Impact of Menopausal Hormone Therapy on cognitive function and dementia Risk: A systematic review and meta-analysis focusing on the timing hypothesis and administration route

Abstract

The relationship between menopausal hormone therapy (MHT) and the trajectory of cognitive function in midlife and later life remains one of the most consequential and contested questions in women’s health. The “critical window” or timing hypothesis posits that the cognitive consequences of exogenous estrogen depend on when therapy is initiated relative to the menopausal transition, and on which formulation, route, dose, and duration are used. After two decades of randomised trials and large pharmacoepidemiological cohorts, the evidence has matured but remains heterogeneous, and clinical guidance still varies widely across regions. We aimed to (i) systematically synthesise contemporary evidence on the effect of MHT on cognitive performance, mild cognitive impairment (MCI), and incident all-cause dementia or Alzheimer’s disease (AD); (ii) quantitatively examine effect modification by timing of initiation, formulation (estrogen-only versus estrogen plus progestogen), route (oral, transdermal, vaginal), and duration; and (iii) evaluate the methodological quality of the evidence base and identify clinically actionable recommendations. We searched PubMed/MEDLINE, Embase, the Cochrane Central Register of Controlled Trials, and Web of Science from inception to 7 May 2026 for randomised controlled trials (RCTs), prospective cohort studies, nested case–control studies, and high-quality systematic reviews investigating MHT and cognitive or dementia outcomes. Two reviewers independently screened citations and extracted data; risk of bias was assessed with the Cochrane RoB 2 tool for RCTs and the Newcastle–Ottawa Scale for observational studies. Where ≥ 2 estimates per pre-specified subgroup were available, we pooled adjusted relative risks (aRRs), hazard ratios (HRs), or odds ratios (ORs) using inverse-variance random-effects meta-analysis with the DerSimonian–Laird estimator and report between-study heterogeneity using Cochran’s Q and Higgins’ I2. Pre-specified subgroups were defined by timing of initiation (< 5 years versus > 5 years from final menstrual period), formulation (estrogen-only versus combined estrogen–progestogen), route (oral, transdermal, vaginal), and duration (< 5 versus 5–9 versus ≥ 10 years). Of 7,271 records identified, 65 studies met the inclusion criteria for qualitative synthesis and 12 studies contributing 18 effect estimates were eligible for quantitative meta-analysis. The pooled aRR for any MHT use versus never-use on incident all-cause dementia was 1.06 (95% confidence interval [CI] 0.97 to 1.15; k = 4; I2 = 96%), reflecting marked clinical and methodological heterogeneity that was substantially explained by stratification on timing, formulation, and duration. Initiation within five years of menopause was associated with a clinically meaningful reduction in dementia risk (pooled aRR 0.66, 95% CI 0.53 to 0.83; k = 3; I2 = 0%; p = 0.0003), whereas late initiation showed no protective effect (pooled aRR 1.02, 95% CI 0.80 to 1.29; k = 2; I2 = 0%). Longterm combined estrogen–progestogen use was associated with elevated dementia risk (pooled aRR 1.22, 95% CI 1.10 to 1.36; k = 4; I2 = 89%; p = 0.0002), while estrogen-only long-term use yielded a heterogeneous null estimate (pooled aRR 0.97, 95% CI 0.76 to 1.23; I2 = 95%). An exploratory sub-analysis restricted to transdermal estradiol regimens yielded a non-significant neutral estimate (pooled aRR 0.89, 95% CI 0.74 to 1.07; k = 2; I2 = 68%). Vaginal-only estrogen carried no detectable risk increment (pooled aRR 0.99, 95% CI 0.96 to 1.02; I2 = 0%). The evidence is consistent with a clinically operative critical window during which short-term, predominantly estrogenonly or transdermal MHT initiated near the time of menopause is associated with neutral or modestly protective cognitive trajectories, whereas long-duration, oral, combined estrogen–progestogen regimens initiated late are associated with increased dementia risk. However, the pooled combined-therapy estimates are dominated by legacy oral CEE/MPA exposures (~ 72% of contributing data) and cannot be assumed to generalise to contemporary transdermal 17β-estradiol/micronized progesterone regimens, which differ fundamentally in pharmacokinetic and neurobiochemical profiles. We propose a Risk-Stratified Cognitive-Safety Framework that integrates timing, formulation, route, baseline vascular risk, and apolipoprotein E (APOE) genotype to guide individualised counselling. Pre-registered, formulation- and route-specific randomised trials with cognitive primary endpoints, biomarker substudies, and at least 10-year follow-up are urgently needed. PROSPERO: pending registration update; protocol available from the corresponding author.