Development of Algorithms for A Quantitative Real Time-Pcr Assay to Determine Egfrviii Status in Glioblastoma Patients from Rindopepimut Clinical Trials.

e14004 Background: The Epidermal growth factor receptor variant III (EGFRvIII) is a commonly expressed biomarker in glioblastoma and a target for anti-EGFRvIII therapeutic. Determination of EGFRvIII status is important for identifying patients for the treatment, as patients with the mutation are more likely to respond to the treatment. A qRT-PCR assay was developed and validated for comparison of the expression of EGFR wildtype and vIII mutant forms. However, an algorithm is needed to determine the mutation status using the Ct and delta Ct. The goal was to develop and evaluate algorithms for determining the EGFRvIII status using this assay. Methods: Results of the qRT-PCR assays of ~4,900 clinical specimens from Rindopepimut clinical trials were used to develop the algorithms. From the test results, 1763 samples were used as training set, with a subset of samples verified by bi-directional Sanger sequencing and used as test set in predictive modeling. Nine predictive models including Support Vector Machine (SVM), Partition Tree as well as an existing cut-off were tested to evaluate the prediction performance. Results: Among the nine models, the SVM and Partition Tree algorithms, demonstrated best performance with overall accuracy of 99.7% (99.93% for the positive and 98.64% for the negative) and 99.2% (99.87% for the positive and 95.93% for the negative) respectively. The Partition Tree model produced a delta CT cut-off of 10.98 that is close to the previously established cut-off of 11, and confirmed this cut-off is sound for patient sample evaluations. In combination with this cut-off, the partition tree and the SVM algorithms allow the leverage of the Ct values in addition to the delta Ct to increase the consistence and accuracy. Conclusions: A Partition Tree algorithm and a SVM model were developed and tested. Either one of the two algorithms can be added to the existing cut-off to ensure the consistency and accurate assessment of EGFRvIII in Glioblastoma, and may aid in the development of strategies for stratifying Glioblastoma patients for EGFRvIII-directed therapies. Clinical trial information: NCT00458601, NCT01480479.