This analysis reports long-term incidence and mortality from primary liver cancer in participants of the Women’s Health Initiative (WHI) randomized hormone therapy trials. Two randomized, double-blind, placebo-controlled trials evaluated conjugated equine estrogen (CEE) alone (in women with prior hysterectomy) and CEE plus medroxyprogesterone acetate (MPA) (in women with a uterus). After nearly 24 years of follow-up and 74 incident liver cancers, neither CEE-alone nor CEE plus MPA significantly altered liver cancer incidence or, for CEE plus MPA, mortality in the overall trial cohorts. Select subgroup analyses suggested possible increased risk with CEE plus MPA among younger participants and prior oral contraceptive users, but event counts were small and results are hypothesis-generating.
Sex differences in liver cancer incidence and outcomes have prompted investigation into hormonal influences. Observational and mechanistic studies have implicated estrogen signaling pathways in liver disease progression and hepatocarcinogenesis, but epidemiologic evidence on menopausal hormone therapy and primary liver cancer risk has been inconsistent. Prior pooled observational data and meta-analyses reported mixed associations—some suggesting increased risk with surgical menopause and variable associations with exogenous hormone use—leaving uncertainty about causal effects. No randomized trial data had been available to address this question before the present WHI randomized-trial follow-up analysis.
The WHI included two parallel randomized, double-blind, placebo-controlled hormone therapy trials conducted at 40 U.S. centers between 1993 and 1998. Eligibility criteria included postmenopausal women aged 50–79 years without recent invasive cancer and with anticipated survival ≥3 years. Women with a uterus (n = 16,608) were randomized 1:1 to daily oral CEE 0.625 mg combined with MPA 2.5 mg or matching placebo. Women with prior hysterectomy (n = 10,739) were randomized 1:1 to CEE 0.625 mg alone or matching placebo. Randomization used a centralized permuted-block algorithm stratified by center, age group, and hysterectomy status. Blinding was maintained with matching tablets; adherence was assessed by pill counts and a placebo run-in period preceded randomization.
Trials were monitored for pre-specified primary outcomes (including coronary heart disease benefit and breast cancer harm) and several secondary outcomes. The CEE plus MPA trial was stopped early in 2002 (median ~5.6 years) and the CEE-alone trial in 2004 (median ~7.2 years) due to findings unrelated to liver cancer. For the present analysis, follow-up continued through December 31, 2023. Incident cancers, including primary liver cancers, and cause-specific deaths were centrally adjudicated by reviewers blinded to treatment assignment; mortality ascertainment was supplemented by National Death Index linkage with near-complete coverage through 2023.
This liver cancer analysis was not protocol-specified. Analyses followed the intention-to-treat principle. Time-to-event analyses used stratified Cox proportional hazards models with baseline hazard functions stratified by age group, dietary modification trial randomization, and study period. Two-sided stratified log-rank tests were used for group comparisons and Kaplan–Meier methods described temporal patterns. Given the small number of events, subgroup analyses were limited principally to the CEE plus MPA trial and accompanied by sensitivity checks. An overall trial-combined HR was computed using an inverse-variance weighted average of trial-specific log-HRs. Statistical analyses used SAS 9.4 and R 4.4; analytic code is available from the corresponding author on request.
Across the two randomized WHI hormone therapy trials and nearly 24 years of follow-up, 74 incident primary liver cancers were documented. In the CEE-alone trial, there were 11 incident liver cancers in the active arm and 15 in the placebo arm (hazard ratio [HR] 0.75; 95% confidence interval [CI], 0.34–1.63), indicating no statistically significant effect. In the CEE plus MPA trial, there were 30 incident liver cancers in the active arm and 18 in the placebo arm (HR 1.63; 95% CI, 0.91–2.92), a non-significant elevation in the point estimate.
Liver cancer mortality in the CEE plus MPA comparison included 27 deaths in the active arm and 22 in the placebo arm (HR 1.18; 95% CI, 0.67–2.07), likewise not statistically significant. The abstract reported subgroup findings within the CEE plus MPA trial: more liver cancers were observed among women aged 50–59 randomized to active therapy versus placebo (10 vs 0 cancers; P-trend 0.01), and prior oral contraceptive users showed a higher HR (4.30; 95% CI, 1.46–12.72) with evidence of interaction (P-interaction 0.01). These subgroup results are based on small numbers and require cautious interpretation.
Baseline characteristics by randomization arm were generally balanced. The source text notes some cohort differences across trials (for example, a lower prevalence of bilateral oophorectomy in the CEE-alone active arm than placebo), but further baseline detail and the remainder of the original Results section text were not reported in the available source.
In this long-term randomized-trial follow-up of the WHI hormone therapy interventions, there was no clear overall effect of menopausal hormone therapy—neither CEE-alone nor CEE plus MPA—on incidence of or mortality from primary liver cancer in the full trial populations. Observed subgroup signals for increased liver cancer with CEE plus MPA among younger women at randomization and prior oral contraceptive users are hypothesis-generating but limited by small event counts and were not pre-specified; they warrant further study in other datasets but do not establish causation.
ClinicalTrials.gov identifier: NCT00000611.
Note: Some trial details and additional analyses referenced in the WHI article were not included in the provided source text and therefore are not summarized here.