Correspondence to Dr Denice S Feig; [email protected]
STRENGTHS AND LIMITATIONS OF THIS STUDY
This systematic review and meta-analysis took a pragmatic approach and included children with metformin exposure for any pregnancy insulin-resistant condition to comprehensively capture all available data on this subject to date.
In addition to commonly reported adiposity measures of body mass index (BMI) z-score, BMI and weight, we also reported on clinically important categorical outcomes of childhood overweight and obesity following in utero metformin exposure.
Risk of bias was a significant concern in all included studies due to high attrition rates in randomised-controlled trial follow-up studies and concerns of ascertainment bias (exposure misclassification) in observational studies.
The oldest age at follow-up was up to 11 years, so this study is unable to assess the effects of intrauterine metformin exposure on adiposity trends in postpubertal adolescents or adults.
Introduction
Metformin is increasingly used in pregnancy for several conditions, including gestational diabetes (GD) and type 2 diabetes (T2D).1 When compared with usual care (ie, insulin) or placebo, metformin treatment in pregnancy did not increase congenital malformations and had several maternal and neonatal benefits.2–4 While its short-term safety and efficacy have been demonstrated in landmark randomised-controlled trials (RCTs), the long-term effects of intrauterine exposure metformin on offspring are less certain.1
Several follow-up studies have reported higher body mass index (BMI) and increased adiposity in children exposed to metformin in utero compared with children without exposure,5 6 raising concern for potential long-term metabolic harm, yet other studies have found no difference between the exposed and unexposed groups.7 In a prior systematic review and meta-analysis performed in 2019, Tarry-Adkins et al reported that prenatal exposure to metformin in GD pregnancies resulted in smaller neonates, and that metformin-exposed neonates were heavier as infants and had a higher BMI by mid-childhood, compared with those treated with insulin.8 However, this review only focused on studies of metformin use for GD women and included only follow-up studies of RCTs, two studies (n=411) of infants aged 18 months to 2 years and two studies (n=301) of children aged 5–10 years in the meta-analysis.8 The overall small number of studies available limited the authors’ ability to draw definitive conclusions.8 Furthermore, as other maternal indications for metformin use aside from GD were excluded from the review, such as T2D and polycystic ovary syndrome (PCOS), it is unknown whether the concerns are generalisable to all offspring exposed to metformin in utero.
To further explore the risk of long-term metabolic harm in offspring exposed to metformin in utero, we conducted a systematic review and meta-analysis using a pragmatic approach by including children of women with GD, T2D, PCOS and non-diabetic overweight/obesity, and following them up to 11 years of age. In addition to RCTs and their follow-up studies, we also included cohort studies to capture the full scope of literature published on this topic to date.
Objectives
The aim of this systematic review was to compare long-term adiposity outcomes in those with and without intrauterine metformin exposure among offspring of insulin-resistant pregnancies. Eligible pregnancies included those with PCOS, GD, T2D and non-diabetic overweight/obesity.
Our primary objective was to evaluate whether offspring of insulin-resistant pregnancies who were exposed to metformin in utero differed in BMI z-score/BMI/weight, overweight, obesity, waist circumference and sum of skinfold thickness compared with those who were not exposed at 12 months to 11 years of age.
The secondary objectives were (1) to compare BMI z-score between metformin-exposed and non-exposed children at three age ranges: (a) 12 months to 3 years, (b) 3–6 years and (c) 6–11 years; and (2) to assess the extent to which the difference in BMI z-score between metformin-exposed and non-exposed children differs depending on the maternal diagnosis (diabetes, PCOS or non-diabetic overweight/obesity).
Research design and methods
This systematic review was conducted in accordance with the Cochrane Handbook for Systematic Reviews of Intervention9 and reported in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) 2020 Statement and Meta-analysis of Observational Studies in Epidemiology guidelines.10 11 The protocol was registered with the PROSPERO International Prospective Register of Systematic Reviews (http://www.crd.york.ac.uk/prospero/) with identifier CRD42023394464.
Eligibility criteria
Peer-reviewed publications, including RCTs, offspring follow-up studies of RCTs, case–control studies and other prospective and retrospective comparative cohort studies, were included. Studies without a comparison group, case reports, case series, review articles, commentaries and other meta-analyses were excluded. Conference abstracts, posters and non-published theses were excluded. The comparator group may have received usual care (ie, treatment with insulin, treatment with another oral glucose-lowering agent aside from metformin such as a sulfonylurea or lifestyle management alone), placebo or no treatment.
Studies included must have reported on offspring of pregnancies complicated by a maternal condition that provided a potential indication for metformin use, including GD, T2D, non-diabetic overweight/obesity and PCOS. Studies included must have also explicitly mentioned metformin use in pregnancy that continued beyond the first trimester. Any dosage of metformin could have been used. We excluded studies with metformin use only in the preconception setting or limited to the first trimester as exposures in these settings are not the focus of our review. Studies included must have reported on offspring adiposity outcomes, assessed between 12 months and 11 years of age, in the groups with and without metformin exposure. Studies reporting on offspring of pregnancies before 12 months of age only were excluded as neonatal/infant, and short-term outcomes of metformin exposure were not the focus of our review.
Adiposity outcomes of interest were BMI z-score, BMI, weight, at risk of overweight, overweight, obesity, waist circumference and sum of skinfolds thickness. At risk for overweight, overweight and obesity were defined as BMI z-score>1 SD, >2 SD or >3 SD above the WHO standard growth curve, respectively, unless otherwise defined by the authors. Waist circumference, a measure of central adiposity, was defined by taking the circumference in the standing position at the umbilicus or the midpoint between the lower rib and top of the iliac crest. Sum of skinfold thickness, a measure of subcutaneous fat, was determined by the thickness of two layers of subcutaneous fat pinched together, taken at two or more sites using a skinfold calliper.
The following online databases were searched from inception to 1 November 2023: Medline (OVID interface, 1948 onwards), Embase (OVID interface, 1980 onwards) and Cochrane Central Register of Controlled Trials (Wiley interface, current issue). References of all eligible articles were also reviewed for additional studies to be screened. An updated search was conducted on 4 October 2024 prior to publication.
Literature search strategies were developed in collaboration with an information specialist using both medical subject headings (MeSH) and non-MeSH search terms. Three main concepts were reflected: concept 1, population of interest (eg, infants, children or paediatric); concept 2, metformin exposure (eg, metformin or biguanide); and concept 3, period of exposure (eg, prenatal, maternal or intrauterine). The three concepts were then combined with the ‘AND’ operator to obtain final search results. We limited studies to only those studying human subjects. No year of publication or language restrictions were applied. The final search strategy for Medline, Embase and Cochrane Central is provided in online supplemental materials table 1-3.
Two reviewers (JF and NT) independently screened all eligible title/abstracts for full-text review and independently reviewed full-text articles for final inclusion according to the prespecified eligibility criteria using Covidence software.12 Discrepancies in study selection were resolved through discussion with a third reviewer (DSF).
Data were independently extracted from the included studies by two reviewers (JF and NT) using a standardised data collection form on Microsoft Excel. The initial data collection form was piloted on two studies to ensure robustness. Reviewers resolved disagreements by discussion, and unresolved disagreements were adjudicated by a third reviewer (DSF).
The following study characteristics were extracted: study design, age or age range of study subjects, country of study, maternal diagnosis, dosage and duration of metformin exposure, details of the comparator exposure (placebo or another medication with the inclusion of dosage and duration if applicable), reported adiposity outcomes and whether any covariates were adjusted for in the analysis.
Risk-of-bias assessment of individual studies
The quality of all included studies was assessed independently by two study authors (JF and NT). Differences in ratings were resolved through discussion with a third author (DSF). For RCTs and follow-up studies, the Cochrane Risk-Of-Bias tool V.2 for randomised trials (RoB V.2) was used.13 For observational studies, the Risk of Bias in Non-Randomised Studies of Exposure (ROBINS-E) tool was used.14 Traffic-light plots were generated using the Risk-of-bias VISualization tool.15
Effect measures for outcomes
For the primary outcome of BMI z-score, BMI or weight at the oldest age reported and the secondary outcomes of BMI z-score, BMI or weight at 12 months to 3 years, 3–6 years and 6–11 years, a stepwise approach was used based on the information available:
BMI z-score, along with SD and sample size, was extracted from eligible studies wherever reported.
If a study did not report BMI z-score, BMI, along with SD and sample size, was extracted instead of BMI z-score. If a study did not report either BMI z-score or BMI, weight, along with SD and sample size, was extracted instead.
To allow for pooling of both BMI and BMI z-scores, the standardised mean difference (SMD) was used as the summary measure in the meta-analysis.
For binary secondary outcomes (overweight, obesity), pooled OR and 95% CIs were estimated, and for continuous secondary outcomes (waist circumference and sum of skinfolds thickness), weighted mean differences and 95% CIs were calculated. For studies that reported only median and IQR, mean and SD were estimated from these parameters using the methodology proposed by Wan et al.16
Data synthesis
Meta-analyses were conducted on any prespecified outcome that was reported in at least three studies using the generic inverse variance method in a random-effects model. Review Manager 5.4 software17 was used to generate forest plots and calculate summary effect measures, along with 95% CI, using the DerSimonian and Laird method.
When multiple publications reporting data from the same cohort of children met eligibility criteria, we included only the study reporting the oldest age group for our primary outcome to avoid duplication as we were most interested in the long-term effects of metformin exposure. For the secondary outcomes which are divided by the age of the child, studies reporting outcomes at earlier age groups were included in the appropriate follow-up age intervals. For superiority RCTs, whenever possible, we used effect sizes from intention-to-treat analysis over per-protocol analysis. Whenever possible, covariate-adjusted effect sizes were used over crude effect sizes.
To assist with the interpretation of results and understanding the clinical significance of findings, SMDs for the primary outcome were then back-converted to the original scales of BMI and BMI z-score by calculating the pooled baseline SD of all studies reporting the outcome using the following formula: MD=SMD × (pooled baseline SD).
χ2 test and the I2 statistic were used to assess statistical heterogeneity. Subgroup and sensitivity analyses were performed to further characterise sources of heterogeneity. A prespecified subgroup analysis was performed based on maternal diagnosis. Studies reporting on BMI z-score or BMI were divided into three subgroups based on whether the maternal diagnosis was GD or T2D (diabetes subgroup), PCOS or obesity. The Instrument for Assessing the Credibility of Effect Modification Analyses (ICEMAN) tool was used to assess the quality of subgroup effect, with the overall effect assigned a credibility rating ranging from very low to high.18
Several sensitivity analyses were performed to examine the robustness of study results:
Low risk-of-bias studies only: studies determined to be ‘high risk of bias’ or ‘very high risk of bias’ were excluded.
By study design: only follow-up studies of RCTs were included.
Funnel plots were constructed for visual inspection and Egger’s tests19 were performed if ≥10 studies were available for evidence of publication bias.
Assessment of overall certainty of evidence
The Grading of Recommendations Assessment, Development and Evaluation (GRADE) instrument was used to determine the quality of evidence for all outcomes.20 Quality was determined as follows: high (ie, further research is very unlikely to change our confidence in the estimate of effect), moderate (ie, further research is likely to have an important impact on our confidence in the estimate of effect and may change the estimate), low (ie, further research is very likely to have an important impact on our confidence in the estimate of effect and is likely to change the estimate) or very low (ie, very uncertain about the estimate of effect).
Patient and public involvement
Patients or the public were not involved in the design, or conduct, or reporting, or dissemination plans of our research.
Results
Study selection
Systematic search strategies generated a total of 856 abstracts, out of which 198 were duplicates. After reviewing 658 non-duplicate abstracts, 30 studies were included for full-text review. Out of these, three studies were excluded for reporting on the same offspring cohorts as other included studies, two were excluded for not reporting on any of the primary or secondary outcomes of interest, one was excluded due to being a conference abstract without accompanying peer-reviewed publication, five were excluded for having either an incorrect exposure or comparison group, and one was excluded for being a study protocol only without results available. The PRISMA flow diagram is shown in online supplemental figure 1 based on the original search from November 2023. No additional eligible studies were identified following the search update in October 2024.
Study characteristics
A total of 18 studies reporting on 7975 children exposed to metformin in utero and >1 million children without metformin exposure in utero were included in at least one meta-analysis.5–7 21–35 14 of the studies5–7 22–24 26 27 30–35 were follow-up studies of 9 RCTs (four conducted in mothers with GD,2 32 36 37 one in T2D,3 one in PCOS38 and three in obesity).39–41 Four studies21 25 28 29 were cohort studies conducted using information from administrative databases. The oldest offspring age at follow-up was 11 years. Table 1 provides the characteristics of all included studies.
Table 1Characteristics of included studies
| Study ID | N (T/C) | Study design | Maternal diagnosis | Data source | Treatment versus control | Offspring age at FU | Outcomes reported |
| Rowan et al33 | 154/164 | RCT-FU | GD | MiG trial (Australia/NZ; n=751) | Metformin versus insulin* | 2 years | Weight, waist circumference |
| Rowan et al6 | 58/51 | RCT-FU | 7–9 years | BMI, waist circumference | |||
| Ijas et al5 | 45/48 | RCT-FU | Two RCTs in Finland (n=100 at Oulu University Hospital; n=221 at Turku University Hospital) | Metformin versus insulin* | 1.5 years (Turku only) | Obesity | |
| Tertti et al34 | 25/27 | RCT-FU | 3–6 years (Oulu only) | BMI z-score | |||
| Paavilainen et al30 | 82/90 | RCT-FU | 9 years (combined Turku and Oulu RCTs) | BMI, overweight, obesity, waist circumference | |||
| Landi et al28 | 1996/1932 | Retrospective cohort | Administrative database (NZ) | Metformin versus insulin* | 4 years | BMI z-score, overweight, obesity | |
| Paul et al32 | 41/37 | RCT-FU | RCT in India (n=159) | Metformin versus glibenclamide* | 9 years | BMI, waist circumference | |
| Martine-Edith et al29 | 76/420 | Retrospective cohort | Single-centre birth cohort (UK) | Metformin versus insulin | 5 years | BMI z-score | |
| Feig et al24 | 111/115 | RCT-FU | T2D | MiTy trial (Canada/Australia; n=502) | Metformin versus placebo† | 2 years | BMI z-score, overweight, obesity |
| Carlsen et al22 | 102/94 | RCT-FU | PCOS | PregMet trial (Norway; n=257 in main trial, n=40 in pilot study) | Metformin versus placebo | 1 year (main trial only) | Weight |
| Hanem et al27 | 81/79 | RCT-FU | 4 years (main trial+pilot) | BMI z-score, overweight | |||
| Hanem et al26 | 69/70 | RCT-FU | 5–10 years (main trial only) | BMI z-score, overweight, obesity, waist circumference | |||
| Rø et al7 | 12/13 | RCT-FU | 7–9 years (pilot only) | BMI z-score | |||
| Deussen et al23 | 156/147 | RCT-FU | Obesity | GRoW trial (Australia; n=524) | Metformin versus placebo | 3 years | BMI z-score, overweight, obesity |
| Panagiotopoulou et al31 | 77/74 | RCT-FU | MOP trial (UK; n=400) | Metformin versus placebo | 3 years | BMI, waist circumference | |
| Yang et al35 | 19/21 | RCT-FU | EMPOWaR trial (UK; n=449) | Metformin versus placebo | 5 years | BMI, obesity | |
| Brand et al21 | 3950/5264 | Retrospective cohort | GD, T2D or PCOS | Administrative database (Finland) | Metformin versus insulin | 1–11 years | Obesity |
| Fornes et al25 | 7899/1 020 342 | Retrospective cohort | PCOS or non-PCOS | Administrative database (Sweden) | Metformin versus no treatment | <11.5 years | Obesity |
*In the GD trials, participants randomised to non-insulin antihyperglycaemics (eg, metformin or glibenclamide) were started on insulin in addition if glycaemic targets were not met.
†In the MiTy trial, all participants were already on insulin prior to randomisation to either metformin or placebo.
BMI, body mass index; GD, gestational diabetes; N (T/C), sample size (treatment group/control group); NZ, New Zealand; PCOS, polycystic ovarian syndrome; RCT-FU, randomised-controlled trial follow-up study; T2D, type 2 diabetes.
Risk-of-bias assessment
All 14 RCT follow-up studies were rated as ‘high risk of bias’ overall when assessed using the RoB V.2 tool. Concerns arose primarily from domain 3 (ie, missing outcome data) as attrition rates in all studies were high and ranged from 22% to 76.5% of eligible children and there was no convincing evidence that attrition rates did not differ based on the true value of the outcome. In addition, some concerns arose from domain 5 (ie, bias in selected reporting of outcomes) as only 4 of the 14 studies were clearly preplanned at the time of the original trial and others were post hoc analyses. Traffic-light plots of the included studies assessed using RoB V.2 are provided in online supplemental figure 2.
Four cohort studies conducted using administrative database information were assessed using the ROBINS-E tool. All scored ‘high risk of bias’ overall based on issues in domain 2 (ie, bias arising from the measurement of the exposure), as all relied on prescription dispensary records which may not indicate actual usage of the medication during pregnancy, and domain 5 (ie, bias due to missing data). Traffic-light plots of the included studies assessed using ROBINS-E are provided in online supplemental figure 3.
Main results
In the analysis of the primary outcome BMI z-score, BMI or weight at the oldest age reported, 11 studies reporting on 12 independent cohorts of children were included (figure 1). Children exposed to metformin in utero did not differ in BMI z-score, BMI or weight compared with their non-exposed peers (SMD −0.02; 95% CI: −0.11, 0.07). The estimated weighted mean difference (MD) for BMI was −0.06 kg/m2 (95% CI: −0.31, 0.20) and for BMI z-score was −0.06 (95% CI: −0.13, 0.08). No significant heterogeneity was detected in the studies included (I2=24%). The weighted mean age at follow-up was 4.4 years for metformin-exposed children and 4.5 years for controls.
Enlarge this image.Figure 1. Forest plot of body mass index (BMI) z-score/BMI in children with and without metformin exposure in utero at the oldest age reported.
In the analysis of the secondary outcome by age at follow-up (figure 2A–C), BMI z-score, BMI or weight was higher in metformin-exposed children compared with their peers at 1 to <3 years of age (SMD 0.15; 95% CI: 0.04, 0.27). No difference was found at 3 to <6 years of age (SMD −0.04; 95% CI: −0.18, 0.11) or at 6 to <11 years of age (SMD 0.14; 95% CI: −0.04, 0.33). No significant heterogeneity was detected in the studies reporting these outcomes.
Enlarge this image.Figure 2. Forest plot of body mass index (BMI) z-score/BMI in children with and without metformin exposure in utero by age at follow-up. (A) 12 months to 3 years, (B) 3-6 years and (C) 6-11 years.
In the analysis of other secondary adiposity outcomes, no significant difference was found in overweight (pooled OR 0.92; 95% CI: 0.77, 1.09), obesity (pooled OR 1.33; 95% CI: 0.98, 1.83) or waist circumference (weighted MD 0.90 cm; 95% CI −1.12, 2.92) among metformin-exposed children and controls at the oldest age reported (online supplemental figures 4–6). A high degree of heterogeneity was found between the studies for waist circumference (I2=68%), likely owing to the different ages of the children reported by the four studies (eg, Panagiotopoulou et al31 reported waist circumference at 3 years, while the other three studies reported this outcome at 5–10 years). Only two studies reported the sum of skinfolds,23 24 thus meta-analysis was not completed. Both studies individually reported no difference in this outcome in metformin-exposed and non-exposed children.23 24
In the prespecified subgroup analysis by maternal diagnosis (figure 3), a significant difference was found in BMI z-score and BMI across the subgroups (p=0.03). In the diabetes subgroup (weighted mean age 4.4 years), no difference was found between metformin-exposed and non-exposed children (SMD −0.02; 95% CI: −0.11, 0.07). In the PCOS subgroup (weighted mean age 7.6 years), metformin-exposed children were heavier than non-exposed peers (SMD 0.31; 95% CI: 0, 0.62), while in the obesity subgroup (weighted mean age 3.5 years), metformin may have had a protective effect as exposed children had statically non-significant lower BMI z-score and BMI than their peers (SMD −0.17; 95% CI: −0.34, 0.01). The weighted MD for BMI was 0.88 kg/m2 (95% CI: 0, 1.77) and −0.48 kg/m2 (95% CI: −0.97, 0.03) for the PCOS and obesity subgroups, respectively. Using the ICEMAN tool, the subgroup effect by maternal diagnosis was found to be of low credibility, as all comparisons were between-trials rather than within-trial, there was an overall small number of studies in each subgroup (eg, PCOS subgroup only contained the PregMet trial) and p value was between 0.01 and 0.05 on the interaction test.
Enlarge this image.Figure 3. Subgroup analysis of body mass index (BMI) z-score/BMI in children with and without metformin exposure in utero by maternal diagnosis.
Within the diabetes subgroup, metformin-exposed children did not differ in the primary adiposity outcome from non-exposed peers when only RCTs were included (SMD 0.03; 95% CI −0.13, 0.18; online supplemental figure 7), when only GD studies were included (SMD −0.02; 95% CI −0.12, 0.09; online supplemental figure 8) or when only RCTs studying GD were included (SMD 0.09; 95% CI −0.13, 0.32; online supplemental figure 9).
For sensitivity analyses, the prespecified analysis using only studies with a low risk of bias could not be completed as all studies reporting the primary outcome had a high risk of bias. Excluding non-RCT study design (ie, removal of the studies by Landi et al28 and Martine-Edith et al29) did not alter the results of the primary outcome as no difference was found among metformin-exposed children and controls in BMI z-score or BMI at the oldest age reported (SMD 0.03; 95% CI: −0.11, 0.16).
Reporting biases
Funnel plots were constructed and Egger’s regression tests were performed for any outcome comparison involving 10 or more studies and provided in online supplemental figure 10. This was satisfied only by the primary outcome. No publication bias was found on visual inspection or based on Egger’s regression test (p=0.23).
Certainty of Evidence using GRADE
Assessment of certainty of evidence using GRADE is shown in online supplemental table 4. Owing primarily to a very serious risk of bias in all the included studies, there is very low to low certainty of evidence to support the finding of no difference in long-term adiposity outcomes among children with and without metformin exposure in utero.
Discussions
Main findings
This meta-analysis of RCT follow-up and observational cohort studies reporting on long-term offspring adiposity outcomes following intrauterine exposure to metformin found no difference in continuous outcomes of BMI z-score and BMI as well as categorical outcomes of overweight and obesity in children exposed to metformin compared with their peers at the oldest age reported of each available cohort to date. When studies were stratified by age of offspring at follow-up, while metformin-exposed offspring were heavier in infancy, the difference dissipated by mid to late childhood. When stratified by maternal diagnosis, metformin exposure was not associated with an increase in childhood adiposity in women with diabetes or non-diabetic overweight/obesity but was increased in offspring of mothers with PCOS.
Prior to our study, two meta-analyses were published that explored the relationship between intrauterine metformin exposure and childhood BMI or BMI z-score.8 42 The Tarry-Adkins et al meta-analysis found that metformin-exposed offspring of GD mothers had higher BMI than their non-exposed peers at 5–9 years8, and the van Weelden et al meta-analysis found metformin-exposed offspring of GD and PCOS mothers were heavier.42 Contrary to their findings, we found no difference in the primary adiposity outcome of BMI/BMI z-score/weight between metformin-exposed and non-exposed children of mothers with GD, PCOS and non-diabetic obesity at the oldest age reported. The discrepancy could be explained by broadening inclusion criteria to include offspring of mothers with non-diabetic overweight/obesity and the inclusion of observational studies as well as newer RCT follow-up studies published since 2019. When we replicated the analysis by Tarry-Adkins et al8 by restricting studies to only RCT follow-up studies of GD, we also found no difference in BMI z-score/BMI between metformin-exposed children and controls, which can be explained by the inclusion of an additional RCT follow-up study by Paul et al32 not included in their analysis and replacing the adiposity data from Tertti et al34 (follow-up at 3–7 years) with that reported by Paavilainen et al30 (the same cohort of children, followed to 9 years).
Another notable finding of our study is that, while metformin-exposed children were heavier compared with non-exposed peers before age 3, the difference in adiposity parameters did not persist beyond this age. This finding of increased adiposity in infancy is consistent with the Tarry-Adkins et al meta-analysis, which found metformin-exposed babies were born smaller but underwent accelerated postnatal growth and were heavier by 18–24 months8. However, contrary to their finding of continued increased adiposity of metformin-exposed offspring by 5–9 years8, we found that intrauterine metformin exposure had a time-limited effect on offspring adiposity that dissipated by mid-childhood. This was true for both metformin-exposed offspring of GD mothers and metformin-exposed offspring of any diagnosis. While this can be interpreted as a reassuring finding, early growth acceleration and catch-up growth in infancy can be harmful as they have been associated with long-term risks of obesity and metabolic disease in adulthood.43 As well, in older children there was a trend to increased adiposity although this was not statistically significant. Longer term follow-up studies, particularly beyond puberty, are needed to definitively address this concern.
When studies were stratified by maternal diagnosis, we found that metformin-exposed children were heavier than non-exposed peers in the PCOS subgroup, while in the obesity subgroup, metformin appears to have had a protective effect on exposed children with a non-significant lower BMI z-score and BMI. The reason for these differences is not clear. This subgroup difference is unlikely to be explained by the duration of in utero metformin exposure as both the PCOS and obesity RCTs began metformin exposure at gestational age 12–18 weeks, whereas only children of GD mothers would have had metformin exposure limited to the third trimester. One possible explanation for this subgroup effect is baseline differences in maternal BMI. Among the three subgroups, maternal BMI was the highest in the overweight/obesity trials followed by the GD/T2D trials and lowest in the PCOS subgroup. Metformin-treated women tend to have less gestational weight gain in pregnancy; therefore, it is conceivable that a higher proportion of women with lower BMI may have gained less gestational weight than outlined in guideline recommendations for their BMI category. As speculated by Rowan et al in the MiG TOFU follow-up study6 and demonstrated in several mouse models,44 45 metformin may have adverse effects on offspring adiposity if the nutrient environment for the fetus is restricted. Metformin-exposed offspring of mothers with PCOS are potentially more likely to be ‘undernourished’ in utero compared with the diabetes or obesity subgroups, and hence experience a more pronounced difference in childhood adiposity, although this is speculative. Another explanation could be that the PCOS subgroup was followed longer than other subgroups, with a weighted mean age of 7.6 years, compared with 4.4 years for the diabetes subgroup and 3.5 years for the obesity subgroup. This age difference could be clinically significant as a child’s BMI trajectory usually entails rapid increases within the first year of life, followed by a progressive decline reaching a nadir around 6 years of age, and then a second rise throughout later childhood known as the adiposity rebound.46 While children in the PCOS subgroup have passed their adiposity rebound, children in the diabetes and obesity subgroups are likely to have yet to undergo this stage and may reveal longer term adiposity impacts of metformin exposure. It is worth noting also that the PCOS subgroup had a smaller sample size (n=164 compared with n=494 in the obesity subgroup and n=5108 for the diabetes subgroup) and was drawn from only one cohort, the PregMet cohort.7 26 Therefore, more data are needed to confirm these findings.
Our study has several strengths. Compared with other meta-analyses conducted on this subject to date,8 42 we included data from five additional RCT follow-up studies published since 2019 on offspring of GD mothers and nine additional observational studies or RCT follow-up studies with maternal diagnoses other than GD. To our knowledge, this is the first systematic review and meta-analysis exploring the relationship between intrauterine metformin exposure and risk of childhood overweight and obesity as categorical outcomes. We also assessed infant exposure according to maternal diagnosis and according to study design. By expanding our review to also include children of other maternal insulin-resistant conditions in pregnancy (eg, T2D, PCOS and non-diabetic overweight/obesity), our conclusions are generalisable to all offspring with metformin exposure in utero. However, there are some limitations to our study. Out of the 14 RCT follow-up studies included, all had a high risk of bias on the RoB V.2 tool on the basis of high attrition rates and concerns of selective outcome reporting. This may be reflective of the difficulty in conducting follow-up studies >5 years past the original trial and getting child participants to return to study centres for adiposity measurements. Many of the studies do not report on baseline characteristics between those lost to follow-up vs those included in the study. In addition, there are other well-established factors such as parental BMI, maternal gestational weight gain, nutrition, exercise, socioeconomic status and screen time that can impact childhood adiposity. These important factors could have had a confounding effect on the study outcomes. Out of the included studies, only five studies adjusted for these covariates, raising the possibility that the difference in adiposity seen in early childhood may be explained by confounders rather than intrauterine metformin exposure. At last, we acknowledge there are also methodological issues common to all meta-analyses, such as publication bias, which may result in an under-reporting of negative or neutral observational studies. However, based on funnel plot and Egger’s test, no obvious publication bias was observed.
Conclusions
Our study found that there was no significant difference in offspring adiposity among children with and without intrauterine metformin exposure followed up until 11 years of age. This adds to the growing body of evidence supporting the safety of metformin use in pregnancy on long-term offspring outcomes. However, the overall certainty of evidence is low as studies included had a high risk of bias. More research is needed with more diverse patient populations in terms of ethnicity and socioeconomic status, and longer follow-up of the offspring, particularly into adolescence and adulthood, before definitive conclusions can be drawn about the long-term safety of intrauterine metformin exposure.
The authors would like to thank Elizabeth Uleryk, Information Specialist at the University of Toronto, for her help in developing and performing the initial search strategy.
Data availability statement
Data are available upon reasonable request.
Ethics statements
Patient consent for publication
Not applicable.
Ethics approval
Not applicable.
Contributors JF and DSF conceived and designed the research question. JF and NT independently screened studies for inclusion and conflicts were resolved through consensus or by discussion with DSF. JF and GT performed statistical analysis. All authors have made substantial contributions to the analysis and interpretation of data; have been involved in drafting the manuscript or revising it critically for important intellectual content; have given final approval of the version to be published; have participated sufficiently in the work to take public responsibility for appropriate portions of the content and have agreed to be accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved. DSF is responsible for the overall content as the guarantor.
Funding The authors have not declared a specific grant for this research from any funding agency in the public, commercial or not-for-profit sectors.
Competing interests DSF reports in-kind donations of metformin and placebo from Apotex for the MiTy trial. Other unrelated grants include those from the Canadian Institute of Health Research, an investigator-initiated grant from Dexcom, in-kind donations from Dexcom and Tandem and honoraria from Novo Nordisk for speaking and from Ypsomed for participating in an Advisory Panel. JH participates in unrelated industry trials (Rhythm Pharmaceuticals, Eli Lilly) and has received honoraria for participation in advisory panel (Novo Nordisk, Rhythm Pharmaceuticals). JH holds the University of Toronto Mead Johnson Chair in Nutritional Science which provides unrestricted research funds. JF, NT, KEM and GT have no conflicts of interest to declare.
Patient and public involvement Patients and/or the public were not involved in the design, or conduct, or reporting, or dissemination plans of this research.
Provenance and peer review Not commissioned; externally peer reviewed.
Supplemental material This content has been supplied by the author(s). It has not been vetted by BMJ Publishing Group Limited (BMJ) and may not have been peer-reviewed. Any opinions or recommendations discussed are solely those of the author(s) and are not endorsed by BMJ. BMJ disclaims all liability and responsibility arising from any reliance placed on the content. Where the content includes any translated material, BMJ does not warrant the accuracy and reliability of the translations (including but not limited to local regulations, clinical guidelines, terminology, drug names and drug dosages), and is not responsible for any error and/or omissions arising from translation and adaptation or otherwise.
1 Newman C, Dunne FP. Metformin for pregnancy and beyond: the pros and cons. Diabet Med 2022; 39: e14700. doi:10.1111/dme.14700
2 Rowan JA, Hague WM, Gao W, et al. Metformin versus insulin for the treatment of gestational diabetes. N Engl J Med 2008; 358: 2003–15. doi:10.1056/NEJMoa0707193
3 Feig DS, Donovan LE, Zinman B, et al. Metformin in women with type 2 diabetes in pregnancy (MiTy): a multicentre, international, randomised, placebo-controlled trial. Lancet Diabetes Endocrinol 2020; 8: 834–44. doi:10.1016/S2213-8587(20)30310-7
4 Abolhassani N, Winterfeld U, Kaplan YC, et al. Major malformations risk following early pregnancy exposure to metformin: a systematic review and meta-analysis. BMJ Open Diabetes Res Care 2023; 11: e002919. doi:10.1136/bmjdrc-2022-002919
5 Ijäs H, Vääräsmäki M, Saarela T, et al. A follow‐up of a randomised study of metformin and insulin in gestational diabetes mellitus: growth and development of the children at the age of 18 months. BJOG 2015; 122: 994–1000. doi:10.1111/1471-0528.12964
6 Rowan JA, Rush EC, Plank LD, et al. Metformin in gestational diabetes: the offspring follow-up (MiG TOFU): body composition and metabolic outcomes at 7-9 years of age. BMJ Open Diabetes Res Care 2018; 6: e000456. doi:10.1136/bmjdrc-2017-000456
7 Rø TB, Ludvigsen HV, Carlsen SM, et al. Growth, body composition and metabolic profile of 8-year-old children exposed to metformin in utero. Scand J Clin Lab Invest 2012; 72: 570–5. doi:10.3109/00365513.2012.712319
8 Tarry-Adkins JL, Aiken CE, Ozanne SE. Neonatal, infant, and childhood growth following metformin versus insulin treatment for gestational diabetes: A systematic review and meta-analysis. PLoS Med 2019; 16: e1002848. doi:10.1371/journal.pmed.1002848
9 Cochrane handbook for systematic reviews of interventions. Available: https://training.cochrane.org/handbook/current [Accessed Feb 2023 ].
10 Page MJ, McKenzie JE, Bossuyt PM, et al. The PRISMA 2020 statement: an updated guideline for reporting systematic reviews. BMJ 2021; 372: n71. doi:10.1136/bmj.n71
11 Stroup DF, Berlin JA, Morton SC, et al. Meta-analysis of observational studies in epidemiology: a proposal for reporting. Meta-analysis Of Observational Studies in Epidemiology (MOOSE) group. JAMA 2000; 283: 2008–12. doi:10.1001/jama.283.15.2008
12 Covidence systematic review software, Available: www.covidence.org
13 Sterne JAC, Savović J, Page MJ, et al. RoB 2: a revised tool for assessing risk of bias in randomised trials. BMJ 2019; 366: l4898. doi:10.1136/bmj.l4898
14 Risk of bias tools - ROBINS-E tool, Available: https://www.riskofbias.info/welcome/robins-e-tool [Accessed 17 Feb 2023 ].
15 McGuinness LA, Higgins JPT. Risk-of-bias VISualization (robvis): An R package and Shiny web app for visualizing risk-of-bias assessments. Res Synth Methods 2021; 12: 55–61. doi:10.1002/jrsm.1411
16 Wan X, Wang W, Liu J, et al. Estimating the sample mean and standard deviation from the sample size, median, range and/or interquartile range. BMC Med Res Methodol 2014; 14: 135. doi:10.1186/1471-2288-14-135
17 The Cochrane Collaboration. Review manager (revman). 2020.
18 Schandelmaier S, Briel M, Varadhan R, et al. Development of the Instrument to assess the Credibility of Effect Modification Analyses (ICEMAN) in randomized controlled trials and meta-analyses. CMAJ 2020; 192: E901–6. doi:10.1503/cmaj.200077
19 Egger M, Davey Smith G, Schneider M, et al. Bias in meta-analysis detected by a simple, graphical test. BMJ 1997; 315: 629–34. doi:10.1136/bmj.315.7109.629
20 GRADE handbook. Available: https://gdt.gradepro.org/app/handbook/handbook.html [Accessed Feb 2023 ].
21 Brand KMG, Saarelainen L, Sonajalg J, et al. Metformin in pregnancy and risk of adverse long-term outcomes: a register-based cohort study. BMJ Open Diabetes Res Care 2022; 10: e002363. doi:10.1136/bmjdrc-2021-002363
22 Carlsen SM, Martinussen MP, Vanky E. Metformin’s effect on first-year weight gain: a follow-up study. Pediatrics 2012; 130: e1222–6. doi:10.1542/peds.2012-0346
23 Deussen AR, Louise J, Dodd JM. Childhood follow-up of the GRoW randomized trial: Metformin in addition to dietary and lifestyle advice for pregnant women with overweight or obesity. Pediatr Obes 2023; 18: e12974. doi:10.1111/ijpo.12974
24 Feig DS, Sanchez JJ, Murphy KE, et al. Outcomes in children of women with type 2 diabetes exposed to metformin versus placebo during pregnancy (MiTy Kids): a 24-month follow-up of the MiTy randomised controlled trial. Lancet Diabetes Endocrinol 2023; 11: 191–202. doi:10.1016/S2213-8587(23)00004-9
25 Fornes R, Simin J, Nguyen MH, et al. Pregnancy, perinatal and childhood outcomes in women with and without polycystic ovary syndrome and metformin during pregnancy: a nationwide population-based study. Reprod Biol Endocrinol 2022; 20: 30. doi:10.1186/s12958-022-00905-6
26 Hanem LGE, Salvesen Ø, Juliusson PB, et al. Intrauterine metformin exposure and offspring cardiometabolic risk factors (PedMet study): a 5-10 year follow-up of the PregMet randomised controlled trial. Lancet Child Adolesc Health 2019; 3: 166–74. doi:10.1016/S2352-4642(18)30385-7
27 Hanem LGE, Stridsklev S, Júlíusson PB, et al. Metformin Use in PCOS Pregnancies Increases the Risk of Offspring Overweight at 4 Years of Age: Follow-Up of Two RCTs. J Clin Endocrinol Metab 2018; 103: 1612–21. doi:10.1210/jc.2017-02419
28 Landi SN, Radke S, Engel SM, et al. Association of Long-term Child Growth and Developmental Outcomes With Metformin vs Insulin Treatment for Gestational Diabetes. JAMA Pediatr 2019; 173: 160–8. doi:10.1001/jamapediatrics.2018.4214
29 Martine-Edith G, Johnson W, Petherick ES. Associations between maternal gestational diabetes metformin or insulin treatment and offspring growth trajectories from birth to 60 months of age: Findings from the Born in Bradford (BiB) study. Diabet Med 2023; 40: e15204. doi:10.1111/dme.15204
30 Paavilainen E, Tertti K, Nikkinen H, et al. Metformin versus insulin therapy for gestational diabetes: Effects on offspring anthropometrics and metabolism at the age of 9 years: A follow-up study of two open-label, randomized controlled trials. Diabetes Obes Metab 2022; 24: 402–10. doi:10.1111/dom.14589
31 Panagiotopoulou O, Syngelaki A, Georgiopoulos G, et al. Metformin use in obese mothers is associated with improved cardiovascular profile in the offspring. Am J Obstet Gynecol 2020; 223: 246. doi:10.1016/j.ajog.2020.01.054
32 Paul PG, Kingsbury BR, Yenuberi H, et al. The Study of Offspring and Mothers with Gestational Diabetes Treated with Metformin or Glibenclamide in a Randomized Controlled Trial after 9 Years. Curr Med Issues 2023; 21: 50–6. doi:10.4103/cmi.cmi_95_22
33 Rowan JA, Rush EC, Obolonkin V, et al. Metformin in gestational diabetes: the offspring follow-up (MiG TOFU): body composition at 2 years of age. Diabetes Care 2011; 34: 2279–84. doi:10.2337/dc11-0660
34 Tertti K, Toppari J, Virtanen HE, et al. Metformin Treatment Does Not Affect Testicular Size in Offspring Born to Mothers with Gestational Diabetes. Rev Diabet Stud 2016; 13: 59–65. doi:10.1900/RDS.2016.13.59
35 Yang L, Lacey L, Whyte S, et al. Metformin in obese pregnancy has no adverse effects on cardiovascular risk in early childhood. J Dev Orig Health Dis 2022; 13: 390–4. doi:10.1017/S2040174421000301
36 Ijäs H, Vääräsmäki M, Morin-Papunen L, et al. Metformin should be considered in the treatment of gestational diabetes: a prospective randomised study. BJOG 2011; 118: 880–5. doi:10.1111/j.1471-0528.2010.02763.x
37 Tertti K, Ekblad U, Koskinen P, et al. Metformin vs. insulin in gestational diabetes. A randomized study characterizing metformin patients needing additional insulin. Diabetes Obes Metab 2013; 15: 246–51. doi:10.1111/dom.12017
38 Vanky E, Stridsklev S, Heimstad R, et al. Metformin versus placebo from first trimester to delivery in polycystic ovary syndrome: a randomized, controlled multicenter study. J Clin Endocrinol Metab 2010; 95: E448–55. doi:10.1210/jc.2010-0853
39 Chiswick C, Reynolds RM, Denison F, et al. Effect of metformin on maternal and fetal outcomes in obese pregnant women (EMPOWaR): a randomised, double-blind, placebo-controlled trial. Lancet Diabetes Endocrinol 2015; 3: 778–86. doi:10.1016/S2213-8587(15)00219-3
40 Dodd JM, Louise J, Deussen AR, et al. Effect of metformin in addition to dietary and lifestyle advice for pregnant women who are overweight or obese: the GRoW randomised, double-blind, placebo-controlled trial. Lancet Diabetes Endocrinol 2019; 7: 15–24. doi:10.1016/S2213-8587(18)30310-3
41 Syngelaki A, Nicolaides KH, Balani J, et al. Metformin versus Placebo in Obese Pregnant Women without Diabetes Mellitus. N Engl J Med 2016; 374: 434–43. doi:10.1056/NEJMoa1509819
42 van Weelden W, Wekker V, de Wit L, et al. Long-Term Effects of Oral Antidiabetic Drugs During Pregnancy on Offspring: A Systematic Review and Meta-analysis of Follow-up Studies of RCTs. Diabetes Ther 2018; 9: 1811–29. doi:10.1007/s13300-018-0479-0
43 Singhal A. Long-Term Adverse Effects of Early Growth Acceleration or Catch-Up Growth. Ann Nutr Metab 2017; 70: 236–40. doi:10.1159/000464302
44 Schoonejans JM, Blackmore HL, Ashmore TJ, et al. Maternal Metformin Intervention during Obese Glucose-Intolerant Pregnancy Affects Adiposity in Young Adult Mouse Offspring in a Sex-Specific Manner. Int J Mol Sci 2021; 22: 8104. doi:10.3390/ijms22158104
45 Salomäki H, Heinäniemi M, Vähätalo LH, et al. Prenatal metformin exposure in a maternal high fat diet mouse model alters the transcriptome and modifies the metabolic responses of the offspring. PLoS One 2014; 9: e115778. doi:10.1371/journal.pone.0115778
46 Kang MJ. The adiposity rebound in the 21st century children: meaning for what? Korean J Pediatr 2018; 61: 375–80. doi:10.3345/kjp.2018.07227
© 2025 Author(s) (or their employer(s)) 2025. Re-use permitted under CC BY-NC. No commercial re-use. See rights and permissions. Published by BMJ Group. http://creativecommons.org/licenses/by-nc/4.0/ This is an open access article distributed in accordance with the Creative Commons Attribution Non Commercial (CC BY-NC 4.0) license, which permits others to distribute, remix, adapt, build upon this work non-commercially, and license their derivative works on different terms, provided the original work is properly cited, appropriate credit is given, any changes made indicated, and the use is non-commercial. See: http://creativecommons.org/licenses/by-nc/4.0/ . Notwithstanding the ProQuest Terms and Conditions, you may use this content in accordance with the terms of the License.
Details
; Feig, Denice S 6
1 Department of Medicine, University of Toronto Faculty of Medicine, Toronto, Ontario, Canada
2 Pharmacy, Sinai Health System, Toronto, Ontario, Canada
3 Department of Medicine, University of Toronto Faculty of Medicine, Toronto, Ontario, Canada; University Health Network, Toronto, Ontario, Canada
4 Obstetrics & Gynaecology, Sinai Health System, Toronto, Ontario, Canada; Obstetrics & Gynaecology, University of Toronto, Toronto, Ontario, Canada
5 Department of Pediatrics, University of Toronto, Toronto, Ontario, Canada; Division of Endocrinology, SickKids Hospital, Toronto, Ontario, Canada
6 Department of Medicine, University of Toronto Faculty of Medicine, Toronto, Ontario, Canada; Division of Endocrinology, Sinai Health System, Toronto, Ontario, Canada