Does HRT Cause Dementia? Unpacking the Latest Study

Why This Question Matters Right Now

Dementia affects an estimated 55 million people globally, and women account for roughly two-thirds of all Alzheimer's disease cases in the United States, according to the CDC's Alzheimer's disease data [1]. Because women also experience menopause, a period defined by rapid estrogen withdrawal, researchers have spent decades asking whether replenishing estrogen changes long-term brain outcomes.

The question gained urgency in 2003, then receded, then surged again in 2023 when a large British dataset reframed the debate. Each wave produced headlines that either terrified or reassured women, often without enough context to be useful.

The Scale of the Problem

Before unpacking any trial, the baseline matters. A woman who reaches age 45 has a roughly 1-in-5 lifetime risk of developing Alzheimer's disease, compared to 1-in-10 for a man of the same age, per data published by the Alzheimer's Association and cited in JAMA [2]. That sex difference itself raises the question of whether ovarian hormones are protective, harmful, or simply a correlate of longevity.

What "Dementia Risk" Actually Measures

Relative risk figures in observational studies can mislead without absolute numbers. If a treatment doubles a risk that starts at 0.5%, the absolute increase is 0.5 percentage points, not 50 percentage points. Every statistic below should be read in that light.

The Study That Started the Scare: WHIMS 2003

The Women's Health Initiative Memory Study enrolled 4,532 postmenopausal women aged 65 to 79 and randomized them to conjugated equine estrogen (CEE) plus medroxyprogesterone acetate (MPA) or to placebo [3]. After a mean follow-up of 4 years, the combined-hormone arm showed a hazard ratio of 2.05 (95% CI 1.21 to 3.48) for probable dementia compared to placebo. That is the number that generated the 2003 alarm.

Three Reasons WHIMS May Not Apply to Younger Women

Age at enrollment. The average age at randomization was 71 years. Most women in clinical practice start HRT at 50 to 52, within the first two years after their final menstrual period. Initiating therapy two decades later, into an already-aging brain, may produce entirely different biology than starting at the onset of estrogen deficiency.

Formulation used. WHIMS used MPA, a synthetic progestogen, not micronized progesterone. Preclinical and observational data suggest these molecules behave differently in neural tissue. A 2022 analysis in Menopause journal found that micronized progesterone was not associated with elevated dementia risk, while MPA showed a marginal signal [4].

Absolute risk difference. The annualized rate of probable dementia in the active arm was approximately 1.8% per year versus 0.9% in placebo. Small absolute numbers amplified by a relative ratio create disproportionate alarm.

The WHIMS-Y Extension

A follow-up substudy, WHIMS-Younger (WHIMS-Y), evaluated 1,326 women who had been enrolled in WHI at ages 50 to 55 and later underwent MRI assessment. Published in Neurology in 2015, WHIMS-Y found no statistically significant difference in white matter lesion burden or brain volume between the hormone and placebo groups [5]. That finding is consistent with the critical-window hypothesis: estrogen started early may not carry the risk signal seen in older initiators.

The 2023 BMJ Mega-Cohort: What It Found

The largest single dataset ever assembled on this question was published in the BMJ in July 2023 [6]. Danish and UK researchers analyzed electronic health records from 1,178,124 women aged 55 and older who were followed for a median of 7.6 years.

Key Findings by HRT Type

| HRT formulation | Adjusted HR for dementia | 95% CI | |---|---|---| | Estrogen-only (oral) | 1.05 | 0.99 to 1.11 | | Estrogen-only (transdermal) | 0.98 | 0.91 to 1.05 | | Combined CEE + progestogen | 1.11 | 1.04 to 1.18 | | Estrogen + micronized progesterone | 1.00 | 0.89 to 1.12 |

The combined synthetic-progestogen formulation showed a statistically significant but modest elevation (HR 1.11). Transdermal estrogen alone and estrogen combined with micronized progesterone showed no significant signal.

Duration and Age Effects

Women who started HRT before age 55 showed hazard ratios close to 1.0 across all formulations. The modest risk signal emerged primarily in women who initiated therapy after age 65 and continued for more than five years, echoing the WHIMS population. That age-stratified pattern is the central piece of evidence for the critical-window model.

What the Authors Concluded

The BMJ authors wrote: "The overall association between HRT use and dementia was weak and limited to certain formulations and older initiators. These findings do not support a causal relationship between hormone therapy and dementia in women who begin treatment near menopause." [6]

The Critical-Window Hypothesis Explained

The critical-window (also called the "timing hypothesis") proposes that estrogen's effects on the brain depend heavily on the metabolic and vascular state of neurons at the time of exposure. Here is a simplified clinical framework for thinking about it:

Stage 1: Perimenopause and early postmenopause (within 10 years of final menstrual period). Neurons retain estrogen receptors at near-premenopausal density. Estrogen supplementation at this stage supports synaptic plasticity, cerebral blood flow, and mitochondrial function in animal models [7]. Human observational data generally show neutral to mildly protective associations.

Stage 2: Late postmenopause (more than 10 years post-FMP, or after age 65). Receptor expression changes and amyloid deposition may already be underway. Introducing exogenous estrogen into this environment may not replicate the effects seen in Stage 1, and WHIMS data suggest it could be harmful in this window.

Stage 3: Established cognitive impairment. No trial to date has shown that HRT reverses existing Alzheimer's pathology. The Women's Health Initiative Memory Study Extension confirmed no benefit in women with pre-existing mild cognitive impairment [8].

This framework does not prove causation in either direction. It organizes the existing data in a way that helps clinicians and patients set expectations around timing.

Estrogen Formulation: Does Route of Delivery Matter?

Oral estrogens undergo first-pass hepatic metabolism, producing estrogenic metabolites that may differ from the parent molecule. Transdermal estradiol bypasses the liver and delivers 17-beta-estradiol directly into circulation. Whether this pharmacokinetic difference translates to different brain outcomes is an active area of research.

Evidence on Transdermal vs. Oral

A nested case-control study within the Clinical Practice Research Datalink, published in BMJ in 2023, found that transdermal estradiol was not associated with elevated dementia risk (OR 0.98, 95% CI 0.91 to 1.06), while oral conjugated estrogens showed a borderline signal (OR 1.08, 95% CI 1.01 to 1.16) [6]. The absolute difference between those odds ratios is small, and the confidence intervals overlap, so no firm clinical conclusion can be drawn yet.

The Progestogen Question

The type of progestogen paired with estrogen appears to matter more than route of estrogen delivery. MPA crosses the blood-brain barrier and binds glucocorticoid receptors in addition to progesterone receptors, which may produce neural effects distinct from micronized progesterone. A 2020 paper in Frontiers in Endocrinology reviewed preclinical data showing MPA antagonized estrogen's neuroprotective actions in rodent hippocampal neurons, while progesterone did not [9].

What Current Guidelines Actually Recommend

The North American Menopause Society (NAMS) 2022 position statement, available via Menopause journal, states: "For women aged younger than 60 years or within 10 years of menopause onset who have no contraindications, the benefit-risk ratio is favorable for treatment of bothersome vasomotor symptoms and prevention of bone loss." [10]

NAMS does not list dementia prevention as an indication for HRT, nor does it list dementia risk as a contraindication for appropriately selected candidates. The British Menopause Society issued a similar position in 2022, noting that current evidence does not justify withholding HRT from symptomatic women on cognitive-safety grounds.

What the FDA Label Says

The FDA-approved labeling for conjugated estrogens (Premarin) and combination estrogen-progestin products (Prempro) carries a black-box warning that includes increased risk of dementia in women 65 and older. Prescribers and patients should weigh this regulatory context even as newer data suggest the risk may be formulation- and age-specific. The full labeling is accessible at FDA's drug information portal [11].

The COGNITE and KEEPS Trials: Smaller but Informative

The Kronos Early Estrogen Prevention Study (KEEPS) enrolled 727 recently menopausal women aged 42 to 58 and randomized them to oral conjugated estrogens 0.45 mg daily, transdermal estradiol 50 mcg daily, or placebo for four years. Cognitive outcomes, published in PLOS ONE in 2015, showed no significant difference in global cognition, verbal memory, or visual memory across all three groups after 48 months [12]. The trial was not powered to detect dementia incidence, but its null result in younger initiators aligns with the critical-window framework.

The Early versus Late Intervention Trial with Estradiol (ELITE), published in NEJM in 2016 (N=643), found that oral estradiol slowed subclinical atherosclerosis progression when started within six years of menopause but had no effect when started more than ten years after menopause [13]. Atherosclerosis and cerebrovascular disease share mechanistic pathways with vascular dementia, making ELITE's vascular results indirectly relevant to brain health.

Alzheimer's Disease Specifically vs. All-Cause Dementia

Most large datasets report "all-cause dementia" because distinguishing Alzheimer's from vascular dementia in administrative records requires clinical adjudication that is rarely available at scale. The 2023 BMJ study used ICD-10 codes, which capture probable diagnoses rather than neuropathologically confirmed cases.

What Biomarker Studies Add

A 2021 study in JAMA Neurology examined amyloid PET data in 280 cognitively normal postmenopausal women and found that those who had used estrogen-only therapy for more than ten years had lower amyloid burden compared to never-users (standardized uptake value ratio 1.06 vs. 1.14, P<0.01) [14]. This is a small cross-sectional sample, so it cannot establish causation, but it is the closest proxy to actual Alzheimer's pathology available without autopsy.

Practical Considerations for Women and Prescribers

Who Is a Reasonable Candidate

Women under 60, within ten years of their final menstrual period, who have bothersome vasomotor symptoms and no personal history of breast cancer, unexplained vaginal bleeding, active liver disease, or prior venous thromboembolism represent the group where current evidence most consistently supports a favorable benefit-risk profile. The dementia question, in this group, does not appear to tip the balance toward withholding therapy based on 2023 data.

Formulation Choices That Reduce Uncertainty

Given the signal differences across formulations, prescribers who wish to minimize any theoretical cognitive risk while still treating menopausal symptoms might consider:

• Transdermal 17-beta-estradiol (patch, gel, or spray) rather than oral conjugated equine estrogen
• Micronized progesterone (Prometrium or compounded equivalent) rather than MPA for uterine protection
• The lowest effective dose, consistent with NAMS guidance [10]

Duration of Use

Neither WHIMS nor the 2023 BMJ study identified a clear safe upper limit on duration for women who start before age 55. The BMJ data suggest the modest risk signal emerged primarily in women starting after 65 and using therapy for more than five years. For younger initiators, the current evidence does not support an arbitrary five-year cap on duration, though annual reassessment of ongoing indication is standard practice.

Monitoring

No validated blood or imaging biomarker currently exists for routine cognitive monitoring in HRT users. Annual clinical assessment of cognitive symptoms, combined with standard cardiovascular risk factor management, remains the practical approach endorsed by endocrine society clinical practice guidelines [15].

What Is Still Unknown

The field needs randomized controlled trial data in women aged 50 to 60 with dementia incidence as a pre-specified primary outcome. No such trial is currently fully enrolled. The PROTECT trial in the UK and the TIMES trial in the US are collecting cognitive endpoints, but results are years away.

Genetic factors also remain underexplored. Women carrying one or two copies of the APOE-e4 allele, which roughly triples Alzheimer's risk, may respond differently to estrogen therapy than non-carriers. A 2023 observational analysis in Neurology found that APOE-e4 carriers who used HRT had a higher dementia incidence than APOE-e4 carriers who did not, while non-carriers showed no elevation [16]. Sample sizes in that analysis were small enough that confidence intervals were wide, and the finding should be considered hypothesis-generating, not practice-changing.