The safety and efficacy of metformin on metabolic syndrome parameters in prostate cancer patients receiving androgen deprivation therapy: A systematic review and meta-analysis

ABSTRACT

Background

Prostate cancer patients are more likely to develop metabolic syndrome because androgen deprivation therapy (ADT), a standard treatment for advanced prostate cancer, causes major metabolic syndrome components and associated disorders. Metformin, a low-cost oral antidiabetic with a favorable safety profile, may be useful for preventing metabolic issues associated with ADT. This meta-analysis aims to assess the efficacy and safety of metformin on metabolic syndrome parameters in prostate cancer patients receiving ADT.

Methods

In November 2025, we conducted a systematic search of PubMed, Scopus, and the Cochrane Library for randomized and non-randomized controlled trials. Studies were considered eligible if they included adult men (≥18 years) with prostate cancer receiving ADT and assessed metformin versus placebo/standard care on metabolic outcomes. For the meta-analysis, we used R 4.5.0 with R Studio 2024.12.1+563 using a random-effects mode.

Results

We included ten trials with a total sample size of 2,642 patients. There were no significant differences in HbA1c (MD = -0.01, 95% CI -0.1; 0.2, P = 0.8), body weight (MD = 0.42, 95% CI -3.31; 4.16, P = 0.8), body mass index (MD = -0.01, 95% CI -0.1; 0.2, P = 0.8), or waist circumference (MD = -1.67, 95% CI -4.6; 1.26, P = 0.26). There was no significant difference in blood pressure, fasting glucose, triglycerides, or HDL cholesterol levels. Metformin was well tolerated but was related with a greater risk of diarrhea (RR = 2.55, 95% CI 1.02; 6.39). The Trial sequential analysis (TSA) stated that the current evidence is insufficient and underpowered.

Conclusion

According to our data, metformin did not provide a substantial improvement in metabolic syndrome indices for prostate cancer patients on ADT. Its efficacy as a stand-alone remedy appears to be limited. Future large-scale clinical trials should investigate the role of metformin in combination with structured lifestyle changes.

Introduction

The World Health Organization (WHO) states that prostate cancer is the third most common diagnosed cancer worldwide [1]. Prostate cancer patients are more likely to experience metabolic syndrome and its associated conditions [2]. Combination therapy can enhance results for cases with metastatic hormone-sensitive prostate cancer [3]. The standard treatment for metastatic hormone-sensitive prostate cancer is androgen deprivation therapy (ADT) plus an androgen receptor pathway inhibitor (ARPI) with or without docetaxel [3]. However, ADT monotherapy is frequently prescribed in older and frail patients. While ADT and ADT combination therapy are effective, they can lead to severe metabolic side effects such as weight gain, muscle or bone mass loss, diabetes, metabolic syndrome, cardiovascular diseases, and mortality [4–6].

Elevated insulin levels and body composition can increase mortality from prostate cancer-related and non-cancer diseases such as diabetes and cardiovascular disease [7,8], reducing the effectiveness of ADT for PCa. After one year of ADT, weight increase ranged from 1.8 to 3.8%, with fat mass ranging from 9.4 to 23.8% [9,10]. Incidence of prostate cancer is higher in older patients and those with pre-existing diseases including cardiovascular disease, diabetes, and renal impairment [11,12]. Furthermore, ARPIs are expensive and thus limiting their use in low- and middle-income nations. In these countries, the number of people diagnosed with advanced prostate cancer is predicted to rise significantly [13]. Cost-effective and safe therapeutic options for common diseases, like prostate cancer, are needed.

Metformin, a widely known antidiabetic agent, has been reported to improve survival for patients with diabetes or impaired fasting glucose while controlling hyperglycemia, weight gain, and macrovascular adverse events [14,15]. Furthermore, metformin has been shown in preclinical research to have both direct and indirect anticancer effects [16]. It has a low-cost and favorable safety profile that may be used to prevent metabolic problems of ADT while also improving prostate cancer outcomes.

The effectiveness of metformin in treating prostate cancer has been investigated for its potential in mitigating metabolic syndrome in this group. Previous studies on metformin’s dual role in prostate cancer patients receiving ADT has revealed benefits in cancer-specific outcomes and short-term metabolic outcome changes [17–19]. However, these were small-scale studies and conducted in a variety of settings, including monotherapy and combination therapy.

The STAMPEDE trial was a phase 3 clinical trial that included patients with locally advanced or metastatic cancer [20]. This trial found no meaningful evidence of an overall survival advantage from adding metformin to standard of therapy in patients with metastatic hormone-sensitive prostate cancer. However, the metformin group reported considerably fewer unfavorable metabolic side effects with ADT than the standard care.

Based on these findings, pooled analysis of studies assessing the role of metformin in prostate cancer patients receiving ADT is warranted. Therefore, this meta-analysis aims to assess the safety and efficacy of metformin on metabolic syndrome parameters associated with ADT in prostate cancer patients.

Methods

Study registration

The study protocol was registered on PROSPERO (CRD420251241140). This systematic review adhered to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) statement standards [21] and is conducted in accordance with the guidelines of the Cochrane handbook for systematic reviews [22]. PRISMA checklist is provided in supplementary appendix.

Literature search and study selection

In November, 2025, we conducted a thorough search of the PubMed, Cochrane Library, and Scopus databases using the term (prostate cancer OR prostatic neoplasm) AND (androgen deprivation therapy OR ADT OR GnRH agonist OR castration) AND (metformin). Detailed search strategies are displayed in Table S-1.

To determine which papers satisfied the inclusion and exclusion criteria, two distinct authors independently evaluated the search results based on title and abstract of each study. We obtained full texts of potentially pertinent papers for additional evaluation after our first screening of abstracts and titles. In order to identify other potential studies that met our inclusion criteria, we manually searched for relevant articles. Throughout the screening process, disagreements were settled by consensus.

Eligibility criteria

We included studies that met the following PICO criteria:

• Population: Adult men (≥18 years) diagnosed with prostate cancer at any stage who were undergoing androgen deprivation therapy (ADT), either as monotherapy or combination therapy.

• Intervention: Oral metformin administered at any dose and for a minimum intervention period of 3 months.

• Control: Placebo or standard of care without metformin.

• Outcomes: At least one of the following outcomes: HbA1c, body weight, body mass index (BMI), waist circumference, fasting blood glucose, fasting triglycerides, high-density lipoprotein (HDL) cholesterol, low-density lipoprotein (LDL) cholesterol, total cholesterol, systolic and diastolic blood pressure, and adverse events.

• Study design: Randomized and non-randomized controlled trials were included.

Observational cohorts, case reports, reviews, and editorials were excluded. Also, we excluded studies that did not meet any other part of our inclusion criteria.

Data extraction

Data retrieved from included studies were gathered by two authors using an excel spreadsheet. We extracted data on the study’s design, blinding, country, sample size, inclusion and exclusion criteria, intervention, control, and follow-up duration. We also gathered baseline demographic and clinical information, including age, BMI, waist circumference, HbA1c, diabetes status, Gleason score, cancer stage, baseline PSA, prior ADT exposure, and ADT type. Regarding the extracted primary outcomes, these included HbA1c, body weight, waist circumference, systolic and diastolic blood pressure, fasting blood glucose, fasting triglycerides, HDL cholesterol, LDL cholesterol, and total cholesterol at multiple follow-up time points, in addition to adverse events.

Quality assessment

Two authors used the Cochrane tool for RCTs (ROB2) [23] to evaluate the quality of the included studies. We employed ROBINS-I (‘Risk Of Bias In Non-randomized Studies – of Interventions’) for evaluating risk of bias in non-randomized trials [24]. The Robvis web tool was used to generate quality assessment Figure 2 [25]. Any disagreements were settled through consensus.

Figure 2.

Summary and traffic-light graph of risk of bias of RCTs.

Statistical analysis

We utilized R 4.5.0 [26] with R Studio 2024.12.1 + 563 [27] to conduct our analysis. We used risk ratios (RRs) and 95% confidence intervals for dichotomous data while continuous data were analyzed using mean difference (MD) and 95% CI. In addition to I-squared (I2) and chi-squared (Chi2) statistics, the forest plot was visually examined to assess statistical heterogeneity between trials. I2 values of 50% showed high heterogeneity. All findings were examined using a random-effects model. We employed the Egger test [28] and funnel plot to examine publication bias and small-study effects where there were ten or more studies; otherwise, we used the DOI plot and LFK index [29]. We utilized the trim-and-fill method to estimate the number of missing studies and update the pooled estimates [30]. In order to formally identify significant outliers, we used Baujat plots [31].

Trial Sequential Analysis (TSA) was used to reduce the likelihood of random errors caused by repeated significance tests in cumulative meta-analysis. A computer software program called Trial Sequential Analysis Viewer (TSA Viewer) was utilized (Version 0.9.5.10 Beta. Copenhagen: Copenhagen Trial Unit, Centre for Clinical Intervention Research, Rigshospitalet. 2016) [32]). A two-sided design was utilized, with a type I error (α) of 0.05 and a type II error (β) of 0.10 (power of 90%). Both the alpha- and beta-spending functions were set to the empirical O’Brien-Fleming type (esOF). Futility boundaries were not used. The required information size (RIS) was calculated without regard for heterogeneity and then raised by 3% to accommodate for the sequential design. The sequential monitoring boundaries for benefit and harm were determined, and the cumulative Z-curve was plotted against these boundaries to see if firm evidence was obtained before reaching the RIS. The analysis was done retrospectively, and the study order was decided based on their appearance in the dataset.

Results

Literature search

We conducted a comprehensive literature search using PubMed, Scopus, and the Cochrane Library and identified 358 articles.Of which, 84 articles were removed as duplicates. After that, 274 articles were assessed based on their title and abstract, and 246 were excluded because they did not fulfill the inclusion criteria.From 28 articles that underwent full-text screening, 18 studies were excluded leaving eight articles for final inclusion. Manual screening retrieved two articles, raising the total number of finally included studies to ten [19,20,33–40] (Figure 1).

Figure 1.

PRISMA flow chart.

Study characteristics

We included ten clinical trials, comprising nine RCTs [19,20,34–40] and one prospective non-randomized trial [33]. The overall sample size comprised a total of 2642 adult patients with prostate cancer, with 1388 in the metformin groups and 1254 in the control groups. The mean age groups varied from 67 to 75 years. BMI averages ranged from 23.6 kg/m² to 31.1 kg/m². The baseline waist circumference had mean values ranging from 89 cm to 110 cm. Glycemic status at baseline varied from exclusively non-diabetic patients with mean HbA1c levels of roughly 5.6% to diabetics with HbA1c levels greater than 8.5%. (Table 1-2)

Table 1.

Characteristics of included studies.

Study	Design	Blinding	Country	Total N	Metformin/Control N	Daily Dose of Metformin	Titration	Placebo	Inclusion Criteria	Exclusion Criter
Aboelnaga et al. [33]	Prospective, Interventional Non-randomized Trial	Open-label	Egypt	95	48/47	2000 mg	Stepwise over 2 weeks	No	• Confirmed diagnosis of adenocarcinoma of PC at any stage. • Indicated for androgen deprivation therapy (ADT). • Type 2 diabetes mellitus (T2DM) or pre-diabetes.	• History of Metformin therapy within the last six months. • Renal impairment (GFR < 40 mL/min). • Hepatic impairment. • Other pathological types of PC. • Type 1 DM or secondary diabetes.
Alagizy et al. [34]	Phase 2 RCT	Open-label	Egypt	132	66/66	2000 mg	Stepwise over 2 weeks	No	• Histological diagnosis of prostatic adenocarcinoma. • Stage IV disease. • WHO Performance Status ≤ 2. • No prior use of ADT, chemotherapy, metformin, or other diabetes mellitus treatment.	• Patients with diabetes mellitus. • Advanced renal, hepatic, or cardiac disease.
Alghandour et al. [35] (MANSMED)	RCT	Open-label	Egypt	124	62/62	1700 mg	850→1700 mg over 1 month	No	• High localized PC (NCCN high-risk: T3/T4, PSA > 20, Gleason ≥ 8, or primary Gleason grade 5) OR node-positive/metastatic hormone-sensitive disease. • Pathologically confirmed prostatic adenocarcinoma. • Performance Status 0, 1, or 2. • Diabetic and non-diabetic patients included. • Adequate organ function (AST/ALT ≤ 2.5x ULN, Bilirubin ≤ 1.5x ULN, eGFR > 45). • Intended for long-term ADT.	• Prior use of metformin or its analogues. • Hypersensitivity to metformin. • History of diabetic ketoacidosis.
Birzniece et al. [36]	RCT, Crossover	Double-blind	Australia	30	15/15	1500 mg	500→1500 mg over 3 weeks	Yes	• Men aged 50–80 years. • Histologically confirmed PC (localized or advanced with bone-only mets). • ECOG Performance Status 0–1. • On stable ADT (GnRH analogues) for >6 months.	• Visceral metastases. • Castrate-resistant PC. • History or current diagnosis of type 1 or type 2 diabetes. • Current or prior use of metformin/other anti-diabetic drugs within the last year. • Hypersensitivity to metformin. • Renal (eGFR < 60) or hepatic impairment. • History of lactic acidosis, other malignancies, etc.
Eigl et al. [37] (PRIME)	Phase 3 RCT	Double-blind	Canada	166	112/54	1700 mg	850→1700 mg over 1 month	Yes	• Tissue diagnosis of PCa initiating ADT or started ADT within 45 days. • Initiating ADT for localized disease, metastatic disease, or biochemical recurrence.	• ADT within 12 months prior to enrollment. • Diagnosis of diabetes mellitus (per Canadian Diabetes Association criteria). • On metformin or other diabetes-mitigating medications. • History of lactic acidosis or conditions that predispose to it. • History of malignancy.
Gillessen et al. [20] (STAMPEDE)	Phase 3 RCT	Open-label	UK/Switzerland	1874	936/938	1700 mg	850→1700 mg over 4–6 weeks	No	• High-risk locally advanced or metastatic adenocarcinoma of the prostate. • Non-diabetic (HbA1c < 6.5%). • Adequate renal function (GFR ≥ 45 ml/min/1.73 m2). • WHO Performance Status 0–2. • Planned to receive long-term ADT.	(Refer to protocol for full list, but key exclusions are implied by inclusions: diabetes, poor renal function, poor performance status).
Mahalingam et al. [38]	Phase 2 RCT	Double-blind	USA	36	19/17	1500 mg	Not specified	Yes	(Specific criteria not explicitly detailed in the provided text, but the population consisted of men with metastatic, castration-sensitive PC who were eligible for chemotherapy before the standard of care changed.)	(Specific criteria not explicitly detailed in the provided text. Two patients failed screening for not meeting inclusion criteria.)
Nobes et al. [39]	RCT (Pilot)	Open-label	UK	40	20/20	1700 mg	850→1700 mg over 2 weeks	No	• Hormone-naïve patients due to receive ADT for any stage of prostate cancer. • WHO Performance Status 0–1.	• Pre-existing diagnosed diabetes mellitus. • Renal impairment (serum creatinine > 130 μmol/L, GFR < 40 mL/min). • Cardiac failure, MI within 6 months. • Hepatic impairment (LFTs > 1.5x ULN). • History of alcohol dependency (>21 units/week).
Usmani et al. [19]	Phase 2 RCT	Double-blind	Canada	83	39/44	1500 mg	500→1500 mg over 4 weeks	Yes	• Normoglycemic men (fasting glucose ≤ 6.9 mmol/L, HbA1c < 6.5%). • Unfavorable intermediate-risk or high-risk prostate cancer. • Fit for radical treatment with ADT and radiotherapy (ECOG 0 or 1).	• Metastatic disease (pelvic lymph nodes below aortic bifurcation allowed). • Preexisting diabetes. • Prior metformin use. • Risk of lactic acidosis (renal/liver impairment, alcohol abuse, severe infection). • Prior history of malignancy (with specific exceptions). • ADT initiated > 3 months before registration.
Zhu et al. [40]	RCT (Pilot)	Open-label	China	62	31/31	1500 mg	Gradual over 6 days	No	• Newly diagnosed PCa at any stage, pathologically confirmed. • Hormone-naïve and suitable for ADT. • Patients with untreated diabetes or hyperglycemia were included.	• Patients who had already received metformin or antidiabetic medications. • Severe medical/psychiatric diseases (severe heart failure, renal/liver failure, severe lung disease, infections, alcohol abuse, other cancers).

ADT = Androgen Deprivation Therapy; PC = Prostate Cancer; PCa = Prostate Cancer; T2DM = Type 2 Diabetes Mellitus; DM = Diabetes Mellitus; GFR = Glomerular Filtration Rate; eGFR = estimated Glomerular Filtration Rate; WHO = World Health Organization; ECOG = Eastern Cooperative Oncology Group; NCCN = National Comprehensive Cancer Network; PSA = Prostate-Specific Antigen; AST = Aspartate Aminotransferase; ALT = Alanine Aminotransferase; ULN = Upper Limit of Normal; LFTs = Liver Function Tests; GnRH = Gonadotropin-Releasing Hormone; RCT = Randomized Controlled Trial; MI = Myocardial Infarction; HbA1c = Glycated Hemoglobin; RT = Radiotherapy.

Table 2.

Baseline characteristics of included studies.

Study ID	Group	Age, year	BMI, kg/m2	WC, cm	HbA1c, %	Diabetes Status	Gleason Score	Stage	Baseline PSA, ng/ml	Prior ADT	ADT Type
Aboelnaga et al. [33]	Metformin	69.8 (5.4)	28.9 (4.3)	105.2 (11.6)	8.5 (1.9)	Diabetic/Pre-Diabetic	>7: 60.4%	Any	26.5 (63.5)	NR	LHRH agonist/Orch + Bicalutamide
	Control	68.6 (5.7)	28.6 (4.3)	104.7 (10.7)	8.8 (1.8)	Diabetic/Pre-Diabetic	>7: 63.8%	Any	27 (73)	NR	LHRH agonist/Orch + Bicalutamide
Alagizy et al. [34]	Metformin	69.5 (9.3)	NR	NR	4.80 (0.67)	Non-Diabetic	High: 51.5%	Stage IV	321.1 (313.5)	Naïve	LHRH agonist/Orch + Bicalutamide
	Control	70.6 (9.5)	NR	NR	4.76 (0.74)	Non-Diabetic	High: 56.1%	Stage IV	312.7 (328.6)	Naïve	LHRH agonist/Orch + Bicalutamide
Alghandour et al. [35]	Metformin	67 [46-81] §	Obese: 42%	NR	NR	Diabetic/Non-Diabetic	≥8: 47.5%	IIIA-IVB	95 (9-2348] §	NR	LHRH agonist/Orch ± Bicalutamide
	Control	69 [46-87] §	Obese: 48.4%	NR	NR	Diabetic/Non-Diabetic	≥8: 61.7%	IIIA-IVB	95 (1-1880] §	NR	LHRH agonist/Orch ± Bicalutamide
Birzniece et al. [36]	Metformin/Control	70.3 (1.6)	31.1 (1.1)	NR	NR	Non-Diabetic	Not Specified	Not Specified	0.4 (0.3)**	Yes	GnRH analogues
Eigl et al. [37], (PRIME)	Metformin	71 (7.41)	28 (4.5)	102 (11)	5.6 (0.3)	Non-Diabetic	NR	Not Specified	20 (31)**	≤45 days	LHRH agonist/antagonist
	Control	69 (6.66)	31 (4)	109 (12)	5.7 (0.3)	Non-Diabetic	NR	Not Specified	21 (47)	≤45 days	LHRH agonist/antagonist
STAMPEDE Gillessen et al. [20]	Metformin	69 (8.15)	NR	NR	<6.5	Non-Diabetic	NR	Metastatic	87 (288.15)	New/Relapsing	LHRH agonist/antagonist
	Control	69 (7.41)	NR	NR	<6.5	Non-diabetic	NR	Metastatic	80 (201.48)	New/Relapsing	LHRH agonist/antagonist
Mahalingam et al. [38]	Metformin	66.9 (10.1)	29.0 (5.1)	41.1 (5.3)	NR	Non-diabetic	NR	Advanced	79.4 (129.6)	≤6 weeks	LHRH agonist/Orch
	Control	70.1 (13.5)	27.0 (3.4)	40.5 (3.9)	NR	Non-diabetic	NR	Advanced	161.7 (418.1)	≤6 weeks	LHRH agonist/Orch
Nobes et al. [39]	Metformin	70.5 (58 – 80) §	27.0 (3.1)	100.1 (6.8)	5.51 (0.20)	Non-Diabetic	Not Specified	Not Specified	11 (5.0 – 50) §	Naïve	GnRH agonist + Bicalutamide
	Control	69.5 (56 – 84) §	25.9 (3.7)	101.6 (9.6)	5.37 (0.27)	Non-Diabetic	Not Specified	Not Specified	17.73 (0.2 – 726) §	Naïve	GnRH agonist + Bicalutamide
Usmani et al. [19]	Metformin	68.7 (52.4–79.3)*	29.4 (25.8–33.4)*	110.1 (82–154)*	5.6 (4.9–6.3)*	Non-Diabetic	Not Specified	High-risk	13.11 (0.8–80.6)*	≤3 months	GnRH agonist + Bicalutamide
	Control	70.7 (48.7–82.8)*	28.2 (26.5–32.1)*	107.6 (92–134)	5.6 (4.9–6.4)*	Non-Diabetic	Not Specified	High-risk	16.5 (0.3–88.4)*	≤3 months	GnRH agonist + Bicalutamide
Zhu et al. [40]	Metformin	74.98 (7.34)	23.58 (22.62–24.54)†	89.02 (87.00–91.04) †	NR	Diabetic	Not Specified	Any stage	43.60 (18.08–69.13) †	Naïve	GnRH agonist/Orch + Antiandrogen
	Control	73.70 (8.50)	24.20 (22.77–25.63) †	88.50 (84.42–92.58) †	NR	Diabetic	Not Specified	Any stage	38.60 (15.53–61.67) †	Naïve	GnRH agonist/Orch + Antiandrogen

ADT, Androgen Deprivation Therapy; BMI, Body Mass Index; HbA1c, Glycated Hemoglobin; IQR, Interquartile Range; LHRH, Luteinizing Hormone-Releasing Hormone; GnRH, Gonadotropin-Releasing Hormone; NR, Not Reported; Orch, Orchiectomy; PSA, Prostate-Specific Antigen; WC, Waist Circumference. Data are presented as Mean (Standard Deviation) unless otherwise specified.*Data present as mean (range), §Data present as median (range), †Data present as mean (95% CI), **Data measured by mg/l.

Quality assessment and certainty of evidence

The overall quality of RCTs included in our review using RoB2 was considered to be of some concerns or high risk of bias [19,20,34–40]. Figure 2 shows the summary and traffic-light plot of the quality assessment. Using ROBIN-I, Aboelnaga et al. was considered to have critical risk of bias [33] (Figure 3).

Figure 3.

Summary and traffic-light graph of risk of bias of non-randomized trial.

Safety and efficacy outcomes

Primary efficacy outcomes

HbA1c

The pooled analysis showed insignificant difference between metformin and placebo in decreasing HBA1c (MD = −0.01, 95% CI −0.1; 0.2, p = 0.8). The pooled results were homogenous (I2 = 0, p = 0.8) (Figure 4(A)). Doi plot showed no asymmetry (LFK index = 0.01) (Supplementary appendix). Subgroup analysis showed insignificant difference in all subgroups (Supplementary appendix).

Figure 4.

(A) forest plot for HBA1c, (B) forest plot for BMI, (C) forest plot for body weight (kg), (D) forest plot for waist circumference (cm).

BMI

The pooled analysis showed insignificant difference between metformin and placebo in decreasing BMI (MD = −0.01, 95% CI −0.1; 0.2, p = 0.8). The pooled results were homogenous (I2 = 0, p = 0.8) (Figure 4(B)). Doi plot showed minor asymmetry (LFK index = 1.27) (Supplementary appendix). The trim-and-fill – adjusted random-effects meta-analysis, which included five studies after imputing two potentially missing ones, demonstrated a significant overall reduction in the outcome favoring the intervention (MD = −1.15; 95% CI −1.89 to −0.41; p = 0.0023). Heterogeneity was minimal (I² = 0%), and the Q-test indicated no significant between-study variability (p = 0.81), suggesting that the included studies were highly consistent. Although the trim-and-fill procedure added two studies with more negative effect estimates, the adjusted pooled effect remained significant and directionally stable, indicating that the observed benefit of the intervention is robust and unlikely to be explained by publication bias (Supplementary appendix).

Body weight (kg)

The pooled analysis showed insignificant difference between metformin and placebo in decreasing body weight (MD = 0.42, 95% CI −3.31; 4.16, p = 0.8). The pooled results were heterogenous (I2 = 63.3, p = 0.04) (Figure 4(C)). Leave one out meta-analysis showed that the heterogeneity resolved after removing Usmani et al. [19] (MD = −1.77, 95% CI −5.21; 1.67, I2 = 0%) (Supplementary appendix). Doi plot showed major asymmetry (LFK index = −5.68) (Supplementary appendix). The Baujat plot shows that Usmani et al. [19] exerts the greatest influence on the overall pooled effect, indicating that this study has the strongest impact on the meta-analytic estimate despite contributing minimally to heterogeneity. In contrast, Aboelnaga et al. [33] contributes the most to heterogeneity, although its influence on the overall effect size is comparatively modest. The remaining studies, including Nobes et al. [39] and Mahalingam et al. [38], demonstrate both low influence and low contribution to heterogeneity (Supplementary appendix).

The trim-and-fill – adjusted random-effects meta-analysis, which included six studies (two imputed), showed a non-significant pooled mean difference of 3.04 (95% CI: −1.33 to 7.40; p = 0.17), indicating no clear evidence of an overall effect. However, the results were characterized by substantial heterogeneity (I² = 72.8%; Q = 18.36, p = 0.0025) (Supplementary appendix).

The TSA indicates that the currently accumulated sample size is insufficient to confirm a statistically significant or robust treatment effect. The required information size (shown in red on the right side of the plot) is approximately 1765 participants (Figure 5). Subgroup analysis showed insignificant difference in all subgroups (Supplementary appendix).

Figure 5.

TSA graph for body weight (kg).

Waist circumference (cm)

The pooled analysis showed insignificant difference between metformin and placebo in decreasing waist circumference (MD = −1.67, 95% CI −4.6; 1.26, p = 0.26). The pooled results were heterogenous (I2 = 54.1, p = 0.08) (Figure 4(D)). Leave one out meta-analysis showed that the heterogeneity not resolved (Supplementary appendix). Doi plot showed minor asymmetry (LFK index = −1.4) (Supplementary appendix). The Baujat plot shows that Zhu et al. [40] exerts the greatest influence on the overall pooled effect, indicating that this study has the strongest impact on the meta-analytic estimate despite contributing minimally to heterogeneity. In contrast, Aboelnaga et al. [33] contributes the most to heterogeneity, although its influence on the overall effect size is comparatively modest (Supplementary appendix).

The random-effects meta-analysis of four studies showed a non-significant pooled mean difference of −1.67 (95% CI: −4.60 to 1.26; p = 0.26). The heterogeneity was moderate (I² = 54.1%). No studies were imputed by the trim-and-fill method, indicating no detectable publication bias (Supplementary appendix).

The TSA suggests that the accumulated evidence is insufficient to conclude a significant treatment effect on waist circumference, and additional data would be required to reach a firm decision (Figure 6). Subgroup analysis showed significant decrease in waist circumference at 8–9 months and 18 months (MD = −1.5, 95% CI (−2.92; −0.07), MD = −1.64, 95% CI (−3.25; −0.04); respectively) (Supplementary appendix)

Figure 6.

TSA graph for waist circumference (cm).

Systolic blood pressure (SBP)

The pooled analysis showed insignificant difference between metformin and placebo in decreasing SBP (MD = −3.18, 95% CI −8.29; 1.93, p = 0.22). The pooled results were homogenous (I2 = 0, p = 0.4) (Figure 7(A)). Doi plot showed minor asymmetry (LFK index = −1.01) (Supplementary appendix). Subgroup analysis showed insignificant difference in all subgroups (Supplementary appendix)

Figure 7.

(A) Forest plot for systolic blood pressure, (B) forest plot for diastolic blood pressure.

Diastolic blood pressure (DBP)

The pooled analysis showed insignificant difference between metformin and placebo in decreasing DBP (MD = 0.99, 95% CI −2.49; 4.46, p = 0.57). The pooled results were homogenous (I2 = 0, p = 0.8) (Figure 7(B)). Doi plot showed minor asymmetry (LFK index = 1.73) (Supplementary appendix). Subgroup analysis showed insignificant difference in all subgroups (Supplementary appendix)

Secondary efficacy outcomes

Fasting blood glucose (mmol/L)

The pooled analysis showed insignificant difference between metformin and placebo in decreasing fasting blood glucose at all time points (Figure 8)

Figure 8.

Forest plot fasting blood glucose (mmol/l).

Fasting triglycerides (mmol/L)

The pooled analysis showed insignificant difference between metformin and placebo in decreasing fasting triglycerides at all time points (Figure 9)

Figure 9.

Forest plot for fasting triglyceride (mmol/l).

HDL cholesterol (mmol/L)

The pooled analysis showed insignificant difference between metformin and placebo in increasing HDL at all time points (Figure 10)

Figure 10.

Forest plot for HDL cholesterol (mmol/L).

LDL cholesterol (mmol/L)

The pooled analysis showed insignificant difference between metformin and placebo in decreasing LDL at all time points except at 12 months metformin significantly decrease LDL (MD = −0.16, 95% CI −0.25; −0.06) (Figure 11)

Figure 11.

Forest plot for LDL cholesterol (mmol/L).

Total cholesterol (mmol/L)

The pooled analysis showed insignificant difference between metformin and placebo in decreasing total cholesterol at all time points except at 12 months metformin significantly decrease total cholesterol (MD = −0.21, 95% CI −0.32; −0.11) (Figure 12)

Figure 12.

Forest plot for total cholesterol (mmol/L).

Safety outcomes: adverse events

Adverse events

All adverse events showed insignificant difference between Metformin and placebo except Diarrhea was significantly high in metformin group (RR = 2.55, 95% CI 1.02; 6.39) (Figure 13).

Figure 13.

Forest plot for adverse events.

Discussion

Summary of findings

This meta-analysis comprised ten clinical trials that compared metformin to placebo on a variety of metabolic outcomes in patients with prostate cancer undergoing ADT. Our findings showed that metformin had no significant advantage over placebo in terms of HbA1c, BMI, waist circumference, or body weight. Subgroup analyses of waist circumference revealed a substantial reduction at 8–9 and 18 months. We also found no significant difference in systolic and diastolic blood pressure, fasting blood glucose, fasting triglycerides, or HDL cholesterol, with the exception of LDL and total cholesterol at 12-month follow-up. We also found a significant increase in the adverse event of diarrhea in the metformin group.

Interpretation of findings

ADT is a crucial part of the treatment of metastatic and locally advanced prostate cancer [41–43]. Patients with prostate cancer now have far better results thanks to recent developments in hormonal treatment; however, this has resulted in prolonged exposures to ADT and the toxicities that go along with it. ADT toxicity is caused by unfavorable metabolic and anthropometric changes, resulting in exacerbation of MS symptoms and related consequences [6,8,44]. A recent study verified that patients had significant weight gain, increased abdominal obesity, and worsening glucose metabolism within nine months of beginning ADT [37]. These anthropometric and metabolic changes occur especially when taken for extended periods of time, as in the case of locally advanced prostate cancer. These alterations continued and were typically more noticeable after a longer period of time [37]. As patients live longer with novel systemic treatments, preventing adverse consequences becomes increasingly important in clinical practice.

Understanding the metabolic disorders related to ADT has led to recommendations that it should only be administered to men who are expected to benefit significantly from it and that the overall treatment duration should be maintained to a minimum as possible [45]. ASCO guidelines [46] recommend the use of ADT for 18 months based on findings of a main trial that failed to show survival benefit of 36-month over 18-month regimen duration [45]. Metformin has proved ability to lower hepatic gluconeogenesis which is probably due to the reduction in glucose and HbA1c [47]. AMPK-linked reduction in hepatic lipogenesis and cholesterol is probably responsible for metformin’s ability to manage lipid levels [48].

According to Usmani et al. [19], during the first year of treatment, males on ADT gained weight and developed abdominal obesity, consistent with available evidence [9]. Additionally, Usmani et al. showed that five months after starting ADT, metformin was able to reduce this weight gain and, to a lesser extent, increases in waist circumference [19]. At 9, 12, and 18 months, the PRIME trial found that the metformin arm had a lower HbA1c than the placebo arm [37]. Furthermore, at nine months, metformin was shown to significantly lower TC and show a positive trend toward lower LDL [37].

However, these alterations were temporary and did not persist with further follow-up, which are in accordance with our findings. Metformin could not mitigate these effects with prolonged follow-up, indicating that these alterations are unlikely to be clinically relevant in this patient population. In accordance with our findings, a previously published meta-analysis of the role of metformin in the same cohort as our study found no significant differences in body weight, HbA1c, BMI, systolic and diastolic blood pressure, or prostate-specific antigen [49]. While we subgrouped patients at different follow-up durations, identified differences were not maintained across lengthy follow-up durations in outcomes such as waist circumference, LDL, or total cholesterol.

Evidence has shown that after five years, 20% of diabetics on metformin alone lose glycemic control [50]. Even though earlier loss of response to metformin should be less common, even a small loss in effect could lower the difference in metabolic syndrome prevalence between the two study groups, decreasing the likelihood of a statistical benefit. Also, patients on ADT lead less active lives during therapy due to fatigue, depression, and other unfavorable side effects, which may lessen the likelihood of favorable and sustained changes in metformin outcomes. Since prior research examining the effects of similar lifestyle modifications in patients using metformin has demonstrated improved weight, abdominal circumference, and BMI after six months, other factors including food and exercise may also have had an impact on the study’s findings [39].

Nobes et al. compared metformin and controlled lifestyle interventions, including a low glycemic index diet and exercise regimen, to standard treatment during a 6-month period [39]. The STAMPEDE study 23 showed significant changes in body weight and metabolic markers (HbA1c, total cholesterol, and LDL) without observed differences in fasting triglycerides or HDL cholesterol levels [20]. The study found a 5.37 kg difference in favor of the intervention arm after 6 months of follow-up, indicating that metformin combined with lifestyle changes may have a greater impact than metformin alone [39]. Our study suggests that metformin may need to be combined with other methods, such as exercise or nutritional treatments, to produce clinically significant blunting of increases in mean weight more than 4 kg and sustain this change over a prolonged follow-up. The incidence of adverse events from metformin in our study was low, in accordance with previous randomized studies [37]. This confirms that metformin is safe and well tolerated among patients receiving ADT.

Strengths and limitations

To the best of our knowledge, this is the most complete pooled analysis with the largest sample size to evaluate the effect of metformin on metabolic syndrome indices in prostate cancer patients undergoing ADT. We used rigorous analysis approaches such as TSA and trim-and-fill, as well as leave-one-out and Baujat plots in cases with heterogeneous results.

However, our findings must be interpreted with caution because the significant heterogeneity observed for crucial outcomes cannot be overlooked. Furthermore, TSA reports inconclusive evidence, indicating the necessity for future large-scale, randomized trials with the needed sample size.

Clinical implications and future recommendations

According to our meta-analysis, metformin did not significantly enhance the major metabolic parameters of prostate cancer patients following ADT. Our findings add to existing evidence by integrating the STAMPEDE and PRIME trials, which indicated significant changes in metabolic markers, but the pooled analysis failed to show a benefit of metformin. As a result, the existing data does not support the routine use of metformin to treat ADT-induced metabolic syndrome. Metformin may need to be used in conjunction with dietary changes and supervised exercise, among other structured lifestyle treatments and this should be investigated in future large-scale randomized trials.

Conclusion

Our meta-analysis found that metformin had no significant effect on the metabolic parameters of prostate cancer patients undergoing ADT, such as HbA1c, BMI, waist circumference, or body weight, systolic and diastolic blood pressure, fasting blood glucose, and various lipid profile parameters. While it has a safe and well-tolerated profile, its efficacy for this particular use in the prostate cancer population is limited. Metformin’s clinical benefit in this instance appears to be low when administered alone. Hence, future trials should investigate the combination of medication like metformin and structured lifestyle interventions.

Author contributions

Conceptualization, A.M.A. (Abdullah M. Alharran) and M.A.A. (Munawer A. Alsaid); Methodology, A.M.A. (Abdullah M. Alharran), M.A.A. (Munawer A. Alsaid), and A.A.H. (Ali A. Hussain); Software, A.M.A. (Abdullah M. Alharran); Validation, H.A. (Husain Alaradi) and F.A.A. (Faisal A. Aladwani); Formal analysis, A.M.A. (Abdullah M. Alharran); Investigation, M.A.A. (Munawer A. Alsaid), A.A.H. (Ali A. Hussain), and F.A.A. (Faisal A. Aladwani); Resources, A.A.H. (Ali A. Hussain) and F.A.A. (Faisal A. Aladwani); Data curation, M.A.A. (Munawer A. Alsaid) and A.M.A. (Abdullah M. Alharran); Writing – original draft preparation, A.M.A. (Abdullah M. Alharran); Writing – review and editing, M.A.A. (Munawer A. Alsaid), A.A.H. (Ali A. Hussain), F.A.A. (Faisal A. Aladwani), and H.A. (Husain Alaradi); Visualization, A.M.A. (Abdullah M. Alharran); Supervision, A.M.A. (Abdullah M. Alharran); Project administration, A.M.A. (Abdullah M. Alharran). All authors have read and approved the final manuscript.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Data availability statement

Data available on request from the corresponding author.

Supplementary material

Supplemental data for this article can be accessed online at https://doi.org/10.1080/20905998.2026.2637371