The effect of customization and use of a fetal growth standard on the association between birthweight percentile and adverse perinatal outcome

BACKGROUND: It has been proposed that correction of offspring weight percentiles (customization) might improve the prediction of adverse pregnancy outcome; however, the approach is not accepted universally. A complication in the interpretation of the data is that the main method for calculation of customized percentiles uses a fetal growth standard, and multiple analyses have compared the results with birthweight-based standards. OBJECTIVES: First, we aimed to determine whether women who deliver small-for-gestational-age infants using a customized standard differed from other women. Second, we aimed to compare the association between birthweight percentile and adverse outcome using 3 different methods for percentile calculation: (1) a noncustomized actual birthweight standard, (2) a noncustomized fetal growth standard, and (3) a fully customized fetal growth standard. STUDY DESIGN: We analyzed data from the Pregnancy Outcome Prediction study, a prospective cohort study of nulliparous women who delivered in Cambridge, UK, between 2008 and 2013. We used a composite adverse outcome, namely, perinatal morbidity or preeclampsia. Receiver operating characteristic curve analysis was used to compare the 3 methods of calculating birthweight percentiles in relation to the composite adverse outcome. RESULTS: We confirmed previous observations that delivering an infant who was small for gestational age (<10th percentile) with the use of a fully customized fetal growth standard but who was appropriate for gestational age with the use of a noncustomized actual birthweight standard was associated with higher rates of adverse outcomes. However, we also observed that the mothers of these infants were 3-4 times more likely to be obese and to deliver preterm. When we compared the risk of adverse outcome from logistic regression models that were fitted to the birthweight percentiles that were derived by each of the 3 predefined methods, the areas under the receiver operating characteristic curves were similar for all 3 methods: 0.56 (95% confidence interval, 0.54-0.59) fully customized, 0.56 (95% confidence interval, 0.53-0.59) noncustomized fetal weight standard, and 0.55 (95% confidence interval, 0.53-0.58) noncustomized actual birthweight standard. When we classified the top 5% of predicted risk as high risk, the methods that used a fetal growth standard showed attenuation after adjustment for gestational age, whereas the birthweight standard did not. Further adjustment for the maternal characteristics, which included weight, attenuated the association with the customized standard, but not the other 2 methods. The associations after full adjustment were similar when we compared the 3 approaches. CONCLUSION: The independent association between birthweight percentile and adverse outcome was similar when we compared actual birthweight standards and fetal growth standards and compared customized and noncustomized standards. Use of fetal weight standards and customized percentiles for maternal characteristics could lead to stronger associations with adverse outcome through confounding by preterm birth and maternal obesity.