Normal Tissue Expression Predicts Survival. Within the good prognosis and the poor prognosis groups, clinicopathologic parameters could no longer separate patients into partitions of different survival outcome. Meanwhile, adjacent normal tissue gene expression provided extra information to further refine the prognosis prediction. Using a LOO framework with the dimension reduction and multivariate Cox model, we assigned predicted hazard for each patient. The hazard was able to further separate the good prognosis group, but not the poor prognosis group.