Addition of SNAP to perinatal risk factors improves the prediction of bronchopulmonary dysplasia or death in critically ill preterm infants

To investigate the early prediction of BPD or death, the potential perinatal risk factors obtained during the first 72 hours of life were studied. Maternal data including chorioamnionitis, premature rupture of the membranes (PROM) >24 hours, gestational diabetes mellitus (GDM), pre-GDM, hypertension, pre-eclampsia, antenatal steroids, and mode of delivery, as well as neonatal data including gestational age, birth weight, gender, the Apgar scores, and delivery room resuscitation course were collected. Antenatal steroids were considered to have been administered if at least one dexamethasone injection was given 12 hours before delivery. Gestational age was calculated from the mother’s menstrual history and was confirmed by ultrasonography.

Clinical status, medication, and therapeutic interventions of the infants were recorded daily during the first 72 hours of life. We evaluated all infants for the presence of perinatal asphyxia, respiratory distress syndrome (RDS), transient tachypnea of the newborn, apnea of prematurity, pneumothorax, pulmonary hemorrhage, PDA, early onset sepsis, necrotizing enterocolitis (NEC), hypoxic-ischemic brain damage, intraventricular hemorrhage (IVH), and acute kidney injury.

In addition, we examined the use of postnatal steroids and surfactants at any time during the first 72 hours of life. Respiratory support, including oxygen, continuous positive airway pressure (CPAP), invasive mechanical ventilation, and the duration of therapies, were recorded. We assigned the respiratory support as none, nasal cannula, nasal CPAP, or invasive mechanical ventilation.

The original SNAP score was prospectively calculated for each infant based on 28 items collected during the first 24 hours of admission, including blood pressure, heart rate, respiratory rate, temperature, partial pressure arterial oxygen/fractional inspired oxygen (PaO₂/FiO₂) ratio, PaO₂, partial pressure of carbon dioxide (PaCO₂) in arterial blood, oxygenation index, hematocrit, white blood cell count, immature total ratio, absolute neutrophil count, platelet count, blood urea nitrogen, serum creatinine (Cr), urine output, serum indirect bilirubin, serum direct bilirubin, serum sodium, serum potassium, serum ionized calcium, serum total calcium, serum glucose, serum bicarbonate, arterial pH, seizure, apnea, and stool guaiac [11]. All the data were prospectively collected by trained graduate students under the supervision of the attending neonatologists and the chief of neonatology.

The outcome measures included the development of BPD or death before the time of BPD screening [20].

Infants were categorized as having mild, moderate, or severe BPD, according to the National Institutes of Health consensus definition of BPD [5, 22]. Mild BPD was defined as receiving supplemental oxygen for ≥28 days but not at a postmenstrual age (PMA) of 36 weeks or at discharge; moderate BPD as receiving oxygen for ≥28 days plus treatment with <30% oxygen at 36 weeks’ PMA; and severe BPD as receiving oxygen for ≥28 days plus ≥30% oxygen and/or positive pressure at 36 weeks’ PMA.

Apnea of prematurity was defined as cessation of breathing that lasts more than 15 seconds and is accompanied by bradycardia or hypoxia [23]. Diagnosis of PDA was based on echocardiogram and color Doppler test. PDA was considered as significant when the following criteria were met, as described previously: (a) a ductal diameter > 1.5 mm; (b) a left ventricular output index of more than 300 ml/kg/min; (c) left atrium to aortic root ratio greater than 1.5; (d) retrograde or absent diastolic flow in the cerebral anterior artery or in the descending thoracic aorta [24]. The criteria for the diagnosis of RDS were in accordance with the European consensus on the management of neonatal RDS in preterm infants. The diagnosis was confirmed on chest x-ray with a classical ‘ground glass’ appearance and air bronchograms. Infants with or at high risk of RDS were given a natural surfactant preparation [25, 26]. Perinatal asphyxia was defined according to the following criteria: (1) pH ≤ 7.0 or base deficit ≥ 16 mmol/l in cord blood or during the first 1 hour after birth; or (2) If the pH was between 7.01 and 7.15, or the base deficit was between 10 and 15.9 mmol/L, or an arterial blood gas value was not available, either a 10 min-Apgar score ≤ 5 or assisted ventilation for ≥ 10 minutes from birth was required [27]. Once these criteria were met, all neonates underwent a standardized neurologic examination performed by neonatologists. Encephalopathy was defined as the presence of one or more signs in at least three of the following six categories (1) level of consciousness; (2) spontaneous activity; (3) posture; (4) tone; (5) primitive reflexes; and (6) autonomic nervous system signs [28]. Preterm infants who met the criteria for diagnosis of encephalopathy were considered as having hypoxic-ischemic brain damage, and diagnosis was confirmed by MRI examination. Diagnosis of IVH was based on the findings on head ultrasound and/or MRI according to the criteria of Papile et al. [29]. A head ultrasound was routinely performed in all preterm infants and in infants diagnosed with perinatal asphyxia. A brain MRI was conducted only in preterm infants who presented with signs and symptoms suggestive of encephalopathy or brain damage during the hospitalization, as described in our previous study [30]. The NEC diagnosis was made according to the modified Bell’s staging criteria. Infants with stage II or greater were defined as having NEC [31, 32]. The diagnostic criteria of sepsis were in accordance with our previous study. Infants with positive clinical and/or laboratory screen and positive culture were defined as having sepsis [33]. EOS refers to sepsis presenting in the first 72 hours of life [34]. The diagnosis of AKI was based on serum Cr >1.5 mg/dl (132.6 μmol/l) and/or a ≥150% increase in serum Cr from baseline [33, 35, 36].

The indication for oxygen therapy in the newborn was PaO₂ <50 mm Hg (6.6 kPa) or transcutaneous oxygen saturation <85% in room air, except in the presence of cyanotic congenital heart disease. Inability to maintain oxygenation with nasal cannula (FiO₂ >60% to maintain PaO₂ >50 mm Hg or arterial blood pH <7.25) or apnea unresponsive to stimulation is an indication for CPAP. The indications for mechanical ventilation were respiratory failure in newborns: PaO₂ <50 mm Hg on FiO₂ >60%, PaCO₂ >60 mm Hg (8.0 kPa), arterial pH <7.20 or prolonged or repetitive unresponsive apnea associated with bradycardia or cyanosis.

Statistical analyses were performed with SPSS 13.0. Bivariate and multivariate logistic regression was performed to identify potential perinatal risk factors associated with adverse outcome. Variables with p <0.2 in the bivariate analysis were entered into a backward stepwise multivariate regression model. Model fit was assessed with the Hosmer-Lemeshow goodness-of-fit test. A non-significant value for the Hosmer-Lemeshow Chi-square test suggests an absence of biased fit. Analysis of the area under the curve (AUC) of the Receiver Operating Characteristic (ROC) curve was constructed to assess the prognostic performance. The nonparametric method of Delong was used to compare significant difference between AUCs (Sigmaplot 10.0 software). Sensitivity and specificity were also determined. All probability values are two-sided. Differences with p values <0.05 were considered to be statistically significant.

Preterm infants with adverse outcome of BPD or death were more likely than controls to have a lower gestational age at delivery, lower birth weight, lower 5 min-Apgar score, and higher SNAP score (Table 1). The impact of SNAP (Odds ratio [OR] per 1-point increase = 1.27, 95% confidence interval [CI] 1.13-1.41, p <0.001) and 5 min- Apgar score (OR 0.72, 95% CI 0.57-0.91, p =0.005) persisted when controlling for gestational age. Birth weight, however, did not remain associated with BPD or death when adjusted for gestational age. Gender and 1 min- Apgar score were not associated with BPD or death at the unadjusted bivariate level. Furthermore, SNAP (OR 1.21, 95% CI 1.07-1.39, p =0.004), 5 min-Apgar score (OR 0.59, 95% CI 0.43-0.81, p =0.001) and gestational age (OR 0.37, 95% CI 0.23-0.58, p <0.001) remained associated with BPD or death after controlling for the severity of BPD.

Among infants with BPD or death, apnea of prematurity, RDS, and PDA were observed more frequently than in the control group. The association of RDS and PDA with BPD or death remained significant after controlling for gestational age. In contrast, the presence of early onset sepsis, hypoxic-ischemic brain damage, IVH, NEC or pneumothorax was not associated with BPD or death.

Variables with p <0.2 in the bivariate analysis were entered into a backward stepwise multivariate logistic regression analysis to identify the factors independently associated with increased risk of BPD or death in critically ill preterm infants. The final model retained gestational age, SNAP, the presence of apnea of prematurity and PDA, and surfactant use as significant predictors, with 5 min- Apgar score as a confounder (p =0.071) (Table 2). The Hosmer-Lemeshow goodness-of-fit test for the final model was not significant (p =0.333), indicating that the model fits the data adequately.

Based on the above-mentioned association between SNAP and adverse outcome of BPD or death, SNAP could be a useful prognostic predictor in critically ill preterm infants. To assess the potential to predict BPD or death, we compared the predictive value of SNAP with other prognostic predictors demonstrated by multivariate regression analysis (Table 3). Gestational age and SNAP appeared to be significant predictors, and achieved AUC of 0.83 (95% CI 0.76-0.89, p <0.001) and 0.78 (95% CI 0.70-0.85, p <0.001) respectively, for predicting the development BPD or death. When combining SNAP and gestational age, the predictive performance improved (AUC 0.88, 95% CI 0.84 to 0.94, p <0.001) over that of gestational age alone, but not reaching statistical significance (p =0.178).

Multivariate logistic regression analysis resulted in a superior final model containing variables of SNAP, gestational age, the presence of apnea of prematurity and PDA, and surfactant use as independent risk factors. With an AUC of 0.92 (95% CI 0.87-0.96, p <0.001), the prognostic accuracy of the final model was superior to the accuracy of the other predictors. By using the nonparametric method of Delong, the prognostic performance of the final model was significantly better than gestational age (p =0.026) or SNAP (p =0.002) alone. Figure 1 shows the ROC curves and AUC for the final model and for SNAP and gestational age.

The sensitivity and specificity of risk factors for the development of BPD or death are shown in Table 4. SNAP displayed sensitivity of 79% and specificity of 71% to predict BPD or death at the optimal cut-off score of 10.5, and the positive and negative likelihood ratios were 2.7 and 0.29, respectively. Gestational age displayed sensitivity 81% and specificity 73% at the optimal cut-off value of 31 weeks for predicting the development of BPD or death, with the positive and negative likelihood ratios of 3.0 and 0.26, respectively. When combining the variables of SNAP and gestational age, the sensitivity and specificity increased to 84% and 77%, with the positive and negative likelihood ratios of 3.8 and 0.20, respectively. Meanwhile, when we combined the variables of SNAP, gestational age, apnea of prematurity, PDA, and surfactant use in the final model, specificity increased to 90%, and positive likelihood ratio increased to 7.4.