Metformin and endometrial cancer risk: A Meta-analysis

Objective: To assess the association between metformin and endometrial cancer risk by conducting a meta-analysis. Methods: A comprehensive search of electronic databases, conference abstracts, and clinical trial registers was performed to identify available evidence. Studies that evaluated the association between metformin use and endometrial cancer risk were considered. We pooled relative risk (RR) estimates with 95% confidence intervals (CIs) by using either a fixed-effects or a random-effects model. Results: Nine studies (1 randomized controlled trial, 5 cohort studies, and 3 case-control studies) involving more than 1.20 million participants and 7,762 cases of endometrial cancer met eligibility criteria. There was no evidence of an association between metformin use and endometrial cancer risk (RR, 0.96; 95% CI, 0.80 to 1.16). Similarly, the pooled data showed that metformin use was not significantly associated with endometrial cancer risk in patients with diabetes (RR, 0.93; 95% CI, 0.78 to 1.11). In subgroup analyses, the results were stable across different study designs and comparisons, and remained unchanged after adjusting for age and BMI or obesity. Conclusions: The current meta-analysis suggests that metformin has no chemopreventive effect against endometrial cancer.


Introduction
Endometrial carcinoma is the most common gynaecologic cancer in developed countries, with 60,050 new cases in the United States alone in the year 2016 [1].
Several risk factors have been reported to be related to an increased risk of endometrial cancer, including reproductive factors, obesity, and diabetes mellitus [2 − 4].
Diabetes mellitus has been linked to associated with an increased risk of endometrial cancer via mechanisms involving insulin resistance and secondary hyperinsulinemia [5,6]. In the condition of insulin resistance, the elevated insulin may have growth stimulation effects and directly and indirectly increase the risk of endometrial cancer [7,8]. The direct mechanism includes activation of key signaling pathways, while in the indirect mechanism, insulin interacts with insulin-like growth factor-1 (IGF-1) and affects sex hormone levels, which in turn may promote cell growth [7 − 9]. Insulin resistance is associated with a spectrum of disease, such as abnormal glucose metabolism, type II diabetes mellitus, obesity, polycystic ovary syndrome (PCOS) [10,11].
Metformin, an oral biguanide hypoglycaemic drug, is the the first line treatment of type II diabetes [12]. It is a well-tolerated and effective drug that normalizes hyperglycemia by reducing insulin resistance and lowering insulin levels [12,13].
Metformin also has beneficial effects in women with PCOS, obesity, and metabolic syndrome [14 − 16].
A growing body of evidence suggests that metformin may have a chemopreventive role against cancer. Several meta-analyses have shown that metformin use may decrease the incidence of site-specific cancers, including colorectal, pancreatic, and liver cancer [17 − 19], while others have not shown this effect, including prostate and breast cancer [20 − 21]. Recently, several studies have investigated the relationship between metformin use and endometrial cancer risk, with inconsistent findings. Some studies have shown that that metformin use may be associated with a reduced risk of endometrial cancer, whereas others have found no protective effect. Therefore, we carried out a meta-analysis of existing evidence to determine the effect of metformin on endometrial cancer risk.

MATERIALS AND METHODS
This meta-analysis was prepared in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) statement [22].

Eligibility criteria
After removing duplicate records, two reviewers screened the titles and abstracts independently. The potentially relevant studies were evaluated in detail to determine their eligibility. The inclusion criteria were as follows: (1) randomized controlled trials (RCTs) or non-randomized studies; (2) evaluated exposure to metformin and endometrial cancer risk; (3) reported risk ratio (RR) and a 95% confidence interval (CI) or provided data for their calculation. Articles were excluded if they were: (1) editorials, letters, reviews, and case reports; (2) studies without appropriate data for determining an estimate of RR and a 95% CI. In cases of multiple publications from the same population, only the most comprehensive studies were included.
Data extraction and quality assessment The following information was collected from each study: publication data (ie, first author's name, year of publication, and study location), study design, study population, type of comparison, sample size, definition of metformin exposure, RRs and 95% CIs, and control for potential confounders by matching or adjustments.
When multiple RR estimates were presented, the most adjusted estimate was extracted.
Two reviewers assessed the quality of included studies independently. The risk of bias in RCTs was evaluated according to the Cochrane Collaboration's tool [23]. The methodological quality of non-randomized studies was assessed according to the Newcastle-Ottawa scale [24]. Studies with a score of 7 or higher were considered as high quality studies.

Statistical analysis
Heterogeneity among study results was explored using the Chi-square (χ 2 , or Chi 2 ) and I 2 test. If substantial heterogeneity (p value < 0.10 or I 2 > 50%) was found, a random-effects model was used to calculate the pooled effect with appropriate cautious interpretation; otherwise, a fixed-effects model was applied. Heterogeneity was explored in subgroup analyses by study design, reference therapy, and adjustment for age and BMI or obesity. Sensitivity analysis was performed to assess the stability of the results. The analyses were performed using Stata version 12.0 software (Stata Corporation, College Station, TX).

Description of studies
A total of 744 records were screened of which 13 articles were assessed for eligibility. Of these, 4 studies were excluded for the following reasons: 1 used overlapping data [25], and 3 did not have available data [ (26 − 28]. No additional studies were identified from reference lists. Finally, 9 studies fulfilled the eligibility criteria and were included in the meta-analysis [29 − 37]. The flow diagram for study selection is showed in Fig. 1. The characteristics of the included 9 studies (1 RCT, 5 cohort studies, and 3 casecontrol studies) are described in Table 1. The studies were published between 2010 and 2017. Of these, 1 study was carried out in Europe and Australasia [29], 3 in USA [30,32,35], 2 in UK [31,33], 1 in Taiwan [34], 1 in Italy [36], and 1 in Canada [37]. These 9 studies included more than 1.20 million participants. The number of endometrial patients was 14 in the RCT, ranged from 71 to 2885 in cohort studies, and from 376 to 2,554 in case-control studies. All non-randomized studies [30 − 37] were controlled for potential confounders by matching or adjustments. Of the included 9 studies, 7 evaluated the association between metformin use and endometrial cancer risk in diabetic patients [29, 30, 32 − 34, 36, 37], while 2 in both diabetic and non-diabetic patients [31,35].
The risk of bias in the included RCT is shown in Fig. 2. The study [29] was considered at low risk of selection bias, attrition bias, and other bias, but unclear risk of detection bias, high risk of performance bias and reporting bias. The NOS scores calculated for non-randomized studies are shown in Table 1. Five studies were considered as high-quality studies (scores of 7 or higher) [30, 32 − 35].

Metformin use and endometrial cancer risk
In meta-analysis of all studies involving 7,762 cases of endometrial cancer, no evidence for an association between metformin use and endometrial cancer risk was observed (RR, 0.96; 95% CI, 0.80 to 1.16). The results showed significant heterogeneity among the studies (p = 0.000; I 2 = 75.5%) (Fig. 3).
After stratifying the data into subgroups based on study design, we found that metformin had no effect on endometrial cancer risk in RCT

Sensitivity analysis
In sensitivity analysis, each study was excluded sequentially and the influence was accessed by recalculating the pooled RR. The analysis confirmed the stability of our findings because the results did not change substantially after sequentially excluding each study.

DISCUSSION
With the high incidence of endometrial cancer and extremely poor prognosis associated with advanced-stage disease, it is important to identify potential chemopreventive agents. The present comprehensive meta-analysis included all available clinical studies, involving more than 1.20 million participants with 7762 cases of endometrial cancer. In this meta-analysis, metformin use was not associated with a substantially decreased or increased risk of endometrial cancer.
Similarly, for women with diabetes mellitus, metformin use did not affect endometrial cancer risk.
A climbing amount of laboratory studies have shown that metformin has direct and indirect anti-tumorigenic effects. Metformin may exert direct anti-tumor effects mainly by activating LKB1 dependent AMP-activated protein kinase (AMPK) [38,39].
AMPK activation leads to the regulation of multiple downstream signaling pathways, including down-regulation of PI3K/Akt/mTOR pathway and the block of lipid biosynthesis through inhibition of the acetyl co-carboxylase [ACC] [40 − 42]. In addition, metformin may indirectly affect tumor growth by reducing insulin resistance, lowering insulin and IGF-1 levels [12,43]. Recent laboratory studies have shown that metformin has several anti-tumorigenic effects in endometrial cancer cell lines, such as inhibits cell proliferation, induces apoptosis, attenuates invasion and metastasis, and reverses progestin resistance [44 − 47].
Apart from the results from laboratory studies, the present meta-analysis did not support the metformin have clinical chemopreventive potential against endometrial cancer. Luo et al. [32] contributed the difference to that metformin may not be a effective chemopreventive agent against endometrial cancer, but has a therapeutic role. Our findings are consistent with a meta-analysis of RCTs suggesting a neutral effect of metformin on the overall incidence of cancer [48]. A recently published meta-analysis found that metformin use was associated with a decreased risk of endometrial cancer among patients with diabetes (RR = 0.87, 95% CI, 0.80 to 0.95) [49]. However, there were only 5 studies included in this meta-analysis.
Compared with the existing study [49], our meta-analysis has several advantages.
We carried out a comprehensive search, including electronic databases, conference abstracts and clinical trial registers for available published and unpublished evidence with no language restrictions. Hence, the likelihood of important publication bias was small. As a result, we identified a much larger number of studies and patients, which may provide more precise results. Since metformin can have benefits and be prescribed for multiple indications in women, we evaluated the association between metformin use and endometrial cancer risk in both diabetic and non-diabetic patients, and assessed the association exclusively in diabetic patients.
In addition, besides study design, we performed additional subgroup analyses by comparisons and adjustments for age and BMI or obesity. As obesity is an established risk factor for endometrial cancer [50], it is important to adjust for the confounding effect of BMI when evaluating the association between metformin use and endometrial cancer risk. In subgroup analyses, our results were stable across different study designs and comparisons, and remained unchanged after adjusting for age and BMI or obesity.
Nevertheless, several limitations of the present meta-analysis should be acknowledged. First, most of the included studies were non-randomized studies.
Meta-analysis of RCTs allows for a more objective evaluation of the evidence because they are more likely to provide unbiased findings than other study designs.
However, non-randomized studies should be included in the meta-analysis when the available RCTs are limited. Second, there was significant heterogeneity among the studies, which might partly due to the diversity of study design, definition of drug exposure, and control for potential confounders. To minimize the biases in this meta-analysis, we extracted the most adjusted estimates to pool the effects, performed subgroup analyses and sensitivity analysis. Third, the limited data made it impractical to evaluate dose, duration effects of metformin.
In conclusion, this comprehensive meta-analysis does not support a chemopreventive effect of metformin against endometrial cancer at the population level. However, the results should be interpreted with caution given the limited data and possibility of residual confounding. Therefore, further randomized chemoprevention trials are needed to validate our findings.    Figure 1 Study flow diagram.