Estrogen Replacement Reduces Risk and Increases Survival Times of Women With Hepatocellular Carcinoma

Abstract

Background & Aims

Environmental factors have been identified that affect risk of hepatocellular carcinoma (HCC), but little is known about the effects of sex hormones on liver cancer development or outcome. We investigated whether menopause hormone therapy (MHT) affects risk, age at onset, or outcome of HCC.

Methods

We performed a case–control study of 234 female patients treated for HCC at a tertiary medical center and with 282 healthy women (controls), from January 1, 2004 through May 31, 2015. We collected detailed information on environmental exposures, ages of menarche and menopause, hysterectomies, and uses of birth control and MHT. We performed multivariable logistic and Cox regression analyses to determine the independent effects of factors associated with women on risk and clinical outcome in HCC. The primary outcomes were effect of MHT on HCC risk, the relationship between MHT with hepatitis virus infection on HCC development, and effect of MHT on age at HCC onset or survival after diagnosis of HCC.

Results

The estimated adjusted odds ratio (AOR) for HCC in women who ever used estrogen was 0.53 (95% CI, 0.32–0.88). This association was supported by the older age of HCC onset among estrogen users (mean, 64.5±0.9 years) vs non-users (mean, 59.2±1.1 years) (P=.001) and the reduced risk of HCC among long-term users (more than 5 years) (AOR, 0.36; 95% CI, 0.20– 0.63). Users of estrogen also had a reduced risk for hepatitis-associated HCC: AOR for users, 4.37 (95% CI, 1.67–11.44) vs AOR for non-users, 17.60 (95% CI, 3.88–79.83). Estrogen use reduced risk of death from HCC (hazard ratio, 0.55; 95% CI, 0.40–0.77) (P=.01). Median overall survival times were 33.5 months for estrogen users (95% CI, 25.7–41.3 months) and 24.1 months for non-users (95% CI, 19.02–29.30 months) (P=.008).

Conclusion

In a case–control study of women with HCC vs female controls at a single center, we associated use of estrogen MHT with reduced risk of HCC and increased overall survival times of patients with HCC. Further studies are needed to determine the benefits of estrogen therapy for women and patients with HCC, and effects of tumor expression of estrogen receptor.

INTRODUCTION

Irrespective of the worldwide variation in the incidence of hepatocellular carcinoma (HCC) ¹, HCC is a male-dominant disease ². In the United States, gender disparity in HCC has been observed not only in disease incidence, ^{3, 4} but also in etiological factors ^{5, 6}, progression to cirrhosis ⁷, and survival ⁸; where male-to-female ratio is 3:1 with poorer prognosis and is commonly associated with chronic viral hepatitis, cigarette smoking, and alcohol consumption. The sex difference has also been observed in transgenic mice with hepatitis B– or C–induced HCC.^{9, 10}

In view of the notable male predominance of HCC, several investigators raised the question about the importance of sex hormones in HCC risk and prognosis. The liver expresses estrogen and androgen receptors, both of which may act as transcription factors and may regulate expression of several regulatory genes involved in several pathways including those associated with cell proliferation and immune response. ^{11, 12}

According to the National Health Statistics Report, in the United States, the percentage of women using contraception increases with age, with 75% of women aged 40–44 years now classified as users ¹³. The association between contraception and HCC was not shown to be conclusive by a meta-analysis of 12 case-control studies. The null association was later confirmed by a U.S. liver cancer pooling project with an OR of 1.12 (95% CI, .82–1.55) ¹⁴. Despite the available literature about the association between contraception and HCC, very little has been published about the association between MHT and risk of HCC.^{14, 15}

This case-control study aimed at integrating clinical and epidemiological data to assess 1) the effect of MHT on HCC risk in females, 2) the relationship between MHT with hepatitis virus infection on HCC development, and 3) the effect of MHT on age at HCC onset or HCC survival.

METHODS

The current investigation is part of an ongoing hospital-based case-control study, which was approved by the Institutional Review Board at The University Texas MD Anderson Cancer Center. Written informed consent for participation was obtained from each participant.

Cases were new patients with pathological or radiological evidence of HCC who were treated at MD Anderson. The control subjects were healthy and genetically unrelated family members (i.e., spouses) of patients at M. D. Anderson who had cancers other than liver, gastrointestinal, lung, or head and neck cancer.

Between January 1, 2004, and May 31, 2015, 234 female patients (cases) with HCC and 282 female controls were eligible for the current investigation. HCC patients and controls were US residents and were interviewed simultaneously in person for demographic features and HCC risk factors with the use of a structured and validated questionnaire.

All participants were asked about their age of menarche, age of menopause, history of hysterectomy, their age when they underwent a hysterectomy, and whether one or both ovaries were removed during their hysterectomy.

Each female was interviewed for ever-use of various birth control types including pills, implant, or injection and the duration of use of various forms of contraception. Participants were also questioned about use of exogenous hormones including estrogen, progesterone, and combined estrogenprogesterone. Methods of use (oral pills, skin patch, injection, and vaginal) and duration of each method were documented. We missed to collect parity information from cases and controls. However, for case patients, we extracted the history of pregnancy, number of pregnancies, and number of children from the institutional epidemiological database of cancer patients. In addition, baseline clinical variables were retrieved from patients’ medical records.

Statistical Methods

Stata software (Stata Corp, College Station, TX) was used for statistical analysis. We performed multivariate unconditional logistic regression analyses. We calculated the adjusted odds ratio (AOR) and 95% confidence interval (CI) values, using maximum likelihood estimation after controlling for confounding effect of demographic and HCC risk factors.

Overall survival (OS) was defined as the time between HCC diagnosis and death or end of follow-up. Median survival was estimated by using the Kaplan-Meier product-limit method, and significant differences between the survival times were determined by using the log-rank test ¹⁶. To identify independent prognostic factors for OS, hazard ratios (HR) and 95% CIs were calculated by using Cox proportional hazard models with a backward stepwise selection.

Analysis of covariance was used to analyze patients’ mean age at HCC onset by hormonal exposure. Linear regression models were used to estimate the mean differences in age at HCC onset associated with use of birth control and exogenous hormones after adjusting for other factors associated with age at onset in this study population.

RESULTS

Table 1 shows that cigarette smoking was not associated with HCC risk in women. Consistent with our previous reports race, HCV, HBV, alcohol use, diabetes, hypothyroidism, early adulthood obesity, and positive family history of cancer were significant risk factors for HCC in US females.^{6, 17, 18}

Table 1.

Multivariate-adjusted Odds Ratio (AOR)^* and 95% Confidence Interval (CI) for HCC risk factors

Demographic variables	HCC patients		Controls		AOR (95% CI)	P value
	N =234	(%)	N =282	(%)
Age (years)
<50	41	(17.5)	42	(14.9)	1 (Reference)
≥50	193	(82.5)	240	(85.1)	0.76 (0.42–1.36)	.4
Race
White	171	(73.1)	261	(92.6)	1 (Reference)
Non-White	63	(26.9)	21	(7.4)	0.27 (0.14 – .55)	<.0001
Educational level
< College Education	174	(74.4)	182	(64.5)	1 (Reference)
≥ College Education	60	(25.6)	100	(35.5)	1.38 (0.86– 2.21)	.2
HCV (Anti-HCV+)†
No infection	174	(74.4)	256	(90.8)	1 (Reference)
HCV infection	60	(25.6)	1	(0.4)	71.6 (9.6 – 536.04)	<.0001
HBV
No infection	224	(95.7)	249	(88.3)	1 (Reference)
HbsAg	10	(4.3)	1	(0.4)	13.95 (1.28–151.58)	.03
Anti-HBc	26		7	(2.5)	2.98 (1.1–8.07)	.03
Unavailable	0		25	(8.9)	---
Cigarette smoking‡
No smoking	128	(54.7)	191	(67.7)	1 (References)
Smokers	106	(45.3)	91	(32.3)	1.43 (0.89–2.31)	.1
Alcohol drinking¥
No drinking	109	(46.6)	198	(70.2)	1 (Reference)
Drinkers	125	(53.4)	84	(29.8)	2.9 (1.81–4.64)	<.0001
Prior history of diabetes
No diabetes	179	(76.5)	256	(90.8)	1 (Reference)
Diabetes	55	(23.5)	26	(9.2)	3.84 (1.96–7.5)	<.0001
Prior BMI (age20–40)
Normal/Slim	153	(65.9)	225	(80.1)	1 (Reference)
Overweight	47	(20.3)	39	(13.9)	1.24 (0.67–2.3)	.5
Obese	32	(13.8)	17	(6)	2.35 (1.0–5.18)	.03
Hypothyroidism
No	176	(75.2)	239	(84.8)	1 (Reference)
Yes	58	(24.8)	43	(15.2)	2.43 (1.43– 4.16)	.001
Family history of cancer
No	52	(22.2)	91	(32.3)	1 (Reference)
Yes	182	(77.8)	191	(67.7)	1.79 (1.07–2.98)	.03

^*Adjusted for age, race, education level, HCV, HBV, alcohol drinking, cigarette smoking, history of diabetes, family history of cancer, obesity at age (20–40), and hypothyroidism;

^†Serological evidence of HCV and HBV were not determined in 25 control females;

^‡Smokers are as subjects who had smoked ≥100 cigarettes during their lifetime;

^¥Drinkers were subjects who had consumed at least 4 alcoholic drinks each month for 6 months in their lifetime.

Table 2 shows female characteristics in cases and controls. A significant impact was observed only among estrogen users yielding a 50% reduction in HCC risk compared with non-users; the estimated AOR (95% CI) was .50 (.29–.86) (Table 3). Long-term use of estrogen alone (>5 years) was reported by 64.2% of control subjects and 55.5% of case patients, yielding a significant reduction in HCC risk (AOR = .36, 95% CI [.20–.63]) compared with never-users. Estrogen use was mainly postmenopausal in cases (87/94) and in controls (177/179). Restricted analysis among white cases and controls did not meaningfully change the observed reduced risk of HCC among estrogen users.

Table 2.

Distribution of Female Characteristics by disease status (HCC cases and Healthy controls)

Demographic variables	HCC patients		Controls		P value
	N =234	(%)	N =282	(%)
State of residency
Texas	117	(50)	161	(57.1)	.06
Other states	117	(50)	121	(42.9)
Marital Status
Single	83	(35.5)	25	(8.9)	<.0001
Married	151	(64.5)	257	(91.1)
Age of menarche
Mean (± SD)	12.86 ± 1.7		12.91 ± 1.7		.7
Hysterectomy
No	122	(52.1)	136	(48.2)	.2
Yes	112	(47.9)	146	(51.8)
Oophorectomy
No	25	(22.3)	46	(36.7)	.01
Yes	87	(77.7)	100	(63.3)
One ovary	14	(16.1)	15	(15)	.4
Two ovaries	73	(83.9)	85	(85)
Age of Hysterectomy
Mean (± SD)	39.3± 1.1		41.4±.8		.1
Menopause
No	35	(15)	39	(13.8)	.4
Yes	199	(85)	243	(86.2)
Age of menopause†
Mean (± SD)	48.78 ± 5.41		49.88 ± 4.48		.1
Prior history of Cancer
None	195	(83.3)	282	(100)
Breast	20	(8.5)	0
Endometrial	6	(2.6)	0
Cervix	4	(1.7)	0
Others	9	(3.8)	0

^†Exclude women who had oophorectomy

Table 3.

Association between oral contraceptives and exogenous hormonal replacement with HCC

Demographic variables	HCC patients		Controls		OR* (95% CI)	P value
	N =234	(%)	N =282	(%)	All
Contraception Use
No	88	(37.6)	74	(26.2)	1 (Reference)
Yes	146	(62.4)	208	(73.8)	.70 (.42–1.18)
Contraception Duration†
≤5 years	51	(35.7)	96	(46.2)	.41 (.22– .79)	.01
6–10	26	(18.2)	47	(22.6)	.54 (.26– 1.15)	.18
>10	66	(46.2)	65	(31.3)	1.38 (.74– 2.60)	.3
Estrogen Use
Never	140	(59.8)	103	(36.5)	1 (Reference)
Ever	94	(40.2)	179	(63.5)	.53 (.32– .88)	.01
Estrogen Alone	63	(26.9)	130	(46.1)	.50 (.29–.86)	.01
Estrogen & Progesterone	31	(13.3)	49	(17.4)	.62 (.31–1.27)	.19
Ever Estrogen Duration‡
≤5 years	37	(44.6)	64	(35.8)	.56 (.34–1.02)	.06
6–10	13	(15.7)	36	(20.1)	.32 (.13– .77)	.01
>10	33	(39.8)	79	(44.1)	.45 (.24– .85)	.01
Progesterone Use
Never	196	(83.8)	219	(77.7)	1 (Reference)
Ever	38	(16.2)	63	(22.3)	.84 (.47– 1.52)	.57
Progesterone Alone	7	(3)	14	(5)	1.22 (.35–4.22)	.75
Progesterone & Estrogen	31	(13.2)	49	(17.3)	1.38 (.45–4.29)	.56
Ever Progesterone duration§
≤5 years	21	(60)	33	(52.4)	1.08 (.51– 2.31)	.84
6–10	4	(11.4)	11	(17.5)	.57 (.14– 2.32)	.43
>10	10	(28.6)	19	(30.2)	.57 (.20–1.63)	.30

Abbreviations:

^*AOR, multivariate-adjusted odds for age, race, education level, marital status, HCV, HBV, alcohol drinking, cigarette smoking, history of diabetes, family history of cancer, obesity at age (20–40), hypothyroidism, oophorectomy, and marital status

^†DuraYon of OC use was unknown in 3 HCC cases

^‡DuraYon of estrogen use was unknown in 11 HCC cases

^§Duration of progesterone use was unknown in 3 HCC cases

The mean age at HCC onset among case patients who recalled estrogen use (± SE) was 64.5 ± .9 years, significantly higher than the mean age at onset of those with no estrogen use, which was 59.2 ± 1.1 years (P = .001). The mean difference in age at HCC onset between estrogen users and non-users was statistically significant after adjusting for other factors associated with age at HCC onset including smoking, alcohol, obesity, hypothyroidism, diabetes, HCV, HBV, marriage, history of pregnancy, educational level, and family history of cancer. The estimated coefficient was 4.65 (95% CI, 1.59–7.70) (P < .003). Considering the years of estrogen exposure in a continuous variable multiple linear regression analysis showed that predicted mean age at HCC onset increased with duration of estrogen use (Figure 1).

Figure 1.

Predicted mean age (years) at HCC onset and duration of estrogen use by linear regression; for example, the predicted mean ages at HCC onset at 6 years, 16 years, and 31 years of estrogen exposure were

As compared to no estrogen users without hepatitis virus infection, the ORs (95% CI) for estrogen use in the absence of hepatitis infection was 0.44 (.27–.74) and for hepatitis virus infection in the absence of estrogen use was 17.60 (3.88–79.83). However, estrogen use attenuated the magnitude effect of hepatitis virus infection on HCC risk, yielding an OR of 4.37 (95% CI, 1.67–11.44).

A total of 39 case patients recalled a prior history of cancer, especially breast cancer (N = 20) (Table 2), whereas additional 3 cases reported a prior history of nonalcoholic steatohepatitis, primary biliary cirrhosis, and autoimmune hepatitis. Restricted analysis among 192 cases and 282 controls without a prior history of cancers or chronic liver diseases did not change the observed reduced risk of HCC among estrogen users.

Figure 2A shows that OS of HCC patients was significantly longer among estrogen users than among non-users, P = .008. Figure 2B shows the univariate HRs (95% CI) of estrogen use and the clinical features of HCC at the time of diagnosis. Multivariate Cox regression analysis of the significant factors related to HCC prognosis indicated that estrogen use was significantly associated with 45% reduced mortality (Adjusted Hazard Ratio (AHR) = .55; 95% CI, .40–.77) (P = .01) after controlling for all confounding factors of HCC OS (Figure 2C).

Figure 2.

(A) Median OS (95%CI) by estrogen use.

(B) Univariate HRs (95% CI) of HCC prognostic factors.

(C) Multivariate AHRs (95% CI) of estrogen use (.55, .40–.77) after adjusting for significant confounding factors of survival including race, hysterectomy, oophorectomy, multi-nodular tumor, cirrhosis, extra-hepatic metastasis, >50% liver involvement, AFP, vascular invasion, TNM staging, and treatment type.

Prior history of pregnancy and pregnancy numbers were not significantly associated with HCC prognosis. Among HCC cases, we found that 66 (28.2%) never get pregnant, 142 (60.7%) had ≤ 3 pregnancies, and 26 (11.1%) had > 3 pregnancies. As compared to no pregnancy the HRs (95%) were .91 (.66–1.26), .86 (.62–1.20), and 1.24 (.74–2.10) for history of prior pregnancy, ≤ 3 pregnancies, and > 3 pregnancies respectively.

DISCUSSION

This study demonstrates 50% reduction in HCC risk development among women who used MHT. The observed reduced risk of HCC among estrogen users in this study was supported by three additional findings: 1) The positive correlation between age at HCC onset and duration of estrogen use. The adjusted linear regression analysis revealed significant coefficients indicating that in women with long-term use of estrogen, HCC tended to be diagnosed at an older age. 2) Attenuation of the magnitude of association between hepatitis virus infection and HCC development among estrogen users compared with non-users. 3) OS improvement in women with HCC who used estrogen compared with survival in non-users. The favorable prognostic observation of estrogen use was independent of the significant baseline clinical features of HCC related to HCC outcome.

Very few studies have investigated the association between MHT and HCC. However, the protective effect that we observed with postmenopausal estrogen use in US women agreed with the results from different populations. The multivariate AOR reported by Yu and colleagues ¹⁵ was .46 (95% CI, .27–.79). In addition, large nested case-control study within the United Kingdom’s Clinical Practice Research Datalink by McGlynn and colleagues ¹⁹ showed that the use of estrogen therapy was associated with a significantly lower risk of HCC (OR=0.44, 95% CI=0.22–0.88).

Findings from population studies with respect to the association between estrogen exposure and other cancers have been contradictory. Although some studies failed to show a significant impact of estrogen on pancreatic ²⁰ or bladder cancers, ²¹ the preventive effect of estrogen against liver cancer in the current study and in others studies was observed for gastric ²² distal large bowel cancer²³, and esophageal cancers ²⁴, with an average risk reduction of 28%, and the estimated ORs (95% CI) were .77 (.64–.92) and .68 (.48–.97), respectively. In contrast, estrogen use was significantly associated with increased risk of breast cancer. ²⁵ Our statistical analysis after excluding HCC women with prior cancers continued to show the protective effect of estrogen use.

Whereas other epidemiological studies focused on the relationship between HCC and some reproductive factors that may modify endogenous levels of female hormones such as age at menarche, age at menopause, hysterectomy, oophorectomy, and parity ^{14, 15, 26}, we observed no significant independent effect of oophorectomy on HCC risk.

There are potential lines of evidence that estrogen may protect against HCC development: the anti-inflammatory effect of estrogen via inhibition of the NF-kB pathway ²⁷ and possible suppression of the release of several pro-inflammatory cytokines that may deregulate the inflammatory and oxidative stress pathways involved in the carcinogenic process ^{28, 29}. Given the key role of IL-6 in carcinogenesis and poor outcome in HCC patients, Naugler et al. ³⁰ showed that estrogen treatment inhibited IL-6 production from Kupffer cells in female mice, leading to a reduction in liver cancer induction. Others have suggested that estrogen may inhibit hepatic tumor growth and progression by acting as a suppressor for alternative activation of tumor-associated macrophages and inhibiting the Jak1-Stat6 signaling pathway ³¹. A recent report provided new insight into the protection of estrogen in HCC via the regulation of NLRP3 inflammasome by estrogen through the ERβ/MARK pathway ³².

Several systemic reviews have shown a negative impact of tamoxifen treatment in HCC ^33–37. A possible explanation for the tamoxifen failure is the lack of defined eligible patients for treatment according to hormonal receptor expression and the possibility that tamoxifen may not be a candidate therapy for patients with variant estrogen receptors.

Given the anti-inflammatory role of estrogen, our finding of the attenuated risk of hepatitis infection among estrogen users versus non-users may not be surprising and is possibly explained by suppression of hepatitis-related hepatic inflammation and steatohepatitis ³⁸ and observed antifibrotic effect of estrogen in animal studies ^{39, 40}

Similar to the natural history of HCC,^{41, 42} most of our HCC patients presented with advanced-stage disease. In addition, heathy control subjects were selected to represent the population from which case patients were ascertained. Only U.S. patients and controls were included, and the geographic distribution of their residential states was similar. Moreover, age of natural menopause in our controls were similar to the general US population.⁴³ We choose not to use patients with cirrhosis as control subjects. We argue that this may lead to differential selection bias due to the significant association between estrogen and other environmental factors with fibrosis progression.

Given the U.S. Food and Drug Administration contraindication for MHT in patients with active liver diseases ⁴⁴ we found that majority of women with MHT had preserved liver function at time of HCC diagnosis.

In conclusion, this study provides robust epidemiological evidence for the benefits of postmenopausal use of estrogen replacement against HCC development and has been corroborated by previous studies. However, this study is the first to highlight survival improvement among women with HCC who used estrogen replacement, after controlling for clinical prognostic factors, which raises the questions of whether similar effects can be observed in men who ever experienced hormonal exposure and whether estrogen can be used in targeted therapy for a selected population based on tumor expression and types of estrogen receptors.

Abbreviation

HCC

hepatocellular carcinoma

MHT

menopausal hormonal therapy

CI

confidence interval

AFP

alpha-fetoprotein

HBV

hepatitis B virus

HCV

hepatitis C virus

AOR

adjusted odds ratio

OS

overall survival

HR

hazard ratio

SE

standard error

AHR

adjusted hazard ratio