Journal of Perinatology (2006) 26, 93–99 r 2006 Nature Publishing Group All rights reserved. 0743-8346/06 $30 www.nature.com/jp ORIGINAL ARTICLE New insights into spontaneous intestinal perforation using a national data set: (1) SIP is associated with early indomethacin exposure JT Attridge1, R Clark2,4, MW Walker2,3 and PV Gordon1,4 1 University of Virginia Children’s Hospital, Charlottesville, VA, USA; 2Pediatrix Medical Group Inc., Sunrise, FL, USA and 3The Children’s Hospital, Greenville Hospital System, Greenville, SC, USA Background: Spontaneous intestinal perforation (SIP) is increasingly common in the premature infant and is associated with significant morbidity. Indomethacin use has been implicated as a co-risk factor for SIP when combined with glucocorticoids, but previous evidence argued against indomethacin being an independent risk factor when used prophylactically. Objectives: (1) To establish a homogeneous cohort of SIP patients in a national data set and to contrast them to patients with surgical necrotizing enterocolitis (NEC). (2) To test the hypothesis that early postnatal indomethacin is independently associated with SIP. Methods: A large de-identified data set was retrospectively queried by diagnosis, and then multiple antenatal and post-natal variables were tested by both univariate and multivariate analysis to identify associations with SIP. Sub-analyses were also performed to look at the timing of drug administration. Results: There were 2105 patients evaluated in the data set. Patients were divided into matched controls (n ¼ 581), those with SIP without report of NEC (n ¼ 633) and those with NEC requiring surgery (n ¼ 891). Infants with SIP were more likely to have a patent ductus arteriosus and more likely to be treated with vasopressors than either control or NEC patients. Compared to infants with NEC, patients with SIP were smaller, less mature and required more support. SIP was also diagnosed earlier than NEC (median of 7 vs 15 days). Patients with SIP were more likely to be treated with indomethacin, hydrocortisone or both on days of life 0–3 than controls. Conclusions: (1) Surgical NEC and SIP have significant differences in presentation, demographics and morbidity. (2) A detailed look at drug timing revealed that early post-natal indomethacin is independently associated with SIP. Correspondence: Dr PV Gordon, University of Virginia, Box 800386, Charlottesville, VA 22908, USA. E-mail address: pvg4n@virginia.edu 4 These authors contributed equally to the design, oversight and publication of this project, and should be considered co-senior authors for the purpose of citation. Received 15 August 2005; revised 18 October 2005; accepted 21 October 2005; published online 1 December 2005 Journal of Perinatology (2006) 26, 93–99. doi:10.1038/sj.jp.7211429; published online 1 December 2005 Keywords: steroids; neonate; necrotizing enterocolitis; intestinal perforation; indomethacin Introduction Spontaneous intestinal perforation (SIP), also known as focal small bowel perforation, is an acquired form of neonatal bowel disease that is distinct from necrotizing enterocolitis (NEC) and has been documented within the low-birth-weight population for over 20 years.1–4 Despite this history, distinguishing SIP from NEC in large clinical databases has not been straightforward and there have been no previous examinations of SIP that use a nationally available neonatal database. We sought to accomplish this in the Pediatrix Medical Group (Pediatrix) community by making the diagnostic categories of SIP and NEC mutually exclusive within its electronic administrative database. Our purpose was twofold: first, to test the hypothesis that SIP and NEC are significantly different in their demographics, presentation and outcomes, and second, to create a database large enough for complex analyses of risk factors. Within this data set the potential association of indomethacin as a risk factor for SIP was examined. As early as 1981, indomethacin administration for a patent ductus arteriosus (PDA) was suggested to be associated with SIP in case reports.5 However, the largest multicenter, randomized, controlled trial of indomethacin to date (the TIPP trial) did not find the prophylactic use of indomethacin to be associated with an increased incidence of SIP (Figure 1).6 This trial was well-designed, but despite its size, was not sufficiently powered for the secondary outcome of SIP. In contrast, two randomized controlled trials of post-natal glucocorticoids found that the combination of early post-natal indomethacin with glucocorticoids resulted in a synergistic increase in SIP.7,8 These findings are perplexing, since the Watterberg study also found that elevated endogenous cortisol levels were associated with SIP.8 If one presumes that the etiologies for elevated cortisol would be normally randomized, it is unclear why a trend towards SIP was not observed Insight into spontaneous intestinal perforation J Attridge et al 94 Figure 1 Graph of odds ratios for SIP with indomethacin or indomethacin þ glucocorticoids in three prospective multicenter, randomized, controlled trials. The TIPP trial (Schmidt et al.6) demonstrates no effect of prophylactic indomethacin (n ¼ 68 cases of SIP/1202 total enrollment). The NICHD trial (Stark et al.7) demonstrates a significant effect with indomethacin and dexamethasone, both given within the first 24 h of life (n ¼ 18 cases of SIP/220 total enrollment). The PROPHET trial (Watterberg et al.8) demonstrates a significant effect with indomethacin and early post-natal hydrocortisone (n ¼ 21 cases of SIP/360 total enrollment). Sixty-three percent of all indomethacin was given within the first 48 h within the PROPHET trial. The odds ratios for the TIPP trial are the same as published. The odds ratios for the NICHD and PROPHET trials are calculated based on comparison cohorts created by combining patients treated with study drug and indomethacin against all other study patients. in the TIPP trial. We hypothesized that an independent association between early post-natal indomethacin and SIP would be detected in our data set, which contains a larger SIP cohort than all currently published reports of SIP. Methods The de-identified administrative data set Clinical data on these neonates were recorded during the time that care was provided in the NICU (as described previously).9 Admission, discharge and daily progress notes were generated using a computer-assisted tool, and the data were stored in an electronic database. These data were then consolidated within the Pediatrix data warehouse where they were de-identified, made HIPAA compliant and configured into tables that could be joined and queried for statistical analyses. The use of the data reported here and certification of the data as de-identified was approved by the University of Virginia Human Investigation Committee (Charlottesville, VA, USA) and the Wichita Medical Research and Education Foundation (Wichita, KS, USA). Data on estimated gestational age represented the best estimate based on both obstetrical data and neonatal examination. Data on mother’s race were based on the options contained in the database: white, black, Hispanic, Native American and Asian populations. The study cohort A retrospective cohort of neonates from the Pediatrix de-identified administrative data set was assembled in order to compare neonates diagnosed with intestinal perforation to a control group. This was accomplished by searching for all reports of ‘NEC’, ‘perforated bowel’, ‘ileal perforation’ and/or ‘isolated perforation Journal of Perinatology Figure 2 Schematic demonstrating our cohort selection strategy for SIP and surgical necrotizing enterocolitis (NEC). The control cohort was matched for weight (within 40 g, 0.02 to þ 0.02), gestational age and gender with that of the SIP cohort. Not shown are 59 gastric perforations in our database, which were excluded from this study. of the bowel’ within specific tables: ‘Patients’, ‘Admissions’, ‘Medications’ and ‘Diagnoses’. Based on this search, each patient was assigned to one of four diagnostic groups: isolated or ileal perforation with no report of NEC (generally diagnosed by surgical record), perforated bowel with no report of NEC, NEC with perforation and NEC with a report of surgical treatment (see Figure 2). Statistical evaluation of these groups showed that neonates with isolated and ileal perforations, and those with bowel perforation but no report of NEC were statistically the same for each of the demographic and treatment characteristics we describe (data not shown). Similarly, neonates with NEC that required surgical intervention were very similar to neonates with NEC and a report of perforation. It became apparent that a third distinctive group of neonates was present in the data set – those with bowel perforation and a report of a major anomaly. These patients were excluded from subsequent analyses. A control group was identified to define the risk factors associated with isolated bowel perforation by matching a random sample of controls to the patients with a report of isolated or ileal perforations. Patients were matched by gestational age, birth weight (within 40 g, 0.02 to þ 0.02) and gender. This matching algorithm was designed to allow us to evaluate the effect of medical intervention on the occurrence of bowel perforations. To evaluate the use of medications previously reported to be associated with bowel perforation,8,10–14 the ‘Medications’ table was queried for any report of ‘hydrocortisone’, ‘dexamethasone’ or ‘indomethacin’ within each of the first 14 days of life in both control and perforation cohorts to look at each day of exposure. Insight into spontaneous intestinal perforation J Attridge et al 95 This search strategy allowed us to compare the drug exposures antecedent to perforation and to compare them with the matched control cohort within discrete windows of exposure. For early (defined as 0–3 days of life) and intermediate (4–7 days of life) windows of drug exposure, direct comparisons were performed (e.g., 0–3 days vs 0–3 days and 4–7 days vs 4–7 days) as the cohorts remained comparable in size (with the intermediate window losing 20% of the SIP cohort due to incidence of perforations in the early window). However, for our 2-week exposure window, there was no time of diagnosis to retrospectively search from in the control group. Had we simply started at day 0 in the SIP cohort, we faced the issue of increasing patient drop-out (statistical censoring of data). To address these issues, we employed the following strategy: for the control group, days of life 0–14 were queried for a report of the medications of interest, and for the SIP group, the 14-day window prior to diagnosis was queried (or back until day of life 0). Fourteen days was chosen after examining the distribution of SIP diagnosis. Two standard deviations worth of SIP occurred within 0–14 days, thus providing a well-matched temporal overlap between the two cohorts and blanketing the clinical window during which SIP is most prevalent. Univariate analysis Study populations were compared using both univariate and bivariate techniques. Continuous variables (estimated gestational age and birth weight) were evaluated with two-tailed t-tests. Categorical variables (e.g., race and gender) were evaluated with a two-tailed w2 test. Nonparametric continuous data were assessed with a Kruskal–Wallis analysis of variance. Multivariate analysis After univariate analysis, multivariate logistic regression was used to identify factors independently associated with a diagnosis of SIP without NEC (patients with perforation compared to matched controls). The demographic variables found by univariate analysis to be significantly different for the two groups (P<0.1) were used in the logistic regression analysis. Variables were entered into the model with a stepwise selection (P-value for entry and retention <0.1). Multivariate analysis was also used to assess for independence between drugs that were found to be associated with a diagnosis of SIP when given within the same treatment window. Results Creation of the data set Between January 1, 1996 and June 1, 2004, care was provided by clinicians affiliated with Pediatrix to 227 711 neonates (see Electronic Supplement A). Using search techniques, we identified 1779 patients with a report of perforation or NEC treated surgically. Neonates with reports of gastric perforation and major congenital anomalies (n ¼ 255) were excluded. Major anomalies associated with a report of bowel perforation were heart anomalies (n ¼ 61), cystic fibrosis (n ¼ 26), gastroschisis (n ¼ 23), intestinal atresia (n ¼ 23), Hirschsprung disease (n ¼ 13), chromosomal abnormality (n ¼ 14), multiple congenital anomalies (n ¼ 11), galactosemia (n ¼ 7), hydrops fetalis (n ¼ 7), omphalocele (n ¼ 4), thalassemia (n ¼ 4) and diaphragmatic hernia (n ¼ 3). These exclusions left a total of 1524 patients with perforations. There was no significant change in the combined reported incidence of acquired bowel disease (i.e., there was no change in the incidence of the sum of cases of NEC and SIP) over time and there was no significant change in the reported incidence of cases (NEC or perforated bowel without NEC) over time (Electronic Supplement A). However, there was an increase in the use of the term, ‘isolated perforation’, and a decrease in the use of the terms, ‘bowel perforation with no report of NEC’, over time (data not shown). Comparison of SIP to matched control and NEC cohorts (Table 1 and Electronic Supplement B) When compared to matched controls for demographic variables, neonates with SIP were more often outborn, less often received antenatal steroids; more often reported to have had a PDA and be treated for it, and more likely to have received vasopressor support. Infants with SIP also had significantly higher mortality than matched controls. When compared to neonates with surgical NEC, infants with SIP were smaller and more immature, were diagnosed earlier in life (median age ¼ 7 vs 15 days), were more often diagnosed with a PDA requiring treatment and were more likely to have received surfactant vasopressors and assisted ventilation. Infants with NEC also had significantly higher mortality than infants with SIP. When compared to control patients (all variables in Table 1 and Electronic Supplement B evaluated), the factors found in multivariate analysis to be independently associated with SIP were being outborn, a report of a PDA as a diagnosis, a report of the use of pressors within 14 days of birth (controls) or within 14 days of the diagnosis of SIP (Table 2). Analysis of pharmacologic variables between SIP and controls Associations between SIP and individual drug interventions were assessed in four separate treatment windows between the control and SIP cohorts (Table 3A): antenatal, day of life 0–3, day of life 4–7 and day of life 0–14 for control vs 14 days prior to diagnosis for SIP. In the antenatal window, we were only able to derive antenatal steroids by report, but there was a significant reduction in antenatal steroid exposure in the SIP cohort when compared to controls. In the early window (days of life 0–3), indomethacin and hydrocortisone were both found to be associated with SIP by univariate analysis. This association did not hold true for either indomethacin or hydrocortisone when comparing exposure during days of life 4–7. Within the 14-day windows, infants with SIP were Journal of Perinatology Insight into spontaneous intestinal perforation J Attridge et al 96 Table 1 Demographic and discharge data Variable Control (Cont) Perforation (SIP) 26 0.83 0.04 5 8 39 129 198 509 Race, n (%) Black Hispanic White 131 (23.1) 118 (20.8) 298 (52.6) 140 (22.9) 126 (20.6) 307 (50.2) 241 (28.5) 218 (25.8) 340 (40.2) Delivered by C-section, n (%) 351 (60.6) 379 (60.1) 494 (55.9) Discharge type, n (%) Died Home Transfer Transfer of service 93 377 81 28 (16.1) (65.1) (14) (4.8) 137 322 123 49 (21.7) (51) (19.5) (7.8) 316 385 147 43 (35.5) (43.2) (16.5) (4.8) X Y (0–8.5) (49–92.5) (0–49) (2–75.3) 17 92.5 12 78 (10–31.5) (66–117) (4–59) (30–104.5) 27 96 34 105 (15–47.5) (65–123) (13–94) (48–139) X X 2 72 16 16 Table 2 Factors independently associated with SIP (SIP vs matched controls) Variable Odds ratio (CI) P-value Outborn Report of PDA Dopamine or dobutaminea 2.34 (1.7–3.2) 1.52 (1.1–2.0) 1.47 (1.1–1.9) <0.0001 0.0032 0.0041 Early treatment (0–3 days) Indomethacin Hydrocortisone 1.86 (1.4–2.5) 1.77 (1.2–2.8) <0.0001 0.0113 CI ¼ confidence interval; PDA ¼ patent ductus arteriosus. a Reported within first 14 days after birth for control and within 14 days of diagnosis for SIP group. again more often exposed to indomethacin and to hydrocortisone. Finally, the pharmacologic factors found in multivariate analysis to be independently associated with SIP (when compared to control patients) were use of pressors within 14 days prior to the diagnosis Journal of Perinatology 26 0.81 0.02 5 7 40 136 220 473 633 (24–28) (0.67–1.05) (0.68 to 0.64) (3–7) (6–9) (6.4) (23.8) (34.8) (75.1) 27 0.94 0.16 6 8 43 197 346 667 891 (25–30) (0.71–1.35) (0.9 to 0.5) (3–7) (6–9) (4.9) (24.6) (38.8) (74.9) SIP vs Cont Number of patients, n Estimated gestational age (weeks), median (25th–75th percentile) Birth weight (kg), median (25th–75th percentile) Birth weight Z-score, median (25th–75th percentile) Apgar score at 1 min, median (25th–75th percentile) Apgar score at 5 min, median (25th–75th percentile) Apgar score at 5 min p3, n (%) Multiple gestation, n (%) Female, n (%) Inborn, n (%) Age at discharge, median (25–75th percentile) Died Home Transfer Transfer of service 581 (24–28) (0.69–1.06) (0.41 to 0.55) (3–7) (6–9) (6.8) (23.3) (34.1) (87.6) NEC (NEC) SIP vs NEC X X X Y X Y X Y X for SIP (as compared to the 14 days after birth for controls); indomethacin exposure in the first 3 days of life and hydrocortisone exposure in the first 3 days of life. Using univariate analysis, the effect of combination therapy (both indomethacin and any glucocorticoid) was evaluated (Table 3B). When examining early exposure (days 0–3), indomethacin in combination with glucocorticoids was found to be significantly associated with SIP. This association was lost in the 2week window. In multivariate analysis, we were not able to demonstrate an interaction between exposure to both indomethacin and hydrocortisone and the occurrence of SIP. This may have been due to the fact that concurrent use of indomethacin and steroids was rare outside of the early window within the SIP cohort. We also note that the effect of early post-natal dexamethasone on neurodevelopment was widely publicized during the time period these data were collected. In comparison to the previous study using this data set, the use of early post-natal dexamethasone was much lower during this epoch.9 This practice trend limited the ability to test for an association between dexamethasone and SIP, Insight into spontaneous intestinal perforation J Attridge et al 97 Table 3 Perforations associated with (A) Timing of individual drugs and (B) drug combinations Control (Cont) Perforation (prior to Dx; SIP) SIP vs Cont 581 633 Pp0.05 Antenatal drug exposure Antenatal steroids reported, n (%) 373 (65) 368 (58.1) Y Early (1/2 the median time to diagnosis) drug exposure: days of life 0–3 Any report of indomethacin, n (%) Any report of hydrocortisone, n (%) Any report of dexamethasone, n (%) 158 (27.2) 36 (6.2) 13 (2.2) 273 (43.1) 78 (12.3) 23 (3.6) X X 43 (7.4) 9 (1.6) 2 (0.0) 32 (5.1) 12 (1.9) 16 (2.5) Days of life 0–14 222 (38.2) 61 (10.4) 42 (7.2) 14 days prior to SIPa 280 (44.2) 97 (15.3) 63 (10) (A) Timing of individual drugs Number of patients, n Intermediate drug exposure: days of life 4–7 Any report of indomethacin, n (%) Any report of hydrocortisone, n (%) Any report of dexamethasone, n (%) Any drug exposure (87.3% of SIP occurred in the first 15 days) Any report of indomethacin, n (%) Any report of hydrocortisone, n (%) Any report of dexamethasone, n (%) (B): Drug combinations Early drug exposure: days of life 0–3 Indomethacin and steroids both, n (%) Indomethacin alone, n (%) Steroids alone, n (%) Neither, n (%) Any drug exposure Indomethacin and steroids both, n (%) Indomethacin alone, n (%) Steroids alone, n (%) Neither, n (%) 17 141 31 392 (2.9) (24.3) (5.3) (67.5) Day of life 0–14 59 (10.2) 163 (28.1) 34 (5.9) 325 (55.9) 58 215 42 318 X X (9.2) (34) (6.6) (50.2) X X Y 14 days prior to SIPa 93 (14.7) 187 (29.5) 57 (9) 296 (46.8) Y X ¼ Pp0.05 where the trend is greater in the SIP cohort; Y ¼ Pp0.05 where the trend is less in the SIP cohort. a The window of 14 days prior to SIP is based on the date of reported diagnoses and is limited to 14 days or the number of days of life prior to diagnosis if less than 14. but also potentially prevented dexamethasone, which has been a dominant risk factor in the literature, from being an over-riding confounder of other variables. Discussion The study of SIP has been challenging. Much of what we know about risk factors for SIP has been derived either from retrospective single-center studies or prospective randomized trials that were stopped because SIP was an undesirable outcome. In addition, the potential for diagnostic confusion between SIP and NEC has delayed the application of national databases towards studying this disease. We sought to advance the understanding of SIP by creating a large, nationally derived database using the Pediatrix electronic administrative note system. Our first objective was to test for differences in demographics, presentation and mortality between SIP and surgical NEC cohorts. We found that infants with surgical NEC were of older gestation, larger, less likely to be Caucasian and presented with their diagnoses at significantly later days of life than infants with SIP. Conversely, infants with SIP were more likely to have a PDA and to receive treatment for it, more likely to receive surfactant and more likely to require pressors. All of these are expected complications of the more premature infant. Both NEC and SIP patients suffered significantly greater mortality than the control cohort, but surgical NEC patients had significantly greater mortality when compared to SIP patients. When taken together, these findings paint two very Journal of Perinatology Insight into spontaneous intestinal perforation J Attridge et al 98 distinct clinical pictures and confirm the hypothesis that NEC and SIP are different diseases. This is despite the fact that they both occur with similar prevalence in low birth weight infants. Our second objective was to utilize our refined SIP data set to examine drug associations with SIP. The available literature suggests that prophylactic indomethacin exposure is not associated with SIP as an independent variable (reviewed in Figure 1).6 However, when combined with early post-natal glucocorticoid exposure, the available literature paradoxically suggests that indomethacin exposure in the first days of life is associated with SIP.7,8 SIP patients were more likely to have received indomethacin during days 0–3 of life, but there was no such association in the subsequent window from days of life 4–7. Within the 14-day window, indomethacin was found to be significantly associated with SIP; however, this association is difficult to evaluate since 87.3% of the 14-day indomethacin was administered during the early window. Multivariate analysis confirmed that indomethacin was an independent variable when administered within the 0- to 3day window. Hydrocortisone was also found to be associated with SIP during days 0–3 by both univariate and multivariate analysis, as was the presence of a PDA, and the need for pressors within 14 days prior to perforation. Conversely, antenatal steroid use was less prevalent in the group of infants with SIP. We also found an association between outborn birth status and SIP. This is the second multicenter study to demonstrate these associations (outborn status and less likely to have received antenatal steroids).8,15 The neonatal transport environment remains one of the most physiologically stressful situations for ELBW infants. Taken together, these variables suggest a role that endogenous adrenocorticoid stress may be as important as exogenous administration of glucocorticoids. While many of these variables could be postulated to be cotraveling with increased severity of illness and associated with intestinal hypoperfusion, being outborn seems an unlikely associate. Since cortisol levels were not available in this data set, we cannot rule out the possibility that our observed association between indomethacin and SIP is dependent upon elevated endogenous cortisol levels. Indomethacin was also associated with SIP when combined with steroids in the 0- to 3-day window by univariate analysis. The persistent co-traveling of indomethacin, with markers of glucocorticoid exposure or their effect, provides a potential explanation for the uncommon frequency of SIP despite the high rate of indomethacin exposure. In essence, a two-hit phenomenon may be required for the genesis of SIP, thereby explaining why all ELBW infants who receive indomethacin do not perforate. There are limitations to this data set. First, this database is retrospective and cannot be used to directly test causality. Second, this data set is one of reported variables and not of actual laboratory and pharmacy values. The inability to capture antenatal Journal of Perinatology indomethacin exposure is one such manifestation of this limitation. The strength of this study is that it represents a large cohort of SIP cases (greater than all of those currently published added together), and thus is well powered for investigation of multiple variables. In addition, this is the first national data set to demonstrate a cohort of SIP patients distinct from surgical NEC patients (including those with NEC and perforation). In summary, SIP is a distinct clinical disease entity from that of surgical NEC and was diagnosed almost as frequently in preterm infants during the last 3 years in our data set (see Electronic Supplement A). We found that SIP patients were more likely to be exposed to early post-natal indomethacin and had multiple associations that clustered around stress-related and cortisolinfluencing variables. Our findings are consistent with a harmful synergism hypothesis involving both early post-natal indomethacin and glucocorticoids in the etiology of SIP. Acknowledgments Dr Gordon was supported by NIH NIDDK Grant 1KO8DK/HD61553-01. We are grateful to Sheryl Mroz and Sarabeth Gordon, MS for their editorial support and to all the caregivers within Pediatrix for their daily attentiveness to the electronic note system. References 1 Alpan G, Eyal F, Vinograd I, Udassin R, Amir G, Mogle P et al. Localized intestinal perforations after enteral administration of indomethacin in premature infants. J Pediatr 1985; 106: 277–281. 2 Aschner JL, Deluga KS, Metlay LA, Emmens RW, Hendricks-Munoz KD. Spontaneous focal gastrointestinal perforation in very low birth weight infants. J Pediatr 1988; 113: 364–367. 3 Scholz TD, McGuinness GA. Localized intestinal perforation following intravenous indomethacin for patent ductus arteriosus. J Pediatr Gastroenterol Nutr 1988; 7: 773–775. 4 Meyer CL, Payne NR, Roback SA. Spontaneous, isolated intestinal perforations in neonates with birth weight less than 1000 g not associated with necrotizing enterocolitis. J Pediatr Surg 1991; 26: 714–717. 5 Nagaraj HS, Sandhu AS, Cook LN, Buchino JJ, Groff DB. Gastrointestinal perforation following indomethacin therapy in very low birth weight infants. J Pediatr Surg 1981; 16: 1003–1007. 6 Schmidt B, Davis P, Moddemann D, Ohlsson A, Roberts RS, Saigal S et al. Long-term effects of indomethacin prophylaxis in extremely-low-birthweight infants. N Engl J Med 2001; 344: 1966–1972. 7 Stark AR, Carlo WA, Tyson JE, Papile LA, Wright LL, Shankaran S et al. Adverse effects of early dexamethasone in extremely-low-birth-weight infants. National institute of child health and human development neonatal research network. N Engl J Med 2001; 344: 95–101. 8 Watterberg KL, Gerdes JS, Cole CH, Aucott SW, Thilo EH, Mammel MC et al. Prophylaxis of early adrenal insufficiency to prevent bronchopulmonary dysplasia: a multicenter trial. Pediatrics 2004; 114: 1649–1657. Insight into spontaneous intestinal perforation J Attridge et al 99 9 Guthrie SO, Gordon PV, Thomas V, Thorp JA, Peabody J, Clark RH. Necrotizing enterocolitis among neonates in the United States. J Perinatol 2003; 23: 278–285. 10 Fujii AM, Brown E, Mirochnick M, O’Brien S, Kaufman G. Neonatal necrotizing enterocolitis with intestinal perforation in extremely premature infants receiving early indomethacin treatment for patent ductus arteriosus. J Perinatol 2002; 22: 535–540. 11 Garland JS, Alex CP, Pauly TH, Whitehead VL, Brand J, Winston JF et al. A three-day course of dexamethasone therapy to prevent chronic lung disease in ventilated neonates: a randomized trial. Pediatrics 1999; 104: 91–99. 12 Vermont Oxford Network Steroid Study Group. Early postnatal dexamethasone therapy for the prevention of chronic lung disease. Pediatrics 2001; 108: 741–748. 13 14 15 Gordon P, Rutledge J, Sawin R, Thomas S, Woodrum D. Early postnatal dexamethasone increases the risk of focal small bowel perforation in extremely low birth weight infants. J Perinatol 1999; 19: 573–577. Gordon PV, Young ML, Marshall DD. Focal small bowel perforation: an adverse effect of early postnatal dexamethasone therapy in extremely low birth weight infants. J Perinatol 2001; 21: 156–160. Watterberg KL. Weighing statistical certainty against ethical, clinical, and biologic expediency: the contributions of the Watterberg trial tip the scales in the right direction: in reply. Pediatrics 2005; 115: 1447. Supplementary information is available on the Journal of Perinatology website (http://www.nature.com/jp) Journal of Perinatology