555: The association of glycosylated hemoglobin (HbA1c) with infant birthweight among women with gestational diabetes: a classification and regression tree (CART) analysis

Abstract

with infant birthweight among women with gestational diabetes: a classification and regression tree (CART) analysis Erica K Berggren, Thomas Stewart, Kim Boggess UNC Hospitals, Obstetrics & Gynecology, Chapel Hill, NC, UNC Gillings School of Global Public Health, Biostatistics, Chapel Hill, NC, University of North Carolina at Chapel Hill, Obstetrics and Gynecology, Chapel Hill, NC OBJECTIVE: To determine if third trimester HbA1c is associated with infant birthweight in women with gestational diabetes (GDM), using traditional regression analysis as well as a classification and regression tree (CART) analysis that examines non-independent relationships of variables associated with birthweight. STUDY DESIGN: We conducted a retrospective cohort study of women diagnosed with GDM using National Diabetes Data Group criteria between April 1996 and May 2010. A subset of women with GDM had a HbA1c collected during the third trimester, in addition to routine screening and diagnostic test results. This subset comprised our sample for analysis. We used linear regression to estimate the association between birthweight, body mass index (BMI), and log transformed measures of all glucose values. We performed a CART analysis to develop a decision tree that offers data-drive thresholds and explores potential interactions and non-linear relationships between variables and infant birthweight. RESULTS: Over 14 years, 1082/33179 (3.3%) of women screened were diagnosed with GDM. A subset of 114 had complete data that included HbA1c. Mean birthweight was 3506 g (SD 678g). Maternal BMI, one-hour screening glucose load result, and fasting diagnostic OGTT were independently associated with infant birthweight (p 0.01, 0.02, 0.02, respectively). HbA1c increased, but not significantly, with increasing birthweight (p 0.05). CART analysis classified each pregnancy into one of four groups. The analysis selected HbA1c 5.7 as the most discriminating variable to predict birthweight, followed by BMI 29.8mg/kg2, and fasting GDM diagnostic test result of 126mg/dL. Values above these hierarchical thresholds denote a group with higher birthweight. (Figure). CONCLUSION: CART analysis, allowing for variable interaction, suggests HbA1c may be associated with infant birthweight, beyond associations identified by a traditional linear regression analysis. Prospective studies examining the predictive use of HbA1C for birthweight is warranted based on this analysis. 556 The NICHD-MFMU GDM study: differences in perinatal outcomes between Hispanic women and Non-Hispanic White women with and without gestational diabetes Erica Berggren For the Eunice Kennedy Shriver National Institute of Child Health and Human Development, MFMU Network, Bethesda, MD OBJECTIVE: Racial/ethnic differences in pathophysiology of gestational diabetes (GDM) may impact perinatal outcomes among these women. We evaluated differences in perinatal outcomes between Hispanic women and non-Hispanic White women with treated or untreated GDM and with glucose intolerance. STUDY DESIGN: We evaluated perinatal outcomes among Hispanic women and Non-Hispanic White women enrolled in a multicenter study of mild GDM treatment. Racial/ethnic comparisons were performed using multiple regression for 1) 931 women assigned to an observational cohort with glucose intolerance but not GDM (50 gram 1-hr glucose load result 135-199 mg/dl with subsequent negative 100 gram 3-hr oral glucose tolerance test), 2) 473 women with mild gestational diabetes and assigned to usual prenatal care, and 3) 485 women with mild gestational diabetes and assigned to treatment. Outcomes included: composite perinatal outcome (stillbirth, neonatal death, hypoglycemia, hyperbilirubinemia, neonatal hyperinsulinemia, birth trauma), gestational age at delivery, birth weight, and hypertensive disorders of pregnancy. RESULTS: The study cohort was 59.8% Hispanic and 27.5% non-Hispanic White. Outcomes are shown in the Table. Within the observational cohort, Hispanic women had more frequent composite perinatal outcome (aOR 1.64 95%CI 1.14, 2.37) and delivered at a greater gestational age (0.38 / 0.14 weeks, P .008), compared with nonHispanic White women. Similarly, among those diagnosed with but not treated for mild GDM, Hispanic women delivered at a greater gestational age (0.52 / 0.21 weeks, P .014). Perinatal outcomes were similar for Hispanic and non-Hispanic White women with mild GDM, regardless of treatment. CONCLUSION: Treated and untreated Hispanic women and non-Hispanic White women with mild GDM have comparable outcomes. Composite adverse perinatal outcome is more frequent among Hispanic women with glucose intolerance but without a diagnosis of GDM. PosterSessionIV Epidemiology, Infectious Disease, Intrapartum Fetal Assessment, Operative Obstetrics, Obstetric Quality & Safety, Public Health-Global Health www.AJOG.org