The Women’s Health Initiative enrolled postmenopausal women aged 50 to 79 years at 40 United States clinical centers between 1993 and 1998 in two randomized, double-blind, placebo-controlled trials. Women with a uterus, numbering 16,608, were assigned to daily oral conjugated equine estrogen at 0.625 milligrams per day plus medroxyprogesterone acetate at 2.5 milligrams per day, or to matching placebo. Women who had undergone hysterectomy, numbering 10,739, were randomized to estrogen alone or placebo. Assignment used a centralized permuted-block algorithm stratified by clinical center, age group, and hysterectomy status, and blinding was maintained with identical-appearing tablets. The estrogen-plus-progestin trial was stopped early in 2002 after 5.6 years because overall health risks outweighed benefits, while the estrogen-alone trial stopped in 2004 after 7.2 years because of increased stroke risk without a clear coronary benefit.

Because liver cancer was not a protocol-specified outcome, the investigators treated the new analysis as a secondary examination, following participants through December 31, 2023. All incident cancers, including liver cancers, were confirmed by central medical record review conducted by adjudicators blinded to randomization assignment. Mortality information was near complete at 98 percent, thanks to serial queries of the National Death Index through 2023. Statistical analysis followed the intention-to-treat principle, using Cox proportional hazards models stratified by age group, dietary modification trial randomization, and study period, with two-sided stratified log-rank tests to compare groups.

A total of 74 incident primary liver cancers were identified across both trials, 26 in the estrogen-alone trial and 48 in the combined-therapy trial. In the estrogen-alone trial, with median cumulative follow-up of 16.3 years for incidence, the hormone arm showed no significant difference from placebo: 11 cancers versus 15, corresponding to an annualized rate of 0.012 percent versus 0.016 percent, and a hazard ratio of 0.75 with a 95 percent confidence interval of 0.34 to 1.63. The result was unchanged after adjustment for bilateral oophorectomy status, and there was no evidence that the proportional hazards assumption was violated.

The combined-therapy trial told a more textured story. Over 19.5 years of cumulative follow-up, women assigned to estrogen plus progestin experienced 30 liver cancers compared with 18 in the placebo group, an annualized rate of 0.019 percent versus 0.012 percent, yielding a hazard ratio of 1.63 with a confidence interval of 0.91 to 2.92, which just missed conventional significance. Crucially, the analysis detected evidence of non-proportional hazards, meaning the treatment effect changed over time. During the intervention phase itself, the hazard ratio was 0.66, but in the post-intervention period it rose to 2.11 with a confidence interval of 1.06 to 4.18, statistically elevated. In other words, the excess liver cancer risk emerged years after women had stopped taking the pills.

Subgroup analyses, restricted to the combined-therapy trial because of limited events elsewhere, sharpened the signal. Women aged 50 to 59 years at randomization who received the combined therapy developed 10 liver cancers, while none occurred in their placebo counterparts, a trend with a P value of 0.01. Prior users of oral contraceptives randomized to the combined regimen faced a hazard ratio of 4.30, with a confidence interval of 1.46 to 12.72 and a significant interaction test. Excess cases appeared among younger women both with and without prior oral contraceptive use, and the association was not explained by adjustment for contraceptive history. Current drinkers showed a somewhat stronger association, particularly those consuming one or more drinks per week, though the interaction test was not significant. No consistent effect modification was seen by smoking, body mass index, or prior menopausal hormone therapy.

The authors urge caution in interpreting these subgroup findings. Age and oral contraceptive use were not randomized variables, so they identify possible treatment-effect heterogeneity rather than causal mechanisms. Chance remains a serious consideration: liver cancer was not a protocol-specified endpoint, the overall effect was not statistically significant, and the findings rest on a modest number of events. Recent large prospective data and meta-analyses do not support a strong association between ever use of oral contraceptives and liver cancer risk, and the concentration of cases among contraceptive users does not establish an additive hormonal burden. Nonetheless, the randomized signals may be valuable for guiding future cohort studies and meta-analyses.

What makes the trial results so striking is how sharply they diverge from the observational literature. A 2016 systematic review and meta-analysis of 1,795 liver cancer cases reported roughly 40 percent lower risk among menopausal hormone therapy users, with a relative risk of 0.60. A United Kingdom nested case-control study using the Clinical Practice Research Datalink and a recent Swedish population-based cohort similarly reported 30 to 60 percent lower liver cancer risk among hormone therapy users. The neutral overall Women’s Health Initiative estimate, and the directionally unfavorable combined-therapy finding, stand in direct contrast to these reports.

This discordance echoes a familiar pattern in hormone therapy research. Observational cohort studies suggested cardiovascular benefit from menopausal hormone therapy, yet the randomized trials found no coronary heart disease prevention and early harm with the combined regimen. Cohort studies linked estrogen-alone therapy to higher breast cancer incidence, while the randomized trials showed significantly lower breast cancer incidence and mortality with estrogen alone. The discrepancies likely reflect self-selection of healthier women who initiate hormone therapy, and depletion of susceptible women who develop early outcomes or discontinue because of intolerable side effects. Randomized assignment, double blinding, and central adjudication of outcomes eliminate these biases. Notably, United States population trends after the sharp decline in hormone therapy use following 2002, including lower breast and ovarian cancer incidence and higher endometrial cancer incidence, have been consistent with the randomized evidence rather than the observational findings.

Mechanistically, the picture remains unsettled. Experimental models suggest that estrogen can protect the liver through anti-inflammatory pathways, including suppression of interleukin-6 via estrogen receptors on Kupffer cells, the liver’s resident macrophages, which could help explain why women develop liver cancer less often than men. Estrogen and progestin signaling can also exert tissue- and pathway-specific effects, and prior Women’s Health Initiative results have differed by formulation, with the addition of medroxyprogesterone acetage offsetting estrogen-alone effects on endometrial, ovarian, and breast cancers, possibly through apoptotic mechanisms. Formulation-specific liver cancer findings are therefore biologically plausible, even if the event numbers here are too small for firm inference. Known viral and non-viral risk factors for liver cancer, including chronic hepatitis, cirrhosis, smoking, alcohol, obesity, diabetes, and thrombocytopenia, were well balanced across randomization groups, limiting residual confounding. The bottom line for clinicians and patients is clear: the randomized evidence argues against menopausal hormone therapy as a strategy for liver cancer prevention, while highlighting younger women and prior oral contraceptive users as subgroups that warrant closer study in future research.

Subject of Research: Menopausal hormone therapy and primary liver cancer: long-term follow-up of the women’s health initiative randomized trials

Article Title: Menopausal hormone therapy and primary liver cancer: long-term follow-up of the women’s health initiative randomized trials

Article References: Pichardo, M. S., Aragaki, A. K., Manson, J. E., Pan, K., Jung, S. Y., Rohan, T. E., Mszar, R., & Chlebowski, R. T. (2026). Menopausal hormone therapy and primary liver cancer: long-term follow-up of the women’s health initiative randomized trials. British Journal of Cancer. https://doi.org/10.1038/s41416-026-03621-9

Image Credits: AI Generated

DOI: 10.1038/s41416-026-03621-9

Keywords: liver cancer, menopausal hormone therapy, Women's Health Initiative, estrogen, hepatocellular carcinoma, randomized trials, Menopausal, hormone, therapy, primary, liver, cancer