Approval for this retrospective study was obtained from the institutional review board at The University of Texas MD Anderson Cancer Center. We included adult patients who developed IMC and received SIT in 2018. We excluded patients with colitis due to other etiologies, including infectious colitis, graft-versus-host disease, and neutropenic colitis. Institutional pharmacy, endoscopy, oncology, and gastroenterology databases were searched to identify eligible patients. Thereafter, a comprehensive chart review was conducted to extract variables of interest.

Data pertaining to IMC were date of onset, time from ICI initiation to onset, duration of symptoms, presenting and peak grades of diarrhea and colitis, treatment, and outcomes. IMC was graded using the Common Terminology Criteria for Adverse Events (CTCAE) version 5.0 at different times to assess response of IMC to treatment [16]. Onset of IMC was categorized into abrupt (peak IMC grade was recorded at first presentation) or gradual (peak IMC grade was recorded after the first presentation). Improvement of symptoms was defined as significant lessening of clinical symptoms of at least one CTCAE grade, as reported by the patient and treating physician. Rapid improvement was defined as the improvement’s occurring within 72 h of IMC treatment (as recommended by the current guidelines).

Collected information included patients’ demographics; clinical and oncologic history; and IMC treatment, clinical presentation, and outcomes. Demographics included age at the time of first ICI infusion, gender, and race/ethnicity. Comorbidities according to the Charlson Comorbidities Index were reported [17]. Cancer type, stage, and treatment were documented as well. Type of ICI, number of infusions, duration of treatment, and non-gastrointestinal irAEs were collected. Also, we reported if ICI therapy was resumed after the first IMC episode. Cancer status at the time of staging and before and after SIT treatment was recorded according to Immune-Modified Response Evaluation Criteria In Solid Tumors (imRECIST) and Immune Response Evaluation Criteria In Solid Tumors (iRECIST) [18, 19].

Treatment of IMC consisted of steroids and a biologic agent (infliximab or vedolizumab or both). Initial dosage of steroids was given following the recommendations from treatment guidelines for irAEs depending on the grade and responsiveness of IMC [1, 2]. Briefly, 1 mg/kg/day is the recommended initial dose of steroid for grade 2 IMC, and 1–2 mg/kg/day is the recommended initial dose of steroid for grade 3–4 IMC. Steroids were administered orally in grade 2 IMC and intravenously in grade 3–4 IMC. Cumulative duration of steroid treatment was measured, but cumulative dose was not because the dosage varies widely and is impractical to calculate. The number of SIT infusions was measured for both infliximab and vedolizumab. In this study, gradual weaning of steroids was started after the first infusion of SIT. The requirement for hospitalization and intensive care unit admissions due to IMC was documented. Additionally, we reported the number of hospital admissions for IMC and the cumulative duration of associated hospitalizations. Steroid tapering was defined as "failed" if the symptoms recurred after decreasing the dose of steroids from the maximum dose. We also recorded the number of steroid tapering attempts before successful cessation of steroids. Time from IMC onset to SIT initiation was noted as well. Recurrence of IMC after complete discontinuation of immunosuppressive therapy was recorded. Survival duration was defined as the time from ICI initiation to last clinical encounter or death.

Fecal calprotectin and lactoferrin values before and after immunosuppressive therapy were recorded. Endoscopic and histologic features at the time of IMC presentation were documented. Endoscopic high-risk features of IMC were defined according to our previous study [7]. Briefly, these high-risk features were one or more of the following: ulcers deeper than 2 mm, ulcers wider than 1 cm, and extensive inflammation involving the colon proximal to the splenic flexure [7]. Repeated endoscopic and histologic evaluations to assess for treatment response were reviewed. Endoscopic remission was defined as the healing of the mucosal ulceration or the resolution of mucosal inflammation. Histologic remission was defined as the absence of active inflammatory features on the repeat mucosal biopsies.

Continuous variables were described by mean and standard deviation (SD) or median and interquartile range (IQR). Categorical variables were described by frequencies and percentages. Fisher exact and χ² tests were used to compare categorical variables. The Wilcoxon rank sum test was used to compare continuous variables. Factors associated with IMC recurrence or steroid taper failure were assessed by univariate logistic regression analyses. To estimate and compare survival durations between subgroups, we used Kaplan-Meier curves and log rank tests. All statistical tests were two-sided. P values of 0.05 or less were considered statistically significant. Statistical analyses were computed by SAS version 9.4 software and SPSS version 24.0 software.

During the study period, 1459 patients received ICI; 179 (12%) of them developed IMC of any grade. Of these 179 patients, 84 (47%) received SIT. The mean age of these 84 patients was 60 years (SD, 14 years). The predominant gender was male (n = 66 [79%]), and the predominant ethnicity was Caucasian (n = 76 [91%]). Melanoma was the most common malignancy (n = 40 [48%]), followed by genitourinary cancer (n = 28 [33%]) and thoracic, head, and neck cancers (11 [13%]). Most patients had stage IV malignancy (n = 72 [87%]). ICI therapy consisted of inhibitors of programmed death protein-1 or its ligand in 33 patients (39%), inhibitors of cytotoxic T-lymphocyte associated protein-4 in 21 patients (25%), and a combination of ICIs in 30 patients (36%). The median clinical follow-up duration of this study was 5 months (range, 1–10 months), and the median endoscopic follow-up time was 4 months (range, 1–7 months).

According to the peak grade of diarrhea, 57 (68%) patients had grade 3 and 13 (16%) had grade 4. Regarding colitis grade, 55% of patients had grade 2 and 45% of patients had grade 3–4. Laboratory assessment of fecal lactoferrin was performed in 56 patients, 53 (95%) of them had positive test. Fecal calprotectin value was obtained in 40 patients with a median of 235 (IQR, 91–479). Of the 64 patients who had endoscopic evaluation with biopsy, 40 (63%) had high-risk features of IMC on endoscopy and 60 (94%) had features of active inflammation on histology. SIT treatment consisted of infliximab in 50 patients (60%) and vedolizumab in 34 patients (40%, Fig. 1).Fig1Fig. 1

Patient characteristics by SIT

The median time from IMC onset to SIT initiation was 10 days (IQR, 5–23 days). Patients who received SIT within 10 days of IMC onset (n = 44 [52%]) required fewer hospitalizations (P = 0.026), had shorter duration of symptoms (P = 0.002), were less likely to experience steroid taper failure after SIT therapy (P = 0.033), and consequently had fewer failed steroid tapering attempts (P < 0.001) compared with patients who received SIT after 10 days of IMC onset (40 [48%]; Table 1). The two groups had similar severity of diarrhea and colitis, endoscopic and histologic features, and ICI type.Tab1

Clinical characteristics stratified by the timing of selective immunosuppressive therapy initiation (SIT)

Steroids were weaned successfully without return of IMC symptoms after SIT therapy in 50 (60%) patients. Compared with these patients, patients in whom steroid tapering failed after SIT had higher peak grades of IMC (P = 0.016), were more likely to have a positive fecal lactoferrin test at time of IMC onset (P = 0.028), were more likely to present with mucosal ulceration (P = 0.015), had longer duration of IMC symptoms (P < 0.001), received longer duration of steroid treatment (P < 0.001), had slower improvement in response to steroid treatment before SIT (P = 0.009), were more likely to experience failed steroid therapy before SIT (P = 0.002), initiated SIT later (P = 0.002), were hospitalized for longer durations (P = 0.001), and were more likely to require multiple hospitalizations (P < 0.001; Table 2).Tab2

Clinical characteristics stratified by the resistance to steroid

Most patients received one or two infusions of SIT (n = 54 [64%]); the median number of infusions was 3 (IQR, 3–4). Compared with patients who received three or more SIT infusions, patients who received one or two SIT infusions were more likely to experience failure of steroid tapering and recurrent IMC (P = 0.030 and P = 0.008, respectively; Table 3). Additionally, patients who received one to two SIT infusions commonly received infliximab rather than vedolizumab (P < 0.001) and had higher levels of fecal calprotectin after SIT (P = 0.011). However, the peak CTCAE grades of diarrhea and colitis were comparable between the two groups.Tab3

IMC outcomes stratified by the number of SIT doses

Of the 79 patients who received steroids before SIT, 38 had improvement of IMC symptoms and 41 had no improvement. Compared to patients who had improvement of IMC symptoms with corticosteroid treatment before SIT, patients whose symptoms did not improve on steroids received steroids for a longer duration (P < 0.001), were more likely to experience steroid taper failure after SIT (P = 0.010), and consequently required more steroid tapering attempts (P = 0.014) and were more likely to be hospitalized multiple times (P = 0.051) and for longer durations (P = 0.001, Table 4). This difference in clinical outcomes was seen despite the fact that the two groups had similar grades of diarrhea and colitis.Tab4

Response to steroid treatment before SIT and outcomes

Among the 41 patients who had poor response to steroids before SIT, 17 began SIT within 10 days and 24 after 10 days of IMC onset. In these patients, early introduction of SIT was associated with a lower number of hospitalizations (P = 0.014; Table 5). Among the 38 patients who had good response to steroids before SIT, 23 received SIT within 10 days and 15 more than 10 days after IMC onset. In these patients, early introduction of SIT was associated with less resistance to steroid tapering after SIT (P = 0.003), fewer failed steroid tapering attempts (P < 0.001), and shorter duration of symptoms (P = 0.016).Tab5

Effect of timing of SIT initiation after IMC onset on outcomes among patients who had different response to steroid therapy

In our cohort, 16 patients received steroids for less than 6 weeks and began SIT within 10 days of IMC onset. These patients, compared with the 30 patients who received steroids for longer than 6 weeks and began SIT more than 10 days after IMC onset, had shorter duration of symptoms (P < 0.001), had fewer steroid tapering attempts (P < 0.001), and were hospitalized less frequently (P < 0.001) and for shorter duration (P = 0.034; Additional file 1: Table S1).

Sixteen patients (19%) developed IMC recurrence after SIT therapy and complete steroid weaning. Logistic regression analysis showed that the predictors of IMC recurrence included failure of steroid weaning after SIT (P = 0.017), infliximab therapy rather than vedolizumab therapy (P = 0.017), 2 or fewer infusions of SIT compared with 3 or more infusions (P = 0.023), multiple hospitalizations (P = 0.002), persistent endoscopic (P = 0.025) and histologic (P = 0.033) inflammation after SIT, lower number of steroid tapering attempts (P = 0.001), shorter duration of steroids (P = 0.022), higher calprotectin values after SIT (P = 0.014), longer duration of symptoms (P = 0.008), and longer duration of hospitalization (P = 0.001; Table 6).Tab6

Outcomes after SIT

Fourteen patients resumed ICI after IMC resolution. Eight of them received vedolizumab concurrently with ICI infusions and six did not (Additional file 1 Figure S1). Only one patient who received concurrent vedolizumab experienced IMC recurrence, whereas three patients who did not receive vedolizumab experienced IMC recurrence. IMC recurrence necessitated immunosuppressive therapy in the one patient who received vedolizumab and in two of the patients who did not.

At the end of study period, 12 patients were dead; 11 because of cancer progression and one because of hypophysitis and sepsis. Survival analyses indicated a comparable impact of vedolizumab and infliximab on overall survival rates (P = 0.151; Additional file 1 Figure S2). Patients who had their ICI treatment course interrupted early (i.e., one or two infusions) had worse survival rates compared with patients who received ICI for more than three or more infusions (P = 0.008; Additional file 1 Figure S3).