Statistical analysis
The number of subjects selected for the study was based on power calculations of GDF15 differences between NW or OB in previous studies in nonpregnant subjects16, 17; a minimum of 8 subjects was needed for a power of 0.95 and alpha= 0.05. Differences in ordered categorical variables between the NW and OB groups were assessed with chi-square or Fisher´s exact test; between-group differences for continuous variables were assessed with a t test, and within-group comparisons for continuous variables were assessed with a paired t test. Associations between GDF15 and changes in glucose, insulin, HOMA-IR, or HOMA-B were analysed with Pearson correlation within groups or linear regression models, adjusted for gestational age and maternal BMI and age at T1, for the whole population. The diagnostic plots for residuals were reviewed and met assumptions of normality and homoscedasticity. Categorical variables were expressed as number or percentage, and continuous variables as mean ± SD. All tests were two-tailed;p < 0.05 was considered statistically significant.