[Data mining and toxicity profile analysis of immune checkpoint inhibitor-related skin toxicity events based on...

The widespread use of immune checkpoint inhibitors (ICIs) has led to breakthrough advances for patients with various advanced solid tumors. As the skin is an important target organ of immune responses, it is the most commonly affected site of treatment-related adverse events associated with ICIs, with a relatively high incidence of ICI-related skin toxicity events. Immune-related adverse events induced by ICIs are increasingly becoming a bottleneck limiting their clinical application. To collect post-marketing adverse events and medication errors related to drugs and therapeutic biological products and to evaluate real-world drug safety, the United States Food and Drug Administration (FDA) established the FDA Adverse Event Reporting System (FAERS) database. Based on the FAERS database, this study aims to systematically evaluate differences in the risk of skin toxicity events among different drug subtypes, cytotoxic-T-lymphocyte-associated antigen-4 inhibitors, programmed death-1 (PD-1) inhibitors, and programmed death-ligand 1 (PD-L1) inhibitors, and to explore the limitations and potential improvements of existing pharmacovigilance methods.

Skin toxicity event data from the FAERS database between 2004 and 2024 were cleaned, standardized, and screened to identify adverse events associated with the target drugs. Pharmacovigilance signal detection methods, including the reporting odds ratio (ROR) method and Bayesian confidence propagation neural network (BCPNN) method, were used for the signals of ICIs-related skin toxicity event in the data, with stratified analyses by age and sex. For the ROR method, a skin toxicity event reporting count ≥3 and a lower bound of the 95% confidence interval (CI) >1 were used as criteria for a positive signal; for the Bayesian method, the information component was used as the core parameter. Subsequently, a systematic statistical analysis of the frequencies of different types of adverse events induced by different drugs was conducted, and outcomes associated with different drugs were summarized. Pharmacovigilance signal detection methods were applied for data analysis.

A total of 15 768 reports of skin-related adverse events were collected. The reported population was predominantly male, with most patients aged ≥65 years, and a higher proportion of cases from Europe and the United States. Among the reported indications, the 3 most common were malignant melanoma, non-small cell lung cancer, and metastatic melanoma. Most adverse events occurred within 30 days after drug administration, during which the number of reports was the highest. Using the two signal detection methods, positive signals were identified for 5 of the 8 target drugs: nivolumab (ROR=1.20, 95% CI 1.17 to 1.23), pembrolizumab (ROR=1.31, 95% CI 1.27 to 1.35), ipilimumab (ROR=1.82, 95% CI 1.74 to 1.90), atezolizumab (ROR=1.06, 95% CI 1.01 to 1.12), and tislelizumab (ROR=3.05, 95% CI 2.71 to 3.43). Further analysis showed that the PD-1 inhibitors pembrolizumab and nivolumab had higher numbers of reported cases of immune-mediated dermatitis, vitiligo, psoriasis, and other skin toxicity events than other ICIs. Comparison of outcomes of skin toxicity reactions caused by different drugs revealed that nivolumab-related cases had the highest numbers of reports of hospitalization, death, and life-threatening, followed by pembrolizumab.

Five ICIs may induce skin toxicity events, which exhibit specific population characteristics, temporal patterns, and toxicity profiles. When using the PD-1 inhibitors nivolumab or pembrolizumab, particular vigilance is required for severe cutaneous toxicity to avoid unfavorable outcomes. Expanding the sample size and incorporating machine learning in the future may improve the precision and clinical translatability of signal detection, providing important evidence for optimizing clinical monitoring strategies for ICIs and establishing toxicity early-warning models.