Metformin Versus Insulin for Gestational Diabetes: Cognitive and Neuropsychological Profiles of Children Aged 9 years.

Only 159 of the 311 children in the original trials (51%) completed the neuropsychological assessment, so a difference between the groups could have been missed. The authors' power calculation, run after the study, found it could not detect small differences in full-scale IQ; only a difference as large as 10 IQ points could be ruled out. Children were assessed in the school year they turned 9, so some were in grade 2 and some in grade 3, which the authors say may have affected the level of school support recorded. Five different psychologists or final-stage psychology students carried out the assessments, although no difference was found between their median scores. Story memory scores were below average in both groups; the authors suggest this may reflect a ceiling effect and tiredness from a long day of testing. In the metformin group, 22 of 82 mothers also received insulin, so that group was not metformin alone. More metformin-group children scored below average on IQ and more received school support, and with a sample this size those differences cannot be dismissed as definitely not real, even though neither was statistically significant.

The authors declare no conflicts of interest. No funding source is stated in the text provided.

Do not read the higher rate of below-average IQ in the metformin group (28.6% versus 16.5% in the insulin group) as evidence that metformin harms children's thinking. That difference did not reach significance (p = 0.070), and the same goes for the higher rate of school support in the metformin group (14.3% versus 8.5%, p = 0.063). Nor is the null result proof that the two drugs are equivalent: the study could rule out only a difference as large as about 10 IQ points.