Gender and menopause impact severity of fibrosis among patients with nonalcoholic steatohepatitis

Authors


  • Potential conflict of interest: Nothing to report.

  • Drs. Diehl, Abdelmalek, Guy, Pang and Suzuki received support through an American Recovery and Reinvestment Act (ARRA) grant from the NIAAA: 5RC2 AA019399 (Anna Mae Diehl, Principal Investigator). Drs. Diehl and Abdelmalek received funding support from NIH/NIDDK grant U01-DK57149. Dr. Abdelmalek was supported by a NIH/NIDDK K23 Career Development Award (K23-DK062116).

Abstract

Estrogens inhibit stellate cell activation and fibrogenesis. Thus, gender and reproductive states may influence the degree of fibrosis in patients with nonalcoholic steatohepatitis (NASH). To investigate the association between gender, menopause, and the severity of liver fibrosis in patients with NASH, we analyzed 541 adult patients enrolled from our Duke Liver Clinics (n = 338) and the Duke Metabolic and Weight Loss Surgery Program (n = 203) who had a histologic diagnosis of NASH. Multiple ordinal logistic regression models were used to assess the association between gender, menopause, and severity of liver fibrosis. Overall, men, premenopausal, and postmenopausal women composed 35.1%, 28.4%, and 36.5% of the population, respectively. The mean age was 48 years and 22% had advanced fibrosis. After adjusting for covariates (enrolling site, grades of portal inflammation, and hepatocyte ballooning) and potential confounders (race, body mass index, diabetes/prediabetes, hypertension), adjusted cumulative odd ratio (ACOR) and 95% confidence interval (CI) for greater fibrosis severity was 1.4 (0.9, 2.1) (P = 0.17) for postmenopausal women and 1.6 (1.0, 2.5) (P = 0.03) for men, having premenopausal women as a reference. There was borderline interaction between gender and age group divided by age 50, the average age at menopause in the U.S. (P = 0.08): ACOR and 95% CI of having greater fibrosis severity in men compared to women was 1.8 (1.1, 2.9) for patients with age <50 years (P = 0.02) and 1.2 (0.7, 2.1) for patients with age ≥50 years (P = 0.59). Conclusion: Men are at a higher risk of having more severe fibrosis compared to women before menopause, while postmenopausal women have a similar severity of liver fibrosis compared to men. These findings may be explained by the protective effects of estrogen against fibrogenesis. (Hepatology 2014;59:1406-1414)

Abbreviations
ACOR

adjusted cumulative odds ratio

BMI

body mass index

COR

cumulative odds ratio

CI

confidence interval

DUHS

Duke University Health System

HCV

hepatitis C virus

HRT

hormone replacement therapy

MOLR

multiple ordinal logistic regression model

NAFLD

nonalcoholic fatty liver disease

NASH

nonalcoholic steatohepatitis

NASH CRN

NASH clinical research network

NHANES

National Health and Nutrition Examination Survey

Nonalcoholic fatty liver disease (NAFLD) is a spectrum of liver disorders ranging from simple hepatic steatosis to inflammation, fibrosis, and cirrhosis. The prevalence of diabetes, obesity, and metabolic syndrome have been rapidly increasing in the U.S.[1] The prevalence of NAFLD, the hepatic manifestation of insulin resistance, has been increasing accordingly in a parallel way.[2] NAFLD is the most common liver disease in the U.S. and the disease burden of NAFLD is growing rapidly. The prevalence of NAFLD has been estimated to be around 10%-30%.[3] However, recent large single-center prospective study showed that the prevalence of NAFLD and nonalcoholic steatohepatitis (NASH) was 46% and 12.2%, respectively.[4] Moreover, NAFLD was reported to be the third most common cause of the liver cancer in the U.S.[5, 6]

The majority of patients with NAFLD have simple hepatic steatosis (fat >5% of liver weight) and these patients are asymptomatic and do not have increased mortality.[7] However, some patients with NAFLD present with NASH, which was shown to increase overall mortality.[8] Necroinflammation in the liver in patients with NASH can contribute to hepatic fibrogenesis, which can progress to cirrhosis. Age, diabetes, and obesity were reported to be associated with severe liver fibrosis.[9] Among the histologic features of NAFLD, hepatocyte ballooning and portal inflammation have been shown to be strong predictors of advanced fibrosis.[10, 11] Knowledge of factors modifying the disease progression of NAFLD, however, remains insufficient.

Women after menopause have increased risk of insulin resistance, hyperlipidemia, and visceral fat accumulation, all of which are known risk factors for NAFLD.[12] Previous epidemiological studies showed that gender and menopause impact the prevalence and incidence of NAFLD.[13] During reproductive age, the prevalence of NAFLD is higher in men compared to women, while, after menopause, the protective effect in women appears to be eliminated.[14, 15] Hormone replacement therapy (HRT) was shown to be protective against having NAFLD after menopause.[14, 16] In patients with chronic hepatitis C virus (HCV), postmenopausal women are associated with an increased risk of having advanced fibrosis compared to premenopausal women and HRT was protective against hepatic fibrosis.[17] The protective effect of premenopausal women or HRT against fibrogenesis in HCV-related liver disease appears to be mediated by estrogen. Estrogen was shown to prevent stellate cell activations in vitro and regression of fibrogenesis in vivo.[18] Currently, how gender and reproductive state impact the severity of fibrosis in patients with NAFLD remains largely unknown.

We theorize that premenopausal women are at a decreased risk of fibrosis progression compared to men, while postmenopausal women are at a similar risk for fibrosis progression as men due to the protective effect of estrogen among adult women patients with NAFLD during their reproductive age. We aimed to assess, in adult patients with NASH, whether gender and menopause are associated with the severity of fibrosis after taking into account other factors affecting the severity of liver fibrosis.

Materials and Methods

Study Design and Data Source

This was a cross-sectional study designed to investigate the effect of gender and menopause on the severity of liver fibrosis in patients with NASH. We used a dataset from the Duke University Health System (DUHS) NAFLD Clinical Database, which is a prospective open-enrolling database established in 2007 to collect clinical and research information from patients with NAFLD. The detailed design of this database project has been described in our previous publication.[19] Briefly, patients with clinical suspicion of NAFLD and age 18 or older who had clinical indications of liver biopsy were approached and enrolled through the Duke Liver Clinic and Duke Metabolic and Weight Loss Surgery Program. All the liver biopsy slides were reviewed and scored by hepatopathologists at Duke University using the NASH clinical research network (CRN) scoring system.[20] The DUHS NAFLD clinical database project was approved by the Duke University Institutional Review Board, and provided deidentified datasets to conduct this analysis.

Study Population

This study included a total of 541 patients with histological diagnosis of NASH, age 18 or over, and enrolled in the DUHS NAFLD clinical database between January 2007 and April 2010. Histologic diagnosis of NASH was defined as follows using the histologic scores recorded as described earlier: 1) steatosis (≥grade 1) accompanied by hepatocyte ballooning (≥grade 1) or lobular inflammation (≥grade 2); or 2) steatosis (≥grade 1) accompanied by fibrosis (≥stage 1) with any lobular inflammation (≥grade 1) or hepatocyte ballooning (≥grade 1). Patients with a history of alcohol abuse or significant alcohol use (more than 14 servings per week for men and more than 7 servings for women), serologic or histologic evidence of coexisting other diseases (viral hepatitis, primary biliary cirrhosis, autoimmune hepatitis, hemochromatosis, Wilson's disease, or alpha-1 antitrypsin deficiency) were excluded in our study.

Liver Histology

All liver biopsy specimens were stained with hematoxylin-eosin and Masson trichrome stains and reviewed by the hepatopathologists using a standardized scoring form created based on the NASH CRN scoring system.[20]

The primary outcome of this study was the severity of hepatic fibrosis. Stages 3 and 4 were considered advanced fibrosis and combined as a single category. Histologic degrees of ballooned hepatocytes and portal inflammation have been strongly associated with severity of fibrosis among patients with NASH.[21] The clinical associations have been endorsed by previous animal experiments which mechanistically linked ballooned hepatocyte/necroinflammation and portal inflammation to fibrogenesis.[22-24] Based on the current understanding, to investigate the potential impact of gender/menopause on fibrogenesis and/or fibrosis regression, we treated grades of hepatocyte ballooning and portal inflammation as covariates in the models and assessed potential associations between fibrosis severity and gender/menopause at a given degree of necroinflammation/portal inflammation.

Gender and Reproductive Variable

The primary predictor in this study was the gender/menopause classification: men, premenopausal women, and postmenopausal women. Menopausal state was classified based on self-reported reproductive information and history of oophorectomy, which were available via the DUHS NAFLD clinical database. When self-reported reproductive information or history of oophorectomy was not available (N = 49 of the 541 subjects), an age of 50 years, the average age at menopause in the U.S. was used to define menopausal state.[25]

Other Study Variables

Clinical information was collected at the time of liver biopsy using standardized questionnaires and medical chart review as a part of the DUHS NAFLD Clinical Database project. Other study variables analyzed in this study include enrollment site (Duke Liver Clinic versus Duke Metabolic and Weight Loss Surgery Program), age, race, body mass index (BMI), smoking history, alcohol consumption, and the presence of diabetes mellitus, prediabetes, and hypertension. Information on alcohol consumption was collected using a self-reported questionnaire (i.e., averaged quantity and frequency) and expressed as average servings per week. Self-reported smoking history was classified into current smokers versus others (including past smokers). Medication lists including estrogen were collected using standardized questionnaires and medical chart review. The presence of comorbid conditions was defined by combining the physician's diagnosis, medication history, both of which were obtained via chart review and the standardized questionnaires, and relevant laboratory data obtained within 30 days from the date of liver biopsy. Diabetes mellitus was defined by fasting blood glucose of ≥126 mg/dL, blood glucose at 2 hours (75 g glucose tolerant test) of ≥200 mg/dL, hemoglobin A1c ≥6.5%, physician's diagnosis, and/or the presence of any diabetic treatments. Prediabetes was defined by fasting blood glucose of ≥100 mg/dL and <126 mg/dL, blood glucose at 2 hours (75 g glucose tolerant test) of ≥140 mg/dL and <200 mg/dL, and/or the physician's diagnosis. Diabetes mellitus and prediabetes were analyzed as a combined variable. Hypertension was defined by physician's diagnosis and/or use of an antihypertensive.

Statistical Analyses

Data are reported as the mean ± standard deviation for continuous variables and as a percentage for categorical variables. Clinical characteristics of the study population were compared among the gender/menopause categories (i.e., men, premenopausal women, and postmenopausal women) using analysis of variance (ANOVA) for continuous variables or chi-square test for categorical variables. As our study population contains two populations enrolled at different sites, the variable of enrolled site was first assessed for potential interaction in the analyses. When no interaction was noted, the variable of enrolled site was included in the models as a covariate. Otherwise, separate models were developed for enrollment sites.

To assess the associations between gender/menopause categories and fibrosis severity, two different modeling approaches were used, taking into consideration enrollment site, grades of hepatocyte ballooning and portal inflammation, age, race, BMI, diabetes mellitus/prediabetes, and hypertension: recursive partitioning conditional inference tree model and multiple ordinal logistic regression model (MOLR).

The former, a nonparametric decision tree model, was used to identify factors that best classify fibrosis severity in a hierarchal manner among all the above-listed variables.[26] Conditional inference trees allow for complex interaction among variables and work as follows: 1) tests the global null hypothesis of independence between predictors and outcome, and implements a binary split for the selected variable with strongest association with response; 2) stop if hypothesis cannot be rejected; otherwise repeat step (1).

In the MOLR analysis, fibrosis stage was set as an outcome, the gender and menopause classification as the primary predictor, and grades of hepatocyte ballooning and portal inflammation and the enrolled sites as covariates. The variables of race, BMI, diabetes mellitus/prediabetes, and hypertension were also included in the models as potential confounders. Significance of the associations was determined using likelihood ratio tests. Magnitude of association was expressed as adjusted cumulative odds ratio (ACOR) and 95% confidence interval (CI), computed using premenopausal women as a reference group.

Further, in order to assess whether gender differently affects the severity of fibrosis before and after menopause, we developed a model including an interaction term (gender * age of <50 or ≥50) using an age cutoff of 50 years, the average age at menopause in the U.S.[25] Separate models for age < and ≥50 were also developed using the above-mentioned variables. In order to investigate the potential benefit of HRT against liver fibrosis, association between HRT and severity of liver fibrosis was assessed in postmenopause women.

Statistical analyses were performed using JMP statistical software v. 9.0 (SAS Institute, Cary, NC) and R v. 2.13.1 (http://www.r-project.org/), and considered differences statistically significant when the P < 0.05. P-values were not adjusted for multiple comparisons.

Results

Patient Characteristics

Between January 2007 and April 2010, there were 722 patients with NAFLD who were enrolled in the database study and whose NAFLD histologic scores were available. Three-quarters of the patients (N = 541) were classified as NASH according to our study definition and served our study population. Among the 541 patients, 338 patients (62%) were enrolled at Duke Liver Clinic and 203 patients (38%) at Duke Metabolic and Weight Loss Surgery Program. Clinical characteristics of the study population are summarized in Table 1. Overall, men, premenopausal, and postmenopausal women composed 35%, 28%, and 37% of the population.

Table 1. Clinical Characteristics of the Study Population
 Premenopausal WomenPostmenopausal WomenMenP Value
(N = 153)(N = 199)(N = 189)
  1. a

    Information on alcohol consumption was missing in 207 subjects while information on smoking status was missing in 201 subjects. The variables were not considered in further analyses due to the high frequency of the missing data.

Age, years (SD)40.1[7.8]56.0 [7.4]45.9 [11.7]<0.01
Enrolling site, metabolic and weight loss surgery %84 (54.9 %)68 (34.2 %)51 (27.0 %)<0.01
Race, Black28 (18.3 %)19 (9.6 %)11 (5.8 %)<0.01
BMI, kg/m244.6+10.639.2+9.437.6+8.4<0.01
Diabetes/impaired glucose tolerance, %76 (49.7 %)115 (57.8 %)82 (43.4 %)0.02
Hypertension, %94 (61.4 %)159 (79.9 %)133 (70.4 %)<0.01
Alcohol consumption,a servings per week1.11±0.130.41±0.120.23±0.15<0.0001
Current smokinga7 (6.0%)5 (3.5%)3 (3.8%)0.59
Hepatocyte ballooning   <0.01
Grade 057(38.0 %)43(21.7 %)43(22.8 %) 
Grade 159(39.3 %)90(45.5 %)112(59.3 %) 
Grade 234 (22.7 %)65 (32.8 %)34 (18.0 %) 
Lobular inflammation   0.93
Grade 04 (2.7 %)3 (1.5 %)3 (1.6 %) 
Grade 1103 (68.2 %)138 (70.4 %)132 (71.0 %) 
Grade 2-344 (29.1 %)55 (28.1 %)51 (27.4 %) 
Steatosis   <0.01
Grade 162 (40.5%)99 (49.8 %)72 (38.1 %) 
Grade 250 (32.7 %)60 (30.2 %)88 (46.6 %) 
Grade 341 (26.8 %)40 (20.1 %)29 (15.3 %) 
Portal inflammation   0.06
Grade 086 (55.7%)105 (53.8%)121 (65.0%) 
Grade 166 (44.3%)90 (46.2%)66 (35.0%) 
Fibrosis   0.02
Stage 017 (11.1%)18 (9.1%)18 (9.5%) 
Stage 177 (50.3%)70 (35.2%)70 (37.0%) 
Stage 237 (24.2%)56 (28.1%)59 (31.2%) 
Stage 3-422 (14.4 %)55 (27.6 %)42 (22.2 %) 

The mean age of the group was the highest in postmenopausal women (56 ± 7 years), followed by men (46 ± 12 years) and premenopausal women (40 ± 8 years) (P < 0.01). Compared to the postmenopausal women and men, premenopausal women had a higher prevalence of Black race (P < 0.01) and higher BMI (P < 0.01) and were more frequently enrolled in the Duke Metabolic and Weight Loss Surgery Program (P < 0.01). On the other hand, the proportion of the patients with diabetes/prediabetes (P = 0.02) and hypertension (P < 0.01) was higher in postmenopausal women. Stage of hepatic fibrosis (P = 0.02), grades of hepatocyte ballooning (P < 0.01), and steatosis (P < 0.01) were significantly different among the gender/menopause categories.

Relationship of Fibrosis Severity to Grades of Other Histologic Features and Clinical Variables

After adjusting for the site of enrollment, grade of hepatocyte ballooning, lobular inflammation, steatosis, and portal inflammation had positive associations with the severity of liver fibrosis.(Table 2) Among histologic variables, the grade of portal inflammation and hepatocyte ballooning were independently associated with the severity of liver fibrosis (Supporting Table 1). In addition to histologic variables, age, diabetes/prediabetes, and hypertension had positive associations with the severity of liver fibrosis, while Black race had an inverse association with the severity of liver fibrosis, independent of the site of enrollment.

Table 2. Relationship of Fibrosis Stage to Other Histologic Features and Clinical Variables Adjusted for Enrolled Sites
 COR95% CIP Value
  1. All the above analyses were performed with the adjustment for enrolled site.

  2. a

    There was significant site-interaction (P < 0.01). For portal inflammation, COR (95%CI) of grade 1 vs. grade 0 in having more advanced fibrosis was 1.47 (0.87-2.51, P = 0.15) in the Metabolic and Weight Loss Surgery Program and 4.52 (2.94-7.03, P < 0.01) in Duke Liver Clinic. For lobular inflammation, COR (95%CI) of grade 1 vs. grade 0 in having more advanced fibrosis was 12.00 (1.75-100.18, P < 0.01) in the Metabolic and Weight Loss Surgery Program and 2.16 (0.32-15.56, P = 0.43) in Duke Liver Clinic while COR of grade 2-3 vs. grade 0 was 6.55 (0.88-58.32, P = 0.07) in the Metabolic and Weight Loss Surgery Program and 4.78 (0.69-35.13, P = 0.11) in Duke Liver Clinic.

Hepatocyte ballooning  <0.01
Grade 0 
Grade 11.541.03-2.31 
Grade 25.143.16-8.44 
Lobular inflammationa  <0.01
Grade 0 
Grade 14.651.18- 18.75 
Grade 2-37.431.85-30.69 
Steatosis  0.05
Grade 1 
Grade 21.130.79-1.60 
Grade 31.661.10-2.50 
Portal inflammationa  <0.01
Grade 0 
Grade 12.902.09-4.03 
Age (per 5 year)1.161.05-1.22<0.01
Race, Black  <0.01
No 
Yes0.440.26-0.72 
BMI, kg/m21.011.01-1.030.21
Diabetes/prediabetes  <0.01
No 
Yes2.101.53-2.87 
Hypertension  <0.01
No 
Yes1.731.22-2.44 

Relationship of Fibrosis Severity to Gender, Menopause, and Estrogen Replacement Therapy

After adjusting for the site of enrollment, grade of portal inflammation, and hepatocyte ballooning, the gender/menopause categories were significantly associated with fibrosis stage (P < 0.04) and postmenopausal women and men had a 1.6-fold (95% CI, 1.0-2.4), 1.7-fold (95% CI, 1.1-2.6) increased risk of having greater severity of liver fibrosis than premenopausal women, respectively. This suggests that at any given degree of hepatocyte ballooning and portal inflammation, postmenopausal women and men have a 60%-70% increased risk of having a more advanced fibrosis stage when compared to premenopausal women (Table 3). When Black race, BMI, diabetes/prediabetes, and hypertension were added to the model, there was a trend toward an association between the gender/menopause categories and fibrosis stages (P = 0.10) and postmenopausal women and men had 1.4-fold (95% CI, 0.9-2.1), 1.6-fold (95% CI, 1.0-2.5) increased risk of having greater severity of liver fibrosis than premenopausal women (Table 3). After excluding the 49 women whose reproductive information was not available, we ran the full model. The association between the gender/menopause categories and fibrosis stages was consistent and even slightly stronger (P = 0.05); postmenopausal women and men had 1.5-fold (95% CI, 1.0-2.4), 1.8-fold (95% CI, 1.1-2.9) increased risk of having greater severity of liver fibrosis than premenopausal women (Supporting Table 2).

Table 3. Association Between Gender/Menopause and Fibrosis Severity With and Without Adjusting for Other Variables
 Model 1P ValueModel 2P ValueModel 3P Value
COR [95% CI]COR [95% CI]COR 95% CI
  1. Model 1: adjusted for the site of enrollment.

  2. Model 2: adjusted for the site, grade of portal inflammation, hepatocyte ballooning.

  3. Model 3: adjusted for the site, grade of portal inflammation, hepatocyte ballooning, Black race, BMI, diabetes/prediabetes, and hypertension.

  4. Age was not included in the multivariable model due to the strong correlation between menopause and age.

Gender and menopause 0.03 0.04 0.10
Premenopausal women  
Men1.36 [0.91-2.03] 1.66 [1.09-2.54] 1.62 [1.04-2.53] 
Postmenopausal women1.70 [1.15-2.52] 1.56 [1.03-2.35] 1.35 [0.88-2.07] 
Hepatocyte ballooning   <0.01 <0.01
Grade 0   
Grade 1 1.65 [1.08-2.51] 1.48 [0.97-2.26] 
Grade 2 4.73 [2.87-7.85] 4.15 [2.50-6.94] 
Portal inflammation   <0.01 <0.01
Grade 0   
Grade 1 2.68 [1.91-2.68] 2.85 [2.02-4.05] 
Enrolling siteMetabolic and Weight Loss Surgery <0.01 <0.01 0.02
No   
Yes0.51 [0.37-0.71] 0.61 [0.42-0.89] 0.56 [0.35-0.91] 
Race, Black     <0.01
No    
Yes  0.38 [0.22-0.65] 
BMI, kg/m2  1.01 [0.99-1.03]0.33
Diabetes/prediabetes     <0.01
No    
Yes  2.17 [1.54-3.08] 
Hypertension     0.30
No    
Yes  1.23 [0.84-1.80] 

The result of the conditional inference tree model is summarized in Fig. 1. The model identified the category of premenopausal women versus others as a significant factor in classifying the severity of fibrosis in patients with NASH, along with hepatocyte ballooning, diabetes/prediabetes, portal inflammation, age, hypertension, and enrolling sites (Fig. 1).

Figure 1.

Conditional inference tree model. The figure demonstrates a decision tree developed with the same set of variables used for ordinal logistic regression models. In the tree, variables with the strongest association of fibrosis severity were split in a hierarchical manner. A detailed methodological explanation and relevant references were provided in the Methods section. Numbers in blue: median fibrosis stage in the groups. DM: diabetes mellitus; Pre-MP: premenopausal women.

Further, to assess whether gender would differentially affect the severity of fibrosis before and after the age 50 years (the average age of menopause in the U.S.), we developed a model including an age-gender interaction term. There was a borderline interaction between gender and the age groups in the relationship to fibrosis stage (P = 0.08). To compare the different gender effect before and after age 50 years, we developed separate models as shown in Table 4. Before age 50, men had 1.8-fold (95% CI, 1.1-2.9; P = 0.02) increased risk of having greater severity of fibrosis than women after adjusting for the site of enrollment, grade of portal inflammation, and hepatocyte ballooning, Black race, BMI, diabetes/prediabetes, hypertension, and age. In contrast, after age 50 the protective effect observed in women appeared to be eliminated (1.2 [95% CI, 0.7-2.0; P = 0.59]). Of note, 16% (N = 31) of postmenopausal women were <50 years old, while 11% (N = 16) of premenopausal women were ≥50 years old in this population. After removing the 47 subjects, overall results were essentially the same, with a smaller P value of 0.05 for the interaction term between gender and age groups (Supporting Table 3).

Table 4. Gender and Severity of Liver Fibrosis: Subgroup Analysis According to the Age Above or Below 50 Years Old
 <50 years ≥ 50 years 
Adjusted COR & 95%CIP ValueAdjusted COR & 95%CIP Value
  1. Models were adjusted for the site, grade of portal inflammation, hepatocyte ballooning, Black race, DM/prediabetes, and age.

  2. Interaction of gender*age group (< and ≥ average age of menopause, 50 years old): P = 0.08. = P-values: likelihood ratio tests.

  3. In this cohort, 16% (N = 31) of postmenopausal women were <50 years old, while 11% (N = 16) of pre-menopausal women were ≥50 years old.

Women  
Men1.77 [1.08-2.90]0.021.17 [0.67-2.04]0.59

Furthermore, the effect of age ≥50 was assessed in men and women separately. After adjusting for the covariates (enrolling site, degrees of portal inflammation, and hepatocyte ballooning), age ≥50 was associated with a significantly increased risk of having more advanced fibrosis only among women (ACOR = 1.8 [1.2, 2.7], P < 0.01) but not men (ACOR = 1.0 [0.6, 1.8], P = 0.89). After adding the other potential confounders in the model (age, Black race, BMI, diabetes/prediabetes, and hypertension), the results remained similar: there was a trend that age of ≥50 was associated with an increased risk of having more advanced fibrosis only among women (ACOR = 1.5 [1.0, 2.4], P = 0.06) but not men (ACOR = 0.9 [0.5, 1.7], P = 0.80).

As an exploratory analysis, we examined the effect of estrogen replacement therapy on fibrosis severity among postmenopausal women. A total of 23 out of 199 postmenopausal women (12%) were under estrogen replacement therapy. After adjusting for the covariates (the site of enrollment, grade of portal inflammation, and hepatocyte ballooning), estrogen replacement was associated with a 50% risk reduction, although the association did not reach statistical significance (ACOR = 0.5; 95% CI: 0.2-1.2; P = 0.11). After adding the other potential confounders in the model (age, Black race, BMI, diabetes/prediabetes, and hypertension), ACOR did not change significantly: 0.6; 95% CI: 0.3-1.4; P = 0.23.

Discussion

We performed a cross-sectional study investigating the effect of gender and menopause on the severity of liver fibrosis using a large single-center prospective database that included patients with a histologic diagnosis of NASH. To address the impact of gender/menopause on hepatic fibrosis, we applied a hypothesis-driven analytic approach to adjust for two strong histologic predictors of advanced fibrosis (i.e., hepatocyte ballooning and portal inflammation), both of which have been mechanistically linked to fibrogenesis.[21-23] In this study we showed that postmenopausal women and men were associated with an increased risk of having greater severity of liver fibrosis than premenopausal women, at any given degree of hepatocyte ballooning and portal inflammation. The conditional inference tree models also identified premenopausal women (versus others) as a significant factor in classifying the severity of fibrosis in patients with NASH. There was a trend for an age-gender interaction; before age 50, men had an increased risk of having severe fibrosis compared to women, while after the age 50, the protective effect observed in women was markedly eliminated. Further, the age effect (age ≥50) on fibrosis severity appeared to be observed among women, but not men. Lastly, although underpowered, estrogen replacement among postmenopausal women was associated with a 50% decreased risk of having more advanced fibrosis compared to no estrogen replacement (although statistical significance was not reached). Taken together, these findings suggest that, at any given degree of hepatocyte ballooning and portal inflammation, premenopausal women (or age younger than 50 years) are protected from having more advanced fibrosis compared to men, although this protective effect seems to be eliminated after the age of menopause. Estrogen replacement among postmenopausal women appeared to be protective against developing more advanced fibrosis. Collectively, our findings support the aforementioned study hypothesis and suggest protective estrogen effects on hepatic fibrosis among patients with NASH.

The protective effect of estrogen was demonstrated particularly in patients with HCV. A single-center prospective study with a total of 251 women with HCV showed that menopause and hormone replacement therapy affect the severity of liver fibrosis: postmenopausal women were more likely to have moderate-severe liver fibrosis (odds ratio [OR]: 3.7, P < 0.01) compared to premenopause women but the probability of moderate-severe fibrosis was lower for postmenopausal women receiving HRT (OR: 0.35, P = 0.01) compared to women not receiving HRT after adjusting for duration of infection, severity of steatosis, and BMI.[17] Another retrospective study examined the effect of menopause and HRT on liver fibrosis progression in 472 HCV-infected women.[27] Postmenopausal women were more likely to have higher mean fibrosis score (AOR: 9.3, P = 0.02) and rate of fibrosis progression (P < 0.05) compared to premenopausal women after adjusting for age, BMI, and necroinflammatory activity. The mean fibrosis score (P < 0.05) and estimated rate of fibrosis progression (P = 0.02) were lower in women who received HRT compared to those who did not receive HRT among the postmenopausal women.

Several studies investigated effect of estrogen/antiestrogen in the development of NAFLD. In a large prospective randomized controlled chemoprevention trial with tamoxifen in Italy (n = 5,408), taxomifen use was associated with an increased risk of developing NAFLD (hazard ratio, 2.0; P = 0.04).[28] A study using the National Health and Nutrition Examination Survey (NHANES) showed that the risk of NAFLD is lower in postmenopausal women taking HRT compared to postmenopausal women not taking HRT (OR: 0.69, P < 0.05).[14] Similarly, a randomized controlled trials with 50 diabetic women showed that low-dose HRT for 6 months resulted in a significant reduction in liver enzymes.[16] The protective effect of estrogen against liver fibrosis was also suggested in pediatric populations. Our previous study showed that patients at or beyond puberty were less likely to have high-grade steatosis, severe portal inflammation, or a high stage of fibrosis than patients who had not entered puberty and we speculated that increased estrogen release at or beyond puberty are primarily responsible for this observation.[29]

Several experimental studies have shown the protective effect of estrogen against liver fibrosis. One study showed the effect of oophorectomy and estrogen replacement on hepatic fibrosis using rats.[18] Hepatic fibrosis induced by dimethylnitrosamine (DMN) was more prominent in male rats compared to female rats. Estrogen decreased collagen synthesis in male rats exposed to DMN. Female rats with oophorectomy had increased expression of procollagen. Estrogen replacement has reversed enhanced fibrogenesis observed in female rats with oophorectomy. Stellate cells incubated with estrogen resulted in decreased collagen synthesis, suggesting that estrogen suppresses the activity of hepatic stellate cells, thereby decrease the collagen production and progression of liver fibrosis. Estrogen was shown to decrease the generation of reactive oxygen species (ROS), transforming growth factor (TGF) beta-1 expression, hepatic stellate cell activation and proliferation, which are integral processes of hepatic fibrogenesis.[30]

The major strength of our study is the large number of patients with NASH in a prospectively designed database. The diagnosis of NASH was based on liver histology, and it was confirmed by experienced pathologists based on current guidelines. Also, we applied multiple modeling approaches to address our research question, which provided complementary information to enrich our data interpretation. Our study, however, has several limitations. First, due to the cross-sectional nature of our study we were unable to prove causality between gender/menopause and severity of fibrosis. Second, referral bias is also a potential limitation, as we included patients from a single academic institution and patients enrolled in the DUHS NAFLD database. Third, our study population did not contain enough subjects who were under estrogen replacement therapy, which precluded us from conclusive analysis on a potential beneficial association between estrogen replacement and severity of fibrosis among postmenopausal women. Our database did not provide anthropometric measures which could potentially have confounded our findings. Lastly, 49 of 352 women had no data on reproductive information and their menopause category was classified based on their age relative to the average age at menopause in the U.S. population. Based on our sensitivity analysis after excluding the 49 women, our estimates of ACOR might have been underestimated due to the misclassification, and the true effect may be even larger than what was observed in the current study.

In summary, our study revealed that postmenopausal women and men are associated with an increased risk of having more advanced fibrosis compared to premenopausal women at a given degree of hepatocyte ballooning and portal inflammation. Considering the findings from previous animal experiments, the observed association between premenopausal women and a decreased risk of hepatic fibrosis may be explained by protective effects of estrogen in fibrogenesis. Further studies are warranted to investigate the impact of estrogen on fibrosis progression in patients with NASH and potential preventive and/or therapeutic effects of estrogen among postmenopausal women with NASH.

Acknowledgment

The authors thank our colleagues in the Duke Division of Gastroenterology, Duke Metabolic and Weight Loss Surgery Program, and Center of Human Genetics for assistance with patient recruitment, acquisition, and/or management of DUHS NAFLD Clinical Database. The authors also thank the NAFLD clinical research manager, study coordinators, clinical support staffs' research and data management personnel, and the patients and their families without whom this study would not have been possible.

Ancillary

Advertisement