Predicting the Likelihood of Immune-related Adverse Events in Breast Cancer Patients

摘要 免疫肿瘤药物明显提高了三阴性乳腺癌(TNBC)的应答率 病人。然而，这些改善是有代价的--10%-25%的患者将经历与免疫相关的 不良事件(IrAES)。这些声学效应似乎与反应无关，而且看起来很特殊。 例如，肾上腺功能不全可能在患者出现严重症状并有皮质醇时出现得较晚。 如果治疗不当，可能导致死亡的接近零的风险。识别单个患者或患者亚组的能力 那些处于这些毒性增加或高风险的人将在几个方面改善结果并减少危害。 风险最高的人可能会避免接受某些免疫疗法的治疗，而风险更高的人可能会 被标记为进行更密切的监测或实施预防性干预以避免或降低AE的等级。它的用途 以这种方式收集人口、生物和遗传信息符合精确肿瘤学的努力。 我们正在解决的癌症焦点问题是，我们是否能够预测个体 在癌症免疫治疗后经历严重的免疫相关不良事件，使用年龄，合并疾病， 电子健康记录(EHR)数据、生活质量(QOL)、不良事件(AE)和遗传变异。我们 假设可以通过预测性分析产生对哪些患者将经历irAEs的早期洞察 嵌入决策支持框架。这项提议的总体目标是：(1)及早破译 哪些患者会经历甲状腺疾病、肺炎、瘙痒、结肠炎、肝毒性或肾上腺？ 失败并最终影响生活质量；以及(2)更好地了解患者基础上的基因图谱 发生irAEs的风险。 我们将使用来自I-SPY2早期乳腺癌试验的丰富的多维数据。由于其自适应能力 平台设计，i-SPY2提供了在同一研究中研究多种免疫疗法的机会， 跨多个站点使用标准方法。我们建议：1)开发和评估一个整体的 方法和由此产生的决策支持算法，专为帮助管理 接受免疫治疗的患者的护理，2)确定新的和注释的单核苷酸 与irAEs相关的多态(SNP)，以及3)在两个新的 试验性的手臂。我们的计算模型将根据500例I-SPY2乳腺癌的信息进行训练 正在接受免疫治疗的试验患者。这项工作的圆满完成将增进我们对 临床、患者报告和遗传因素是导致irAEs的潜在因素，并使早期预测谁处于危险之中成为可能 在治疗开始之前。

ABSTRACT Immuno-oncology agents have clearly improved rates of response in triple negative breast cancer (TNBC) patients. However, these improvements come at a cost -- 10-25% of patients will experience an immune-related adverse event (irAEs). These AEs do not appear to be associated with response and appear idiosyncratic. Adrenal insufficiency, for example, can appear late when patients are extremely symptomatic and have a cortisol near zero that can lead fatality if improperly treated. The ability to identify individual patients or subsets of patients who are at increased or high risk of these toxicities will improve outcomes and reduce harm in several ways. Those at highest risk may avoid treatment with certain immunotherapies, while those at increased risk could be flagged for closer monitoring or placed upon prophylactic interventions to avoid or downgrade the AE. The use of demographic, biologic and genetic information in this way is in keeping with precision oncology efforts. The cancer-focused question we are addressing is whether we can predict the likelihood of individuals experiencing serious immune-related adverse events following cancer immunotherapy using age, comorbidities, electronic health record (EHR) data, quality of life (QOL), adverse events (AE), and genetic variations. We hypothesize that early insight into which patients will experience irAEs can be generated by predictive analytics embedded within a decision-support framework. The overall goals of this proposal are to: (1) decipher early which patients are going to experience thyroid disease, pneumonitis, pruritus, colitis, hepatoxicity, or adrenal failure and ultimately affect quality of life; and (2) better understand the genetic profile that underlies patients' risk of developing irAEs. We will use a rich multidimensional data from the I-SPY2 trial in early breast cancer. Due to its adaptive platform design, I-SPY2 provides the opportunity to study multiple immunotherapies within in the same study, using standard methodologies across multiple sites. We propose to: 1) develop and evaluate a holistic approach and resulting decision support algorithm, designed for clinician-researchers who help manage the care of patients undergoing immunotherapy, 2) determine both novel and annotated single nucleotide polymorphisms (SNPs) associated with irAEs, and 3) validate the decision support algorithm in two new experimental arms. Our computational models will be trained on information from 500 I-SPY2 breast cancer trial patients undergoing immunotherapy. Successful completion of this work will increase our understanding of the clinical, patient-reported, and genetic factors underlying irAEs and enable early prediction of who is at risk before therapy is initiated.