Impact on survival of severe immuno-related adverse events in deficient mismatch repair/microsatellite instable-high digestive cancers treated with immunotherapy Authors: David Tougeron, Charles Kauffmann, Margherita Ambrosini, Olayide Boussari, Sara Lonardi, Frank A Sinicrope, Marwan Fakih, Michael J Overman, Elena Elez, Anthony Turpin, Marie Decraecker, Antoine Hollebecque, Simon Pernot, Thibault Mazard, Marie Dutherage, Romain Cohen, Francesco Sclafani, Mohamed Ben Abdelghani, Thomas Aparicio, Clélia Coutzac, Vincent Hautefeuille, Marie Muller, Chiara Cremolini, Rosine Guimbaud, Julien Taieb, Filippo Pietrantonio, Raphael Olivier, Emily Alouani Published: 2026 ABSTRACT AbstractBackgroundDeficient DNA mismatch repair/microsatellite instability-high (dMMR/MSI-H) cancers are very sensitive to immune checkpoint inhibitors (ICIs), yet their use is frequently complicated by immune-related adverse events (irAEs). Our study analyzed the impact of grade ≥3 irAEs (immune-related severe adverse events (irSAEs)) on survival in patients with dMMR/MSI-H digestive cancers treated with ICIs.MethodsWe conducted an international, multicenter, ambispective study involving 1,175 patients from 34 centers. The primary endpoint was the correlation between the occurrence of irSAEs and progression-free survival (PFS). Secondary endpoints included factors associated with irSAEs, correlation between irSAEs and overall response rate (ORR) and overall survival (OS). Prespecified landmark and time-dependent survival models were used to account for the time to irSAE occurrence.ResultsAmong 1,175 patients treated with ICIs for digestive cancers, 49.1% were female, median age was 66.9 years (IQR 53.7–77.3) and 82.9% had colorectal cancer. Overall, 382 patients (32.5%) had an irAE, including 117 (10.0%) irSAEs. Median time to irSAEs occurrence was 3.78 months (IQR 1.97–7.62). The most frequent irSAEs were gastrointestinal (3.4%), hepatic (1.8%) and dermatologic (1.0%). The only factor associated with irSAEs was the use of an ICI combination versus monotherapy (20.0% vs 11.4%, p=0.03). Among patients with irSAEs, 71.5% received oral and 29.8% intravenous corticosteroids, and 15.7% immunosuppressive agents. Overall, in 26.9% of patients, ICIs were resumed following irSAEs, with an irAE recurrence rate of 57.1%. irSAE was associated with a better ORR (48.6% vs 34.6%, p<0.0001). The multivariable Cox model revealed that irSAE as a time-dependent variable was not associated with better PFS (HR=1.280 (95% CI 0.898 to 1.824), p=0.172) or OS (HR=1.043 (95% CI 0.712 to 1.528), p=0.828). Likewise, landmark analysis at 3 months and 6 months also showed that irSAEs were not associated with better PFS or OS.ConclusionsThe occurrence of irSAE was not independently associated with better survival in patients with dMMR/MSI-H digestive cancers treated with ICIs.Keywords: Immune related adverse event - irAE, Immune Checkpoint Inhibitor, Microsatellite, Mismatch repair - MMR, Gastrointestinal Cancer WHAT IS ALREADY KNOWN ON THIS TOPIC Deficient mismatch repair (dMMR)/microsatellite instability-high (MSI-H) digestive cancers are sensitive to immune checkpoint inhibitors (ICIs), but some of these patients treated with ICIs develop immune-related severe adverse events (irSAE). It has been suggested that the occurrence of irSAE is associated with better outcomes in certain cancers, but this has not yet been evaluated in dMMR/MSI-H digestive tumors. Deficient mismatch repair (dMMR)/microsatellite instability-high (MSI-H) digestive cancers are sensitive to immune checkpoint inhibitors (ICIs), but some of these patients treated with ICIs develop immune-related severe adverse events (irSAE). It has been suggested that the occurrence of irSAE is associated with better outcomes in certain cancers, but this has not yet been evaluated in dMMR/MSI-H digestive tumors. WHAT THIS STUDY ADDS In patients treated with ICIs for digestive dMMR/MSI-H cancers, the only factor associated with irSAEs was the use of ICI combinations. Recurrence rates for irAEs following ICI re-introduction were high. Time-dependent multivariable models revealed that irSAEs were not associated with better survival. In patients treated with ICIs for digestive dMMR/MSI-H cancers, the only factor associated with irSAEs was the use of ICI combinations. Recurrence rates for irAEs following ICI re-introduction were high. Time-dependent multivariable models revealed that irSAEs were not associated with better survival. HOW THIS STUDY MIGHT AFFECT RESEARCH, PRACTICE OR POLICY The occurrence of irSAEs in patients with dMMR/MSI-H digestive cancers was not associated with better survival, and the high risk of irSAE recurrence following ICI re-introduction supports the discontinuation of ICIs after irSAEs in patients with disease control. ICI re-introduction could be considered only in cases of disease progression. The occurrence of irSAEs in patients with dMMR/MSI-H digestive cancers was not associated with better survival, and the high risk of irSAE recurrence following ICI re-introduction supports the discontinuation of ICIs after irSAEs in patients with disease control. ICI re-introduction could be considered only in cases of disease progression. Introduction Immune checkpoint inhibitors (ICIs) targeting programmed death 1 (PD-1) or programmed death ligand 1 (PD-L1) and cytotoxic T-lymphocyte antigen 4 (CTLA-4) have revolutionized the prognosis of many cancers. About 15% of early-stage colorectal cancers (CRC) have a deficient DNA mismatch repair/microsatellite instability-high (dMMR/MSI-H) phenotype, while the prevalence of this phenotype in non-colorectal digestive cancers varies from 30% in small bowel adenocarcinomas to 1% in pancreatic adenocarcinomas.1–7 dMMR/MSI-H tumors could occur either in a hereditary context (Lynch syndrome) or more frequently as sporadic cases. Determination of MMR/MSI tumor status has important prognostic and therapeutic implications and is recommended at diagnosis for all CRC, gastric and small bowel adenocarcinomas. In the metastatic setting, even though dMMR/MSI-H digestive tumors have historically been associated with a poor prognosis and chemoresistance,8–12 they are associated with frameshift mutations, which generate immunogenic neoantigens, making them highly sensitive to ICIs.13 Three randomized trials, the KEYNOTE-177, CHECKMATE 8HW and SAMCO-PRODIGE 54 have shown the superiority of ICIs over chemotherapy in metastatic dMMR/MSI-H CRC.14–16 In metastatic dMMR/MSI-H, data for non-colorectal tumors are more limited, but non-randomized studies have demonstrated the efficacy of ICIs versus chemotherapy.17 18 A recent meta-analysis comparing ICIs alone or in combination with chemotherapy versus standard of care in patients with advanced MSI-H gastric cancer suggested that ICI use procured an improvement in overall survival (OS) and progression-free survival (PFS).19 ICIs stimulate the immune system against the tumor, but can also cause immune-related adverse events (irAEs).20 21 Severe grade 3–5 irAEs occur in 10–20% of patients with ICI monotherapy and in 20–40% with ICI combinations, leading to approximately 10% of treatment discontinuations and 3% of immune-related toxic deaths.22–24 Rates of irAEs associated with ICIs are well-described, but factors associated with these irAEs are poorly studied. It is challenging to compare the frequency of irAEs according to the type of ICI or primary tumor, and predictive factors for severe irAEs and their correlations with outcomes are largely unknown.25 26 In addition, data concerning irAEs in dMMR/MSI-H gastrointestinal tumors are limited due to the rarity of these tumors, in particular the factors associated with severe grade ≥3 irAEs (immune-related severe adverse events (irSAEs)). Moreover, the impact of these irAEs on prognosis is not well-known. Several studies have shown that survival in patients with irAEs from ICIs is longer than that in patients with no irAEs, even following ICI discontinuation, especially in melanoma and non-small-cell lung cancer.27–30 By contrast, some studies found no association between irAEs and treatment efficacy.31–33 To our knowledge, there are no data concerning the association between irAEs and treatment efficacy in dMMR/MSI-H digestive cancers. Herein, we report an international multicenter ambispective study to describe irAEs in dMMR/MSI-H digestive cancers treated with ICIs and analyze the correlation between the occurrence of irSAEs and ICI efficacy. Materials and methods Population This is an international, multicenter, ambispective study involving 34 centers in France, Italy, the USA and Belgium. All consecutive patients with locally advanced or metastatic unresectable dMMR/MSI-H digestive cancer treated with immunotherapy (ICI monotherapy or in combination) regardless of the treatment line between January 2014 and December 2023 were included. The cut-off date was June 30, 2025. Inclusion criteria were age ≥18 years, MSI-H (by molecular biology using PCR and/or next-generation sequencing as per standard institutional practices) and/or dMMR (by immunohistochemistry of the four MMR proteins) digestive cancer, and treatment with ICI (monotherapy or in combination). Exclusion criteria were ICIs combined with chemotherapy or targeted therapy and discordance between MSI and MMR status (ie, proficient mismatch repair (pMMR)/MSI-H or dMMR/microsatellite stable (MSS) tumors). The study was conducted in accordance with the Declaration of Helsinki and was approved by the French regulatory requirements (Commission Nationale de l’Informatique et des Libertés, CNIL, reference 2222018 v0) and local ethics committees. Patient and tumor characteristics Data on patients (age, sex and Eastern Cooperative Oncology Group performance status (ECOG PS)) and tumors (primary tumor site, tumor node metastasis stage, tumor grade, locally advanced vs metastatic disease, sites of metastases at ICI start) were collected. BRAF and RAS mutational status were also collected for CRC. The treatment line and type of ICI (anti-PD-(L)1 monotherapy vs anti-CTLA-4-based combination), the duration of treatment and the reason for ICI discontinuation were collected. Adverse events irAEs were defined as AEs occurring after ICI start and considered by treating physicians to result from immunological dysfunction. irAEs were graded according to the National Cancer Institute Common Terminology Criteria for Adverse Events V.4.0. All AEs, principally endocrine, gastrointestinal, hepatic, dermatological, pulmonary, rheumatological and nephrological, were collected, but the AEs of interest were grade ≥3 irAEs (irSAEs), which were used to classify patients with and without irSAEs. Information on the management and outcome of irSAEs was collected, including the use of oral or intravenous corticosteroids and immunosuppressive agents (according to the American Society of Clinical Oncology (ASCO) guidelines). Data on treatment decisions, including ICI discontinuation and possible reoccurrence of irAEs after treatment re-introduction, were also collected. Outcomes and objectives Patients were divided into two groups: with irSAE (grade ≥3 irAEs) or without irSAE (no irAE or irAE grade 1 or 2). The main objective of the study was to determine whether irSAEs were associated with PFS. The secondary objectives concerned the correlation between irSAEs and overall response rate (ORR) and OS. The predictive factors for irSAEs were also determined in univariable and multivariable analyses. The ORR was defined as the percentage of patients with a complete or partial response according to Response Evaluation Criteria in Solid Tumors V.1.1. The disease control rate (DCR) was defined as the proportion of patients with a complete/partial response or stable disease as the best response. PFS was defined as the time from the start of ICI to the first documented disease progression or death from any cause, whichever occurred first. Patients alive and without progression, whether included retrospectively or prospectively, were censored at the end of the follow-up (June 30, 2025). OS was defined as the time from the start of the ICI to the date of death from any cause. Patients alive were censored at the end of the follow-up. Statistical analysis Age (continuous variable) was described according to its median and IQR. Categorical variables (ie, sex, ECOG, primary tumor site, grade, metastatic sites) were described as frequencies and percentages. Factors associated with the occurrence of irSAEs were evaluated using the Wilcoxon-Mann-Whitney test for age and using Fisher’s exact test for categorical variables. A p value of <0.05 denotes a significant association. PFS and OS were estimated using the Kaplan-Meier method and described using medians or percentages, with their 95% CIs, at specific time points. Median follow-up was estimated using the reverse Kaplan-Meier method. Associations between patients’ and tumor characteristics on the one hand and PFS and OS on the other were examined using a Cox regression model, and variables with a p value ≤0.1 were entered into a multivariable Cox regression model with HR and 95% CI. To consider the time to irSAE occurrence and immortal time bias, prespecified landmark analyses were done at different time points depending on the median time to irSAE occurrence. Also, an additional prespecified analysis using toxicity as a time-dependent variable in the multivariable Cox model was carried out. A p value of <0.05 was considered statistically significant. All statistical tests were two-sided. All analyses were done using R V.4.4.2 (R Foundation for Statistical Computing, Vienna, Austria). This observational study follows the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) guidelines. Results Characteristics of patients and tumors Overall, 1,175 patients with dMMR/MSI-H digestive cancers treated with ICIs were included. The median age was 66.9 years (IQR 53.7–77.3) and 49.1% of patients were women (table 1). The most frequent primary tumor was CRC (82.9%), followed by esophagogastric cancer (7.7%). Most tumors were metastatic at ICI initiation (93.8%). Most patients (70.5%) had previously undergone resection of the primary tumor. Most patients (85.2%) had a good performance status (ECOG PS 0 or 1) at baseline. Among colorectal adenocarcinomas, 30.2% harbored the RAS mutation (n=268/887) and 36.6% had the BRAFV600E mutation (n=327/894). Table 1Characteristics of patients in the overall population, the population with and without serious immune-related adverse eventsCharacteristicOverall populationN=1175*Population without irAE or irAE ≤grade 2N=1058*Population with irAE ≥grade 3N=117*P value†Sex (female)577 (49.1)515 (48.7)62 (53.0)0.4Age (year)0.8 Median (IQR)66.9 (53.7–77.3)66.9 (53.9–77.2)66.8 (51.8–77.5)Primary tumor site0.8 Colorectal974 (82.9)874 (82.6)100 (85.5) Esophagogastric91 (7.7)83 (7.8)8 (6.8) Pancreatic26 (2.2)23 (2.2)3 (2.6) Other84 (7.1)78 (7.4)6 (5.1)Grade0.5 Poorly differentiated211 (39.1)179 (38.1)32 (46.4) Moderately differentiated235 (43.6)208 (44.3)27 (39.1) Well differentiated93 (17.3)83 (17.7)10 (14.5) Missing data63658848Metastatic disease at ICI start1096 (93.8)987 (93.7)109 (94.0)>0.9 Missing data65*Peritoneal carcinomatosis511 (43.8)452 (43.0)59 (51.3)0.09 Missing data972Liver metastasis393 (33.7)352 (33.5)41 (35.7)0.7 Missing data972Lung metastasis213 (18.3)190 (18.1)23 (20.0)0.6 Missing data972Lymph node metastasis644 (55.2)581 (55.3)63 (54.8)>0.9 Missing data972Surgery of primary tumor823 (70.5)735 (69.9)88 (75.2)0.3 Missing data770Line of ICI treatment0.6 First-line527 (48.0)465 (47.4)62 (53.0) Second-line353 (32.1)316 (32.2)37 (31.6) Third-line140 (12.8)129 (13.1)11 (9.4) Fourth-line and more78 (7.1)71 (7.2)7 (6.0) Missing data77770ECOG PS at ICI start0.5 0421 (36.5)375 (36.1)46 (40.0) 1562 (48.7)508 (48.8)54 (47.0) 2135 (11.7)121 (11.6)14 (12.2) 3–437 (3.2)36 (3.5)1 (0.9) Missing data20182RAS mutation among CRC268 (30.2)244 (30.7)24 (26.4)0.5 Missing data87789BRAF mutation among CRC327 (36.6)289 (36.0)38 (41.8)0.3 Missing data80719Type of ICI0.3 Pembrolizumab545 (67.6)486 (68.4)59 (62.1) Nivolumab93 (11.5)84 (11.8)9 (9.5) Ipilimumab plus nivolumab26 (3.2)22 (3.1)4 (4.2) Other142 (17.6)119 (16.7)23 (24.2) Missing data36934722Number of ICI molecules0.03 ICI monotherapy706 (87.6)630 (88.6)76 (80.0) ICI combination100 (12.4)81 (11.4)19 (20.0) Missing data36934722* n (%).†Fisher’s exact test or Wilcoxon-Mann-Whitney test.CRC, colorectal cancer; ECOG PS, Eastern Cooperative Oncology Group Performance Status; ICI, immune checkpoint inhibitor; irAE, immune-related adverse events. Characteristics of patients in the overall population, the population with and without serious immune-related adverse events * n (%). †Fisher’s exact test or Wilcoxon-Mann-Whitney test. CRC, colorectal cancer; ECOG PS, Eastern Cooperative Oncology Group Performance Status; ICI, immune checkpoint inhibitor; irAE, immune-related adverse events. Nearly half of the patients received ICIs in the first-line setting (48.0%) and one-third (32.1%) in the second-line setting. The most common ICI used was pembrolizumab (67.6%) and most patients (87.6%) received ICI monotherapy. The median duration of ICI treatment was 9.7 (IQR 2.7–22.4) months and treatment was ongoing in 207 patients (17.6%) at the end of the follow-up. Overall, 82.4% patients permanently discontinued ICI treatment (n=931/1130), mainly due to disease progression (n=413/931, 44.4%), completion of treatment (n=249/931, 26.7%) or irAEs (n=102/931, 11.0%). Immune-related adverse events Overall, 382 patients (32.5%) had at least one irAEs, including 117 (10.0%) with at least one irSAEs. Median time to irSAE occurrence was 3.78 months (IQR 1.97–7.62) (figure 1). Time to serious immune-related adverse events. The most frequent irAEs of any grade were cutaneous (11.5%), dysthyroidism (10.3%) and gastrointestinal events (9.9%) (table 2). Other irAEs occurred less frequently (5% or less). irSAEs were mainly digestive, hepatic and cutaneous AEs (3.4%, 1.8% and 1.0%, respectively). Other AEs accounted for a minority (<1%) of irSAEs. Five grade 5 AEs were observed, corresponding to 4.3% of patients with irSAEs and 0.4% of the overall population. Fatal irAEs included one hepatocellular failure (related to anti-PD-1), one colitis (related to anti-PD-1 plus anti-CTLA-4), one autoimmune lung disease (related to anti-PD-1), one myocarditis (related to anti-PD-1) and one nephrotoxicity (related to anti-PD-1 plus anti-CTLA-4). Table 2Immune-related adverse eventsn, %Immune-related adverse events (n=1175)Grade 1–2Grade 3–4-5All gradesPatient with at least one irAE265 (22.6%)117 (10.0%)382 (32.5%)Dermatological135 (11.5%)12 (1%)147 (12.5%)Digestive116 (9.9%)40 (3.4%)156 (13.2%)Dysthyroidism121 (10.3%)2 (0.2%)123 (10.5%)Rheumatological54 (4.6%)9 (0.8%)63 (5.4%)Hepatic36 (3.1%)21 (1.8%)57 (4.9%)Pulmonary17 (1.4%)7 (0.6%)24 (2%)Adrenal insufficiency20 (1.8%)4 (0.3%)24 (2%)Hypophysitis10 (0.9%)8 (0.7%)18 (1.5%)Pancreatic9 (0.8%)4 (0.3%)13 (1.1%)Nephrological9 (0.8%)3 (0.3%)12 (1.0%)Muscular6 (0.5%)2 (0.2%)8 (0.7%)Hematological3 (0.3%)4 (0.3%)7 (0.6%)Infusion-related reaction2 (0.2%)1 (0.1%)3 (0.3%)Ophthalmologic3 (0.3%)03 (0.3%)Cardiac1 (0.1%)2 (0.2%)3 (0.3%)Neurological2 (0.2%)1 (0.1%)2 (0.2%)Diabetes mellitus02 (0.2%)2 (0.2%)irAE, immune-related adverse events. Immune-related adverse events irAE, immune-related adverse events. The only factor significantly associated with the occurrence of irSAEs was the use of an ICI combination (20.0% vs 11.4% with anti-PD-(L)1 monotherapy, p=0.03) (table 1). The primary tumor site, ECOG PS, age and line of treatment were not significantly associated with the occurrence of irSAEs. Management of immune-related adverse events A total of 121 irSAEs were recorded in 117 patients, including 4 patients with two events. Regarding the management of irSAEs, 71.5% of patients received oral corticosteroids (n=60/84), 29.8% intravenous corticosteroids (n=25/84) and 15.7% immunosuppressive therapy, including antitumor necrosis factor (n=13/83). The irSAEs resolved (grade ≤1) in most cases (86.2%). Among the 12 patients whose irSAE did not resolve, 5 died, 3 were still being treated for their irSAE, 3 had their irSAE controlled by other treatments (1 diabetes mellitus treated with insulin therapy and 2 hypophysitis managed with hydrocortisone) and 1 was still receiving ICI (grade 3 lipase increase). After the first irSAEs, 21 patients (26.9%, n=21/78) resumed ICIs and 12 patients (57.1%, n=12/21) experienced recurrence of the irAE. Outcomes according to immune-related adverse events In the overall population, median follow-up was 34.1 months (IQR 18.91–55.29). The ORR and DCR rates were 36.1% (n=357/987) and 78.5% (n=775/987), respectively. ORR was significantly higher in patients with irSAEs (48.6% for patients with irSAE vs 34.6% for patients without irSAE (p=0.005)) as was DCR (88.0% vs 77.3%, p=0.014) (table 3). When irSAEs were treated as a time-fixed covariate, PFS was significantly longer in patients with irSAEs than in those without irSAE (median PFS: not reached vs 43.3 months, 95% CI 34.2 to 54.1 and HR=0.594, 95% CI 0.424 to 0.834, p=0.003) (online supplemental figure 1 and table 1). PFS at 3 years was 66.1% in patients with irSAE versus 52.6% in patients without irSAE. However, when irSAEs were modeled as a time-dependent variable in multivariable Cox analysis, no association with improved PFS was observed (HR=1.280, 95% CI 0.898 to 1.824, p=0.172) (table 4). Landmark analyses were done at 3 and 6 months since the median time to irSAE onset was 3.78 months. Similarly, landmark analysis at 3 months (HR=0.896, 95% CI 0.456 to 1.759, p=0.750) and 6 months (HR=1.054, 95% CI 0.585 to 1.899, p=0.861) did not show an association between irSAE and improved PFS (table 5, online supplemental table 2 and figure 2,3). In multivariable analysis, the presence of peritoneal carcinomatosis, liver metastasis, lung metastasis, a later line of ICI treatment and worse ECOG PS at ICI initiation were independently associated with worse PFS (table 4). Table 3Tumor response according to the presence or absence of grade 3 and more immune-related adverse eventsPatient without irSAE (n=878)Patients with irSAEs (n=109)Overall (n=987)Complete response19 (2.2%)10 (9.2%)29 (2.9%)Partial response285 (32.4%)43 (39.4%)328 (33.2%)Stable disease375 (42.7%)43 (39.4%)418 (42.4%)Disease progression199 (22.7%)13 (12.0%)212 (21.5%)irSAE, immune-related serious adverse events. Tumor response according to the presence or absence of grade 3 and more immune-related adverse events irSAE, immune-related serious adverse events. Table 4Factors associated with progression-free survival using Cox model with irSAE as a time-varying covariateCharacteristicUnivariateMultivariate (N=1076)NHR95% CIP valueHR95% CIP valueSex (female)11750.8930.752 to 1.0600.196Age (1 year more)11751.0041.000 to 1.0080.0471.0051.000 to 1.0090.050Primary tumor site1175 Colorectal———— Esophagogastric0.7300.504 to 1.0570.0950.7330.500 to 1.0740.111 Pancreatic0.8100.419 to 1.5680.5320.7580.386 to 1.4890.422 Other1.1000.796 to 1.5200.5651.1110.796 to 1.5510.536Grade539 Poorly differentiated—— Moderately differentiated1.0440.781 to 1.3950.771 Well differentiated0.9710.658 to 1.4340.883No metastatic disease at ICI start*11690.4140.243 to 0.7040.001Peritoneal carcinomatosis11661.4501.221 to 1.7220.0011.4811.232 to 1.7810.001Liver metastasis11661.4271.197 to 1.7020.0011.3791.138 to 1.6720.001Lung metastasis11661.8051.482 to 2.1990.0011.5131.211 to 1.8900.001Lymph node metastasis11660.9830.826 to 1.1690.843Surgery of primary tumor11681.0110.833 to 1.2270.908Line of ICI treatment1098 First-line———— Second-line1.2921.045 to 1.5980.0181.3821.111 to 1.7190.004 Third-line1.7561.350 to 2.2850.0011.7021.297 to 2.2350.001 Fourth-line and more2.0981.542 to 2.8550.0011.7901.286 to 2.4920.001ECOG PS at ICI start1155 0———— 11.5061.234 to 1.8370.0011.4521.173 to 1.7990.001 22.3221.764 to 3.0580.0012.3841.789 to 3.1780.001 3–43.4562.264 to 5.2750.0013.3362.159 to 5.1560.001RAS mutation among CRC*8870.8190.660 to 1.0170.070BRAF mutation among CRC*8941.0730.879 to 1.3100.486Type of ICI806 Pembrolizumab—— Nivolumab0.9620.693 to 1.3340.815 Ipilimumab plus Nivolumab0.3070.098 to 0.9610.042 Other0.9680.737 to 1.2720.817Number of ICI molecules806 ICI monotherapy—— ICI combination0.6320.432 to 0.9240.018irSAE as a time-dependent variable11751.3050.925 to 1.8410.1291.2800.898 to 1.8240.172*The “No metastatic disease at ICI start” variable was not included in the multivariate since due to interactions with the variables from different metastatic sites. RAS and BRAF mutations were not included in the multivariate model since they are relevant only in CRC.CRC, colorectal cancer; ECOG PS, Eastern Cooperative Oncology Group performance status; ICI, immune checkpoint inhibitor; irSAE, immune-related serious adverse events; Ref, reference. Factors associated with progression-free survival using Cox model with irSAE as a time-varying covariate *The “No metastatic disease at ICI start” variable was not included in the multivariate since due to interactions with the variables from different metastatic sites. RAS and BRAF mutations were not included in the multivariate model since they are relevant only in CRC. CRC, colorectal cancer; ECOG PS, Eastern Cooperative Oncology Group performance status; ICI, immune checkpoint inhibitor; irSAE, immune-related serious adverse events; Ref, reference. Table 5Factors associated with progression-free survival using Cox model with landmark at 3 monthsCharacteristicUnivariateMultivariate (N=830)NHR95% CIP valueHR95% CIP valueSex (female)9080.7640.605 to 0.9660.0250.7120.553 to 0.9160.008Age (1 year more)9081.0071.001 to 1.0120.0131.0101.004 to 1.0160.001Primary tumor site908 Colorectal—— Esophagogastric0.7360.444 to 1.2210.235 Pancreatic0.6860.255 to 1.8420.454 Other1.3850.917 to 2.0910.122Grade425 Poorly differentiated—— Moderately differentiated0.9510.646 to 1.4000.800 Well differentiated0.9010.537 to 1.5120.694No metastatic disease at ICI start9020.5890.313 to 1.1080.101Peritoneal carcinomatosis8991.2741.008 to 1.6100.0431.3291.034 to 1.7070.026Liver metastasis8991.0670.832 to 1.3690.610Lung metastasis8991.8791.432 to 2.4660.0011.6741.235 to 2.2690.001Lymph node metastasis8990.9550.754 to 1.2090.699Surgery of primary tumor9021.0320.788 to 1.3500.820Line of ICI treatment850 First-line———— Second-line1.0350.776 to 1.3810.8131.1770.875 to 1.5840.281 Third-line1.4761.027 to 2.1220.0351.5081.042 to 2.1820.029 Fourth-line and more1.9961.324 to 3.0090.0012.0891.355 to 3.2190.001ECOG PS at ICI start890 0———— 11.2871.002 to 1.6540.0491.1720.896 to 1.5340.248 21.3170.851 to 2.0370.2161.3280.854 to 2.0650.209 3–41.1430.466 to 2.8040.7700.9420.382 to 2.3270.897RAS mutation among CRC6770.8380.626 to 1.1220.236BRAF mutation among CRC*6821.2560.960 to 1.6430.097Type of ICI617 Pembrolizumab—— Nivolumab0.7910.490 to 1.2760.337 Ipilimumab plus nivolumab0.2560.036 to 1.8320.175 Other1.1890.841 to 1.6810.327Number of ICI molecules617 ICI monotherapy—— ICI combination0.7730.475 to 1.2580.300irSAE9080.8170.420 to 1.5860.5500.8960.456 to 1.7590.750*BRAF mutation was not included in the multivariate model since it is relevant only in CRC.CRC, colorectal cancer; ECOG PS, Eastern Cooperative Oncology Group performance status; ICI, immune checkpoint inhibitor; irSAE, immune-related serious adverse events; Ref, reference. Factors associated with progression-free survival using Cox model with landmark at 3 months *BRAF mutation was not included in the multivariate model since it is relevant only in CRC. CRC, colorectal cancer; ECOG PS, Eastern Cooperative Oncology Group performance status; ICI, immune checkpoint inhibitor; irSAE, immune-related serious adverse events; Ref, reference. jitc-14-7-s001.pdf When irSAEs were treated as a time-fixed covariate, OS tended to be longer in patients with irSAE (median not reached vs 84.3 months 95% CI 67.5 to not applicable (NA), p=0.061) (online supplemental table 3). OS at 3 years was 70.6% in patients with irSAE versus 63.0% in patients without irSAE. At the end of the follow-up, 34.5% of patients without irSAE had died versus 25.6% of patients with irSAE. In multivariable analysis, using irSAEs as a time-dependent variable, irSAE was not associated with better OS (HR=1.043, 95% CI 0.712 to 1.528, p=0.828) (table 6). Using landmark analysis at 3 months (HR=1.186, 95% CI 0.698 to 2.017, p=0.528) and 6 months (HR=0.968, 95% CI 0.558 to 1.679, p=0.908) irSAE was not associated with better OS (online supplemental table 4, 5). In multivariable analysis, the presence of peritoneal carcinomatosis, liver metastasis, lung metastasis, a later line of ICI initiation and worse ECOG PS at ICI start were associated with worse OS (table 6). Additionally, we found that patients alive and progression-free at 3 months included in the landmark analysis at 3 months were less likely to have peritoneal carcinomatosis, liver metastasis and lung metastasis, and were more likely to have been treated in a first-line setting and to have a better ECOG PS, all of which are associated with better PFS and OS (online supplemental table 6). Table 6Factors associated with overall survival using Cox model with irSAE as time-dependent variableCharacteristicUnivariateMultivariate (N=1076)NHR95% CIP valueHR95% CIP valueSex (female)11750.9240.759 to 1.1260.433Age (1 year more)11751.0051.000 to 1.0090.0411.0040.999 to 1.0100.126Primary tumor site1175 Colorectal—— Esophagogastric0.8100.531 to 1.2360.328 Pancreatic0.7270.324 to 1.6300.439 Other1.2500.870 to 1.7970.228Grade539 Poorly differentiated—— Moderately differentiated0.9140.653 to 1.2800.600 Well differentiated0.8530.539 to 1.3500.498No metastatic disease at ICI start**11690.4470.239 to 0.8390.012Peritoneal carcinosis11661.5541.275 to 1.8930.0011.6411.326 to 2.0310.001Liver metastasis11661.5471.267 to 1.8900.0011.5911.280 to 1.9790.001Lung metastasis11661.7221.373 to 2.1590.0011.3951.079 to 1.8030.011Lymph node metastasis11660.9280.761 to 1.1320.462Surgery of primary tumor11680.8730.702 to 1.0850.221Line of ICI treatment1098 First-line———— Second-line1.2570.980 to 1.6120.0721.3191.022 to 1.7010.033 Third-line1.6331.202 to 2.2190.0021.5291.117 to 2.0920.008 Fourth-line and more2.5111.803 to 3.4980.0012.1471.507 to 3.0600.001ECOG PS at ICI start1155 0———— 11.7681.392 to 2.2470.0011.6561.278 to 2.1460.001 23.5582.620 to 4.8320.0013.6822.670 to 5.0770.001 3–45.0463.205 to 7.9460.0014.8363.018 to 7.7490.001RAS mutation among CRC*8870.7470.580 to 0.9620.024BRAF mutation among CRC8941.1220.893 to 1.4090.322Type of ICI806 Pembrolizumab—— Nivolumab1.0830.760 to 1.5430.660 Ipilimumab plus nivolumab0.5100.163 to 1.5990.248 Other0.8610.625 to 1.1860.360Number of ICI molecules806 ICI monotherapy—— ICI combination0.6530.422 to 1.0100.056irSAE as time-dependent variable11751.1660.805 to 1.6890.4161.0430.712 to 1.5280.828*The “No metastasis disease at ICI start” variable was not included in the multivariate since due to interactions with the variables from different metastatic sites. RAS mutation was not included in the multivariate model since it is relevant only in CRC.CRC, colorectal cancer; ECOG PS, Eastern Cooperative Oncology Group Performance Status; ICI, immune checkpoint inhibitor; irSAE, immune-related serious adverse events. Factors associated with overall survival using Cox model with irSAE as time-dependent variable *The “No metastasis disease at ICI start” variable was not included in the multivariate since due to interactions with the variables from different metastatic sites. RAS mutation was not included in the multivariate model since it is relevant only in CRC. CRC, colorectal cancer; ECOG PS, Eastern Cooperative Oncology Group Performance Status; ICI, immune checkpoint inhibitor; irSAE, immune-related serious adverse events. Discussion Our study did not identify grade ≥3 irAE as a predictor of better survival outcomes in patients with dMMR/MSI-H digestive cancers treated with ICIs. To our knowledge, this is the first study to specifically evaluate the relationship between irSAEs and outcomes in dMMR/MSI-H digestive cancers. In addition, our series represents the largest real-world cohort to date, including 1,175 patients with dMMR/MSI-H digestive cancers treated with ICIs. Patients’ and tumor characteristics from our series are consistent with those reported in previously published trials and cohorts of dMMR/MSI-H digestive cancers.14–19 It is worth noting that 48.0% of patients received ICI in a first-line setting, and that the majority (87.6%) were treated with ICI monotherapy. Overall, 32.5% of patients experienced at least one irAE, including 10.0% of irSAEs. Five patients (0.4%) experienced a grade 5 adverse event. Only 11.0% of patients permanently discontinued ICI treatment due to irAEs. These rates are consistent with the literature, according to which the expected rate of grade ≥3 irAEs in ICI monotherapy is 10–20%, rising to 20–40% with ICI combination therapy, with about 10% of permanent discontinuation and 3% of immune-related toxic deaths in other tumor types.22–24 The rates of AEs in our study are comparable to those in the KEYNOTE-177 trial (pembrolizumab vs chemotherapy), with 31% of irAEs and 9% of grade ≥3 irAEs (no death) with anti-PD-1 monotherapy.34 Similarly, these results align with those of the CHECKMATE-8HW trial, which reported 14% of grade ≥3 irAEs with nivolumab (one death) versus 22% with nivolumab plus ipilimumab combination (two deaths).15 In our study, irSAEs were mainly gastrointestinal, hepatic and cutaneous AEs (3.4%, 1.8% and 1.0%, respectively). Approximately 90% of patients recovered from irSAE with adequate management. This toxicity profile is consistent with that reported in pivotal trials. In the KEYNOTE-177 trial, the most common grade ≥3 irAEs were colitis (3%) and hepatitis (3%).34 In the CHECKMATE-8HW study, the most common grade ≥3 irAEs in the nivolumab plus ipilimumab and nivolumab groups were diarrhea or colitis (3% and 2%, respectively), hypophysitis (3% and 1%, respectively), and adrenal insufficiency (3% and <1%, respectively).15 To our knowledge, few studies have investigated predictors of irSAEs, and none have focused on patients with dMMR/MSI-H digestive cancers treated with ICIs. IrSAEs were reported to occur more frequently with ICI doublet therapy than with monotherapy.15 In our cohort, the only factor significantly associated with irSAEs was ICI combination therapy compared with monotherapy (20.0% vs 11.4%, respectively). A major risk factor for irAEs is a personal history of chronic inflammatory disease, but unfortunately, this information was not collected in our study. Recent studies, mainly in lung cancer, identified a younger age, brain metastases, a history of smoking, high PD-L1 expression and first-line treatment as factors associated with a higher incidence of irSAEs.35–37 These discrepancies likely reflect tumor-specific differences in immune toxicity profiles, as previously suggested.25 26 The median time to irSAE onset was approximately 3.5 months in our study. While KEYNOTE 177 and CHECKMATE-8HW did not report time to irAEs onset, our results are consistent with other randomized trials and real-world cohorts across tumor types, where median onset typically ranges from 2 to 3 months, depending on the type of toxicity.38–40 Skin toxicities usually occur first (3–6 weeks), followed by colitis, whereas hepatitis and endocrinopathies tend to appear later (6–24 weeks). When irSAEs were analyzed as a time-fixed covariate, an approach known to induce bias, irSAEs appeared to be associated with improved outcomes, including higher ORR (48.6% vs 34.6%), higher DCR (88.0% vs 77.3%), longer PFS (3-year PFS: 66.1% vs 52.6%), and a trend toward improved OS (3-year OS: 70.6% vs 63.0%). However, irSAEs are by nature time-dependent events that occur during treatment. The higher ORR and DCR observed in patients with irSAEs should be interpreted with caution, as these endpoints are susceptible to guarantee-time bias. Patients who develop irSAEs must remain on treatment long enough to both respond to treatment and develop toxicity, while those with early progression may not. When accounting for this using time-dependent Cox models and landmark analyses, irSAEs were no longer associated with improved PFS or OS. This suggests a classic immortal time bias; patients with favorable prognostic factors survive longer and therefore have more time to develop irSAEs, whereas patients with a poor prognosis may progress or die before having the opportunity to experience ICI toxicity. Indeed, we found that patients who progressed or died within 3 months were more likely to have peritoneal carcinomatosis, liver metastasis, lung metastasis, to be treated with ICIs in later lines and to have a worse ECOG PS, all of which are associated with shorter PFS and OS. Landmark analyses could be subject to selection bias by excluding patients with early progression or death. The approach in the time-dependent Cox analyses is a well-known alternative that avoids not only the problem of selecting a landmark time (which excludes patients with early events) but also any misclassification errors regarding irSAE status. Several studies in other tumor types, including melanoma, renal cell carcinoma, non-small-cell lung cancer, hepatocellular carcinoma, and non-MSI-H gastric cancer, have reported associations between irAEs and improved PFS, OS, or response to treatment.27–30 33 In more than one thousand patients with resected melanoma treated with adjuvant pembrolizumab, irAEs were associated with better recurrence-free survival (HR, 0.61; 95% CI 0.39 to 0.95; p=0.03).27 Similar associations have been reported in metastatic renal cell carcinoma and hepatocellular carcinoma.28 30 41 However, results remain inconsistent, with several studies, particularly in melanoma, failing to identify irSAEs as predictors of improved outcome, and many not accounting for the time-dependent nature of irSAEs.40 42–44 Moreover, some data suggest that only specific toxicities, such as endocrine or cutaneous irAEs, may be associated with better outcomes.35 44 We also studied the management of irSAEs. Most patients received corticosteroids (70%) and about 15% required immunosuppressive agents. In about 90%, irSAE resolved with appropriate management, with no detrimental impact on survival, even when ICIs were discontinued. Previous studies in other tumor types also demonstrated that ICI discontinuation after irAE did not negatively affect survival, as was the case with the use of corticosteroids.29 32 40 43 In our study, only one-third of patients resumed immunotherapy after the occurrence of an irSAE, and about half of these patients experienced a recurrence of the same irAE. This rate is in line with prior studies, which reported irAE recurrence rates ranging from 30 to 50%.20 35 45 A recent meta-analysis identified several factors associated with irAE recurrence, including the type of irAE, the type of cancer, a short time to re-introduction (less than 10 weeks), gastrointestinal irAEs and ICI combination therapy.20 These findings may help to guide the management of dMMR/MSI-H patients with irSAEs. Given that most patients achieve disease control and that the risk of irAE recurrence is substantial, ICI reintroduction should be considered only in patients with disease progression after ICI discontinuation and discussed on a case-by-case basis within a multidisciplinary team. The main limitations of our study include its ambispective design and the potential under-reporting of low-grade irAEs. It is for this last reason that we decided to look only at the links between grade ≥3 irAEs and outcomes. However, major strengths include the large sample size: the largest existing cohort of dMMR/MSI-H digestive tumors, supported by multiple international centers. In addition, data concerning tumor and patients’ characteristics, treatment response and survival are reliable with little missing information. Importantly, we applied appropriate time-dependent analytical methods, including landmark analyses, to account for the delayed occurrence of irSAEs. Conclusion Together, these findings indicate that while irAEs may reflect immune activation, irSAEs should not be interpreted as reliable surrogate markers of treatment efficacy in dMMR/MSI-H digestive cancers since when appropriate time-dependent methods were applied there was no association with survival. Among patients who experienced irSAEs, most achieved durable disease control, despite permanent ICI discontinuation and the use of corticosteroids and immunosuppressive agents. These findings suggest that a watch-and-wait strategy, without systematic reintroduction of ICIs after irSAEs, may be appropriate in patients who maintain disease control. Acknowledgments The authors would like to thank the AGEO as well as all the clinical research assistants who helped to collect the clinical data. Footnotes Contributors: Conceptualization: DT, CK, RO and EA. Acquisition of the data: All authors. Analysis and interpretation of the data: DT, CK, MA, OB, RO and EA. Writing: DT, CK, RO and EA. Review and editing: All authors. Guarantor: DT. All authors contributed to the preparation, editing, and review of this manuscript. All authors read and approved the final manuscript. RO and EA shared co-last authorship. Funding: The authors have not declared a specific grant for this research from any funding agency in the public, commercial or not-for-profit sectors. Competing interests: DT reports roles as consultant/advisor and/or honoraria, travel grants, and research grants from Amgen, Bayer, Roche, Merck Serono, Sanofi, Pierre Fabre, MSD, AstraZeneca, BeiGene, Bristol Myers Squibb, Merck Health KGAA, Merck Sharp & Dohme, Servier, and Taiho and Pierre Fabre, and reports institutional research funding from Amgen, MSD, Servier, AstraZeneca, Pierre Fabre, Gilead, Takeda. EA reports consultant/advisor honoraria from Merck Serono and Servier, research grants from Roche and travel grants from Pierre Fabre. AT reports personal fees from Servier, Incyte Bioscience, Novocure, BMS, Taiho, Norgine, Pierre Fabre and Merck, as well as travel grants and personal fees from AstraZeneca, Pierre Fabre, Takeda and MSD outside the submitted work. MO reports consulting from Exelixis, Summit, GSK, Merus, 3T Biosciences, Roche, BMS, Merck, Moma, Amgen, Takeda, Pfizer, Natera, Agenus, Regeneron, Astellas; research funding from Summit, Moma, Takeda, Roche, Lilly, Merck, BMS, Nouscom, Phanes. FS reports consultancy, advisory roles, honoraria: AMAL Therapeutics, Amgen, Bayer, BMS, Dragonfly Therapeutics, GSK, Medimix, Merck, Nordic Pharma, Roche, Servier; research funding and educational grants: Amgen, Astellas, AstraZeneca, Bayer, BMS, Fondation Contre le Cancer, Knowledge Centre of Excellence, Merck, MSD, Pierre Fabre, Roche, Sanofi, Servier; travel expenses: Amgen, AstraZeneca, Bayer, Lilly, Merck, Roche, Servier; leadership roles as secretary of the EORTC Gastrointestinal Tract Cancer Group. TM has received honoraria for speaking or consulting roles from Servier, Pierre Fabre, Merck Serono, Takeda, Galapagos, and research grants from Amgen and Roche. He has also received travel grants from Pierre Fabre, Merck Serono, MSD and Servier. VH has received honoraria for speaking or consulting roles from AdaCap/Novartis, Deciphera, Pierre Fabre, Merck Serono, Ipsen, as well as travel grants from Pierre Fabre, Merck Serono, and Ipsen. Filippo Pietrantonio reports receiving research funding (to Institution) from Lilly, BMS, Incyte, AstraZeneca, Amgen, Agenus, Rottapharm, Johnson & Johnson, GSK, Tempus, BeOne. Personal honoraria as an invited speaker from BeOne, Daiichi-Sankyo, Seagen, Astellas, Ipsen, AstraZeneca, Servier, Bayer, Takeda, Johnson & Johnson, BMS, MSD, Amgen, Merck-Serono, Pierre Fabre, Incyte, AstraZeneca. Advisory/Consultancy from BMS, MSD, Amgen, Pierre Fabre, Johnson & Johnson, Servier, Bayer, Takeda, Astellas, GSK, Daiichi-Sankyo, Pfizer, BeOne, Jazz Pharmaceuticals, Incyte, Rottapharm, Merck-Serono, Italfarmaco, Gilead, AstraZeneca, Agenus, Revolution Medicines, 3T Biosciences. Travel expenses from Amgen, Merck-Serono, Pierre Fabre, Servier, Astellas, Incyte, Johnson & Johnson. JT declares compensated consultancy or advisory role for Amgen, Astellas, BMS, Merck, MSD, Novartis, Pfizer, Pierre Fabre, Rottapharm, Sanofi, Servier, Takeda; travel grants from Servier, Takeda, Pierre Fabre, Merck; research funding from Novartis, Servier, Pierre Fabre, Natera and speaker fees for Amgen, Astellas, BMS, Merck, MSD, Natera, Novartis, Pierre Fabre, Servier, Takeda. TA reports roles as consultant/advisor and/or honoraria and travel grants from Amgen, Gilead, Pierre Fabre, AstraZeneca, Merck, BMS, MSD. RC reports honoraria from Amgen, AstraZeneca/Daiichi Sankyo, Bristol Myers Squibb, Exeliome Bioscience, Merck, MSD Oncology, Servier; travel fees and congress invitations from Bristol Myers Squibb, Enterome, MSD Oncology, Servier. None of the remaining authors have any competing interests to disclose. Provenance and peer review: Not commissioned; externally peer reviewed. Supplemental material: This content has been supplied by the author(s). It has not been vetted by BMJ Publishing Group Limited (BMJ) and may not have been peer-reviewed. Any opinions or recommendations discussed are solely those of the author(s) and are not endorsed by BMJ. BMJ disclaims all liability and responsibility arising from any reliance placed on the content. Where the content includes any translated material, BMJ does not warrant the accuracy and reliability of the translations (including but not limited to local regulations, clinical guidelines, terminology, drug names and drug dosages), and is not responsible for any error and/or omissions arising from translation and adaptation or otherwise. Data availability statement Data are available upon reasonable request. Data are available upon reasonable request. The authors declare that the data presented in this article are available as raw data. This supplementary material of this study is available from the corresponding author upon reasonable request. Ethics statements Patient consent for publication Not applicable. Ethics approval The study was approved by the Commission Nationale de l’Informatique et des Libertés, CNIL, reference 2222018 v0. Participants gave informed consent to participate in the study before taking part. References Associated Data Supplementary Materials jitc-14-7-s001.pdf Data Availability Statement Data are available upon reasonable request. Data are available upon reasonable request. The authors declare that the data presented in this article are available as raw data. This supplementary material of this study is available from the corresponding author upon reasonable request.