Differential Dermatologic Adverse Events Associated With Checkpoint Inhibitor Monotherapy and Combination Therapy: A Meta-Analysis of Randomized Control Trials

Background: As immune checkpoint inhibitors (ICIs) transition to the forefront of cancer treatment, a better understanding of immune related adverse events (IRAEs) is essential to promote safe clinical practice. Dermatologic adverse events are the most common IRAEs and can lead to drug withdrawal and decreased quality of life. This meta-analysis aimed to investigate the risk of the most prevalent dermatologic adverse events (pruritus and rash) among various ICI treatment regimens. Methods: A systematic search of electronic databases was performed to identify qualified randomized controlled trials (RCTs). Data for any grade and high grade pruritus and rash were extracted for meta-analysis. Two reviewers independently assessed methodological quality. The relative risk summary and 95% confidence interval were calculated. Results: 50 RCTs involving 29941 patients were analyzed. The risk of pruritus (2.15 and 4.21 relative risk respectively) and rash (1.61 and 3.89 relative risk respectively) developing from CTLA-4 or PD-1/-L1 inhibitor were increased compared to placebo, but this effect was not dose-dependent. PD-1/-L1 plus CTLA-4 inhibitor was associated with increased risk of pruritus (1.76 and 0.98 relative risk respectively) and rash (1.72 and 1.37 relative risk respectively) compared to either monotherapy. Compared with CTLA-4 inhibitor, PD-1/-L1 inhibitor had a significantly decreased risk of pruritus and rash in both monotherapy and combination therapy (0.65 and 0.29 relative risk respectively). No significant difference was found between PD-1/-L1 inhibitor combined with chemotherapy and PD-1/-L1 monotherapy in any grade and high grade rash (0.84 and 1.43 relative risk respectively). In subgroup analyses, PD-1 inhibitor was associated with reduced risk of pruritus and rash compared to PD-L1 inhibitor. Conclusion: Our meta-analysis demonstrates a better safety profile for PD-1/-L1 inhibitor compared to CTLA-4 inhibitor in terms of pruritus and rash among both monotherapy and multiple combination therapies. PD-L1 inhibitor may contribute to an increased risk of pruritus and rash compared to PD-1 inhibitor.


INTRODUCTION
The application of immune checkpoint inhibitors (ICIs) is a significant milestone for clinical strategies in cancer. Due to increased activation of the immune system, ICIs can cause a spectrum of IRAEs that affect multiple organ systems and can even lead to death (Fausto et al., 2020). Dermatologic toxicities appear to be the most prevalent IRAEs, both with Programmed cell death protein 1/Programmed cell death-ligand 1 (PD-1/ PD-L1) inhibitor and Cytotoxic T lymphocyte associate protein 4 (CTLA-4) inhibitor, and occur in more than a third of patients treated with ICI monotherapy (Sibaud et al., 2016). Consequently, decreased quality of life due to dermatologic adverse events may contribute to unnecessary drug withdrawal by patients. Additionally, management of serious dermatologic adverse events, including oral and topical steroids, may result in reduced drug efficacy (Geisler et al., 2020). Among dermatologic IRAEs manifestations, pruritus and rash are the most common (Boutros et al., 2016;Ellis et al., 2020;Geisler et al., 2020). Indeed, clinical studies demonstrate that pruritus may occur in 11-21% of patients treated with anti-PD-1/-L1 inhibitor, 24.4-35.4% of patients treated with CTLA-4 inhibitor, and 33.2-47% of patients in dual CTLA-4/PD-1 blockade (Geisler et al., 2020;Nishijima et al., 2017;Sibaud et al., 2016). For rash, incidence ranges as high as 20% for patients receiving PD-1 inhibitor, 14-26% for patients receiving CTLA-4 inhibitor, and 28.4-55% for patients receiving dual anti-CTLA-4/PD-1 blockade therapy (Geisler et al., 2020;Sibaud et al., 2016). Therefore, to balance the benefits and risks among multiple ICI treatment patterns in clinical strategy, an improved understanding of dermatologic IRAEs is essential (Collins et al., 2017;Ellis et al., 2020).
Combination immunotherapy has become a popular treatment option due to its superior clinical efficacy. However, ICI combination therapy is associated with toxic effects resulting from unbalanced activation of the immune system (Da et al., 2020). As mentioned above, combination of anti-CTLA-4 and anti-PD-1 therapy is associated with more frequent, more severe, and earlier dermatologic IRAEs compared to monotherapy (Almutairi et al., 2020;Sibaud et al., 2016). However, few studies have assessed dermatologic adverse events resulting from various ICI treatment regimens. Although previous meta-analysis (Nishijima et al., 2017;Yang et al., 2019) evaluated the incidence of selected dermatologic and mucosal adverse effects associated with PD-1/-L1 inhibitors, the authors included chemotherapy and ipilimumab as the only control arms. Other studies investigated the incidence and risk of IRAEs (including dermatologic adverse events) due to ICI monotherapy and combination therapy (Almutairi et al., 2020;Velasco et al., 2017;Wang et al., 2021), yet the patients included in their analysis were limited to a single tumor such as melanoma or lung cancer. Moreover, direct comparisons of the risk of dermatologic IRAEs between combination therapy and ICI monotherapy are lacking due to a dearth of head-to-head clinical trials. Therefore, a better understanding of dermatologic adverse events in this context is still needed. In the current study, we focused on the two most common dermatologic adverse events, pruritus and rash (Braun et al., 2020;Golian et al., 2016), in patients receiving ICI monotherapies and combination therapies including chemotherapy, targeted therapy, and other ICI treatment regimens. All the data used in this meta-analysis are derived from published literature and clinical trials.

Search Strategy and Eligibility Criteria
Two investigators (Yang Ge and Hui-Yun Zhang) independently searched PubMed, Embase, Web of Science, and the Cochrane Library. The last search was performed on January 20, 2020. The following terms were used: (Nivolumab or Opdivo or ONO-4538 or ONO 4538 MDX-1106 or BMS-936558 or pembrolizumab or lambrolizumab or Keytruda or cemiplimab or Pidilizumab or camrelizumab or SHR-1210 or JS001 or sintilimab or Durvalumab or MEDI4736 or atezolizumab or avelumab or Bavencio or tremelimumab or ticilimumab or Ipilimumab) and (Carcinoma or Neoplasia or Tumor or Cancer or Malignancy) and randomized controlled trials.
The following inclusion criteria were used: 1) studies included either ICI monotherapy or ICI combination therapy with chemotherapy/targeted therapy/ICIs in patients diagnosed with solid tumor; 2) studies investigated the following dermatologic adverse events: pruritus and rash; 3) randomized controlled clinical trials published in English. The following exclusion criteria were used: 1) phase I clinical trials; 2) studies without related data; 3) studies reporting dermatologic adverse events which are not related to ICIs; 3) editorials, letters, case reports, expert opinions, or reviews; and 4) duplicate publications.
"any grade" referred to CTCAE grades 1-5; "low grade" referred to CTCAE grades 1-2; "high grade" referred to CTCAE grades 3-5. The dosage of ICIs was also extracted to investigate if high dose ICIs are associated with increased IRAEs. Less than or equal to 3 mg/kg of PD-1/CTLA-4 was identified as "low dose", while greater than or equal to 10 mg/kg was identified as "high dose". The extraction was performed by two investigators (Yang Ge and Huiyun Zhang) independently and any controversies were resolved by discussion.
Quality assessment was performed using Review Manager 5.3. Risk of bias for the eligible study was evaluated according to following items recommended by the Cochrane Collaboration: randomization, allocation concealment blinding of participant, blinding of outcome assessors, incomplete outcome data, selective reporting, and other bias.

Statistical Analysis
We conducted the meta-analysis using Review Manager 5.3. Risk ratio (RR) and 95% confidence interval (95% CI) were applied to evaluate the risk of pruritus and rash for both experimental and control arms. Relative risk ratio (RRR) with 95% CIs between different treatment regimens were calculated using RRs and 95% CIs. Heterogeneity was tested by the I 2 and Q test. When p > 0.1 and I 2 ≤ 50%, it was considered to indicate no significant heterogeneity and the fixed-effect model was applied. Otherwise, the random-effects model was applied. Begg's and Egger's tests were performed using Stata 16.0 to estimate publication bias. Subgroup analyses were performed to explore the sources of heterogeneity according to the different ICI class and tumor types.

Search Results and Study Characteristics
14,819 publications were initially identified from the database and plus 11 from other sources. After excluding duplicates, 13,777 publications were assessed for review of title and abstract. 336 articles were further assessed for full-text review. Finally, 50 RCTs (n 29,941 patients) were included in this meta-analysis ( Figure 1). Most of the included studies involved patients with melanoma (N 15) and none small cell lung carcinoma (NSCLC) (N 12). The others were focused on renal cell carcinoma (RCC) (N 5), head and neck squamous cell carcinoma (HNSCC) (N 4), small cell lung cancer (SCLC) (N 3), gastric cancer or gastrooesophageal junction cancer (GC/GOJC) (N 3), prostate cancer (N 2), urothelial cancer (UC) (N 2), malignant mesothelioma (N 1), triple-negative breast cancer (TNBC) (N 1), hepatocellular carcinoma (HCC) (N 1), and pancreatic cancer (N 1). Details of characteristics of the included studies are shown in Table 1.

High Dose Vs Low Dose Programmed Cell Death Protein 1/Programmed Cell Death-Ligand 1 Inhibitor
In this section, five qualifying studies with 2,015 patients total were analyzed. Compared to low dose groups, RRs for any grade pruritus and any grade rash developed after high dose PD-1/PD-L1 inhibitor therapy were 0.84 (95%CI 0.63-1.14, p 0.26) (Supplementary Figure 4A) and 0.79 (95%CI 0.56-1.11, p 0.17) respectively (Supplementary Figure 4B).
For comparison of PD-1/-L1 plus CTLA-4 inhibitor to CTLA-4 inhibitor monotherapy, we included four studies with 1,813 patients total. Only any grade rash was more frequent in patients administered CTLA-4 inhibitor combined with PD-1/-L1 inhibitor, in comparison to CTLA-4 inhibitor monotherapy [any grade pruritus RR:0.98 (95% CI 0.80-1.19, p 0.81) ( Figure 6E), any grade rash RR:1.37 (95% CI 1.07-1.74, p 0.01) ( Figure 6F)]. Data for high grade pruritus and high grade rash are not reported because only two studies identified included these categories, which was not sufficient for a qualified metaanalysis.

Subgroup Analyses
Programmed Cell Death Protein 1 Vs Programmed Cell Death-Ligand 1 Inhibitor Subgroup analysis was performed to identify the relative impact of PD-1 and PD-L1 inhibitor on pruritus and rash. 20,769 patients from 42 studies were included in this analysis. Risks of any grade pruritus (RR: 1.93 (95% CI 1.40-2.67) p < 0.00001 Supplementary Figure 5A) and any grade rash [RR: 1.28 (95% CI 1.03-1.58) p < 0.00001 Supplementary Figure 5B] developed during PD-1 inhibitor therapy were decreased compared to PD-L1 inhibitor. When assessing high grade rash between PD-1 inhibitor and PD-L1 inhibitor therapies, no statistically significant difference was found [RR: 0.67 (95% CI 0.39-1.17) p 0.46 Supplementary Figure 5C].  Figures 6A-D). Sensitivity analysis showed that no single study could significantly affect the aggregated estimates (Supplementary Figure 9). However, there was mild asymmetry for RRs of pruritus and rash (Supplementary Figure 10). The Egger's test (Supplementary Figure 11) shown some evidence of publication bias for pruritus (p 0.005/p 0.006) and high grade rash (p 0.001), while the Begg's test revealed no evidence of publication bias (Supplementary Figure 12).

DISCUSSION
With the growing number of patients receiving ICIs, there is significant need to understand associated adverse events in order to improve therapy management. In clinical practice ICIs have shown significant efficiency in multiple tumors, both as monoand combination therapies. The unique ICI mechanism of action (Sibaud et al., 2016) is also accompanied with a series of IRAEs, which are distinguishable from traditional adverse effects of cancer treatment. Dermatological reactions, especially pruritus and rash, are some of the most common IRAEs, and can severely affect the quality of life and psychological well-being of patients (Sibaud et al., 2016). High grade rash can impact ICI treatment efficacy through dose-limiting effects or even result in treatment discontinuation (Geisler et al., 2020). To achieve better clinical efficacy, ICI combination therapy has become more commonly used. However, few studies have been conducted to assess the risk of dermatological-specific IRAEs among multiple treatment patterns. To our knowledge, the current study is the first comprehensive assessment of the relative risk of pruritus and rash among various ICI treatment regimens.
We first compared ICI monotherapy to placebo, and both PD-1/-L1 and CTLA-4 inhibitor were associated with increased risk of any grade pruritus and rash. Notably, CTLA-4 inhibitor was associated with higher risk of high grade pruritus and rash. A comparison between PD-1/-L1 inhibitor and CTLA-4 inhibitor monotherapy was also conducted. RRs for pruritus and rash developed after PD-1/-L1 inhibitor were decreased compared to CTLA-4 inhibitor, which is in line with the current mainstream consensus that CTLA-4 inhibitor is more likely to lead to pruritus and rash (Almutairi et al., 2020;Geisler et al., 2020;Hansen et al., 2017;Sibaud et al., 2016).
Whether the risk of developing pruritus and rash correlated with different dose regimens of immune checkpoint inhibitor is an important area of focus given issues regarding patient quality of life and treatment discontinuation. Previous studies have shown no significant correlation between PD-1/-L1 inhibitor dosage and incidence of pruritus and rash (Hansen et al., 2017;Robert et al., 2014). On the contrary, a retrospective study suggested that the frequency of IRAEs (pruritus and rash included) developed after Ipilimumab increased with dose. Another study reached a similar conclusion (Golian et al., 2016) that cutaneous IRAEs related to ipilimumab are dose-related. In the current study, compared with the low dose group, RRs for any grade pruritus and rash developed after PD-1/-L1 inhibitor in the high dose group were not significantly increased. The corresponding comparison between CTLA-4 inhibitor high dose and low dose group could not be carried out because of insufficient data. Overall, given the discrepancies among findings in studies assessing dose-dependencyof rash and pruritus, further efforts should be made to investigate the problem and instruct clinical application, both in terms of mechanism and clinical research. In order to increase the percentage of patients benefiting from ICI treatment and reduce the occurrence of IRAEs, efforts are currently being made to combine current ICIs with new checkpoint inhibitors or other treatment methods to achieve synergistic effects (Kon and Benhar, 2019). In clinical practice, PD-1/-L1 and CTLA-4 inhibitor are being combined with other anti-cancer drugs including chemotherapy, targeted therapy, radiotherapy and other immunotherapies. Although traditionally regarded as immunosuppressive agents, some preclinical studies have shown that chemotherapy may have immune-stimulatory properties (Postow et al., 2015a). Some studies indicate combination chemotherapy leads to more general adverse events (Lynch et al., 2012;Wang et al., 2021), while other studies report severe side effects (Chamoto et al., 2020) . We used the relative risk ratio (RRR) to indirectly compare the risk of pruritus and rash. RRR was used to compare PD-1/PD-L1 inhibitor monotherapy with combined chemotherapy based on PD-1/PD-L1 inhibitor, and showed that the risk of pruritus, but not rash, was increased ( Table 2). These results suggest that PD-1/-L1 inhibitor combined with chemotherapy may have a tolerable dermatologic adverse profile in terms of pruritus and rash, indicating that increased efficacy through combining ICIs with chemotherapy may be feasible. Targeted therapies for oncogenic signaling pathways are also attractive partners in combination with immune checkpoint blockade (Postow et al., 2015a). Unfortunately, only 2 RCTs comparing PD-1/-L1 inhibitor and targeted therapy resulted from our database search, and RRR for PD-1/PD-L1 inhibitor plus targeted therapy compared to PD-1/-L1 monotherapy could not be calculated. When more data becomes available, further analysis of this aspect may provide useful information.
Since CTLA-4 inhibitor monotherapy showed increased risk of pruritus and rash relative to PD-1/-L1 inhibitor according to our data, RRR was calculated to investigate the difference between PD-1/-L1 plus chemotherapy and CTLA-4 plus chemotherapy. When contrasted with PD-1/-L1 inhibitor combination chemotherapy, CTLA-4 inhibitor combination chemotherapy was associated with a much higher risk of pruritus and rash ( Table 2). The mechanism leading to this is not yet fully understood. The major physiological role of CTLA-4 seems to be through distinct effects on the two main subsets of cluster of differentiation four positive (CD4 + ) T cells: down modulation of helper T cell activity and enhancement of regulatory T (Treg) cell immunosuppressive activity (Bylicki et al., 2020;Cancela et al., 2020;Peggs et al., 2009). Blockade of the PD-1 pathway may enhance antitumor immune responses by diminishing the number and/or suppressive activity of intratumoral Treg cells (Arigami et al., 2020). It is thought that PD-1 predominantly regulates effector T cell activity within tissue and tumors, whereas CTLA-4 predominantly regulates T cell activation (Arigami et al., 2020). Although dermatologic adverse events observed with ICIs used in combination are more frequent, more severe, and longer lasting (Sibaud et al., 2016), combination immunotherapy has more extensive clinical applications due to improved efficacy. Therefore, our data suggest that PD-1/-L1 inhibitor may be preferable in patients who have suffered from previous dematologic problems. Moreover, in the case of severe dermatologic IRAEs with CTLA-4 therapy, re-challenge with an agent of a different class may be a good treatment strategy.
Subgroup analysis was performed to investigate if there was any difference in the incidence of pruritus and rash between PD-1 and PD-L1 inhibitor. Based on the known interactions of PD-1 ligands, PD-1 antibodies may have different biological activities than PD-L1 antibodies. PD-1 antibodies prevent PD-1 from interacting with PD-L1 and Programmed cell death-ligand 2(PD-L2), but do not prevent the interaction between PD-L1 and Cluster of differentiation 80(CD80). In contrast, most PD-L1 antibodies prevent the interaction between PD-L1 and CD80 and between PD-L1 and PD-1, but not the interaction between PD-1 and PD-L2. Therefore, it is possible that depending on which interaction predominates in a particular cancer, PD-1 and PD-L1 antibodies may not have redundant activity (Arigami et al., 2020). Results from subgroup analysis showed that any grade pruritus and rash developed from PD-1 inhibitor were decreased FIGURE 6 | Forest plots of the relative risks and 95% CIs for pruritus and rash in comparison of combined immunotherapy and either monotherapy: (A) any grade pruritus for PD-1/-L1 plus CTLA-4 inhibitor compared to PD-1/-L1 inhibitor; (B) high grade pruritus for PD-1/-L1 plus CTLA-4 inhibitor compared to PD-1/-L1 inhibitor; (C) any grade rash for PD-1/-L1 plus CTLA-4 inhibitor compared to PD-1/-L1 inhibitor; (D) high grade rash for PD-1/-L1 plus CTLA-4 inhibitor compared to PD-1/-L1 inhibitor; (E) any grade pruritus for PD-1/-L1 plus CTLA-4 inhibitor compared to CTLA-4 inhibitor; (F) any grade rash for PD-1/-L1 plus CTLA-4 inhibitor compared to CTLA-4 inhibitor.
Frontiers in Pharmacology | www.frontiersin.org July 2021 | Volume 12 | Article 640099 14 compared to PD-L1 inhibitor, while the comparison in high grade (3-5) rash did not reach a statistically significant level. Therefore, PD-1 inhibitor may be recommended in terms of decreased dermatologic adverse events (pruritus and rash) for clinical applications. In cancer type subgroup analysis, we found that patients with HNSCC may have better tolerability overall as evidenced by a lower risk for any grade pruritus. Since only 1 RCT of HNSCC was included in high grade subgroup, more efforts are needed to validate this observation.
Our study has some notable strengths. To the best of our knowledge, this is the first and most comprehensive analysis that investigated the risk of pruritus and rash among different ICI treatment regiments in multiple solid tumors. In addition, the 50 clinical trials included in our meta-analysis were all highly qualified randomized control trails, which supports the credibility of our study. Morever, we investigated the risk of not only all grade but also high grade pruritus and rash, for the management of these two side effects of differing severity. Finally, since head-to-head comparison of PD-1/PD-L1 inhibitor combination therapies and PD-1/PD-L1 inhibitor alone were not available, we used the relative risk ratio (RRR) to indirectly compare the risk of pruritus and rash. The results of our RRR analysis indicate that the added skin toxicity of chemotherapy is manageable in combination immunotherapy, which may have clinical implications.
This meta-analysis also has some limitations. Mild heterogeneity was found among the included studies. The heterogeneity may result from differences in cancer type, line of therapy, follow-up time, or other unspecified factors. Study design, blinding, dosage and frequency of drug administration in both intervention and control arm could also have resulted in heterogeneity. Thus, we utilized the random-effect model and subgroup analyses for high heterogeneity to explore possible variation in the outcomes of the included studies. What`s more, since patients included in our meta-analysis were from RCTs with strict inclusion criteria, risk of pruritus and rash could be underestimated because of their better health condition, compared with patients in real world application.

CONCLUSION
In summary, we identified that PD-1/-L1 inhibitor is associated with decreased risk of pruritus and rash in comparison to CTLA-4 inhibitor in both monotherapy and combined immunotherapy regimens. Additionally, pruritus and rash developed from PD-1/-L1 inhibitor are not dosedependent. Moreover, compared to PD-1/-L1 inhibitor alone, the combination of chemotherapy with PD-1/-L1 inhibitor may not significantly increase the risk of pruritus and rash. As the most prevalent and obvious IRAEs, dermatologic adverse events such as rash and pruritus should be further studied to help manage such events and enhance patient benefits from ICI therapy.