Kim Woorim, Oh Soo-Jin, Kim Hyun-Jeong, Kim Jun-Hyeob, Gil Jin-Yeon, Ku Young-Sook, Kim Joo-Hee, Kim Hyoun-Ah, Jung Ju-Yang, Choi In-Ah, Kim Ji-Hyoun, Kim Jinhyun, Han Ji-Min, Lee Kyung-Eun
College of Pharmacy, Kangwon National University, Chuncheon 24341, Republic of Korea.
College of Pharmacy, Chungbuk National University, Cheongju 28160, Republic of Korea.
J Clin Med. 2024 Jul 11;13(14):4050. doi: 10.3390/jcm13144050.
Rheumatoid arthritis (RA) is a chronic inflammatory disorder primarily targeting joints, significantly impacting patients' quality of life. The introduction of tumor necrosis factor-alpha (TNF-α) inhibitors has markedly improved RA management by reducing inflammation. However, these medications are associated with adverse skin reactions, which can vary greatly among patients due to genetic differences. This study aimed to identify risk factors associated with skin adverse events by TNF-α in RA patients. A cohort study was conducted, encompassing patients with RA who were prescribed TNF-α inhibitors. This study utilized machine learning algorithms to analyze genetic data and identify markers associated with skin-related adverse events. Various machine learning algorithms were employed to predict skin and subcutaneous tissue-related outcomes, leading to the development of a risk-scoring system. Multivariable logistic regression analysis identified independent risk factors for skin and subcutaneous tissue-related complications. After adjusting for covariates, individuals with the TT genotype of rs12551103, A allele carriers of rs13265933, and C allele carriers of rs73210737 exhibited approximately 20-, 14-, and 10-fold higher incidences of skin adverse events, respectively, compared to those with the C allele, GG genotype, and TT genotype. The machine learning algorithms used for risk prediction showed excellent performance. The risk of skin adverse events among patients receiving TNF-α inhibitors varied based on the risk score: 0 points, 0.6%; 2 points, 3.6%; 3 points, 8.5%; 4 points, 18.9%; 5 points, 36.7%; 6 points, 59.2%; 8 points, 90.0%; 9 points, 95.7%; and 10 points, 98.2%. These findings, emerging from this preliminary study, lay the groundwork for personalized intervention strategies to prevent TNF-α inhibitor-associated skin adverse events. This approach has the potential to improve patient outcomes by minimizing the risk of adverse effects while optimizing therapeutic efficacy.
类风湿关节炎（RA）是一种主要累及关节的慢性炎症性疾病，严重影响患者的生活质量。肿瘤坏死因子-α（TNF-α）抑制剂的引入通过减轻炎症显著改善了RA的治疗。然而，这些药物与皮肤不良反应有关，由于基因差异，患者之间的反应差异很大。本研究旨在确定RA患者中与TNF-α相关的皮肤不良事件的危险因素。进行了一项队列研究，纳入了接受TNF-α抑制剂治疗的RA患者。本研究利用机器学习算法分析基因数据并识别与皮肤相关不良事件相关的标志物。采用各种机器学习算法预测皮肤和皮下组织相关结局，从而开发出一种风险评分系统。多变量逻辑回归分析确定了皮肤和皮下组织相关并发症的独立危险因素。在调整协变量后，与具有C等位基因、GG基因型和TT基因型的个体相比，rs12551103的TT基因型个体、rs13265933的A等位基因携带者和rs73210737的C等位基因携带者发生皮肤不良事件的发生率分别高出约20倍、14倍和10倍。用于风险预测的机器学习算法表现出色。接受TNF-α抑制剂治疗的患者发生皮肤不良事件的风险因风险评分而异：0分，0.6%；2分，3.6%；3分，8.5%；4分，18.9%；5分，36.7%；6分，59.2%；8分，90.0%；9分，95.7%；10分，98.2%。这项初步研究的结果为预防TNF-α抑制剂相关皮肤不良事件的个性化干预策略奠定了基础。这种方法有可能通过将不良反应风险降至最低同时优化治疗效果来改善患者预后。
