We identified participants for this population-based, retrospective cohort study from the Washington state longitudinal births database, which is comprised of linked birth certificate data for all women with at least 2 singleton births in Washington State between 1992 and 2005. The Institutional Review Board of the Washington State Department of Health approved the use of these data for the current study. For our analysis, eligible subjects included women with two consecutive live births between 1992 and 2005, diagnosis of GDM at the baseline pregnancy, and vaginal delivery of a live infant during the baseline pregnancy. GDM diagnosis and vaginal delivery were identified by checkbox format and abstracted from the birth certificates. The diagnosis of GDM is typically made when an abnormal response to an oral glucose load is identified at a prenatal visit between 24 and 28 weeks of gestation. We cannot verify how the screening for and diagnosis of GDM were made, as specific practices were not documented. We excluded women with established diabetes at their baseline pregnancy and women whose baseline pregnancy did not result in a live singleton birth. We excluded women who had undergone cesarean delivery at their baseline pregnancy, although we could not exclude women who in theory may have had a cesarean delivery prior to 1992 or a cesarean delivery out-of-state prior to our baseline ascertainment. In addition, we excluded women with a medical indication for cesarean delivery during their subsequent pregnancy (genital herpes, non-vertex or breech presentations, placenta previa and abruptio placenta).
Our exposure of interest was interpregnancy weight change. The interpregnancy weight change for each woman was calculated (prepregnancy weight at subsequent pregnancy - prepregnancy weight at baseline pregnancy) and assigned to one of the following three categories: weight loss (greater than 10 lbs), weight stable (±10 lbs), or weight gain (greater than 10 lbs). Prepregnancy weight is typically the weight measured at the first prenatal visit. However, it is possible that for some women in our cohort prepregnancy weights were self-reported. The outcome of cesarean delivery for the subsequent birth was identified using the checkbox format on the birth certificate.
Using multiple logistic regression, we calculated the odds ratio (OR) and 95% confidence interval (CI) for cesarean delivery separately comparing the weight-loss and weight-gain groups to the weight-stable group (reference group). We identified important variables to adjust for in our analyses a priori, based on current evidence. Specifically, we included the following confounding variables in our regression analyses (for the subsequent birth unless otherwise noted): maternal age (<25, 25-34, ≥35 years), maternal race/ethnicity (White, Black, Hispanic, Asian or other), maternal education (<12, 12-15, ≥16 years), interbirth interval (<12, 12-35, ≥36 months), pre-pregnancy weight at the baseline pregnancy (<100, 100-149, 150-199, ≥200 lbs), weight gain during the baseline and subsequent pregnancy (loss, 0-14, 15-24, 25-34, ≥35 lbs), smoking during pregnancy (no/yes), and year of birth (subsequent infant).
We evaluated the dose-response relationship between interpregnancy weight gain tertiles on the risk of cesarean delivery. Finally, a sub-analysis was performed examining the effects of interpregnancy weight change using change in BMI between baseline pregnancy and subsequent pregnancy (prepregnancy BMI at subsequent pregnancy - prepregnancy BMI at baseline pregnancy) among the 83% of subjects for whom BMI data were available, as an additional means to explore our hypothesized associations between interpregnancy weight change and cesarean delivery. BMI categorization was based on a previous study by Villamor et al
evaluating the relationship of interpregnancy BMI change with risk of adverse pregnancy outcomes in a large Swedish cohort (18
). All statistical analyses were performed using Stata 10.0 (StataCorp LP, College Station, TX).