718 Comparison of growth curves for predicting small for gestational age infants among contemporary US populations

Abstract

To compare frequency of small for gestational age (SGA) infants using popular growth curves versus race/ethnicity-specific growth curves and investigate differences in performance by week of gestational age (GA) at birth. Retrospective population-based cohort study of all US live births, stillbirths, and neonatal deaths from 35-40 weeks GA using US vital statistics data, 2012-2016. Birth outcomes were compared between 4 strata of maternal race/ethnicity: non-Hisp White (NHW), non-Hisp Black (NHB), Hispanic (Hisp), and non-Hisp Asian (NHA). The primary outcome was percent of births diagnosed as SGA <10th percentile for GA, defined by the Alexander, Hadlock, and NICHD ethnicity-specific growth curves. 17,750,507 live births, stillbirths, and neonatal deaths occurred in the U.S. from 2012-2016 from 35-40 weeks. Of those, 12.1% were categorized as SGA by either Alexander, Hadlock, or NICHD curves. The frequency of SGA diagnosis by Alexander was lowest overall for NHW(8.4%), followed by Hisp(9.3%), NHA(13.4%), and highest for NHB(16.3%), p<0.001. Hadlock yielded the lowest overall frequency of SGA for NHW(8.5%), followed by Hisp(9.5%), NHA(13.7%), and highest for NHB(16.4%). The NICHD race/ethnicity specific curves yielded the lowest frequency of SGA in Hisp(6.9%), followed by NHA(8.7%), NHB(11.5%), and highest in NHW(12.4%). NICHD curves estimated highest rates of SGA in NHW compared to Alexander and Hadlock. However, for NHB, Hisp, and NHA births, NICHD curves estimated the lowest rates of SGA. Among all curves and race/ethnicities, the highest SGA rates were observed at 35 weeks, max estimate 26.1%. The lowest SGA rates were observed at 39-40 weeks, min estimate 5.3%. The largest variation in SGA diagnosis was noted between the Alexander and NICHD curves in the NHB group, rate difference 7.5% (CI 7.3%-7.7%). SGA diagnoses are over-represented at earlier gestational ages. Using ethnicity-specific growth curves may lower the rate of over-estimation and reduce the rate of unnecessary intervention. Updating growth curves may also improve overall diagnostic capabilities.