Pneumatosis Intestinalis in the Pediatric Oncology Population: An Eleven Year Retrospective Review at Memorial Sloan Kettering Cancer Center

Abstract

Background:

Pneumatosis intestinalis (PI) is characterized by the presence of intramural gas in the gastrointestinal (GI) tract. The overall aim of this study was to review risk factors and outcome of pediatric oncology patients at our institution who developed PI.

Procedure:

Patients diagnosed with PI between 2007–2018 were identified from ICD-10 coding of radiology reports at Memorial Sloan Kettering Kids, a tertiary pediatric oncology center. Outcomes of interest were (1) resolution and time to resolution of PI, (2) surgical intervention within two weeks of diagnosis of PI or (3) death secondary to PI. To capture the resolution of PI, we defined the “time-to-recovery (TTR),” as the time elapsed between date of PI diagnosis to the date of recovery.

Results:

Forty-two patients were identified. Within 30 days of diagnosis of PI, three patients had surgical intervention for PI (7%) and two patients died (5%) due to non-PI causes. Median time-to-recovery of PI was 4.5 days [95% CI: 3 to 7 days]. In univariable and multivariable analyses, only steroid use in the prior 30 days was significantly associated with a faster time to recovery of PI [HR = 2.27 (95% CI: 1.17–4.41), p=0.02].

Conclusions:

This is the largest case series of patients with PI in the pediatric oncology population, which reveals significantly lower surgical and mortality rates than other published PI series. For the majority of patients, conservative medical management is indicated. A prospective study is warranted to define diagnosis and management guidelines for PI in the pediatric oncology population in a cooperative group setting.

INTRODUCTION

Pneumatosis intestinalis (PI) is the presence of intramural gas in the GI tract. While the exact pathophysiology of PI remains unknown [1], several theories have been proposed - bacterial, mechanical, and pulmonary [2]. The bacterial theory postulates that gas-forming bacteria enter through defects in the mucosa from underlying mesenteric ischemia. The mechanical theory suggests that increased intraluminal pressure from obstruction leads to gas entering into the bowel wall from mucosal damage. Pulmonary pathology can lead to alveolar rupture with subsequent air dissection along vessels, eventually reaching the bowel wall.

Published case series on PI include pediatric stem cell transplant (SCT) patients (N=18) [3], non-neonatal pediatric patients (N=32) [4], and general adult patients (N=40) [5], in addition to a prospective cohort study on necrotizing enterocolitis (NEC), the neonatal form of PI (N=17,159) [6]. Poor outcomes in these studies were associated with preceding ischemia, graft-versus-host-disease (GVHD) colitis, pre-existing sepsis, age ≥ 60 years, emesis, and white blood cell (WBC) >12 K/mm³ [4, 5]. Rates of surgical management for PI and death (with variations in the time endpoints for these definitions) occurred in 14–35% and 8–35% of patients, respectively [3–6]. Death due to PI can be from enterocolitis, viscus perforation, or sepsis from translocation of GI anaerobic organisms [2].

We have gathered the largest case series of pediatric oncology patients with PI to define their etiology, clinical course and outcome. The aim of this study was to review our institutional experience and management of pediatric oncology patients who developed PI to establish factors associated with poor outcome of PI (i.e. surgical intervention or death).

METHODS

Patient Selection

This was a single institution retrospective study from Memorial Sloan Kettering (MSK) Kids, a tertiary pediatric oncology center. Eligibility included patients who had radiological evidence of PI between January 1, 2007 to January 1, 2018 that were identified through both hospital records using the ICD-10 code K63.89 (“other specified diseases of the intestine”) or J98.2 (“interstitial emphysema”) and the terms “pneumatosis” and/or “pneumomediastinum” in a radiology report. Patients were excluded if they were not treated on the MSK Kids Service or had no evidence of PI upon imaging review. Case information was recorded from the electronic medical record. Incidence was estimated based on approximately 400 active patients (receiving chemotherapy, immunotherapy, or surgery) treated at MSK Kids annually. The study was approved by the Institutional Review Board at MSK Cancer Center.

Oncologic and Treatment History Prior to PI Diagnosis

Most recent oncologic diagnosis that received active treatment was recorded. First-line treatment encompasses induction/consolidation, maintenance, or remission for most recent primary diagnosis. All chemotherapy, immunotherapy, radiation, and surgery prior to PI diagnosis were recorded to capture accumulation of prior treatment. Specifically, these factors were tabulated within the 30 days prior to PI diagnosis to attempt to find correlation.

Management and Resolution of PI

Management details that were recorded included dietary and antibiotic changes occurring. Similarly, imaging type and timing were recorded after PI diagnosis, until resolution of PI was noted.

Initially we aimed to capture a complete recovery, as defined with three criteria: absence of PI on imaging (#1), and return to a patient’s baseline diet (#2) and antibiotic regimen (#3). Due to variations in management, this definition of a complete recovery would not accurately capture the recovery of PI in our population. For example, a patient with PI was advanced to a full diet (#3), had his/her antibiotics discontinued (#2), then was discharged from his/her inpatient hospitalization, despite their last imaging modality being an x-ray that demonstrated PI (did not meet #1). However, their next x-ray would not occur until a follow-up outpatient appointment several weeks later. Therefore, a complete recovery (with negative x-ray for PI) after several weeks at a follow-up appointment would not accurately capture the date of that patient’s true recovery.

Therefore, we established the term “time-to-recovery (TTR),” as the time elapsed between date of PI diagnosis to the date of recovery (defined as a minimum of one of three criteria met: absence of PI on imaging, advancement of diet, or de-escalation of antibiotics). All three criteria do not need to be met. De-escalation of antibiotics included removal of one or more antibiotics (while patient remained on other antibiotics), changing from a broad-spectrum antibiotic to a narrower-spectrum antibiotic, and/or changing from intravenous (IV) to per oral antibiotics. When more than one criterion was observed, the date of the first occurring event was used as the date of PI recovery. If two of these outcomes occurred “first” on the same day, they both were recorded as indicative of outcome of PI.

Statistical Analysis

Patient demographics were tabulated using mean and range for continuous variables, and frequencies and percentages for categorical variables. If a patient had a recurrence (more than one distinct episode of PI), then only the first episode was included as part of all statistical analyses. Patients who died before recovery were treated as competing events in the TTR analysis.

Cumulative incidence of PI recovery was estimated using an Aalen-Johansen estimate. Factors prognostic of recovery were assessed in a cause-specific hazard model to account for the competing event. Assessed factors were asymptomatic versus symptomatic diagnosis, antecedent SCT, surgical intervention, presence of a pneumomediastinum, occurrence of acute and chronic GVHD, no changes made in diet, no antibiotics given, GI infection and steroids during the 30 days preceding the PI diagnosis. Variables with a p-value < 0.20 in univariable analysis were introduced to the multivariable model. A variable selection was done to keep variables associated with TTR with a p-value < 0.05.

RESULTS

Patients

Forty-two patients were identified, giving an approximate PI incidence of 1% at MSK Kids in 11 years (Table 1). There was an approximately 1.2:1 male to female ratio. The median age of patients was 8 years (range 1–25). The median time from diagnosis of most recent primary hematologic or oncologic disease to PI diagnosis was 11 months (range 0.4–63.2 months). Five patients had a duration >36 months; their diagnoses were: relapsed neuroblastoma, ALL status post SCT (N=2), relapsed epithelioid sarcoma (3-year off-treatment), and pilomyxoid astrocytoma (no evidence of disease; 5-year off treatment).

TABLE 1.

Patient Characteristics of N=42 Patients with PI. “Diagnosis” and “Disease and Treatment Status at PI Diagnosis” refers to the most recent primary hematologic or oncologic diagnosis that received active treatment. If the primary disease was in remission for >36 months (3 years) at time of diagnosis of PI, the off-treatment status is noted under “Diagnosis.” The term “First-line treatment” encompasses induction/consolidation, maintenance, or remission for most recent primary hematologic or oncologic diagnosis. [ALL=acute lymphocytic leukemia. AML=acute myeloid leukemia. CML = chronic myelogenous leukemia. CNS = central nervous system. DSRCT=desmoplastic small round cell tumor. GI = Gastrointestinal. GVHD = graft versus host disease. PI=pneumatosis intestinalis]

Characteristics	N	Percentage (%)
Age (years)
0–5	10	24
5–10	13	31
>10	19	45
Gender
Male	23	55
Female	19	45
Diagnosis
Hematologic Malignancy	22	52
ALL	14	33
AML	4	10
Primary	2	5
Treatment-related (primary Wilms)	1	2
Treatment-related (primary osteosarcoma)	1	2
MDS	2	5
With GATA2 deficiency	1	2
Treatment-related (primary Ewing)	1	2
CML	1	2
Hodgkin Lymphoma	1	2
Neuroblastoma	8	19
Sarcoma	6	14
DSRCT	2	5
Rhabdomyosarcoma	1	2
Epithelioid (3-year off-treatment relapse, in relapse)	1	2
Osteosarcoma	1	2
Osteosarcoma (15-years after retinoblastoma diagnosis, in remission)	1	2
CNS tumor	4	10
Pilocytic Astrocytoma	1	2
Astrocytoma (5-year off-treatment relapse, in remission)	1	2
Retinoblastoma	1	2
Ependymoma	1	2
Wilms Tumor (5-year off-treatment relapse, in relapse)	1	2
Aplastic anemia	1	2
Disease and Treatment Status at PI Diagnosis
No prior SCT	31	74
Disease status
First-line treatment	21	50
Refractory/relapsed treatment	10	24
History of SCT	11	26
Disease status after SCT
Remission	10	24
Refractory/Relapsed	1	2
GI GvHD
Acute GI GvHD	4	10
Chronic GI GvHD	3	7
No GI GvHD	4	10
Time post BMT
<100 days	3	7
>= 100 days	8	19
Receiving Steroids Within Past 30 Days of Diagnosis of PI
Yes	21	50
For GvHD	10	24
Treatment for ALL or Hodgkins Disease	5	12
For CNS tumor supportive care	2	5
Supportive treatment for ATG or blinatumomab or cytarabine	3	7
For panhypopituitarism	1	2
No	21	50

Presentation of PI

Less than half of the patients [45% (19/42)] had abdominal pain (Table 2). Abdominal tenderness and distension were more common physical findings in 31% (13/42) and 40% (17/42) patients, respectively. There were 10 patients (24%) who were asymptomatic at the time of PI diagnosis (Supporting Information Table S1). Four patients were found to have pneumomediastinum (10%), and six patients were found to have free intraperitoneal air on radiographic findings (14%).

TABLE 2.

Symptoms, Signs, Radiologic, Laboratory, and Infectious Findings of Patients with Pneumatosis Intestinalis (PI) at Diagnosis. For laboratory values, not every patient had a laboratory value at the time of diagnosis, therefore the N for each laboratory value was noted. Infectious Diagnoses were tabulated if they occurred 30 days before or after the PI diagnosis.

Characteristics	N=42	Percentage (%)
Patient Status at time of PI Diagnosis
Outpatient
Remained Outpatient	4	10
Admitted for PI	20	48
Inpatient	18	43
Clinical symptoms
Abdominal pain	19	45
Vomiting	11	26
Diarrhea	13	31
Asymptomatic	10	24
Fever	8	19
Clinical signs
Distension	17	40
Tenderness	13	31
Peritonitis	0	0
Radiologic findings
Pneumoperitoneum	6	14
Ascites	6	14
Pneumomediastinum	4	10
Portal venous gas	3	7
Pneumopericardium	1	2
Laboratory Values	N	Median (Range)
WBC (K/mcL)	41	4.0 (0.0 – 19.9)
ANC (K/mcL)	41	2.9 (0.0 – 18.0)
HCO3- (mEq/L)	38	25 (17 – 33)
Infectious Diagnosis	N	Number of days Prior to PI Diagnosis
Bloodstream	3
*Enterococcus Faecalis	1	10
Enterococcus Faecalis + Methicillin Sensitive  Stapholococcus Aureus + Coagulase negative Stapholococcus	1	Same day
Acinteobacter baumanni + Candida albicans	1	30
Stool	5
Norovirus + Clostridium difficile	1	Norovirus – 1, Clostridium – same day
Clostridium difficile	2	21 (N=1); Same (N=1)
Norovirus	1	Same day (was positive at last check 37 days before, as well)
Cryptosporidium toxic megacolon	1	9
Other	2
Aspiration pneumonia	1	1
Peritoneal fluid growing Enterococcus Faecium and Faecalis	1	Same day

^*Denotes same patient. PI=Pneumatosis Intestinalis.

At presentation of PI, median WBC count was 4.0 K/mcL (range 0–19.9) and median ANC was 2.9 K/mcL (range 0–18.0). One patient had elevated WBC of 19.9. There were 11 patients who were neutropenic (ANC < 1.5 K/mcL) at time of PI diagnosis, eight of whom had severe neutropenia (ANC <0.5 K/mcL). All patients with severe neutropenia had PI recovery, with a median TTR of PI of 7.5 days (range 0–19). Two patients who were severely neutropenic had surgical intervention, and one additional patient with neutropenia had concurrent typhlitis. Median bicarbonate value was 25 mEq/L (range 17–33).

Infections were documented within 30 days prior to the PI diagnosis (Table 2) in 9 patients. Bloodstream infections (N=3) included: Enterococcus Faecalis (N=1, 10 days) and Enterococcus Faecalis + Methicillin Sensitive Staphylococcus Aureus + Coagulase Negative Staphylococcus (N=1, same day), Acinetobacter Baumannii + Candida Albicans (N=1, 30 days). Stool culture results were positive in 5 patients (1 patient had 2 positive cultures) for: both Norovirus (1 day) + Clostridium difficile (same day) (N=1), Norovirus only (same day), Clostridium difficile only (N=2, 21 days and same day), and Cryptosporidium (N=1, 9 days). Other infections included aspiration pneumonia (N=1, 1 day) and peritoneal fluid with Enterococcus Faecalis + Faecium (N=1, same day).

Additionally, one patient was found to have stool culture positive 2 days after PI diagnosis (Vancomycin Resistant Enterococcus Faecium + Klebsiella Pneumoniae Carbapenamase Producer + Escherichia Coli Extended Spectrum Beta Lactamase Producer). Enterococcus was deemed a colonization, as patient (previously noted) was already being treated for Enterococcus Faecalis bacteremia.

Oncologic History Prior to PI Diagnosis

Seven of 42 patients had completed standard of care therapy, including three treatment-related hematologic malignancies, three relapsed primary diseases >3 years off-treatment, and a second solid tumor diagnosis. Ten patients had refractory/relapsed disease (without a SCT) and one patient had a refractory/relapsed disease (with a prior SCT).

Overall, 50% of patients (21/42) received steroids within 30 days prior to their PI diagnosis. Three patients had received prior radiotherapy that included the abdominal field, but none within the prior 30 days. There were low rates of recent (<30 days) tyrosine kinase inhibitors (N=3) and cytarabine (N=2).

Fourteen patients had prior abdominal and/or chest surgeries including a combination of neuroblastoma gross tumor resection (GTR) (N=7), small or large bowel resection (N=2), desmoplastic round small cell tumor (DSRCT) GTR (N=2), gastric (g-tube) or jejunal tube placement (N=2), colostomy placement (N=2), and kidney resection (N=1); one patient had >1 concomitant surgical intervention. These abdominal and/or chest surgeries were performed within a median of 52 days (average 13 months, range 3 days-11.4 years) before PI diagnosis. Five patients had concurrent abdominal pathology, including small bowel obstruction (N=2), pancreatitis (N=1), intussusception (N=1), and recent (<15 days) major abdominal surgery (N=1).

There were 11 patients who underwent a prior SCT. The diagnosis of PI occurred a median time of 6.4 months after SCT (range 62 days-2.8 years). Ten patients had GVHD, seven of whom had GI GVHD. There were four patients with acute GI GVHD, three patients with chronic GI GVHD, and four patients with no GI GVHD. Ten patients were on steroids for GVHD, two patients were on calcineurin inhibitors, and two patients on mycophenolic mofetil at the time of PI diagnosis.

Management at time of PI Diagnosis

At presentation of PI, 18 patients were inpatient and 24 patients were outpatient. Four of the 24 outpatients remained outpatient; they had resolution of PI without complication. The majority of patients were made nil per os (NPO) (24/42, 57%); eleven patients had diet that was reduced to either liquid or elemental. There were seven (17%) patients for whom the diet was not changed (who had uncomplicated clinical courses).

Twenty four of 42 patients were started on IV antibiotics; the median duration of IV antibiotic therapy was 10.5 days (range 7–14 days). Antibiotics that were started /expanded at time of diagnosis included piperacillin-tazobactam, metronidazole, meropenem, imipenem, and clindamycin. Three patients had delays in their oncologic treatment (surgery N=1, chemotherapy N=1, and SCT N=1) due to diagnosis of PI.

Eight children had evidence of pneumoperitoneum; six were treated non-surgically, while two had surgery for PI. A third patient underwent surgery for intussusception. Thus, a total of three patients underwent surgery in the setting of PI.

Two children who underwent surgery during the acute PI episode had evidence of acute systemic decompensation at presentation. Both underwent laparoscopy initially and subsequent laparotomy to resect affected bowel; one patient was diverted with an ileostomy. One was a 3-year old girl with high-risk pre-B cell acute lymphoblastic leukemia (ALL), who developed diarrhea and abdominal pain while neutropenic (Figure 3). She was maintained on modified maintenance chemotherapy with oral methotrexate and 6-mercaptopurine to allow recovery. She had a laparoscopic revision of the ostomy approximately 1 month later, and has been in remission for 11 years.

FIGURE 3.

Time to any form of recovery of the N=40 patients who had a recovery of their PI, regardless of which recovery occurred first. Any (first) resolution N=40, Imaging N=26, Advanced diet N=31, Discontinued Antibiotics N=12, De-escalated Antibiotics N=10. There were 7 extreme variables (>30 days) excluded from the figure for the following modalities: Any resolution (64 days), Imaging (39, 47, 64, 103 and 142 days) and Deescalated AB (1097 days). (AB=antibiotics)

The third child required a right hemicolectomy and ileocolic anastomosis for intussusception 14 days after the diagnosis of PI was made (and after symptoms of PI resolved). She was a 2-year old girl with ALL who was admitted for fever and neutropenia following induction therapy. She developed oliguria, abdominal pain, abdominal distention, and constipation for 3 days, followed by 3 episodes of hematochezia. Abdominal US demonstrated portal venous air, CT revealed PI, and she had stool positive for Clostridium difficile 2 days later. She was alive at the time of analysis.

Outcome of PI

Forty of 42 patients (95%) had resolved PI at the time of analysis; two patients died without resolution of PI (according to last imaging performed), neither death was attributed to the PI. There were no mortalities directly attributable to the diagnosis of PI. At the time of this analysis 38% (16/42) patients have died. Two deaths occurred <30 days after PI diagnosis. One patient was on end-of-life care. Another patient was critically ill with Acinetobacter and Candida sepsis; this patient’s PI resolved prior to his death.

Forty of 42 patients recovered from their PI; Figure 2 demonstrates the cumulative incidence of PI resolution. The TTR ranged from 0 to 64 days, and the median time was 4.5 days [95% CI: 3 to 7 days]. The 10-day and 30-day recovery rates were 84% [95% CI 72–95%] and 93% [95% CI 87–100%], respectively. The most common first criterion for resolution of PI was an advancement in diet (N=29) (Figure 3). Less common first signs of recovery were resolution of PI on imaging (N=9), de-escalation of antibiotics (N=3), and discontinuation of all antibiotics (N=2).

FIGURE 2.

Cumulative Incidence curve of patients with pneumatosis intestinalis (PI) representing time to first sign of recovery (TTR), defined as advancement of diet, resolution of PI on imaging, or de-escalation of antibiotics. N=40 of N=42 patients had a recovery. Patients who died before resolution were treated as competing events. (PI = pneumatosis intestinalis)

The shortest TTR was seen in patients with an advancement to diet (N=31, median TTR=3 days, range 0–12), while the longest recovery was seen via resolution of PI imaging (N=25, median TTR=11 days, range 1–142). The median TTR seen with de-escalation of antibiotics was 7 days (N=9, range 2–17 days) and with discontinuation of antibiotics was 8 days (N=13, range 4–15 days).

In univariable analysis, steroid use in the prior 30 days of PI diagnosis (p=0.02) was correlated with shorter TTR. The following interventions were not associated with a longer TTR: symptomatic cases, SCT, surgical intervention, pneumomediastinum, acute GVHD, chronic GVHD, no change in diet, no antibiotics given, or GI infection in prior 30 days. In multivariable analysis after variable selection, only administration of steroids in the 30 days preceding PI diagnosis was significantly associated with higher chance of recovery of PI (HR = 2.27 [95% CI: 1.17–4.41], p=0.02).

Five patients had recurrent episodes of PI. For the episode to be classified as a second or third episode, there had to be radiological imaging that demonstrated resolution of the prior episode of PI. Their subsequent courses were not included in data tabulation or analysis, but are described in Supporting Information Table S2. Four patients were status post SCT (for GATA2 deficiency, treatment-related MDS, and treatment-related AML) and one patient had metastatic, relapsed disease. Four patients were on steroids. Three of the patients were deceased at time of data analysis; their dates of death were 7–11 months from the first PI diagnosis.

DISCUSSION

In this study, we describe 42 pediatric patients with oncologic disorders who developed PI. We report the estimated incidence of PI of 1% (or about 1 in 100 patients) in the pediatric oncology population. As there is no systemic prospective methods of diagnosis and given that PI was diagnosed incidentally on imaging in about ¼ of our patients, the true incidence may be higher. The most common presenting symptoms and physical findings were abdominal pain, distension, and tenderness, similar to other published reports. In univariable and multivariable analysis, only steroid use in the past 30 days was significantly associated with a faster time to recovery of PI. Our low median TTR of PI of 4.5 days was most commonly seen in advancement in diet. Radiographic resolution of PI was much longer than our other markers of TTR, as imaging lags behind improvement in clinical symptoms. Our low mortality and surgical rates, as compared to other pediatric and adult case series of PI, demonstrate that PI should be studied differently in the pediatric oncologic population.

This study evaluated a number of published factors that were postulated to be associated with PI. We observed that the chance of PI recovery doubled with administration of steroids in 30 days preceding the PI diagnosis. Corticosteroid use is a known therapy associated with the development of PI, postulated to alter the intestinal mucosa [7]. It is possible that physicians managing those patients were more comfortable in advancing diet or de-escalating antibiotics, as steroid use was a previously published association with PI, compared to patients presenting with PI of unknown origin. Given the conflicting findings of our study with those published, a case control study could elucidate if steroids are truly a cause of PI. In contrast with published case reports suggesting causation, we had low rates of recent abdominal surgery, concurrent abdominal pathology, tyrosine kinase inhibitors, or cytarabine, in our population [8] [9].

In our cohort, it appears that anatomic extension of disease or particular radiographic features do not correlate with severity of disease. In evaluating a number of features such as free intraperitoneal air, ascites, mesenteric edema, and portal/mesenteric venous gas, we failed to identify any specific imaging features predictive of more severe disease [10]. Our patients typically were diagnosed with x-ray, however, US (which not many of our patients had) is more sensitive than plain films for portal venous air (a more serious condition). Of our 3 patients with portal venous air, only 1 had surgical intervention. As only 2 of 8 patients with pneumoperitoneum required surgical management, one could argue that free air in PI is not an absolute indication for surgery. This is in stark contrast to the management of infants with NEC (neonatal PI), where pneumoperitoneum is the only definitive indication for surgical management, as it indicates intestinal perforation [11]. None of our N=4 cases of concurrent pneumomediastinum, suggestive of a more systemic disease, required surgery. Our description of simultaneous pneumopericardium and PI has not been previously described in the literature and does not represent a clear anatomic relationship between the two sites. Finally, since radiographic recovery lags significantly behind clinical resolution of symptoms, it is unnecessary to expose patients to repeated x-rays to check for resolution of PI. In concert with our above management recommendations, management of PI should probably be symptom-based, not imaging-based.

This study demonstrated a subgroup of patients who had PI detected incidentally on imaging. Five of these 10 patients had abdominal primary tumors, so they likely underwent more abdominal imaging than non-abdominal primary diseases. The PI resolved without surgical intervention in all 10 asymptomatic cases, though one patient had omission of chemotherapy course and another had delay of SCT. We did note that certain patients with PI should be more carefully managed. Of our five patients who had more than one episode of PI, three died within 7–11 months of their first PI diagnosis. The biologic link between relapsed, more serious oncologic primary course and PI is unclear, but our cohort suggests that multiple episodes of PI is associated with poor prognosis. Additionally, 2 of 3 patients with PI who required surgical intervention were also concurrently neutropenic. The small amount of any new positive microbial studies (stool or blood) in our population at diagnosis of PI, despite 18% of patients presenting with fever, may suggest that antibiotics would only be necessary in certain subgroups such as patients who are febrile, neutropenic, or perforated. However, it is important to note that many clinical notes described the two neutropenic patients who underwent surgery as appearing systemically unwell, thus, clinical presentation of an ill-appearing child is significant in the setting of PI. Overall, patients with relapsed disease, who are post-SCT or who are neutropenic may represent three cohorts of pediatric oncology patients with PI who should be monitored with more conservative management.

The management of pediatric PI at MSK Kids has traditionally been centered around an inpatient admission consisting of intestinal rest, IV broad-spectrum antibiotics and cautious recommencement of enteral intake. Patients who were admitted were kept NPO and commenced on broad-spectrum antibiotics with anaerobic coverage (usually piperacillin/tazobactam). Three patients received high-flow oxygen as treatment for PI though this treatment is no longer utilized by our physicians [12, 13]. At discharge, 13 of 42 patients received oral antibiotic prescriptions as an outpatient. Repeat x-rays were typically taken at every outpatient clinic appointment (at various timepoints) to search for resolution of PI, even in asymptomatic patients.

In the absence of severe symptoms at presentation, we have shown that many patients with PI can be managed conservatively, perhaps a select group on an outpatient basis. In general, it would be vital to avoid unnecessary fasting and prolonged antibiotic courses in a patient population who are nutritionally deplete and are at risk of Clostridium Difficile infection, respectively. Steroids were the only factor significantly associated with a faster recovery of PI, and pediatric oncology patients are typically on steroids to treat conditions associated with high mortality. Given that PI in our population had low surgical and mortality rates, future studies should be performed to confirm whether any oncological-directed therapy would need to be decreased or discontinued solely due to the presence of PI. We were unable to tally the classic endpoints of complete and partial remission, as there is such variation within the ways that physicians manage PI, instead relying on our surrogate endpoint of TTR. Overall, a prospective study is warranted to define diagnosis and management guidelines for PI in the pediatric oncology population in a cooperative group setting.

Supplementary Material

Supinfo

TABLE S1. Clinical Status and Course of Patients who were Asymptomatic at Time of Pneumatosis Intestinalis Diagnosis (N=10). AML= acute myeloid leukemia, ATG =anti-thymocyte globulin, CML=chronic myelogenous leukemia, CT=computed tomography, DSRCT=desmoplastic small round cell tumor, NPO=nil-per-os, PCR=polymerase chain reaction, sAML=secondary acute myeloid leukemia, SCT=stem cell transplant.

TABLE S2. Clinical Status and Course of Patients who Had Recurrent Episodes of Pneumatosis Intestinalis (N=5). MDS = myelodysplastic syndrome, NK=natural killer, PI=pneumatosis Intestinalis, SCT=stem cell transplant

FIGURE 1.

A 3-year-old girl with high risk pre-B cell acute lymphoblastic leukemia. A. Extensive pneumatosis is shown on KUB (arrows). B. Abdominal CT scan confirmed severe pneumatosis of entire right colon (arrows). C. CT abdomen demonstrated intraperitoneal free air (arrows). D. Cross-table lateral x-ray demonstrates small volume free air (arrows). E. CT abdomen with air seen in the portal vein on cross-sectional views (arrows). F. Area of colon with pneumatosis appears distended. G. Air in the bowel wall (red arrow) and omentum (black arrow) over the transverse colon (black arrow). H. Transverse colon shows pneumatosis (red arrow) next to healthy bowel (black arrow).

ABBREVIATIONS KEY

AB

Antibiotics

ALL

Acute Lymphocytic Leukemia

AML

Acute Myeloid Leukemia

ANC

Absolute Neutrophil Count

ATG

Anti-Thymocyte Globulin

CML

mChronic Myelogenous Leukemia

CNS

Central Nervous System

CT

Computed Tomography

GI

Gastrointestinal

GTR

Gross Total Resection

GVHD

Graft-Versus-Host Disease

IV

Intravenous

MDS

Myelodysplastic Syndrome

MSK

Memorial Sloan Kettering

NEC

Necrotizing Enterocolitis

NPO

Nil-Per-Os

PI

Pneumatosis Intestinalis

SCT

Stem Cell Transplant

TKI

Tyrosine Kinase Inhibitors

TTR

Time-To-Recovery

US

Ultrasound

WBC

White Blood Cell

DATA SHARING STATEMENT:

The data that support the findings of this study are available from the corresponding author upon reasonable request.