Large Cohort Study Links Statin Use to Higher Dementia Risk

Statin use was associated with a significantly higher risk of dementia in two of the world’s largest biomedical databases, according to a population-based retrospective cohort study published in Frontiers in Pharmacology. In the All of Us Research Program, statin users had more than double the dementia hazard of non-users after weighting-based adjustment, and the association was replicated in the UK Biobank. The authors stress that the observational design and the conflicting state of the existing literature mean the findings should not yet change how anyone thinks about statin therapy.

The research team, led by investigators at Ewha Womans University and Chungbuk National University in South Korea, analyzed adults aged 40 or older who self-identified as White. The discovery cohort drew on the All of Us Research Program, a U.S. National Institutes of Health initiative, using 83,355 participants of whom 31,513 were statin users, with follow-up data through October 2023. For replication, they turned to the UK Biobank, analyzing 173,282 participants including 43,976 statin users. People with dementia, stroke, Parkinson’s disease, or traumatic brain injury before their index date were excluded, and new users of statins were compared with non-users assigned a randomly sampled pseudo-index date.

Because statin users differed substantially from non-users at baseline, the researchers applied inverse probability of treatment weighting to balance measured characteristics. Before weighting, statin users in the discovery cohort were more frequently male, older, and had a markedly heavier comorbidity burden, including hypertension in 49.5 percent versus 16.1 percent of non-users, diabetes in 18.4 percent versus 3.7 percent, and coronary artery disease in 17.5 percent versus 3.6 percent. After weighting, all standardized mean differences fell below 0.1, and dementia risk was estimated with Cox proportional hazards models adjusted for sex, age group, index year, body mass index, APOE genotype, comorbidities, and concomitant medications.

In the discovery cohort, 5,968 incident dementia cases occurred. Statin use was associated with an adjusted hazard ratio of 2.22 (95 percent confidence interval 2.02 to 2.44), with incidence rates of 64.9 versus 30.7 per 10,000 person-years in users and non-users. Subtype analyses showed consistent patterns: Alzheimer’s disease carried an adjusted hazard ratio of 2.02, vascular dementia 2.30, and other dementia types 2.23. The association persisted across all age groups examined and across all APOE ε4 strata, indicating that the pattern did not depend on this well-established genetic risk factor for cognitive decline.

Dose appeared to matter. After excluding users with missing intensity data, moderate-to-low intensity statin use was associated with an adjusted hazard ratio of 2.08 relative to non-use, while high-intensity use showed an even larger hazard ratio of 2.70, with incidence rates of 59.0 and 84.6 per 10,000 person-years respectively. In a direct comparison within statin users, high-intensity therapy carried 14 percent greater dementia hazard than moderate-to-low intensity therapy (adjusted hazard ratio 1.14, 95 percent confidence interval 1.07 to 1.21). Both lipophilic and hydrophilic statins were associated with elevated risk, with hydrophilic statins showing slightly higher point estimates that did not reach statistical significance in direct comparisons.

The replication cohort told a consistent story at lower absolute rates. Dementia incidence was 11.8 per 10,000 person-years among statin users versus 6.2 among non-users, and statin use was associated with an adjusted hazard ratio of 2.54 (95 percent confidence interval 1.98 to 3.33). The authors attribute the large difference in absolute rates between cohorts partly to follow-up length: when the discovery analysis was restricted to a comparable period ending in May 2017, incidence rates became much more similar between the two datasets. Three sensitivity analyses, restricting to at least three years of follow-up, applying propensity-score overlap weighting, and fitting Fine-Gray models treating death as a competing event, all left the associations intact in both cohorts.

To explore possible biology, the team ran an exploratory mediation analysis in the UK Biobank Pharma Proteomics Project on 38,553 participants, 6,108 of them statin users. Because direct measurements of circulating coenzyme Q10 are unavailable in large proteomic datasets, they used COQ7, a protein involved in the late stages of CoQ10 biosynthesis, as a proxy. Statin users had significantly lower COQ7 levels, a decrease of about 0.11 standard deviations, and lower LDL cholesterol. The estimated indirect association through COQ7 was statistically significant and accounted for 8.2 percent of the total association, while the indirect path through LDL cholesterol was not statistically significant. Separately, higher COQ7 levels were associated with lower odds of dementia, with spline analysis suggesting the association was most pronounced at lower COQ7 levels.

The LDL cholesterol result is notable given decades of debate about whether cholesterol lowering itself affects the brain. Although statin use was strongly associated with reduced LDL cholesterol, and lower LDL cholesterol showed a modest association with higher dementia odds, the mediated effect through LDL cholesterol did not reach significance. The authors note that prior studies of LDL cholesterol and dementia have been inconsistent, and that a recent long-term randomized trial found no evidence of cognitive harm despite sustained very low LDL cholesterol levels, so the observed association in this study was not explained by cholesterol reduction in these analyses.

The limitations are substantial and the authors enumerate them candidly. As an observational study, causal inference is restricted despite robust adjustment; residual confounding and confounding by indication remain possible, since underlying cardiovascular risk may influence both statin prescribing and dementia risk, and smoking and alcohol use could not be included in the prescription-based analyses. In the mediation work, COQ7 was measured only once during regular statin use with no pre-treatment baseline, so temporal ordering could not be established, and the models rest on the unverifiable assumption of no unmeasured confounding. The authors explicitly label the mediation findings hypothesis-generating, noting that the 8.2 percent proportion mediated is a modest statistical association that does not quantify absolute impact on dementia risk.

The broader literature remains contradictory: some observational studies suggest increased dementia risk with statins, others find no association, and a recent meta-analysis reported even protective effects. The U.S. Food and Drug Administration added information about potential cognitive side effects to statin labeling in 2012 and later clarified in 2016 that such events were generally reversible. Against that backdrop, this study’s contribution is its scale, its two-cohort replication, and its consistent sensitivity analyses, alongside a specific, testable hypothesis about CoQ10 biosynthesis-related pathways. The authors conclude that further longitudinal and experimental studies are needed before these associations carry any clinical implications for the millions of adults who take statins to prevent cardiovascular disease.

Subject of Research: Observational association between statin therapy and dementia risk in the All of Us and UK Biobank cohorts with exploratory COQ7 mediation analysis

Article Title: Association between statin therapy and dementia risk: evidence from All of Us and UK Biobank with coenzyme Q7-related mediation analysis

Article References: Lee, D. H., Park, Y.-A., Kim, J. S., Yee, J., Han, J. M., & Gwak, H. S. (2026). Association between statin therapy and dementia risk: evidence from All of Us and UK Biobank with coenzyme Q7-related mediation analysis. Frontiers in Pharmacology, 17, Article 1915633. https://doi.org/10.3389/fphar.2026.1915633

Image Credits: AI Generated

DOI: 10.3389/fphar.2026.1915633

Keywords: statins, dementia, All of Us Research Program, UK Biobank, COQ7, coenzyme Q10, pharmacoepidemiology, LDL cholesterol, observational study, mediation analysis, Alzheimer's disease, cardiovascular prevention