Potential impact of hormone replacement therapy on the risk of hepatocellular carcinoma in women of the PLCO cohort

Introduction

Over 900,000 people in the world developed hepatocellular carcinoma (HCC) in 2020 with the worldwide incidence continuing to rise over the past 50 years, due to the lack of effective prevention approaches [1]. The prognosis for people once diagnosed with HCC is poor with 5-year survival rates of less than 7 % worldwide resulting from the lack of effective treatments [2]. Consequently, liver cancer prevention is a global challenge where any significant inroad would represent a considerable health and cost benefit to a substantial population worldwide.

A major feature of HCC is sexual dimorphism, i.e., males are more susceptible to HCC than females, a common feature in both rodents and humans. It has been reported that oestrogen can prevent hepatic tumorigenesis in rodents and hence, is postulated to play a role in this sexual dimorphism [3], [4], [5], [6], [7]. For example, male rats and mice became resistant to hepatic tumorigenesis and liver injury when supplemented with oestradiol whereas female rats and mice became susceptible to HCC once their ovaries were removed; this was reversed by supplements of oestradiol to reach the biological oestrogen levels in the blood of females [3], [4], [5], [6]. Preclinical studies in male rodents showed more beneficial effects in terms of HCC prevention from oestradiol supplements than in females [3], [4], [5], [6], [7]. Our previous studies showed that oestrogen signalling prevented HCC through multiple pathways of hepatic carcinogenesis, such as carcinogen detoxification, cell proliferation, etc. [7]. Thus, oestradiol supplements may represent an effective approach to HCC prevention in both men and women.

The efficacy of oestradiol supplementation on the reduction of HCC incidence in humans has been reported in many retrospective studies worldwide. As HCC is a rare cancer in women, many studies reported to date have been retrospective studies, as it is difficult to gather enough cases for statistical power using a prospective study design. For example, a nested case-control study within the United Kingdom’s Clinical Practice Research Datalink (CPRD) showed that hormone replacement therapy (HRT) was significantly associated with 42 % less risk of developing HCC (odds ratio [OR]=0.58) in 339 post-menopausal women compared to 1,318 matched controls [8]. A case-control study of 234 HCC cases and 282 unmatched healthy controls recruited at a single cancer centre in the USA showed that ever using exogenous oestrogen (either for contraception or for HRT) was associated with significantly lower risks of HCC (OR=0.53, 95 % confidence interval [CI]: 0.32–0.88, p=0.01) [9]. A retrospective population-based cohort study using data from Taiwan’s National Health Insurance Research Database showed that hepatitis C patients who were treated with oestradiol (n=1,022) had reduced risks of developing HCC (hazard ratio=0.43, 95 % CI: 0.30–0.61, p<0.001) compared to 1,022 matched non-treated controls [10]. The evidence from these retrospective studies is converging on a clear message that HRT is associated with reduced risk of HCC in women.

The downside of retrospective studies is that they have many potential sources of bias and confounding (especially where no attempt is made to match cases and controls). Prospective studies generally have fewer sources of bias and confounding, although, for rare cancers like HCC in women, very large studies are necessary in order to provide sufficient power for statistical analyses. However, studies using prospective cohorts with both HCC and non-HCC cases of liver cancer have shown varied results (sexual dimorphism mainly occurs in HCC but not all liver cancer types). Several meta-analyses of pooled prospective datasets have been conducted to investigate the association of HRT with liver cancer risk, including using the PLCO data, however, only a small set of cases from the PLCO cohort were included in these analyses. For example, only 80 cases (60 controls and 20 HCCs) were used to evaluate the HCC risks and the risks of intrahepatic cholangiocarcinoma in the pooled liver cancer project [11], 12]. Another study from the liver cancer pooling project including 22 HCC cases from the PLCO data found a modestly increased risk of HCC with HRT but the significance disappeared when the analysis of HRT was adjusted for bilateral oophorectomy [13].

The goal of this study aims to complement previous publications on the topic of the association between HRT and HCC incidence with the entire PLCO cohort (53,681 women with data on HRT usage, including 16 HCC cases), the results of which have clinical implications for both guiding future prevention and therapy of HCC, and in determining the risk profile for HRT use in women.

Materials and methods

Results

Of the total PLCO sample (n=78,209) a total of 53,681 women had complete records regarding HRT usage and HCC status (Table 1). Of the 53,681, a total of 16 HCC events (0.03 %) occurred and 35,255 (65.7 %) women reported use of HRT; all were over 50 years old and about 90 % of them self-reported their ancestry as ‘white’; 55.7 % (29,918) had taken oral contraceptives, and 4.6 % (2,459) and 3.2 % (1,693) had comorbidities of diabetes and liver diseases (hepatitis or cirrhosis), respectively. With a median follow-up of 11.3 years, the yearly incidence rate of HCC in this cohort was estimated at 2.64 per 100,000 persons.

In women who reported HRT use (n=35,255), only 9 of them developed HCC compared to 7 HCC events in women not having HRT (n=18,426) (Table 1). Multivariate regression analysis of the whole cohort (Table 2) revealed that HRT reduced the risk of HCC, but that this reduction was not statistically significant (OR=0.61, p=0.367). The only variable significantly associated with an increased risk of HCC (OR=5.09, p=0.034) was liver comorbidity (hepatitis or cirrhosis).

To investigate the combinational effects of HRT with diabetes, liver comorbidity, and oral contraceptive use on HCC risks, we conducted a two-way ANOVA between HRT use and these variables (Table 3). For each of the three variables, the evidence for a synergistic effect (interaction) with HRT was evaluated and the p-value was reported. The results showed that oral contraceptive use was significantly associated with a reduced risk of HCC in the HRT group and a non-significant increased risk of HCC in the no HRT group. There was a nearly-significant interaction between HRT use and oral contraceptive use (p=0.054). In addition, liver comorbidity was significantly associated with an increased risk of HCC in the no-HRT group, but not in the HRT group (where no participants had liver comorbidity and HCC). This interaction between liver comorbidity and HRT was significant (p=0.0001).

Next, we conducted a univariate regression analysis of HRT use broken down by type of hormone used (Table 4). HRT use of any hormone was non-significantly associated with a reduced risk of HCC (OR=0.67, p=0.430) in the univariate analysis and none of the individual hormones were significantly associated with risk of HCC.

Given that the number of HCC cases is small, we expanded our analysis to include all hormone use cases (HRT, oral contraceptive use, or both; n=64,592) and all liver cancer cases (n=31). Multivariate regression analysis revealed that HRT reduced the risk of liver cancer, but that this reduction was not statistically significant (OR=0.67, p=0.45; Table 5). Oral contraceptive use also non-significantly reduced the risk of liver cancer (OR=0.48, p=0.19). The variable ‘Any hormone therapy’ compared those taking any hormone therapy (HRT, oral contraceptives use, or both) with those who took neither, which also non-significantly reduced the risk of liver cancer. The only variable significantly associated with the risk of liver cancer was liver comorbidity, which was associated with an increased risk of liver cancer (OR=4.64, p=0.04, Table 5).

When we examined different combinations of hormone therapies in the sample with complete data for both HRT and OC use (n=52,725; n=17 liver cancer cases, Table 6), there was no significant difference between groups.

Discussion

Different from those retrospective studies solely comparing HCC patients with and without HRT [8], 9], [15], [16], [17], this population study of the PLCO cohort provides a detailed epidemiological overview of HCC incidence and HRT usage in a largely female population in the United States with a median 11.3-year follow-up. The estimated yearly incidence rate of HCC in this cohort (2.64 per 100,000 persons) from 2006 to 2019 is broadly in line with a study reporting increased HCC incidence rates in women from 2.38 per 100,000 population in 2001 to 3.09 in 2020 in the USA [18]. Despite this being a large study, as HCC is a rare cancer in women and also relatively rare in the USA compared to some other countries, event numbers were low. Due to these low numbers, we had low power to detect a significant association between HCC and HRT use. However, although we did not observe associations that were statistically significant, the trends and effect sizes of reduced HCC risk with HRT use were in line with reports from retrospective studies published to date [8], [9], [10]. Although this cohort represents the HCC incidence and HRT usage in the general population in the United States in many ways, it should be noted that 90 % of the women in the study self-reported their ancestry as ‘white’, which is higher than the overall population in the USA and may reflect study site, recruitment process, etc. Nevertheless, our data might provide insights for clinicians when developing a personalised treatment plan for high-risk women (i.e., liver comorbidities), to consider oestradiol supplementation. Despite the reduced power due to small case numbers, this study provides valuable data from a cohort representing the HCC incidence and HRT usage in the general population in the United States, also indicating the importance of considering more specific high-risk populations for future study.

Previous reports on the associations between oral contraceptives and HCC risks have shown mixed results [19], [20], [21], [22], [23], although many studies have reported a trend toward reduced risk of HCC with the use of oral contraceptives. In our cohort, we also saw a non-significant association between oral contraceptive use and reduced risk of HCC (Table 2). When we examined both oral contraceptive use and HRT together, the protective effect on HCC incidence was more pronounced in women who used both oral contraceptives and HRT compared to the use of either of them alone (Table 3). Thus, we speculate that women taking oral contraceptives during child-bearing age and also having HRT during and after menopause, i.e., prolonged exposure to oestradiol supplementation, may achieve the best possible prevention of HCC.

It was observed that women with pre-existing liver diseases, such as hepatitis or cirrhosis, had a higher risk of developing HCC (OR=5.09, Table 2). Interestingly, those women with liver diseases (hepatitis or cirrhosis) had the greatest preventive effects on HCC incidence by HRT as evidenced by zero HCC events in these women with a significant interaction observed between HRT use and liver comorbidities (p<0.001, Table 3). HCC frequently is mainly developed in the cirrhotic liver which in turn stems from hepatitis, including viral, alcoholic, and other causes. It usually takes 20–30 years to develop HCC from hepatitis or cirrhosis, which also shows sexual dimorphism with male dominance. It has been reported that oestrogens also prevent the development of hepatitis or cirrhosis [3], [4], [5], [6], [7]. Thus, we speculate that hormone therapy such as HRT may achieve the best prevention of HCC, particularly when used over a long time period (as suggested by the fact that women on oral contraceptives and HRT appeared to have increased protection from HCC), possibly by intervening at early stages in HCC pathogenesis, namely hepatitis or cirrhosis. It would be of further interest to ascertain whether hormone use might prevent the progression of hepatitis to cirrhosis. This in itself could reduce or prevent the substantial burden of morbidity and mortality caused by cirrhosis even in the absence of HCC.

Given that the number of HCC cases is so small, we conducted similar analyses by including all hormone use cases (n=64,592) and all liver cancer cases (n=31), but the reduction in the risk of liver cancer by HRT was still not significant (Table 5). The only significance was observed in the association of liver comorbidity with an increased risk of liver cancer (p=0.04, Table 5), which is similar to our observations on HCC cases (p=0.034, Table 2). This further suggests that a large sample size is essential for such studies in the future.

Therefore, the key limitation of this study is the small sample size of HCC (n=16) and liver cancer patients (n=31), which resulted in low power to investigate the impact of HRT on the prevention of HCC or liver cancer. Nevertheless, this is one of the largest datasets on such topics for the American/Caucasian population. Different from the American population in the PLCO data, two very recent studies of Asian populations showed promising but varied results [24], 25]. One study on the Chinese population with 258 liver cancer cases in a total of 72,807 women showed that HRT was positively associated with reduced risks of liver cancer but not statistically significant [25]. Another study on the Korean population with 138 liver cancer cases in 29,328 women showed that the risk of liver cancer was significantly reduced with the duration of HRT but not the total HRT usage [24]. A very recent study showed that in the Korean population, HRT significantly reduced the risk of liver cancer [24]. Therefore, the selection of populations and sample sizes are the key for such studies. The lower incidence of liver cancer in the PLCO cohort could be due to that 90 % of participants were white women who generally have the lowest incidence of liver cancer compared to other races [26].

From future perspectives, we expect that men will benefit more from oestradiol supplements in the prevention of HCC than women due to the naturally lower baseline oestrogen levels in males and consequently that supplements of oestrogens would prevent HCC. Given that men have the most HCC, supplements of oestradiol could be used as an effective approach for HCC prevention in both men and women who have low levels of oestrogens.

Conclusions

In this PLCO cohort, non-significant reductions in the risk of HCC were observed for users of HRT and oral contraceptives and these effects could be mediated by preventing liver hepatitis or cirrhosis. The low number of HCC diagnoses (16 events in total in this cohort of 53,681 women), which reflects the wider healthy female population, limited the power to investigate the impact of HRT on HCC incidence. Similar results of reduction trends with no significance were still found when we expanded the data by including all 31 cases of liver cancer intersecting with all hormone use cases. Further large prospective studies of higher-risk populations are required to confirm the relationship between exogenous oestrogens and HCC prevention.