Estradiol is significantly associated with prognosis in non-surgical liver cancer patients: from bench to bedside

Meta-analysis of the influence of gender on the OS of patients with liver cancer

The literature search retrieved 424 unique citations. After screening of titles and abstracts, 87 full-text articles remained for assessment, reporting that females had longer survival than males in different type cancer (pooled HR 0.86 [95% CI: 0.83–0.89]; p < 0.001; I² = 98%) (Supplementary Figure 1). Finally, 6 studies including 98,010 individuals were reviewed to appraise the effect of gender on the OS of liver cancer patients. The forest plot of gender for the efficacy in reducing the risk for prognosis in liver cancer patients is shown in Supplementary Figure 2. Among patients with liver cancer, females demonstrated better prognosis than males (pooled HR 0.93 [95% CI: 0.91–0.94]; p < 0.001; I² = 0%), and homogeneity performed well.

Characteristics of liver cancer patients from the SEER database

Data from 12,312 liver cancer patients housed in the SEER research database, who had not undergone surgery between 2010 and 2015, were included in this study. All participants were randomly divided into the primary (70% [n = 8658]) or validation (30% [n = 3654]) cohort. The baseline characteristics of these cohorts are summarized in Table 1. Among all patients, 27.2% (n = 2375) were female and 72.8% (n = 8966) were male. In total, 17.9% (n = 2200) were ≤ 55 years of age at diagnosis and 82.1% (n = 10,112) were > 55 years of age. The distribution according to race/ethnicity was as follows: white, 69.5% (n = 8553); black, 12.8% (n = 1580); and other (American Indian/AK Native, Asian/Pacific Islander), 17.7% (n = 2179). According to the American Joint Committee on Cancer, 28.5% (n = 3515) of patients were grade I at diagnosis, 47.2% (n = 5809) were grade II, 22.7% (n = 2800) were grade III, and 1.5% (n = 188) were grade IV. Among these individuals, 57.6% (n = 7090) were married, 19.1% (n = 2348) were single, 11.2% (n = 1378) were divorced, and 10.3% (n = 1269) were widowed. According to pathology, 88.7% (n = 10,924) had HCC, 11.1% (n = 996) had intrahepatic cholangiocarcinoma (ICCs), and others were combined. Regarding tumor size, 26.1% (n = 3217) of individuals exhibited lesions < 30 mm and 73.9% (n = 9095) had lesions > 30 mm. Tumor TNM stages are summarized in Table 1. Percentages were similar between the primary and validation cohorts.

Survival analysis

Univariate and multivariate Cox regression hazards models were used to analyze prognostic factors in patients with liver cancer (Tables 2, 3). In the univariate Cox models, age, sex, race, grade, tumor TNM stage, marital status, tumor size, and histological type were significantly associated with OS of non-surgical liver cancer patients (p < 0.05). Multivariate Cox analysis confirmed that all of those variables were independent prognostic factors for OS (p < 0.05).

According to the results of multivariate Cox analysis, patients < 55 years of age had a worse prognosis than those ≤ 55 years of age (HR 1.198 [95% CI: 1.109-1.295; p < 0.001) (Table 3). Compared with Caucasians, other races (i.e., American Indian/AK Native, Asian/Pacific Islander) had a better prognosis (HR 0.805 [95% CI: 0.742-0.873; p < 0.001), but blacks did not (p = 0.068) (Table 3). There were significant prognostic differences between males and females among non-surgical liver cancer patients. Females demonstrated a better prognosis than males (p < 0.001) (Figure 1A), and survived longer than males (32.3 ± 0.7 vs. 30.1 ± 0.4 months, respectively; p = 0.001). Compared with males, females had a lower hazard ratio (HR 0.868 [95% CI: 0.810-0.930; p < 0.001). As the clinical grade increased, patient prognosis worsened (p < 0.001); however, grade 2 was not an independent risk factor for prognosis in non-surgical liver cancer patients (p = 0.586). The HRs for tumor TNM stage are shown in Table 3. Patients who were married demonstrated better prognosis (HR 0.805 [95% CI: 0.805-0.953]; p = 0.011), and widowed patients demonstrated poor prognosis (HR 1.251 [95% CI: 1.029-1.410]; p < 0.001). Compared with ICCs, patients with HCC have a worse prognosis (HR 1.168 [95% CI: 1.060-1.287]; p = 0.002).

Figure 1. Kaplan-Meier survival analysis with log-rank test was performed in non-surgical liver cancer patients. (A) Kaplan-Meier survival analysis with log-rank test was performed in non-surgical liver cancer patients in the primary cohort. (B) Before menopause (i.e., ≤ 55 years of age), women with liver cancer have a better prognosis than men (p<0.0001, log-rank test).

Prognostic nomogram for OS and ROC of the model

Based on independent prognostic factors identified in the multivariate Cox regression analysis, a prognostic nomogram was developed for liver cancer patients to predict 3-and 5-year survival probabilities (Figure 2). In primary cohorts, the area under the ROC curves (AUROCs) of the model was 0.730 (p < 0.001) (Figure 3A), the consistency index of the model was 71.56 (95%CI: 71.17-71.96), and the internal calibration curve for the probability of survival at 3-and 5-years showed a good agreement between the nomogram-predicted probability of OS and actual survival (Figure 3B, 3C). In the validation cohort, AUROCs were 0.734 (p < 0.001) (Figure 3D), the consistency index was 70.74 (95%CI: 70.13-71.35), there was also a well-performed calibration curve for survival prediction (Figure 3E, 3F). Therefore, the nomogram could reliably predict the 3- and 5-year OS probabilities. Furthermore, independent nomograms were developed for patients ≤ 55 and > 55 years of age (Supplementary Figures 3, 4).

Figure 2. A nomogram for predict 3- and 5-year death odds of non-surgical liver cancer patients (established by Cox regression model). The yellow Violin Plot and the light blue box display the distribution of patients in the primary cohort. The size of the light blue box represents the proportion of patients. Abbreviations: HCC, hepatocellular carcinoma; ICCs, intrahepatic cholangiocarcinomas; Combined=HCC+ICCs.

Figure 3. The receiver operating characteristic (ROC) curves and calibration curve of the prognostic nomogram. (A) ROC curves of the prognostic nomogram in the primary cohort. (B, C) The calibration curve of the nomogram-predicted probability in the primary cohort (3-year OS and 5-year OS, respectively). (D) ROC curves of the prognostic nomogram in the validation cohort. (E, F) The calibration curve of the nomogram-predicted probability in the validation cohort (3-year OS and 5-year OS, respectively). Age, sex, race, grade, tumor TNM stage, marital status, tumor size, and histological type are pooled in the primary(a) and validation(b) cohorts. AUC= area under the curve.

Survival differences according to sex and sex hormones: from bench to bedside

The primary hypothesis, that female liver cancer patients have a better prognosis than males is attributable to different estrogen levels, was investigated. Based on previous studies [5, 14, 15], 55 years of age was defined as a surrogate for menopause. As shown in Figure 1B, before menopause (i.e., ≤ 55 years of age), there was an extreme gap between females and males in survival probability (39.4 ± 1.4 vs. 32.7 ± 0.8 months, respectively; p < 0.001). While this gap will be smaller after menopause (i.e., age > 55 years), the difference between females and males (30.5 ± 0.6 vs. 29.3 ± 0.4 months, respectively; p = 0.022) (Supplementary Figure 5) persisted. For further exploration, we obtained 265 male and 139 female liver cancer patients from the TCGA database, differentially expressed genes (DEGs) and KEGG pathway analysis results showed that there were significant differences in steroid hormone biosynthesis between male and female liver cancer patients (Figure 4).

Figure 4. Identification of differentially expressed genes and KEGG enrichment between male and female liver cancer patients. (A) Volcano plot of male and female liver cancer patients in differentially expressed genes (DEGs). (B, C) KEGG enrichment of DEGs. (D) gene network diagram between DEGs and KEGG pathways.

According to a previous study [16], estradiol may be the main source of endogenous estrogen in postmenopausal women. This indicates that gender differences in patient prognosis may be primarily attributable to estrogen in patients with non-surgical liver cancer. To confirm the effect of estrogen on liver cancer, different concentrations of estrone and estradiol were used to stimulate HCC cell lines, including Hep 3B, BEL-7402, and Huh7. The results demonstrated that estradiol inhibited the proliferation of HCC cell lines and increased apoptosis; however, estrone exerted no effect (Figure 5). The most obvious effect was obtained when the concentration of estradiol is 5μmol/L. This result suggests that the difference in prognosis between men and women among liver cancer patients may be mainly attributed to estradiol.

Figure 5. Estradiol inhibited the proliferation and increased the apoptosis of liver cancer cell lines. (A) Quantification of the apoptotic HCC cells population by flow cytometry. (B) Cell proliferation in HCC cells with different treatments was assessed using Cell-Light EdU Apollo 567 (catalog no· C10310-1; RiboBio). EdU positive cell rates were calculated; scale bar 50μm. Control (DMSO). *p<0.05, **p < 0·01, ***p<0·001. P < 0·05 was considered statistically significant. P values were calculated with Student's t-test.

Search strategy and selection criteria

A literature search of the PubMed database, performed using a combination of the keywords ((gender OR sex) AND “Marital status” AND race AND (cancer OR tumor)), retrieved a total of 424 relevant studies. After reviewing abstracts, 337 articles were excluded, and 87 tumor-related articles were used to verify the relationship between gender and tumor prognosis. Among these, 7 studies investigating liver cancer were retained, and one addressed liver transplantation. Ultimately, 6 studies were used to identify and appraise the effect of sex on overall survival (OS) in patients with liver cancer.

Data sources and patient selection

The Surveillance, Epidemiology, and End Results (SEER) Program provides information regarding cancer statistics of the United States. SEER is supported by the Surveillance Research Program (SRP) in the National Cancer Institute’s Division of Cancer Control and Population Sciences (DCCPS) (https://seer.cancer.gov/). Data were based on Incidence-SEER 18 Regs Research Data + Hurricane Katrina Impacted Louisiana Cases (1973-2015), among which 123,806 liver cancer patients were screened. Excluding operable (n = 7188) and unknown (n = 118,682) liver cancer patients, data from 12,312 non-surgical liver cancer patients were analyzed using SEER*Stat software. 265 male and 139 female liver cancer patients were obtained from the TCGA database for differentially expressed genes (DEGs) and KEGG pathway analysis.

In vitro experiments

The HCC cell lines Huh 7, Hep 3B, and BEL-7402 were maintained at 37° C in a humidified incubator with a 5% CO₂ atmosphere. After treatment with different concentrations of estrone, estradiol, and dimethyl-sulfoxide (as control) for 24 h, cell apoptosis and proliferation experiments were conducted. Apoptosis experiments were performed as previously described [30], and flow cytometry was performed using a commercially available kit (Annexin V PE apoptosis kit, BD Biosciences, San Jose, CA, USA). Cell proliferation was assessed using Cell-Light EdU Apollo 567 (catalog no. C10310-1; RiboBio, China), as previously described [31].

Statistical analyses

Statistical analyses were conducted using SPSS version 25.0 (IBM Corporation, Armonk, NY, USA), GraphPad Prism 8 (GraphPad Inc, San Diego, CA, USA) and the package of meta, ggplot, survival, and rms of R version 3.6.1.

Meta-analysis was performed using the package meta in R. The 95% confidence intervals (CI) reported in the available articles were extracted and pooled after calculating TE, seTE, and log transformed. Study heterogeneity was assessed using the I² statistic [32]. Survival curves for both the genders were plotted using the Kaplan-Meier method by R. Univariate and multivariate Cox regression analyses were used to estimate independent risk factors in non-surgical liver cancer patients on OS and to generate hazard ratios (HR) and corresponding 95% CI in SPSS version 25. All results were verified in the validation cohort. A prognostic nomogram was developed based on multivariate Cox regression analyses. The receiver operating characteristic (ROC) curve was plotted using SPSS version 25. All statistical tests were two-tailed and differences with p < 0.05 were considered as statistically significant.