelian randomization study published online in Archives of Medical Science on November 28, 2025 reports genetic evidence for a causal relationship between obstructive sleep apnea (OSA) and dementia, but the direction is the opposite of what most observational research has suggested. Using genome-wide association data from a European-ancestry sleep apnea study (126,695 OSA cases and 303,630 controls) and dementia diagnoses from the Finnish FinnGen consortium (24,864 general dementia cases, coded under ICD-10 F00 to F09), the researchers found that a higher genetic predisposition to OSA was associated with lower odds of a general dementia diagnosis (inverse-variance weighted odds ratio 0.830, 95 percent confidence interval 0.700 to 0.970), after outlier genetic variants were removed. This follows an earlier story in this section reporting no overall link between sleep apnea and a single blood amyloid marker; the new analysis takes a different approach, using genetic variants as instruments across dementia broadly rather than testing one biomarker.

The study also ran the analysis in reverse, treating dementia as the exposure and sleep apnea as the outcome. Genetic liability to general dementia was linked to lower odds of an OSA diagnosis (odds ratio 0.940, CI 0.919 to 0.962) and of a broader "generalized sleep apnea" diagnosis (odds ratio 0.914, CI 0.886 to 0.943); genetic liability to Alzheimer's disease showed a similar pattern with the generalized sleep apnea phenotype (odds ratio 0.968, CI 0.955 to 0.981). An initial signal linking vascular dementia to lower sleep apnea risk did not hold up: once the researchers removed outlier genetic variants flagged by their pleiotropy-detection method, that association was no longer statistically significant.

The authors themselves flag the counterintuitive direction of their headline result. In their own discussion, they note that the protective-looking association contradicts a substantial body of observational research that has reported OSA as a risk factor for dementia, and they offer two possible explanations rather than a conclusion: Mendelian randomization may simply be stripping out confounding that inflates the OSA-dementia link in observational cohorts, or the "unspecified dementia" category used in the Finnish health records, a broad group that includes several miscellaneous diagnoses, may not behave like more specific dementia subtypes. Either way, the paper does not resolve which explanation is correct.

The study leaves the underlying question, whether obstructive sleep apnea changes a person's dementia risk, more open than before rather than settled: a stronger method has produced a result that does not match the direction most cohort studies have reported, and the authors say the discrepancy itself needs to be explained before either finding can be treated as the last word.