Introduction

Prostate cancer (PCa) is the most frequently diagnosed cancer among men in over half of all countries worldwide and is the second leading cause of cancer deaths in men, after lung cancer [1,2,3,4,5]. PCa remains the third most prevalent cancer globally, with over 1.4 million new cases and 370,000 fatalities reported in 2020 [6]. On the other hand, apart from skin cancer, PCa is the most common cancer among men in the Western world [7]. Frequent urination, urinary weakness, urinary incontinence, blood in the urine, burning and persistent pain in the lower back, and abdominal pain are also clinical symptoms of PCa [8]. Various factors such as age, race, genetic factors, environmental factors, and family history play an important role in the progression of PCa [9, 10]. More than 670,000 PCa patients are diagnosed each year. Of these, 225,000 are in Europe and 240,000 in the United States [11]. The incidence of PCa varies between races. For example, 4–7 per 100,000 in Asian countries and 70–100 per 100,000 in European and North American countries [12, 13]. Metformin has multiple mechanisms for reducing cancer and carcinogenesis: direct action (on tumors and the microenvironment) and indirect action (on hosts that may affect tumors). Metformin is generally connected directly or indirectly through the AKT-Mtor pathway [14,15,16,17]. Mechanisms of pathway activation most commonly associated with PCa include deletion of inhibitory PTEN [18], PI3K mutations [19], or activation of growth factor receptors such as insulin [20,21,22]. The ability of metformin to reduce hyperinsulinemia may also indirectly reduce the risk of PCa [23,24,25]. In addition, laboratory evidence suggests that hyperinsulinemia regulates insulin receptors in PCa cells and promotes tumor growth [26, 27]. Reducing insulin levels in the blood stream or direct activation of AMP kinase. In this study, we used systematic evaluation and meta-analysis to investigate the relationship between metformin and PCa risk.

Methods

Study design

This study has been registered (registration number: CRD42023447013) with the PROSPERO database before July 22, 2023 (https://www.crd.york.ac.uk/prospero/display_record.php?RecordID=447013). We used the Cochrane Handbook for Systematic Reviews of Interventions for the preparation and conduct of this meta-analysis [28]. We reported this meta-analysis with reference to Preferred Reporting Items for Systematic Review and Meta-Analyses (PRISMA) guidelines [29].

Search strategy

The literature search was completed before July 22, 2023 for relevant available articles from the following databases: (1) PubMed; (2) Ovid MEDLINE; (3) Scopus; (4) Embase; (5) Cochrane library; (6) Web of Science; (7) Sinomed (CBM); (8) China National Knowledge Infrastructure (CNKI); (9) Wanfang Data Knowledge Service Platform; (10) China Science and Technology Journal VIP Database. The registration search was completed by 22 July 2023 and the relevant data retrieved were from the following registration pools: (1) ClinicalTrials.gov; (2) International Clinical Trials Registry Platform (ICTRP); (3) The EU Clinical Trials Register; (4) Chinese Clinical Trial Registry. The relevant retrieval strategy was as following: (“Metformin” or “ Dimethylbiguanidine” or “ Dimethylguanylguanidine” or “ Glucophage”) and (“Prostatic Neoplasms” or “ Prostate Neoplasms” or “ Prostate Cancer” or “Cancer of Prostate” or “ Prostatic Cancers”). Relevant Chinese technical terms for the Chinese databases were used to search for published articles. Furthermore, references of all relevant articles and reviews were retrieved to search for additional eligible studies.

Inclusion and exclusion criteria

Inclusion criteria

This meta-analysis included studies based on the following criteria: (1) participants with no PCa history were selected for the incidence analysis, while those with a PCa history were chosen for recurrence and mortality analyses; (2) metformin was the exposure factor; (3) studies provided relative risks (RRs), odds ratios (ORs), or hazard ratios (HRs) along with 95% confidence intervals (CIs) or data enabling their calculation; (4) in cases of multiple publications from the same population, the study with the larger sample size or more comprehensive data was chosen; and (5) studies were assessed for quality using the Newcastle Ottawa scale (NOS), requiring a score of at least 6 stars.

Exclusion criteria

(1) Antidiabetic drugs which did not include metformin; (2) comments or letters to the editor, case reports, and abstract-only publications; (3) Preprint servers, such as medRxiv/bioRxiv, etc.

Data extraction

After removing duplicates, two reviewers (Y Liu and Q Zhang) independently screened all abstracts and titles to exclude irrelevant articles. Full texts of potentially relevant studies were then downloaded and reviewed, with those meeting the selection criteria included in this systematic review. Two independent investigators (Y Liu and Q Zhang) extracted data from the included articles. The extracted data comprised the first author’s name, year of publication, study location, study methods, sample size, metformin usage, primary outcomes, raw data of patient numbers in the trial (metformin) and control groups, and adjusted RRs/ORs/HRs with corresponding 95% CIs.

Quality assessment

Two investigators (Y Liu and Q Zhang) independently evaluated the methodological quality of the included case-control and cohort studies using the nine-star NOS [30]. The assessment considered eight items across four categories: selection of cohort studies, comparability, outcomes, or exposure for case-control studies. Studies were rated as low-, moderate-, or high-quality based on their NOS scores (0–3, 4–6, 7–9, respectively). The certainty of evidence was determined using the Grading of Recommendations Assessment, Development and Evaluation (GRADE) framework [31].

Statistical analysis

Statistical analyses were conducted using Stata (version 16.0; Stata Corp, College Station, TX) and RevMan (version 5.3; Cochrane Library) software. The pooled RR with 95% CI was calculated from the extracted raw data to assess the association between metformin use and PCa risk. When multiple RRs were available, the effect value controlling for the most confounding factors was chosen. Subgroup analyses were performed to explore the relationship between metformin use and PCa incidence and mortality, considering study region, study design, dosage type, and diagnosis type. HRs were directly considered as RRs [32, 33], and ORs were converted to RRs using the formula: RR = OR/((1–P0) + (P0 × OR)), where P0 represents the incidence of the outcome in the non-exposed group [34]. The standard error (SE) of the converted RR was calculated using SElog(RR) = SElog(OR) × log(RR)/log(OR). This formula was also applied to determine the upper and lower confidence limits of the CI based on the adjusted odds ratio [35]. To generalize our study results beyond the included studies, a random-effects model was used as it is the most suitable for meta-analysis [36]. Studies reporting data on PC incidence in terms of person-years, number of cases, and metformin dose or duration were included in the dose-response analysis. Restricted cubic splines with three knots at the 10%, 50%, and 90% percentiles of the distribution were used for both linear and non-linear dose-response analyses [37]. The study-specific estimates were then combined using the restricted maximum likelihood method in a multivariate random-effects meta-analysis. Sensitivity analysis was conducted to determine if any single study significantly influenced the results [38]. Publication bias was evaluated qualitatively with funnel plots and quantitatively using Begg’s and Egger’s tests [39]. A P value of less than 0.05 was considered statistically significant in all analyses.

Results

Study selection and characteristics

A search of databases and registries identified 3950 database records and 82 registry records, and 415 potentially eligible studies were selected after removing duplicate information and screening titles and abstracts. Of the 415 potentially eligible studies, a total of 41 studies met the inclusion criteria, including 34 studies of PCa incidence [15, 40,41,42,43,44,45,46,47,48,49,50,51,52,53,54,55,56,57,58,59,60,61,62,63,64,65,66,67,68,69,70,71,72], 5 studies of PCa recurrence [73,74,75,76,77], and 5 studies of PCa mortality [52, 55, 75, 78, 79]. Google Scholar and Baidu Scholar were also searched, with 682 records being identified as potentially relevant to this study. However, these records were excluded as they were duplicates to the studies in the databases and the registries. The detailed process of literature screening is shown in Fig. 1.

Fig. 1: PRISMA flow diagram of search strategy.
figure 1

Flow diagram of the study search and selection process.

Full size image

Finally, forty-one studies with a total of 3,933,414 subjects were included in this meta-analysis. The characteristics of the included studies are shown in Tables 1 and 2.

Table 1 General characteristics of eligible case-control studies.
Full size table
Table 2 General characteristics of eligible cohort studies.
Full size table

Overall meta-analysis of metformin use on PCa risk

The results showed that metformin use was associated with a lower incidence of PCa (RR: 0.82, 95% CI: 0.74–0.91, P < 0.001, I2 = 97%, Fig. 2), and the random effects model was adopted. Meanwhile, metformin use was found to be associated with reduced recurrence (RR: 0.97, 95% CI: 0.81–1.15, P = 0.71, I2 = 0%, Fig. 3) and mortality (RR: 0.94, 95% CI: 0.81–1.09, P = 0.42, I2 = 75%, Fig. 4) in PCa with a random effects model, but the results were not statistically significant.

Fig. 2: Metformin use on PCa risk of incidence.
figure 2

Results of a meta-analysis of metformin use on PCa risk of incidence.

Full size image
Fig. 3: Metformin use on PCa risk of recurrence.
figure 3

Results of a meta-analysis of metformin use on PCa risk of recurrence.

Full size image
Fig. 4: Metformin use on PCa risk of mortality.
figure 4

Results of a meta-analysis of metformin use on PCa risk of mortality.

Full size image

Subgroup analysis of metformin use on PCa risk

The results of subgroup analyses based on the study design showed that metformin administration was associated with a reduced incidence of PCa in the cohort study subgroup (RR: 0.81, 95% CI: 0.73–0.90, P < 0.001, I2 = 97%), and that metformin administration did not increase the risk of recurrence (RR: 0.93, 95% CI: 0.81–1.18, P = 0.77, I2 = 0%) and mortality (RR: 0.94, 95% CI: 0.81–1.09, P = 0.42, I2 = 75%,) from PCa. Meanwhile, the results of the case-control study suggested that metformin was not associated with either PCa incidence (RR: 0.90, 95% CI: 0.78–0.91, P = 0.16, I2 = 81%).

Subgroup analysis by study area showed that metformin was found to be associated with a reduced incidence of PCa in Asia (RR: 0.67, 95% CI: 0.56–0.79, P < 0.001, I2 = 96%) and Europe (RR: 0.89, 95% CI: 0.81–0.97, P = 0.01, I2 = 83%). In the North American study, metformin use was also found to reduce the risk of PCa, but the difference was not statistically significant (RR: 0.83, 95% CI: 0.66–1.05, P = 0.11, I2 = 97%). Meanwhile, in the North American study, metformin use was found to reduce the risk of mortality from PCa, but the difference was not statistically significant (RR: 0.91, 95% CI: 0.77–1.07, P = 0.24, I2 = 78%).

The results of all subgroup analyses are shown in Table 3.

Table 3 Subgroup analysis of metformin use on PCa risk.
Full size table

Dose-response meta-analysis

Three studies with a total of 515,615 participants were included in a dose-response analysis of incidence [49, 66, 69]. Among these studies, metformin exposure was expressed as duration of exposure.

Linear dose-response models showed a significant negative association between duration of metformin exposure and risk of PCa (exb(b): 0.980, P < 0.001). Furthermore, nonlinear dose-response analysis showed a similar association (Coef1 = –0.299, P1 < 0.001, Coef2 = 0.325, P2 < 0.001). Each 1-year increment in metformin exposure was associated with a 22% reduction in the risk of PCa (RR: 0.78, 95% CI: 0.77–0.79, p < 0.001, Fig. 5).

Fig. 5: Dose-response analysis of duration of metformin use and risk of PCa.
figure 5

Dose-response analysis of restricted cubic splines in multivariate random-effects dose-response models for the relationship between duration of metformin use and risk of PCa.

Full size image

Study quality assessment and risk of bias

All included observational studies were considered to be above moderate quality studies, as depicted by NOS ≥ 6. The NOS-based assessment indicated a low to moderate risk of bias, while the GRADE assessment revealed low certainty in the evidence supporting metformin’s ability to reduce PCa incidence, recurrence and mortality. This is mainly due to the retrospective nature of the studies and potential selection and publication biases (See Supplementary Information: Fig. S1). In addition, Begg’s test and Egger’s test found publication bias in some studies. In this study, for PCa incidence, P value of Begg’s test was 0.047, and P value of Egger’s test was 0.004. For PCa recurrence, P value of Begg’s test was 0.806, and P value of Egger’s test was 0.182. For PCa mortality, P value of Begg’s test was 0.825, and P value of Egger’s test was 0.573. Some of the funnel plots were shown to be unsymmetrical (Fig. S2).

Sensitivity analysis

For incidence, recurrence, and mortality, conclusions were unchanged after excluding individual papers and calculating heterogeneity and effect sizes (Fig. S3).

Discussion

In this study, we synthesized evidence from cohort and case-control studies involving 3,933,414 participants in 15 countries and regions. Our findings suggest a potential chemoprotective effect of metformin on PCa incidence, recurrence, and mortality. We reported the effect of metformin use on PCa risk in different types of studies and regions. In addition, our study explored a possible dose-response relationship between metformin use and PCa incidence based on exposure time. Finally, the main findings of this study support the mechanistic hypothesis that metformin use is negatively associated with the risk of PCa incidence, recurrence, and mortality.

The relationship between metformin use and cancer has been widely debated. Several studies have explored metformin’s potential chemopreventive effects on various tumors, including breast [80], brain [81], and melanoma [82]. This meta-analysis, with a larger participant pool than previous ones [83, 84], further substantiates metformin’s protective effect against PCa.

Several studies suggest that metformin’s antitumor effects may involve multiple mechanisms. It has been reported as an indirect activator of AMP-activated protein kinase (AMPK), inhibiting the growth of PCa cells [85] and being selectively toxic to p53-deficient tumor cells [86]. However, metformin also inhibits the proliferation of most breast cancer cells, regardless of p53 status [87]. Ben Sahra et al. [16] showed that metformin decreases the level of cell cycle protein D1, exerting antitumor effects both in vivo and ex vivo. Huang et al. [18] demonstrated that metformin inhibition might trigger a signaling pathway that effectively inhibits cellular growth.

Existing studies have demonstrated that insulin and insulin-like growth factors are crucial in regulating cellular energy and growth. These hormones and their associated signaling networks significantly contribute to tumor formation [88]. Epidemiological studies have found that insulin-like growth factor-1 (IGF-1) promotes the proliferation of various cancer cells, including breast and prostate cancers. Additionally, it has been demonstrated that insulin can regulate the activity of the IGF-1 receptor [89]. IGF-1 is present in normal cells, but in cells with malignant growth characteristics, it exerts strong mitogenic and anti-apoptotic effects. Thus, the level of IGF-1 in the human body influences tumorigenesis. Tumor cells can produce IGF-1 through autocrine or paracrine secretion, promoting their differentiation and proliferation. When IGF-1 binds to its receptor, it initiates the mitogen-activated protein kinase 2 signaling pathway and the phosphatidylinositol 3-kinase/Akt signaling pathway, which, when activated, promotes cell proliferation and inhibits apoptosis in tumor cells [90]. The mammalian target of rapamycin (mTOR) is a serine/threonine kinase of the phosphatidylinositol 3-kinase-related enzyme family, regulated by intracellular and extracellular signals, including nutrients like glucose and amino acids, as well as growth factors such as insulin and insulin-like growth factors. These factors regulate cell growth. When metformin is used, it reduces glucose and insulin levels in the body, thereby affecting mTOR activity and inhibiting cell growth.

Additionally, confounding mechanisms might explain the link between metformin use and reduced PCa incidence. Type 2 diabetes is a known risk factor for PCa [91]. Metformin users are often obese and have type 2 diabetes, both conditions associated with a lower risk of PCa [92, 93]. Metformin, a common antidiabetic drug, helps control hyperglycemia in type 2 diabetes patients by affecting mitochondrial respiration, leading to energy deficiency and molecular changes [94, 95]. One proposed mechanism for metformin’s antitumor effects is the inhibition of mitochondrial respiratory complex I. This inhibition reduces ATP production, activating AMPK in an LKB1-dependent manner, which then inhibits mTOR, leading to anticancer effects [95]. Additionally, new diabetes treatments, such as GLP-1 inhibitors, have been found to inhibit PCa growth, reducing the risk of PCa [96]. We hypothesize that treating diabetes can further lower the incidence of PCa.

In addition to the mechanisms described above, this may be due to lower testosterone levels in diabetic men than in non-diabetic men [97].

Based on the potential antitumor mechanism of metformin, numerous reports have explored its relationship with PCa risk, but the quality and findings of these studies vary. This article reviews and analyzes 41 studies, encompassing 3,933,414 participants, and finds that metformin use reduces the risk of PCa recurrence and death. However, the results show no significant difference and exhibit high heterogeneity. This variability may stem from differences in study design, such as drug use in control groups, drug combinations, sequential drug use, dosages, follow-up periods, control of confounding factors, duration of drug use, and variations in study populations’ age, occupation, ethnicity, and geographic area. In the subgroup analyses of this study, metformin administration was found to reduce the risk of PCa in Asian and European populations. However, no significant correlation was observed between metformin use and PCa incidence, recurrence, and mortality in North American populations. This may be due to the limited number of studies conducted in North America and the focus of current research on Asia and Europe. Additionally, significant genetic differences and susceptibility loci for PCa between Asian, European, and American populations may influence metformin’s effectiveness in preventing PCa [98,99,100].

Although this meta-analysis showed a potential benefit of metformin for PCa treatment and a better risk-benefit ratio, this study has several limitations. First, there are limitations in that the inclusion of so many retrospective studies does not lead to a reasonable and unbiased conclusion and is prone to bias. Second, in studies examining the association between metformin and PCa recurrence and mortality, there was a trend toward lower risk, but it did not reach statistical significance. Further randomized controlled trials and real-world studies are needed to explore potential dose-response relationships. Third, subgroup analyses of PCa types were not performed in this study; therefore, it was not possible to examine the effect of metformin on different types of PCa. Finally, although some confounders were corrected for in the analysis, there is no guarantee that all potential confounders were considered. Other unreported and unanalyzed confounders may have been present in the original study. However, future randomized, double-blind controlled trials with adequate sample sizes and validated study protocols are still needed to assess and confirm the potential benefits of metformin for PCa prevention and to determine the optimal dose of metformin with a favorable risk-benefit ratio.

Conclusions

This meta-analysis showed that metformin use was independently associated with a reduction in PCa incidence. A duration-dependent relationship was found between metformin and PCa incidence, suggesting that prolonged metformin use is associated with a lower risk of developing PCa. Meanwhile, this study may provide guidance to clinicians to improve the prognosis of PCa patients. In the future, larger prospective cohort studies or even randomized controlled as well as longer follow-up trials are needed to confirm the relationship between metformin use and PCa.