www.elsevier.com/locate/jpedsurg

Journal of Pediatric Surgery (2012) 47, 881–884

An analysis of early nonmortality outcome prediction inesophageal atresiaAbdullah Alshehri, Andrea Lo, Robert Baird⁎

Division of Pediatric General Surgery, The Montreal Children's Hospital, McGill University, Montreal, Quebec, Canada

Received 7 January 2012; accepted 26 January 2012

0d

Key words:Esophageal atresia;Complication;Morbidity;Outcome prediction

AbstractBackground: Advances in care for neonates with esophageal atresia (EA) has improved overall survivalrates. Disease-specific prognostic scores for EA assess mortality risk without assessing patientmorbidity. We undertook an analysis of these and generic scoring systems evaluating their ability topredict early nonmortality outcomes.Methods: We conducted a retrospective review of all patients with EA at our tertiary care children'shospital. Disease-specific (Spitz, Waterston, and Montreal) and generic prognostic scores (Score forNeonatal Acute Physiology II and Score for Neonatal Acute Physiology–Perinatal Extension) werecalculated. Outcomes investigated included mortality and early nonmortality outcomes (leak, stricture,and recurrent fistula). These were assessed individually and as a composite “poor outcome” score.Correlations were sought, and receiver operating characteristic curves were generated.Results: Fifty patients were included for analysis, with 5 deaths (10%) in our series. Eight patientsdeveloped a postoperative leak, 18 developed stenosis requiring dilatation, and 2 developedrefistulization. Overall, 51% of survivors had a poor composite outcome. Although no prognosticscore achieved statistical significance, the generic scores outperformed the disease-specific scores inpredicting early nonmortality outcomes.Conclusions: Postoperative morbidity remains common in patients with EA. Disease-specific,preexisting prognostic scoring systems do not delineate surviving patients at risk for early complicationsand appears to underperform when compared with generic prognostic scores.© 2012 Elsevier Inc. All rights reserved.

Esophageal atresia (EA) with or without tracheoesopha-geal fistula (TEF) is one of the most common surgicallycorrectable life-threatening congenital anomalies, with anoverall incidence of 1 in 2500 to 1 in 3000 [1]. The firstsuccessful repair of an EA with TEF in the neonatal periodwas performed by Haight [2] in 1941. Since then, successivepublications describing EA have demonstrated consistent

⁎ Corresponding author. Tel.: +1 514 412 4438; fax: +1 514 412 4289E-mail address: robert.baird@mail.mcgill.ca (R. Baird).

022-3468/$ – see front matter © 2012 Elsevier Inc. All rights reserved.oi:10.1016/j.jpedsurg.2012.01.041

improvement in patient survival over time. The classicdescription by Myers [3] demonstrated a linear improvementin survival from 34% (1948-1952) to 89% (1968-1972).With advances in neonatology, intensive care, neonatalanesthesia, and surgical technique, survival now approaches100% in the absence of other malformations [4-7]. Since theoriginal description of EA repair by Haight, however, theprocedure has been continuously and immutably linked topostoperative complications. In fact, Haight's originalpublication described the subsequent development of an

Table 1 Demographics of patients included in analysis

Male (%) 27 (54%)Subtype of EA (%)EA and distal TEF 39 (78%)EA with proximal and distal TEF 4 (8%)H-type TEF 4 (8%)Isolated EA 2 (4%)EA with proximal TEF 1 (2%)Birth weight (g), mean ± SD 2827 ± 794Gestational age (wk), mean ± SD 38 ± 3Congenital anomalies (%)Cardiac 36 (72%)Genitourinary 13 (26%)Gastrointestinal 6 (12%)Chromosomal 4 (8%)Musculoskeletal 2 (4%)APGAR score at 1 min, mean ± SD 7.8 ± 1.6SNAP-II score, mean ± SD 9.6 ± 10.7SNAP-PE score, mean ± SD 13 ± 14.4

SNAP-II indicates the Score for Neonatal Acute Physiology – II.SNAP-PE indicates the Score for Neonatal Acute Physiology with

882 A. Alshehri et al.

anastomotic stricture that was successfully dilated [2]. Myers[3], in 1974, demonstrated that 52% of survivors requiredeither endoscopic dilatation or operative repair for treatmentof anastomotic strictures; an identical 52% stricture rate wasdemonstrated by Konkin et al [8] nearly 30 years later.

Several prognostic criteria have been developed in aneffort to predict survival for patients with EA. Publicationsby Waterston, Poenaru, and Spitz all proposed simpledisease-specific prognostic tools based on a combination ofbirth weight, respiratory parameters, and the presence orabsence of associated anomalies. Other non–disease-specificprognostic scoring systems have since been developed andvalidated on large cohorts of neonatal patients, including theScore for Neonatal Acute Physiology II (SNAP-II) and Scorefor Neonatal Acute Physiology–Perinatal Extension (SNAP-PE) [9-11]. The purpose of this study was to assess the abilityof the preexisting scores to predict early nonmortalityoutcomes and to compare the performance of traditionaldisease-specific scores to generic (non–disease-specific)scores [9-15].

Perinatal Extension.

1. Patients and methods

After obtaining institutional review board approval, weconducted a review of all patients with EA and/or TEF over 5years at our tertiary care children's hospital. Demographicinformation was collected for each patient, as well as detailsof the atresia subtype, associated anomalies, and allcomponents of the various prognostic scoring systems. Thedisease-specific (Waterston, McGill, and Spitz) and genericprognostic scores (SNAP-II and SNAP-PE) were calculatedfor each patient. Outcomes investigated included mortalityand early nonmortality outcomes (leak, stricture, recurrentfistula). We defined early complications as those occurringwithin 1 month after surgery. These complications wereassessed individually and designated as either present orabsent. These were also summarized in a composite “pooroutcome” score, with the presence of a single earlycomplication defining a poor outcome. Follow-up waslimited to 1 year and was based on the surgeon clinicnotes. Correlations were sought with Spearman correlationcoefficient. Receiver operating characteristic (ROC) curveswere generated, and the area under the curve was calculatedto assess score performance.

2. Results

Over the 5-year study period, 50 patients with EA with orwithout TEF were included in our database (Table 1). All ofthese met inclusion criteria, and survivors were followed upfor a minimum of 1 year after surgery. The mean birth weightand gestational age were 2.8 kg and 37 weeks, respectively.Twenty-seven (54%) of the patients were males. Thirty-nine

patients had an EA and distal TEF (78%); associatedcongenital anomalies were identified in 44 patients (86%).The mean SNAP-II and SNAP-PE scores were 9.6 and 13,respectively. There were 3 twins in our data set, all of whomsurvived. Five patients were intubated before operation, 2 ofwhom died. The median age at repair was the second day oflife (range, 0-59). All patients were maintained onpostoperative antireflux medication, and 89.5% of patientshad enteric feeds initiated early in the postoperative periodvia a transanastomotic feeding tube. The median length ofstay was 26 days with a range of 1 to 383 days.

There were 5 deaths (10%) in our series. Two deathsoccurred intraoperatively and were caused by trachealagenesis and subsequent respiratory failure. Two patientsdied of sepsis at 29 and 270 days of life, respectively. Onepatient weighing 1800 g died of intractable respiratorydistress syndrome at 9 days of life. Correlation of mortalitywith predictive scores demonstrated no significant results:Waterston (P = .08), McGill (P = .08), Spitz (P = .17),SNAP-II (P = .16), and SNAP-PE (P = .10).

Overall, 51% of survivors developed at least 1 earlypostoperative complication. Eight patients developed apostoperative leak that was managed conservatively. Ofthese 8 patients, 5 went on to develop an anastomoticstricture requiring dilation, whereas 13 patients developed astricture de novo. Patients requiring esophageal dilation forstricture underwent a median of 5 procedures (range, 0-18dilatations) with all patients being maintained on oral intakealone for nutritional support at 1 year of age. Two patientsdeveloped a recurrent fistula that required a secondoperation. No significant correlation between the compositeoutcome and prognostic scores was demonstrated in our dataset (Table 2). The ROC curves comparing the performance of

Table 2 Correlation coefficients and area under the receiveroperator curve (AUC) for composite nonmortality outcomes ofsurviving patients with EA

Disease-specific Correlation AUC

Spitz P = .57 0.48Waterston P = .77 0.46Montreal P = .65 0.48GenericSNAP-II P = .25 0.63SNAP-PE P = .19 0.60

AUC indicates area under the receiver operator curve.

883Analysis of early nonmortality outcome prediction in EA

the 5 different scoring systems are displayed in Fig. 1. Noneof the scores were able to significantly predict nonmortalityoutcomes, although the generic scores (SNAP-II and SNAP-PE) outperformed disease-specific scores.

3. Discussion

Survival of children with EA and/or TEF has improveddramatically over the last 5 decades and is mainly caused bythe important advances in the perioperative care of thesefragile patients. This has included the widespread use ofgentle ventilation techniques, perioperative total parenteralnutrition, targeted antibiotic prophylaxis, and the improvedmanagement of associated congenital anomalies [16].Several scoring systems have been developed and validatedto predict survival in these patients. In 1962, Waterston andAlberdeen [12] proposed a scoring system to stratify childrenwith TEF and/or EA based on birth weight, presence of

Fig. 1 Receiver operating characteristic curves of the 3 disease-specific prognostic scores (Waterston, McGill, and Spitz) and the 2generic prognostic scores (SNAP and SNAP-PE) for patients withEA. Note: the diagonal reference line indicates no discriminationbetween poor and good outcomes.

pneumonia, or other congenital anomalies to estimate risk ofdeath. The overall reported death rate in this report was 50%.Subsequent analysis of their data set by the McGill groupsuggested that respiratory failure and the need for preoper-ative ventilation, along with the presence of severecongenital malformation, better correlated with mortality[13]. The superiority of the McGill over the Waterston scoreto predict mortality was shown in a separate cohort by Teichet al [17]. Spitz et al [15], in 1994, proposed a modifiedprognostic score emphasizing low birth weight and thepresence of major cardiac anomalies as 2 major factorsinfluencing survival. Generic scores designed to predictneonatal mortality have also been designed and validated,based mainly on the initial physiologic status of the patient.The SNAP-II and SNAP-PE have been shown to be robustprognostication scores [9-11] and have also been evaluatedin surgical subsets of patients [18,19].

Survival for patients with EA is associated with asignificant complication rate and the potential for long-term burden of disease. Early complications include thedevelopment of an anastomotic leak, recurrent fistula, andstricture requiring intervention. Late complications includegastroesophageal reflux disease, tracheomalacia, esophagealdysmotility, dysphagia, and growth retardation [4,8,14,20].Although many case series investigating outcomes ofpatients with EA report complication rates, there have beenfew publications that have specifically investigated therelationship between preoperative factors and the develop-ment of complications. The presence of tension at the level ofthe anastomosis has long been known to correlate withpostoperative complications. This was confirmed byKosloske et al, although the perception of anastomotictension remains subjective and extremely difficult toquantify [21]. A recent report investigated the developmentof primarily long-term complications, including chronicrespiratory disease and gastroesophageal reflux disease.Factors associated with the development of a complicationincluded twin birth, preoperative tracheal intubation, birthweight less than 2500 g, long gap atresia, anastomotic leak,and postoperative tracheal intubation for 5 days or more.This study provides insight into the spectrum of complica-tions facing patients with EA but fails to differentiate earlyfrom late complications. Indeed, it lists a postoperative leakas a risk factor rather than a complication [16].

This study evaluated the ability of existing prognosticscoring systems (both disease specific and generic) to predictthe development of an early postoperative complication. Themortality rate in our data set was 10% and was attributedmostly to the presence of life-threatening anomalies orpostoperative sepsis and is comparable with publishedreports [4-7]. Fifty-one percent of patients in our reportdeveloped an early complication including anastomotic leak,recurrent fistula, or stricture. This complication rate is alsosimilar to previous studies and has changed little since the1960s [16,22]. None of the 5 scores (both disease andnondisease specific) could predict nonmortality outcomes in

884 A. Alshehri et al.

a statistically significant manner. However, SNAP-II andSNAP-PE appeared to outperform disease-specific scoresbased on both their correlation coefficient and the area underthe ROC of their respective ROC curves. In particular, thegeneric scores demonstrated superior sensitivity for a givenspecificity, suggesting their trend toward superior discrim-ination in ruling out patients at risk for an early complication.

Improvement in survival for patients with EA has notbeen accompanied by a concomitant decrease in surgicalcomplication rates. Although we have not demonstratedthe ability of mortality scoring systems to predictnonmortality outcomes, generic scores (SNAP-II andSNAP-PE) appear to be at least equivalent and potentiallysuperior to traditional scores. This suggests the possibilitythat initial patient physiology may inform the futuredevelopment of early postoperative complications. Thisreview is subject to the weaknesses of retrospectiveanalyses and is therefore prone to significant bias.Furthermore, the focused time scale (chosen to minimizepractice variance) has resulted in a relatively small samplesize that may obscure true correlations. Further investiga-tion of variations in surgical technique (suture material,use of transanastomotic feeding tubes), specific physiolog-ic variables, and evaluation of traditional risk factors in aprospective manner is warranted. Determination andreduction of these risk factors for early complicationsafter EA repair could help parents and treating physiciansanticipate patient outcomes and potentially improve thelong-term prognosis of these children.

References

[1] Shaw-Smith C. Oesophageal atresia, tracheo-oesophageal fistula, andthe VACTERL association: review of genetics and epidemiology.J Med Genet 2006;43:545.

[2] Haight CTH. Congenital atresia of the esophagus with tracheoeso-phageal fistula: extrapleural ligation of fistula and end to endanastomosis of esophageal ends. Surg Gynecol Obstet 1943;76:672-88.

[3] Myers NA. Oesophageal atresia: the epitome of modern surgery. AnnR Coll Surg Engl 1974;54:277-87.

[4] Goyal A, Jones M, Couriel J, et al. Oesophageal atresia and tracheooesophageal fistula. Arch Dis Child Fetal Neonatal Ed 2006;91:F381-4.

[5] Spitz L, Kiely E, Brereton R, et al. Management of esophageal atresia.World J Surg 1993;17:296-300.

[6] Okamoto T, Takamizawa S, Arai H, et al. Esophageal atresia:prognostic classification revisited. Surgery 2009;145:675-81.

[7] Lilja HE, Wester T. Outcome in neonates with esophageal atresiatreated over the last 20 years. Pediatr Surg Int 2008;24:531-6.

[8] Konkin DE, O'Hali WA, Webber EM, et al. Outcomes in esophagealatresia and tracheoesophageal fistula. J Pediatr Surg 2003;38:1726-9.

[9] Gagliardi L, Cavazza A, Brunelli A, et al. Assessing mortality risk invery low birthweight infants: a comparison of CRIB, CRIB-II, andSNAPPE-II. Arch Dis Child Fetal Neonatal Ed 2004;89:F419-22.

[10] Points TH. SNAP-II and SNAPPE-II: simplified newborn illnessseverity and mortality risk scores. J Pediatr 2001;138:92-100.

[11] Richardson DK, Gray JE, McCormick MC, et al. Score for NeonatalAcute Physiology: a physiologic severity index for neonatal intensivecare. Pediatrics 1993;91:617-23.

[12] Waterston DJBCR, Alberdeen E. Oesophageal atresia: tracheooeso-phageal fistula. A study of survival in 218 infants. Lancet 1962:819-22.

[13] Poenaru D, Laberge JM, Neilson I, et al. A new prognosticclassification for esophageal atresia. Surgery 1993;113:426-32.

[14] Spitz L. Esophageal atresia and tracheoesophageal fistula in children.Curr Opin Pediatr 1993;5:347.

[15] Spitz LKE, Morecroft JA, et al. Oesophageal atresia: at-risk groups forthe 1999′s. J Pediatr Surg 1994;29:723-5.

[16] Castilloux J, Noble AJ, Faure C. Risk factors for short- and long-termmorbidity in children with esophageal atresia. J Pediatr 2010;156:755-60.

[17] Teich S, Barton DP, Ginn-Pease ME, et al. Prognostic classification foresophageal atresia and tracheoesophageal fistula: Waterston versusmontreal. J Pediatr Surg 1997;32:1075-80.

[18] Skarsgard ED, MacNab YC, Qiu Z, et al. SNAP-II predicts mortalityamong infants with congenital diaphragmatic hernia. J Perinatol2005;25:315-9.

[19] Baird R, Puligandla PS, St Vil D, et al. The role of laparotomy forintestinal perforation in very low birth weight infants. J Pediatr Surg2006;41:1522-5.

[20] Holland AJA, Ron O, Pierro A, et al. Surgical outcomes of esophagealatresia without fistula for 24 years at a single institution. J Pediatr Surg2009;44:1928-32.

[21] Jane McKinnon LKA. Prediction and prevention of anastomoticcomplications of esophageal atresia and tracheoesophageal fistula.J Pediatr Surg 1990;25:778.

[22] Little D, Rescorla F, Grosfeld J, et al. Long-term analysis of childrenwith esophageal atresia and tracheoesophageal fistula. J Pediatr Surg2003;38:852-6.