Optimal Predictors of Response to Trastuzumab

最近，曲妥珠单抗在佐剂环境下的试验已于#年完成。 美国(NCCTG/NSABP)和欧洲(HERA试验)都表明这种药物可以降低HER2阳性乳腺癌患者的复发率。然而，在这两个试验中，仍然有相当大比例的患者复发使用该药物。同样，许多早期对转移性癌症患者的试验也显示缺乏反应，即使他们有HER2表达的肿瘤。这些事实，再加上现在有针对她的家庭信号通路的新药可用，表明可以为曲妥珠单抗开发新的、更具体的伴随诊断方法，从而提高选择患者进行这种治疗的特异性。 这一提议的基本假设是，对结果的最佳预测 对于服用曲妥珠单抗的患者，可以通过组合预测疗效的多个标志物来实现。我们建议使用一套新的技术来询问接受曲妥珠单抗和化疗治疗的HER2肿瘤，我们可以定义最佳预测因子，并在前瞻性试验中验证该分类器。这些技术将包括评估DNA、RNA和蛋白质，以确定最佳预测的最佳模式。蛋白质将使用多重免疫荧光进行评估，RNA将使用转录微阵列分析进行评估，DNA将以拷贝数分析的形式使用高密度SNP阵列进行评估。这些检测方法中的每一种都有可能被转化为乳腺癌患者可用的配套诊断方法。在这个修订后的两年版本中。对于这笔赠款，我们建议保留对所有3种模式的分析，但只完成赠款的培训集方面。我们设想在18-24个月内提交后续报告，显示该项目努力的结果模型，然后建议在CALGB 40601队列或类似队列上验证模型(S)。调整后的目标包括： 目的1)利用HER途径相关蛋白、下游信号蛋白和HER2的异二聚化或同源二聚化状态，建立一个最佳的多重预测模型。 这个目标将使用定量多路技术(称为AquA)来准确地原位测量一系列10-25 HER2途径相关蛋白的蛋白表达，以构建一系列模型来预测CALGB 9840试验中曲妥珠单抗的反应(紫杉醇和曲妥珠单抗在一线转移环境中的试验)。 目的2)利用基于ILumina DASL的基因表达谱建立最优的多重预测模型。 这一目标将使用CALGB 9840队列中Illumina DASL定制阵列(1536个基因)平台上的定制基因集来评估基因表达，以确定多路预测模型的候选预测因子。具体地说，与曲妥珠单抗反应相关的候选扩增片段和与这些候选扩增片段相关的基因将被识别为包括在模型中。 目的3)对来自目标1和目标2的组合数据进行计算建模，以找出最适合训练集(CALGB 9840)数据的3-5个模型。 这一目标将创建一系列最佳模型，以便最好地从CALGB培训集中的非应答者中选择应答者。模型的验证将通过省略一个交叉验证方法来完成，以期在随后的研究中在独立的队列中进行更可靠的验证。

Recently, trials of trastuzumab in the adjuvant setting have been completed in both the US (NCCTG/NSABP) and in Europe (the HERA trial) that have shown this drug can decrease the recurrence rate in patients with HER2 positive breast cancer. However in both trials, there are still a significant percentage of patients that recur on the drug. Similarly, numerous earlier trials in patients with metastatic cancer also showed a lack of response even though they had HER2 expressing tumors. These facts, combined with the facts that new drugs are now available that target HER family signaling pathways, suggest that new, more specific, companion diagnostics could be developed for trastuzumab that increase the specificity of selection of patients for this therapy. The underlying hypothesis of this proposal is that an optimal predictor of outcome for patients on trastuzumab can be achieved by combining multiple markers which predict response. We propose that using a set of novel techniques to interrogate HER2 tumors being treated with trastuzumab and chemotherapy, we can define the optimal predictor and validate this classifier in a prospective trial. The techniques will include assessment of DNA, RNA and protein to determine the best modality for optimal prediction. Protein will be assessed using multiplexed immunofluoresence, RNA will be assessed using transcriptional microarray profiling and DNA will be assessed in the form of copy number analysis using high density SNP arrays. Each of these assays has the potential to be translated into a usable companion diagnostic assay for breast cancer patients. In this revised 2 year version. of this grant we propose to keep the analysis of all 3 modalities, but to only complete the training set aspects of the grant. We envision a follow-up submission in 18-24 months that shows the model resulting for the efforts of this project, then proposing validation of the model(s) on the CALGB 40601 cohort or similar. The scaled aims include: AIM 1) To develop an optimal multiplexed predictive model that uses HER pathway related proteins, downstream signaling proteins and hetero- or homodimerization state of HER2. This aim will use the quantitative multiplexing technology (called AQUA) for accurate in situ measurement of protein expression on a series of 10-25 HER2 pathway related proteins to construct a series of models that predict response to trastuzumab in the CALGB 9840 trial (A trial of taxol and trastuzumab in in the first line metastatic setting) AIM 2) To use Ilumina DASL-based Gene Expression Profiling to develop an optimal multiplexed predictive model. This aim will assess gene expression using a custom gene set on the Illumina DASL custom array (1536 genes) Platform in the CALGB 9840 cohort to identify candidate predictors for the multiplexed predictive model. Specifically, candidate amplicons, and genes associated with these candidate amplicons, which are associated with response to trastuzumab will be identified for inclusion in the model. AIM 3) To do computational modeling of the combined data from Aim I and 2 to discover the best 3-5 models that fit the training set (CALGB 9840) data. This aim will create a series of optimal models that best select responders from non-responders in the CALGB training set. Validation of the models will be done by Leave One Out Cross Validation methods in anticipation of future more robust validation in an independent cohort in a subsequent study.