Background Several studies have reported the incidence of immune-related adverse events (irAEs) as a predictor of the efficacy of anti-programmed cell death protein 1 antibodies in patients with cancer. However, immortal time bias has not always been fully addressed in these studies. In this retrospective multicenter study, we assessed the association between the incidence of irAEs and the efficacy of pembrolizumab in urothelial carcinoma (UC) using time-dependent analysis, an established statistical method to minimize immortal time bias.
Methods The study included 176 patients with advanced UC who underwent pembrolizumab treatment at seven affiliated institutions between January 2018 and July 2020. Patients with irAEs were compared with those without irAEs in terms of overall survival (OS) and cancer-specific survival (CSS). Immortal time bias was eliminated by using time-dependent analysis.
Results Of the 176 patients, irAEs occurred in 77 patients (43.8%), with a median of 60 days. The irAEs (+) cohort showed significantly favorable OS and CSS compared with the irAEs (−) cohort (p=0.018 and p=0.005, respectively), especially in the cohort with grade 1–2 irAEs (OS and CSS; p=0.003 and p=0.002, respectively). Multivariate analyses identified any irAEs and grade 1–2 irAEs as independent favorable prognostic factors for OS and CSS.
Conclusion Even after minimizing immortal time bias by time-dependent analysis, the incidence of irAEs, especially grade 1–2 irAEs, could be a significant predictor of favorable prognoses in patients with UC who have undergone pembrolizumab treatment.
- tumor biomarkers
- urinary bladder neoplasms
- urologic neoplasms
Data availability statement
Data are available upon reasonable request. The raw data underlying this study are available from the corresponding author upon reasonable request.
This is an open access article distributed in accordance with the Creative Commons Attribution Non Commercial (CC BY-NC 4.0) license, which permits others to distribute, remix, adapt, build upon this work non-commercially, and license their derivative works on different terms, provided the original work is properly cited, appropriate credit is given, any changes made indicated, and the use is non-commercial. See http://creativecommons.org/licenses/by-nc/4.0/.
Statistics from Altmetric.com
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.
Immune checkpoint inhibitors (ICIs), including programmed cell death protein 1 (PD-1) antibodies, have led to remarkable advances in the treatment of various types of cancers. Pembrolizumab, a PD-1 antibody, has been shown to prolong overall survival (OS) in patients with advanced urothelial carcinoma (UC) following disease progression after platinum-containing chemotherapy.1 Pembrolizumab has thus been established as a second-line treatment for advanced UC.2 3 However, ICIs can produce immune-related adverse events (irAEs), such as rash, colitis, hepatitis, endocrinopathies, pneumonitis, myositis, and nephritis.4 5
Several previous studies have reported that the incidence of irAEs is associated with a good therapeutic response to ICIs in malignancies,6 7 including melanoma,8 non-small cell lung cancer,9 10 gastric cancer,11 12 and UC.13–15 However, immortal time bias has not always been fully addressed in these studies. We previously reported the correlation between irAEs and favorable efficacy of pembrolizumab in patients with UC using single-center preliminary data13; however, immortal time bias was not eliminated.
Immortal time bias occurs when a group of patients does not survive long enough to receive an intervention.16–18 Thus, irAE development and patient survival occur because the shorter the survival, the lower the chance of an irAE development. Time-dependent analysis (or called Mantel–Byar method) and landmark analysis are both established methods for avoiding immortal time bias, whereas the former is much superior to the latter in minimizing the immortal time bias.16 To the best of our knowledge, no previous studies have conducted time-dependent analysis to assess the association between irAEs and response to ICIs. In the present study, we assessed the association between irAEs and the efficacy of pembrolizumab in patients with UC using multicenter data. Immortal time bias was eliminated by conducting time-dependent analysis.
This retrospective study included 176 patients with advanced (metastatic or locally advanced) UC who underwent pembrolizumab treatment. Of the 176 patients, 175 patients were treated with pembrolizumab as a second-line therapy following disease progression after first-line chemotherapy, and the remaining one patient was treated as a first-line therapy because he was on hemodialysis and ineligible for platinum-containing chemotherapy. The study cohort includes 150 patients who were analyzed in our previous article, which assessed the association of the albumin-to-globulin ratio with conventional oncological outcomes in the setting of advanced UC treated with pembrolizumab.19 All patients received pembrolizumab at a dose of 200 mg/body intravenously every 3 weeks at seven affiliated institutions between January 2018 and July 2020.
The patients’ clinical data were retrospectively extracted as follows: sex, age, occurrence and grade of irAEs, use of glucocorticoids for irAEs, OS, and cancer-specific survival (CSS). The time for irAEs to occur following the start of pembrolizumab treatment was also noted. The grade of irAEs was evaluated according to the Common Terminology Criteria for Adverse Events V.5.0.20 Furthermore, to guarantee the quality of data on irAE grades, an investigator at each institution retrospectively reviewed and interpreted clinical records to determine irAE grades, and another investigator at the same institution double-checked the results. Follow-up information was obtained in October 2020.
The maximum therapeutic effect and frequency of glucocorticoid use was analyzed using the χ2 test or Fisher’s exact test. Survival data of patients were analyzed using time-dependent analysis, whereby the survival time of each patient who experienced irAEs from time of starting pembrolizumab treatment to time of final observation was divided into time from the start of pembrolizumab to the onset of initial irAE and time after the onset of initial irAE. The former time period was assigned to the irAEs (−) cohort and the latter period to the irAEs (+) cohort. Meanwhile, regarding patients who did not experience irAEs, all of the survival time was assigned to the irAEs (−) cohort (figure 1). Survival curves were generated using the Kaplan-Meier method and compared using log-rank tests. Similar time-dependent analyses were performed for each group of irAE grades (grade 1–2 and grade 3–5), each type of irAEs, and each route of glucocorticoid administration. Time-dependent Cox model was used for multivariate analyses of OS and CSS. All statistical analyses, except time-dependent multivariate Cox analyses, were performed using the JMP Pro software (V.15.0.0, SAS Institute). Time-dependent multivariate Cox analyses were conducted by a biostatistician (YU) using SAS V.9.4. All p values were two-sided and considered significant at p<0.05.
The clinical characteristics of patients at the start of pembrolizumab treatment are shown in table 1. A total of 176 patients comprised 132 men (67.2%) and 44 women (32.8%) with a median age of 71 years (IQR, 66–76 years). The median follow-up duration was 8.1 months (IQR, 4.0–15.2 months). No patients had any baseline immune dysfunction, immunosuppressive medication, or pre-existing autoimmune condition.
Incidence of irAEs
As shown in table 2, irAEs occurred in 77 patients (43.8%) with a median onset of 60 days (IQR, 25–126 days) from the initiation of pembrolizumab treatment. Grade 1–2 and grade 3–5 irAEs occurred in 35.8% and 11.9% of patients, respectively. The most common irAEs were skin disorders (eg, pruritus, rash, and dermatitis; 22.2%), followed (in decreasing order of prevalence) by endocrine disorders (eg, thyroid dysfunction, adrenal insufficiency, and type 1 diabetes; 13.6%), respiratory disorders (eg, interstitial pneumonia; 5.7%), intestinal disorders (eg, diarrhea and colitis; 5.1%), hepatitis, nephritis, stomatitis, myositis, and myocarditis. When glucocorticoid administration was required due to irAEs, treatments were administered externally, orally, or intravenously (33.8%, 28.6%, and 14.3%, respectively). Patients with respiratory disorders required frequent oral or intravenous administration of glucocorticoids (60% and 40%, respectively).
The Kaplan-Meier curves with survival analyses are shown in figures 2 and 3 (and in online supplemental figures 1 and 2). Time-dependent univariate analyses of patient survival among characteristics of irAEs are shown in table 3. The irAEs (+) cohort showed significantly favorable OS and CSS compared with the irAEs (−) cohort (HR=0.59, p=0.018 and HR=0.52, p=0.005, respectively; figure 2). The grade 1–2 irAEs (+) cohort showed significantly favorable OS and CSS compared with the cohort without grade 1–2 irAEs (HR=0.49, p=0.003 and HR=0.47, p=0.002, respectively; figure 3). Among the types of irAEs, the cohort with respiratory disorders showed significantly poor OS compared with the cohort without respiratory disorders (HR=2.66, p=0.011; online supplemental figure 1). Among the routes of glucocorticoid administration, the cohort with irAEs requiring intravenous administration showed significantly poor OS compared with cohort without the intravenous administration (HR=2.12, p=0.029; online supplemental figure 2). Multivariate analyses of patient survival using time-dependent Cox model are shown in table 4. The incidence of any irAEs and grade 1–2 irAEs was identified as independent favorable prognostic factor for OS and CSS along with Eastern Cooperative Oncology Group Performance Status of 0, primary site of the bladder, and number of lesions of 4 or less.
Immortal time bias is an often-overlooked phenomenon in time-dependent studies. Landmark analysis is a method that was proposed by Anderson et al21 in the 1980s. A fixed time after the initiation of therapy is selected as the landmark, and patients who die before the landmark time are excluded from the analysis. Although this approach is effective in removing the immortal time bias, it is inferior in that the results differ depending on the choice of the landmark.18 Time-dependent analysis (proposed by Mantel and Byar22) has proved to be superior to landmark analysis in minimizing immortal time bias in observational studies of survival outcomes.16 Previous studies that assessed the incidence of irAEs and the efficacy of ICIs have either not considered immortal time bias at all6 7 13–15 or they have avoided using landmark analysis.8–12 The present study is the first study to assess the association between the incidence of irAEs and the efficacy of ICIs using time-dependent analysis. In the present study, the survival time of irAEs (+) cohort was calculated only after the onset of initial irAE according to time-dependent analysis. Given that the median time from the start of pembrolizumab to the onset of initial irAE was 60 days, the survival time of irAEs (+) cohort might be estimated to be approximately 2 months shorter than its actual duration. Time-dependent analysis could therefore be considered as a more ‘conservative’ approach than landmark analysis because it should underestimate the survival time of the irAE (+) group. In other words, showing the superiority of the irAE (+) group by time-dependent analysis could be regarded as more ‘impressive’ than by landmark analysis.
ICIs generate irAEs by unbalancing the immune system in patients. From another point of view, irAEs may suggest the good responsiveness of the patient’s immune system to ICIs. In this study, we found that the incidence of irAEs correlated with the favorable therapeutic effect of pembrolizumab in patients with advanced UC. The irAEs (+) cohort showed significantly favorable OS and CSS compared with the irAEs (−) cohort. Our result is significant in that the incidence of irAE correlates with favorable survival even if the immortal time bias has been eliminated by means of time-dependent analysis as detailed above.
We found that the cohort with grade 1–2 irAEs showed significantly favorable OS and CSS, whereas the cohort with grade 3–5 irAEs showed a tendency towards poor OS. The greater severity of irAEs in the latter cohort may account for the shortening of OS in the cohort. Moreover, the cohort with respiratory disorders and the cohort with intravenous administration of glucocorticoids both showed significantly poor OS. Poor OS in the latter cohort may be a coincidental reflection of the requirement to administer intravenous glucocorticoids when irAEs become life threatening. Faje et al23 reported that high-dose glucocorticoids for irAEs were associated with reduced survival in patients who received ICIs. Our results are consistent with those of their study because intravenous administration of glucocorticoids is usually high dose. It is controversial whether high-dose glucocorticoids reduce the antitumor effects of ICIs.23 24 In our study, the cohort with intravenous administration of glucocorticoids did not show significantly poor CSS. However, it is imperative that intravenous administration of glucocorticoids be performed urgently because delayed treatment can be fatal. Similarly, poor OS in the cohort with respiratory disorders is because such disorders are frequently life-threatening and require intravenous glucocorticoid treatment. Although the proportion of grade 3–5 respiratory disorders was similar to that of endocrine disorders (40.0% vs 41.7%, respectively), respiratory disorders were the more life-threatening ones of the irAEs, resulting in requiring more frequent intravenous glucocorticoid use.
Interestingly, multivariate analyses of patient survival revealed that primary site of the upper tract had significantly poor OS and CSS compared with the bladder. Advanced upper tract urothelial carcinoma (UTUC) has different molecular and genetic features from the most common carcinoma of the bladder,25 suggesting a possible different sensitivity to ICI. However, UTUC is under-represented in clinical trials because of its minority.26 The significant results obtained in this study may be due to the high UTUC ratio of 44%.
Our study had some limitations. First, it was retrospective and did not represent prospective clinical trials. Second, the sample size was small; although compared with previous studies that have reported on the correlation between irAEs and efficacy in UC, this study has been the largest. Further studies with larger populations are required to validate our results.
Even after minimizing immortal time bias by time-dependent analysis, the incidence of irAEs, especially irAEs of grade 1–2, could be a significant predictor of favorable prognoses in patients with UC who have undergone pembrolizumab treatment.
Data availability statement
Data are available upon reasonable request. The raw data underlying this study are available from the corresponding author upon reasonable request.
Patient consent for publication
This study involves human participants and was approved by the Institutional Review Board of The University of Tokyo (approval number: 10565). Because of the retrospective design of the study, written informed consent was waived. Each participant was given the opportunity to decline participation in the study through the opt-out form on our website.
We would like to thank Editage (www.editage.com) for English language editing.
Contributors T Kawai and S Taguchi contributed to the conception, study design, analysis, and interpretation of data, and drafted the first manuscript. TN contributed to the conception and study design, supervised the study, and revised the manuscript critically for important intellectual content. JK, YN, DO, KY, T Kaneko, SK, MT, and YS contributed to acquisition of data. YU contributed to the analysis and interpretation of data. TF, HF, YE, HN, S Takahashi, and HK supervised the study, helped to draft the manuscript, and were involved in revising it critically for important intellectual content. All authors read and approved the final manuscript. T Kawai and S Taguchi accept full responsibility for the overall work as the guarantors.
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 None declared.
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.