PEDIATRICS Vol. 108 No. 2 August 2001, pp. 402-406

Pneumatosis Intestinalis in Non-neonatal Pediatric Patients

Amethyst C. Kurbegov, MD, MPH and Judith M. Sondheimer, MD

From the Department of Pediatrics and Section of Pediatric Gastroenterology, Hepatology and Nutrition, University of Colorado Health Sciences Center, Children's Hospital of Denver, Denver, Colorado.

	ABSTRACT

Top Abstract Methods Results Discussion Conclusion References

Objectives.  To describe fully pneumatosis intestinalis (PI) in non-neonatal pediatric patients and to characterize those patients with higher risk of poor outcome, including need for surgery and death.

Methods.  A retrospective chart review was conducted of all patients 30 days of age and older with PI in a tertiary care children's hospital during an 8-year period. Underlying medical condition, presenting signs and symptoms, radiologic grade of pneumatosis, and events that immediately preceded the onset of PI were reviewed, and their correlation with outcome was assessed.

Results.  Thirty-seven episodes of PI occurred in 32 patients. Seventy-eight percent of patients were male, and the median age was 29 months. Major patient diagnostic groups identified with PI included healthy children (22%), patients with organ and bone marrow transplant (22%), patients with decompensated congenital heart disease (12.5%), motility disorders (12.5%), gastroschisis (9%), and short bowel syndrome (6%). The most common events that immediately preceded the onset of PI were noninfectious colitis (32%), acute enteric infection or toxin (27%), bowel ischemia (20%), and gastrointestinal dysmotility (17%). Resolution of PI with medical management occurred in 78% of episodes (good outcome). Twenty-two percent of episodes resulted in a poor outcome: patient death (8%) or surgery (14%). The presence of portal venous gas and low mean serum bicarbonate concentration were the only clinical factors that correlated significantly with poor outcome. Only 25% of patients with pneumoperitoneum required surgery. Poor outcome was seen most commonly in 2 patient diagnostic groups: transplant patients (43% of patients) and decompensated cardiac disease (50% of patients). The event that preceded PI also had an impact on outcome. PI preceded by ischemia or graft versus host disease colitis was associated with poor outcome in 50% and 75% of cases, respectively.

Conclusions.  PI is a radiologic sign that occurs in a variety of settings in non-neonates. PI preceded by bowel ischemia or graft versus host disease colitis has the worst prognosis, and the presence of portal venous gas and acidosis correlate with poor outcome. Not all patients with pneumoperitoneum require surgical intervention. Overall, outcome of PI in non-neonatal patients was better than that reported in neonates with necrotizing enterocolitis.pneumatosis intestinalis, necrotizing enterocolitis, non-neonatal.

Pneumatosis intestinalis (PI) is a radiologic finding that, in pediatrics, is associated most commonly with necrotizing enterocolitis (NEC) in neonates. Because many pediatricians' experience with PI derives entirely from neonates, PI frequently is viewed as an ominous finding, requiring immediate surgical attention. Isolated reports and short series have suggested, however, that PI may be a more benign finding outside the neonatal period.

Previous reports identified infection, anatomic abnormalities, inflammatory bowel disease, intestinal motility disorders, congenital heart disease (CHD), chronic lung disease, and gastrointestinal allergy as medical conditions associated with a higher incidence of PI. Children with immunosuppression as a result of chemotherapy, organ transplantation, AIDS, and congenital causes also have been reported to be at higher risk of PI.^1-18 The relative importance of these conditions to PI in non-neonates is not clear because most previous reports are small and limited to a specific group of patients who share an underlying diagnosis. In this study, we reviewed all cases of PI in non-neonatal (30 days of age) pediatric patients who were treated during an 8-year period at a pediatric tertiary care hospital. The patients therefore represented the entire spectrum of pediatric conditions from those encountered in the emergency department and general pediatric clinic to those in the subspecialty surgical and medical services.

	METHODS

Top Abstract Methods Results Discussion Conclusion References

A retrospective chart review of all patients with PI at the Children's Hospital of Denver between January 1992 and December 1999 was completed. Patients were identified through admitting and discharge diagnoses and/or radiographic reports. All discharge diagnoses and radiologic diagnoses have been computerized at the Children's Hospital since 1992.

The following patient data were obtained: age, gender, underlying medical condition as documented in patient medical records, symptoms and signs at time of diagnosis (fever, vomiting, diarrhea, abdominal pain, distention, gastrointestinal bleeding, peritoneal signs, absolute neutrophil count, and serum bicarbonate concentration), grade of pneumatosis, and characteristics of stool. Treatment modality, including bowel rest, antibiotics, and surgical intervention, was documented. Because the underlying medical condition of each patient was not necessarily the precipitating cause of PI, we identified, as specifically as possible, the immediate preceding event or events associated with each episode of PI that seemed to be the mechanism that produced the intramural gas. Our list of potential precipitating events was obtained from a recent review of PI in adults, which identified 5 mechanisms by which gas dissects into the intestinal wall.¹⁹ In our patients, these included acute enteric infection, bowel wall ischemia, noninfectious colitis, intestinal dilation with gas, and gas dissection from free air in the chest. There were several entities grouped as noninfectious colitis. They included colitis associated with graft versus host disease (GVHD), short bowel colitis, allergic colitis and nonspecific colitis. GVHD colitis is characterized by diarrhea clinically and biopsy findings of apoptosis in the lamina propria and crypt disappearance. Short bowel colitis is a bloody diarrhea seen in patients with >50% small intestinal resection, usually occurring when enteral feedings are advanced beyond the capacity of the shortened intestine.²⁰ Allergic colitis occurs in young infants with sensitivity to dietary protein, most often milk, and can be diagnosed by rectal biopsy or by response withdrawal of the suspect protein.²¹ We also documented the number of days to radiologic resolution of PI, the incidence of recurrence, the elapsed days between repeat episodes, and patient survival of each episode.

Authoritative pediatric radiology texts as well as several review articles describe the patterns of pneumatosis in bowel wall and portal venous system, but there is no standard system for grading PI on abdominal plain film or CT scan.^1,17,22 Therefore, we designed a grading system for this study based on the supine abdominal radiograph. All radiographs were reviewed by pediatric radiologists as well as the authors. We identified 4 grades of PI. In grade I, pneumatosis was observed in 1 or 2 abdominal quadrants. In grade II, more than 2 quadrants with PI were present. In grade III, free abdominal air was noted in the presence of either grade I or II pneumatosis. The presence of portal venous gas was considered to be grade IV pneumatosis.

We defined an outcome for each episode of PI. We defined a "good" outcome as resolution of gastrointestinal symptoms and radiologic findings of PI with medical management alone. Patients with "poor" outcome were those who required surgical intervention or those who died with pneumatosis present. We compared the presenting features, grade of pneumatosis, and clinical management in these 2 broadly defined groups. Generalized Estimating Equations method was used to compare good and poor outcome for statistically significant differences. P  .05 was considered significant.

	RESULTS

Top Abstract Methods Results Discussion Conclusion References

Underlying Medical Condition

Thirty-seven episodes of PI were identified in 32 patients age 30 days and older from 1992 to 1999. Five patients had 2 episodes of PI. Twenty-five of the patients were male (78%). The mean age of the patients was 69 months (range: 1-240 months). The median age of the group was 29 months. The underlying medical conditions of the 32 patients are shown in Table 1.

Preceding Events Associated With PI

There were a total of 41 preceding events for the 37 episodes of PI. Two equally likely events were identified in 4 of the episodes. In some patients, PI occurred during an acute worsening of the underlying medical diagnosis. This was especially true for patients with CHD, tissue transplantation, and dysmotility.

Thirteen of the 37 episodes of PI were preceded by noninfectious colitis, making it the most common preceding event. Four of 13 were GVHD colitis, 4 were short bowel colitis, 3 were allergic colitis, and 2 were a nonspecific colitis documented by biopsy after resolution of the PI episode. Stool examinations in these patients for viruses, pathogenic bacteria, parasites, and C. difficile toxin were negative with the exception of 1 GVHD patient with both colitis and rotavirus infection. There were no cases of Crohn disease or ulcerative colitis in this series.

The next most common preceding event was infection, which occurred in 11 of the 37 episodes. These included 4 episodes with C. difficile toxin in the stool, 3 with rotavirus, 3 with enteric adenovirus, and 1 with both rotavirus and C. difficile toxin. Eight of the episodes of PI were preceded by intestinal ischemia, 5 of which occurred in patients with decompensated CHD. Seven of the 37 episodes of PI were associated with gastrointestinal obstruction and distension, resulting from mechanical obstruction in 2 and pseudoobstruction in 5. One episode of PI was preceded by a pneumothorax, which occurred after dilation of an esophageal stricture in a child with Down syndrome. Finally, in 1 patient who presented to the emergency department with abdominal pain and distention, we were unable to identify any preceding event. Constipation was suspected by radiographs in this previously healthy child but was not confirmed by history.

Of the 4 episodes of PI with 2 suspected preceding events, 3 involved infection: 2 episodes with both inflammatory colitis and rotavirus and 1 episode with C. difficile and obstruction. The fourth episode with 2 preceding events was associated with noninfectious colitis and ischemia.

In the 5 patients with 2 episodes of PI, the preceding events usually were the same for both episodes. Short bowel colitis preceded both PI episodes in the 2 patients with short bowel syndrome. Ischemia preceded both episodes in the patient with decompensated heart disease. Massive intestinal distension preceded both episodes in the patient with pseudo-obstruction. Finally, in 1 patient with a bone marrow transplant, adenovirus infection occurred with the first episode and GVHD colitis with the second. Recurrences occurred 1 to 38 days after resolution of the first episode.

In Table 2 are shown the events that preceded PI in the 4 largest patient diagnostic groups: healthy children, transplant patients, infants with decompensated CHD, and patients with chronic gastrointestinal dysmotility. Except for the healthy children, most episodes were related to the patients' underlying conditions. In the 7 healthy children, 8 preceding events were identified, the most common being infection. No healthy child had more than 1 episode. In the 7 transplant patients, there were 8 PI episodes with 9 preceding events; 1 patient had 2 episodes. Noninfectious colitis secondary to GVHD (4) and infection (3) were the most common preceding events in transplant patients. There were 5 PI episodes in the 4 patients with CHD, all preceded by bowel ischemia secondary to low cardiac output. In the 4 patients with gastrointestinal dysmotility, there were 5 episodes of PI, all associated with acute exacerbations of the primary diagnosis and characterized by worsening of intestinal dilation.

Radiologic Grading of PI

By radiologic examination, 6 episodes of PI were grade I, 19 were grade II, 8 were grade III, and 4 were grade IV. One patient with both free air and portal air was classified as grade IV.

Surgery was required only in 2 of the 8 patients with pneumoperitoneum. These 2 patients had obvious intestinal perforation at the time of laparotomy, 1 from mechanical obstruction and 1 from severe GVHD colitis. The remaining 6 episodes were treated with bowel rest and/or antibiotics. The pneumatosis in the 6 who recovered with medical management alone was associated with functional obstruction (3), adenovirus infection (1), and short bowel colitis (2). Surgical consultation was obtained in all cases of pneumoperitoneum, and the surgeons' and gastroenterologists' decision not to operate in all cases took into account the absence of peritoneal signs or abdominal tenderness on physical examination.

Outcome was not significantly related to PI grades I to III, only to grade IV. All patients with grade IV PI had necrotic bowel on pathology, and all either required surgery or died before surgical intervention.

Therapy and Outcome

Twenty-nine of the 37 episodes of PI were treated medically (78%). Twenty episodes were treated with bowel rest and systemic antibiotics, 5 with bowel rest alone, 3 with antibiotics only, and 1 with no therapy. The untreated episode occurred in a previously healthy, afebrile, 14-year-old boy who had abdominal pain and distension and whose pneumatosis resolved in 7 days. Pneumatosis grades in patients with good outcome were 5 grade I, 18 grade II, 6 grade III, and 0 grade IV. Five patients had surgery, all of whom recovered. Pneumatosis grades in these patients were 1 grade I, 0 grade II, 2 grade III, and 2 grade IV. Three patients died with PI present, none of whom underwent surgery. The underlying diagnosis in 2 was decompensated CHD and in 1 was bone marrow transplant. All of these patients had necrotic bowel at autopsy, and in all, ischemia preceded PI. In these, the onset of PI was part of a preterminal multiorgan system failure. Surgery was not recommended in these 3 patients because of their poor overall prognosis. Two of the 3 patients who died had grade IV pneumatosis, and 1 was grade II.

Correlates of Outcome

There were 29 episodes of PI with good outcome and 8 with poor outcome (Table 3). Gender, fever, vomiting, diarrhea, blood in stool, abdominal pain, abdominal distention, and absolute neutrophil count were not significantly different between the 2 outcome groups. The mean serum bicarbonate concentration was significantly lower in the patients with poor outcome than in the patients with good outcome (P < .02). Low serum bicarbonate may have been caused either by necrotic bowel tissue or by other conditions associated with a sicker patient. A retrospective analysis did not allow for a distinction between these 2 issues.

Grade IV pneumatosis was found in 50% of the episodes with poor outcome and in none of the episodes with good outcome (P < .03). Of the 8 patients with poor outcome, 4 were precipitated by ischemia, 3 by GVHD colitis, and 1 by C. difficile and obstruction. The highest prevalence of episodes with poor outcome was found in 2 diagnostic groups: transplant patients (43%) and decompensated cardiac disease patients (50%).

	DISCUSSION

Top Abstract Methods Results Discussion Conclusion References

Most studies of PI in pediatric patients have focused on its incidence in specific diagnostic groups of patients. In this study, we reviewed an entire children's hospital population during an extended period to obtain a more realistic spectrum of the non-neonatal patients with PI. The male predominance (78%) in our 32 patients was unexpected and does not seem to be a consistent finding in smaller published series of non-neonates with PI or neonates with NEC. Although not statistically significant, an even higher proportion of our patients with poor outcome were male (87%). Our patients ranged in age from 30 days to 20 years, and there was no significant association of outcome with age.

The majority of PI episodes (78%) resolved with conservative management. The mortality rate was only 8%. In contrast, one major pediatric text reports a mortality of 15% in neonates with NEC in tertiary care nurseries.²³ Among our non-neonatal patients, those with transplants and those with an underlying diagnosis of decompensated CHD were particularly likely to have a poor outcome. In this combined group of 11 patients, the mortality was 27%, and 45% of episodes were classified as having poor outcome. Healthy children with acute infectious diarrhea as the preceding event, although symptomatic with abdominal distension, pain, and diarrhea, uniformly had good outcome.

Although experience with neonates with NEC suggests that all patients with free air in the abdomen and PI require surgery for intestinal perforation, only 2 of our 8 patients with free air (in the absence of portal venous gas) were considered ill enough to require surgery. The others, based on benign clinical assessment and known underlying medical condition, were treated with bowel rest and antibiotics and did well. There are several reports in the surgical literature that recognize "benign pneumoperitoneum" as a consequence of PI in non-neonatal patients. These reports suggest that the free air in the abdomen in these cases probably represents the dissection of sterile intramural gas through the serosal surface of the intestine without through and through bowel wall perforation.^24-27 In our series, the features that distinguished patients with free air who required surgery from those who did not were evidence of mechanical obstruction, high-risk underlying condition (CHD or transplant), and portal venous gas. Portal venous gas, with or without free peritoneal air, was a universally ominous sign in our patients. All 4 episodes of PI with portal venous gas had necrotic bowel on pathologic review and a poor study outcome, including 2 deaths.

Peritoneal signs were found only in episodes with poor outcome, but this association did not reach statistical significance in our study probably because of the small number of patients with this finding. It may be of significance that several of the patients with poor outcome were heavily sedated or paralyzed at the time that PI developed to facilitate mechanical ventilation, and thus peritoneal signs could not be appreciated. We did note, however, that alert patients without peritoneal signs were highly unlikely to have poor outcome, despite the presence of free peritoneal air.

	CONCLUSION

Top Abstract Methods Results Discussion Conclusion References

This study indicates that there are few absolute indications for surgery in the non-neonate with PI and that experienced clinical judgment remains an important factor in directing management. Our results suggest that portal venous gas and acidosis are significantly associated with a need for surgery. The number of abdominal quadrants that showed pneumatosis did not correlate with outcome. Patients who were most likely to have a poor outcome were those with the underlying diagnoses of CHD or tissue transplant. The preceding events most commonly associated with a poor outcome were GVHD colitis and ischemia. It may be significant that these preceding events very closely resemble the situation in the neonate with NEC, ie, they produce a deep inflammatory and ischemic bowel wall injury that permits air dissection and a high risk of perforation.

	FOOTNOTES

Received for publication Sep 13, 2000; accepted Feb 19, 2001.

Reprint requests to (J.M.S.) Children's Hospital of Denver, 1056 E 19th Ave, Denver, CO 80218. E-mail: sondheimer.judith{at}tchden.org

	ABBREVIATIONS

PI, pneumatosis intestinalis; NEC, necrotizing enterocolitis; CHD, congenital heart disease; GVHD, graft versus host disease.

	REFERENCES

Top Abstract Methods Results Discussion Conclusion References

1. Knechtle SJ, Davidoff AM, Rice RP Pneumatosis intestinalis: surgical management and clinical outcome. Ann Surg 1990; 212:160-165 [Medline]
2. Boerner RM, Ried DB, Warshauer DM, Isaacs K Pneumatosis intestinalis: two case reports and a retrospective review of the literature from 1985 to 1995. Dig Dis Sci 1996; 41:2272-2285 [CrossRef][Medline]
3. Reynolds JL, Gauderer MWL, Hrabovsky EE, Shurin SB Pneumatosis cystoides intestinalis in children beyond the first year of life: manifestations and management. J Pediatr Surg 1991; 26:1376-1380 [Medline]
4. West KW, Rescorla FJ, Grosfeld JL, Vane DW Pneumatosis intestinalis in children beyond the neonatal period. J Pediatr Surg 1989; 24:818-822 [CrossRef][Medline]
5. Oldham KT, Coran AG, Drongowski RA, Baker PJ, Wesley JR, Polley TZ The development of necrotizing enterocolitis following repair of gastroschisis: a surprisingly high incidence. J Pediatr Surg 1988; 23:945-949 [CrossRef][Medline]
6. John A, Kickey K, Fenwick J, Sussman B, Beeken W Pneumatosis intestinalis in patients with Crohn's disease. Dig Dis Sci 1992; 27:813-817 [CrossRef]
7. Luks FI, Chung MA, Brandt JL, Pneumatosis and pneumoperitoneum in chronic idiopathic intestinal pseudoobstruction. J Pediatr Surg 1991; 26:1384-1386 [CrossRef][Medline]
8. Aziz EM Neonatal pneumatosis intestinalis associated with milk intolerance. Am J Dis Child 1973; 125:560-562 [Abstract/Free Full Text]
9. Jaffe N, Carlson DH, Vawter GF Pneumatosis cystoides intestinalis in acute leukemia. Cancer 1972; 30:239-243 [CrossRef][Medline]
10. Silliman CC, Haase GM, Strain JD, Indications for surgical intervention for gastrointestinal emergencies in children receiving chemotherapy. Cancer 1994; 74:203-216 [CrossRef][Medline]
11. Yeager AM, Kanof ME, Kramer SS, Pneumatosis intestinalis in children after allogeneic bone marrow transplantation. Pediatr Radiol 1987; 17:18-22 [CrossRef][Medline]
12. Burress GC, Ben-Ami T, Whitington PF Pneumatosis in infants after orthotopic liver transplant. J Pediatr Gastroenterol Nutr 1996; 23:577-582 [CrossRef][Medline]
13. Cunnion KJ Pneumatosis intestinalis in pediatric acquired immunodeficiency syndrome. Pediatr Infect Dis J 1998; 17:355-356 [CrossRef][Medline]
14. Gelman SF, Brandt LJ Pneumatosis intestinalis and AIDS: a case report and review of the literature. Am J Gastroenterol 1998; 93:646-650 [CrossRef][Medline]
15. Tang MLK, Williams LW Pneumatosis intestinale in children with primary combined immunodeficiency. J Pediatr 1998; 132:546-549 [CrossRef][Medline]
16. Capitanio MA, Breenberg SB Pneumatosis intestinalis in two infants with rotavirus gastroenteritis. Pediatr Radiol 1991; 21:361-362 [CrossRef][Medline]
17. Silverman FN, Kuhn JP. Caffey's Pediatric X-Ray Diagnosis: An Integrated Imaging Approach. 9th ed. St Louis, MO: Mosby; 1995
18. Fenton LZ, Buonomo C Benign pneumatosis in children . Pediatr Radiol 2000; 30:786-793 [CrossRef][Medline]
19. Pear BL Pneumatosis intestinalis: a review. Radiology 1998; 207:13-19 [Abstract/Free Full Text]
20. Taylor SF, Sondheimer JM, Sokol RJ, Silverman A, Wilson H Noninfectious colitis associated with short bowel syndrome . J Pediatr 1991; 119:24-30 [CrossRef][Medline]
21. Roy CC, Silverman AS, Alagille D. Pediatric Clinical Gastroenterology. 4th ed. St Louis, MO: Mosby; 1995
22. Heng Y, Schuffler MD, Haggitt RC, Rohrmann CA Pneumatosis intestinalis: a review. AJG 1995; 90:1747-1758
23. Wigglesworth JS, Singer DB. Textbook of Fetal and Perinatal Pathology. 2nd ed. Malden, MA: Blackwell; 1998
24. D'Agostino S, Fabbro MA, Musi L, Bozzola L Pneumatosis cystoides intestinalis: a rare cause of nonsurgical pneumoperitoneum in an infant. J Pediatr Surg 2000; 35:1106-1108 [CrossRef][Medline]
25. Williams NMA, Watkin DFL Spontaneous pneumoperitoneum and other nonsurgical causes of intraperitoneal free gas. Postgrad Med J 1997; 73:531-537 [Abstract/Free Full Text]
26. Hoover EL, Cole GD, Mitchell LS, Adams CZ, Hassett J Avoiding laparotomy in nonsurgical pneumoperitoneum. Am J Surg 1992; 164:99-103 [CrossRef][Medline]
27. Khouri MR, Levine MS, Dabezies M, Saul SH Benign pneumoperitoneum in a patient with celiac sprue. J Clin Gastroenterol 1989; 11:70-72 [Medline]