Background: It has been known for long time that pregnancy associated with weight problems women but few important factors have been identified, except for weight gain during pregnancy itself.
Objective: To identify cutoffs for weight gain during pregnancy for identification of women at risk of high weight retention after pregnancy.
Methods: A longitudinal analysis of 563 women who gave birth in 1984-1985 was used. Weight development during pregnancy was used as diagnostic test for high weight retention at 1 and 15 years follow-up. True positives for high weight retention were defined as > 90th percentile at 1 and 15 years. Receiver Operating Characteristic (ROC) analyses were performed to derive cutoffs minimizing the absolute and relative number of misclassifications.
Results: The average weight retention at 1 year (WR1y) and 15 years (WR15y) was 1.1 +/- 3.6 and 7.6 +/- 7.4, respectively. There was a moderately strong correlation between weight gain during pregnancy and weight retention at 1 and 15 years follow-up (r2 = 0.13 and r2 = 0.05, respectively; both P < 0.001) and weight gain during pregnancy as diagnostic test to find high weight retainers performed better than chance both for WR1y (AUC = 0.76 +/- 0.04, P < 0.01) and WR15y (0.63 +/- 0.04, P < 0.05). To minimize the absolute number of misclassifications of high weight retention, a highly specific and insensitive cut-off of around 24 kg weight gain during pregnancy was needed. The best trade-off between sensitivity and specifity, minimizing the relative number of misclassifications, was at a cutoff of around 16 kg, but this resulted in three times as many absolute misclassifications.
Conclusion: Weight gain during pregnancy was found to be a moderately strong diagnostic test for weight retention after pregnancy. In order to implement weight gain recommendations for clinical practice larger studies need to be conducted and the demands on the reference values specified, since decisions about the trade-offs between types and extent of misclassifications need to be made.