Endoscopic and Histologic Features of Immune Checkpoint Inhibitor-Related Colitis

Abstract

Background

Diarrhea and colitis are the second most common immune checkpoint inhibitor (ICPI)-induced adverse events. However, a comprehensive characterization of the endoscopic and histologic features of ICPI-induced diarrhea and colitis is lacking. Therefore, we aimed to describe endoscopic and histologic features of ICPI-induced gastrointestinal toxicities and to assess their association with patients’ clinical characteristics and outcomes.

Methods

We retrospectively reviewed records of 53 patients with ICPI-related diarrhea/colitis between 2011 and 2017. We collected data on demographics, diarrhea/colitis grade, treatment, and endoscopic and histologic findings. Long-term follow-up included repeat endoscopy findings, diarrhea recurrence, and overall survival. We compared groups by treatment, endoscopic and histologic findings, and constructed Kaplan-Meier survival curves.

Results

Most patients had grade 2 or higher diarrhea (87%) and colitis (60%). Thirty-one patients were successfully treated with corticosteroids, and 22 additionally required infliximab. On endoscopy, 21 (40%) patients had ulcerations and 22 (42%) had nonulcerative inflammation. Patients with ulcerations had more steroid-refractory disease (P = 0.044) and high-grade diarrhea (P = 0.033). Histology showed mostly acute (23%) or chronic (60%) inflammation. During mean follow-up duration of 18.9 months, 19 (36%) developed recurrent diarrhea. Most patients had persistent endoscopic (8/13, 62%) and histologic (9/11, 82%) inflammation. Patients with higher-grade adverse events had improved survival. Higher-grade colitis was associated with endoscopic inflammation (P = 0.039), but grade of diarrhea was not associated with endoscopic inflammation or grade of colitis.

Conclusion

immune checkpoint inhibitor, diarrhea, colitis, adverse event, endoscopy, histology, immunotherapy

INTRODUCTION

Immune checkpoint inhibitors (ICPIs) target the host immune system so that malignant cells cannot evade it. This class of drugs has rapidly expanding indications for the treatment of advanced malignancies.^1–3 ICPIs have shown great promise in improving patient survival and delaying the progression of an increasing number of malignancies, including melanoma, nonsmall cell lung cancer, renal cell carcinoma, and squamous cell cancers of the head and neck cancer.^{4, 5} The currently available ICPIs, antibodies inhibiting cytotoxic T-lymphocyte-associated protein-4 (CTLA-4), programmed cell death protein-1 (PD-1), and programmed death-ligand 1 (PD-L1), target pathways that inhibit cytotoxic T-cell activation, proliferation, and function.^{1, 2} These agents strengthen native adaptive T-cell immunity against neoplastic cells and induce loss of antitumor immune tolerance.^{1, 2}

Whereas the augmentation of patients’ native immune systems by ICPIs leads to improved tumor control, nonspecific immune activation can also cause damage to normal tissue, resulting in the development of immune-related adverse events affecting a variety of organ systems, including the skin, endocrine organs, gastrointestinal (GI) tract, liver, and lungs.^{3, 6–9} Findings from clinical trials show that immune-related adverse events occur frequently, with the most common severe or life-threatening (grades 3 and 4) events occurring in the GI tract.⁸ Immune-related adverse events involving the GI tract have been reported in approximately 21% to 44% of patients treated with CTLA-4 inhibitors and less frequently in patients treated with PD-1/PD-L1 inhibitors. Typically, patients present with diarrhea that is often self-limiting; however, if it is not properly managed, ICPI-related colitis can lead to serious morbidities and mortality.^{7, 9, 10} Treatment for ICPI-related colitis is typically based on symptom severity and involves immunosuppression with corticosteroids for moderate to severe symptoms; in steroid-refractory cases, biologics such as infliximab and vedolizumab have been used.^{9, 11, 12} However, long-term outcome data are lacking.

ICPI-related GI toxicities became recognized in our daily practice as the use of ICPIs is increasing rapidly over the past 5 to 10 years.^6–11 Our knowledge about the pathogenesis, clinical presentation, severity assessment, evaluation modality, and management strategy of ICPI-related GI toxicities is still very limited. Clinical practice is mainly based on the experience of individual practitioners, practice group, and institution. The few available practice guides from experts in the field are mostly based on case reports and case series.^{7, 9, 11} Recently, this lack of guidelines spurred multiple professional oncology societies to develop practice guidelines for general oncologists in the community.¹³ However, very little published literature describes the endoscopic and histologic characteristics of ICPI-related colitis and long-term patient outcomes. To optimize the care of this treatment-limiting adverse event, we need to better understand its endoscopic and histologic features and identify factors and markers that could be used to identify high-risk populations for more aggressive treatment.

In this study, we aimed to systematically describe the endoscopic and histologic characteristics of a large cohort of patients who developed diarrhea and colitis following treatment with anti–CTLA-4 or anti–PD-1 ICPIs. We also aimed to determine whether radiologic, endoscopic, and histologic findings correlated with clinical characteristics and to analyze patients long-term outcomes.

MATERIALS AND METHODS

Patient Selection and Data Collection

We performed a retrospective study using electronic medical records of patients who received treatment with ICPIs and underwent endoscopy and biopsy for GI symptoms eg, diarrhea, bleeding, abdominal pain, at The University of Texas MD Anderson Cancer Center between March 2011 and March 2017. Approval for this study was obtained from the Institutional Review Board at MD Anderson. Informed consent requirement was waived for this retrospective study.

The included patients were adults who (1) had received an anti–CTLA-4 agent, an anti–PD-1 agent, or a combination of both agents for the treatment of a malignancy; (2) had diarrhea or colitis with onset after initiation of ICPI treatment; (3) had endoscopy and biopsy for the diarrhea or colitis workup; and (4) discontinued ICPI therapy and initiated immunosuppressive therapy because of ICPI-related GI toxicity. Patients whose diarrhea was attributed to other etiologies were excluded.

A total of 53 patients were included. We collected data regarding patient characteristics, medical and oncologic history, clinical characteristics of and treatment history for ICPI-related colitis, computed tomography (CT) scans, endoscopy and pathology reports, and long-term follow-up information. Patient characteristics included age, sex, presence of comorbidities, smoking history, and current or prior (within 3 months of ICPI initiation) use of nonsteroidal antiinflammatory agents (NSAIDs). Comorbidities included coronary artery disease, congestive heart failure, chronic obstructive pulmonary disease, human immunodeficiency virus, atrial fibrillation, graft versus host disease (GVHD), asthma, hypertension, diabetes mellitus, dyslipidemia, and hypocorticolism. Oncologic history included type of malignancy, cancer stage based on the American Joint Committee on Cancer (AJCC) Cancer Staging System, ICPI regimen(s), and additional chemotherapy received (chemotherapeutic agents, targeted therapies, biotherapeutic agents, hormonal therapy, miscellaneous agents, and clinical trial agents).

Clinical Evaluations

The clinical information collected included diarrhea and colitis grading, type of immunosuppressive therapy received for ICPI-related diarrhea/colitis, and the duration of therapy. Diarrhea and colitis were graded using the National Cancer Institute Common Terminology Criteria for Adverse Events (CTCAE), Version 4.03. Diarrhea was graded as follows: grade 1: increase of < 4 stools per day over baseline; grade 2: increase of 4 to 6 stools per day over baseline; grade 3: increase of >=7 stools per day over baseline; grade 4: life-threatening consequences; and grade 5: death. Colitis was graded as follows: grade 1: asymptomatic, pathologic, or radiographic findings only; grade 2: abdominal pain, mucus, or blood in stool; grade 3: abdominal pain, fever, change in bowel habits with ileus, or peritoneal signs; grade 4: life-threatening consequences such as perforation, bleeding, ischemia, necrosis, or toxic megacolon; grade 5: death.

Endoscopy

Original endoscopy reports were reviewed for the type of endoscopy conducted, the gross description, distribution of colitis, and inflammation pattern. The gross description included the presence of mucosal ulcerations (ulcers of any size), nonulcerative inflammation (erythema, exudate, loss of vascular pattern, edema, friability, erosions), or normal appearance. The distribution of colitis was classified as left colon only, both left and right colon, or ileum. As a summary measure, we used “extensive colitis” to describe inflammation involving both left and right colon and/or the colon and terminal ileum. The pattern of inflammation, categorized as diffuse, patchy, or segmental features, was also recorded.

Histology

The pathology reports from biopsy samples taken via colonoscopy or sigmoidoscopy performed at the time of diarrhea/colitis onset were used as the baseline. Inflammation patterns were categorized as acute, chronic, lymphocytic, or normal. Acute patterns of inflammation were exhibited with the following descriptions: neutrophilic or eosinophilic infiltrate, cryptitis, crypt abscess, and apoptosis. Chronic inflammation was described as basal lymphocytic infiltrate, cryptic architecture distortion, and Paneth cell metaplasia. The lymphocytic pattern referred to the presence of increased intraepithelial infiltration of lymphocytes, which is similar to a subtype of microscopic colitis. There existed cases with a combination of different abnormal histologic features. To simplify the classification, we ranked their significance in the following decreasing order: chronic inflammation, lymphocytic colitis, and acute inflammation. In the event any combination occurred, that case was listed under the histologic feature with a higher order.

Radiology

Available CT scan reports were reviewed to determine whether signs of colitis were present on CT scan of abdomen and pelvis.

Long-term Follow-up

Information about diarrhea recurrence, repeat colonoscopy/sigmoidoscopy, and treatment regimens for diarrhea recurrence were collected for each included patient until the last follow-up visit. Overall survival (OS) was defined as the time from initial exposure to an ICPI until death or last follow-up.

Statistical Analysis

The distribution of continuous variables was summarized by the mean, standard deviation, and range. The distribution of categorical variables was summarized using frequencies and percentages. Continuous variables were compared between treatment groups using the Wilcoxon rank-sum test or Kruskal-Wallis test (for comparisons of more than 2 groups). The Fisher exact test or the chi-square test was conducted to evaluate associations between categorical variables. Spearman correlation was used to assess the relationship between 2 ordinal variables. Kaplan-Meier curves were used to estimate unadjusted OS. Log-rank tests were used to compare OS between groups. All statistical tests were 2-sided. P values less than or equal to 0.05 were considered statistically significant. Statistical analyses were carried out using SAS version 9.4 (SAS Institute, Cary, NC) and SPSS version 24.0 (SPSS Inc., Chicago, IL).

RESULTS

Colitis Clinical Features and Treatment

A total of 53 patients with ICPI-related diarrhea and colitis were included in this analysis. The baseline characteristics of the patients are shown in Table 1. The GI symptoms of 31 (58%) patients responded to treatment with corticosteroids alone, and 22 (42%) patients received and responded to infliximab for steroid-refractory disease course. Similar demographic and clinical characteristics were found in patients whose diarrhea/colitis responded to steroids and those who had steroid-refractory disease; no significant differences in age, sex, comorbidities, type of malignancy, or type of ICPI used were found (Table 1). Advanced cancer (stage IV) was more prevalent in the steroid-only group than the steroid + infliximab group (94% vs 68%, P = 0.025; Table 1). The duration of follow-up ranged from 1.8 months to 46.4 months, with mean of 18.9 months (SD 17.2). Most of the patients (46/53; 87%) had CTCAE grade 2 to 4 diarrhea and grade 2 to 3 colitis (32/53; 60%) (Table 2). No patients had grade 4 colitis. The grades of diarrhea and colitis were comparable in the steroid-only and the steroid + infliximab groups (P = 0.218 and P = 0.779, respectively; Table 2).

Table 1:

Association of Patient Characteristics with Treatment Group

Characteristic	Steroid Only No. = 31	Steroid + Infliximab No. = 22	P value
Age in years, mean (SD)	60.2 (17.0)	59.5 (14.5)	0.630
Male sex, no. (%)	18 (58)	17 (77)	0.239
Comorbidities present, no. (%)	20 (65)	15 (68)	1.000
Smoking, no. (%)	14 (45)	4 (18)	0.076
Malignancy type, no. (%)			0.376
Melanoma	20 (65)	17 (77)
Other	11 (35)	5 (23)
Cancer stagea, no. (%)			0.025
Stages III	2 (6)	7 (32)
Stage IV	29 (94)	15 (68)
Checkpoint inhibitor type, no. (%)			0.750
Ipilimumab	21 (68)	15 (68)
Combination b	6 (19)	3 (14)
Pembrolizumab	4 (13)	3 (14)
Tremelimumab	0 (0)	1 (5)
Additional chemotherapy, no. (%)	13 (42)	6 (27)	0.385
Follow-up duration in months, mean (SD)	22.5 (22.2)	16.4 (10.0)	0.196

^aAmerican Joint Committee on Cancer (AJCC) Cancer Staging System

^bCombination: ipilimumab + nivolumab

Table 2:

Association of Colitis Characteristics with Treatment Group

Characteristic	Steroid Only No. = 31	Steroid + Infliximab No. = 22	P value
Diarrhea grade, no. (%)			0.218
Grade 1	6 (19)	1 (5)
Grades 2–4	25 (81)	21 (95)
Colitis grade, no. (%)			0.779
Grade 1	13 (42)	8 (36)
Grades 2–3	18 (58)	14 (64)
Endoscopic findings, no. (%)			0.017
Ulcer	8 (26)	13 (59)
Nonulcer inflammation	14 (45)	8 (36)
Normal	9 (29)	1 (5)
Distribution of colitis, no. (%)			0.094
Ileum alone	0 (0)	1 (5)
Right + left colon	9 (29)	8 (36)
Left colon alone	9 (29)	9 (41)
Ileum + colon	3 (10)	3 (14)
Normal	10 (32)	1 (5)
Pattern of endoscopic inflammationa, no. (%)			0.563
Diffuse	13 (59)	9 (43)
Patchy	8 (36)	10 (48)
Segmental	1 (5)	2 (10)
Histologic inflammation pattern, no. (%)			0.264
Chronic	17 (55)	15 (68)
Acute	6 (19)	6 (27)
Lymphocytic	4 (13)	0 (0)
Normal	4 (13)	1 (5)
Apoptosis present, no. (%)	6 (19)	6 (27)	0.524

^a Total number of patients who had endoscopic inflammation is 43 and another 10 patients had normal endoscopy that were not included in this variable.

Endoscopic Findings

Overall endoscopic features

All 53 patients underwent endoscopic evaluation via either colonoscopy (35 patients) or flexible sigmoidoscopy (18 patients) and biopsy. Colitis with ulceration was found in 21 (40%) patients, nonulcerative inflammation in 22 (42%) patients, and no gross inflammation in 10 (19%) patients (Table 2). Among the 43 patients with inflammation on endoscopy, 18 (43%; from 12 sigmoidoscopies and 6 full colonoscopies) patients had documented colitis in the left colon, 17 (40%) had colitis with both left and right colon involvement (from 21 full colonoscopies and 2 sigmoidoscopies), 6 (14%) had ileocolonic disease, and 1 (2%) had inflammation confined to the ileum only (Table 2). Most of the inflammation had a diffuse pattern (22, 51%); patchy (18, 42%) and segmental (3, 7%) patterns were less common (Table 2). Generally, the endoscopic features were comparable in patients who were treated with steroids alone and those who were treated with combined steroid and infliximab (Table 2).

Endoscopic features of ICPI-related colonic inflammation

ICPI-related colonic inflammation had a spectrum of presentations on endoscopic images, including large, deep ulcerations, diffuse or patchy erythema, inflammatory exudate, loss of vascular pattern, aphthae, edema, friability, erosions, and normal appearance (Fig. 1).

FIGURE 1.

Representative endoscopic findings: (A) large deep ulceration, (B) diffuse erythema with exudate, (C) patchy erythema, (D) aphtha (indicated by yellow arrow), (E) edema, and (F)normal.

Ulcers on endoscopy

Patients who had ulcerations on endoscopy had steroid-refractory disease course significantly more often than did patients with a normal-appearing colon and those with nonulcerative inflammation (62% vs 31%; P = 0.044; Table 3). NSAID use did not significantly trigger the ulcer development (29% vs 41%; P = 0.388). Persistent ulcers were more commonly found on repeat endoscopy in patients who had ulcerations on the initial endoscopy (60% vs 0%; P = 0.033), whereas more nonulcerative inflammation was seen on follow-up endoscopy in the group of patients with nonulcerative inflammation in the initial evaluation (71% vs 0%; P = 0.033; Table 3). Patients who had ulcers on initial endoscopy had a significantly higher rate of high-grade diarrhea than did those who had no ulcers (100% vs 79%; P = 0.033; Table 3).

Table 3:

Association of Patient, Diarrhea, and Colitis Characteristics with Endoscopic Findings

Characteristic	Ulcers No. = 21	No Ulcersa No. = 29	P value
Cancer stage, no. (%)			0.025
Stage III	7 (33)	2 (7)
Stage IV	14 (67)	27 (93)
Checkpoint inhibitor type, no. (%)			0.113
Ipilimumab	18 (86)	16 (55)
Combinationb	2 (9)	6 (21)
Pembrolizumab	1 (5)	6 (21)
Tremelimumab	0 (0)	1 (3)
NSAID use, no. (%)	6 (29)	12 (41)	0.388
Colitis treatment, no. (%)			0.044
Steroids	8 (38)	20 (69)
Infliximab + steroids	13 (62)	9 (31)
Diarrhea grade, n o.(%)			0.033
Grade 1	0 (0)	6 (21)
Grade 2–4	21 (100)	23 (79)
Colitis grade, n (%)			0.760
Grade 1	6 (29)	13 (45)
Grade 2–3	15 (71)	16 (55)
Histologic findings, no. (%)			0.951
Chronic	14 (67)	18 (62)
Acute	5 (24)	7 (24)
Lymphocytic	1 (5)	3 (10)
Normal	1 (5)	1 (3)
Recurrent diarrhea, no. (%)	8 (38)	9 (31)	0.763
Repeat endoscopic findingsc, no. (%)			0.033
Ulcer	3 (60)	0 (0)
Nonulcerative inflammation	0 (0)	5 (71)
Normal	2 (40)	2 (29)

^aIncludes nonulcerative endoscopic inflammation cases (n = 22) and normal endoscopies with positive histologic inflammation (n = 7)

^bCombination: ipilimumab + nivolumab

^cThe total number of patients who underwent repeat endoscopy is 5 in the ulcers group and 7 in the no ulcers group

Distribution of colonic inflammation

Left-sided colitis was mainly identified by flexible sigmoidoscopy (12/18, 67%), and extensive colitis was more commonly reported via colonoscopy (21/23, 91%; P < 0.001). Otherwise, the distribution of colitis was not significantly associated with any other clinical factor, eg, cancer stage, diarrhea/colitis grade and treatment, or endoscopy findings.

Histologic findings and images

Histologic evaluations of the initial biopsy specimens showed that most patients had either acute (12/53; 23%) or chronic (32/53; 60%) inflammation. Four (8%) patients had increased lymphocytic infiltration in the epithelium, and 5 (9%) had normal histology (Tables 2, 23, 24). Apoptosis as a feature of acute inflammation was seen in only 12 (23%) of the biopsy samples (Table 2). Six (11%) patients with acute or chronic inflammation on histology had a grossly normal colon on endoscopy (Table 4). There were no significant differences among histology groups in terms of colitis treatment, diarrhea grade, or colitis grade (Table 4). Patients who had initial histology findings of acute or chronic inflammation tended to have persistent inflammation on repeat colonoscopy and repeat histologic evaluation (Table 4).

Table 4:

Association of Clinical Characteristics with Initial Histologic Findings

	Initial Histology Findings
Characteristic	Acutea No. = 12	Chronicb No. = 32	Lymphocyticc No. = 4	Normal No. = 5	P value
Cancer stage, no. (%)					0.434
Stage III	1 (8)	5 (16)	1 (25)	2 (40)
Stage IV	11 (92)	27 (84)	3 (75)	3 (60)
Checkpoint inhibitor type, no. (%)					0.685
Ipilimumab	10 (83)	20 (62)	3 (75)	3 (60)
Combinationd	0 (0)	6 (19)	1 (25)	2 (40)
Pembrolizumab	2 (17)	5 (16)	0 (0)	0 (0)
Tremelimumab	0 (0)	1 (3)	0 (0)	0 (0)
Diarrhea grade, no. (%)					0.444
Grade 1	3 (25)	3 (9)	0 (0)	1 (20)
Grade 2–4	9 (75)	29 (91)	4 (100)	4 (80)
Colitis grade, no. (%)					0.479
Grade 1	5 (42)	11 (34)	3 (75)	2 (40)
Grade 2–3	7 (58)	21 (66)	1 (25)	3 (60)
Colonoscopy findings, no. (%)					0.339
Ulcer	5 (42)	14 (44)	1 (25)	1 (20)
Nonulcerative Inflammation	5 (42)	14 (44)	2 (50)	1 (20)
Normal	2 (17)	4 (13)	1 (25)	3 (60)
Repeat colonoscopy findingse, no. (%)					0.518
Ulcer	2 (40)	1 (14)	0	0
Nonulcerative inflammation	1 (20)	4 (57)	0	0
Normal	2 (40)	2 (29)	0	1 (100)
Repeat histology findingsf, no. (%)					0.149
Chronic	4 (80)	1 (20)	0	0
Acute	1 (20)	2 (40)	0	0
Lymphocytic	0 (0)	1 (20)	0	0
Normal	0 (0)	1 (20)	0	1 (100)

^aAcute: findings of acute inflammation alone without chronic or lymphocytic colitis features

^bChronic: findings of chronic inflammation alone and/or any combination of histologic features that includes chronic inflammation

^cLymphocytic: findings consistent with lymphocytic colitis alone or its combination with acute colitis features

^dCombination: ipilimumab + nivolumab

^eThe total number of patients who underwent repeat endoscopy are 5 in acute group, 7 in chronic group, 0 in lymphocytic group, and 1 in normal group

^fThe total number of patients who underwent repeat biopsy are 5 in acute group, 5 in chronic group, 0 in lymphocytic group, and 1 in normal group

The spectrum of histologic features of ICPI- related colitis included acute colitis with neutrophilic infiltration, cryptitis, crypt abscess, and intraepithelial apoptosis; chronic colitis with basal lymphocytic infiltration and crypt distortion; and lymphocytic colitis with increased intraepithelial lymphocytic infiltration (Fig. 2).

FIGURE 2.

Representative histologic images: (A) active colitis with crypt architectural distortion and dropout (H&E, 20×); (B) active colitis with neutrophilic cryptitis, erosion, and crypt abscess (H&E, 20×); (C) neutrophilic cryptitis with intraepithelial apoptosis (H&E, 20×); and (D) increased intraepithelial lymphocytic infiltration (H&E, 40×).

Clinical and endoscopic follow-up

During the mean follow-up period of 18.9 months (SD 17.2), 19 of the 53 patients (36%) developed recurrent diarrhea despite initial successful treatment. For their recurrent diarrhea, 3 patients were treated with antimotility agents only, 10 received a repeat course of steroids, and 6 required infliximab. Factors such as initial diarrhea or colitis grade, presence of inflammation on initial endoscopy or histology, initial infliximab requirement, and duration of steroid use were found not to be associated with diarrhea recurrence. Thirteen patients underwent repeat endoscopy with biopsy, 8 for recurrent diarrhea or colitis symptoms and 5 for other indications, including anemia, melena, or removal of a mass or polyp. Patients who required infliximab as part of their initial treatment for ICPI-related colitis were more likely to require repeat endoscopic evaluation for colitis-related indications than were those who were initially treated with a steroid only (P = 0.032).

Of the 13 patients who underwent repeat endoscopy, persistent gross inflammation was found in 8 (62%), with histology revealing acute inflammation in 3 patients and chronic inflammation in 5 patients. Among the 5 patients with initial acute inflammation who underwent repeat endoscopy, the inflammation had become chronic in 4, and 1 patient had persistent acute inflammation (Table 4). Inflammation found on repeat endoscopic and histologic examination was not associated with clinical recurrence of diarrhea. In fact, 4 of 6 patients with no clinical recurrence of diarrhea were found to have gross inflammation on repeat endoscopy.

Long-term survival

Patients with extensive colonic inflammation involving both the left and right colon and/or both the colon and ileum had significantly longer OS durations than did patients with inflammation limited to the left colon (P = 0.010; Fig. 3A). These survival differences cannot be explained by disease stage, as OS did not differ by this variable (Fig. 3B). Patients with ulcers on colonoscopy, recurrent diarrhea, higher-grade colitis, and acute inflammation on histologic examination trended towards better survival (P > 0.05; Figs. 3C-3F).

FIGURE 3.

Kaplan-Meier curves comparing overall survival by (A) distribution of inflammation on endoscopy; (B) cancer stage; (C) presence of mucosal ulceration on endoscopy; (D) diarrhea recurrence; (E) colitis grade; and (F) type of inflammation on histology.

Association of clinical grade with inflammation findings

Of the 21 patients with grade 1 colitis, 14 (67%) had inflammation on endoscopy, and 19 (90%) had inflammation on histologic examination (Table 5). Of the 7 patients with grade 1 diarrhea, 4 (57%) showed inflammation on endoscopy, and 6 (86%) had histologic inflammation. A finding of endoscopic inflammation (29/32 patients, 91%) was significantly associated with higher-grade (grades 2–3) colitis (P = 0.039; Table 5), and most (39/46; 85%) patients with endoscopic inflammation reported grade 2 or higher diarrhea (P = 0.114). We found no significant associations between colitis grade or diarrhea grade and the presence of histologic inflammation (Table 5).

Table 5:

Association of Colitis Grade with Endoscopic and Histologic Findings

	Colitis Grade 1 No. = 21	Colitis Grade 2–3 No. = 32	P value
Endoscopic findings, no. (%)			0.039
Inflammation	14 (67)	29 (91)
Normal	7 (33)	3 (9)
Histology features, no. (%)			1.000
Inflammation	19 (90)	29 (91)
Normal	2 (10)	3 (9)

Sensitivity and specificity of CT imaging for identifying colitis

Of 36 patients with inflammation on endoscopic imaging, 17 (47%) had normal CT findings, whereas 2 of 21 (10%) patients whose CT findings indicated colitis had normal endoscopic findings. The sensitivity of CT for identifying colitis was, therefore, 53%, and the specificity was 78%.

Association and correlation of diarrhea and colitis grades

Among the 53 patients included in the study, no significant association was found between diarrhea grade and colitis grade (P = 0.18; Rho = 0.19; Fig. 4).

FIGURE 4.

Scatter plot of colitis grade and diarrhea grade. Spearman correlation Rho = 0.19 (P = 0.18).

DISCUSSION

This is by far the largest scale single center study to characterize the endoscopic and histologic features of ICPI-related GI toxicities with long-term follow-up. In this retrospective study, we found that colonic ulceration is a distinguishing feature associated with a steroid-refractory disease course and that ulceration is likely to persist long term even when initial treatment is successful. However, more extensive and higher-grade colitis, the presence of ulcers on endoscopy, a finding of acute histologic inflammation, and development of recurrent diarrhea were likely associated with better OS. Our comparison of different evaluation tools for ICPI-related GI toxicity demonstrated that the grades of diarrhea and colitis were not correlated or interchangeable. There was also poor association between grades of colitis/diarrhea and inflammation observed on radiologic and histologic studies. Endoscopic inflammation, in contrast, was associated with higher-grade of colitis, but not with higher-grade diarrhea.

Our results demonstrated that clinical symptoms did not always correspond clearly to endoscopic inflammation. The distribution of colonic inflammation varied from the small bowel to the colon, with a range of endoscopic patterns. Most of the extensive colitis cases were identified by full colonoscopy, whereas most of the left-sided colitis cases were found by flexible sigmoidoscopy. However, this means that the extent of colitis could have been underestimated in some patients who were diagnosed with left-sided colitis because they were evaluated with flexible sigmoidoscopy only. Endoscopic inflammation—ulcers and nonulcerative inflammation—was found in over 80% of the study patients, even though only 60% met the CTCAE criteria for grade 2–3 colitis. On the other hand, the presence of ulcers on endoscopic examination was associated with higher diarrhea grade and a higher rate of infliximab requirement, which was consistent with a previous case study.¹⁴ The inconsistencies between the endoscopy reports and clinical grading raise the concern that the use of a solely clinical symptom-based grading system may provide an incomplete evaluation of the overall presentation and severity of ICPI-related GI toxicities.

On histologic examination of biopsy specimens, 23% of the patients in this study had acute inflammation similar in character to that reported in previous studies,^{15, 16} with neutrophilic infiltrate, cryptitis, apoptosis, and/or crypt abscesses. However, we also observed a high percentage of cases exhibiting signs of chronic inflammation, including basal lymphocytic infiltrate, crypt architecture distortion, and Paneth cell metaplasia. Surprisingly, 3 patients who met the CTCAE criteria for grade 2–3 colitis and had inflammation visible on endoscopy had normal histology. Biopsy sampling error may explain these conflicting results.

The pathogenesis of the different histologic patterns is not quite clear from our data. It appears that both acute and chronic inflammation can present at the time of colitis diagnosis; however, the exact onset time of histologic inflammation before the clinical diagnosis is difficult to determine. Therefore, whether the chronic inflammation we found represented a later stage of initial acute inflammation is unknown. Moreover, the histologic features of the chronic inflammation we observed completely overlap with those of inflammatory bowel disease (IBD), leading us to ask whether this specific histologic subtype could be an early phase of IBD. These are areas that require further clarification in the future.

Long-term follow-up data on ICPI-related colitis treatment are limited, although most studies show complete or near-complete recovery.^{10, 17–19} In our cohort, with a follow-up duration of up to 46.4 months, recurrent diarrhea developed in over one-third of patients, including more than half of those who initially received infliximab. The recurrence of diarrhea was not associated with any specific immunotherapy type, diarrhea and colitis grade, or initial endoscopic or histologic findings. Interestingly, active inflammation was found on some repeat endoscopies (obtained for any indications) in patients with and without recurrent diarrhea. Indeed, repeat endoscopy revealed persistent endoscopic and histologic inflammation in a majority of patients, regardless of whether they had clinical recurrence of diarrhea or colitis. We could not tell whether inadequate follow-up duration in these specific patients could explain the lack of diarrhea recurrence. At this writing, there is no clearly defined optimal follow-up duration based on published studies that can guarantee no residual effect in the immune system after ICPI cessation. On histologic follow-up, 4 of 5 patients with acute inflammation on the initial histologic evaluation showed signs of a transition to a chronic inflammatory pattern. The initial chronic inflammatory subtype could persist, switch to a different chronic subtype (eg, lymphocytic colitis), or recover completely and be followed by a new onset of acute inflammation. Moreover, the progression from an acute to a chronic inflammation pattern on long-term follow-up again raises concerns about the development of late-onset IBD secondary to ICPI treatment. More clinical data are needed to further evaluate this group of patients to characterize the disease behavior.

Our survival analyses suggested that patients with more severe GI toxicities, represented by more extensive colitis, recurrent diarrhea, mucosal ulcerations on endoscopy, and higher-grade colitis, tended to have better OS than did those with less severe GI effects. Although the small number of patients in each subgroup made our analysis significantly underpowered, we could still clearly see this trend. The chronic inflammatory pattern could represent a subtype characterized by prolonged damage from the autoimmune insult, which may lead to worse survival outcomes. However, again, these differences did not reach statistical significance and require further study in larger cohorts.

Currently, the main evaluation tool for ICPI-related GI toxicity is the CTCAE criteria version 4.03 for diarrhea and colitis, a grading system relying solely on clinical symptoms and signs. In general clinical practice, radiologic and endoscopic evaluations are performed on a case-by-case basis to obtain more information that can guide treatment. However, these evaluation tools are not included in the CTCAE criteria for diagnosis and severity assessment. In our study, 85% of patients underwent CT scans for their GI symptoms. Surprisingly, the sensitivity of CT diagnosis of colitis was only 53%, with a specificity of 78%, using endoscopic inflammation as the gold standard. With its high false-negative rate and low sensitivity, radiologic evaluation may not be an ideal initial screening tool for colitis workup. We found a strong association between higher colitis grade and the presence of endoscopic inflammation, but not histologic inflammation. Even in patients with grade 1 colitis (no clinical signs of colitis), there were a few patients with endoscopy-proven inflammation. In IBD, we commonly find that the onset of endoscopic and histologic inflammation occurs before the onset of clinical symptoms. A similar pattern also appears to be the case in ICPI-related colonic inflammation. Even in the absence of clinical signs of colitis (grade 1 colitis), more than 90% of patients had already developed histologic inflammation, and two-thirds had developed endoscopic inflammation. Once the clinical signs and symptoms of colitis manifested (grade ≥ 2), the endoscopic and histologic inflammation were already well established. A similar pattern was observed with regard to diarrhea grade. Again, more than 85% of patients with only grade 1 diarrhea had proven histologic inflammation, and more than half had endoscopic inflammation—a similar prevalence as that found among patients with higher-grade diarrhea. We suspect that the inflammatory process at the tissue and cellular level long precedes the onset of clinical symptoms of higher-grade diarrhea. To be able to obtain more accurate assessment of ICPI-related GI toxicity and make more informed decisions on its treatment, detailed endoscopic and histologic information is critical.

The CTCAE criteria for diarrhea and colitis have been used interchangeably in much of the published literature, creating some confusion when discussing ICPI-related GI toxicities. It is difficult to completely separate these 2 conditions, as they frequently overlap with each other. Therefore, most management recommendations are based on a combination of the diarrhea and colitis grades, which in many cases are not the same. Therefore, in this study, we examined whether diarrhea and colitis grades were correlated to determine whether they could be combined to represent distinct risk groups. Our findings argue against using a combination of these 2 evaluation criteria to determine the management strategy. Nonetheless, it remains difficult to completely separate the overlapping GI presentations of diarrhea and colitis in the grading system, as they coexist in many cases and may represent different levels of risk for complications. A comprehensive grading system is needed to incorporate the clinical signs and symptoms and the endoscopic and histologic components, similar to systems that are well established in the IBD field (eg, the Mayo Clinic Ulcerative Colitis Scoring System).

It should be noted that there were limitations to this study. Firstly, given the nature of retrospective study, the detailed information regarding overall accumulated steroid dosage used was not very accurately recorded. As an alternative, we used steroid duration for our analysis that may not be the best measurement for evaluation of its risk for complication. Secondly, the endoscopic, pathologic, and radiological data were all based on the original reports in patients’ chart. There could be interobserver variance that affected our results. Thirdly, as no standard practice guidelines for the management of this condition were issued by any oncology societies over the past 5 years, the study patients were mostly cared for on the basis of the experience and comfort level of the individual providers and institution. Likely owing to the conservative practice patterns at our institution and limited resources, only a small number of patients with ICPI-related GI toxicities received endoscopic and radiologic evaluation at the time of diagnosis, and long-term follow-up was not routinely performed after initial GI toxicity treatment. In addition, some patients with normal endoscopic findings did not undergo routine biopsies, given our limited knowledge about this new medical challenge. Although we were able to collect the largest sample size to date despite these limitations, our subgroup analyses were still underpowered.

CONCLUSIONS

In conclusion, this is by far the largest scale study focused on endoscopic and histologic descriptions of ICPI-related colitis with long-term follow-up. Our study showed that the presence of ulcers on endoscopic examination is a predictor of a steroid-refractory disease course, which most likely requires stronger immunosuppressive agents early on. Patients with more severe toxicity (ie, more extensive endoscopic inflammation or ulcers, higher-grade colitis, recurrent diarrhea) tend to have improved OS. Patients with chronic histologic inflammation tend to have worse long-term survival outcomes than do those with acute inflammation. In addition, widely used evaluation tools, including CTCAE criteria, radiologic imaging, endoscopy, and histology, have poor correlation with each other. A more accurate and comprehensive grading system for ICPI-related GI toxicity is, therefore, desirable to avoid basing treatment strategies on clinical symptoms alone. Future studies with larger sample sizes and longer follow-up duration are warranted for better clarification of this critical and challenging condition.

Conflicts of Interest: No conflicts of interest declared from all the authors in this manuscript.

Supported by: This study was not supported by any funding.

ACKNOWLEDGMENTS

Medical writing support for development of this paper was provided by Department of Scientific Publications at MD Anderson Cancer Center.

Author notes

Co-senior authors