Supplementary Material for: External validation of a multivariate model for targeted surfactant replacement

Background and Aim: Early targeted surfactant therapy for preterm infants is recommended but the best criteria to personalize treatment are unclear. We validate a previously published multivariate prognostic model based on gestational age (GA), lung ultrasound score (LUS) and oxygen saturation to inspired oxygen fraction ratio (SatO2/FiO2) using an independent data set. Design: Pragmatic, observational study in 10 Italian and Spanish NICUs, including preterm babies (250 and 336 weeks divided in 3 gestational age intervals) with clinical signs of respiratory distress syndrome and stabilized on CPAP. LUS and SatO2/FiO2 were collected soon after stabilization. Their prognostic accuracy was evaluated on the subsequent surfactant administration by a rigorously masked physician. Results: One hundred seventy-five infants were included in the study. Surfactant was given to 74% infants born at 25 -27 weeks, 38.5% at 28-30 weeks and 26.5% at 31-33 weeks. The calibration curve comparing the validation and the development populations showed significant overlap with an intercept=0.08, 95%C.I. (-0.34; 0.5) and a slope=1.53, 95% C.I. (1.07-1.98). The validation cohort had a high predictive accuracy. Its ROC curve showed an AUC=0.95, 95%C.I. (0.91-0.99) with sensitivity=0.93, 95%C.I. (0.83-0.98), specificity = 0.81, 95% C.I. (0.73-0.88), PPV = 0.76, 95%C.I. (0.65-0.84), NPV=0.95, 95%C.I. (0.88-0.98). LUS≥9 demonstrated the highest sensitivity (0.91, 95%C.I. (0.82-0.97]) and specificity = 0.81, 95%C.I. (0.72-0.88) as individual predictor. LUS and SatO2/FiO2 prognostic performances varied with gestational age. Conclusions: We validated a prognostic model based on LUS and Sat/FiO2 to facilitate early, customized surfactant administration that may improve respiratory management of preterm neonates研究背景与目的：早产患儿早期靶向肺表面活性物质治疗已被临床推荐，但用于个体化治疗的最佳评估标准尚不明确。本研究利用独立数据集，对既往发表的一款基于胎龄（gestational age, GA）、肺超声评分（lung ultrasound score, LUS）以及血氧饱和度/吸入氧浓度比（oxygen saturation to inspired oxygen fraction ratio, SatO2/FiO2）的多变量预后模型进行验证。 研究设计：一项在意大利与西班牙10家新生儿重症监护病房（Neonatal Intensive Care Unit, NICU）开展的务实性观察研究，纳入胎龄25~33+6周（划分为3个胎龄区间）、存在呼吸窘迫综合征临床体征且经持续气道正压通气（continuous positive airway pressure, CPAP）病情稳定的早产婴儿。于患儿病情稳定后即刻采集LUS与SatO2/FiO2数据，由严格实施盲法的医师评估二者对后续肺表面活性物质给药的预测准确性。 研究结果：本研究共纳入175例受试婴儿。其中25~27周胎龄患儿中74%接受了肺表面活性物质治疗，28~30周胎龄患儿的给药占比为38.5%，31~33周胎龄患儿的给药占比为26.5%。校准曲线对比验证队列与模型开发队列显示，二者存在显著重叠：截距为0.08，95%置信区间（CI）为(-0.34, 0.5)，斜率为1.53，95%CI为(1.07~1.98)。验证队列展现出较高的预测准确性：其受试者工作特征（Receiver Operating Characteristic, ROC）曲线下面积（Area Under Curve, AUC）为0.95，95%CI为(0.91~0.99)；灵敏度为0.93，95%CI为(0.83~0.98)；特异度为0.81，95%CI为(0.73~0.88)；阳性预测值（Positive Predictive Value, PPV）为0.76，95%CI为(0.65~0.84)；阴性预测值（Negative Predictive Value, NPV）为0.95，95%CI为(0.88~0.98)。当LUS≥9分时，作为单一预测指标，其灵敏度可达0.91（95%CI为0.82~0.97），特异度为0.81（95%CI为0.72~0.88）。此外，LUS与SatO2/FiO2的预测性能随胎龄存在显著差异。 研究结论：本研究验证了一款基于LUS与SatO2/FiO2的预后模型，可助力实现早期个体化肺表面活性物质给药，有望改善早产新生儿的呼吸管理策略。