Efficacy and safety of variable-dose metformin as adjunctive therapy to insulin in adolescents with type 1 diabetes mellitus: a systematic review and network meta-analysis

Abstract

Background

Adolescent Type 1 Diabetes (T1D) is complicated by insulin resistance, dyslipidemia, and heightened cardiovascular risk. As adjunctive metformin therapy lacks optimized dosing, we conducted a network meta-analysis (NMA) to rigorously assess the dose-dependent efficacy and safety of metformin combined with insulin in this population.

Methods

Totally 764 adolescents (aged 10–19 years) met inclusion criteria were involved. We assessed five metformin regimens: 1.0 g/day, 1.7 g/day, 2.0 g/day, weight-based (≤ 60 kg:1.0 g/day;≥60 kg:2.0 g/day), and multi-tiered weight-based (< 50 kg:1.0 g;50–75 kg:1.5 g;≥75 kg:2.0 g). Outcomes included HbA1c, BMI/BMI-Z, insulin dose, lipid profile, and adverse events. A Bayesian NMA was performed using R 4.4.1, with effect sizes reported as mean differences (MD) or relative risks (RR) with 95% confidence intervals (CI). Surface under the cumulative ranking curve (SUCRA) values ranked interventions.

Results

Consequently, metformin 2.0 g/day significantly reduced BMI (MD=-0.6 kg/m², 95%CI:-0.68, -0.52) and LDL-C (MD=-12.78 mg/dL, 95%CI:-21.17, -0.49) versus placebo. Doses of 1.0 g/day, 2.0 g/day, and the 60 kg weight-based regimen significantly reduced daily insulin requirements. Metformin 2.0 g/day and the 50 kg weight-based regimen reduced total cholesterol. No regimen significantly lowered HbA1c or triglycerides versus placebo. All doses demonstrated safety profiles comparable to placebo regarding gastrointestinal events, hypoglycemia, diabetic ketoacidosis (DKA), and transaminase elevations (RRs not significant).

Conclusion

To sum up, adjunctive metformin at 2.0 g/day offers significant benefits in weight management, insulin dose reduction, and lipid improvement (LDL-C, total cholesterol) for adolescents with T1D, with a favorable safety profile across all doses, which will represent a viable therapeutic option.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12902-025-02043-7.

Introduction

Type 1 Diabetes (T1D) is an autoimmune disorder characterized by the progressive destruction of pancreatic β-cells, leading to a deficiency in insulin secretion. This condition significantly affects the health and quality of life of millions of adolescents globally [1, 2]. The conventional treatment involves the administration of exogenous insulin to regulate blood glucose levels; however, achieving long-term glycemic control remains challenging. Insulin monotherapy often often results in frequent hypoglycemic episodes and significant glucose variability, and it has not been demonstrated to effectively reduce the risk of long-term complications [3, 4].

Adolescents with T1D are particularly susceptible to reduced insulin sensitivity due to elevated levels of glucagon-like hormones and suboptimal dietary habits, which can contribute to insulin resistance, excessive weight gain, and subsequent lipid metabolic disorders. These factors increase the potential risk for long-term cardiovascular complications [5–7]. Metformin, traditionally used for Type 2 Diabetes Mellitus (T2DM), is being explored for Type 1 Diabetes management due to its ability to lower liver glucose production, enhance glucose uptake, and improve insulin sensitivity. It also aids in appetite suppression, weight loss, and lipid regulation [8].

However, there is no consensus on the optimal dose, effectiveness, or safety for adolescents with T1D [9]. Studies have employed various metformin dosing regimens, showing mixed results on blood glucose control, weight, insulin needs, and side effects. These variations affect clinical decisions and limit the broader use of metformin as an adjunct therapy.

Consequently, a network meta-analysis is imperative to rigorously evaluate the efficacy and safety of integrating variable-dose metformin into insulin regimens for adolescents with T1D. This study aims to systematically aggregate and synthesize extant literature using advanced statistical methodologies, with the primary objective of quantifying the effects of divergent metformin dosages on glycemic control, weight dynamics, insulin dose requirements, and health-related quality of life. Additionally, the investigation will comprehensively characterize the safety landscape of metformin, focusing on gastrointestinal adverse events, hypoglycemia risk, and hepatic enzyme elevations.

Materials and methods

PROSPERO registration number: CRD42024594507. This study followed the PRISMA guidelines [10] for literature retrieval, data organization, quality control, and result interpretation.

Data sources and retrieval

Electronic searches were performed in PubMed, Embase, Cochrane Library, and Web of Science from database inception to June 30, 2024. A combination of subject-specific and free-text search strategies was employed, with randomized controlled trials (RCTs) using the search strategy provided by the Cochrane Collaboration. The search keywords used were (“metformin”) AND (“Type 1 diabetes”) AND (“RCT”). The full, detailed search string for each database (PubMed, Embase, Cochrane, Web of Science) in Supplementary file Table S1.

Inclusion criteria

1. Study Design: Randomized controlled trials (RCTs) or quasi-RCTs with parallel or crossover design.

• (2)Population: Adolescents (aged 10–19 years) with confirmed type 1 diabetes (T1D) based on clinical criteria (e.g., insulin dependence, presence of islet autoantibodies).

• (3)Intervention: Studies evaluating metformin monotherapy or combination therapy with insulin, including fixed-dose regimens (e.g., 1.0 g/day, 1.7 g/day, 2.0 g/day) or weight-based dosing 50 kg and 60 kg group).

• (4)
  Outcomes:

  • a
    Efficacy Measures: Glycated hemoglobin (HbA1c), body mass index (BMI), BMI-Z score, daily insulin requirements (U/kg/day), insulin sensitivity indices (e.g., HOMA-IR, QUICKI), lipid profile [total cholesterol (TC), triglycerides (TG), low-density lipoprotein cholesterol (LDL-C)].
  • b
    Safety Measures: Incidence of gastrointestinal adverse events (e.g., nausea, diarrhea), hypoglycemia episodes, diabetic ketoacidosis (DKA), and transaminase elevations (ALT/AST >2× upper limit of normal).

• (5)Language: Publications in English.

Two researchers pairs screened titles and abstracts retrieved from the searches, independent from each other, and assessed these for eligibility against the above inclusion criteria (PICOS).Two researchers pairs screened titles and abstracts retrieved from the searches, independent from each other, and assessed these for eligibility against the above inclusion criteria (PICOS).

Exclusion criteria

1. Population: Studies involving other types of diabetes, such as T2DM, Latent Autoimmune Diabetes or Mixed-Type Diabetes (LADA).

2. Data Availability: Case reports, letters, guidelines, animal studies, duplicate publications, studies with non-eligible interventions, age groups, or inaccessible full texts, or studies from which data could not be extracted.

Literature screening

Two researchers independently screened titles and abstracts using Endnote X9 following PICOS criteria. Studies passing the initial screening were downloaded for full-text review, followed by a secondary screening to exclude studies not meeting the inclusion criteria. Any discrepancies were resolved by a third researcher.

Data extraction

Two researchers independently extracted data from the full-text studies, with any discrepancies adjudicated by a third researcher. The extracted data included: author information, publication year, country, sample size, mean age, gender, intervention details, metformin dosage, comorbidities, and follow-up duration. The outcome measures assessed included: HbA1c, BMI, BMI-Z score, total daily insulin dose, insulin sensitivity, TC, TG, and LDL-C. Adverse outcome measures included gastrointestinal events, hypoglycemia incidence, DKA, and transaminase levels. For continuous variables, the mean difference (MD) and standard deviation (SD) between baseline and endpoint were calculated, with data being transformed using statistical tools [11, 12]. For categorical variables, the risk ratio (RR) and 95% confidence intervals (CI) were calculated.

Quality assessment

The researchers (Cheng Li, Lingyan Qiao) independently utilized the Cochrane Collaboration’s Risk of Bias tool, ROB2.0 [13]), to assess the risk of bias for each RCT. Any discrepancies were resolved by the investigator (Tang Li). The evaluation of bias risk was performed across various domains, including random sequence generation, allocation concealment, participant and personnel blinding, outcome assessment blinding, handling of incomplete data, selective reporting, and other sources of bias. All judgments were classified as “Yes” (low risk of bias), “Unclear” (indicating insufficient or uncertain information for assessing bias), or “No” (high risk of bias).

Data analysis

A Bayesian network meta-analysis was conducted using R 4.4.1 to evaluate the efficacy and safety of various metformin dosages as an adjunctive treatment to insulin in adolescents with T1D. The analysis encompassed five distinct metformin dosing regimens: ① 1.0 g/day; ② 1.7 g/day; ③ 2.0 g/day; ④<60 kg 1.0 g/day, ≥ 60 kg 2.0 g/day; ⑤<50 kg 1.0 g, 50–75 kg 1.5 g, ≥ 75 kg 2.0 g. The impart of these dosages on outcomes such as glycated hemoglobin, BMI, BMI-Z score, total daily insulin dose, insulin sensitivity, TC, TG, LDL-C, as well as adverse events such as gastrointestinal issues, hypoglycemic episodes, DKA, and liver enzyme levels were analyzed.

For continuous variables, the effect size was expressed as the mean difference (MD) with a 95% confidence interval (CI), whereas for binary variables, it was expressed as the relative risk (RR) with a 95% CI. Publication bias was assessed using a comparison-adjusted funnel plot. In the absence of publication bias or small sample effects, studies were expected to be symmetrically distributed around the regression line in the comparison-adjusted funnel plot. All results were presented as network evidence plots (where line thickness and circle size represent the number of studies and sample size, respectively) and forest plots. Heterogeneity among the included studies was assessed using the I² test. If I² < 50%, indicating low heterogeneity, a fixed-effects model was used for analysis; otherwise, a random-effects model was applied. A Bayesian random-effects model was constructed using the Markov Chain Monte Carlo (MCMC) method, with 20,000 pre-iterations and 50,000 actual iterations. The model’s convergence was evaluated by the potential scale reduction factor (PSRF), with values between 1.00 and 1.05 indicating satisfactory convergence. The cumulative ranking probability plot was generated, and the surface under the cumulative ranking (SUCRA) was calculated. A higher SUCRA value indicated a higher probability of a given intervention being the most effective. Pairwise comparisons of efficacy and safety risks for different doses of metformin were displayed using league tables, with continuous variables compared using MD and 95% CI. When 95% CI was entirely < 0 or > 0, a statistically significant difference in efficacy was indicated; when 95% CI crossed 0, no significant difference was observed. For categorical variables, RR and 95% CI were compared, with 0.90 < RR < 1.10 indicating similar risk, and RR with 95% CI entirely < 1.00 or > 1.00 indicating a statistically significant difference in risk.

Results

Literature selection

A total of 10,329 potentially relevant articles were initially identified through the pre-specified search strategy. Of these, 1,293 duplicates were removed using Endnote X9’s deduplication function. Following title/abstract screening against PICOS criteria, 8,999 non-relevant articles were excluded, yielding 37 studies for full-text review. Subsequent full-text evaluation led to the exclusion of 9 studies due to age mismatch (participants > 19 years or < 10 years), 10 due to inaccessible full-text availability, and 4 for missing outcome data (e.g., no reported HbA1c or BMI). Ultimately, 14 eligible RCTs were included in the final network meta-analysis. The study selection process is documented in the PRISMA-compliant flowchart provided in Fig. 1.

Fig. 1.

PRISMA Flow Diagram

Study characteristics

This study ultimately incorporated 14 studies [14–26], encompassing a total of 764 adolescents with diabetes. The patient distribution across different groups was as follows: 72 adolescents with diabetes in the 1.0 g/day metformin plus insulin group, 13 adolescents with diabetes in the 1.7 g/day metformin plus insulin group, 156 adolescents with diabetes in the 2.0 g/day metformin plus insulin group, 118 adolescents with diabetes in the weight-stratified metformin group (≤ 60 kg: 1.0 g/day; ≥60 kg: 2.0 g/day plus insulin), and 39 adolescents with diabetes in the multi-tiered weight-based metformin group (< 50 kg: 1.0 g; 50–75 kg: 1.5 g; ≥75 kg: 2.0 g plus insulin). The remaining 366 adolescents with diabetes were assigned to the placebo group. The baseline characteristics of the included studies are shown in Table S2.

Quality assessment

Quality assessment via the Cochrane Risk of Bias Tool revealed that 12 studies exhibited low risk across all domains, including random sequence generation, allocation concealment, participant/personnel blinding, outcome assessment blinding, incomplete outcome data handling, and selective reporting. Conversely, the study conducted by Yang et al. [22] was classified as high risk due to the absence of a clear method for allocation concealment. Similarly, the study by Nadeau et al. [21] was deemed potentially at high risk, as it failed to specify whether missing values were addressed (Fig. 2).

Fig. 2.

ROB Quality Assessment

Network meta-analysis of the efficacy of different doses of metformin

Network plot

The evidence network for efficacy outcomes across metformin doses is presented in Table S3 and Fig. 3. Nodes represent study sample sizes, with larger nodes indicating greater participant numbers. Solid lines between nodes denote direct comparisons of interventions, where line thickness corresponds to the number of studies supporting each comparison—thicker lines indicate more robust evidence.

Fig. 3.

Network Diagram. A. HBA1C; B. BMI; C. BMI-Z Score; D. TIDD; E. CACTI; F. TC; G. TG; H. LDL-C. Notes: HbA1c: Glycated Hemoglobin; BMI: Body Mass Index; TIDD: Total Insulin Dose per Day; CACTI: Insulin Sensitivity Indice; TC: Total Cholesterol; TG: Triglycerides; LDL-C: Low-Density Lipoprotein Cholesterol; Metformin weight1: < 60kg: 1.0g, ≥ 60kg: 2.0g; Metformin weight2: < 50kg: 1.0g, 50-75kg: 1.5g, ≥ 75kg: 2.0g

League table and cumulative probability ranking

HbA1c reduction

The network meta-analysis league table showed that all metformin dose groups had mean differences (MD) with 95% confidence intervals (CI) overlapping zero for HbA1c, indicating no statistically significant reduction compared to placebo (P > 0.05 for all). The SUCRA indicates that metformin at a dose of 2.0 g (76.11%) is superior to other dosages, including metformin 50 kg (66.57%), metformin 1.0 g (50.32%), metformin 1.7 g (34.89%), and placebo (25.91%) (Fig. 4).

Fig. 4.

Cumulative Probability Plot. A HBA1C; B BMI; C BMI-Z Score; D TIDD; E CACTI; F TC; (G) TG; (H) LDL-C Notes: HbA1c: Glycated Hemoglobin; BMI: Body Mass Index; TIDD: Total Insulin Dose per Day; CACTI: Insulin Sensitivity Indice; TC: Total Cholesterol; TG: Triglycerides; LDL-C: Low-Density Lipoprotein Cholesterol; Metformin weight1: < 60 kg: 1.0 g, ≥ 60 kg: 2.0 g; Metformin weight2: < 50 kg: 1.0 g, 50–75 kg: 1.5 g, ≥ 75 kg: 2.0 g

BMI and BMI-Z score reduction

For BMI (Table S4), the 2.0 g/day dose significantly reduced BMI compared to placebo (MD = −0.6 kg/m², 95% CI: −0.68 to −0.52; P < 0.05), with SUCRA ranking 2.0 g/day (78.32%) as most effective, followed by 1.0 g/day (61.68%), 50 kg weight-based dosing (45.83%), 1.7 g/day (42.51%), and placebo (21.67%) (Fig. 4). For BMI-Z scores (Table S5), the 60 kg weight-based dose (61.83%) outperformed 2.0 g/day (56.33%) and placebo (31.84%) (Fig. 4).

Daily insulin requirement reduction

A network meta-analysis league table comparing the effects of different doses of metformin on daily insulin usage (Table S6) reveals that, compared to Placebo, metformin at 1.0 g/day (MD = −0.14, 95% CI: −0.27 to 0; P < 0.05) as well as 2.0 g/day (MD = −0.09, 95% CI: −0.18 to 0; P < 0.05), and metformin dose groups based on a body weight of 60 kg (MD = −0.15, 95% CI: −0.27 to −0.02; P < 0.05) all contributed to a reduction in daily insulin usage, with statistically significant differences (P < 0.05). The SUCRA for insulin usage reduction ranksmetformin 60 kg (69.43%) > metformin 1.0 g (66.59%) > metformin 50 kg (62.30%) > metformin 1.7 g (49.58%) > metformin 2.0 g (43.97%) > Placebo (8.13%), indicating that the metformin dose based on a 60 kg body weight yielded the most optimal reduction in daily insulin usage (Table S7, Fig. 4).

Insulin sensitivity

Three studies statistically assessed the effects of varying doses of metformin on insulin sensitivity markers. This study employed a pairwise network meta-analysis, with the finding detailed in the league table (Table S8). Compared to placebo, metformin at a dosage of 1.0 g/day (based on a 60 kg body weight) did not result in a statistically significant improvement in insulin sensitivity. The SUCRA were as follows: metformin 60 kg (96.56%) > Placebo (29.06%) > metformin 1.0 g (24.39%) (Fig. 4).

Lipid profile changes

Additionally, a network meta-analysis league table comparing cholesterol-lowering effects of different metformin doses is shown in Table S9. Compared with placebo, both 2.0 g/day metformin (MD = −6.36 mg/dL, 95% CI: −8.60 to −4.16; P < 0.05) and the 50 kg weight-based regimen (MD = −7.82 mg/dL, 95% CI: −12.05 to −3.56; P < 0.05) significantly reduced total cholesterol (TC) levels. Cumulative probability rankings (SUCRA) demonstrated: metformin50kg (80.36%) > metformin2.0 g (65.07%) > metformin60kg (63.14%) > metformin1.0 g (29.16%) > placebo (12.23%), indicating the 50 kg weight-based dose was most effective for TC reduction (Fig. 4).

For triglycerides (TG, Table S9), no metformin dose group achieved statistically significant TG reduction compared to placebo (P > 0.05 for all). SUCRA rankings showed: metformin60kg (75.06%) > metformin1.0 g (61.29%) > metformin50kg (50.51%) > metformin2.0 g (33.65%) > placebo (29.50%), though these differences lacked clinical significance (Fig. 4).

In LDL-C analysis (Table S10), the 2.0 g/day metformin group exhibited the largest reduction (MD = −12.78 mg/dL, 95% CI: −21.17 to −0.49; P < 0.05), with SUCRA values ranking: metformin2.0 g (93.05%) > metformin60kg (68.20%) > placebo (41.19%) > metformin50kg (28.30%) > metformin1.0 g (19.25%), confirming its superiority in LDL-C lowering (Fig. 4).

Network meta-analysis of the safety of different doses of metformin

Network diagram

The nodes in the diagram represent the sample size of the included studies, with larger nodes corresponding to larger sample sizes. Solid lines connecting two points denote the interventions being compared, with the thickness of the solid lines reflecting the number of studies included; thicker lines indicate a greater number of studies (Fig. 5).

Fig. 5.

Network Diagram. A GIAE Network Diagram, B Hypoglycemia Network Diagram, C DKA Network Diagram, D ALT Network Diagram Notes: GIAE: Gastrointestinal Adverse Events; DKA: Diabetic Ketoacidosis; ALT: Alanine Aminotransferase; Metformin weight1: < 60 kg: 1.0 g, ≥ 60 kg: 2.0 g; Metformin weight2: < 50 kg: 1.0 g, 50–75 kg: 1.5 g, ≥ 75 kg: 2.0 g

League table and cumulative probability ranking

The primary adverse effects of metformin are manifested in gastrointestinal reactions, hypoglycemia, DKA, and elevated transaminases. This study analyzed the aforementioned data, and found that, compared to Placebo, the relative risk (RR) values for gastrointestinal reactions at different doses of metformin all included 1, indicating no statistically significant difference. Moreover, different doses of metformin did not lead to an increased incidence of gastrointestinal adverse events (Table S11). The cumulative probability rankings (SUCRA) are as follows: placebo (76.94%) > metformin 50 kg (60.55%) > metformin 1.0 g (39.80%) > metformin 2.0 g (39.89%) > metformin 60 kg (32.82%) (Fig. 6). Additionally, compared to Placebo, different doses of metformin did not lead to an increase incidence of hypoglycemia or DKA, nor did they significantly elevate transaminase levels. Comprehensive data are shown in Tables S12-14. Regarding hypoglycemia, the SUCRA indicate the following order: metformin 1.0 g (84.64%) > placebo (74.94%) > metformin 60 kg (24.12%) > metformin 2.0 g (16.30%) (Fig. 6). For DKA, the SUCRA are as follows: placebo (67.84%) > metformin 60 kg (67.08%) > metformin 2.0 g (44.91%) > metformin 1.0 g (20.17%) (Fig. 6). In terms of ALT, the SUCRA are as follows: metformin 60 kg (85.32%) > placebo (46.59%) > metformin 50 kg (18.09%) (Fig. 6).

Fig. 6.

Cumulative Probability Plot. A Gastrointestinal Reactions with Different Doses of Metformin; B Association Between Different Doses of Metformin and Hypoglycemia; C Association between Different Doses of Metformin and DKA; D Association between Different Doses of Metformin and ALT Notes: SUCRA: Surface Under the Cumulative Ranking Curve. GIAE: Gastrointestinal Adverse Events; DKA: Diabetic Ketoacidosis; ALT: Alanine Aminotransferase; Metformin weight1: < 60 kg: 1.0 g, ≥ 60 kg: 2.0 g; Metformin weight2: < 50 kg: 1.0 g, 50–75 kg: 1.5 g, ≥ 75 kg: 2.0 g

Publication bias

The funnel plot for the efficacy and safety outcomes of different doses of metformin from 14 studies is shown below (Fig. 7). The symmetrical distribution of each outcome indicator on both sides of the central axis of the funnel plot suggests a minimal likelihood of publication bias in the included studies.

Fig. 7.

Funnel Plot. A HBA1C, B BMI, C BMI-Z Score, D TIDD, E CACT, F ALT, G TC, H TG, I LDL, J GIAE, K. Hypoglycemia, L. DKA Note: HbA1c: Glycated Hemoglobin; BMI: Body Mass Index; TIDD: Total Insulin Dose per Day; CACTI: Insulin Sensitivity Indice; ALT: Alanine Aminotransferase; TC: Total Cholesterol; TG: Triglycerides; LDL-C: Low-Density Lipoprotein Cholesterol; GIAE: Gastrointestinal Adverse Events; DKA: Diabetic Ketoacidosis; Metformin weight1: < 60 kg: 1.0 g, ≥ 60 kg: 2.0 g; Metformin weight2: < 50 kg: 1.0 g, 50–75 kg: 1.5 g, ≥ 75 kg: 2.0 g

Discussion

T1D is defined by autoimmune-mediated pancreatic β-cell destruction, resulting in absolute insulin deficiency. Among adolescents with T1D undergoing long-term insulin replacement therapy, excessive weight gain and insulin resistance are prevalent, exacerbating metabolic disturbances and elevating the risk of cardiovascular complications [27]. The potential role of metformin, an insulin sensitizer, in the management of T1D has attracted increasing scholarly interest. The minimum effective dose of metformin is 500 mg/day, with the maximum dosage of the standard adult tablet formulation being 2550 mg/day. However, there is currently no consensus regarding the appropriate pediatric dosage of metformin. This study is pioneering in its analysis of the specific effects of metformin at different doses as an adjunct to insulin therapy in adolescents with T1D, thereby providing crucial evidence for clinical decision-making.

Metformin functions by reducing hepatic glucose production, enhancing peripheral glucose uptake and utilization, and acting on the intestines to inhibit glucose absorption by enterocytes, which promotes glucose excretion via the gastrointestinal tract, thereby improving insulin sensitivity in children with T1D [28, 29]. Additionally, metformin has been shown to suppress appetite, reduce energy intake, lower both baseline and postprandial insulin levels, and enhance leptin sensitivity, all of which contribute to weight reduction [30, 31]. In this study, a meta-analysis of 14 relevant studies revealed that a daily administration of 2.0 g of metformin resulted in a significant reduction in BMI or BMI-Z scores, consistent with previous research [17, 32].

Currently, there is no definitive conclusion regarding whether metformin can lower HbA1c levels in adolescents with T1D. A randomized, double-blind, placebo-controlled trial by Nadeau et al. reported no significant difference in HbA1c levels between the metformin group and the baseline after 6 months of treatment [23]. In contrast, Libman et al. observed a greater reduction in HbA1c levels in the metformin group compared to the placebo group at 13 weeks, although this difference was not sustained at 26 weeks. Additionally, study perfomed by Setoodeh et al. showed a significant reduction in HbA1c levels in the metformin group after 12 months of treatment [18]. In this study, the meta-analysis revealed that administration of metformin at doses of 1.0 g/day and 2.0 g/day, irrespective of whether the dosage was adjusted for a body weight of 60 kg–50 kg, resulted in a reduction in HbA1c levels, although the difference was not statistically significant.

Beyond its effects on reducing BMI and improving insulin resistance, metformin has been shown to enhance pancreatic β-cell function by promoting β-cell proliferation, reducing inflammatory responses, optimizing lipid metabolism, and alleviating endoplasmic reticulum stress. These mechanisms collectively contribute to the partial restoration of partial insulin secretion capacity [33, 34]. The present study corroborates previous findings by Meng [35], Liu [36], and Khalifah [37], demonstrating that metformin, when administered alongside insulin at doses of 1.0 g/day, 2.0 g/day, and body-weight-adjusted doses for a 60 kg individual, can reduce the requirement for exogenous insulin in adolescents with T1D. However, only three studies have reported using CACTIexa as an indicator of insulin sensitivity, with no significant improvement in insulin sensitivity observed when comparing the metformin groups with the placebo group, likely due to a small sample size.

Persistent insulin resistance frequently results in secondary mixed hyperlipidemia. This condition enhances the activity of hormone-sensitive lipase in adipose tissue, promoting TG breakdown and releasing free fatty acids, thereby raising blood lipid levels [38, 39]. Furthermore, insulin resistance can increase the production of very low-density lipoprotein (VLDL) and LDL in the liver while simultaneously inhibiting LDL receptor (LDL-R) expression, thus disrupting normal lipid metabolism [40, 41]. Elevated LDL accumulation in the arterial wall, along with the complex immune-inflammatory response it triggers [42], not only initiates atherosclerotic cardiovascular disease (ASCVD) but also plays a crucial role in the eventual formation of thrombi associated with atherosclerosis [43]. The results of this meta-analysis show that 2.0 g/day of metformin significantly reduces LDL levels, offering cardiovascular protection, in agreement with the studies by Bjornstad et al. [44–46]. No similar effect was observed with other doses of metformin. In addition to the prominent role of elevated LDL in ASCVD, TC elevation is an important risk factor for coronary artery disease (CAD). The Japanese JACC identified that individuals with TC levels ≥ 6.72 mmol/L exhibited a 3.74-fold increased risk of CAD mortality compared to those with TC levels < 4.14 mmol/L [47]. Similarly, the JPHC study showed that, in a multifactorial analysis, Japanese males with TC levels ≥ 6.21 mmol/L had a 1.63-fold higher risk of ischemic stroke compared to those with TC levels < 4.65 mmol/L, with the relative risk of large artery occlusion being 2.86 times higher [48]. This meta-analysis found that 2.0 g/day of metformin, or doses adjusted for a 50 kg body weight, significantly reduced TC levels, whereas other dosages did not exhibit a similar effect. TG, another critical risk factor for cardiovascular and cerebrovascular mortality in individuals with diabetes [49, 50], are implicated in various pathophysiological processes, including endothelial dysfunction, platelet aggregation, hypercoagulability, and free radical formation. However, this study found that none of the doses of metformin significantly reduced serum TG levels, which is consistent with the findings of Meng’s research [35].

Metformin is widely recognized for frequently inducing gastrointestinal reactions, such as nausea, vomiting, diarrhea, and dyspepsia, which may be linked to altered intestinal glucose uptake, microbiota dysregulation, and increased lactate production [51, 52]. However, this meta-analysis found no significant increase in GI adverse events with any metformin dose combined with insulin compared to placebo (P > 0.05). While GI tolerance is generally favorable, clinicians should monitor adolescents with T1D with preexisting gastrointestinal dysfunction.

Although subcutaneous insulin therapy is life-saving for T1D, it carries a risk of hypoglycemia in children. A prior study by Meng et al. [35] demonstrated that the metformin-insulin combination therapy might increase hypoglycemia risk in pediatric T1D, but our meta-analysis showed no significant elevation in hypoglycemic events with 1.0 g/day, 2.0 g/day, or 60 kg weight-based metformin doses, aligning with findings by Liu et al. [36] and Khalifah et al. [37]. Diabetic ketoacidosis (DKA) is the most common acute complication in adolescents with T1D and a leading cause of death in children with T1D [53, 54]. The progression of pancreatic β-cell dysfunction correlates with an elevated risk of DKA [55]. This study found that metformin doses of 1.0 g/day, 2.0 g/day, and those adjusted for a body weight of60kg did not increase the incidence of DKA, corroborating previous research findings [35]. Metformin is not metabolized by the liver, which reduces its potential for hepatotoxicity. However, adolescents with T1D with insulin resistance are prone to non-alcoholic fatty liver disease (NAFLD), characterized by hepatic fat accumulation and inflammatory responses that may elevate transaminase levels, thereby impairing liver function and limiting lactate clearance, which increases the risk of lactic acidosis. Therefore, children with serum transaminases levels exceeding three times the normal upper limit or with severe hepatic dysfunction should avoid metformin [56]. This meta-analysis showed that metformin, regardless of dosage (adjusted by weight of 60 kg–50 kg), did not adversely affect liver function, consistent with the findings of Jemma et al. [21].

In summary, compared to insulin monotherapy, metformin add-on therapy did not significantly reduce HbA1c but improved anthropometric and lipid profiles: 2.0 g/day metformin reduced BMI/BMI-Z scores, while 1.0 g/day, 2.0 g/day, and 60 kg weight-based doses lowered daily insulin requirements. Although insulin sensitivity remained unchanged, 2.0 g/day metformin significantly reduced LDL-C and total cholesterol, potentially mitigating cardiovascular risk. Across all doses, metformin demonstrated a favorable safety profile with no increased risk of GI events, hypoglycemia, DKA, or transaminitis.

Notably, this study is subject to several limitations. First, some heterogeneity in effect sizes in this meta-analysis may be related to differences in the study populations, including ethnicity, which affects the generalizability of the results and requires cautious interpretation. Second, the various studies employed different indices for calculating insulin sensitivity, making meta-analysis difficult. Third, the duration of the included RCTs was relatively short, and the long-term effects of metformin on T1D glucose metabolism, lipid metabolism, and complications are not yet conclusive. Of note, CGM-derived glycemic outcomes in this study were not presented in the included study reports. Therefore, the study results may not necessarily apply to all populations. For example, while metformin improved BMI and LDL-C (2.0 g/day), HbA1c remained unchanged and CGM outcomes were not available. Lastly, substantial heterogeneity in age distribution, sex ratios, and baseline BMI across cohorts hindered subgroup analyses exploring differential treatment responses by demographic or anthropometric characteristics.

To address these gaps, future research should prioritize large-scale, long-term RCTs with standardized insulin sensitivity metrics to validate our findings. Additionally, age- and BMI-stratified trials are warranted to characterize metformin’s effects on metabolic outcomes in specific T1D pediatric subgroups, refine therapeutic windows, and inform personalized treatment strategies for adolescents with varying baseline characteristics.

Conclusion

Metformin is a viable treatment option for adolescent T1D, with a daily dose of 2.0 g proving to be the most effective. It can reduce body weight, lower daily insulin requirements, improve metabolic indices such as LDL-C and TC, and is well-tolerated with good safety.

Authors’ contributions

All authors contributed to the study conception and design. Writing - original draft preparation: Cheng Li; Writing - review and editing: Cheng Li; Conceptualization: Cheng Li; Methodology: Cheng Li, Lingyan Qiao; Formal analysis and investigation: Cheng Li, Lingyan Qiao, Tang Li; Funding acquisition: Cheng Li; Resources: Cheng Li; Supervision: Cheng Li, and all authors commented on previous versions of the manuscript. All authors read and approved the final manuscript.

Funding

This research was supported by the Shandong Provincial Natural Science (ZR2021QH257).

Data availability

All data generated or analysed during this study are included in this published article.

Declarations

Ethics approval and consent to participate

Not applicable.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.