Clinical Prediction Score for Nasal CPAP Failure in Pre-term VLBW Neonates with Early Onset Respiratory Distress

Transcription

1 JOURNAL OF TROPICAL PEDIATRICS, VOL. 57, NO. 4, 2011 Clinical Prediction Score for Nasal CPAP Failure in Pre-term VLBW Neonates with Early Onset Respiratory Distress by Mrinal S. Pillai, 1 Mari J. Sankar, 1 Kalaivani Mani, 2 Ramesh Agarwal, 1 Vinod K. Paul, 1 and Ashok K. Deorari 1 1 Division of Neonatology, Department of Pediatrics, All India Institute of Medical Sciences, New Delhi, India 2 Department of Biostatistics, All India Institute of Medical Sciences, New Delhi, India Correspondence: Dr Ashok Deorari, Division of Neonatology, Department of Pediatrics, All India Institute of Medical Sciences, New Delhi , India. Tel: þ , Fax: þ <ashokdeorari_56@hotmail.com>. Summary We prospectively observed 62 pre-term very low birth weight neonates initiated on nasal continuous positive airway pressure (CPAP) for respiratory distress in the first 24 h of life to devise a clinical score for predicting its failure. CPAP was administered using short binasal prongs with conventional ventilators. On multivariate analysis, we found three variables gestation <28 weeks [adjusted odds ratio (OR) 6.5; 95% confidence interval (CI) ], pre-term premature rupture of membranes [adjusted OR 5.3; CI ], and product of CPAP pressure and fraction of inspired oxygen 1.28 at initiation to maintain saturation between 88% and 93% [adjusted OR 3.9; CI ] to be independently predictive of failure. A prediction model was devised using weighted scores of these three variables and lack of exposure to antenatal steroids. The clinical scoring system thus developed had 75% sensitivity and 70% specificity for prediction of CPAP failure (area under curve: 0.83; 95% CI ). Key words: CPAP failure, prediction score, pre-term, very low birth weight, respiratory distress. Introduction Pre-term very low birth weight (VLBW) neonates are at high risk of developing respiratory distress in the immediate post-natal period. Traditionally, these infants are managed by intubation and mechanical ventilation. In recent years, use of continuous positive airway pressure (CPAP) has gained immense popularity as the primary mode of respiratory support in these infants. Its use has been associated with reduced need for mechanical ventilation [1] and associated lung injury [2, 3]. In addition, the need for up-transfers in non-tertiary care settings has been found to be significantly reduced following use of CPAP [4]. However, not all pre-term neonates with respiratory distress respond to CPAP therapy. A recent trial that evaluated CPAP as the primary mode of support in extremely pre-term infants (gestational age <28 weeks) found a failure rate of 46% in the first few days of life [5]. The present study was therefore designed to the identify the group of infants who are likely to fail nasal CPAP and require mechanical ventilation so that the attending physician/staff are vigilant during CPAP administration and be prepared with the necessary facilities for mechanical ventilation and/or referral. Methods This prospective observational study was conducted between August 2007 and April 2009 at a level 3 neonatal unit in New Delhi, India. VLBW (<36 weeks) neonates who developed respiratory distress within 24 h of life were eligible for enrolment. We excluded infants with lethal malformations and severe perinatal asphyxia (1-min Apgar score <3). Informed written consent was obtained from one of the parents before enrolling the infants. The study was approved by the ethics committee of the institute. All pre-term VLBW infants were shifted to the neonatal intensive care unit after initial stabilization in the delivery room (unit protocol). They were monitored for development of respiratory distress over the next 24 h; nasal CPAP was initiated if the infant developed distress. Infants with severe respiratory distress (Silverman score >7), bradycardia (<100 b.p.m.), prolonged capillary refill time (>3 s), and apnoea/poor respiratory efforts were intubated and ventilated without a trial of CPAP. CPAP was delivered using short binasal prongs (Argyle prongs; Sherwood Medical St. Louis MO) with conventional ventilators SLE 2000 (Specialised Laboratory South Croydon, UK) and Babylog 8000 (Dra ger Medical Lu beck, Germany). We used initial pressures of 4 5 cm H 2 O and fraction of inspired oxygen (FiO 2 ) of The CPAP pressure and/or FiO 2 were then adjusted to maintain the oxygen saturation (SpO 2 ) between 88% and 93%: pressure was ß The Author [2010]. Published by Oxford University Press. All rights reserved. For Permissions, please journals.permissions@oup.com 274 doi: /tropej/fmq047 Advance Access published on 16 June 2010

2 increased in steps of 1 cm H 2 O up to a maximum of 7 cm H 2 O and FiO 2 in steps of 0.05 up to a maximum of 0.6. Adequacy of CPAP was monitored both clinically (respiratory rate, Silverman score, chest expansion and SpO 2 ) and radiologically (lung volume in chest X-ray). Exogenous surfactant was administered only if the infant required mechanical ventilation (early/late rescue therapy). InSurE (intubate surfactant extubate) approach was not routinely used during the study period. Infants were weaned off from CPAP when they were comfortable, had no signs of distress and maintained SpO 2 of 88 93% while on FiO 2 of <0.3. Primary outcome of the study was CPAP failure defined as requirement of mechanical ventilation for one or more of the following indications: (i) hypoxia (partial pressure of oxygen (PaO 2 ) <50 mmhg; oxygen saturation <85%); (ii) prolonged apnoea (>20 s) or recurrent apnoea (more than two episodes in 24 h); (iii) persistent/worsening respiratory distress (Silverman score 4); (iv) hypercarbia (PaCO 2 >65 mmhg) and/or acidosis (ph <7.2) despite maximum CPAP pressure of 7 cm H 2 O and FiO 2 of 0.6; and (v) systemic hypotension requiring vasopressors. Baseline characteristics such as birth weight, gender, gestational age (based on mother s last menstrual period date and/or first trimester ultrasound), exposure to antenatal steroids (ANS), mode of delivery, Apgar at 1 and 5 min, need for resuscitation, FiO 2 and CPAP pressure at initiation as well as maximum requirement, SpO 2 and Silverman score at initiation, and maternal variables like gestational hypertension, diabetes complicating pregnancy, preterm premature rupture of membranes (PPROM), abnormal umbilical blood flow, etc. were recorded in a pro forma. The cause of respiratory distress and severity of respiratory distress syndrome (RDS) based on radiological findings were noted. Other clinical data such as early onset sepsis (clinical/culture positive in the first 72 h of life), pneumothorax, patent ductus arteriosus, chronic lung disease (continuing oxygen requirement during first 28 days of life), duration of CPAP and mortality were also collected. All enrolled infants were monitored till 7 days of life for the need of mechanical ventilation. They were observed periodically for grunting, retractions, apnoea, cyanosis, abdominal girth, SpO 2, blood pressure and capillary refill time. The nasal interface (cannula) was checked 2 4 hourly for fixation, blockage and any evidence of local injury. Nursing care was provided in the form of minimal handling, ensuring fixation of prongs and patency of prongs and airway, decompression of stomach by inserting an orogastric tube, maintaining asepsis and frequent changes in position. We conducted training sessions for nursing staff every 6 months to improve success with implementation of CPAP. Statistical analysis Patient information was collected in a pro forma and data entry was done using Epi-Info v (Centers for Disease Control and Prevention, Atlanta, GA, USA). Statistical analysis was carried out using Stata 9.0 (College Station, TX, USA). Data were presented as number (%) or mean SD/median (range) as appropriate. Bivariate analysis was done using Student s t-test/wilcoxon rank-sum test for continuous data and chi-square/fisher s exact test for categorical variables. Continuous variables such as FiO 2 /Silverman score at initiation of CPAP were dichotomized at cut-offs obtained by receiver operating characteristic (ROC) curve with CPAP failure as outcome. To find the predictors of CPAP failure, univariate logistic regression analysis followed by multivariable logistic regression with backward stepwise elimination variable selection method were used [6]. In the final prediction model, we also included another variable (lack of ANS) thought to be clinically important but not found to be statistically significant in the multivariate analysis, and the results were reported as adjusted OR with 95% CI. We then devised a weighted prediction score using b-coefficients of these variables derived from logistic model (multiplied by 10 and rounded off to the nearest integer) [7]. The discriminatory power and calibration of the prediction score were evaluated using the area under the ROC curve [8] and Hosmer Lemeshow goodness-of-fit test respectively [9]. According to Hosmer Lemeshow, areas under the curve between 80% and 90% are consistent with excellent discriminatory power and calibration is adequate if the Hosmer Lemeshow statistics has a p-value of >0.05 [9]. We also calculated the optimal cut-off for the weighted score using the ROC curve and estimated the sensitivity, specificity and positive likelihood ratio of the cut-off. Results Out of the 186 neonates born at <36 weeks of gestation and weighing <1500 g at birth, 92 developed respiratory distress within 24 h of life. While 17 of these infants were started on intermittent mandatory ventilation without a trial of CPAP, 9 were managed with only oxygen by hood. After excluding 4 infants with congenital anomalies, we enrolled 62 neonates who were initiated on nasal CPAP. Baseline characteristics of the enrolled neonates are outlined in Table 1. A total of 16 (25.8%) neonates failed CPAP and required mechanical ventilation in the first week of life. While 14 infants had failure in the first 72 h, two required intubation later. Compared to those neonates who did not require ventilation (CPAP success group), those who failed CPAP ( failure group) were born at an earlier gestation: 50% in the latter group Journal of Tropical Pediatrics Vol. 57, No

3 TABLE 1 Baseline characteristics of study population (n ¼ 62) Variable Gestational age, mean (SD) (weeks) 30 (2) Birth weight, mean (SD) (grams) 1120 (238) Female gender 28 (45) Gestation <28 weeks 14 (23) Birth weight <1000 g 25 (40) ANSs Completed one course 52 (84) Single dose (partial treatment) 5 (8) No dose 5 (8) Apgar at 1 min <6 5 (8) Cause of respiratory distress RDS 44 (71) TTNB 16 (25.8) Pneumonia 2 (3.2) Duration of CPAP, median (range) (h) 12 ( ) Data presented as n (%), unless specified otherwise. TTNB, transient tachypnoea of newborn. Variable TABLE 2 Comparison of CPAP failure and success groups were born at <28 weeks compared to 13% in the success group. Among the maternal/antenatal variables, only PPROM was found to be significantly different between the two groups (Table 2). The underlying cause of respiratory distress was different between the groups almost all those who failed CPAP had either RDS or pneumonia as the underlying condition, while about one-third in the other group had delayed adaptation (transient tachypnoea of newborn) as the cause of respiratory distress. Neonates who failed CPAP were sicker than the other group, as reflected by a significantly high FiO 2 /pressure requirement and Silverman score at the time of initiation of CPAP. Also, they were more likely to be diagnosed to have sepsis and patent ductus arteriosus (Table 2). On multivariate analysis, only three variables gestation <28 weeks, PPROM, and product of CPAP pressure and FiO 2 at initiation 1.28 (used at the time of initiation of CPAP to maintain CPAP failure (n ¼ 16) CPAP success (n ¼ 46) Neonatal variables Gestational age, mean (SD) (weeks) 28.7 (2.0) 30.3 (1.9) Gestation <28 weeks 8 (50.0) 6 (13.0) Birth weight, mean (SD) (g) 1035 (289) 1150 (216) 0.11 Birth weight <1000 g 9 (56.3) 16 (34.8) 0.13 Male gender 7 (43.8) 27 (58.7) 0.30 Delivery by caesarean section 9 (56.3) 33 (71.7) 0.25 Apgar at 1 min, mean (SD) 7.4 (1.4) 7.4 (1.5) 1.00 Maternal/antenatal variables No ANS 3 (18.7) 2 (4.3) 0.07 PPROM 7 (43.8) 6 (13.0) Gestational hypertension 7 (43.8) 15 (32.6) 0.42 GDM/diabetes complicating pregnancy 2 (12.5) 3 (6.5) 0.60 Underlying condition and severity Cause of respiratory distress 0.01 Pneumonia 2 (12.5) 0 RDS 13 (81) 31 (67.4) Delayed adaptation/ttnb 1 (6.3) 15 (32.6) X-ray grading of RDS: Mild RDS 7/13 (53.8) 31/31 (100) Moderate/severe RDS 6/13 (46.2) 0 Silverman score at initiation of CPAP, mean (SD) 4.1 (1.1) 3.4 (0.7) 0.01 FiO 2 at initiation, mean (SD) 39.2 (2.9) 29.5 (1.3) CPAP pressure at initiation, mean (SD) 4.8 (0.8) 4.3 (0.7) 0.02 Product of CPAP pressure and FiO 2 at initiation, mean (SD) 1.9 (0.8) 1.3 (0.4) Product of pressure and FiO2 at initiation (68.8) 15 (32.6) 0.01 Patent ductus arteriosus 5 (35.7) 2 (4.2) Early onset sepsis 4 (25) 1 (2.2) 0.01 p Data presented as n (%), unless specified otherwise. GDM, gestational diabetes; TTNB, transient tachypnoea of newborn. 276 Journal of Tropical Pediatrics Vol. 57, No. 4

4 adequate chest expansion and SpO 2 of 88 93%) were found to be predictive of CPAP failure (Table 3). Scoring system for prediction of CPAP failure We used four variables the three variables found to be significant in the multivariate model along with lack of ANS to devise a scoring system for prediction of CPAP failure. We calculated weights for each variable from the b-coefficients derived from the logistic regression analysis. The final weighted score was calculated as follows: 20 (gestation <28 weeks) þ 18 (PPROM) þ 18 (no ANSs) þ 11 (product of CPAP pressure and FiO 2 at initiation 1.28) with each variable being assigned a value of 1 if present and 0 if absent (Table 3). Based on the distribution of infants in the two groups with respect to the weighted score and the ROC curve analysis (Table 4), the optimal cut-off score for prediction of CPAP failure was determined. The sensitivity, specificity, positive predictive value and negative predictive value of the cutoff of 18 were 75, 70, 46 and 89%, respectively; the positive and negative likelihood ratios were 2.46 and 0.36, respectively. ROC curve was then plotted to evaluate the predictive ability of the weighted score with area under the curve (AUC) being 83% [95% confidence interval (CI) 71 95%], the score had excellent model discriminatory power (Fig. 1). In addition, the goodnessof-fit test value of 6.43 with 7 df demonstrated excellent model calibration for the observed vs. predicted outcome (p ¼ 0.49). Discussion The present study is an attempt to ascertain the factors associated with failure of CPAP and to devise a prediction model/score using these factors. This could help in early identification of neonates who are at risk of failure and, therefore, require intense monitoring and supervision during CPAP administration. To the best of our knowledge, no attempt has been made so far to devise such scoring system. Variable Previous studies on predictors of CPAP failure are mostly retrospective in nature and have not attempted to develop any model for prediction [10, 11]. The prediction score developed in the present study includes four variables that can be easily ascertained at the bedside either by history (gestation, PPROM and lack of exposure to ANS) or by monitoring (product of CPAP and FiO 2 at initiation of CPAP) and is therefore a very simple tool to use. While three of these variables were derived from the multivariate model, one variable no ANS was added despite lack of statistical significance in the regression analysis. This was done because (i) in a developing country like India, usage of ANS could vary significantly a recent study on bubble CPAP from the country found that 44% of pre-term neonates did not receive steroids [12] as opposed to only 8% in our study and (ii) the lack of significance could be due to the small number of women who did not receive ANS. We used b-coefficients of the variables included in the model to develop the final weighted score. This method has been reported to be much superior TABLE 3 Logistic regression analysis of the variables related to CPAP failure Unadjusted OR (95% CI) TABLE 4 Distribution of infants with CPAP failure and success with respect to the scores obtained Risk score Adjusted OR (95% CI) CPAP failure n (%) p CPAP success n (%) Estimated regression coefficient (SE) Gestation <28 weeks 6.7 ( ) 6.5 ( ) (0.8) 20 PPROM 5.2 ( ) 5.3 ( ) (0.8) 18 Product of CPAP pressure and FiO ( ) 3.9 ( ) (0.7) 11 Not received ANSs 5.1 ( ) 5.8 ( ) (1.2) 18 Total n (%) 0 1 (4.2) 23 (95.8) 24 (100) 11 3 (25) 9 (75) 12 (100) 18 2 (33.3) 4 (66.7) 6 (100) 20 1 (20) 4 (80) 5 (100) 29 1 (20) 4 (80) 5 (100) 31 2 (50) 2 (50) 4 (100) 38 1 (100) 0 1 (100) 47 1 (100) 0 1 (100) 49 4 (100) 0 4 (100) A cut-off of 18 had 75% sensitivity and 70% specificity to predict CPAP failure. Score a (points) Final weighted score ¼ 20 (gestation <28 weeks) þ 18 (PPROM) þ 18 (no ANSs) þ 11 (product of CPAP pressure and FiO 2 at initiation >1.28) with each variable being assigned a value of 1 if present and 0 if absent. a Obtained by multiplying the regression coefficient by 10. Journal of Tropical Pediatrics Vol. 57, No

5 to the traditional model based on adjusted ORs [13]. The final score thus developed had excellent discriminatory power to predict the outcome (AUC of 83%; Fig. 1). The cut-off of 18 derived from the distribution of infants between the two groups at different scores (Table 4) and the ROC curve had reasonable sensitivity and specificity. The positive likelihood ratio of 2.5 indicates that an infant with any of the three variables namely gestation <28 weeks, PPROM or no exposure to ANS is about 2.5 times likely to require mechanical ventilation as opposed to one without them (Table 3). The individual variables associated with failure of CPAP in our study were different from that reported in previous studies: while Ammari et al. [11] reported need for positive pressure ventilation at delivery, alveolar arterial oxygen gradient (A-aDO 2 ) >180 mmhg in first arterial blood gas and severe RDS were risk factors, the other case control study from Malaysia found moderate/severe RDS, septicaemia and pneumothorax to be associated with CPAP failure [10]. In a recent prospective study from India, no or partial exposure to ANS, whiteout on the chest X-ray, patent ductus arteriosus, sepsis/pneumonia and Downes score >7 or FiO 2 >50% after min of CPAP were found to be predictive of CPAP failure [14]. This disparity could be due to the difference in the study population, varied nursing practices, coverage of ANS, nature of CPAP device used (bubble CPAP in other studies vs. ventilator-derived CPAP in ours) and the statistical methods used (univariate analysis in most of the other studies vs. multivariate model in our study). While infants born at lower gestation [5] and those who were not exposed to ANS [12] are usually FIG. 1. ROC curve for the prediction score. AUC (83%) and the goodness-of-fit test value (6.43 with 7 df; p ¼ 0.49) indicate an excellent discriminatory power and model calibration for the observed vs. the predicted outcome. thought to be at high risk for mechanical ventilation, the third variable identified in the present study, PPROM,has not so far been associated with CPAP failure. Interestingly, sepsis did not turn out to be a significant factor in the multivariate analysis. We find it difficult to explain this phenomenon the small number of infants with sepsis (as opposed to those with PPROM) could have resulted in such contradictory results. Though the fourth parameter product of CPAP pressure and FiO 2 is not conventionally used in clinical practice, we included it as a composite measure of the only two variables that are relevant during administration of CPAP. The cut-off of 1.28 for the product is, however, too low: any infant with requirement of 33% FiO 2 would then be at risk of mechanical ventilation (assuming the CPAP pressure to be 4 cm H 2 O). Fortunately, the score for this variable is less compared to that of others the optimal cut-off of 18 for the overall weighted score warrants the presence of one of the other three variables as well to reliably predict CPAP failure. The strengths of our study include: (i) a prospective follow-up of all enrolled neonates administered CPAP for its failure; (ii) devising a simple clinical prediction score; and (iii) representing the ground reality in developing countries where infants are often managed without surfactant and the number of trained nursing personnel is limited. The major limitation of our study is the small sample size a larger number of infants with the outcome of interest (CPAP failure) could have identified more variables thus enabling to devise a better score. Moreover, we administered CPAP using conventional ventilators and short binasal cannulae and InSurE approach was not employed. Therefore, we need to be cautious before generalizing the results to other units that use bubble CPAP or employ InSurE approach regularly. Additional research addressing some of these limitations will be needed to validate our findings. To conclude, a simple clinical score comprising four variables namely, gestational age <28 weeks, PPROM, lack of exposure to ANS, and product of CPAP pressure and FiO would predict failure of nasal CPAP in pre-term VLBW infants with reasonable accuracy. References 1. Roberton NR. Early nasal CPAP reduces the need for intubation in VLBW infants. Eur J Pediatr 1998;157: Attar MA, Donn SM. Mechanisms of ventilatorinduced lung injury in premature infants. Semin Neonatol 2002;7: Apisarnthanarak A, Holzmann-Pazqal G, Hamvas A, et al. Ventilator-associated pneumonia in extremely 278 Journal of Tropical Pediatrics Vol. 57, No. 4

6 preterm neonates in a neonatal intensive care unit. Pediatrics 2003;112: Singh M, Deorari AK, Khajuria RC, Paul VK. A four year study on neonatal morbidity in a New Delhi hospital. Indian J Med Res 1991;94: Morley CJ, Davis PG, Doyle LW, et al. COIN Trial Investigators. Nasal CPAP or intubation at birth for very preterm infants. N Engl J Med 2008;358: Hosmer DW, Lemeshow S. Applied Logistic Regression. 2nd edn. New York: John Wiley & Sons Inc, Makdisse M, Ramos L. A risk score for predicting peripheral arterial disease in individuals 75 years or older. Arq Bras Cardiol 2007;88: Hanley JA, McNeil BJ. The meaning and the use of the area under a receiver operating characteristic (ROC) curve. Radiology 1982;143: Lemeshow S, Hosmer DW. A review of goodness of fit statistics for use in the development of logistic regression models. Am J Epidemiol 1982;115: Boo NY, Zuraidah AL, Lim NL, Zulfiqar MA. Predictors of failure of nasal continuous positive airway pressure in treatment of preterm infants with respiratory distress syndrome. J Trop Pediatr 2000;46: Ammari A, Suri M, Milisavljevic V, et al. Variables associated with the early failure of nasal CPAP in very low birth weight infants. J Pediatr 2005;147: Urs PS, Khan F, Maiya PP. Bubble CPAP - a primary respiratory support for respiratory distress syndrome in newborns. Indian Pediatr 2009;46: Moons KG, Harrell FE, Steyerberg EW. Should scoring rules be based on odds ratios or regression coefficients? J Clin Epidemiol 2002;55: Koti J, Murki S, Gaddam P, Reddy A, et al. Bubble CPAP for respiratory distress syndrome in preterm infants. Indian Pediatr 2009 May 20. [Epub ahead of print]. Journal of Tropical Pediatrics Vol. 57, No