The study setting is Kaiser Permanente of Northern California, which provides comprehensive medical services to more than 3 million members located in a 14-county region in Northern California. The methods used to identify this cohort have been described in detail elsewhere.13,14
Briefly, we identified all pregnancies resulting in a singleton live birth between January 1, 1996, and June 30, 1998, to women without recognized diabetes before pregnancy by using the Kaiser Permanente of Northern California pregnancy glucose tolerance registry,15
which has been reported to be 99.4% accurate.16
To be eligible, women could not have had GDM in a previous pregnancy and had to be screened for GDM between 24 and 28 weeks of gestation (according to the earliest ultrasonogram obtained) in the index pregnancy. Trained medical record abstractors completed chart review on the randomly selected women in the case and control groups to confirm the presence of the inclusion criteria. Chart review was conducted on 437 GDM cases identified in the electronic databases, with 388 eligible for study inclusion. We also reviewed the medical records for 1,000 potential women for the control group, of whom 972 met the eligibility criteria. Women were classified as having GDM if two or more plasma glucose values obtained during the 100-g, 3-hour oral glucose tolerance test were abnormal according to the National Diabetes Data Group criteria.17
The trained medical record abstracters recorded the pregnancy weights noted on the prenatal form, which included self-reported prepregnancy weight, any measured weights during the 1 year before pregnancy, weight measured at the first prenatal visit, and weight measured at or before the patient had her 50-g, 1-hour glucose challenge screening test for GDM. Additional information obtained from the medical records included height, last menstrual period (LMP), parity, smoking during pregnancy, blood pressure at the first prenatal visit, and gestational age estimated by the earliest ultrasonogram obtained before 24 weeks of gestation. Blood pressure at the first prenatal visit was categorized according to the American Heart Association’s criteria.18
Preeclampsia was considered present if a woman had a physician’s diagnosis in the medical chart. Women’s self-reported race/ethnicity and education were obtained from the neonate’s birth certificate.
The rate of gestational weight gain, per week, before the glucose screening test was calculated as the weight measured at or before the glucose screening test minus prepregnancy weight divided by the weeks of gestation attained at the time of the weight measurement. On average, the glucose screening test weight was measured less than 1 week before the test (mean [±standard deviation (SD)]0.8 [±1.1] weeks, range 0–4 weeks). We calculated the rate of gestational weight gain in the first trimester as the weight at first prenatal visit (12 [±1.5] weeks, range 6–13 weeks) minus self-reported pregravid weight divided by the weeks of gestation attained at the first prenatal visit. Rate of weight gain in the second trimester was defined as the weight at glucose screening test (performed on average at 26 [±1.8] weeks) minus weight at the first prenatal visit divided by the number of weeks between measurements. The gestational week assigned to each maternal weight measurement was based on the earliest ultrasonogram obtained before 24 weeks of gestation.
The 2009 IOM recommendations provide pre-gravid BMI-specific recommendations for a range of absolute weight gain in the first trimester and a range of rate of weight gain per week for the second and third trimesters.12
Based on the observed weight gain at the time of the 1-hour glucose challenge test, we determined whether each woman met, exceeded, or was below the IOM recommendations. We subtracted 13 weeks from the gestational age at the time of the glucose screening test and multiplied this value by the BMI-specific recommended rate of weight gain for the second and third trimesters. Weight gained after the first trimester was then added to the BMI-specific absolute weight gain recommended by the IOM for the first trimester.
Self-reported prepregnancy weight was missing for 226 of the eligible women (15%). For those missing a self-reported prepregnancy weight, the measured weight closest to the woman’s LMP but no more than 12 months before her LMP, was used. To validate this method of estimating prepregnancy weight, we compared the self-reported prepregnancy weight to a weight measured within 12 months of the LMP among 507 women (44.7%) for whom both measurements were available. The intraclass correlation coefficient between the two weights was 0.967. The mean difference (self-reported minus measured weight) was 1.2 (±3.5) kg for normal weight women, 0.8 (±4.9) kg for overweight women, and −1.3 (±5.0) kg for obese women. These findings are similar to previous reports,19–21
except that unlike previous studies,19,21
the obese women in our study were less likely to under-report their pregravid weight.
Of the eligible 388 women with GDM and 972 women in the control group for whom chart review was completed, the following were excluded from all analyses because of missing data: pregravid weight (10.6% of women in the case group and 15.4% of women in the control group), height (0% of women in the case group and 0.7% of women in the control group), and no weight measured at a prenatal visit within 4 weeks of screening test (1.7% of women in the case group and 2.8% of women in the control group). This left 341 women with GDM and 793 women in the control group for analyses. Women for whom gestational weight gain data were missing were similar to those included in the analysis regarding age, race/ethnicity, parity, and education. This study was approved by the human subjects committee of the Kaiser Foundation Research Institute and the Committee for the Protection of Human Subjects, California Health and Human Services Agency.
We categorized women into tertiles based on the distribution of the rate of weight gain in the women in the control group. We examined differences between women in the case group and women in the control group in the distributions of categorical variables using χ2
tests. Unconditional logistic regression was used to obtain odds ratios (ORs) as estimates of the relative risk of GDM in relation to category of rate of gestational weight gain. Women in the lowest third of the distribution of rate of gestational weight gain were used as the reference group. For models using the IOM recommendations as the exposure, we compared women who exceeded the recommendations with those who met or were below the recommendations because so few women were below the recommendations (46 women in the case group [13%] and 104 women in the control group [13%]). Variables evaluated for confounding included maternal age, race/ethnicity, pregravid BMI, blood pressure at the first prenatal visit, parity, maternal education, and first-degree family history of diabetes. To assess confounding, we entered covariates into a logistic regression model one at a time and then compared the adjusted and unadjusted odds ratios.22
Final logistic regression models included covariates that altered unadjusted odds ratios for gestational weight gain by at least 10%, as well as those covariates of a priori interest (ie, parity and blood pressure). The multivariable adjusted models included the following covariates: age, race/ethnicity, parity, blood pressure, and prepregnancy BMI. Sensitivity analyses were conducted to examine first-trimester weight gain and weight gain up to the time of glucose screening among those women with a measured weight that was within 6 months of their LMP and further adjusted for the length of time between the prepregnancy weight measurement and the index pregnancy. In addition, we also conducted a sensitivity analysis examining the rate of gain in the first trimester among women who had their first prenatal visit performed at 12–14 weeks of gestation to assess whether measuring weight at the end of the first trimester would change our results. To assess the potential modifying effects of pregravid BMI (continuous) and race/ethnicity (non-Hispanic white compared with African-American, Asian, and Hispanic women) and age (younger than 30 years compared with 30 years or older), we examined interaction terms and repeated the analyses within these subgroups.