287: Diagnostic accuracy of clinical newborn assessment versus ultrasound to define preterm birth: the PROMISE trial

Abstract

Among HIV-infected pregnant women enrolled in PROMISE 1077BF/1077FF, antenatal antiretroviral therapy was associated with higher preterm birth (PTB) compared to zidovudine alone, when determined by New Ballard Score (NBS). In a subset of PROMISE participants, we evaluated the performance of NBS against a gold standard of ultrasound-based dating. PROMISE was a multi-country trial that compared different evidence-based antiretroviral strategies to reduce vertical HIV transmission and improve maternal health. Some enrolling sites—mostly from South Africa, Uganda, and India—provided obstetric sonography as part of the standard of care, while others did not. PTB was defined as birth <37 weeks gestation, and then secondarily at thresholds of <34 and <32 weeks. To assess the diagnostic performance of NBS for PTB (relative to ultrasound), we calculated the sensitivity, specificity, positive predictive value, and negative predictive value. In sensitivity analyses, we considered only those participants with dating ultrasounds performed <24 weeks gestation. Among 3,423 enrolled HIV-infected pregnant women, 724 (21.1%) singleton pregnancies had NBS and ultrasound data. Of these, 353 (48.7%) had ultrasounds performed <24 weeks gestation. The median gestational age at the time of ultrasound was 24.0 weeks (IQR: 19.0, 28.8), and the median gestational age at entry based on ultrasound was 25.5 weeks (IQR: 20.4, 30.7). The proportion of births categorized as <37 weeks was higher when ultrasound was used compared to NBS (18.2% vs. 15.3%), though this difference diminished when confined to women in whom ultrasound was performed <24 weeks (18.4% vs 17.8%). The NBS was highly specific (92.1%) but not sensitive (48.5%) at correctly classifying PTB <37 weeks compared to ultrasound. This finding remained consistent at PTB thresholds of <34 weeks (sensitivity: 28.9%; specificity: 99.4%) and <32 weeks (sensitivity: 40.0%; specificity: 99.7%). We observed a similar pattern among participants with ultrasounds performed <24 weeks gestation (see Table). Across different PTB thresholds, the NBS is a highly specific, but poorly to moderately sensitive tool for identifying PTB. Where ultrasound is not routinely available, NBS alone may significantly underestimate the frequency of PTB. Accurate determination of gestational age at birth can decrease the burden of perinatal morbidity and is urgently needed in many resource-limited settings.