Can we improve the gestation-adjusted projection (GAP) method for prediction of birth weight in morbidly obese women?

Objective: The gestation-adjusted projection (GAP) is a method to predict birthweight using population birth data and third trimester ultrasound fetal weight. This method usually utilizes population birth weight data from almost 40 years ago. In 2011, a large cohort of racially diverse infants across the US was included to validate updated birth curves. Our objective was to determine if the updated data would improve the accuracy of the GAP method during the third trimester among obese women. Methods: This secondary analysis of a cohort study included singleton pregnancies of obese women who had fetal growth assessment(s) in the third trimester. The first subgroup (N = 235) included women with a BMI >40 kg/m2 who had ultrasounds during 30 + 0-35 + 0 weeks (EARLY) and greater than 35 + 0 weeks (LATE). The second subgroup (N = 431) included women with a BMI 30-35, 40-50, or >50 kg/m2 who had an ultrasound during 34 + 0-36 + 6 weeks. Mean absolute percent error was calculated for all GAP methods and compared using paired t-tests. Sensitivity, specificity, and area under the curve for diagnosis of birth weight >4000 grams were also estimated for each GAP method. Results: The mean absolute percent error for the first subgroup (N = 235) using historical population birth weights was 7.4-7.9%. After using updated population birth weight curves using all neonates, the mean absolute percent error for the first subgroup ranged between 7.6 and 9.4%. GAP predictions using all neonates, as well as male and female-specific birth data compared to the historical population data during both the EARLY and LATE periods were significantly worse (p < .01). The mean absolute percent error for the second subgroup (N = 431) using historical population birth weights ranged from 7.2 to 7.9%. The absolute percent error using gender-specific compared to historical data was significant in the BMI 30-35 group (male 8.1% versus historical 7.6%, p < .01, female 8.1% versus historical 7.6%, p < .01). The differences in absolute percent error between historical and updated population data became less evident in the BMI 40-50 and >50-kg/m2 groups (p = .05 and p = .15, respectively) though still overall performed worse with the updated data. Conclusions: Prediction of birth weight using the GAP method does not seem to be improved among obese women after using updated population data. Alternatively, modeling techniques may need to be applied to improve the accuracy of the GAP method.

Duke Scholars

Author Samantha Thomas  

Author Brita Katherine Boyd Obstetrics and Gynecology, Maternal Fetal Medicine