Figure 1.

1A, A prognostic signature can separate patients into those with a superior survival “good signature” and those with an inferior survival “poor signature” compared to the entire group as illustrated. The signature may divide the patients into 3 or 4 groups if this is clinically desirable compared to a division into 2 groups. The signature performance may differ by histology and stage and generally the differences are less prognostic in validation sets. 1B. A signature that is prognostic and predictive requires a prospective randomized trial in which those with both a “good” and a “poor” signature would be separately randomized to receive or not receive adjuvant chemotherapy following surgical resection. Survival would be superior in those randomized to the chemotherapy.