The rising incidence of cancer is considered a major threat to population health globally; it is estimated that by 2020, about 16 million cancer patients will be diagnosed as new cases each year [1
]. Advances in knowledge for appropriate early detection and treatment with follow-up care, and also the identification of specific cancer biomarkers, may be an effective approach to reducing the burden of cancer. Recently, a number of measurable biomolecules have been defined as biomarkers in cancer, including cyclin D-1, Ki-67 antigen, breast cancer 1 or 2 (BRCA1 or BRCA2), alpha-fetoprotein (AFP), human epidermal growth factor receptor-type2-new (HER2-new) protooncogene, p53 gene, human epididymis protein 4 (HE4
), and mucin 1 (MUC1) to detect cancer [2
]. The focus on combining therapeutics with cancer diagnostics and related biomarkers may play an effective role in the development of cancer medicine.
Metformin, a biguanide anti-diabetic, is usually considered the drug of choice for diabetics. Over the last decade, the extensive use of metformin in human studies has indicated that metformin is useful in reducing mortality and morbidity in diabetic patients caused by cancers such as breast, endometrial, ovarian, prostate, liver, pancreatic, lung, medullary thyroid, gastric, and colon cancer [3
]. The metformin impact on cancer cell proliferation has been related to adenosine mono-phosphate-activated protein kinase (AMPK) activation, decreased mTOR signaling, and protein synthesis. Metformin could potentiate the antitumor effect of several MEK inhibitors in lung cancer cells [5
Research (in-vitro) has indicated that metformin inhibits the growth of various human cancer cell lines by controlling the PI3K-AKT-mTOR signaling pathway as well as glucose metabolism [7
]. Metformin can induce activation of adenosine mono-phosphate-activated protein kinase (AMPK), which leads to suppression of the mTOR pathway [9
]. However, due to the use of high doses of metformin in humans, the results of in-vitro studies cannot be attributable to humans and therefore these data do not confirm the beneficial impact of metformin on all types of cancer cells for clinical purposes. Several human studies pointed to the effects of metformin on tumor growth and showed a decrease in the expression of Ki-67 (a biomarker of proliferation) in patients with breast [10
], prostate [12
], and endometrial cancer [13
]. Additionally, metformin regulated some indicators of cancer with poor prognoses such as blood glucose and insulin levels [16
]. There are several debates about the findings of human research on the impact of metformin in cancer patients, as most of them did not compare the metformin group with the placebo or non-drug group.
The aim of the current review was to present the results of clinical trials on the influence of metformin on circulating cancer-related biomarkers and metabolic factors including insulin, fasting blood sugar (FBS), homeostatic model assessment for insulin resistance (HOMA-IR), body mass index (BMI), and Ki-67.
This systematic study was conducted in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analysis (PRISMA) recommendations.
2.1. Search Strategy
A systematic search of the literature was conducted until February 2020 in the following databases: Ovid-Medline, Scopus, PubMed, Web of Science, and Cochrane. Two groups of search terms were used, which were boolean (AND or OR) without language restrictions. The keywords were as follows:
Metformin (metformin, dimethylbiguanidine, dimethylguanylguanidine, glucophage, metformin hydrochloride, and metformin HCl);
Cancer and metabolic biomarkers.
The population, intervention, comparison, and outcome (PICO) framework in the present study includes the following information:
P (population): non-diabetic cancer patients;
I (intervention): metformin;
C (comparison): placebo or no drugs;
O (outcome): cancer and metabolic biomarkers.
2.2. Inclusion Criteria and Selection Study
Two collaborators (T.F. and H.A.) reviewed the abstract to determine the quality of the studies for the selection of all papers. The abstracts and titles collected from the databases were saved in Endnote, Reference Manager X9, and eliminating duplicates. Two authors then removed unrelated studies by title, abstract, and full text, respectively. Any inconsistencies in the selected studies were discussed and evaluated with third co-workers.
The studies were selected according to the criteria listed below:
The research was based on clinical trial;
Non-diabetic patients with breast or endometrial cancers were selected as participants for the intervention and control groups;
Metformin was administrated for the intervention group;
The endpoint of each study was cancer biomarkers;
The language of articles had to be English.
2.3. Exclusion Criteria
Studies with unclear methodology and unclear results were not considered;
Duplicate studies were excluded;
Animals, cohort, case-report, case-control, letters, erratum, and conference papers were excluded.
2.4. Data Extraction
Four authors (T.F., H.A., B.R., and F.S.) extracted following data after reading the full text of the selected articles: first author, year of publication, country where RCT was conducted, participants (cancer type, sample size, age, and gender), metformin dose, placebo dose, duration of study, and mean ± standard deviation (SD) or mean ± standard error of mean (SEM) of biomarkers at baseline and after intervention for each study. Discrepancies with the selected studies were clarified through discussion with a third-fourth author.
2.5. Quality Assessment
The quality of the studies was independently assessed by three authors (T.F., B.R., and H.A.) using the “Version 2 of the Cochrane risk-of-bias tool for randomized trials (RoB 2)”. Six risks of bias items were individually assessed in the RCTs by answering each question that determines the high (H), unknown (U) or low risk (L) of bias. Performance bias was determined by two domains including random housing and blinding. Detection bias were done according to the random outcome assessment and blinding outcome assessment. Attrition bias was revealed according to the incomplete outcome data. Reporting bias considered how selective outcome reporting was and what was found. Other sources of bias in result and method sections consisted of any important concerns about bias not covered by other domains in the tool. High risk of bias indicates low quality of articles. Low risk of bias indicates high quality of articles. Unclear risk of bias indicates that the limited information restricted correct judgment to identify low or high quality of the article.
2.6. Statistical Analysis
The heterogeneity of the selected articles was evaluated using the “Chi-square test”. The I2 statistic was used to examine the total variation percentage between articles. To estimate the pooled effect, a model with random-effects was used when I2 was higher than 50% and a model with fixed-effects model if I2 was lower than 50%. The pooled effect of metformin on cancer biomarkers was indicated as effect size, odds ratio (OR), and standardized mean differences (SMD) with 95% confidence intervals (CI). Overall results showed a significant difference between the two groups when the level of p-value was less than 0.05 (p < 0.05). The Egger regression test was used to determine the publication bias. A sensitivity analysis was performed to determine the effect of each included study on the size of the pooled effect. The meta-analysis in this study was performed with “STATA 13 software” (R3.3.1, University of Auckland, Auckland, New Zealand).
The present study aimed to demonstrate the scientific evidence on the impact of metformin on biomarkers that could be linked to the occurrence of cancer. There was evidence from nine RCTs investigating the impact of metformin in patients with breast and endometrial and non-diabetic cancer. Systematic findings on the nine publications indicated metformin decreased insulin levels in three studies, FBS in one, BMI in three, Ki-67 in three, and HOMA-IR in one.
Our pooled analysis findings showed that metformin therapy did not have a significant effect on measured biomarkers such as insulin, glucose, HOMA-IR, or BMI. We also found that metformin could decrease Ki-67 in patients with operable endometrial cancer. However, due to the high heterogeneity of the included studies, the data from the current study should be interpreted with caution.
It is important to consider that the high heterogeneity among studies evaluating the effect of metformin on insulin, glucose, HOMA-IR, BMI, and Ki-67 affected our data. The RCTs included varied in study population characteristics such as overweight/obese patients, not-overweight/obese patients, patients with two cancer, and patients with stage 1–3 cancer, as well as in sample size, duration and dose of metformin treatment, and adjustment for confounders.
Our findings are similar to data from a meta-analysis of observational studies conducted by Tang et al. (2018), which examined the impact of metformin on breast cancer incidence and mortality [26
]. They found no association between metformin therapy in diabetic breast cancer patients and interpreted that the bias of observational studies prevented a safe conclusion. They suggested that it was necessary to design RCTs to prove the exact association between metformin and breast cancer.
Three of the RCTs included in the current meta-analysis [10
] did not report a significant reduction effect of metformin on Ki-67 in non-diabetic breast cancer patients compared to their controls. This is in contrast to the results of a meta-analysis conducted by Rahmani et al. (2019) on nine RCTs [27
]. They reported that metformin decreased insulin, glucose, leptin, C-reactive protein (CRP), HOMA-IR, and Ki-67 levels, which attenuated the outcome in breast cancer patients. The difference between our findings and the study by Rahmani et al. (2019) is related to another article included in two meta-analyses. In two included articles in the studies of Rahmani et al. (2019), those of Laskov et al. (2014), and Sivalingam et al. (2016)), it was reported that metformin decreased circulating insulin, glucose, and Ki-67 levels in cancer patients who did not meet our inclusion criteria such as English language and RCTs restriction. Due to the limited number of available studies on the effect of metformin on the Ki-67 biomarker in the endometrial cancer patient, no subgroup analysis was performed.
Ki-67 is a deterministic prognostic biomarker for the detection of breast cancer [28
], but there are few documents on its use as a diagnostic biomarker for endometrial cancer. It was found that high-stage cancer had higher levels of Ki-67. Several investigations reported an association between high levels of Ki-67, insulin, and glucose, and a poor prognosis in endometrial cancer [31
]. In two RCTs [13
], metformin was found to significantly decrease the expression of Ki-67 in patients with endometrial cancer. Metformin changed in phosphorylated mTOR proteins and serum markers of insulin resistance. It was suggested that metformin could decrease Ki-67 expression only in patients with insulin resistance.
In cancerous cells, the levels of blood circulating insulin and IGF increases, thereby the growth of these cells is induced. Metformin can decrease hepatic gluconeogenesis and lipogenesis and also elevate fatty acid oxidation, insulin sensitivity, and inhibiting gluconeogenesis. This ameliorates blood glucose control, reducing food consumption and intestinal glucose absorption [28
]. However, the mechanisms related to its effect on cancer biomarkers have not fully understood. Metformin could decrease Ki-67 in patients with operable endometrial cancer prior to surgery and increase apoptosis, leading to inhibit cancer progression. For breast cancer patients, it seems that metformin may be effective in early stage. However, several factors such as genetic polymorphisms may cause various response to metformin in cancer patients. It is necessary to find target populations for its use.
Our finding did not confirm or reject the protective effect of metformin in woman with endometrial and breast cancer, as most RCTs in our meta-analysis did not consider insulin resistance as the most important confounding factor.
Our meta-analysis study has one advantage, including the use of RCTs to assess the association between the metformin therapy and biomarkers changes.
However, this study has several limitations. First, high heterogeneity of the articles and the bias of the publications affected our data. Second, all included studies had a small sample size and poor quality. Third, the small number of selected studies limited us to performing the subgroup analysis for duration and dose of intervention of metformin effects on cancer-related biomarkers and also age of subjects. Fourth, we could not find sources of heterogeneity between included publications in this meta-analysis.