Metformin

Pregnancy

MotherToBaby says metformin is not expected to increase the chance of birth defects or miscarriage, while poorly controlled diabetes can raise these risks. The Glumetza label says the available human pregnancy studies have not reported a clear association with major defects or miscarriage, but their small samples and inconsistent comparison groups cannot rule out a drug-related risk. The label also reports no adverse developmental effects in rats and rabbits at the tested exposures; those animal findings do not replace human evidence. In a prospective cohort, Scherneck et al. (2018) compared 336 first-trimester exposed pregnancies with 1,011 matched controls. Adjusted estimates did not show a statistically significant increase in major birth defects (OR 0.58, 95% CI 0.3 to 1.3) or spontaneous abortion (HR 0.95, 95% CI 0.6 to 1.5). Panchaud et al. (2018) studied 392 exposed pregnancies and 431 unexposed references. Overall major defects were reported in 20/392 (5.1%) versus 9/431 (2.1%) (adjusted OR 1.70, 95% CI 0.70 to 4.38); pregnancy losses were 20.8% versus 10.8% (adjusted HR 1.57, 95% CI 0.90 to 2.74). Neither overall comparison was statistically significant. In the pre-existing diabetes subgroup, the authors saw higher rates and concluded that this appeared due to the underlying diabetes, while the non-diabetes indication group did not show meaningfully increased risks. Metformin crosses the placenta. Eyal et al. (2010) studied 35 pregnant women, collected cord blood from 12, and found delivery cord-plasma levels from below detection to 1,263 ng/mL. In the MiG TOFU follow-up, Rowan et al. (2018) assessed 208 children at 7 or 9 years. Body-fat percentage and measured metabolic markers were similar between the metformin and insulin groups, although children in the 9-year metformin subgroup were larger on some measures. Ryssdal et al. (2025) conducted post-hoc analyses of two PCOS randomized trials and a follow-up study, including 634 women and 145 offspring. At age 8, the analysis reported higher odds of allergy (OR 4.83, 95% CI 1.47 to 21.8) and eczema (OR 2.42, 95% CI 1.14 to 5.33) in offspring exposed in utero. This is an association from a small follow-up group, not proof of cause. Mayadunne et al. (2025) reviewed long-term offspring outcomes and described the overall findings as limited and inconsistent. A 2026 systematic review of 10 GDM randomized trials reported lower odds of neonatal hypoglycemia (OR 0.65, 95% CI 0.46 to 0.92) and a mean birth weight about 69 g lower with metformin; long-term child outcomes remained uncertain (Brinkmann et al., 2026). A separate 2026 meta-analysis of randomized and cohort evidence reported lower overall congenital-malformation risk with metformin versus insulin (RR 0.83, 95% CI 0.71 to 0.99), with additional genetic analyses described as hypothesis-generating (Ji et al., 2026). A 2026 patient-level meta-analysis of 10 placebo-controlled trials in pregnancies without diabetes found no reduction in gestational diabetes; its prevention question does not establish treatment effects in people with diabetes (Mousa et al., 2026). A 2026 randomized trial studied short-term metformin after antenatal betamethasone in 169 women without diabetes who had risk of preterm delivery. It reported neonatal hypoglycemia in 10/48 versus 23/58 preterm infants, but the setting and short treatment window differ from ongoing treatment for diabetes (Yefet et al., 2026).

Breastfeeding

LactMed, revised August 15, 2026, says milk levels are low and reported infant doses are less than 0.5% of the mother's weight-adjusted dose. MotherToBaby also describes transfer into milk as very small. In Eyal et al. (2010), six women provided milk samples; estimated daily infant intake was 0.13 to 0.28 mg and the relative infant dose was below 0.5%. Cord-blood levels in the same study ranged from undetectable to 1,263 ng/mL, which is pregnancy exposure rather than milk exposure. Briggs et al. (2005) started seven women on metformin after cesarean delivery; two were excluded, the estimated infant dose in milk was 0.43% to 1.08% of the mother's weight-adjusted dose, and three nursing infants had glucose readings measured, all within the reported normal range of 47 to 77 mg/dL. Glueck et al. (2006) followed 61 nursing and 50 formula-fed infants born to 92 mothers with PCOS who used metformin throughout pregnancy and lactation. At 3 and 6 months, the study found no differences in weight, height, motor-social development, or illness. Because mothers used metformin during pregnancy as well as while nursing, this study cannot isolate effects from milk exposure alone. LactMed and MotherToBaby advise caution for newborns, premature infants, and infants with kidney impairment. LactMed says timing breastfeeding around doses offers little benefit. The Glumetza label says available studies are small and have limited infant-outcome data, and that there is not enough information to determine effects on the breastfed infant or milk production. Husby et al. (2026) found no difference in exclusive breastfeeding at one month between women with PCOS randomized to metformin or placebo; this measured breastfeeding behavior, not infant effects from drug in milk.

What was tested

Pregnancy evidence includes observational cohorts of first-trimester exposure compared with unexposed pregnancies, randomized trials comparing metformin with insulin or placebo for gestational diabetes or PCOS, and follow-up of children after in-utero exposure. Studies measured birth defects, pregnancy loss, preterm birth, birth size, newborn blood sugar, and later body composition or metabolic markers. Eyal et al. measured metformin in maternal blood, cord blood, and breast milk. Lactation studies measured milk and infant levels, infant glucose, growth, development, or illness. A 2026 trial also tested a short course after antenatal steroids in a distinct preterm-birth setting, while another 2026 patient-level meta-analysis studied gestational-diabetes prevention in pregnancies without diabetes.