Patients were identified at two participating institutions: Massachusetts General Hospital (MGH; n=113) and Vanderbilt University Medical Center (n=17). All patients had advanced malignant melanoma and developed biopsy-proven ICI colitis after receiving at least one dose of an ICI between 2011 and 2017. ICI regimens included single agent PD-1 inhibition with pembrolizumab or nivolumab (n=12), single agent CTLA-4 inhibition with ipilimumab (n=83) or combination therapy with ipilimumab and nivolumab (n=35). Patients with GI infection at time of onset were excluded.

Medical records were retrospectively reviewed, and data were extracted. Baseline demographic data and detailed clinical treatment histories were collected for each patient. Positive smoking status was defined as current or former smoker with more than a 10 pack-year history. Cases of colitis were graded based on the CTCAE V.4.0.21 Cases were graded solely on the frequency of loose stool since other ICI-induced colitis symptoms (abdominal pain, presence of blood and/or mucus in stool) were infrequent and not consistently documented. Additional information pertaining to colitis characteristics and management were extracted including time of onset, diagnostic evaluation performed (endoscopy, biopsy), management (dose/duration of steroids, use of infliximab) time to resolution (grade ≤1) and rates of flare. Flare was defined by need to resume or escalate immunosuppressant regimen. Steroid duration was defined as time between initial administration and date of taper to ≤10 mg of prednisone (or its equivalent). Cancer-related data such as tumor stage, prior treatments, and ICI type/duration were recorded. Objective responses were determined using the Response Evaluation Criteria in Solid Tumors (RECIST) V.1.1.22

In patients with available endoscopic data, observational data were collected. In a subset of patients with endoscopic and pathologic data, endoscopic data were scored using the Mayo Endoscopic Score (MES). The MES is part of a clinical system devised at the Mayo Clinic, Rochester, Minnesota, that is used to quantify the degree of inflammation in the GI tract for patients with ulcerative colitis.23 The score ranges from 0 to 3: 0 indicates no features of macroscopic inflammation; 1 indicates mild inflammation, characterized by mild friability, erythema, and decreased vascularity; 2 indicates moderate inflammation, characterized by friability, marked erythema, absent vascular patterns, and presence of erosions; and 3 indicates severe inflammation, in which ulcerations and spontaneous bleeding are present. Scores were formulated through review of archived images and endoscopic reports by one gastroenterologist (MD).

To evaluate the impact of infliximab on antitumor outcomes, progression-free survival (PFS) and overall survival (OS) were evaluated. This analysis excluded patients receiving adjuvant ipilimumab and patients receiving ICI before January 1, 2015, a date signifying a shift in our institutional practice pattern to employ infliximab early in the course of colitis, specifically in cases refractory to steroids after 48–72 hours. All data were updated in November 2018.

Patient characteristics were summarized using proportions for categorical variables and medians with range for continuous data. Comparisons of ulceration rates and steroid use according to MES s was based on Fisher’s exact tests. PFS was measured from time of ICI initiation to clinical/radiographic progression or death. OS was measured from time of ICI initiation to death. The follow-up of patients who neither died nor progressed was censored at the last follow-up date. Multivariable Cox proportional hazards regression models were implemented to evaluate the impact of important baseline patient and disease information on time-to-event outcomes. To address the potential for guarantee-time bias in the analyses of the effect of infliximab versus steroids only on PFS or OS, two approaches were used: an extended Cox model with infliximab initiation as a time-dependent covariate and an 11-week conditional landmark approach. Eleven weeks was chosen as the landmark time because the maximum time recorded to commence infliximab was 10.6 weeks after colitis onset. Patients who were alive and progression free 11 weeks after the onset of colitis were followed forward in time and compared according to infliximab use or steroids only. The distributions of all time-to-event outcomes were estimated using Kaplan-Meier methods and compared between groups using stratified log-rank tests. Analyses were performed using R (V.3.5.3) and SAS (V.9.4).

We identified 130 patients with ICI-induced colitis across two institutions. Baseline characteristics of these patients are summarized in table 1. Most patients were male (63%), treatment-naïve (66%), BRAF wild-type (70%), non-smokers (55%) with median body mass index (BMI) of 28 (range 14–53). The median age at ICI initiation was 65 years (range 22–95). Underlying autoimmune conditions were present in a subset of patients (18%), with most common conditions involving endocrine organs and rheumatologic disease, although there were four patients with underlying GI disease (three cases of ulcerative colitis, one celiac disease).

Ten patients received adjuvant ipilimumab (10 mg/kg) for high-risk, resected stage III/IV disease, while 120 received ICI in treatment for unresectable, metastatic disease (table 2). Most patients (64%) received anti-CTLA-4 monotherapy (adjuvant therapy or in treatment of metastatic disease), followed by 27% treated with combination therapy—including patients treated with sequential therapy (anti-CTLA-4 followed by anti-PD-1 started within 6 weeks or vice versa)—and 9% with single agent PD-1 inhibition. The median follow-up duration since initial dose of ICI was 185 weeks (42.5 months) (95% CI 159 to 229) based on inverse Kaplan-Meier estimation.

Colitis developed after a median of 6 weeks on therapy (range 0.1–145), with 86% of patients experiencing a grade ≥2 toxicity. Grade 3–4 colitis occurred in 12 cases (9%) with no grade 5 events. The primary presenting complaint was loose, frequent stool with 17 cases of documented hematochezia. Colitis onset was defined as the date symptoms developed, not the date medical evaluation for symptoms occurred. Endoscopic assessment with biopsy was performed in 123 cases (95%) with companion upper endoscopy performed in 17 cases (13%). Most patients required inpatient admission for management, with 81% of patients (n=106) hospitalized at least once and 28% (n=37) requiring two or more admissions for management of symptoms. Five patients developed a perforation of the colon, for which they underwent surgery.

All patients were treated with corticosteroids with a median time between clinic-documented symptoms and start of steroids of 3 days (range 0–34 days). Eighty patients (68%) were initially treated with high-dose intravenous corticosteroids (median dose 1 mg/kg) and 92 (71%) received intravenous steroids at some point during their treatment course. Patients required steroids for a median of 8.3 weeks (range 1–146 weeks). Steroid duration was not associated with the grade of diarrhea nor the presence or absence of hematochezia.

Additional and/or alternative immunosuppression was employed in 59 cases. In three cases of microscopic colitis, budesonide alone was an effective monotherapy.13 Four patients received adjunctive budesonide when colitis flared on their high-dose steroid taper, while two patients benefited from the addition of mesalamine and one from budesonide suppositories. Fifty-two patients (42%) required at least one infusion of infliximab 5 mg/kg. In our cohort, 27 of the 52 patients received their first dose of infliximab as an inpatient; 19 of which were treated prior to January 2015 and only 8 after January 2015. The median number of infliximab doses was one (average 1.7), with a range of 1–6. Twenty-three patients received ≥2 doses of infliximab; 13 received two doses; 7 received three doses; and 3 received >3 doses. Sixty-three per cent (15/23) of the patients who received two or more doses were treated after January 1, 2015. The majority of patients that required infliximab received treatment with CTLA-4 antagonism with or without PD-1 blockade (64% and 14%, respectively). There was no statistically significant association between initial grade of colitis or associated hematochezia and infliximab administration. All patients received infliximab while on a steroid taper to control persistent and/or progressive symptoms. Infliximab was administered in twenty-seven patients (27/52;52%) when steroids were unable to be tapered below 1 mg/kg of prednisone (or its equivalent), whereas 25 patients (25/52;48%) required infliximab only when colitis symptoms flared after successfully beginning a taper. The median duration from dose one of infliximab to resolution of colitis symptoms was 2.7 weeks (0.14–16.4 weeks). Receipt of infliximab was associated with a longer duration of steroid use with a median duration of 10.1 weeks (range 3.4–146) compared with the non-infliximab cohort of 6.8 weeks (range 0.4–55.7).

One hundred twenty-three (95%) patients had resolution of colitis to grade ≤1. The median time to resolution from onset of colitis to grade 1 symptom in the entire cohort was 4.4 weeks (95% CI 3.7 to 5.7 weeks). In the cases where symptoms did not resolve to grade ≤1 (n=7); two patients were lost to follow-up, four patients died of progressive disease and one patient died of ICI-induced neurotoxicity prior to colitis resolving.

Sixty patients (46%) experienced persistent symptoms or a flare of their colitis symptoms during or after the completion of their steroid taper. Most patients (60%) had received single agent ipilimumab (adjuvant therapy or front line), followed by 27% who received front-line combination anti-CTLA-4/PD-1 therapy and 8% front-line PD-1 inhibition. The remaining 5% had received second-line PD-1 inhibition (2%) or second-line ipilimumab (3%)

Side effect rates from immunosuppression were low, with 15% (n=20) of patients experiencing a definite complication related to steroid and/or infliximab use, and three additional patients experiencing an adverse event likely due to immunosuppression (figure 1). Definitely related complications (aside from transiently elevated glucose) included secondary adrenal insufficiency (n=6), Cushing syndrome (n=2), steroid induced mania (n=1) and infection (n=11). Infectious complications included symptomatic oral candidiasis (n=5), pneumocystis jiroveci pneumonia (n=2) bacteremia (n=2), herpes simplex virus reactivation (n=1), and tuberculosis reactivation, (n=1). Cases in which immunosuppression was likely to be a contributing factor included one patient with biopsy-confirmed Steven-Johnson’s Syndrome (SJS) due to Bactrim prophylaxis while on prolonged steroids and in two cases where steroid use was thought to play a role in colonic perforation. There were no grade 5 events attributable to immunosuppression.

Endoscopic assessment with biopsy was performed in 123 cases of suspected colitis (95%). Endoscopic features of ICI-induced colitis ranged from normal appearing mucosa to congestion with erythema to frank ulceration. In the MGH cohort (n=113), 17 patients (15%) had a companion upper endoscopy.24 Reasons for concomitant esophagogastroduodenoscopy (EGD) included nausea (n=6), dysphagia/dyspepsia (n=4), melena (n=2), and anorexia (n=1). Four patients underwent EGD in initial work-up of diarrhea with no documented upper GI complaint. In 11 cases, a GI tract melanoma was identified. Pathologic assessment of biopsies taken from the stomach and duodenum confirmed three cases of concomitant erosive gastritis and eight cases of duodenitis, respectively.

To accomplish the aims outlined in this study, several subgroup analyses were performed (figure 2) both to evaluate the prognostic role of endoscopy and to assess the impact of infliximab in the management of the toxicity, particularly the effect of TNF-α inhibition on colitis and cancer outcomes.

A subset of patients (n=44) had sufficient endoscopic and pathologic data available, including high-resolution endoscopic images, to enable an independent review of the endoscopic diagnosis. A comprehensive visual and histopathologic assessment was performed and MES grading was tabulated with scores ranging from 0 (no visible inflammation) to 3 (colonic ulceration). Presence of ulceration (MES 3) was associated with use of infliximab (p=0.008), consistent with previously published data (figure 3A).14 15

We next evaluated the impact of infliximab on outcome. We performed a subgroup analysis on the impact of infliximab on cancer outcome. After excluding patients receiving adjuvant ipilimumab and patients receiving ICI before January 1, 2015—a date signifying a shift in our institutional practice pattern to employ infliximab early in the course of colitis, specifically in cases refractory to steroids after 48–72 hours—42 patients remained for analysis. We evaluated the impact of infliximab, comparing the 18 of whom had received infliximab and 24 who had not. After adjusting for baseline covariates, there was no significant difference in steroid duration or cancer-related outcomes. Neither the timing of infliximab administration nor number of doses impacted PFS or OS. We then further stratified outcome by timing of infliximab with early defined as <4 weeks after symptom onset and late, >4 weeks after onset. There was no significant difference in steroid duration or cancer-related outcomes of PFS or OS in the early versus late subgroup. In an additional landmark analysis, we examined a subset of 82 patients who were alive and progression free at 11 weeks to assess for differences in survival according to treatment with steroids alone versus infliximab. There were 47 patients who received steroids alone and 35 who received steroids and infliximab. There was no statistically significant difference in OS when comparing these cohorts (figure 5B).