Long-term systemic outcomes of postmenopausal hormone replacement therapy by age at menopause: a real-world cohort study
Ting-Fang Lu, Chien-Hsing Lu, Yu-Hsiang Shih, Yen-Fu Chen, Chun-Ting Fan, Shao-Jing Wang, Shih-Tien Hsu, Chih-Ku Liu, Lou Sun, Sheau-Feng Hwang, Tsung-Hsien LeeAbstract
STUDY QUESTION
Does postmenopausal hormone replacement therapy (HRT), initiated within one year after menopause diagnosis, confer differential long-term health effects according to age at menopause?
SUMMARY ANSWER
Long-term postmenopausal HRT was associated with age-dependent clinical outcomes, with broadly consistent cardiovascular and skeletal associations across age groups, but more variable metabolic and oncological associations according to age at menopause.
WHAT IS KNOWN ALREADY
The "timing hypothesis" suggests that HRT initiated closer to menopause may have more favourable cardiometabolic associations than later initiation, but evidence across distinct age-at-menopause strata remains limited.
STUDY DESIGN, SIZE, DURATION
This retrospective new-user cohort study, utilized de-identified data from the TriNetX Research Network, which provided access to electronic medical records for approximately 111 million patients globally. To establish a relatively healthy baseline cohort, we included women aged 30–90 years without prior major systemic diseases who had a documented diagnosis of menopause. To ensure adequate time for the development of hard clinical endpoints, we required that the index event (initial menopause diagnosis) occurred at least 10 years before data extraction (December 31, 2025), thereby guaranteeing a minimum follow-up period of one decade. To reduce immortal time bias and survivor bias inherent in prevalent-user designs, we employed a new-user, time-to-event cohort design. Women with a pre-existing history of type 2 diabetes mellitus (T2DM), cardiovascular disease (CVD), osteoporosis, compression fracture, venous thromboembolism (VTE), breast cancer or colorectal cancer on or before the index date were excluded.
PARTICIPANTS/MATERIALS, SETTING, METHODS
Women who initiated postmenopausal HRT containing estrogen and/or progestogen within one year after menopause diagnosis were propensity score-matched 1:1 with non-users using baseline demographic, clinical and laboratory covariates. Primary matching yielded 83,670 matched pairs in women who experienced menopause at age ≤60 years and 12,175 matched pairs in women who experienced menopause at age >60 years. Time-to-event analyses using Cox proportional hazards regression were conducted across four clinically defined menopausal age strata: premature ovarian insufficiency (POI, <40 years), early menopause (40–45 years), normal menopause (46–60 years), and late menopause (>60 years). Hazard ratios (HRs) were calculated for incident T2DM, CVD, compression fracture, breast cancer, colorectal cancer, and VTE. Interaction p-values were calculated using the normal menopause group (46–60 years) as the reference for interaction testing. Sensitivity analyses were conducted in both the overall ≤60-year and >60-year cohorts using an alternative 1:1 matching strategy based only on age, race and baseline body mass index (BMI), excluding laboratory parameters.
MAIN RESULTS AND THE ROLE OF CHANCE
In the overall matched cohort of women experiencing menopause at age ≤60 years, postmenopausal HRT was associated with lower hazards across all evaluated systemic endpoints. Compared with non-users, postmenopausal HRT was associated with lower hazards of T2DM (events: 3,517 vs. 4,580; crude incidence: 4.20% vs. 5.47%; HR: 0.595, 95% confidence interval (CI): 0.575–0.615) and CVD (events: 3,960 vs. 4,413; crude incidence: 4.74% vs. 5.29%; HR: 0.699, 95% CI: 0.688–0.709). Skeletal health was preserved, evidenced by a lower hazard of compression fractures (events: 426 vs. 441; crude incidence: 0.51% vs. 0.53%; HR: 0.780, 95% CI: 0.719–0.846). Furthermore, postmenopausal HRT was associated with lower hazards of breast cancer (events: 1,281 vs. 1,504; crude incidence: 1.53% vs. 1.80%; HR: 0.786, 95% CI: 0.743–0.831), colorectal cancer (events: 222 vs. 276; crude incidence: 0.27% vs. 0.33%; HR: 0.847, 95% CI: 0.738–0.971), and VTE (events: 1,247 vs. 1,346; crude incidence: 1.49% vs. 1.61%; HR: 0.859, 95% CI: 0.811–0.910). In the primary age-stratified model, the >60-year cohort showed a higher breast cancer hazard (HR: 1.151, 95% CI: 1.057–1.255); in the corresponding >60-year group sensitivity analysis, using alternative matching, the estimate remained directionally similar (HR: 1.218, 95% CI: 1.027–1.444). Across the remaining >60-year sensitivity outcomes, the direction of association was preserved, including continued lower hazards of T2DM, CVD, VTE and compression fracture, while the colorectal cancer association remained neutral.
LIMITATIONS, REASONS FOR CAUTION
Important limitations include residual confounding, incomplete socioeconomic and screening data, limited details on HRT formulation, route, dose and duration, unavailable breast cancer subtype information, unavailable exact maximum follow-up duration, and the inability to directly incorporate standardized patient-level geography into the matching or regression models. Region-restricted inference was additionally constrained because the Europe, Middle East and Africa (EMEA) >60-year HRT cohort contained only 48 exposed users, precluding stable matched analysis.
WIDER IMPLICATIONS OF THE FINDINGS
While cardiovascular and skeletal associations appeared broadly consistent across menopausal ages, the associations with breast cancer in the older age group highlight the critical need for individualized prescribing. These real-world findings are most consistent with earlier initiation of postmenopausal HRT for symptomatic women and those with POI, while urging extreme caution with initiation in late-onset menopause.
FUNDING
No specific external funding was received. The study was supported by departmental funds from the Department of Obstetrics and Gynecology, Taichung Veterans General Hospital.
DISCLOSURES
The authors declare no competing interests.
TRIAL REGISTRATION NUMBER
N/A.