The Optimal Dosage and Duration of Metformin for Prevention and Treatment of Antipsychotic-Induced Weight Gain: An Updated Systematic Review and Meta-Analysis

Abstract

Background

Weight gain and metabolic complications are substantial adverse effects associated with second-generation antipsychotics. However, comprehensive guidelines for managing antipsychotic-induced weight gain are lacking.

Methods

This review included all double-blind, placebo-controlled studies investigating metformin’s effectiveness in addressing antipsychotic-related weight gain. We systematically searched PubMed, Embase, the Cochrane Central Register of Controlled Trials, Google Scholar, and ClinicalTrials.gov for relevant studies from the inception to 2024. A random-effects model was used for the meta-analysis.

Results

This meta-analysis, including 20 studies with 1070 patients, revealed that metformin significantly surpassed placebo in attenuating weight gain in patients receiving antipsychotics. The mean weight change with metformin was −3.32 kg [95% confidence interval (CI): −4.57 to −2.07]. Additionally, metformin use resulted in a marked decrease in body mass index [−1.24 kg/m² (95% CI: −1.70 to −0.77)]. Metformin could maintain the effects from 12 to 24 weeks.

Conclusions

This updated meta-analysis investigated the durations and dosages of metformin use in patients with schizophrenia experiencing antipsychotic-induced weight gain. The findings highlight the need for additional large-scale research to validate our findings.

Introduction

Second-generation antipsychotics (SGAs) are widely used because the incidence of extrapyramidal side effects is lower with SGAs than with first-generation antipsychotics. Although they are essential for treating schizophrenia, second-generation antipsychotics have severe adverse effects, especially weight gain and metabolic syndrome.¹ Metabolic syndrome involves a cluster of conditions, including high blood pressure, elevated blood glucose levels, and abnormal cholesterol levels.² These conditions also increase the risk of mortality; for instance, they double the risk of stroke-related death.³ Metabolic syndrome is common in people with schizophrenia, and its prevalence increases with age. Metabolic syndrome can affect cognition and quality of life. Furthermore, in long-term-stable patients with schizophrenia, metabolic syndrome may have minor effects on performance in symbolic encoding and spatial span tests. However, the overall cognitive performance of patients with schizophrenia and normal metabolism was found to be better than that of patients with schizophrenia and poor metabolic status.^4,5 Thus, most individuals receiving an SGA regimen develop obesity and related conditions in addition to their primary illness, and these conditions adversely affect patients’ quality of life and public health.⁶

The precise mechanisms underlying the side effects of SGAs are not fully understood but are believed to involve neurotransmitters and genetic factors.⁷ Although some studies have supported an association of these side effects with histamine, muscarinic, and adrenergic receptors,⁸ other studies have suggested that polymorphisms in the leptin gene and leptin receptor gene may also play a role.^9–11 Researchers have evaluated the ability of a wide range of medications to address metabolic syndrome in patients with psychiatric illnesses, with these medications including orlistat, glucagon-like peptide-1 (GLP-1) receptor agonists, naltrexone-bupropion, and metformin.¹² Of the most frequently studied medications, metformin is recommended because of its safety profile, potential effect on cognitive function,¹³ and numerous metabolic benefits, particularly its ability to promote weight loss.¹⁴

Clinical trials and meta-analyses have demonstrated the efficacy of metformin in preventing antipsychotic-induced weight gain.^15,16 However, these studies and the current clinical guidelines¹² have not identified a precise dosage and treatment duration or quantified the potential side effects. This study determined the most appropriate dosage and duration of metformin treatment to prevent and manage antipsychotic-induced weight gain. The findings of this study can provide valuable guidance for medical experts.

Methods

This review was conducted in accordance with methods outlined in the Cochrane Handbook for Systematic Reviews of Interventions and adhered to the updated version of the Preferred Reporting Items for Systematic Reviews and Meta-Analyses reporting guideline statement.^17,18 Two reviewers (T.R.P. and J.A.C) screened all titles and abstracts independently and identified relevant articles. Any disagreements were resolved through discussion. Moreover, we manually searched available bibliographies and reviewed articles to identify additional eligible studies. This study was prospectively registered in PROSPERO (registration number: CRD42023486319).

Literature Search Strategy

We searched the PubMed, Google Scholar, Cochrane Central Register of Controlled Trials, Clinicaltrial.gov, and Embase databases for studies published from the inception of each database to June 18, 2024, that investigated the efficacy of metformin in preventing antipsychotic agent–induced weight gain. The search terms used were as follows: randomized controlled trial OR clinical trial AND metformin AND antipsychotic agents OR dopamine antagonists OR atypical antipsychotics OR antipsychotic-induced weight gain OR second-generation antipsychotics. The details of the search terms are presented in Supplementary Table S1.

Inclusion and Exclusion Criteria

A study was included in this meta-analysis if it met the following criteria: (1) it was a randomized controlled trial (RCT) with a double-blind assessment design; (2) it enrolled patients who had received a schizophrenia diagnosis based on the criteria outlined in the Diagnostic and Statistical Manual of Mental Disorders, Fourth Edition (DSM-IV), DSM-V, or International Classification of Diseases, Tenth Edition and thereafter were prescribed an antipsychotic medication; and (3) it specifically investigated the effectiveness of metformin in reversing or preventing weight gain induced by antipsychotic treatment. The following types of studies were excluded: (1) open-label studies and observational studies, (2) studies lacking available data for analysis, and (3) duplicate reports of the same study.

Data Extraction and Quality Assessment

Two reviewers (T.R.P. and J.A.C.) independently assessed the methodological quality of each study by using the revised risk-of-bias (version 2.0) method.¹⁹ The following domains were assessed: the randomization process, deviations from the intended interventions, missing outcome data, and selected outcome reporting.

The following data were extracted from included studies: the first author; year of publication; study duration; study type (blinding and design); number of patients; patients’ characteristics; prescribed antipsychotic(s) and the dose; the control, comparator, or placebo group; concomitant drugs; all outcomes of interest; and adverse drug reactions.

Outcome Assessments

In this study, the primary outcome was weight change, and the secondary outcomes were body mass index (BMI), fasting blood glucose level, and the insulin resistance index. According to the definition provided by De Silva et al.,¹⁵ our definition of prevention included the administration of metformin before the onset of chronic schizophrenia, including its use before initiation of antipsychotic therapy and during the first episode of schizophrenia.

Fasting blood glucose is a widely recognized measure of metabolic health and serves as a critical indicator of glucose homeostasis; however, the insulin resistance index, despite being less familiar, can play an essential role in diagnosing and managing conditions associated with insulin resistance, such as type 2 diabetes and metabolic syndrome. This index can help health-care professionals understand how effectively an individual’s body responds to insulin, the key hormone involved in regulation of blood sugar. The insulin resistance index enables a detailed assessment of the metabolic implications of schizophrenia treatment and the preventative potential of metformin.

Statistical Methods and Data Synthesis

Statistical analysis was conducted using Review Manager version 5.4 (Cochrane Review Manager; version 5.4, Oxford, UK) and Comprehensive Meta-Analysis (CMA) software (CMA, version 3). The random-effects models of DerSimonian and Laird were employed because of large variations in participants and interventions in the included studies; such variations could cause deviations in the results.²⁰ We evaluated average variations in participants’ body weight, BMI, fasting glucose level, and insulin resistance index. If standard deviations were not reported, we calculated them by using standard errors for specific subgroups or from reported confidence intervals. We used Cochran’s Q to determine whether heterogeneity existed between the included studies and calculated the I² statistic to estimate the degree of heterogeneity (heterogeneity was considered significant when I² > 75%). To calculate I², the Cochran’s Q test is often employed. This test calculates Cochran’s Q by determining the weighted sum of the squared deviations of each study’s effect from the overall effect across all studies. However, using only the Q test cannot fully reveal how much of the total variation is attributable to heterogeneity as opposed to chance. I² is derived from Q and the number of degrees of freedom (df). The formula for calculating I² is as follows:

$${\displaystyle \begin{array}{l}{\displaystyle I^2=\left(\frac{Q-df}{Q}\right)\times 100\mathrm{\%}}\end{array}}$$

where Q is Cochran’s heterogeneity statistic, and df is usually calculated as the number of studies minus one (N −1). Publication bias was identified using a funnel plot and Egger’s test.

Subgroup analysis was performed to compare weight and BMI for adults versus children (age cutoff: 18 years²¹), those never given an antipsychotic agent versus those receiving an antipsychotic agent for a first episode versus those receiving an antipsychotic agent for a chronic condition, different dosages, different durations, patients with and those without diabetes, and adverse events. To evaluate the effect of heterogeneity, we performed sensitivity analysis in which individual studies were systematically removed. This methodology enabled a rigorous assessment of the robustness of the findings and helped identify whether any single study affected the overall results.

Results

Characteristics of Included Trials

We initially identified 4087 records from electronic databases. Subsequently, 119 records were eliminated because they were duplicates; this left 3968 records. All of these records were reviewed, and the eligibility of 19 full-text publications was determined. After full-text screening, we excluded 1 trial involving only 1 participant,²² 5 open-label studies,^23–27 5 conference letters or case reports,^16,28–31 and 1 research protocol.³² Ultimately, 8 additional studies^33–40 were identified and combined with 12 trials^41–52 included in a previous review¹⁵ for qualitative analysis (Figure 1). Table 1 summarizes the key features of the 20 studies included. The daily metformin dosage in these studies ranged from 500 to 2550 mg. Chiu et al.³⁴ compared 2 dosages of metformin (500 and 1000 mg). To prevent data duplication, we included only the experimental group receiving 1000 mg and the placebo group. The duration of metformin treatment varied, with regimens ranging from 12 to 24 weeks. All participants in the studies had received a diagnosis of schizophrenia based on the DSM-IV criteria and were receiving stable antipsychotic treatment. The mean age at diagnosis ranged from 13 to 47 years. The baseline mean weight and BMI ranged from 55.5 to 114.8 kg and from 21.32 to 42.4 kg/m², respectively (Supplementary Table S2). The percentage of enrollees with overweight and obesity were 35% and 30%, respectively. Three studies focused on children and adolescents.^40,43,50 A total of 1070 patients were enrolled in the 20 studies. The outcomes of the risk-of-bias evaluation for selected trials are depicted in Supplementary Figure S1. We discovered variation in the risk of bias among the included studies. Although all of the studies were RCTs, 6 were deemed to be of low quality.

Figure 1.

Preferred Reporting Items for Systematic Reviews and Meta-Analyses Flow Diagram for Study Selection

Table 1.

Characteristics of Included Trials

Study	Participants	Age, mean (SD)	Males, n(%), metformin/placebo	Country	Numbers	Intervention	Duration	Outcomes measured
Arman 2008	Age <20 years; On risperidone 2-6 mg	NA	NA	Saudi Arabia	Metformin N = 16 Placebo N = 16	Metformin 500 mg twice daily	12 weeks	Weight, BMI
Baptista 2006	Age >18 years; Olanzapine monotherapy >4 months	47.6 (8.6)	10 (52.6)/12 (66.7)	Venezuela	Metformin N = 19 Placebo N = 18	Metformin 850-1750 mg	14 weeks	Weight, BMI, Fasting glucose
Baptista 2007	Age >18 years; Olanzapine monotherapy >4 months	NA	23 (63.9)/19 (52.8)	Venezuela	Metformin N = 36 Placebo N = 36	Metformin 850-2550 mg	12 weeks	Weight, BMI, Fasting glucose, HOMA-IR
Carrizo 2009	Clozapine treatment >3 months	39.1 (16.1)	20 (83.3)/23 (76.7)	Venezuela	Metformin N = 31 Placebo N = 30	Extended-release metformin 500-1000 mg/day	14 weeks	Weight, BMI, HOMA-IR
Chen 2012	Clozapine treatment >3 months; BMI >24 or 1 metabolic syndrome criteria	41.6 (8.7)	13 (46.4)/15 (55.6)	Taiwan	Metformin N = 28 Placebo N = 27	Metformin 1500 mg/day	24 weeks	Weight, BMI, Fasting glucose, HOMA-IR
De Silva 2015	Age >18 years; Weight gain >10% of body weight	34.4 (10.2)	6 (17.6)/8 (25)	Sri Lanka	Metformin N = 34 Placebo N = 32	Metformin 500 mg twice daily or placebo	24 weeks	Weight, BMI, Fasting glucose
Jarskog 2013	Age 18-65 years; BMI >27; Duration of illness >1 year	43.2 (11.0)	52 (69.3)/49 (69.0)	United States	Metformin N = 75 Placebo N = 71	Metformin 500 mg twice daily increased up to a maximum of 2000 mg/day	16 weeks	Weight, BMI, Fasting glucose
Klein 2006	Age 10-17 years; Gained >10% body weight	13.1 (2.4)	12 (67.3)/9 (50)	United States	Metformin N = 34 Placebo N = 32	Metformin 850 mg twice daily	16 weeks	Weight, BMI, Fasting glucose
Wang 2012	Age 18-60 years; Gained >7% of body weight	26.4 (4.5)	15 (47)/19 (56)	China	Metformin N = 32 Placebo N = 34	Metformin 500 mg twice daily	12 weeks	Weight, BMI, and the HOMA-IR
Wu 2012	Age 18-40 years; First episode	26.4 (4.5)	NA	China	Metformin N = 42 Placebo N = 42	Metformin 1000 mg/day	24 weeks	Weight, BMI, HOMA-IR
Wu 2008a JAMA	Age 18-45 years; First episode patient who gained >10% of bodyweight	26.3 (0.8)	16 (50)/16 (50)	China	Metformin N = 32 Placebo N = 32	Metformin 750 mg or placebo (also metformin + lifestyle and lifestyle + placebo groups)	12 weeks	Weight, BMI, HOMA-IR
Wu 2008b AM J	Age 18-50 years; First episode patients on olanzapine	25.4 (3.7)	10 (55.6)/10 (52.6)	China	Metformin N = 18 Placebo N = 19	Metformin 250 mg thrice daily	12 weeks	Weight, BMI, Fasting glucose, HOMA-IR
Rado 2016	Age >18; Administration of clozapine or olanzapine was not permitted within 3 months	36.4 (9.0)	7 (58.3)/5 (38.5)	United States	Metformin = 12 Placebo = 13	Metformin 500-2000 mg	24 weeks	Weight, BMI, Fasting glucose, HOMA-IR
Agarwal 2020	BMI >25; Ages 17–45; <5 years of schizophrenia, schizoaffective disorder, or bipolar disorder	31.6 (6.4)	23.8 (7.66)/21.3 (4.90)	Canada	Metformin N = 21 Placebo N = 9	Metformin initiated at 500 mg QD, increased to 750 mg BID	16 weeks	Weight, BMI, Fasting glucose, HOMA-IR
Khan 2020	Olanzapine ≥3 m; Age: 17~60	NA	41(60)/43(62)	Pakistan	Metformin N = 69 Placebo N = 69	Olanzapine (5-20 mg) and Metformin (500 mg twice a day)	12 weeks	Weight
Sabaghi 2019	Age between 18-65; BMI ≥27; Duration of illness ≥1 year; at least 1 antipsychotic (no dose or type change in 1 m)	47.6 (10)	24 (72.7)/20 (60.6)	Iran	Metformin N = 33 Placebo N = 33	Metformin 500 mg twice daily	12 weeks	Weight, BMI, Fasting glucose
Sisikind 2021	Age 18-64; Clozapine ≤2 weeks; Fasting Blood Glucose Level ≤6.0 mmol/L; BMI 18-40	34.3 (11.1)	7 (70)/10 (100)	Australia	Metformin N = 8 Placebo N = 5	Metformin XR starting at 500 mg daily and titrating up to 2000 mg daily	24 weeks	Weight, BMI, Fasting glucose, HOMA-IR
Tang 2020	Ages 16-40 Gained ≥5% body weight; BMI ≥18.5	24.5 (5.1)	4 (50)/5 (55.6)	Singapore	Metformin N = 8 Placebo N = 9	Metformin XR starting at 500 mg daily and titrating up to 2000 mg	24 weeks	Weight, BMI
Chiu 2016	Aged 20–65 years; clozapine treatment ≥3 months	46.3 (8.8)	8 (42.1)/8 (44.4)	Taiwan	Metformin (500 mg) N = 18 Metformin (1000 mg) N = 19 Placebo N = 18	Metformin 1000 mg or 500 mg daily	12 weeks	Weight, BMI, Fasting glucose
NCT00617240	Ages 10-17; second-generation antipsychotic-naïve or less than 2 weeks exposure to any second-generation antipsychotic, except ziprasidone	14.2 (2.0)	3 (40)/2 (50)	United States	Metformin N = 5 Placebo N = 4	Metformin in doses from 250 mg to 2000 mg/day for 26 weeks	24 weeks	Weight

Synthesis of Results

Body Weight.

Our meta-analysis, involving 20 studies and 1070 participants, revealed that individuals undergoing antipsychotic treatment exhibited a significantly greater reduction in weight when administered metformin in comparison with a placebo. The mean difference (MD) in weight reduction was −3.32 kg [95% confidence interval (CI) −4.57 to −2.07, P < .001), as depicted in Figure 2.

Figure 2.

Forest Plot of Difference in Mean Weight Change for Metformin Versus Placebo

BMI.

Some of the included trials observed a significant difference in BMI between their metformin and placebo groups. In an evaluation of 18 studies, metformin treatment resulted in a considerably greater reduction in BMI than did placebo in individuals receiving antipsychotics (MD: −1.24 kg/m², 95% CI −1.70 to −0.77, P < .001; Figure 3).

Figure 3.

Forest Plot of Mean Change in BMI for Metformin Versus Placebo

Fasting Blood Glucose.

Figure 4 presents a forest plot derived from a meta-analysis examining the change in fasting blood glucose level; the plot was derived from 15 studies involving 844 participants in total. The meta-analysis revealed a significant reduction in fasting blood glucose level in the metformin group compared with the placebo group (MD: −0.47 mg/dl, 95% CI −0.92 to −0.01, P = .049; Figure 4).

Figure 4.

Forest Plot of Mean Change in Fasting Blood Glucose Level for Metformin Versus Placebo

Insulin Resistance Index.

Figure 5 displays a forest plot from the meta-analysis assessing changes in the insulin resistance index. Of the 12 studies analyzed, 7 demonstrated a significant discrepancy in changes in this index between the metformin and placebo groups. Compared with placebo, metformin significantly improved insulin resistance in patients receiving antipsychotic medication (MD: −1.43, 95% CI −2.17 to −0.69, P = .0002; Figure 5).

Figure 5.

Forest Plot of Mean Change in the Insulin Resistance Index for Metformin Versus Placebo

Subgroup Analysis

Adults Versus Children.

Data synthesized from 3 studies^40,43,53 focusing on a young patient demographic revealed a significant advantage in weight reduction when taking metformin instead of placebo (MD: −3.94 kg, 95% CI −7.17 to −0.70, P < .001). The pooled weight reduction for adults was significant (MD: −3.29 kg, 95% CI −4.61 to −1.97, P = .02). However, a subgroup analysis involving subgroups of pediatric and adult patients revealed no significant differences in the weight reduction benefit of metformin between these 2 subgroups (χ² = 0.13, df = 1, P = .72, I² = 0%). We also evaluated the effect of metformin on BMI in children versus adults. As summarized in Table 2, the adult and child subgroups exhibited BMI reductions of 1.22 and 1.47 kg/m², respectively (adults: MD: −1.22 kg/m², 95% CI −1.71 to −.74, P < .001; children: MD: −1.47 kg/m², 95% CI −2.57 to −0.36, P = .009). Similar trends were noted in the subgroups, and the differences between them were not significant (χ² = 0.16, df = 1, P = .69).

Table 2.

Summary of Subgroup Analyses

Subgroup categories	Number of studies	Participants	Effect estimate, MD (95% CI)	Heterogeneity (I2)
Change of weight in adults versus children
Adult	17	999	−3.29 (−4.61, −1.97)	87%
Children	3	71	−3.94 (−7.17, −0.70)	0%
Overall	20	1070	−3.32 (−4.57, −2.07)	84%
Change of BMI in adults versus children
Adult	16	861	−1.22 (−1.71, −0.74)	85%
Children	2	62	−1.47 (−2.57, −0.36)	0%
Overall	18	923	−1.24 (−1.70, −0.77)	83%
Change of weight among patients who never given an antipsychotic agent versus those receiving an antipsychotic agent for a first episode versus those receiving an antipsychotic agent for a chronic condition
Patients who never given an antipsychotic agent	4	131	−2.87 (−5.37, −0.37)	63%
Patients receiving an antipsychotic agent for a first episode	4	231	−6.10 (−6.97, −5.22)	0%
Patients receiving an antipsychotic agent for a chronic condition	14	770	−2.84 (−4.32, −1.36)	81%
Overall	22	1432	−3.32 (−4.57, −2.07)	84%
Change of BMI among patients who never given an antipsychotic agent versus those receiving an antipsychotic agent for a first episode versus those receiving an antipsychotic agent for a chronic condition
Patients who never given an antipsychotic agent	5	227	−1.80 (−3.14, −0.47)	84%
Patients receiving an antipsychotic agent for a first episode	4	231	−1.55 (−2.78, −0.33)	89%
Patients receiving an antipsychotic agent for a chronic condition	10	560	−1.10 (−1.65, −0.55)	84%
Overall	19	1081	−1.37 (−1.96, −0.79)	91%
Change of weight in different metformin durations
12 weeks	8	541	−4.11 (−6.47, −1.74)	92%
24 weeks	7	269	−3.05 (−4.03, −2.07)	0%
Overall	15	810	−3.77 (−5.29, −2.24)	85%
Change of weight in different dosages of metformin
No more than 1000 mg daily	11	790	−3.85 (−5.64, −2.06)	90%
More than 1000 mg daily	9	280	−2.05 (−2.85, −1.24)	0%
Overall	20	1070	−3.32 (−4.57, −2.07)	84%
Change of weight in patients with diabetes versus patients without diabetes
Patients with diabetes	1	22	−2.96 (−7.07, 1.16)	NA
Patients without diabetes	19	1048	−3.34 (−4.62, −2.05)	85%
Overall	20	1070	−3.32 (−4.57, −2.07)	84%
Change of BMI in patients with diabetes versus patients without diabetes
Patients with diabetes	1	22	−0.75 (−5.10, 3.60)	NA
Patients without diabetes	17	1001	−1.34 (−1.50, −1.18)	84%
Overall	18	1023	−1.34 (−1.50, −1.18)	83%

Patients Who Never Given an Antipsychotic Agent Versus Those Receiving an Antipsychotic Agent for a First Episode Versus Those Receiving an Antipsychotic Agent for a Chronic Condition.

Metformin was found to reduce body weight by −2.87 kg (95% CI −5.37 to −.37, P = .02) in patients who had never taken an antipsychotic drug. Analysis of patients receiving antipsychotics for a first episode or a chronic schizophrenia revealed similar trends (first episode: MD: −6.10 kg, 95% CI −6.97 to −5.22, P < .001; chronic condition: MD: −2.84 kg, 95% CI −4.32 to −1.36, P < .001). In addition, metformin treatment resulted in a reduction in BMI (never taken an antipsychotics: MD: −1.80 kg/m², 95% −3.14-−0.47, P = .008; first episode: MD: −1.55 kg/m², 95% CI −2.78 to −0.33, P < .01; chronic condition: MD: −1.10 kg/m², 95% CI −1.65 to −0.55, P < .001).

Different Treatment Durations.

We further conducted an analysis of subgroups with different durations of metformin use. Defining 12-week and 24-week regimens, we observed a nonsignificant difference between the 2 subgroups (χ = .65, df = 1, P = .42). As summarized in Table 2, the use of metformin was favored over placebo in both regimens. Eight studies employed 12 weeks of treatment and discovered that metformin had a significant effect (MD: −4.11 kg, 95% CI −6.47 to −1.74, P < .001). The longer-duration regimen studies also demonstrated the efficacy of metformin (MD: −3.05 kg, 95% CI −4.03 to −2.07, P < .001).

Different Dosages.

We performed an analysis involving subgroups with different metformin dosages. We used 1000 mg as a cutoff and thereby separated the included studies into 2 subgroups: ≤1000 and >1000 mg. As summarized in Table 2, the result suggested that the use of metformin was favored over that of placebo in both groups of antipsychotic users. The low-dose subgroup included 11 studies, and a weight decrease of 3.85 kg was observed in this subgroup (95% CI −5.64 to −2.06, P < .001). The high-dose subgroup had a similar outcome—a reduction in weight of 2.05 kg (MD: −2.05, 95% CI −2.85 to −1.24, P < .001). The results for the 2 subgroups were no significantly different (χ = 3.23, df = 1, P = .07).

Patients With Versus Those Without Diabetes.

In this subgroup analysis, metformin significantly reduced body weight by −3.32 kg in the patients without diabetes (95% CI −4.62 to −2.05, P < .001) but did not do so in patients with diabetes. Similar results were found for BMI (Table 2).

Safety Profile

Supplementary Figure S2 displays the safety profile results. The most frequently reported side effects were vomiting, nausea, and diarrhea. No evidence was obtained that associated metformin with more frequent episodes of nausea and vomiting (results of 11 trials; MD: 1.18, 95% CI 0.88 to 1.57, P = .27). According to the meta-analysis of available information on diarrhea, metformin did not result in a higher incidence of this side effect than did placebo (MD: 1.20, 95% CI 0.75 to 1.91, P = .44).

Sensitivity Analysis

To further determine the robustness of the results of our primary outcome, the studies regarded as being high risk in the risk of bias assessment were excluded. Despite this exclusion, the outcome still favored metformin, thereby demonstrating that the efficacy of metformin is robust (MD: −2.70 kg, 95% CI −3.32 to −2.07, P < .001). Significant heterogeneity was observed among the studies (P < .001, I² = 84%). Thus, we removed each study separately and found that the removal of 4 specific studies^35,36,41,44 significantly reduced the heterogeneity from 84% to 0%, as depicted in Supplementary Figure S3 (P = .46, I² = 0%).

Publication Bias

We examined publishing bias through a funnel plot, which is presented in Supplementary Figure S4, indicating a symmetric outcome. The treatment effect is represented on the horizontal axis, and the standard error is represented on the vertical axis. The results of Egger’s regression test revealed a two-tailed P value of approximately 0.14, which exceeds 0.1, suggesting an absence of bias in the results.

Discussion

We conducted a thorough data analysis of metformin use in patients with schizophrenia and its effects on antipsychotic-induced weight gain. The results revealed a marked decrease in weight, BMI, fasting glucose level, and insulin resistance index in the metformin group. Specifically, the mean weight change across 20 studies was −3.32 kg (95% CI −4.57 to −2.07). A similar trend was noted in the 12-study meta-analysis conducted by de Silva et al.,¹⁵ who reported a mean weight change of −3.27 kg in the metformin group (95% CI −4.66 to −1.89). Notably, the study conducted by Ellul et al.⁵³ focused on children and adolescents, determining a weight change of −1.83 kg (95% CI −2.47 to −1.18) after 12 weeks of therapy, whereas a six-study meta-analysis by Solymár et al.⁵⁴ involving participants over 60 years old found a weight change of −2.23 kg (95% CI −2.84 to −1.62). Moreover, on the basis of information from 4 trials, the Cochrane Review, which was published in 2022¹⁶ and focused on patients using metformin before taking an antipsychotic agent, reported that metformin may prevent these individuals from gaining 4.03 kg (95% CI 2.28 to 5.78). In the present meta-analysis, we not only evaluated patients who never taken an antipsychotic agent but also separated patients taking an antipsychotic agent without weight gain (first episode) from the patients with antipsychotic-induced weight gain. Our results indicated that metformin reduced weight and BMI in all 3 of these patient groups. Therefore, metformin is suitable for not only prevention but also treatment of antipsychotic-induced weight gain. In addition, the metabolic effects of metformin on body weight loss could improve cognitive functions.⁵⁵ Battini et al. indicated that, to some degree, metformin improved cognitive, and other dimensions of schizophrenia symptoms in patients with schizophrenia.⁵⁶ Overall, our review provides a comprehensive summary of the current evidence, confirming that metformin is effective in reversing antipsychotic-induced weight gain, with the results being consistent with those of earlier trials conducted across different populations and treatment durations.

Regarding sensitivity analysis, upon exclusion of 4 specific trials (Baptista 2006,⁴⁴ Khan 2020,³⁵ Sabaghi 2019,³⁶ and Wu 2008b⁴²), we found that the I² decreased from 84% to 0%. This substantial decrease in heterogeneity suggests that the excluded studies contributed to the high degree of variation when all of the results were pooled. Further examination of these studies revealed that these trials shared several characteristics that may have introduced bias and affected the overall analysis. Specifically, the study by Baptista⁴⁴ exhibited a high risk of bias and involved the concurrent use of olanzapine and metformin. The studies by Khan³⁵ and Wu⁴² involved concurrent use without incorporating behavioral and nutrition interventions. Sabaghi³⁶ also employed no behavioral or nutrition intervention. Thus, 3 studies involved the concurrent use of metformin and olanzapine, 3 did not include behavior or nutrition interventions, and 1 was deemed high risk. This sensitivity analysis emphasizes the importance of examining the specific characteristics of each study and enhances the strength and reliability of our conclusions.

In addition to pooling all the available data, we investigated the crucial aspect of metformin treatment duration. The subgroup analysis of different treatment durations revealed a substantial weight change of −4.11 kg (95% CI −6.47 to −1.74) after 12 weeks of therapy and a comparable −3.05 kg (95% CI −4.03 to −2.07) after 24-weeks of therapy; no significant difference in the outcome was observed between the 2 durations. This finding suggests that the benefits of metformin are achieved within 12 weeks, with its effect plateauing thereafter. It may also indicate that the effect of metformin could be maintained for 24 weeks. By contrast, the meta-analysis published by Ellul et al.,⁵³ which also explored short-term metformin use over various durations, compared weight loss at weeks 4, 12, and 16 and discovered loss of 0.98, 1.83, and 3.23 kg, respectively; the weight loss thus increased with an increase in the treatment duration, without a plateau effect as was found in our analysis. Despite the substantial heterogeneity among the subgroup analyses, our findings offer valuable insights into the optimal duration of metformin therapy, indicating that a 12-24 week regimen may be a reasonable approach to achieving significant weight reduction.

The current guidelines¹² recommend starting with a dosage of 500 mg twice daily, progressively increasing to a maximum of 2000 mg per day. However, the quality of the evidence supporting this specific dosing regimen is relatively low, and the optimal dosage has not been conclusively determined. Our meta-analysis results indicate that doses of ≤1000 mg/day result in a weight change of −3.85 (95% CI −5.64 to −2.06), whereas doses of >1000 mg/day result in a smaller change of −2.05 (95% CI −2.85 to −1.24), with the difference between these subgroups being nonsignificant. A network meta-analysis conducted by Fu et al.⁵⁷ focused on the role of metformin in individuals with obesity but without diabetes and utilized a similar approach. They categorized metformin doses into ≤1000, 1000–2000, and ≥2000 mg/day and compared the reduction in BMI in the 3 subgroups to identify the most suitable dosage and intervention period for using metformin in adolescents and adults.

We aimed to update the research of De Silva et al.¹⁵ for 2 main reasons. First, it includes only double-blind, placebo-controlled RCTs, which are considered the gold standard in clinical research. Such trials follow a robust protocol, adhering to PRISMA guidelines and ensuring systematic and transparent reporting. To differentiate our meta-analysis from previous meta-analyses, we concentrated on determining the optimal metformin treatment dosage and duration and quantifying the potential side effects of this treatment. Furthermore, we performed sensitivity analysis to rigorously examine the impact of heterogeneity. Second, to assess publication bias, we used not only funnel plots but also Egger’s test. Although some may argue that the interpretation of funnel plots can be subjective, Egger’s regression test offers a solution to this problem by detecting publication bias through comprehensive statistical analysis.

Several drugs have been approved especially for the treatment of obesity and for weight management (such as orlistat, GLP-1 agonists, and naltrexone-bupropion), and each of these drugs has its own advantages and limitations, which should be carefully considered. RCTs have shown that orlistat reduces weight by approximately 3 kg over 1 year, but the high dropout rate with this treatment limits its long-term usage.⁵⁸ GLP-1 agonists can be expensive and are not covered by all insurance plans. The data in favor of naltrexone-bupropion are conflicting; only 1 RCT has been published, and its weight reduction outcomes were nonsignificant.⁵⁹

In this updated meta-analysis, we systematically investigated durations, dosages, and adverse events when using metformin to treat patients with schizophrenia and antipsychotic-induced weight gain. Although our outcomes are promising, our study is not without limitations. First, 6 out of the 20 included studies were assessed as having a high risk of bias, and this may have affected the robustness of our findings. Second, substantial heterogeneity was detected in the subgroup analyses, which may also have biased our results. Third, we included studies with very small participant groups, such as the NCT00617240 study,⁴⁰ which had only 9 participants; these small-scale trials likely affected our estimate of the effect of metformin. Fourth, the body weight and BMI of children have different significance to those of adults, and so does the dosage of metformin. Therefore, caution should be taken when interpreting the results of metformin in children. Finally, long-term treatment is necessary to maintain weight loss and achieve clinical benefits. Weight may be regained if metformin is discontinued, especially because patients with schizophrenia are unlikely to discontinue their antipsychotic treatment. However, this cannot be addressed by the current results.

Some aspects of this research field still need to be clarified, and future research must involve larger groups of participants and compare a series of timeframes and metformin dosages. Such research could assess the efficacy and safety of metformin in patients with schizophrenia and antipsychotic-induced weight gain and thereby contribute to the understanding and management of this clinical challenge, ultimately enhancing the quality of life in these individuals.

Conclusion

Our meta-analysis reveals significant reductions in weight, BMI, fasting glucose level, and insulin resistance index in the metformin group. The efficacy of metformin is consistent across various treatment durations, with no significant difference in weight loss observed between 12 and 24 weeks of therapy. Regarding dosing, our findings suggest that both lower (≤1000 mg/day) and higher (>1000 mg/day) doses are effective. Future research should focus on larger, more diverse populations and explore different metformin dosages and treatment periods to further validate and refine these findings. Such research will enhance clinical management and improve the quality of life for patients with schizophrenia and associated metabolic complications.

Supplementary material

Supplementary material is available at https://academic.oup.com/schizophreniabulletin/.

Author contributions

T.R.P. and J.A.C. wrote the first draft of the manuscript. T.R.P. and C.P.H. reviewed and collected the data from patients. J.A.L. and M.C.L. performed the statistical analysis. L.M.C. and C. H. M. critically revised the manuscript. All authors contributed to the final version of the manuscript.

Funding

This study was not funded by any institutions.

Conflicts of interest

All authors declare that they have no competing interests.