Menopausal Hormone Replacement Therapy and the Risk of Ovarian Cancer: A Meta-Analysis

Background: Findings by epidemiologic studies on menopausal hormone replacement therapy (HRT) and the risk of ovarian cancer are inconsistent. This study aimed to assess the association of menopausal HRT with the risk of ovarian cancer by histological subtype. Methods: A literature search was performed in PubMed, Web of Science, and EmBase for relevant articles published from inception to August 2018. Pooled relative risk ratios (RRs) with 95% confidence intervals (CIs) were determined with a random-effects model. Results: Thirty-six studies involving 4, 229, 061 participants were included in this meta-analysis. The pooled RR of ovarian cancer was 1.29 (95%CI 1.19–1.40, I2 = 57.4%) for menopausal HRT. In subgroup analysis by study design, pooled RRs of ovarian cancer in cohort and case-control studies were 1.35 (95%CI 1.19–1.53) and 1.24 (95%CI 1.11–1.38), respectively. In subgroup analysis by continent, association of menopausal HRT with ovarian cancer was significant for North America (1.41 [1.23–1.61]), Europe (1.22 [1.12–1.34]), and Asia (1.76 [1.09–2.85]), but not Australia (0.96 [0.57–1.61]). Association differed across histological subtypes. Increased risk was only found for two common types, including serous (1.50 [1.35–1.68]) and endometrioid (1.48 [1.13–1.94]) tumors. Conclusion: This meta-analysis suggests that menopausal HRT may increase the risk of ovarian cancer, especially for serous and endometrioid tumors.


INTRODUCTION
Ovarian cancer is known as the most lethal genital system malignancy (1). It is also the fifth leading cause of cancer-related deaths in American women (1). In 2018, the estimated new ovarian cancer cases and deaths will be 22,240 and 14,070 in the US, respectively (1). In 2018, the age standardized incidence rate of ovarian cancer is 6.6 per 100,000 in world 1 . Ovarian cancer can be divided into five histologic subtypes: serous tumor, mucinous tumor, endometrioid tumor, clear cell tumor, and other type of ovarian cancer. And the different histologic types of ovarian cancer may has different protective factors or pathogenic factors. Breastfeeding (2) and oral contraceptives (3) have been confirmed as protective factors in ovarian cancer. However, other exposures such as obesity (4,5), diabetes (6), miscarriage (7) and a family history of breast/ovarian cancer (8) are demonstrated risk factors for ovarian cancer.

Literature Search Strategy
We performed a literature search to identify relevant available articles from PubMed, Web of Science and EmBase from inception to August 2018 with no restrictions. Search terms included "hormone replacement therapy" (or "HRT") and "ovarian cancer" (or "ovarian neoplasms" or "ovarian carcinoma" or "ovary cancer"). The reference lists of the included studies were also reviewed for potential relevant studies.

Inclusion Criteria
Inclusion criteria were: (1) original report from observational studies; (2) menopausal HRT as the exposure of interest; (3) ovarian cancer as the outcome of interest; (4) relative risk ratio (RR) with 95% confidence interval (CI) provided. The most recent and complete study was selected if studies from the same population were repeated.
Two investigators searched and reviewed all relevant studies independently. Any disagreement was resolved by consensus with the involvement of a third reviewer.

Data Extraction
The following information were extracted from each study by two investigators independently: first author's name, published year, country, study design, follow-up duration, age range or mean age at baseline, sample size and number of cases, histological subtype of ovarian cancer, the types of hormones used in the study population, RR (we presented all results as RR for simplicity) with 95%CI and adjustment for potential confounders. We extracted RRs adjusted for the most confounding factors in the original studies. We prioritized the RRs for highest vs. lowest duration category of HRT use. If the study did not provide RRs for highest vs. lowest duration category of HRT use, we extracted the RRs for "use vs. non-use."

Statistical Analysis
The Newcastle-Ottawa Scale was used to assess the quality of studies included in this meta-analysis. Pooled data were obtained as the inverse variance-weighted means of the logarithm of RRs with 95%CI to assess the associations of menopausal HRT and the risk of different histological subtypes of ovarian cancer, respectively. The DerSimonian and Laird random effects model (REM) was used to combine study-specific RRs (95%CIs). The I 2 statistic was adopted to assess heterogeneity among studies (I 2 -values of 0, 25, 50, and 75% represented no, low, moderate and high heterogeneity, respectively). Metaregression with restricted maximum likelihood estimation was performed to explore the important covariates that might have significant impact on between-study heterogeneity. Subgroup analyses were stratified on study design, geographic location and the types of hormones used in the study population. Sensitivity analysis was performed with one study removed at a time to assess whether the results could have been affected markedly by a single study. The funnel plot and Egger's test were performed to explore the smallstudy effect.
All statistical analyses were performed with STATA version 14.0 (Stata Corporation, College Station, TX, United States). All reported probabilities (P-values) were two-sided, with a statistical significance level of 0.05.

Literature Search Results
We identified 2,445 articles by literature search, of which 2,387 were excluded after title and abstract review (Figure 1). Three additional articles were found by searching the reference lists of included articles. Eleven articles with duplicate data from the same population, 13 reports without RR and/or 95%CI and one article assessing the risk of ovarian cancer mortality were excluded. Finally, 36 published articles were eligible for this meta-analysis.

Characteristics of Studies
For the association of menopausal HRT with the risk of ovarian cancer, 34 articles (11-28, 30-42, 44-46) (15 cohort and 19 casecontrol studies) were included, involving 3,305,108 participants. The Newcastle-Ottawa Scale indicated that most of the studies included in this meta-analysis were of high quality (thirty of them scored more than seven). Among these studies, 15 were performed in Europe, 15 in North America, 2 in Asia and 2 in Australia. For the association of menopausal HRT and the risk of ovarian cancer by histological subtype, 12 studies (21,29,35,(38)(39)(40)(41)(42)(43)(44)(45)(46)

Quantitative Synthesis
The association of menopausal HRT with the risk of ovarian cancer is summarized in Table 2.

Meta-Regression and Sensitivity Analysis
To assess between-study heterogeneity, we performed univariate meta-regression with the covariates of study design, publication year and continent. However, none of these covariates was found to have a significant impact on between-study heterogeneity. After excluding two study (16,20) (RR > 3.0) in the ovarian cancer assessment, the heterogeneity remained at a moderate level (I 2 = 52.1%, P heterogeneity < 0.001), and the pooled RR was 1.27 (95%CI 1. 17-1.37).
In sensitivity analysis excluding one study at a time, pooled RRs (95%CIs) of the association of menopausal HRT with the risk of ovarian cancer ranged from 1.

Publication Bias
Visual inspection of the funnel plot ( Figure S3) and Egger's test (P ovarian cancer = 0.083) showed no evidence of significant small-study effect for the association of menopausal HRT with the risk of ovarian cancer. Egger's test also provided no evidence of significant small-study effect for the association of menopausal HRT with the risk of ovarian cancer by histologic subtype (P serous tumor = 0.762, P endometrioid tumor = 0.550, P mucinous tumor = 0.655, P clear call tumor = 0.349, P other types of ovarian cancer = 0.892).

DISCUSSION
The current meta-analysis assessed associations of menopausal HRT with the risk of ovarian cancer in various histologic subtypes. Findings of this meta-analysis indicated a positive association of menopausal HRT with the risk of ovarian cancer. In subgroup analysis by study design, significant positive associations were observed in both cohort and case control studies. In subgroup analysis by histologic subtypes, we found that menopausal HRT may increase the risk of serous and endometrioid tumors. In subgroup analysis by the hormones types, significant positive associations were observed for both HRT and ERT. The pooled RR indicated that there might be a stronger association in ERT users, but the result might be not true enough with the insufficient studies about ERT.
The mechanism underlying the association of menopausal HRT with ovarian cancer is not well-understood. A theory suggests that high levels of gonadotropins during menopause act as a promoter on the affected ovarian tissue (49). These findings imply that menopausal HRT might decrease the risk of cancer by reducing the levels of gonadotropins. However, these benefits might be outweighed by estrogen-induced ovarian cell proliferation (50). Estrogen and progesterone receptors are found in normal ovarian surface and most of ovarian tumors are estrogen receptor-positive (51,52). Estrogen could stimulate the proliferation of ovarian surface epithelial cells and progesterone could promote the apoptosis of ovarian cells. The weaker risk effect of HRT than ERT may be because progesterone counteract the proliferative effect of estrogen on ovarian cells (52)(53)(54). Between-study heterogeneity is common in metaanalysis. It is necessary to explore the potential sources of between-study heterogeneity. In this meta-analysis, a moderate between-study heterogeneity was found. However, meta-regression analysis with the covariates of study design, published year and continent revealed no source of between-study heterogeneity. After excluding two study (20) (RR > 3.0) in the analysis of menopausal HRT and ovarian cancer, between-study heterogeneity was slightly reduced, but results did not change substantially. This indicated that the results were stable and credible. This meta-analysis had some advantages. The first is the sufficient sample size that made the study had high statistical power to detect even small associations. Secondly, we extracted RRs reflecting the highest degree of control for potential confounders in the original studies. This will help us to get a real connection between the factors and the disease. Thirdly, sensitivity analysis showed that no individual study had excessive effects on pooled data for menopausal HRT and the risk of ovarian cancer by histologic subtypes. Fourthly, after excluding two study (RR>3.0) in ovarian cancer analysis, between-study heterogeneity was slightly reduced, and the results did not change substantially, suggesting that they were stable. Fifthly, in FIGURE 3 | Forest plot of menopausal HRT and the risk of ovarian cancer in subgroup analysis stratified by study design. The size of a gray box is proportional to the weight assigned to the respective study, and horizontal lines represent 95% confidence intervals (CIs). subgroup analysis stratified by the hormones types, we found both HRT and ERT could increase the risk of ovarian cancer.
However, there were still some deficiencies in this metaanalysis. First, the authors adjusted for confounders such as age, parity, duration of oral contraceptive use, tubal ligation, age at natural menopause and age at menarche etc. in original studies, but we dare not deny whether some unknown confounders might lead to exaggerating or underestimating the association. In addition, confounders adjusted for in various studies were different, which might affect the observed association. Some common biases such as selection bias, recall bias and lost to follow-up etc. in observational studies might also affect the authenticity of the results. Secondly, follow-up durations in various cohort studies differed. Some potential cases might not be observed due to limited follow-up in certain studies. Thirdly, menopausal HRT might be slightly different in each of the papers analyzed. In some papers, the HRT referred to estorgens + progestins, but in others, the HRT referred to only estorgens or estorgens + progestins. This might affect the observed association. Fourthly, the limited amount of studies  assessing histologic subtypes made it difficult to confirm the relationship in terms of the kind of therapy and its association with the histological subtypes of ovarian cancer. Fifthly, although the age of subjects was over 50 years old in most included studies, a few studies covered the data from women from pre-menopausal age. This might bias the results of the meta-analysis. Sixthly, the insufficient available data prevented us from conducting a doseresponse relationship to explore the association between length of HRT use and the risk of ovarian cancer.
In conclusion, this meta-analysis suggests that menopausal HRT may increase the risk of ovarian cancer, especially for serous and endometrioid tumors. This finding requires confirmation by further studies of associations of menopausal HRT with the risk of ovarian cancer in various histological subtypes.

DATA AVAILABILITY STATEMENT
All datasets for this study are included in the article/supplementary material.

SUPPLEMENTARY MATERIAL
The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fendo. 2019.00801/full#supplementary-material Figure S1 | Forest plot of menopausal HRT and the risk of ovarian cancer in subgroup analysis stratified by geographic location. The size of a gray box is proportional to the weight assigned to the respective study, and horizontal lines represent 95% confidence intervals (CIs). Figure S2 | Forest plot of menopausal HRT and the risk of ovarian cancer in subgroup analysis stratified by the hormones types. The size of a gray box is proportional to the weight assigned to the respective study, and horizontal lines represent 95% confidence intervals (CIs). Figure S3 | The funnel plot of menopausal HRT and the risk of ovarian cancer. Each dot represents a distinct study.