Denmark followed 877,000 women for 14 years. Hormone therapy did not raise mortality.

For twenty-two years, the central fact of menopausal hormone therapy in clinical practice was a warning label. The boxed warning the US Food and Drug Administration issued in 2003 was based on secondary outcomes from the Women's Health Initiative trial of oral conjugated equine oestrogens with medroxyprogesterone acetate, and it was generalised across all doses, formulations, and routes — including local vaginal therapies — under a mandate to prescribe the lowest effective dose for the shortest duration [s2]. The Department of Health and Human Services and the FDA announced removal of that warning on November 10, 2025 [s2].

Into that changed landscape, The BMJ published on February 18 a Danish nationwide register study asking the plainest possible question about the therapy: does it increase the risk of dying [s1].

The study

The cohort was Danish women born between 1950 and 1977 and alive at age 45 [s1]. Follow-up began on each woman's 45th birthday and ended on 31 July 2023.

Of 969,424 eligible women, 92,619 were excluded for thrombophilia, liver disease, arterial or venous thrombosis, breast cancer, endometrial cancer, ovarian cancer, previous use of menopausal hormone therapy, or previous bilateral oophorectomy — leaving 876,805 women [s1]. Of those, 104,086 (11.9%) redeemed a prescription for systemic menopausal hormone therapy, and 47,594 (5.4%) died, over a median follow-up of 14.3 years (IQR 7.9–21.0) [s1].

The primary outcome was death as registered in the Central Persons Register; secondary outcomes were cardiovascular, cancer, and other cause-specific mortality [s1].

The two numbers that appear to contradict each other

The crude figures: women who used hormone therapy had 54.9 deaths per 10,000 person-years, against 35.5 per 10,000 person-years in the unexposed group [s1].

The adjusted figure: hazard ratio 0.96 (95% CI, 0.93 to 0.98) [s1].

Those are not in conflict; they are the same data before and after accounting for who takes the drug. The adjustment set included age, calendar year, parity, educational degree, income group quarter, country of birth, diabetes, hypercholesterolaemia, hypertension, atrial fibrillation, valvular disease, heart failure, and having three or more hospital contacts between ages 44 and 45 [s1]. Age is doing most of the work: women starting hormone therapy are, by construction, further into midlife than the full comparison population and accrue person-time at older ages.

Which number a reader should trust depends on whether they believe the adjustment is adequate. The crude rate is not evidence of harm; it is evidence that the exposed and unexposed groups are not comparable without adjustment. Whether they are comparable with it is the study's central assumption and cannot be verified from within an observational design.

Duration

Stratified by cumulative duration of use, the adjusted hazard ratios were [s1]:

• Less than 1 year: 1.01 (95% CI, 0.98 to 1.05)
• 1 to 2.9 years: 0.94 (0.89 to 0.98)
• 3 to 4.9 years: 0.90 (0.84 to 0.95)
• 5 to 9.9 years: 0.89 (0.84 to 0.95)
• 10 years or more: 0.98 (0.90 to 1.07)

The middle of that range sits below 1 with confidence intervals excluding it; the two ends do not. A reader inclined to see a duration-response relationship will see one from one to ten years. A reader inclined to caution will note that the longest-duration stratum returns to the null and that duration of use is itself a marker of tolerating the drug well — women who do badly stop.

No unequivocal differences in cause-specific mortality were found between groups [s1].

The oophorectomy subgroup

Among 703 women who underwent bilateral oophorectomy between ages 45 and 54, those who used hormone therapy experienced a 27–34% lower mortality hazard than those who did not, with a median age at death of 60.9 years (IQR 55.3–66.6) versus 56.6 years (52.9–62.0) [s1].

This is the subgroup where surgical loss of ovarian function is abrupt and complete, and where replacement has the strongest physiological rationale. It is also a subgroup of 703 women — small enough that it should be read as consistent with existing understanding rather than as independent evidence for it.

What the study does and does not establish

Its conclusion is deliberately negative in form: this nationwide cohort study did not find menopausal hormone therapy was associated with increased mortality [s1]. That is a different statement from finding that it reduces mortality, and the authors do not make the second claim.

The design cannot resolve healthy-user bias — the tendency of people who take preventive medications to be healthier, wealthier, and more engaged with care in ways registers capture only partly, though the adjustment for education, income, and prior hospital contacts is an attempt at exactly that. Exposure is defined by redeemed prescriptions, which is closer to actual use than a written prescription but is not adherence. And the cohort is Danish women born from 1950 onward, using the formulations and doses prescribed in Denmark over that period.

What to watch

Whether comparable register analyses in other national systems reproduce the null, and whether prescribing patterns shift in the wake of the label change now that the warning that shaped two decades of practice has been removed [s2].

Nothing here is a recommendation to start, continue, or stop hormone therapy. Decisions about menopausal hormone therapy involve individual risk factors and belong with a clinician.

Sources

• [s1] Mikkelsen AP, Bergholt T, Lidegaard Ø, Scheller NM. Menopausal hormone therapy and long term mortality: nationwide, register based cohort study. The BMJ, published online 2026-02-18.
• [s2] Sriprasert I, Hodis HN, Mack WJ, et al. Elimination of the Black Box Warning on Menopausal Hormone Therapy. Obstetrics & Gynecology, published online 2026-02-13.