Advertisement

Analysis of the association between prospectively collected immune-related adverse events and survival in patients with solid tumor treated with immune-checkpoint blockers, taking into account immortal-time bias

Open AccessPublished:August 10, 2022DOI:https://doi.org/10.1016/j.ctrv.2022.102452

      Highlights

      • The association between immune-related adverse events (irAEs) and survival in patients with solid tumor is still unclear.
      • Published studies not dealing appropriately with the immortal-time bias should be interpreted with caution.
      • Our real-life study based on prospectively collected irAEs, used a landmark analysis and a Cox regression model with time-dependent covariate to address the immortal-time bias.
      • IrAEs are associated with improved survival using a time-varying Cox regression model.

      Abstract

      Background

      Numerous retrospective studies and reviews have reported a positive association between immune-related adverse events (irAEs) and survival in non-small cell lung cancer (NSCLC) and melanoma patients treated with immune checkpoint blockers (ICBs). However, some results are controversial and the studies, whose results converge, should be interpreted cautiously because most of them do not deal appropriately with the immortal-time bias. Here, we report an observational real-life study of the association between prospectively collected irAEs and survival of patients treated with ICBs while dealing with the immortal-time bias.

      Methods

      Data from patients treated at Gustave Roussy from June 2014 to October 2017 with anti-PD-(L)1 antibodies for a melanoma or NSCLC have been prospectively collected in the REISAMIC database, a pharmacovigilance registry dedicated to irAEs. Adverse events of grade 2 and higher were collected prospectively. To study the association between the occurrence of irAEs and survival, we used both a landmark analysis and a Cox regression model with time-dependent covariate.

      Results

      577 patients were treated with anti-PD-(L)1 antibodies for melanoma (60.3 %) or NSCLC (39.7 %). The occurrence of an irAE was significantly associated with improved overall survival (OS): HR 0.56, 95 % CI [0.41; 0.75], p = 0.0001 and progression-free survival (PFS): HR 0.63, 95 % CI [0.47; 0.83], p = 0.001 using a Cox regression model with time-dependent covariate. In a 12-week landmark analysis, median OS was 21.2 months (95 % CI, 12.2 to 35.7) and 16.4 months (95 % CI, 12.4 to 21.3) p = 0.26 and median PFS was 14.3 months (95 % CI, 9.5 to 24.6) and 13.4 months (95 % CI, 10.2 to 18.3) p = 0.66, for patients with and without irAEs, respectively.

      Conclusions

      In our real-life study of patients with melanoma and NSCLC treated with anti-PD-(L)1 antibodies, we confirm that irAEs are associated with improved survival using a time-varying Cox regression model. Analysis with a landmark method showed no difference in OS or PFS between patients who experienced irAE during the first 12 weeks of treatment and those who did not. Retrospective analysis and reviews including studies that do not deal with the immortal-time bias and studies insufficiently powered for a landmark analysis should be interpreted with caution.

      Keywords

      Introduction

      The development of immune checkpoint blockers (ICBs) such as anti-programmed death-1 (anti-PD-1) and programmed death ligand-1 (anti-PD-L1) antibodies have changed the landscape of cancer treatment. ICBs have improved the clinical outcome of patients in various types of cancer including melanoma and non-small cell lung carcinoma (NSCLC) and the list of their indications has been extended [
      • Vaddepally R.K.
      • Kharel P.
      • Pandey R.
      • et al.
      Review of Indications of FDA-Approved Immune Checkpoint Inhibitors per NCCN Guidelines with the Level of Evidence.
      ]. Immune-related adverse events (irAEs) are immune-mediated secondary effects to ICBs. Although the exact pathophysiology of irAEs remains undetermined, it has been hypothesized that certain patients who experience irAEs, would have reinvigoration of their immune system and therefore could derive greater benefit from ICBs [

      Das S, Johnson DB. Immune-related adverse events and anti-tumor efficacy of immune checkpoint inhibitors. J Immunother Cancer 2019;7. doi:10.1186/s40425-019-0805-8.

      ].
      Numerous studies have previously reported a positive association between irAEs and survival in patients with NSCLC and melanoma treated with ICBs [

      Eggermont AMM, Kicinski M, Blank CU, et al. Association between immune-related adverse events and recurrence-free survival among patients with stage III melanoma randomized to receive pembrolizumab or placebo: a secondary analysis of a randomized clinical trial. JAMA Oncol Published Online First: 2 January 2020. doi:10.1001/jamaoncol.2019.5570.

      ,
      • Freeman-Keller M.
      • Kim Y.
      • Cronin H.
      • et al.
      Nivolumab in resected and unresectable metastatic melanoma: characteristics of immune-related adverse events and association with outcomes.
      ,
      • Haratani K.
      • Hayashi H.
      • Chiba Y.
      • et al.
      Association of immune-related adverse events with nivolumab efficacy in non–small-cell lung cancer.
      ]. This association remains unclear because of conflicting results published by other teams [
      • Owen D.H.
      • Wei L.
      • Bertino E.M.
      • et al.
      Incidence, risk factors, and impact on survival of immune-related adverse events in patients with non-small cell lung cancer.
      ,
      • Sato K.
      • Akamatsu H.
      • Murakami E.
      • et al.
      Correlation between immune-related adverse events and efficacy in non-small cell lung cancer treated with nivolumab.
      ]. Part of the controversy might be related to immortal-time bias, sample size bias, bias induced by retrospective analysis, and to methodological differences between studies.
      Four systematic reviews and meta-analysis showed improved progression free-survival (PFS) and overall survival (OS) in patients experiencing an irAE [
      • Fan Y.
      • Xie W.
      • Huang H.
      • et al.
      Association of immune related adverse events with efficacy of immune checkpoint inhibitors and overall survival in cancers: a systemic review and meta-analysis.
      ,
      • Hussaini S.
      • Chehade R.
      • Boldt R.G.
      • et al.
      Association between immune-related side effects and efficacy and benefit of immune checkpoint inhibitors – A systematic review and meta-analysis.
      ,
      • Petrelli F.
      • Grizzi G.
      • Ghidini M.
      • et al.
      Immune-related adverse events and survival in solid tumors treated with immune checkpoint inhibitors: a systematic review and meta-analysis.
      ,
      • Zhou X.
      • Yao Z.
      • Yang H.
      • et al.
      Are immune-related adverse events associated with the efficacy of immune checkpoint inhibitors in patients with cancer? A systematic review and meta-analysis.
      ]. However, these meta-analyses have included studies dealing inappropriately with the immortal-time bias and retrospective observational studies that may limit the quality of the analysis. Even though their results converge, these should be interpreted with caution.
      Here, we report the first observational real-world study of the association between prospectively collected irAEs and survival of patients treated with ICBs using both a landmark analysis and a Cox regression model with time-dependent covariate.

      Methods

      Patients

      Patients treated at Gustave Roussy from June 1st, 2014 to October 30th, 2017 with anti-PD-(L)1 antibodies for a solid tumor, have been included prospectively in our pharmacovigilance database REISAMIC (https://www.gustaveroussy.fr/en/reisamic-register). Patients with a malignant haematologic disease, and solid tumors other than melanoma and NSCLC were excluded from our analysis. All patients received at least one dose of treatment. Patients were treated until disease progression or unacceptable toxicity according to the referring physician.

      Data collection

      Patients’ characteristics and adverse events of grade 2 and higher according to the Common Terminology Criteria for Adverse Events version 4 were reported prospectively. Dates of progression or death and status at last follow up were collected retrospectively. The end of data collection and follow-up period was December 31st, 2017. We defined irAEs as immune-related adverse events following the initiation of treatment with anti-PD-(L)1 antibodies. For complex adverse events, a multi-disciplinary committee with an independent panel of experts in immunological toxicities determined the relationship between the adverse event and the treatment and assessed causality using the World Health Organization’s Uppsala Monitoring Center scale. IrAEs were managed independently by the patient’s referring physician based on the Gustave Roussy irAEs management guidelines [
      • Champiat S.
      • Lambotte O.
      • Barreau E.
      • et al.
      Management of immune checkpoint blockade dysimmune toxicities: a collaborative position paper.
      ]. The study was conducted in compliance with good clinical practice and the tenets of the Declaration of Helsinki. The REISAMIC registry has been authorized by Gustave Roussy Institutional Review Board and the French National Data Protection Commission (CNIL N° 2098694).

      Efficacy assessments and outcomes

      OS was defined as the time from treatment initiation until either the date of death, the date of last follow-up or the date of end of analysis. PFS was defined as the time from treatment initiation until disease progression or death, whichever occurred first.

      Statistical analyses

      Data analysis was performed from April to September 2018. Because the occurrence of irAEs is a time-dependent factor, we addressed the immortal-time bias by using two methods: a landmark analysis and a Cox regression model with time-dependent covariate. With the landmark analysis, we aimed to detect a difference of OS between patients who experienced irAE within a fixed period after PD-(L)1 initiation and those who did not. With a Cox model with time-depending covariate, we aimed to evaluate the effect of the occurrence of an irAE during patients follow-up. As this study was not designed for specific hypothesis on expected effect, there was no number of required subjects to calculate. For the landmark analysis, we considered a time point at 12 weeks after treatment initiation, based on the data available. Only patients alive at this point were included in our analysis and divided into groups that presented or not with irAEs before this time point. A sensitivity analyses according to histology and irAE type was performed at the 12-week landmark. A sensitivity analyses according to the onset of irAE was performed at 8 and 16 weeks.
      For the Cox regression model with time-dependent covariates, the covariates included were: age, histology, ECOG score, type of treatment received, number of previous lines, presence of metastases, steroids at baseline and the occurrence of irAE. Occurrence of the irAE were considered as a time varying covariate. For the comparison of the impact of corticosteroids on irAE, a Fisher’s exact test was used.
      All analyses were performed using SAS software (version 9.4; SAS Institute Inc).

      Results

      Patients’ characteristics

      577 patients were included in the study. Baseline characteristics are shown in Table 1. 75 % of patients received at least one line of treatment prior to ICBs. 297 patients (51.5 %) were treated with pembrolizumab and 280 (48.5 %) with nivolumab. 348 patients (60.3 %) were treated for melanoma and 229 (39.7 %) for NSCLC. Median (range) treatment duration was 3 months (0; 32) in patients without irAEs and 7 months (0; 32) in patients with irAEs. At the time of analysis, 87 patients were still undergoing treatment, and 490 had stopped for progression (75.5 %), toxicity (9.8 %), efficacy (8 %) or other (6.7 %). Median follow-up period was 17.6 months (95 % CI [15.4–20.2]).
      Table 1Patients characteristics.
      All patients (n = 577)No IrAE patients (n = 408)IrAE patients
      IrAE are of grade ≥ 2.
      (n = 169)
      Age

      <= 65 years

      >65 years
      321 (55.6 %)

      256 (44.4 %)
      230 (56.4 %)

      178 (43.6 %)
      91 (53.8 %)

      78 (46.2 %)
      EGOG performance status

      0–1

      2

      3–4

      Missing
      495 (86.1 %)

      54 (9.4 %)

      26 (4.50 %)

      2
      334 (82.3 %)

      48 (11.8 %)

      24 (5.90 %)

      2
      161 (95.3 %)

      6 (3.50 %)

      2 (1.50 %)

      0
      Ratio Female/Male248/ 329173/ 23575/ 94
      Histology

      Melanoma

      NSCLC
      348 (60.3 %)

      229 (39.7 %)
      220 (53.9 %)

      188 (46.1 %)
      128 (75.7 %)

      41 (24.3 %)
      Metastatic disease499 (86.7 %)359 (87.9 %)140 (82.8 %)
      No of previous lines of treatment



      No previous line

      One or two

      Three or more




      155 (26.9 %)

      342 (59.3 %)

      80 (13.9 %)




      94 (23.0 %)

      254 (62.3 %)

      60 (14.7 %)




      61 (36.1 %)

      88 (52.1 %)

      20 (11.8 %)
      Type of treatment

      Pembrolizumab

      Nivolumab
      297 (51.5 %)

      280 (48.5 %)
      193 (47.3 %)

      215 (52.7 %)
      104 (61.5 %)

      65 (38.5 %)
      Duration of treatment, months

      Median (range)
      4 (0–32)3 (0–32)7 (0–32)
      Baseline corticosteroids

      ≥ 10 mg per day
      136 (23.6 %)113 (27.7 %)23 (13.6 %)
      Abbreviations: irAE, immune-related adverse events; NSCLC, non-small cell lung carcinoma; No, number.
      # IrAE are of grade ≥ 2.

      IrAEs profile

      Among the 577 patients included in the study, 169 patients (29.3 %) developed 241 irAEs of grade 2 and higher. 121 patients presented only one irAE and 48 presented 2 or more irAEs with 1 to 5 organs affected. irAEs appeared after a median of 2 months of treatment with ICBs. irAEs were more commonly observed in skin (31.1 %), thyroid (20.7 %), gastrointestinal tract (10.8 %), liver (6.6 %), lungs (5.4 %) and pancreas (4.6 %) (Supplementary 1). Skin and thyroid toxicities were mainly low grade. In contrast, pulmonary, hepatic and gastro-intestinal irAEs were less frequent but more severe, managed mainly by systemic corticosteroids and often lead to a definitive discontinuation of ICBs. 51 (8.8 %) patients discontinued therapy due to severe irAEs. There were two deaths related to irAEs.

      Impact of irAEs on survival

      In our cohort of patients, median OS was 7.1 months (95 % CI [4.4–9.4]) and median PFS was 2.8 months (95 % CI [2.3–3.6]) for NSCLC patients. Median OS was 12.5 months (95 % CI [10.0–19.7] and median PFS was 6.9 months (95 % CI [5.6–9.5]) for melanoma patients (Supplementary 2). To avoid immortal-time bias when analyzing the association between irAEs and survival, we conducted both a landmark analysis and a time-dependent Cox regression model.

      Landmark analysis

      Number of patients and number of events, overall and at each landmark time, is shown in Table 2. Number of irAEs by tumor type at each landmark time, is shown in Supplementary 3.
      Table 2Number of patients and number of events, overall and at each landmark time.
      irAE during landmark periodNo irAE during landmark periodTotal
      Overall Survival
      No Landmark169408577
      Number of events60277337
      Landmark 2 months74400474
      Number of events34202236
      Landmark 3 months98329427
      Number of events42159201
      Landmark 4 months105277382
      Number of events43127170
      Progression free Survival
      No Landmark169408577
      Number of events73319392
      Landmark 8 weeks72360432
      Number of events39208247
      Landmark 12 weeks91275366
      Number of events46145191
      Landmark 16 weeks91221312
      Number of events41103144
      Abbreviations: irAE, immune-related adverse events.
      In our 12-week landmark analysis, 427 patients were still alive: 98 patients had presented with irAE and 329 patients did not. Median OS was 21.2 months (95 % CI, 12.2 to 35.7) and 16.4 months (95 % CI, 12.4 to 21.3) for patients with and without irAEs, respectively (p = 0.26) (Fig. 1A). In our 12-week landmark analysis, 366 patients were still alive and did not have a progressive disease: 91 patients had presented with irAE and 275 patients did not. Median PFS was 14.3 months (95 % CI, 9.5 to 24.6) and 13.4 months (95 % CI, 10.2 to 18.3) for patients with and without irAEs, respectively (p = 0.66) (Fig. 1B).
      Figure thumbnail gr1
      Fig. 1Analysis of (A) overall survival and (B) progression-free survival in patients with and without irAE with 12-week landmark analysis. Survival was estimated using the Kaplan-Meier method in each group and groups were compared using the log-rank test. Abbreviations: irAE, immune-related adverse events.
      Sensitivity analyses according to histology (melanoma and NSCLC) and irAEs type (skin and thyroid) at 12-week landmark as well as 8 and 16-week landmark sensitivity analyses found no significant association between irAEs and OS or PFS (Supplementary 4 and 5). Multivariable analysis using a Cox model adjusted for age, sex, histology, ECOG, type of immunotherapy, number of previous lines, metastatic disease and, baseline corticosteroids and a 12-week landmark analysis, revealed that no specific irAE (skin, endocrine and gastro-intestinal irAE taken separately), was significantly associated with OS or PFS (Supplementary 6).

      Cox regression model with time-dependent covariate

      The occurrence of an irAE was significantly associated with OS: HR 0.56, 95 % CI [0.41; 0.75], p = 0.0001 (Table 3), and PFS: HR 0.63, 95 % CI [0.47; 0.83], p = 0.001 (Table 4).
      Table 3Analysis of the occurrence of any irAE and the effect of covariates on Overall Survival.
      Hazard Ratio95 % Confidence Intervalsp-value
      Age
      < 65 years old1*[0.73; 1.16]0.49
      ≥ 65 years old0.92
      Histology
      Melanoma1*[1.02; 2.20]0.04
      Lung1.50
      ECOG
      0 or 11*[1.32; 2.73]

      [2.68; 6.67]
      <0.0001
      21.90
      3 or 44.22
      Type of treatment
      Nivolumab1*[0.77; 1.59]0.58
      Pembrolizumab1.11
      Number of previous lines of treatment
      No previous lines1*[0.88; 1.67]

      [0.73; 1.68]
      0.44
      One or two lines1.21
      Three lines or more1.10
      Presence of metastases
      No1*[1.46; 3.47]0.0002
      Yes2.25
      Presence of corticosteroids at baseline
      No1*[1.81; 2.97]<0.001
      Yes2.32
      Occurrence of any IrAE#
      No1*[0.41; 0.75]0.0001
      Yes0.56
      Analysis was performed using a time-dependent Cox regression model Abbreviations irAE, immune-related adverse events# IrAE are of grade ≥ 2.
      Table 4Analysis of the occurrence of any irAE and the effect of covariates on progression-free survival.
      Hazard Ratio95 % Confidence Intervalsp-value
      Age
      < 65 years old1*[0.77; 1.18]0.65
      ≥ 65 years old0.95
      Histology
      Melanoma1*[1.12; 2.27]0.009
      Lung1.59
      ECOG
      0 or 11*[1.03; 2.08]

      [1.83; 4.47]
      <0.001
      21.47
      3 or 42.86
      Type of treatment
      Nivolumab1*[0.77; 1.49]0.70
      Pembrolizumab1.07
      Number of previous lines of treatment
      No previous lines1*[0.04; 1.72]

      [0.93; 2.02]
      0.22
      One or two lines1.28
      Three lines or more1.37
      Presence of metastases
      No1*[1.25; 2.59]0.002
      Yes1.80
      Presence of corticosteroids at baseline
      No1*[1.75; 2.79]<0.0001

      Yes2.21
      Occurrence of IrAE
      IrAE are of grade ≥ 2.
      No1*[0.47; 0.83]0.001
      Yes0.63
      Analysis was performed using a time-dependent Cox regression model. Abbreviations: irAE, immune-related adverse events.
      # IrAE are of grade ≥ 2.
      Sensitivity analyses regarding the main types of irAEs were performed. The occurrence of a cutaneous irAE was significantly associated with OS: HR 0.50, 95 % CI [0.32; 0.80], p = 0.0001 (Supplementary 7), and PFS: HR 0.59, 95 % CI [0.38; 0.90], p = 0.001 independently of the other covariates, in particular of the histological type. Similar results were found with thyroid irAE: OS: HR 0.50, 95 % CI [0.29; 0.86], p = 0.01 (Supplementary 8) and PFS: HR 0.49, 95 % CI [0.29; 0.84], p = 0.008.

      Use of systemic corticosteroids (CS) or immune-modulators and its impact on survival

      136 patients (23.6 %) were on CS at the time of ICBs initiation, among them 104 (76 %) had a dose superior to 20 mg per day. CS were prescribed for brain metastases (41 %), pain management (25 %), severe irAEs on previous ICB (9 %), respiratory symptoms (7 %), auto-immune disease (4 %) or occlusion (2 %) (9 % other indications, 3 % missing data).
      With a similar exposure to ICBs, 23/136 (16.9 %) patients with baseline CS versus 146/441 (33.1 %) patients without baseline CS, developed an irAE: ORR 2.43, 95 % CI [1.48–3.9] (p = 0.002) (Supplementary 9). We included CS at baseline as covariates in the Cox regression model and found a statistical significant association with worse overall survival: HR 2.32, 95 % CI [1.81; 2.97], p = <0.001, and progression-free survival: HR 2.21, 95 % CI [1.75; 2.79], p = <0.0001 (Table 3, Table 4).

      Discussion

      Our study is a large real-world cohort evaluating the association between prospectively collected irAEs and survival of patients treated with ICBs by considering the immortal-time bias using both landmark analysis and a Cox model with time-dependent covariate.
      With the Cox regression model with time-dependent covariate, we found a significant association between the occurrence of grade ≥ 2 irAEs and survival in patients treated with anti-PD-(L)1 for a melanoma or NSCLC. With the landmark analysis, no difference in OS or PFS was detected between patients who experienced grade ≥ 2 irAEs in the 12 first weeks of treatment and those who did not.
      The association between irAEs and survival in both melanoma and NSCLC patients is still unclear due to conflicting data in published studies.
      Our study illustrates well the importance of using a time-dependent method and selecting the appropriately powered method, while analyzing the association between irAEs and survival.
      Because ICB induced irAEs are typically delayed and occur between 3 and 4 weeks and up to 1 year after the first infusion of ICBs [
      • Weber J.S.
      • Hodi F.S.
      • Wolchok J.D.
      • et al.
      Safety profile of nivolumab monotherapy: a pooled analysis of patients with advanced melanoma.
      ], and because only patients who live long enough can develop irAEs, the occurrence of irAEs is a time-dependent variable that could induce an immortal time bias if analyzed as a fixed covariate. Established methodology such as landmark analysis, Cox model with time-varying variable or inverse-probability weighted models are classically used to eliminate this bias [
      • Cho I.S.
      • Chae Y.R.
      • Kim J.H.
      • et al.
      Statistical methods for elimination of guarantee-time bias in cohort studies: a simulation study.
      ,
      • Putter H.
      • van Houwelingen H.C.
      Understanding landmarking and its relation with time-dependent cox regression.
      ,
      • Giobbie-Hurder A.
      • Gelber R.D.
      • Regan M.M.
      Challenges of guarantee-time bias.
      ].
      Hence, previous published studies, not dealing with the time-varying nature of the occurrence of irAEs (no landmark analysis or time-varying Cox regression model) are biased [
      • Abu-Sbeih H.
      • Tang T.
      • Ali F.S.
      • et al.
      The impact of immune checkpoint inhibitor-related adverse events and their immunosuppressive treatment on patients’ outcomes.
      ,
      • Fujii T.
      • Colen R.R.
      • Bilen M.A.
      • et al.
      Incidence of immune-related adverse events and its association with treatment outcomes: the MD Anderson Cancer Center experience.
      ,
      • Judd J.
      • Zibelman M.
      • Handorf E.
      • et al.
      Immune-related adverse events as a biomarker in non-melanoma patients treated with programmed cell death 1 inhibitors.
      ,
      • Kim H.I.
      • Kim M.
      • Lee S.-H.
      • et al.
      Development of thyroid dysfunction is associated with clinical response to PD-1 blockade treatment in patients with advanced non-small cell lung cancer.
      ,
      • Rogado J.
      • Sánchez-Torres J.M.
      • Romero-Laorden N.
      • et al.
      Immune-related adverse events predict the therapeutic efficacy of anti-PD-1 antibodies in cancer patients.
      ,
      • Toi Y.
      • Sugawara S.
      • Kawashima Y.
      • et al.
      Association of immune-related adverse events with clinical benefit in patients with advanced non-small-cell lung cancer treated with nivolumab.
      ]. And although systematic reviews and meta-analysis are usually robust [
      • Fan Y.
      • Xie W.
      • Huang H.
      • et al.
      Association of immune related adverse events with efficacy of immune checkpoint inhibitors and overall survival in cancers: a systemic review and meta-analysis.
      ,
      • Hussaini S.
      • Chehade R.
      • Boldt R.G.
      • et al.
      Association between immune-related side effects and efficacy and benefit of immune checkpoint inhibitors – A systematic review and meta-analysis.
      ,
      • Petrelli F.
      • Grizzi G.
      • Ghidini M.
      • et al.
      Immune-related adverse events and survival in solid tumors treated with immune checkpoint inhibitors: a systematic review and meta-analysis.
      ,
      • Zhou X.
      • Yao Z.
      • Yang H.
      • et al.
      Are immune-related adverse events associated with the efficacy of immune checkpoint inhibitors in patients with cancer? A systematic review and meta-analysis.
      ], these should be interpreted cautiously as they included the above-mentioned studies not dealing with the immortal-time bias.
      The landmark analysis has the advantage to simply visualize survival curves using the Kaplan-Meier method with the complexity to obtain statistical power. Thus, selecting carefully the landmark time is important to detect a sufficient number of events while patients are still under treatment and alive. The Cox regression model with time-dependent covariate has increased statistical power over the landmark method by starting the analysis at the time of cohort entry, including all data and using all study follow up.
      In our study, using a Cox regression model with time-varying variable, we showed a significant association between the occurrence of grade ≥ 2 irAEs and survival in patients treated with anti-PD-(L)1 for a melanoma or NSCLC. In order to avoid bias due to population heterogeneity, histological type was included in the covariates used for the model. Our results are consistent with other studies using the time-dependent Cox regression model [

      Eggermont AMM, Kicinski M, Blank CU, et al. Association between immune-related adverse events and recurrence-free survival among patients with stage III melanoma randomized to receive pembrolizumab or placebo: a secondary analysis of a randomized clinical trial. JAMA Oncol Published Online First: 2 January 2020. doi:10.1001/jamaoncol.2019.5570.

      ,
      • Maillet D.
      • Corbaux P.
      • Stelmes J.-J.
      • et al.
      Association between immune-related adverse events and long-term survival outcomes in patients treated with immune checkpoint inhibitors.
      ].
      In studies dealing with the immortal-time bias by a landmark approach, the results are contradictory, due in part to the limits of the method. An association between irAEs and survival was found in patients treated with nivolumab for NSCLC [
      • Haratani K.
      • Hayashi H.
      • Chiba Y.
      • et al.
      Association of immune-related adverse events with nivolumab efficacy in non–small-cell lung cancer.
      ,
      • Ricciuti B.
      • Genova C.
      • De Giglio A.
      • et al.
      Impact of immune-related adverse events on survival in patients with advanced non-small cell lung cancer treated with nivolumab: long-term outcomes from a multi-institutional analysis.
      ] and melanoma [
      • Freeman-Keller M.
      • Kim Y.
      • Cronin H.
      • et al.
      Nivolumab in resected and unresectable metastatic melanoma: characteristics of immune-related adverse events and association with outcomes.
      ]. Conversely, no significant association between irAEs and survival was found in both NSCLC [
      • Owen D.H.
      • Wei L.
      • Bertino E.M.
      • et al.
      Incidence, risk factors, and impact on survival of immune-related adverse events in patients with non-small cell lung cancer.
      ] and melanoma patients [
      • Hua C.
      • Boussemart L.
      • Mateus C.
      • et al.
      Association of vitiligo with tumor response in patients with metastatic melanoma treated with pembrolizumab.
      ].
      In the landmark analysis of our study, to show a difference of OS with an HR = 0.56 (as found in the Cox model with time-varying covariates), the number of events required for 90 % power at 1-sided a = 0.025, is 126 events. At the landmark of 12 weeks, 201 events have occurred as shown in Table 2, and yet our landmark analysis, in contrast to the Cox regression model, showed no difference in OS or PFS between patients who experienced irAE during the first 12 weeks of treatment and those who did not. In the overall Kaplan-Meier curve with no landmark (Supplementary 10), the curves separate early, with a big survival drop in patients with no irAE. Because patients who died before the landmark time are excluded for the analysis, the landmark method did not measure the difference occurring early between the survival curves. Also, patients with an irAE occurring after the landmark, are classified in the no irAE group and contribute to the improvement of survival in this group. Selecting a different landmark time may change the classification of some subjects. These two points can explain the lack of significant results with the landmark analysis. The cox-regression model seems to be a good alternative to avoid both the problem of selecting a landmark time and any misclassification errors, and allows capturing all events since cohort entry. Another point is that irAE occurring in the first 12 weeks might not have an effect on survival beyond that point.
      Our study has limitations related to the specificities of our cohort. First, the REISAMIC database is an observational study on real-world data encompassing only clinically meaningful irAEs of grade ≥ 2. It waves the possible effect of grade 1 irAEs and underestimates the rates of irAEs. Patients do not undergo a full skin examination at each visit, thyroid functional tests and scan evaluations are less frequent than in clinical trials. There is a bias inherent to the lack of randomization. Finally, since PD-L1 has only been required in France for pembrolizumab use in NSCLC, we were able to retrieve it retrospectively for 81 patients only and could not include it in our multivariable analysis.
      In conclusion, in our real world study of patients with melanoma and NSCLC treated with anti-PD-(L)1 antibodies, we confirm that irAEs are associated with survival using a time-varying Cox regression model. We highlight the importance of addressing the immortal-time bias by using an appropriate methodology according to the number of patients, the type of treatment and disease and the kinetics of irAEs. Reviews including studies not addressing the immortal-time bias should be interpreted cautiously. Now that anti-PD(L)1 are combined either to other ICBs, chemotherapy or tyrosine-kinase inhibitors, studying the association between irAEs and survival is more complicated. Larger prospective studies with time-dependent analysis and immunological analysis are warranted to better elucidate the impact of irAEs on survival.

      CRediT authorship contribution statement

      Maria Kfoury: Conceptualization, Visualization. Marie Najean: Methodology, Software, Formal analysis, Visualization. Ariane Lappara: . Anne-Laure Voisin: Validation, Investigation. Stéphane Champiat: Investigation. Jean-Marie Michot: Investigation. Salim Laghouati: Project administration. Caroline Robert: Resources. Benjamin Besse: Resources. Jean-Charles Soria: Resources. Olivier Lambotte: Investigation. Christophe Massard: Supervision. Aurélien Marabelle: Supervision, Project administration. Matthieu Texier: Methodology, Software, Formal analysis, Visualization.

      Declaration of Competing Interest

      The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

      Acknowledgements

      We would like to thank the patients and their families who have agreed that this clinical data could be collected and reported. We also thank all the caregivers from Gustave Roussy and associated organ experts who treated those patients. We thank the radiology team from Gustave Roussy who performed the radiological assessments from these patients.

      Competing interests

      JCS declares:
      • In the last 2 years consultancy fees and shares from:
        • Relay Therapeutics.
        • Gritstone bio.
      • Board of Directors Hookipa Pharmaceuticals 2018-2021.
      • Full time employee at AstraZeneca 2017-2019.
      • Full time employee at Amgen August 2021-present.

      Funding

      The authors received no financial support for the research, authorship or publication of this article.

      Declarations

      The study was conducted in compliance with good clinical practice and the tenets of the Declaration of Helsinki. Constitution of the REISAMIC registry had been authorized by the Gustave Roussy Institutional Review Board and the French National Data Protection Commission (CNIL N° 2098694).

      Appendix A. Supplementary material

      The following are the Supplementary data to this article:

      References

        • Vaddepally R.K.
        • Kharel P.
        • Pandey R.
        • et al.
        Review of Indications of FDA-Approved Immune Checkpoint Inhibitors per NCCN Guidelines with the Level of Evidence.
        Cancers. 2020; 12: 738https://doi.org/10.3390/cancers12030738
      1. Das S, Johnson DB. Immune-related adverse events and anti-tumor efficacy of immune checkpoint inhibitors. J Immunother Cancer 2019;7. doi:10.1186/s40425-019-0805-8.

      2. Eggermont AMM, Kicinski M, Blank CU, et al. Association between immune-related adverse events and recurrence-free survival among patients with stage III melanoma randomized to receive pembrolizumab or placebo: a secondary analysis of a randomized clinical trial. JAMA Oncol Published Online First: 2 January 2020. doi:10.1001/jamaoncol.2019.5570.

        • Freeman-Keller M.
        • Kim Y.
        • Cronin H.
        • et al.
        Nivolumab in resected and unresectable metastatic melanoma: characteristics of immune-related adverse events and association with outcomes.
        Clin Cancer Res Off J Am Assoc Cancer Res. 2016; 22: 886-894https://doi.org/10.1158/1078-0432.CCR-15-1136
        • Haratani K.
        • Hayashi H.
        • Chiba Y.
        • et al.
        Association of immune-related adverse events with nivolumab efficacy in non–small-cell lung cancer.
        JAMA Oncol. 2018; 4: 374-378https://doi.org/10.1001/jamaoncol.2017.2925
        • Owen D.H.
        • Wei L.
        • Bertino E.M.
        • et al.
        Incidence, risk factors, and impact on survival of immune-related adverse events in patients with non-small cell lung cancer.
        Clin Lung Cancer Published Online First. August 2018; 22https://doi.org/10.1016/j.cllc.2018.08.008
        • Sato K.
        • Akamatsu H.
        • Murakami E.
        • et al.
        Correlation between immune-related adverse events and efficacy in non-small cell lung cancer treated with nivolumab.
        Lung Cancer. 2018; 115: 71-74https://doi.org/10.1016/j.lungcan.2017.11.019
        • Fan Y.
        • Xie W.
        • Huang H.
        • et al.
        Association of immune related adverse events with efficacy of immune checkpoint inhibitors and overall survival in cancers: a systemic review and meta-analysis.
        Front Oncol. 2021; 11633032https://doi.org/10.3389/fonc.2021.633032
        • Hussaini S.
        • Chehade R.
        • Boldt R.G.
        • et al.
        Association between immune-related side effects and efficacy and benefit of immune checkpoint inhibitors – A systematic review and meta-analysis.
        Cancer Treat Rev. 2021; 92102134https://doi.org/10.1016/j.ctrv.2020.102134
        • Petrelli F.
        • Grizzi G.
        • Ghidini M.
        • et al.
        Immune-related adverse events and survival in solid tumors treated with immune checkpoint inhibitors: a systematic review and meta-analysis.
        J Immunother. 2020; 43: 1-7https://doi.org/10.1097/CJI.0000000000000300
        • Zhou X.
        • Yao Z.
        • Yang H.
        • et al.
        Are immune-related adverse events associated with the efficacy of immune checkpoint inhibitors in patients with cancer? A systematic review and meta-analysis.
        BMC Med. 2020; 18: 87https://doi.org/10.1186/s12916-020-01549-2
        • Champiat S.
        • Lambotte O.
        • Barreau E.
        • et al.
        Management of immune checkpoint blockade dysimmune toxicities: a collaborative position paper.
        Ann Oncol. 2016; 27: 559-574https://doi.org/10.1093/annonc/mdv623
        • Weber J.S.
        • Hodi F.S.
        • Wolchok J.D.
        • et al.
        Safety profile of nivolumab monotherapy: a pooled analysis of patients with advanced melanoma.
        J Clin Oncol Off J Am Soc Clin Oncol. 2017; 35: 785-792https://doi.org/10.1200/JCO.2015.66.1389
        • Cho I.S.
        • Chae Y.R.
        • Kim J.H.
        • et al.
        Statistical methods for elimination of guarantee-time bias in cohort studies: a simulation study.
        BMC Med Res Methodol. 2017; 17: 126https://doi.org/10.1186/s12874-017-0405-6
        • Putter H.
        • van Houwelingen H.C.
        Understanding landmarking and its relation with time-dependent cox regression.
        Stat Biosci. 2017; 9: 489-503https://doi.org/10.1007/s12561-016-9157-9
        • Giobbie-Hurder A.
        • Gelber R.D.
        • Regan M.M.
        Challenges of guarantee-time bias.
        J Clin Oncol. 2013; 31: 2963-2969https://doi.org/10.1200/JCO.2013.49.5283
        • Abu-Sbeih H.
        • Tang T.
        • Ali F.S.
        • et al.
        The impact of immune checkpoint inhibitor-related adverse events and their immunosuppressive treatment on patients’ outcomes.
        J Immunother Precis Oncol. 2018; 1: 7https://doi.org/10.4103/JIPO.JIPO_12_18
        • Fujii T.
        • Colen R.R.
        • Bilen M.A.
        • et al.
        Incidence of immune-related adverse events and its association with treatment outcomes: the MD Anderson Cancer Center experience.
        Invest New Drugs. 2018; 36: 638-646https://doi.org/10.1007/s10637-017-0534-0
        • Judd J.
        • Zibelman M.
        • Handorf E.
        • et al.
        Immune-related adverse events as a biomarker in non-melanoma patients treated with programmed cell death 1 inhibitors.
        Oncologist. 2017; 22: 1232-1237https://doi.org/10.1634/theoncologist.2017-0133
        • Kim H.I.
        • Kim M.
        • Lee S.-H.
        • et al.
        Development of thyroid dysfunction is associated with clinical response to PD-1 blockade treatment in patients with advanced non-small cell lung cancer.
        Oncoimmunology. 2017; 7e1375642https://doi.org/10.1080/2162402X.2017.1375642
        • Rogado J.
        • Sánchez-Torres J.M.
        • Romero-Laorden N.
        • et al.
        Immune-related adverse events predict the therapeutic efficacy of anti-PD-1 antibodies in cancer patients.
        Eur J Cancer Oxf Engl. 1990; 2019: 21-27https://doi.org/10.1016/j.ejca.2018.10.014
        • Toi Y.
        • Sugawara S.
        • Kawashima Y.
        • et al.
        Association of immune-related adverse events with clinical benefit in patients with advanced non-small-cell lung cancer treated with nivolumab.
        Oncologist. 2018; https://doi.org/10.1634/theoncologist.2017-0384
        • Maillet D.
        • Corbaux P.
        • Stelmes J.-J.
        • et al.
        Association between immune-related adverse events and long-term survival outcomes in patients treated with immune checkpoint inhibitors.
        Eur J Cancer. 2020; 132: 61-70https://doi.org/10.1016/j.ejca.2020.03.017
        • Ricciuti B.
        • Genova C.
        • De Giglio A.
        • et al.
        Impact of immune-related adverse events on survival in patients with advanced non-small cell lung cancer treated with nivolumab: long-term outcomes from a multi-institutional analysis.
        J Cancer Res Clin Oncol. 2019; 145: 479-485https://doi.org/10.1007/s00432-018-2805-3
        • Hua C.
        • Boussemart L.
        • Mateus C.
        • et al.
        Association of vitiligo with tumor response in patients with metastatic melanoma treated with pembrolizumab.
        JAMA Dermatol. 2016; 152: 45-51https://doi.org/10.1001/jamadermatol.2015.2707