In this long-term follow-up of the WHI hormone therapy trials, neither CEE-alone nor CEE plus MPA significantly influenced liver cancer incidence or liver cancer specific mortality over 25 years. However, in the CEE + MPA trial, evidence of non-proportional hazards was observed, with liver cancer incidence higher in the postintervention period. In subgroup analyses, CEE plus MPA vs placebo was associated with more liver cancers among women aged 50-59 years and prior oral contraceptive users.
The higher incidence of liver cancer among women aged 50–59 years or prior oral contraceptive users that were randomized to CEE + MPA should be cautiously interpreted. Chance remains an important consideration, as liver cancer was not a protocol-specified outcome, the overall effect was not statistically significant, and these findings were based on limited events. Age and oral contraceptive use were not randomized, therefore should be viewed as identifying possible treatment-effect heterogeneity rather than causal mechanisms [15]. Recent large prospective and meta-analytic data do not support a strong association between ever oral contraceptive use and liver cancer risk [16]. Thus, these subgroup findings may reflect chance, susceptibility, or other subgroup-associated characteristics rather than causal effects. The concentration of cases among oral contraceptive users does not establish additive “hormonal burden.” Nonetheless, these randomized trial findings may be useful for future cohort studies and meta-analyses.
Liver cancer HRs were directionally different between HT trials, although the CEE-alone confidence interval was wide and compatible with both lower and higher risk. WHI trial findings have differed by formulation, as addition of MPA offset the unfavorable effects of CEE-alone on endometrial and ovarian cancer [17, 18], as well as offset favorable effects of CEE-alone on breast cancer incidence and breast cancer mortality [19], perhaps mediated by apoptotic effects [20]. Thus, formulation-specific liver cancer findings are biologically plausible, as estrogen and progestin signaling can have tissue- and pathway-specific effects, but should be interpreted cautiously given the limited number of events in this analyses.
In separate analyses of each randomized trial, and in a meta-analytic analysis combining the trial-specific HRs, we found that HT was not associated with liver cancer incidence or liver cancer-specific mortality. This neutral analysis finding differs from several observational reports suggesting lower liver cancer risk with HT. In a systematic review and meta-analysis [5], Zhong et al. reported lower primary liver cancer risk among HT users, with directionally similar associations for estrogen-alone and estrogen plus progestin. Similarly, a UK Clinical Practice Research Datalink nested case-control study reported lower liver cancer risk among menopausal hormone therapy users, particularly estrogen-only users [21] and a recent Swedish population-based cohort reported lower hepatocellular carcinoma risk for both estrogen-only and estrogen-plus-progestogen therapy [22]. These observational findings, generally suggesting approximately 30%–60% lower liver cancer risk among HT users, stand in contrast to the neutral overall WHI HR-estimate and to the CEE + MPA trial finding, which was directionally inconsistent with a large protective effect.
WHI randomized trial findings for HT and female cancers differ from cohort study findings in several areas. In cohort studies, both estrogen-alone and estrogen plus progestin therapies have generally been associated with higher ovarian cancer incidence, whereas only estrogen-alone was significantly associated in the WHI randomized trial [17]. Similarly, cohort studies have shown estrogen-alone to be associated with higher breast cancer incidence and mortality, while WHI randomized trial findings showed significantly lower breast cancer incidence, the protocol-specified primary outcome for safety, and breast cancer mortality with estrogen-alone [11]. This discordance was also seen for coronary heart disease, a protocol-specified primary efficacy outcome in the WHI HT trials; observational studies suggested cardiovascular benefit, whereas WHI trials found no coronary heart disease prevention benefit and early harm with CEE + MPA [23]. These differences may reflect important distinctions in study design, as cohort studies can be influenced by self-selection of healthier women initiating HT, and by depletion of susceptible women who develop early outcomes or discontinue because of intolerable side effects [17, 23]. In contrast, WHI participants were willing to be randomly assigned, in a double-blind manner, to HT or placebo, and outcomes were centrally adjudicated and analyzed regardless of treatment tolerance or adherence. U.S. population trends support the WHI randomized clinical trials evidence whether they agree or disagree with the observational study results. The sharp decline in hormone therapy use after 2002 was followed by lower incidence of breast and ovarian cancers, higher endometrial cancer incidence, and lower incidence of myocardial infarction, consistent with WHI randomized evidence [24,25,26,27].
Mechanisms linking estrogen exposure and liver cancer remain incompletely understood. Experimental models suggest protective effects of estrogen through anti-inflammatory pathways, including IL-6 suppression via activation of estrogen receptors on Kupffer cells [28], but observational data are inconsistent [5, 28]. A 2017 meta-analysis of observational studies (1,795 cases) reported protective associations for menopausal hormone therapy overall (RR, 0.60; 95% CI, 0.37–0.96), though these findings were not significant when stratified by estrogen-only or combined therapy [5]. These inconsistencies highlight the value of randomized trial data, as presented here.
Known viral and non-viral risk factors for liver cancer (chronic hepatitis, cirrhosis, smoking, alcohol, obesity, diabetes, thrombocytopenia) [29] were balanced across WHI randomization groups limiting residual confounding. Reproductive factors (prior menopausal hormone therapy and oral contraceptive use) were also balanced, except for modest differences in oophorectomy in the CEE-alone trial.
In prior WHI reports, CEE plus MPA increased risk of coronary disease, stroke, venous thromboembolism, and invasive breast cancer, with few sustained benefits for chronic disease prevention [30, 31]. While CEE plus MPA reduced colorectal cancer incidence early [32], the finding was not supported with longer follow-up [31]. CEE-alone has shown more favorable profiles particularly in younger women [33] including reduction in breast cancer incidence and breast cancer mortality [11]. The present results extend WHI randomized clinical trial findings regarding menopausal hormone therapy influence on liver cancer, a less common endpoint, and indicate no overall benefit of menopausal hormone therapy. Full presentation of the risks and benefits of menopausal hormone therapy have been recently presented elsewhere [33].
Strengths and limitations
This study has notable strengths including the randomized trial design among a large and well-characterized cohort, with long-term follow up extending nearly 25 years, centralized adjudication of cancer incidence and outcomes. The availability of two hormone therapy trials allowed separate evaluation of CEE-alone and CEE plus MPA. This study has limitations. Primary liver cancer was not a specified secondary study outcome, and the modest number of incident cases limited statistical power. Study findings only apply to the CEE-alone or CEE plus MPA regimens evaluated; however, long-term randomized clinical trial findings for liver cancer are not available for other hormone therapies.
