Personalized medicine in biomarker identification at their optimal cut-off selection

Retrospective exploratory analysis to identify biomarker pairs in the AVAGAST Phase III study.

The main hypothesis of this retrospective exploratory biomarker analysis is the identification of dichotomization levels based on optimal selection driven by the predictive value of single biomarkers. The outcome of interest optimization might unveil additional treatment benefits. Furthermore, testing the biomarker pairs at their optimal cut-off selection might provide the predictive score candidates.

712 plasma 92% and 727 plasma 94% tumor samples of all patients were using Cox model by identifying optimal dichotomization to maximize treatment benefit. A quadrant analysis grouped biomarker pairs into subsets yielded the best clinical benefit. Candidate biomarker score using the nested 2-fold cross-validation method was used to adjust the optimal cut-off selection.

Patients with lower VEGF-R1 at optimal cut-off with low IHER2G showed significant improvement in PFS - first line (HR = 0.62; 95% CI: 0.50 to 0.78). The interaction p-value of the biomarker pair was adjusted at 0.0147094.

The predictive biomarker is a potential candidate for PFS - first line in patients with advanced gastric cancer treated with Bevacizumab.