However, there was significantly less weight gain in those taking metformin in this population (8.7 vs. 11.48 kg; P less than .0001). And newborns had significantly larger head circumferences if their mothers with PCOS took metformin (35.4 vs. 35.0 cm; P = .006).
Dr. Løvvik and her colleagues subsequently performed the same analysis on the first per-protocol group, one that excluded the 46 patients who were early dropouts after randomization. Here, the odds ratio favoring metformin for the composite primary outcome measure was a significant 2.55 (n = 432; 95% CI, 1.10-6.42; P = .03). Results for the final per-protocol analysis that excluded dropouts and those with less than 90% adherence to study medication were similar, with an odds ratio of 2.76 in favor of metformin use during pregnancy (n = 298; 95% CI, 1.0-8.82; P = .05).
In response to an audience question, Dr. Løvvik said that, though a per-protocol analysis of some sort had been predetermined, the decision to perform the narrower analysis on the highly adherent group was made later. However, compared to thresholds for adherence in other studies, “ours was way higher – so it’s too strict,” she acknowledged.