Characteristics, treatment, and outcome of patients with bowel perforation after immune checkpoint inhibitor exposure

Exposure to immune checkpoint inhibitors (ICIs) can predispose to immune-related adverse events (irAEs) involving the gastrointestinal tract. The association between ICIs and bowel perforation has not been well studied. We aimed to describe the clinical course, complications, treatment, and outcomes of patients experiencing bowel perforation during or after ICI treatment. This retrospective, single-center study included adult cancer patients with bowel perforation that occurred between the first dose of ICI treatment and up to 1 year thereafter between 1/1/2010 and 4/30/2021. Patients’ clinical course, imaging, treatment, and outcomes related to bowel perforation were collected and analyzed. Of the 13,991 patients who received ICIs during the study period, 90 (0.6%) met the inclusion criteria. A majority were male (54.4%), the most common cancer type was melanoma (23.3%), and most patients had received PD-1/L1 inhibitor treatment (58.8%). Onset of perforation occurred after a median of four ICI treatment cycles. The most common symptom was abdominal pain (95.5%). The colon was the most common location for the perforation (37.7%). Evidence of diverticulitis, enterocolitis, or appendicitis was seen in 32 (35.6%) patients, and 6 (6.6%) patients had luminal cancer involvement at the time of perforation. The overall hospitalization rate related to perforation was 95.5%, with mortality of 15.5% during the same admission. Antibiotics were given in 95% of our sample; 37.8% of patients also required surgical/interventional radiology intervention. Forty-six patients (51.1%) had perforation-related complications (e.g., sepsis, fistula, abscess), which were associated with a higher mortality rate (30%). Our findings suggest a low incidence of bowel perforation after ICI treatment (0.6%), with 40% of patients having coexisting bowel inflammation as a potential contributing factor. Patients with bowel perforation had an aggressive disease course and high rates of hospitalization, complications, and mortality. Early recognition and prompt intervention is critical to improve patient outcomes. Future studies are warranted to further investigate the cause, predictive markers, and optimal treatment for this patient population.


Introduction
Immune checkpoint inhibitors (ICIs) are paramount in the treatment of various malignancies and improve patient survival. To date, more than eight ICI agents are approved. While conferring an appreciable survival benefit, these agents also produce unique immune-related adverse events (irAEs), with diarrhea and colitis among the most common.
Immune-mediated colitis has been reported in up to 40% of patients treated with ICIs and varies widely in severity (Prieux-Klotz et al. 2017;Cho 2019;Shivaji et al. 2019). Failure in early recognition and delayed or suboptimal treatment early in the disease course can lead to an increased risk of later complications such as bowel perforation .
Bowel perforation of the gastrointestinal tract is a serious medical emergency in oncological patients (Yeun 2022). Bowel perforation can also be observed in patients with infections such as enterocolitis (Bavaro 2002) (which has already been extensively described in patients receiving ICIs (Tian et al. 2018) or other diseases causing inflammation of the gut; patients receiving ICIs may, therefore, be at an increased risk for perforation compared to other patient populations. Bowel perforation in patients receiving ICIs has only been described in case reports (Beck et al. 2019;Beck 2020;Burdine et al. 2014;Dilling et al. 2014;Shah et al. 2017;Zhou et al. 2021;Celli et al. 2018;Yasuda et al. 2017;Delasos et al. 2019;Delasos et al. 2019;Kim et al. 2019). In the available literature, the majority of patients (six of ten) had a diagnosis of colitis prior to the perforation. Perforation following colitis was seen mostly in patients with metastatic melanoma treated with anti-CTLA-1 inhibitors or combination ICI therapy. The treatment of choice was exploratory laparotomy. Of the ten patients in the available reports, two patients died of perforation or its related complications (Beck et al. 2019;Beck 2020;Burdine et al. 2014;Dilling et al. 2014;Shah et al. 2017;Zhou et al. 2021;Celli et al. 2018;Yasuda et al. 2017;Delasos et al. 2019;Delasos et al. 2019;Kim et al. 2019).
It is still unclear what risk factors are associated with bowel perforation in patients receiving ICIs. Notably, combination regimens of ICIs and other cancer treatments are frequently employed in current oncologic practice, which adds to the complexity of this phenomenon. Previous studies have shown the effectiveness of cancer response to the combination of ICIs and VEGF inhibitors such as bevacizumab in renal cancer (Tang et al. 2020), raising questions regarding the risk of bowel perforation in these patients, as bevacizumab is a medication known to cause rare bowel perforation (Chen et al. 2013). Effective management for this severe complication remains under-investigated. Exploration of modifiable risk factors and understanding the disease course are, therefore, essential to improve patient outcomes. To address this knowledge gap, we performed a retrospective descriptive analysis of the clinical course, complications, and outcomes of cancer patients with bowel perforation in the context of ICI exposure.

Study design and population
This retrospective, descriptive, single-center study included ICI-treated adult patients with bowel perforation between January 1, 2010, and April 30, 2021. The study was conducted with approval from the Institutional Review Board. We identified patients aged 18 years or older who were (1) treated with an ICI and (2)

Clinical data
Basic and cancer-related information such as age, sex, primary cancer type, stage, ICI, and other cancer treatment and doses and Charlson Comorbidity Index were collected. Data related to bowel perforation included clinical symptoms, imaging description (including location of perforation), endoscopic findings, coexisting conditions that could be potential contributing factors (such as active colonic inflammation [e.g., colitis, diverticulitis, appendicitis, enteritis], cancer involvement, iatrogenic [endoscopy] or infectious causes), treatment (medical; antibiotics or surgical; type of surgery; interventional radiology drainage), and outcome (hospitalization, surgery, complications [e.g., sepsis, abscess, fistula, death]). For those with a reported gastrointestinal irAE (enterocolitis), severity of diarrhea and colitis was defined based on the National Cancer Institute's Common Terminology Criteria for Adverse Events (CTCAE) version 5.0 (Services UDoHAH 2017). Information on other coexisting irAEs within a 3-month window of the bowel perforation was also documented.

Statistical analysis
A descriptive statistical analysis was performed. The distribution of continuous variables was summarized using median and interquartile range (IQR). The distribution of categorical variables was summarized by frequencies and percentages, and the associations of these variables were evaluated with the Whitney U test and Chi-square test. All statistical evaluations were two-sided, and P values less than 0.05 were considered statistically significant.

Patient population
During the study period, a total of 13,991 patients received at least one of the following ICI agents: ipilimumab, nivolumab, cemiplimab, pembrolizumab, tremelimumab, atezolizumab, durvalumab or avelumab. Of that total, we identified 90 eligible patients who met the inclusion criteria (Fig. 1). The incidence of bowel perforation was 0.6% among the base number of ICI-treated patients in the study window.

Patient characteristics and oncologic history
Among the 90 patients evaluated, the median age was 64 years, 54.4% were male and 73.3% were white (Table 1). The majority of the patients were diagnosed with melanoma (23.3%), followed by gastrointestinal cancer (16.6%). The majority (87.7%) of patients had a diagnosis of stage IV malignancy. Most patients (58.8%) received a PD-1/L1 inhibitor as monotherapy, 15.5% received CTLA-4 inhibitor monotherapy, and 25.5% received a combination of both.
The systemic irAEs occurring in these patients are summarized in Table 1. Colitis was present in seven patients (7.7%) within 3 months of bowel perforation. The next most common irAEs were endocrine (often thyroiditis) and pulmonary in nature, present in six patients (6.6%) each.

Perforation-related characteristics
The median time from ICI initiation to onset of bowel perforation was 3 months (IQR: 1-6), and perforation occurred after a median of four (IQR: 2-7) treatment cycles ( Table 2). The most common clinical symptom was abdominal pain (95.5% of patients). Other pertinent findings on physical exam included guarding in 18 patients (20%) and fever in 15 patients (16.6%). Based on imaging and/or surgical reports, the colon was the most common location for perforation (37.7%), followed by the small bowel (30%) and appendix (12.2%). Potential risk factors contributing to perforations were active colonic inflammation (35.6%), cancer involving the bowel (6.6%) and iatrogenic etiology (2.2%). Active colonic inflammation was due to diverticulitis in 21 patients (23.3%), enterocolitis in 7 patients (7.7%), and appendicitis in 4 patients (4.3%), with perforation occurring in that same location (Table 3). Fifty patients (55.5%) had no identifiable cause of bowel perforation. In our cohort, four patients (4.3%) had a micro-perforation or contained perforation visible only on imaging. Comparison between patients with perforation within 6 months and beyond 6 months after the last dose of ICI did not identify any difference in underlying cancer type, ICI treatment or perforation-related characteristics and outcome.

Treatments and outcomes
All our patients received treatment for bowel perforation. The majority received antibiotics (95.5%), 20 patients underwent surgery (22.2%) and 14 patients (15.5%) underwent an interventional radiology procedure (Table 2). Hospitalization occurred in 95.5% of patients, with a median hospital stay of 7 days. Fourteen patients (15.5%) died during the same admission. Following perforation, 71 patients (78.8%) received no further cancer treatment, 11 patients continued their ICI, and the remaining 8 patients were switched to other treatments. All-cause mortality was 78.8%.

Discussion
ICI treatment has become the mainstay for treating different types of cancer in modern oncology because of their improved survival outcomes. However, irAEs have been recognized as major obstacles to continuing immunotherapy, leading to frequent treatment interruptions. Post-ICI bowel perforation can have fatal consequences; despite this, quality data in the literature are still lacking. Our study is one of the largest to explore this phenomenon.  ICI immune checkpoint inhibitor, IQR interquartile range, IR interventional radiology * The information on perforation location was inconclusive in the charts of 14 patients † The information on the potential cause of bowel perforation was not available in the remaining 50 patients. Only two patients with previous history of iatrogenic perforation produced after endoscopy ‡ None of the remaining 58 patients had any comorbid bowel disease § In four patients (4.3%), an IR procedure (drainage) was followed by surgery at a later date We demonstrated a low bowel perforation incidence of 0.6% among ICI-treated patients, with 40% having coexisting bowel abnormalities as a potential contributing factor. Patients with bowel perforation had an aggressive disease course, high rate of hospitalization, and high complication rate, leading to a higher mortality rate. Bowel perforation can be detrimental for patients with advanced stages of cancer (Banaszkiewicz et al. 2014); mortality in these cases has been reported to be as high as 30% and even 70% in patients with peritonitis (Shin et al. 2016). Bowel perforation has been estimated to occur in 1.0% to 1.5% of patients with melanoma and up to 6% of patients with renal cell carcinoma treated with ipilimumab (Haanen et al. 2017). This is inconsistent with our study finding of 0.6%, which is likely due to the large variety in cancer types and specific ICIs used among our patients. Despite a high complication rate of 51%, only 15.5% of our patients had perforation-related death. Early recognition and prompt intervention may have contributed to this low mortality rate.
Combinations of ICIs with non-ICI agents such as VEGF inhibitors are frequently utilized in practice. The risk and incidence of bowel perforation among patients with combined regimens remain to be investigated. The incidence rate of 0.6% in our ICI-treated cohort is lower compared to reported rates of perforation secondary to VEGF inhibitors such as bevacizumab and lenvatinib alone (Hapani et al. 2009;Badgwell et al. 2008;Valerio et al. 2021;Shaikh et al. 2020). These medications may have higher risk of causing perforation through multiple mechanisms, including impaired wound healing, wound infection, ecchymosis, dehiscence, and surgical site bleeding, which further predispose to complications after surgery. Invasive procedures are, therefore, frequently delayed for 4-8 weeks out of concern for safety. This scenario has not been studied in the setting of ICIs (Adhikari et al. 2021). No known mechanism has been proposed or found for ICI-induced bowel perforation. Among our cohort, six patients received a combination of ICI and bevacizumab, but the incidence of bowel perforation in patients receiving this combination is not clear given the lack of base number on this combination regimen. However, all six patients developed complications related to bowel perforation. There is also evidence suggesting that the combination use of ipilimumab with IL-2 therapy was associated with higher risk of perforation (Smith et al. 2007). The risk of perforation may, thus, be augmented in a select group of patients, especially in the context of more frequent use of combination regimens, for whom heightened attention to monitoring for this complication is warranted in certain cases.
It has been recognized that coexisting bowel conditions could contribute to bowel perforation after ICI exposure among cancer patients. These could take the form of the malignancy itself, treatment-related gastrointestinal irAEs, or the other forms of bowel inflammation mentioned above. Within our cohort, 16.6% of patients had confirmed primary gastrointestinal malignancy or luminal metastasis; at least 31% of the patients had imaging evidence of diverticulitis, appendicitis or enterocolitis at the time of perforation, similar to what has been shown in previous case reports (Beck et al. 2019;Beck 2020;Burdine et al. 2014;Dilling et al. 2014;Shah et al. 2017;Zhou et al. 2021;Celli et al. 2018;Yasuda et al. 2017;Delasos et al. 2019;Delasos et al. 2019;Kim et al. 2019). Perforation secondary to diverticular disease is suggested to comprise up to 15% of cases, while appendiceal perforation has been reported to be more prominent (28.5%) (Chow et al. 1997;Balogun et al. 2019). ICI-related enterocolitis has been recognized as a high-risk condition for perforation. Given the absence of a history of bowel perforation or bowel disorders before starting ICIs in our cohort, cancer diagnosis and/or cancer treatment could lead to de novo changes in the integrity of the gastrointestinal wall, resulting in a lower threshold for certain complications. In addition, our study did not identify a difference in the clinical presentations and outcomes with respect to the timeframe of perforation onset after ICI exposure. The window of potential risk posed from beginning ICI treatment to the pathologic process in the gastrointestinal tract is uncertain; therefore, a 1-year window was used for inclusion of perforation cases, based on the existing evidence that colitis, one of the most common irAEs, has been reported to persist longer than 2 years (Zou et al. 2020). Further studies are needed to expand our knowledge in this domain.
Making an early diagnosis of perforation can be challenging due to its similar presentation to other conditions. Abdominal imaging modalities, the most common methods used for patients in our cohort, are generally adequate and sensitive for establishing the diagnosis if the patient is hemodynamically stable. However, emergent intervention, i.e., surgery, particularly in a hemodynamically unstable patient, should not be delayed if there is high suspicion of perforation, even in the absence of confirmation by imaging (Wu et al. 2021). In the cancer population, many comorbidities may preclude patients from being surgical candidates. In these cases, aggressive antibiotic treatment with potential drainage by interventional radiology may represent an alternate strategy. Unfortunately, our small sample size did not allow for adequate comparison of risks and benefits between surgical and non-surgical options. A study that includes a larger sample size would help to further clarify the optimal treatment for bowel perforation in the cancer population. Regardless, the presence of malignancy alone is a poor prognostic factor for negative outcomes in cancer patients (Wu et al. 2021).
Our study identified one unique case of a patient with perforation located along the greater curvature of the stomach (Fig. 2a-c), a complication not yet described in the literature. The patient had a diagnosis of stage IV neuroendocrine pancreatic carcinoma with metastasis to the liver. Perforation occurred spontaneously after 13 cycles of ipilimumab, which was subsequently permanently discontinued because of the perforation event, and the patient died shortly after the diagnosis. In addition, we observed four patients with contained micro-perforation ( Fig. 2d-f). Despite the unknown pathogenesis of this presentation (Chang et al. 2017), these four patients had a fairly benign disease course compared to the rest of the cohort and had good responses to conservative antibiotic treatment. Of the four patients, three did not require further surgery and survived, although none of them were eligible to resume any type of cancer treatment after the perforation. These cases highlight the wide spectrum of perforation presentations and disease course, necessitating close monitoring and a firm understanding of this disease entity.
Our study represents the largest study on this topic to date but possesses certain limitations. First, it is single center and retrospective in nature; the availability of the information on chart review could have led to a potential underestimation of the real perforation incidence and all the pertinent risk factors of perforation. Second, the small sample size limits the power of subgroup analysis to compare different treatment modalities. Third, the findings may not apply to general practice, given the unique setting of a tertiary cancer center that has (1) a patient population with more complex advanced cancer conditions with comorbidities and (2) more available resources, expertise, and support for alternative treatment options. The general management of the patients at this cancer center compared to the practice at other hospitals could be more expectant and conservative owing to the already negative prognoses of these patients.

Conclusion
Bowel perforation is extremely rare in patients receiving ICIs, with incidence of 0.6% in our sample. However, it could be associated with an aggressive disease course, with serious complications and fatal outcomes despite antibiotics, surgery or interventional radiology procedures. Early recognition and prompt management are essential to mitigate the negative impact to patient outcomes. Further prospective studies are warranted to clarify the triggering factors and the optimal management strategy for this life-threatening condition when treating patients with ICIs.
Author contributions YW was the senior author of the study; she developed the concept, designed the study, interpreted the results, ensured the preservation of data accuracy and integrity at all stages, agreed to be accountable for all aspects of the study, oversaw the overall direction and planning of the study, and contributed to the writing of the manuscript, with input from all authors. APM collected the data for the study, conducted and interpreted the analysis, and wrote the manuscript. MS and CL helped with data collection and analysis. All authors critically revised the final version of the manuscript as well as had access to the study data and reviewed and approved the final manuscript.
Funding This research did not receive any specific grant from funding agencies in the public, commercial, or not-for-profit sectors.

Data availability
The data sets used and analyzed in this study are available from the corresponding author on reasonable request.

Conflict of interest
The authors have not received any financial consideration from any person or organization to support the preparation, analysis, results or discussion of this article.

Fig. 2
Imaging of perforation of the stomach and contained microperforation. a-c, Computed tomography (CT) images of the abdomen demonstrating large defect or perforation along the stomach great curvature (arrows) in coronal (a), sagittal (b), and axial (c) views. d-f, Abdominopelvic CT images demonstrating contained micro-perforation of the colon with small air pockets and adjacent fat infiltration (arrows) in axial (d), sagittal (e), and coronal (f) views