Clinical Predictors and Outcomes for Recurrent Pneumatosis Intestinalis in Children: A Case Control Study

Abstract

Introduction:

Pneumatosis intestinalis (PI) is air collections within the wall of the intestine that can be associated with life threatening conditions. Recurrent episodes of PI have been reported; however, little is known about risks for recurrent disease. This study aims to identify predictors of recurrent PI and evaluate long-term outcomes in patients with recurrent PI.

Methods:

This retrospective case-control study evaluated patients, ages 6 months to 18 years, with imaging evidence of PI over a 30-month period. Images were analyzed by consensus to confirm PI. Recurrent PI was defined as new findings of PI after resolution by imaging and/or at least 1-month interval between episodes of PI. Univariate and multivariable analyses were performed using logistic regression, with significance set to P < 0.05.

Results:

Forty-six children were included. Recurrent PI occurred in 17 (37%) patients, with a total of 39 episodes. Predictors of recurrent PI all related to the segment of bowel involved with small bowel PI predictive of recurrent PI(oddsratio [OR] 1.6; 95%confidenceinterval [CI] 1.01, 2.4)and colonic PI protective (OR 0.6; 95% CI 0.4, 0.87) after adjusting for age. Predictors for surgical intervention included cerebral palsy (OR 17; 95% CI 1.7, 167) and PI involving small bowel (OR 19; 95% 3.1, 114).

Conclusion:

Location of PI is predictive of recurrence. Clinical outcomes were similar between groups with single episode and recurrent PI.

Pneumatos is intestinalis (PI) reflects gas within the intestinal wall and is apparent on imaging as round and/or linear air bubbles. PI has a spectrum of clinical presentations from asymptomatic to life-threatening illness (1,2). Pneumatosis can be a hallmark of bowel ischemia (and sometimes infarction) in adults (1,3-5) and in infants as a presentation of necrotizing enterocolitis (6). The pathogenesis of PI is likely multifactorial and proposed etiologies include: hydrogen-producing bacteria (7), mechanical distention secondary to intestinal obstruction or pulmonary barotrauma (8), and abnormal mucosal barrier immune response (1,9). In pediatrics, PI is most commonly seen in the neonatal period secondary to necrotizing enterocolitis (10) and is less frequently observed in older children (11). Recurrent episodes of PI have been reported in case series at a frequency of 15–80% (10,12,13). To our knowledge, there are no studies specifically evaluating predictors of recurrent PI or defining recurrent episodes of PI. It is unknown if recurrent PI can be predicted at first episode, or if patients with recurrent PI develop additional comorbidities, or have worse mortality.

We hypothesized that pediatric patients with recurrent PI would differ from patients with single episode PI in terms of morbidity and mortality, possibly due to repeated injury to the bowel or due to different underlying risk factors. Our aims were to identify factors predictive of recurrent PI and describe outcomes in the patient population with recurrent PI.

METHODS

Study Design

This was a retrospective single-center case–control study of pediatric patients (< 18 years) with PI diagnosed at Cincinnati Children’s Hospital Medical Center. A convenience sample of consecutive patients was identified. Utilizing SoftTek Illuminate software (Illuminate; Overland Park, KS) and a keyword of “pneumatosis” 533 patients were identified for the period May 2016 through December 2018 (Consort Diagram, Figure1, Supplemental Digital Content, http://links.lww.com/MPG/C401). We excluded patients that had no PI confirmed on imaging, patients >18 years old, and patients <6 months to avoid inclusion of patients with necrotizing enterocolitis. For each confirmed episode of PI, images over the prior 10-year period, and remaining study period were evaluated for PI, as available. Clinical follow-up was a minimum of three months after diagnosis of PI. For the purpose of this study, recurrent PI was defined as distinct episodes of PI identified by imaging, separated by at least one month and/or with complete interval resolution of imaging findings. Clinical and demographic information was obtained from the patient electronic medical record (Epic System Corporation; Verona, WI) including patient demographics, clinical course, laboratory results, and outcomes.

Image Review

All relevant imaging examinations were reviewed in consensus by two board-certified fellowship-trained pediatric radiologists (10 years and 6 years post-fellowship experience). Images available varied by patient and included abdominal radiography (XR), abdominal computed tomography (CT), abdominal magnetic resonance imaging (MRI), and abdominal ultrasound (US).

The following imaging characteristics of PI were evaluated: location (small bowel ± colonic vs colonic only), extent of involved bowel (focal [limited to a single abdominal quadrant] vs diffuse [involvement of more than one abdominal quadrant]), presence of bowel obstruction, free peritoneal fluid (if CT, MRI, or US available), presence of bowel wall thickening, altered bowel wall enhancement (if CT or MRI available), presence of extra-luminal air (pneumoperitoneum and/or portal venous gas), and an estimation of the duration of PI based on available follow-up imaging examinations.

Statistical Analysis

Patients were grouped based on the frequency of PI: single episode (control group) versus recurrent PI (case group). Comparisons between groups was performed using chi-squared, Wilcoxon rank-sum, or Fisher exact tests, as appropriate. Univariate and multivariable logistic regression was used to assess univariate associations between covariables and recurrent PI. Odds ratios were adjusted for variables of significance between groups. Continuous results are expressed as medians and interquartile ranges (IQR) (first quartile, third quartile) or counts (percentages) unless otherwise specified. A P-value <0.05 was considered significant for all inference testing. All statistical analyses were performed using R program (version 6.4.2). This study was approved by our Institutional Review Board, and the requirement of informed consent was waived.

RESULTS

Cohort Description

During the study period, PI was confirmed by consensus review in 46 patients with detailed patient characteristics summarized in Table 1 and Table 1, Supplemental Digital Content, http://links.lww.com/MPG/C399. Forty-four (96%) patients had an underlying chronic medical condition with the most common being: any transplant (n = 19), cerebral palsy (n = 14), and non-transplant related immunosuppression (n = 10). Only two patients were previously healthy, and both of those had only a single episode of PI. Clinical follow-up from first episode of PI for all patients was a median of 24 months (IQR 11–29 months).

TABLE 1.

Demographics

Patients	Total 46‡	Single PI 29§	Recurrent PI 17¶	P
Demographics
Male	25 (54%)	17 (59%)	8 (47%)	0.4
Age at first presentation (y)	4.6 (1.6, 10.3)	4.2 (1.9, 6)	8.9 (1.6, 13.4)	0.3
Body mass index: z score	0.68 (−0.33, 1.47)	1.08 (0.36, 1.93)	0.23 (−0.77, 1.3)	0.3
Clinical history
Healthy	2 (4%)	2 (7%)	0	0.5
Cerebral palsy	14 (30%)	7 (24%)	7 (41%)	0.3
Short bowel syndrome	5 (11%)	3 (10%)	2 (12%)	0.7
Congenital heart disease	7 (15%)	5 (17%)	2 (12%)	0.7
Artificial ventilation	10 (21%)	5 (17%)	5 (29%)	0.5
Transplant hematopoietic stem cell	12 (26%)	9 (31%)	3 (18%)	0.4
Transplant whole organ†	7 (15%)	5 (17%)	2 (12%)	0.7
Non-transplant immune suppression	10 (22%)	6 (21%)	4 (24%)	0.9
Per episode of PI	68‡‡	29§	39§§	p
Precedent or concomitant events
Respiratory viral infection	4 (6%)	0	4 (10%)	0.13
Urinary tract infection	8 (12%)	1 (3%)	7 (18%)	0.13
Bacteremia	7 (10%)	2 (7%)	5 (13%)	0.6
Viral gastroenteritis	9 (13%)	6 (21%)	3 (8%)	0.2
Clostridium difficile infection	2 (3%)	1 (3%)	1 (3%)	0.9
Graft versus host disease (GVHD)	3 (4%)	3 (10%)	0	0.07
Current steroid use	31 (46%)	17 (59%)	14 (36%)	0.11
Current antibiotic use	42 (62%)	19 (66%)	23 (59%)	0.2
Radiologic evaluation
Small bowel ± colonic PI	7 (10%)	1 (3%)	6 (15%)	0.2
Colonic only PI	60 (88%)	28 (97%)	32 (82%)	0.13
Extent: focal	30 (44%)	16 (55%)	14 (36%)	0.2
Extent: diffuse	36 (53%)	13 (45%)	23 (59%)	0.4
Extra-luminal air	16 (24%)	4 (14%)	12 (31%)	0.12
Free peritoneal fluid	10 (15%)	2 (7%)	8 (21%)	0.02*
Portal venous gas	6 (9%)	2 (7%)	4 (10%)	0.3
Small bowel obstruction	4 (6%)	2 (7%)	2 (5%)	0.9
Bowel wall edema	17 (25%)	7 (24%)	10 (26%)	0.9
Duration of pneumatosis (days)	7 (4, 12)	8 (3, 11)	7 (4, 12)	0.9
Outcomes
Time to PI recurrence (days)	N/A	N/A	97 (46, 229)	N/A
Follow-up from first episode (mo)	24 (11, 29)	24 (13, 28)	17 (7, 32)
Average NPO (days)	4.5 (0, 8.5)	4 (0, 9.3)	5.5 (0.3, 7)	0.6
Length of stay (days)	22 (5, 51)	15 (6, 27)	24 (5, 55)	0.2
Change in nutrition (no.)	11 (16%)	4 (14%)	7 (18%)	0.7
PN for more than 7 days (no.)	30 (44%)	15 (52%)	15 (38%)	0.3
Antibiotic duration (days)	7 (1, 10)	7 (1, 10)	6.5 (0.25, 10)	0.5

NPO = nil per os; PN = parenteral nutrition. ^†Kidney, liver, or small bowel transplant. Statistical tests performed between single episode and recurrent PI with Wilcoxon rank-sum test, Fisher exact test, and chi-squared test of independence where appropriate with significance denoted P-value <0.05 (*). Data presented as median (interquartile range) or count (%). ^‡Percentages in this column have the denominator of 46. ^§Percentages in this column have the denominator of 29. ^¶Percentages in this column have the denominator of 17. ^‡‡Percentages in this column have the denominator of 68. ^§§Percentages in this column have the denominator of 39.

Recurrent Pneumatosis and Predictors

Recurrent PI occurred in 17 patients (37% of total cohort) accounting for 39 episodes (range: 2–4 episodes). The median time between episodes of PI was 97 days (IQR 45–229 days). PI episodes typically resolved in a median of 7 days (IQR 3–12 days) from diagnosis regardless of the frequency of PI. At the first episode of PI, small bowel PI was associated with increased odds of recurrent PI (OR 1.6; 95% CI 1.01, 2.4) while colonic PI was protective (OR 0.6; 95% CI 0.4, 0.87), after adjusting for age (Table 2). Recurrent episodes of PI affected the same region of the bowel in 13 (76%) of 17 patients (Fig. 1).

TABLE 2.

Univariate analysis for recurrent pneumatosis intestinalis

Variable	OR (95% CI)	P-value	Adjusted OR (95% CI)	P-value
Age at first diagnosis (y)	1.1 (0.98, 1.27)	0.1
Male sex	0.6 (0.2, 2.1)	0.5	0.9 (0.7, 1.2)	0.5
Body mass index: z score	0.9 (0.5, 1.3)	0.5	1 (0.9, 1.1)	0.9
Healthy	0.3 (0.01, 6.9)	0.9	0.7 (0.4, 1.3)	0.3
Cerebral palsy	2.2 (0.6, 8.2)	0.2	1.1 (0.8, 1.5)	0.6
Short bowel syndrome	1.2 (0.14, 7.7)	0.9	1.2 (0.8, 2)	0.4
Congenital heart disease	0.6 (0.08, 3.4)	0.6	1 (0.7, 1.5)	0.9
Artificial ventilation	2 (0.5, 8.6)	0.3	1.1 (0.8, 1.6)	0.5
Transplant hematopoietic Stem cell	0.5 (0.09, 1.9)	0.3	0.9 (0.6, 1.2)	0.4
transplant whole organ (kidney/liver/small bowel)	0.6 (0.08, 3.4)	0.6	0.9 (0.6, 1.3)	0.6
Non-transplant immune suppression	1.2 (0.3, 4.9)	0.8	1 (0.7, 1.5)	0.8
All immune suppression (transplant and non-transplant related)	0.3 (0.09, 1.07)	0.069	0.8 (0.6, 1.02)	0.08
Graft versus host disease (GVHD)	0.2 (0.02, 4.5)	0.3	0.7 (0.4, 1.3)	0.2
Current steroid use	0.3 (0.07, 1.01)	0.06	0.8 (0.6, 1.01)	0.07
Current antibiotic use	0.5 (0.1, 1.6)	0.2	0.8 (0.6, 1.1)	0.2
Any small bowel	8.6 (0.87, 84.9)	0.065	1.57 (1.01, 2.4)	0.047
Both small bowel + colonic PI	14.2 (0.7, 294)	0.086	1.98 (1.2, 3.4)	0.015
Colonic only	0.1 (0.003, 0.7)	0.02	0.59 (0.4, 0.87)	0.01

Univariate analysis of covariables for recurrent PI for each patient at the time of their first episode of PI. Testing was conducted by logistic regression or contingency table with significance denoted P-value <0.05. CI = confidence interval; OR = odds ratio. Adjusted OR were adjusted for age at first diagnosis.

FIGURE 1.

An example of recurrent colonic pneumatosis. (A) Coronal CT image at initial presentation with focal pneumatosis intestinalis (PI) from cecum to ascending colon (arrows). (B) Frontal abdominal radiograph three months later shows complete resolution of PI. (C) Recurrence of PI 4 months after initial presentation, with diffuse colonic PI involving ascending to descending colon (white arrows) with sparing of the sigmoid/rectum (black arrow). Extraluminal gas is also present in the retroperitoneum (pneumoretroperitoneum) and mediastinum (pneumomediastinum) surrounding the heart (arrow heads). CT = computed tomography.

Preceding Events

Nine episodes of PI were in patients diagnosed with viral gastroenteritis, and four had positive polymerase chain reaction testing for adenovirus or norovirus. Bacteremia in six out of seven patients was secondary to gut derived microbiota (including Enterococcus faecalis, Klebsiella oxytoca, Klebsiella pneumonia, and Escherichia coli). Markers of inflammation and ischemia including c-reactive protein, lactate, and platelets were not significantly elevated in patients with either single or recurrent PI with no difference between groups (Table 1, Supplemental Digital Content, http://links.lww.com/MPG/C399).

Overall Management of Pneumatosis

PI was diagnosed as an outpatient or in the emergency department and managed as an outpatient without changes to feeding regimen or prior antibiotic regimen in seven patients (four with single PI and three with recurrent PI). All other episodes occurred just before or during a hospital admission. Inpatient management of PI was not different between single episode and recurrent PI including duration of nil per os (NPO), antibiotic treatment duration, and changes in enteral nutrition (Table 1). Median length of stay for all patients was 22 days (IQR 5–51 days) without any significant difference between single episode and recurrent PI (P = 0.2). Initiation of enteral feeds was determined based on clinical assessment and not solely based on imaging resolution of PI. A change in feeding regimen before time of discharge or next episode of PI occurred in 11 episodes of PI and included: transition to more hydrolyzed formula (n = 7), supplemental parenteral nutrition (PN) (n = 2), or post-pyloric feeds (n = 2).

In both single episode and recurrent PI, >50% of patients were on prophylactic antibiotics before diagnosis of PI (indications included: urinary tract infection prophylaxis or prophylaxis secondary to their immunocompromised state). After PI diagnosis, patients with 12 episodes of PI continued their current or prophylactic antibiotic regimen without change. In 38 patients with 50 episodes of PI, empiric antibiotic therapy was initiated or escalated from prophylactic coverage. The antibiotic regimen was either monotherapy or combination therapy of: Zosyn (n = 28), metronidazole (n = 16), cefepime (n = 15), and meropenam (n = 7). The median duration of antibiotic treatment was 7 days after PI diagnosis (IQR 1, 10) without difference both groups (P = 0.5).

Extra-luminal air or pneumoperitoneum was present in 16 episodes of PI (four single episodes and 12 episodes in eight patients with recurrent PI, P = 0.12). Management of extraluminal air was by conservative measures (n = 11) or surgery (n = 5); one patient died during surgery. Portal venous gas was observed in six patients, three which required surgery and one subsequently died after a respiratory aspiration event. The other three patients with portal venous gas were managed medically, and one patient was completely asymptomatic.

Surgery and Mortality Outcomes

Seven patients underwent eight surgical interventions related to the diagnosis of PI, two patients with a single episode PI and five patients with recurrent PI (P = 0.13) (Table 1). One patient with recurrent PI had two surgeries which first required jejunal resection due to bowel ischemia with ventriculoperitoneal shunt externalization and the second surgery was an exploratory laparotomy with lysis of adhesions. Other patients underwent only one surgical intervention including exploratory laparotomy with lysis of adhesions (n = 3), ileocecectomy for ileocolic intussusception (n = 1), right hemicolectomy for ischemia (n = 1), and cholecystectomy for hydropic gallbladder (n = 1). Available histopathology of resected intestine was significant for regions of ischemic necrosis with neutrophilic infiltrate in four patients.

A univariate analysis demonstrated increased odds of surgical intervention in the setting cerebral palsy (OR 17; 95% CI 1.7, 167), artificial ventilation at the time of diagnosis (OR 12; 95% CI 2.4, 89), and peritonitis upon presentation (OR 77; 95% CI 3.5, 1690) (Table 2, Supplemental Digital Content, http://links.lww.com/MPG/C400). Imaging findings associated with surgery included small bowel ± colonic PI (odds ratio [OR] 19; 95% confidence interval [CI] 3.1, 114) and extra-luminal air/pneumoperitoneum (OR 6.7; 95% CI 1.4, 36.4). The case fatality rate for the study population was 13% (n = 4 single episode PI and n = 2 recurrent PI, P = 0.2). Death occurred on average 214 days after diagnosis of PI, with a single patient dying within 24 hours of diagnosis of their only episode of PI in context of life-threatening volvulus.

DISCUSSION

Previous literature has described recurrent PI in small case series’ of pediatric patients with cerebral palsy (12) or hematopoietic stem cell transplant (13); however, further details of their clinical course and drivers of recurrence was not evaluated. In this case–control study, we show that 37% of children greater than 6 months of age had recurrent PI and the location of PI at the first episode is predictive of recurrent PI. Specifically, small bowel PI was associated with increased odds of recurrence. Healthy children were rarely diagnosed with PI and they did not develop recurrent PI. Chronic medical conditions of cerebral palsy and disorders requiring immunosuppression were present in >90% of patients in our study, though no specific clinical pre-existing condition was associated with recurrent PI. Surgical intervention and mortality in recurrent PI were no different from single episode PI. Overall, this study provides important findings that small bowel PI is recurrent PI is predicted by small bowel PI, occurs frequently and often affects the same region of bowel.

Pneumatosis has been described in a variety of clinical settings with multiple overlapping theories of pathogenesis which can be grouped into microbial dysbiosis, mucosal injury, ischemia, and mechanical trauma. Microbial dysbiosis, the imbalance of beneficial and pathogenic bacteria, can promote immune dysregulation leading to intestinal inflammation (14,15). Antibiotic therapy and dietary intake are among many other contributors that have been shown to affect the intestinal microbiome. In preterm neonates, pathogenic microbial dysbiosis shifts have been found to precede the development of NEC (16). In our study, antibiotic use before diagnosis with PI was common as prophylactic antibiotics were used in 50% of patients and treatment antibiotics for prior infections in seven patients. In a case series of adults diagnosed with colonic PI, patients were administered Bifidobacterium for presumed bacterial dysbiosis with subsequent symptom improvement (8). In our population, 16% of patients had a change in their diet despite resolution of PI (more hydrolyzed formula, route of administration, or PN). In adults, refractory symptoms have required similar changes to diet which improved symptoms and reduced PI recurrence (17,18). Future microbiome analysis in these settings may provide important insight to the contribution of microbial dysbiosis and antibiotic treatment to the development of PI as well as the role for probiotics.

Mucosal injury secondary to microbial dysbiosis or altered immune response (corticosteroid therapy, immunosuppressive medications, or graft versus host disease) may lead to bacterial translocation and bacteremia (19,20). In the setting of immunosuppressive therapy at the time of PI diagnosis, patients were being treated with either chemotherapeutic medications or other immune modulators in 63% (n = 29) of patients. Three patients with single episode PI were diagnosed with active gastrointestinal graft versus host disease, but none had recurrent PI. Corticosteroids were being used at the time of diagnosis in 46% of episodes but did not have higher odds of recurrent PI or surgery unlike previously reported case series (20). We observed six of the seven episodes of bacteremia associated with PI in our study had gastrointestinal bacteria cultured from blood cultures suggesting bacterial translocation may occur at the time of PI. This suggests that a small subset of patients with known mucosal disease or immune suppression may be at risk for bacteremia in the setting of PI.

Mechanical distention of the bowel secondary to pulmonary disease (10), intestinal obstruction (21), or byproducts of microbial fermentation (1,22-24) could lead to dissection of air into the mucosa. This study identified patients with intestinal obstruction and underlying pulmonary disease associated with PI, however there was not any increased odds of recurrent PI. This may represent a static injury that once resolved does not reoccur.

In the setting of pneumatosis, the extent of mesenteric ischemia has been associated with poor outcomes including surgery and death in addition to pneumoperitoneum and/or portal venous gas (3,5,10,21,25,26); however, both pediatric and adult cases of pneumoperitoneum have been managed conservatively when patients are clinically stable suggesting some finding may represent ruptured cystic structures rather than transmural bowel wall perforation (12,27-30). In this study, conservative management was pursued in >50% of cases in the setting of pneumoperitoneum (11/16 episodes) and portal venous gas (3/6 episodes). Both pneumoperitoneum and portal venous gas occur predominantly in patient with recurrent PI; however, three of the eight patients who underwent surgical intervention with findings of portal venous gas underwent bowel resection for ischemic, confirmed by histologic evaluation. These findings suggest that portal venous gas especially in the setting of pneumoperitoneum should be evaluated closely for possible bowel ischemia.

Management of PI can often be conservative consisting of NPO and antibiotic treatment (10,12), however there are no consensus guidelines for management. In our study, resolution of PI was used clinically in many patients as criteria to advance enteral feedings and discontinue antibiotics. While PI can be identified on many imaging modalities, CT has been shown to accurately determine the extent of PI and may be more sensitive for PI (31) in cases of progressive symptoms. As described above, patients presenting with peritonitis and on mechanical ventilation had the highest odds of undergoing urgent surgery, but the presence of small bowel PI also had increased odds of surgery. We propose a potential algorithm for evaluation of PI to assist management decisions (Fig. 2).

FIGURE 2.

Management algorithm for pneumatosis in pediatric patients.

A strength of this study was the analysis of images for the presence and extent of PI by two experienced fellowship-trained pediatric radiologists. While this study’s sample size is small, it did include consecutive patients enrolled from a large dedicated pediatric medical center. There was a lack of structured prospective follow-up as these were clinical patients, although the median follow-up from initial PI diagnosis was 24 months which improves the reliability of the outcomes. Despite these limitations, we feel that this study adds to the literature by demonstrating outcomes and predictors of single episode and recurrent PI, patients that can be difficult to manage due to their many comorbidities.

In conclusion, PI recurred in approximately 37% of our cohort of 46 children with an initial episode of PI. Recurrent PI was associated only with radiologic features of the bowel segment involved, and there was no significant difference in likelihood of surgery or death between patients with single episode or recurrent PI. While it is unclear whether recurrent PI represents isolated static insults or progressive injury, this study demonstrated patients with recurrent PI overall responded well to conservative management. Predictors of surgical intervention included presence of pneumoperitoneum, portal venous gas, and newly described small bowel PI. There was no difference in mortality between groups. Future larger, prospective studies are needed to identify biomarkers that can guide clinical management of PI that correlates with important clinical symptomatology and outcomes.

What Is Known

• Pneumatosis intestinalis (PI) is uncommon outside the neonatal population.

• PI occurs in a diverse pediatric population and can have both benign and worrisome clinical features.

What Is New

• More than 90% of pediatric patients older than 6 months with PI had cerebral palsy or were taking immunosuppressive medications.

• PI involving the small intestine had increased odds of recurrence.

• Outcomes in patients with recurrent PI were not significantly different from those with a single episode of PI in this cohort.