Heterogeneous effect of gestational weight gain on birth weight: quantile regression analysis from a population-based screening.

PURPOSE: Classical regression models might give an incomplete picture of the associations between predictors and outcomes. We investigated associations between gestational weight gain (GWG) and birth weight along the entire birth weight distribution with quantile regression and estimated effects of hypothetical prevention strategies. METHODS: The GWG-birth weight association was analyzed using quantile and classical regression models on data from a population-based gestational diabetes screening (n = 4760) at the Szent Imre Teaching Hospital in Budapest, Hungary (2002-2005). Birth weight distributions were modeled based on hypothetical GWG changes. RESULTS: At a body mass index of 20 kg/m(2), a 1-kg difference in GWG was associated with a 14.2 g (95% confidence interval, 10.0-20.9) higher birth weight at the fifth percentile of the birth weight distribution and a 29.0 g (21.3-35.6) higher birth weight at the 95th percentile. The coefficient from linear regression was 20.7 (17.5-24.0). Estimates differed modestly between the two regressions at a body mass index of 30 kg/m(2). A population-wide 2-kg decrease in GWG would rather affect the risk of macrosomia (-1.8%) than that of low birth weight (+0.4%). In contrast, a 3-kg decrease in GWG among overweight and obese women would lower macrosomia more modestly (-0.8%). CONCLUSIONS: A population-wide lowering of GWG would lead to greater improvements in the right tail of the birth weight distribution.

Maternal glycemia and risk of large-for-gestational-age babies in a population-based screening.

OBJECTIVE: Gestational diabetes is a risk factor for large-for-gestational-age (LGA) newborns, but many LGA babies are born to mothers with normal glucose tolerance. We aimed to clarify the association of maternal glycemia across the whole distribution with birth weight and risk of LGA births in mothers with normal glucose tolerance. RESEARCH DESIGN AND METHODS: We undertook a population-based gestational diabetes screening in an urban area of Hungary in 2002-2005. All singleton pregnancies of mothers >or=18 years of age, without known diabetes or gestational diabetes (World Health Organization criteria) and data on a 75-g oral glucose tolerance test at 22-30 weeks of gestation, were included (n = 3,787, 78.9% of the target population). LGA was determined as birth weight greater than the 90th percentile using national sex- and gestational age-specific charts. RESULTS: Mean +/- SD maternal age was 30 +/- 4 years, BMI was 22.6 +/- 4.0 kg/m(2), fasting blood glucose was 4.5 +/- 0.5 mmol/l, and postload glucose was 5.5 +/- 1.0 mmol/l. The mean birth weight was 3,450 +/- 476 g at 39.2 +/- 1.2 weeks of gestation. There was a U-shaped association of maternal fasting glucose with birth weight (P(curve) = 0.004) and risk of having an LGA baby (lowest values between 4 and 4.5 mmol/l, P(curve) = 0.0004) with little change after adjustments for clinical characteristics. The association of postload glucose with birth weight (P = 0.03) and the risk of an LGA baby (P = 0.09) was weaker and linear. CONCLUSIONS: Both low and high fasting glucose values at 22-30 weeks of gestation are associated with increased risk of an LGA newborn. We suggest that the excess risk related to low glucose reflects the increased use of nutrients by LGA fetuses that also affects the mothers' fasting glucose.