Braganca Xavier Camila, Guardia Gabriela D A, Alves João Pedro B, Lopes Carlos Diego H, Awni Beatriz M, Campos Eduardo F, Jardim Denis L, Galante Pedro A F
MD Anderson Cancer Center, Houston, TX, 77030, United States.
Hospital Sírio-Libanês, São Paulo, SP, 01308-050, Brazil.
Oncologist. 2025 Apr 4;30(4). doi: 10.1093/oncolo/oyaf078.
Immune checkpoint inhibitors (ICIs) have significantly advanced cancer therapy, yet their efficacy in tumors with low tumor mutational burden (TMB) remains suboptimal. In this study, we aimed to elucidate the impact of somatic mutations on overall survival (OS) in TMB-low patients treated with ICIs and to explore the potential for personalized treatment selection through machine learning.
We conducted a comprehensive analysis of 1172 TMB-low (TMB < 10 mutations per megabase) patients with cancer receiving ICIs, examining the association between specific gene mutations and OS. Additionally, we developed a decision tree model (DTM) to predict OS based on clinical features and tumor mutational profiles.
Our findings reveal that mutations in DAXX, HLA-A, H3C2, IGF1R, CTNNB1, SMARCA4, KMT2D, and TP53 are significantly associated with poorer survival outcomes in the multivariate analysis. Remarkably, for renal cell carcinoma (RCC) patients, VHL mutations predicted improved OS following ICI even when adjusted for age, sex, and microsatellite instability (MSI) status in both multivariate analysis and the DTM model.
These results reinforce the prevailing notion that TMB alone does not predict ICI response, highlighting the critical role of individual gene mutations in TMB-low tumors under ICI therapy. Furthermore, our study demonstrates the promise of machine learning models in optimizing ICI treatment decisions, paving the way for more precise and effective therapeutic strategies in this patient population.
免疫检查点抑制剂（ICIs）显著推动了癌症治疗的进展，但其在肿瘤突变负荷（TMB）较低的肿瘤中的疗效仍不尽人意。在本研究中，我们旨在阐明体细胞突变对接受ICIs治疗的TMB低患者总生存期（OS）的影响，并通过机器学习探索个性化治疗选择的潜力。
我们对1172例接受ICIs治疗的TMB低（TMB<10个突变/Mb）癌症患者进行了全面分析，研究特定基因突变与OS之间的关联。此外，我们基于临床特征和肿瘤突变谱开发了一种决策树模型（DTM）来预测OS。
我们的研究结果显示，在多变量分析中，DAXX、HLA-A、H3C2、IGF1R、CTNNB1、SMARCA4、KMT2D和TP53的突变与较差的生存结果显著相关。值得注意的是，对于肾细胞癌（RCC）患者，在多变量分析和DTM模型中，即使在调整了年龄、性别和微卫星不稳定性（MSI）状态后，VHL突变仍预示着ICI治疗后OS的改善。
这些结果强化了仅TMB不能预测ICI反应的普遍观点，突出了个体基因突变在ICI治疗下TMB低的肿瘤中的关键作用。此外，我们的研究证明了机器学习模型在优化ICI治疗决策方面的前景，为该患者群体中更精确有效的治疗策略铺平了道路。
