Maternal BMI was recorded at 12 wkGA, and the number of women included in the analysis of maternal characteristics was determined by the availability of metabolite data at 20 and 28 wkGA. The total number of women in the analysis at 28 wkGA was 466, including 171 cases of GDM and 295 controls, and at 20wkGA, it was 469, including 171 cases and 298 controls.

Abbreviations: BMI, body mass index; Comp, comparator group; DOR, diagnostic odds ratio; FN, false negative; FP, false positive; LR, likelihood ratio; NPV, negative predictive value; P(GDM), predicted probability of gestational diabetes mellitus from the model, selected to match the specificity of the screening test BMI > 30 kg/m² or to achieve the screen positive rate of ~5%; PPV, positive predictive value; TN, true negative; TP, true positive; wkGA, weeks of gestational age.

^aDue to the case cohort design, the proportion of screen positives was calculated in the comparator group (ie, the random subcohort) among women who had the metabolite measurements available (at 28 wkGA, n = 306 including 11 cases of GDM and 295 noncases, and at 20 wkGA, n = 308 including 10 cases of GDM and 298 noncases), and PPV and NPV were weighted by the inverse of the random subcohort sampling fraction. The proportions of screen positives, sensitivity, specificity, PPV, and NPV are given as percentages.