Cutaneous Adverse Events and Cancer Survival Prognosis With Immune Checkpoint Inhibitor Treatment: A Systematic Review and Meta-Analysis

Key Points

Question  Are cutaneous immune-related adverse events (cirAEs) and their subtypes associated with improved cancer prognosis among patients receiving immune checkpoint inhibitor (ICI) treatment?

Findings  In this systematic review and meta-analysis including 23 articles with a total of 22 749 patients treated with ICIs, those with cirAEs exhibited extended overall survival and progression-free survival (PFS). Consistent results were observed across all subgroups stratified by cirAE subtype, study design, geographic region, ICI type, and cancer type except for PFS analysis in the US.

Meaning  These findings suggest that cirAEs may have prognostic value during ICI treatment and may provide valuable information for diagnostic assessment and subsequent ICI treatment.

Importance  Growing research suggests that the prevalence of cutaneous immune-related adverse events (cirAEs) is associated with favorable outcomes among individuals with cancer who receive immune checkpoint inhibitor (ICI) treatment.

Objective  To identify whether the presence of cirAEs and their subtypes subsequent to ICI administration is associated with enhanced cancer prognosis.

Data Sources  The PubMed, Embase, Cochrane Library, and Web of Science databases were searched for publications examining the association between cirAE development during ICI treatment and subsequent cancer prognosis. The initial search was limited to English-language publications from database inception until December 31, 2022; a subsequent search was performed on May 21, 2023.

Study Selection  Two reviewers independently scrutinized the identical articles and included those that constituted original research evaluating the association between cirAE development and cancer prognosis.

Data Extraction and Synthesis  The search terms, study objectives, and methodological protocols were defined before study initiation. The aforementioned 2 reviewers performed data extraction independently and resolved discrepancies through agreement. This study followed the Preferred Reporting Items for Systematic Reviews and Meta-analysis and the Meta-analysis of Observational Studies in Epidemiology reporting guidelines. The protocol was prospectively registered with PROSPERO. Data analyses were conducted between May 21 and June 1, 2023.

Main Outcomes and Measures  The major outcome end points were overall survival (OS) and progression-free survival (PFS). Subgroup analyses were also conducted according to cirAE type, cancer type, geographic region, study design, and ICI type. Given the heterogeneity inherent in the included studies, a DerSimonian-Laird random-effects model was adopted.

Results  This systematic review and meta-analysis included 23 studies with a total of 22 749 patients treated with ICIs. The occurrence of cirAEs was associated with improved OS (hazard ratio [HR], 0.61 [95% CI, 0.52-0.72]; P < .001) and PFS (HR, 0.52 [95% CI, 0.41-0.65]; P < .001). Consistent results were observed across all subgroups stratified by study design, geographic region, ICI type, and cancer type, aligning with the overall estimate of OS and PFS improvement. However, no statistically significant differences were identified in terms of PFS within studies conducted in the US.

Conclusions and Relevance  In this systematic review and meta-analysis, the presence of cirAEs and their subtypes was associated with improved prognosis for individuals with cancer undergoing ICI treatment. These findings suggest that cirAEs may have useful prognostic value in ICI treatment.

Over the past decade, immune checkpoint inhibitors (ICIs) have revolutionized the therapeutic landscape for patients with advanced malignant tumors. Immune checkpoints, such as cytotoxic T-lymphocyte–associated protein 4 (CTLA-4), programmed cell death 1 (PD-1), and programmed cell death ligand 1 (PD-L1), have been effectively targeted using antagonist antibodies.^1,2 Presently, ICIs assume pivotal roles in the treatment of melanoma, non–small cell lung cancer (NSCLC), and other solid tumors.³

Regardless of their therapeutic effectiveness, ICIs can cause immunotoxicity as a result of increased immunologic activation, which manifests as an immune-related adverse event (irAE). Broadly speaking, irAEs may affect any organ or system at any given point during or subsequent to discontinuation of ICI treatment.² Cutaneous irAEs (cirAEs) are the most commonly reported toxic effects, affecting approximately 30% to 60% of all treated patients.⁴ An extensive array of clinical manifestations has been documented, encompassing exanthems, lichenoid reactions, dermatitis, psoriasis, vitiligo, bullous disorders, and Stevens-Johnson syndrome or toxic epidermal necrolysis.⁵ A vital question arises regarding whether the occurrence of distinct cirAE types is associated with clinical outcomes of patients undergoing ICI treatment. Several studies have noted a positive association of various cirAE subtypes with tumor response and survival.^6-8

Previous meta-analyses have predominantly concentrated on the association between all irAE categories and treatment effectiveness. However, to our knowledge, the precise association between cirAEs, including their distinct subtypes, and cancer survival remains a topic of debate.⁹ As a result, we conducted this comprehensive and up-to-date systematic review and meta-analysis to determine whether the prevalence of cirAEs and their specific subtypes following ICI treatment is associated with improved outcomes among patients with cancer. This study aimed to improve understanding and provide valuable insight into individualized treatment.

Southeast University deemed this systematic review and meta-analysis exempt from review and waived informed consent because only publicly available, previously published studies were used. This research was carried out in compliance with the Preferred Reporting Items for Systematic Reviews and Meta-analysis (PRISMA) and Meta-analysis of Observational Studies in Epidemiology (MOOSE) reporting guidelines. The study protocol was prospectively registered with PROSPERO (CRD42023390967).

We searched the PubMed, Embase, Cochrane Library, and Web of Science databases for studies published between database inception and December 31, 2022. The following search terms were used for the initial search: cutaneous adverse event OR cutaneous immune-related adverse event AND PD-1, PD-L1, CTLA-4, immune-checkpoint inhibitor, OR ICI AND prognosis OR survival. On May 21, 2023, we conducted an updated electronic search using the same 4 databases and expanded the key terms to include the following: cutaneous adverse event, cutaneous immune-related adverse event, dermatological adverse event, OR mucosal adverse event AND PD-1, PD-L1, CTLA-4, immune-checkpoint inhibitor, OR ICI AND prognosis, survival, mortality, OR outcome.

Two researchers (Y.D. and W.W.) independently conducted a rapid screening of titles and abstracts. They then carried out a full-text search to identify relevant studies. Finally, they screened the citations of the relevant studies to identify any additional eligible studies.

The following criteria were considered essential for study inclusion: (1) patients with malignant tumors who received ICI treatment; (2) analyses that provided hazard ratios (HRs) of overall survival (OS) or progression-free survival (PFS) comparing patients with cirAEs with patients without; (3) assessment of adverse events based on the National Cancer Institute Common Terminology Criteria for Adverse Events (version 3.0, 4.0, or 5.0), with cirAE diagnosis and severity determined through clinical examinations, biological markers, and imaging data; (4) be a prospective, retrospective, or randomized clinical trial; (5) availability of HRs with 95% CIs of survival data; and (6) research published in English-language journals.

Reviews, case reports, guidelines, editorials, and conference abstracts were excluded from consideration. Final determination of study eligibility and inclusion was reached through consensus among both reviewers (Y.D. and W.W.).

The same 2 reviewers (Y.D. and W.W.) extracted data independently and resolved any differences through consensus. The following data were extracted for each study: first author’s name, publication year, geographic region of the original study, study design, ICI type, cancer type, number of patients, sex, specific cirAE types, landmark analysis, and outcomes. The primary end points were PFS and OS among patients with cirAEs and patients without cirAEs assessed as HRs, respectively. In cases in which HR data were not directly provided, Kaplan-Meier curves were digitized using Engauge Digitizer open-source software to estimate HRs.^10,11

Newcastle-Ottawa Scale (NOS) criteria were used to assess the design, conduct, and reporting of the included studies. Two researchers (Y.D. and W.W.) independently assessed the quality items and potential differences in scoring. Each study was assigned a score for methodological quality ranging from 0 (poor) to 9 (best) based on criteria related to study cohort selection, comparability, and outcomes. Any discrepancies were resolved through consensus involving the participation of a third author (M.C.). Trials that provided HRs for PFS or OS following ICI treatment, either directly or indirectly, were considered qualified studies.

Pooled HRs and their corresponding 95% CIs were calculated to assess the association of cirAE occurrence with OS and PFS. An HR greater than 1 indicated a poor prognosis, whereas an HR less than 1 indicated a favorable prognosis.

Fixed- and random-effects meta-analyses were both performed. Based on the final collection of trials reporting OS and/or PFS data for patients with or without cirAEs, the general inverse-variance–weighted technique was applied to assess the overall effect size. I² statistics were used to determine the percentage of variance in the data that might be attributable to heterogeneity rather than sampling error. Low heterogeneity was indicated by an I² score of 25% or less. Given the relatively small size of our final research sample and the moderate levels of heterogeneity, all results are presented using the DerSimonian-Laird random-effects model. Subgroup analyses were carried out to investigate the association of cirAEs with OS and PFS based on study design, geographic area, cancer type, and cirAE subtype, as long as at least 2 studies were available. The cirAE subtypes were classified using Delphi consensus definitions,¹² including psoriasis, lichenoid or lichen planus-like lesions, bullous pemphigoid, eczematous, vitiligo, erosive mucocutaneous, Grover disease, exanthem, eruptive atypical squamous proliferation, and pruritus without rash. Vague descriptions of cirAE types were not included in the subgroup analysis.

Assessment of publication bias involved the use of Begg funnel plots and the Egger test. A significance level of P < .05 (2-sided) indicated the presence of publication bias in both tests. Moreover, we used the trim-and-fill method to identify and address potential publication bias.¹³ This involved estimating the number of potentially missing studies and incorporating hypothetical studies into the original analysis, thus deriving an adjusted effect size.

To assess robustness and potential bias in estimating OS, we conducted a sensitivity analysis wherein each study was sequentially excluded. Furthermore, we performed an additional sensitivity analysis excluding studies with a high risk of bias, determined based on study type, sample size, and result reliability.

All statistical analyses were conducted using Stata, version 17.0 (StataCorp). P < .05 was considered statistically significant. Data analyses were conducted between May 21 and June 1, 2023.

A total of 1345 citations were identified in the initial search (Figure 1). Through comprehensive analysis of the titles and abstracts, we identified 38 articles that examined the association between ICI treatment and cirAE prognostic value. Seventeen articles did not meet the inclusion criteria and were thus eliminated. Two additional studies were discovered through a search of the citations of relevant studies. The meta-analysis comprised 23 articles that satisfied the established criteria, involving a total of 22 749 patients.^6,9,14-34

The eTable in Supplement 1 provides an overview of the study characteristics. The majority of studies (n = 21) were retrospective,^6,9,14-34 whereas 2 were prospective^6,28 and 1 was a randomized clinical trial.¹⁶ All included studies examined the association between cirAE development and cancer prognosis. A total of 7 studies included only OS data,^9,14-19 3 studies provided only PFS data,^6,20,21 and 13 studies provided both OS and PFS data.^22-34 In terms of ICI type, most eligible studies focused on PD-1 inhibitors and/or PD-L1 inhibitors^{6,9,14,15,17,18,20-34}; only 2 studies examined CTLA-4 inhibitors.^16,19 With regard to geographic region, 8 studies were conducted in Asia-Pacific countries,^{6,15,18,20,22,23,25,26} 8 in European countries,^{17,21,24,28-30,32,34} and 7 in the US.^{9,14,16,19,27,31,33} In terms of cancer type, 9 studies focused on melanoma,^{6,14-17,22,30,32,34} 7 on NSCLC,^{23-26,28,29,31} and 1 on metastatic renal cell carcinoma²¹; the remaining 6 included mixed tumor types.^{9,18-20,27,33}

We used NOS criteria to evaluate the quality of each study included in the meta-analysis. Ratings ranged from 6 to 9 (mean [SD] rating, 7.26 [0.94]) (eTable in Supplement 1). Based on the NOS criteria, all studies met the minimum quality threshold and were included in our comprehensive meta-analysis.

Twenty articles examined cirAE development and its association with OS among patients receiving ICI treatment.^{9,14-19,22-34} Sixteen articles evaluated the association between cirAEs and PFS.^6,20-34 For example, Min Lee et al²⁷ reported that OS and PFS data were specifically related to the dermatitis subtype. Bottlaender et al³⁰ provided OS and PFS data for 2 cohorts (case and control), along with OS data for 2 distinct cirAE types. Thompson et al³¹ reported OS and PFS data for multiple cirAE types. Nakamura et al,²² Dousset et al,³² and Nardin et al³⁴ included OS and PFS data specifically for the vitiligo subtype. Nelson et al³³ presented OS and PFS data for the bullous pemphigoid subtype. Tang et al⁹ and Zhang et al¹⁹ reported OS data for 17 and 10 cirAE types, respectively.

Our meta-analysis of OS included 16 trials,^{9,14-19,23-31} and an association between cirAE occurrence and longer OS (HR, 0.61 [95% CI, 0.52-0.72]; I² = 70.7%; P < .001) was observed (Figure 2A). The meta-analysis of PFS included 12 articles,^{6,20,21,23-31} and we observed that patients who experienced cirAEs after ICI treatment had better PFS than those who did not (HR, 0.52 [95% CI, 0.41-0.65]; I² = 45.1%; P < .001) (Figure 2B).

Landmark analysis was used to address guarantee-time bias in the evaluation of the association between treatment and survival. Results (using a minimum of 6 weeks) were available for 11 studies with data for OS^{9,14-16,18,19,23,24,26,29,31} and for 8 studies with corresponding data for PFS.^{6,21,23-26,29,31} Meta-analysis of these selected studies revealed a significant advantage in terms of PFS (HR, 0.65 [95% CI, 0.54-0.79]; P < .001), with minimal heterogeneity (I² = 0%). Moreover, patients who experienced cirAEs demonstrated improved OS (HR, 0.69 [95% CI, 0.59-0.80]; I² = 65.8%; P < .001) compared with those who did not (Figure 3).

In the subgroup analysis based on cirAE subtypes,¹² patients who had one of the following cirAE subtypes had a significant advantage in OS compared with participants who did not: eczematous (HR, 0.69 [95% CI, 0.50-0.93]; I² = 0%; P = .02), lichenoid or lichen planus-like skin lesions (HR, 0.51 [95% CI, 0.38-0.69]; I² = 0%; P < .001), pruritus without rash (HR, 0.70 [95% CI, 0.62-0.79]; I² = 0%; P < .001), psoriasis (HR, 0.63 [95% CI, 0.47-0.84]; I² = 6.3%; P = .001), or vitiligo (HR, 0.30 [95% CI, 0.19-0.46]; I² = 55.5%; P < .001). For PFS, vitiligo was the only cirAE subtype with a significant advantage (HR, 0.28 [95% CI, 0.19-0.43]; I² = 16.5%; P < .001). However, no statistically significant difference was observed for the bullous pemphigoid and exanthem subtypes (Table).

The included studies were from the Asia-Pacific, Europe, and US regions. For OS, a significant advantage was observed for patients in Asia-Pacific countries (HR, 0.48 [95% CI, 0.39-0.60]; I² = 10.0%; P < .001), Europe (HR, 0.47 [95% CI, 0.30-0.72]; I² = 48.1%; P = .001), and the US (HR, 0.81 [95% CI, 0.71-0.91]; I² = 49.5%; P = .001) who were treated with ICIs and developed cirAEs. For PFS, a significant advantage was also observed for these patients in Asia-Pacific countries (HR, 0.51 [95% CI, 0.38-0.68]; I² = 0%; P < .001) and Europe (HR, 0.49 [95% CI, 0.34-0.70]; I² = 52.1%; P < .001) but not in the US (HR, 0.56 [95% CI, 0.18-1.81]; I² = 86.0%; P = .34) (Table).

The included studies were classified as retrospective, prospective, or randomized clinical trials. Because of the paucity of data, we only examined retrospective studies for OS (HR, 0.61 [95% CI, 0.51-0.73]; I² = 72.2%; P < .001) and PFS (HR, 0.56 [95% CI, 0.44-0.70]; I² = 39.4%; P < .001). Consistent findings for PFS were observed for the prospective trials (HR, 0.33 [95% CI, 0.16-0.68]; I² = 43.0%; P = .03).

Only 1 study¹⁶ reported on OS or PFS associated with anti–CTLA-4 treatment combined with various ICI types. Consequently, we performed a subgroup analysis specifically focusing on OS and PFS outcomes of PD-1 or PD-L1 inhibitors, and we observed that cirAEs were associated with improved OS and PFS (Table).

Among the cancer types analyzed in this meta-analysis, the most frequently included were melanoma (9 studies) and NSCLC (7 studies). The pooled data analysis revealed a noteworthy association between cirAE incidence and enhanced OS (HR, 0.51 [95% CI, 0.37-0.71]; I² = 55.1%; P < .001) and PFS (HR, 0.45 [95% CI, 0.32-0.65]; I² = 0%; P < .001) among patients with melanoma. Similarly, patients with NSCLC who received ICI treatment also demonstrated prolonged OS (HR, 0.50 [95% CI, 0.33-0.77]; I² = 66.9%; P = .002) and PFS (HR, 0.61 [95% CI, 0.46-0.80]; I² = 45.4%; P < .001).

To assess the association between each study and overall risk estimates, sensitivity analyses were conducted by systematically excluding individual studies. Irrespective of the particular study omitted, the associations between the presence of cirAEs and the prognosis of patients receiving ICI treatment, as determined by OS and PFS, remained consistent (eFigures 1 and 3 in Supplement 1). Furthermore, the results of the sensitivity analysis focusing on studies with low risk of bias were consistent with the overall estimate, supporting the validity of the findings (eFigures 2 and 4 in Supplement 1).

Begg funnel plots and the Egger test were used to evaluate publication bias. In the case of PFS data, visual analysis of the funnel plot (eFigure 6 in Supplement 1) revealed publication bias, which was validated by the Begg test and the Egger test. Similarly, the funnel plot for OS showed asymmetry that was consistent with results of the Egger test but not the Begg test (eFigure 5 in Supplement 1). The trim-and-fill method was used to analyze whether publication bias affected the combined results. Although 4 studies were imputed for OS and 5 studies were imputed for PFS, the resulting output did not substantially deviate from the original results, indicating the robustness and stability of the findings (eFigures 7 and 8 in Supplement 1).

Although numerous systematic reviews and meta-analyses have investigated the emergence of irAEs and their association with treatment effectiveness and survival outcomes, there has been a paucity of reviews examining potential associations between cirAE emergence and prognosis among patients receiving ICI treatment. To date, the precise association between cirAEs and prognosis in the context of various ICI types, cancer types, cirAE subtypes, or geographic regions remains controversial. This extensive meta-analysis included 23 studies involving a total of 22 749 patients. The findings suggest that patients with cancer who were treated with ICIs and developed cirAEs had substantially longer OS and improved PFS compared with those who were not. These findings are consistent with previous meta-analyses that reported an association between cirAE prevalence and improved PFS and OS. Hence, the emergence of cirAEs can serve as a prognostic factor for patients with cancer undergoing ICI treatment.^2,35-38 Additionally, to minimize potential bias associated with the varying durations of ICI exposure, we performed a landmark analysis for both PFS and OS among patients with a minimum exposure period of 6 weeks. The findings supported the hypothesis that the presence of cirAEs is associated with long-term therapeutic advantages (PFS and OS) among patients treated with ICIs.

The association between cirAE subtypes and prognosis was also investigated. The majority of cirAEs manifested as vitiligo, which exhibited the most favorable OS and PFS outcomes. These findings aligned with previous studies examining the association between vitiligo and prognosis among patients with melanoma.³⁹ Although the precise underlying mechanism remains unknown, Zhang et al¹⁹ suggested that epitope spreading leading to clonal diversification of T-cell responses may contribute to robust antitumor activity.

To date, the exact mechanism of irAEs has not been clarified. The CTLA-4 immune checkpoint has been shown to suppress immune response through various mechanisms, including attenuation of T-cell activation during the initial stages of the immune response; in contrast, PD-1 is generally believed to suppress T cells during later stages of the immune response in peripheral tissues.⁴⁰ Uncontrolled immune activation resulting from ICIs targeting these immune checkpoints can lead to a wide range of irAEs, including cirAEs. Consequently, based on the underlying mechanism of irAEs, it is logical to surmise that the emergence of irAEs signifies a more favorable prognosis. However, limited data were available for the CTLA-4 subgroup, necessitating further investigation.

Substantial heterogeneity was observed in this meta-analysis, which can be partially attributed to subgroup analysis based on cancer type. Differences in cirAE development among patients with different cancer types (eg, NSCLC and melanoma) who receive ICIs, as well as variations in ICI types, may have contributed to this heterogeneity. However, the paucity of studies reporting relevant results highlights the need for further research to identify the potential associations.

To our knowledge, this study represents the most extensive systematic review and meta-analysis conducted thus far regarding the association between cirAE occurrence and cancer prognosis among patients undergoing ICI treatment. This investigation incorporated all currently available studies, encompassing both retrospective and prospective designs. Additionally, our research accounted for the various risk factors associated with cirAEs, as defined by Delphi consensus, thus resulting in a comprehensive, precise, and intricate analysis. Notably, this meta-analysis comprehensively examined 2 crucial clinical outcomes: OS and PFS. Moreover, we performed subgroup analyses considering study design, geographic region, ICI type, and cancer type, aiming to minimize confounding variables that may potentially affect the results and provide a more realistic assessment of the association between targeted risk factors and patient outcomes.

Our study has various limitations that should be acknowledged. First, the majority of the data came from retrospective studies, adding possible biases such as recall bias and selection bias. Second, when HRs were not supplied in the original research, we collected data from Kaplan-Meier survival curves and computed HRs, which may have contributed to small statistical errors. Third, the limited availability of studies with specific subgroups of patients, such as studies with patients receiving anti–CTLA-4 treatment and studies with a prospective design, might have yielded less conclusive findings. Finally, the Egger tests uncovered substantial publication bias within the analyses of OS and PFS, suggesting that this may exaggerate the pooled results.

In this systematic review and meta-analysis, an association was observed between the existence of cirAEs and improved cancer prognosis among patients receiving ICI treatment. These data suggest that cirAEs may have useful prognostic value in ICI treatment. More large-scale prospective studies are needed to validate and establish the association between cirAEs and survival outcomes, especially for different grades of cirAEs and interactions with multisystem irAEs.

Accepted for Publication: July 4, 2023.

Published Online: September 6, 2023. doi:10.1001/jamadermatol.2023.3003

Corresponding Author: Fei Wang, MD, Department of Dermatology, ZhongDa Hospital, School of Medicine, Southeast University, Nanjing 210009, China (ffwangfei@163.com).

Author Contributions: Dr Du had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Concept and design: Du, Wu, Dong, Wang.

Acquisition, analysis, or interpretation of data: Du, Wu, Chen, Wang.

Drafting of the manuscript: Du, Wu, Wang.

Statistical analysis: Du, Wu, Wang.

Obtained funding: Du, Wu.

Administrative, technical, or material support: All authors.

Conflict of Interest Disclosures: None reported.

Funding/Support: The preparation, review, and approval of this manuscript was supported by grant SJCX22_0069 from the Postgraduate Research & Practice Innovation Program of Jiangsu Province.

Role of the Funder/Sponsor: The funder had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.

Data Sharing Statement: See Supplement 2.