Advancing Analysis and Interpretation of Adverse Events and PROs in Cancer Clinical Trials

摘要 在这项研究申请和计划中，我们将与NRG肿瘤统计中心合作， 开发分析和图形策略，以研究评估治疗耐受性的新方法， 以及使用来自随机NSABP试验的数据来模拟数据呈现的新方法， 不良事件通用术语标准（CTCAE）数据和高质量患者报告 结果（PRO）数据。我们已经将这些新的分析方法和其他方法应用于NSABP和NRG 肿瘤学III期临床试验，包括PRO-CTCAE项目，以评估与以下相关的治疗毒性： 免疫疗法和其他治疗。在资助期的前四年，我们开发了一部小说， 根据患者的总体毒性经历， 根据CTCAE和PRO-CTCAE进行AE分级评估。TI解释了所有观察到的毒性等级 而不是像传统上那样仅对最严重的一个进行处理。由于其对不同的 总体毒性，我们表明，TI可以识别治疗相关毒性的预测比传统的 汇总分数，如最高分数和平均分数。在这项研究计划中，我们计划（1）应用我们的 利用PRO数据评价内分泌治疗毒性的方法 NSABP P-1临床试验中使用CYP 2D 6血浆样本的基因型和他莫昔芬停药 基因分型，预计在未来六个月内，（2）继续我们的分析的可行性，频繁 ePRO数据评估测试ePRO每周测量相对于以下数据收集的附加值 以临床医生的CTCAE评估为基准，每个周期进行ePRO，（3）进一步评估和传播 乳腺癌症状浏览器可视化在线工具，通过更新内容，功能，技术 通过返回原始焦点小组和招募更多的焦点小组进行进一步的定性分析， 评估，（4）继续我们对绝经后妇女症状轨迹的分析， 与个体轨迹中的成员资格相关的患者宿主因素以及这些因素如何影响治疗 停产和其他成果，（5）继续开展动态风险预测的工作 使用纵向PRO、临床医生评估的CTCAE和 NSABP B-35 III期临床试验的基线临床和人口统计学数据。我们将测试附加值 除了纵向PRO外，还包括纵向临床医生评估的CTCAE数据， 动态模型的性能，评估和比较CTCAE时的预测性能 总结与我们的新TI相对于平均等级，和最大等级，并建立基于实时计算器 这些新的预测模型在闪亮的应用程序，以帮助医疗保健专业人员的决策。

ABSTRACT In this application and program of research, we will collaborate with the NRG Oncology Statistical Center to develop analytic and graphical strategies to investigate novel methods for assessing treatment tolerability, as well as to model new approaches for data presentation using data from randomized NSABP trials that contain both Common Terminology Criteria for Adverse Events (CTCAE) data and high-quality patient reported outcomes (PRO) data. We have applied these new analytic approaches and other methods to NSABP and NRG Oncology phase III clinical trials that include PRO-CTCAE items to assess treatment toxicity associated with immunotherapy and other treatments. During the first four years of the funding period, we developed a novel summary measure, the toxicity index (TI), to discriminate patients based on their overall toxicity experiences as assessed by AE grades according to CTCAE and PRO-CTCAE. TI accounts for all observed toxicity grades rather than only the most severe one, as is conventionally done. Because of its sensitivity to differences in the overall toxicity, we showed that the TI can identify predictors of treatment-related toxicity better than conventional summary scores such as max grade and average grade. In this program of research, we plan to (1) apply our methods for evaluation of endocrine therapy toxicity by use of PRO data to associations between CYP2D6 genotype and tamoxifen discontinuation in the NSABP P-1 clinical trial using plasma samples for CYP2D6 genotyping that are expected within the next six months, (2) continue our analysis of the feasibility of frequent assessment of ePRO data test the added value of weekly measurements of ePROs relative to data collection of ePROs every cycle using clinician’s CTCAE assessment as a benchmark, (3) further evaluate and disseminate the Breast Cancer Symptom Explorer visualization online tool by updating the content, functionality, technical features by returning to original focus groups and recruiting additional focus groups for further qualitative evaluation, (4) continue our analysis of symptom trajectories among postmenopausal women by exploring patient host factors associated with membership in the individual trajectories and how these impact treatment discontinuation and other outcomes, (5) continue our work on developing and building dynamic risk prediction models for treatment discontinuation and efficacy using longitudinal PROs, clinician’s assessed CTCAE, and baseline clinical and demographic data from the NSABP B-35 phase III clinical trial. We will test the added value of including longitudinal clinician’s assessed CTCAE data in addition to longitudinal PROs in the predictive performance of the dynamic model, assess and compare the predictive performance when CTCAE are summarized with our novel TI relative to average grade, and max grade, and build real-time calculators based on these new predictive models in Shiny app to aid healthcare professionals in decision making.