WHAT IS ALREADY KNOWN ON THIS TOPIC

In many studies, including our previous report on the main CLUE study, metformin exposure, compared with insulin exposure, during pregnancy was associated with increased relative risk of being small for gestational age (SGA) and decreased risk of being large for gestational age (LGA).

WHAT THIS STUDY ADDS

Exposure to metformin at any time during pregnancy, regardless of the indication (ie, gestational diabetes mellitus (GDM), pregestational type 2 diabetes mellitus or polycystic ovary syndrome), compared with non-pharmacological antidiabetic treatment (naïve cohort) was not associated with difference in risk of SGA or LGA in the offspring.

Among mothers with GDM, only a decreased risk of LGA was found when metformin cohort was compared with naïve cohort; however, given the observational nature of the data, differences in baseline risk between exposure groups and unmeasured or residual confounding cannot be ruled out as an explanation.

HOW THIS STUDY MIGHT AFFECT RESEARCH, PRACTICE OR POLICY

This follow-up study suggests that exposure to metformin does not translate into difference in the risk of SGA or LGA.

These results add further evidence to support safe use of metformin during pregnancy.

Nonetheless, when antihyperglycemic treatment is needed in a GDM pregnancy complicated by intrauterine growth restriction, metformin should be avoided.

Introduction

Metformin use in pregnancy is increasing worldwide, both for continued use in pregnant women with underlying type 2 diabetes mellitus (T2DM) and for gestational diabetes mellitus (GDM).1 2 Metformin is also commonly prescribed off-label in women with polycystic ovary syndrome (PCOS) to induce ovulation and to improve pregnancy outcomes.3 Metformin crosses the placenta, exposing the fetus to plasma concentrations comparable to maternal circulations.4 However, several studies reported the effectiveness of metformin for glycemia control without short-term adverse effects and with additional potential benefits in the neonatal period.5–7 In early 2022, European Health authorities approved the use of metformin during pregnancy and in the periconceptional period for the originator metformin product (Glucophage, Merck KGaA, Darmstadt, Germany).8 This change was, at least in part, based on the large register-based cohort (main CLUE study), in which maternal pregnancy exposure to metformin was studied and both short-term and long-term adverse outcomes in children (up to 11 years after birth) were assessed.9 The study found no increased long-term risk associated with pregnancy exposure to metformin. However, an increased risk of being small for gestational age (SGA) and a decreased risk of children being large for gestational age (LGA) after in utero exposure to metformin in comparison to in utero exposure to insulin was found. Similar results were observed in children born to mothers with GDM and naïve to in utero exposure to non-pharmacological antidiabetic treatment compared with in utero exposure to insulin.

We hypothesized from our previous study results that the observed increased risk of SGA and decreased risk of LGA may be related to the choice of insulin as the reference for comparison, given the known association between insulin and weight increase.10–12 Based on these findings and to further strengthen the evidence, this follow-up study was performed to investigate if metformin is associated with increased risk of SGA and decreased risk of LGA when compared with non-pharmacological antidiabetic treatment exposure.

Methods

This was a follow-up study performed on our previously published population-based register data cohort study.9 The study population included singleton children born to women, 18–45 years of age at the time of delivery, in Finland during 2004–2016. The children were identified using the Finnish Medical Birth Register,13 which holds information on all births in Finland, including date of birth, gestational age and birth weight. The start date of pregnancy was calculated by subtracting gestational age (recorded in weeks and days) from the date of delivery to obtain the date of the last menstrual period (LMP). The main analyses included two cohorts: children with in utero exposure to metformin (metformin cohort) regardless of the indication (ie, GDM, pregestational T2DM or PCOS) and children of mothers with GDM with non-pharmacological antidiabetic treatment (naïve cohort) (described in online supplemental item 1).

Exclusion criteria were maternal diagnosis of type 1 diabetes mellius, maternal dispensation of systemic glucocorticoids during pregnancy (agents in this drug class are known to interfere with metformin) and maternal dispensation of antidiabetic medications other than metformin (anatomical therapeutic code (ATC) A10BA02) and insulin (ATC A10A) during pregnancy. To ensure adequate capture of information on exposure and baseline characteristics, children born to women not registered in Finland throughout the entire duration of pregnancy were excluded. Information on the selection of the study population and definitions of exclusion criteria are provided in online supplemental item 2 and 3.

Furthermore, a metformin-GDM subcohort was created for additional analyses. This subcohort was identified by excluding women who were assumed to use metformin in the treatment of T2DM or PCOS (details provided in online supplemental item 1).

Maternal exposure to metformin and to other pharmacological antidiabetic treatment was defined as having any dispensed prescription between the LMP and the date of delivery obtained from the Finnish Prescription Register.14 Information to identify children with LGA (birth weight 2 SD above the gestational age-specific and sex-specific reference mean in Finland) and SGA (birth weight 2 SD below the gestational age-specific and sex-specific reference mean) was ascertained from the Medical Birth Register.13

The data sources used to identify covariates were the Medical Birth Register,13 the Prescription Register,14 the Care Register for Healthcare (HILMO),15 the Register of Primary Healthcare Visits (AvoHILMO),16 regional laboratory databases and Statistics Finland.17 A detailed description of the methods and covariate definitions is available in our previous publication.9

The analyses of SGA and LGA were conducted using logistic regression to estimate ORs with 95% CIs, with naïve cohort as the reference. Inverse probability of treatment weighting (IPTW) with stabilized weights based on propensity scores (PS) was used to control for confounding. Separate PS models were estimated for the main analyses with metformin cohort and naïve cohort; and for the additional analyses with metformin-GDM subcohort and naïve cohort. Analyses were conducted after trimming of children outside the overlapping range of the PS. The predictors of the PS models were a broad range of covariates describing the mothers and pregnancies, including demographic factors, comorbidities before and during pregnancy, lifestyle factors, gestational week of GDM diagnosis, region of residence for the child and calendar year of delivery.9 The standardized mean difference was used to assess covariate balance between cohorts after weighting. Covariates that were not balanced after weighting (standardized mean difference ≥0.1) were included as independent variables in the outcome models.

Results

A total of 3964 children in the metformin cohort and 82 675 children in the naïve cohort fulfilled the criteria for inclusion in the main analyses (table 1 and online supplemental item 4). The metformin-GDM subcohort consisted of 2361 children. In the main analyses, 1070 (27.0%) children in metformin cohort and 233 (0.3%) in naïve cohort were excluded due to non-overlapping PS. The corresponding numbers for the metformin-GDM subcohort and naïve cohort in the additional analyses were 11 (0.5%) and 5080 (6.1%), respectively.

Table 1

Baseline characteristics of children with maternal exposure to metformin (metformin cohort) and children of mothers with GDM exposed to non-pharmacological antidiabetic treatment (naïve cohort)

The median for gestational age at birth was 39.4 weeks in the metformin cohort and 39.9 weeks in the naïve cohort (table 1). In the metformin cohort, 43.6% of children were born during the latest time period (2014–2016) and 15.3% during the earliest time period (2004–2008), whereas in the naïve cohort 30.1% of births were born in 2014–2016 and 28.7% in 2004–2008. Maternal median age at delivery was 32 years in the metformin cohort and 31 years in the naïve cohort, while maternal prepregnancy body mass index (BMI) median was 29.7 kg/m2 in the metformin cohort vs 26.9 kg/m2 in the naïve cohort.

Most covariates included in the PS models were balanced after applying IPTW (standardized mean difference <0.1), however gestational week of maternal GDM diagnosis and PCOS remained unbalanced (table 1). GDM was diagnosed earlier in the metformin cohort and later in the naïve cohort (median of 24.4 gestational weeks and 28.3 gestational weeks, respectively) and PCOS frequency was higher in the metformin cohort (11.7%) than in the naïve cohort (1.4%). Both variables were, accordingly, included as adjusting variables in subsequent outcome models.

SGA was found in 2.3% of the children in the metformin cohort and 2.2% in the metformin-GDM subcohort (table 2). The percentage in the naïve cohort was 2.0%. No difference was found in the risk for SGA when metformin cohort was compared with naïve cohort in the main analyses (weighted OR (wOR) 0.97, 95% CI 0.73 to 1.27) or when the metformin-GDM subcohort was compared with naïve cohort in the additional analyses (wOR 1.01, 95% CI 0.75 to 1.37).

Table 2

Risk of being LGA or SGA at birth in the main analyses with metformin cohort and naïve cohort and in the additional analyses with metformin-GDM subcohort and naïve cohort

LGA was found in 4.0% of the children in the metformin cohort and 4.7% in the metformin-GDM subcohort (table 2). The percentage in the naïve cohort was 4.1%. No difference was found in the risk for LGA when metformin cohort was compared with the naïve cohort in the main analyses (wOR 0.91, 95% CI 0.75 to 1.11) but the risk was decreased in the metformin-GDM subcohort, in comparison to the naïve cohort, in the additional analyses (wOR 0.72, 95% CI 0.56 to 0.92).

Discussion

No difference in LGA risk was found in the metformin cohort compared with the naïve cohort in the main analyses of this follow-up study to CLUE, but a decreased risk of LGA was found in the additional analyses with the metformin-GDM subcohort. A decreased risk of LGA birth for metformin versus insulin during pregnancy in this study cohort has already been described (online supplemental item 5).9 The recent large prospective clinical trial MiTy found a reduced risk of LGA in offspring exposed to metformin+insulin compared with placebo+insulin in pregnant women with T2DM.18 Furthermore, a large meta-analysis found no change in LGA incidence between metformin-exposed and insulin-exposed neonates,12 with the caveat that the metformin groups in the underlying studies did not discriminate the use of additional insulin in the metformin treatment groups. In prospective clinical trials in obese non-diabetic women, metformin use versus placebo did not alter the risk of LGA birth,19 20 whereas in a recent cohort study on pregnant women without diabetes and PCOS exposed or non-exposed to metformin, a reduced LGA risk for metformin exposure was found.21 Prepregnancy BMI and gestational hyperglycemia are considered drivers of LGA risk in the offspring.22 In summary, current evidence suggests that metformin may reduce hyperglycemia-related LGA.

Our previous findings in the main CLUE study indicated a significantly increased risk of SGA associated with in utero exposure to metformin when compared with insulin (online supplemental item 5),9 but not for metformin+insulin versus insulin. The MiTy study found an overall shift towards lower birth weight and increased risk of SGA in offspring exposed to metformin+insulin compared with insulin+placebo in pregnant women with T2DM.18 Likewise, in a recent randomized double-blind study, metformin-exposed babies born to GDM mothers were born approximately 100 g lighter than those exposed to placebo.23

A large meta-analysis12 24 found a significant decrease in birth weight after intrauterine exposure to metformin compared with insulin, but no difference in SGA risk.24 Other observational studies have also not found differences in SGA risk when comparing metformin and diet treatment among pregnant women with GDM.25 26 And indeed, no differences were found in this follow-up analysis to CLUE regarding SGA risk in the metformin cohort compared with the naïve cohort, neither in the main analysis nor with the narrowed metformin-GDM subcohort. Metformin crosses the placenta,4 and based on the findings above, it is discussed whether metformin affects feto-placental metabolism27 leading to a reduced birth weight and accelerated catch-up growth later in life.24

However, large prospective studies in non-diabetic patient do not support a general impact of metformin on fetal weight: metformin did not alter median birth-weight z score (percentile) placebo19 28 or SGA risk20 in non-diabetic obese pregnant women at doses of 1500–3000 mg/day. Birth weight and SGA risk were also not altered in babies born to women with pregestational insulin resistance29 or PCOS.30

Thus, it might be discussed that a potential (residual) lack of glucose control with insulin (or other comparator treatments) in GDM pregnancies leads to a relatively higher birth weight,31 32 for example, as suggested by better glycemic control with metformin versus placebo in MiTy.18 In the main CLUE study,9 the risk of SGA birth was increased for drug-naïve GDM women compared with those on insulin with wOR 0.72, 95% CI 0.56 to 0.92 (see online supplemental item 5), which could be indicative of insulin use shifting towards higher birth weight. As large prospective studies like EMERGE (NCT02980276) will be needed to clearly determine the underlying mechanisms, practitioners are well advised to monitor fetal growth and discontinue metformin in case of emerging growth restriction as well as countering overly fetal growth in utero in insulin-treated mothers.

Similar to the main CLUE study,9 this follow-up study has a number of strengths, being a comprehensive study involving national coverage of mothers’ exposure to metformin across the various levels of healthcare, allowing the results to be generalizable to the Finnish population. GDM in this study was identified from multiple sources (diagnosis code or laboratory records), which should provide extensive and precise information. Also, the use of IPTW methods based on PS, including a broad range of maternal characteristics, improved comparability between the cohorts and reduced the potential for confounding.

There were several limitations in this follow-up study that deserve comment. Drug exposure was potentially misclassified due to the non-use of the dispensed drug (eg, due to gastrointestinal side effects of metformin)33 34 and, subsequently, dose stratification was not possible. Confounding by severity could affect the results, since information on PCOS, T2DM and GDM severity was not available in the data. Although IPTW methods were used to account for broad range of characteristics (eg, demographic characteristics, maternal comorbidities, pregnancy-related variables), the prevalence of PCOS, T2DM and GDM differed across treatment cohorts, which potentially caused a residual confounding. While the metformin cohort had the highest proportion of PCOS and the lowest proportion of GDM, the naïve cohort included, by definition, women with GDM diagnosis only. Furthermore, 27% of the metformin cohort was trimmed when IPTW analysis was performed due to PS non-overlapping in the main analyses, likely due to the absence of GDM diagnosis during pregnancy (table 1). Only 0.5% of the metformin cohort was trimmed when the PS procedure was applied in the additional analyses with metformin-GDM subcohort. Therefore, the additional analyses with metformin-GDM subcohort and naïve cohort improved the generalizability and robustness of the results, since a smaller proportion of the children were trimmed.

Conclusion

In conclusion, this CLUE follow-up study supports previous findings that in utero exposure to metformin in comparison to drug-naïve GDM is not associated with an increased risk of SGA or LGA. The observed decrease in the risk of LGA in the analyses focusing on mothers with GDM may imply residual confounding. Regarding SGA risk, the results of this analysis, and even more the neutral outcomes of prospective studies in non-GDM indications add to an unclear relationship between metformin exposure in utero and potential risk of lower birth weight, as seen versus insulin. For the time being, physicians are well advised to consider other treatment options than metformin when (risk of) intrauterine growth restriction is observed. Large prospective studies assessing metformin alone versus placebo in GDM pregnancies, like EMERGE (NCT02980276) will help to clear the sight.

Data availability statement

Data may be obtained from a third party and are not publicly available. All data relevant to the study are included in the article or uploaded as supplementary information.

Ethics statementsPatient consent for publication

Not applicable.

Ethics approval

The Ethics Committee of the Hospital District of Helsinki and Uusimaa, Finland, granted the CLUE study a favorable approval (reference number HUS/1742/2017). This study was a retrospective analysis, therefore, informed consent by participants was not required.

Acknowledgments

The study team would like to thank Ulrike Gottwald-Hostalek (Merck KGaA, Darmstadt, Germany) for enabling the conduction of this analysis. We further acknowledge Saara Hetemäki as the project manager and data access lead of this research study; Rosa Juuti for her contribution on the conceptualization of the study, the study protocol and the study permits; Anna Glans Lundin for her contribution on the study protocol; Minna Vehkala, Juha Mehtälä and Pasi Korhonen for their contribution in the study design and planning and interpreting of the statistical analyses; Henrik Svanström for contributing to interpretation of analyses; Marja Vääräsmäki and Laure Morin-Papunen for their clinical assistance throughout the CLUE study and Catarina Camarinha and Geetika Nirmal for preparing this manuscript.