Background: Evidence for the hypothesis that personalization of therapy optimizes benefit is a subject of debate. We compared efficacy between US Food and Drug Administration (FDA) approved cancer treatments deploying a personalized strategy vs. those without a biomarker-based rationale. Methods: We reviewed registration trials (based on package inserts from Drugs@FDA) that evaluated new agents receiving FDA approval (Sept. 1998 to June 2013) for cancer. Pooled analysis of response rate (RR), progression-free survival (PFS) and overall survival (OS) for all trials and hazard ratios (HRs) for time-to-event endpoints for randomized trials were compared between personalized (defined as using a drug matched to a biomarker or in a population where > 50% of patients harbor the cognate target) vs. non-personalized trials. Results: A total of 58 approved drugs were included (57 randomized [32% personalized] and 55 non-randomized trials [47% personalized]; N=38,104 patients). Personalized trials more often included oral drugs (68% vs. 35%, P=0.001), single agents (89% vs. 71%, P=0.036) and more frequently allowed crossover to experimental arms than non-personalized trials (67% vs. 28%, P=0.009). RR was significantly higher (49% vs. 25%, P<0.001) and PFS was longer (8.5 vs. 5.5 mos, P=0.001) with a personalized approach. In the multilinear regression analysis, personalized therapy was selected as an independent factor predicting a higher RR (P<0.001) and longer PFS (P=0.002). Median HR for PFS of experimental to control arms was 0.37 for personalized vs. 0.58 for non-personalized trials (P=0.003), indicating that, in randomized trials, the personalized experimental arm had significantly more improvement in PFS. Personalized trials also showed a longer OS (median= 18.2 vs. 13.5 mos, P=0.035 in the multilinear analysis), even though crossover was allowed more frequently in personalized trials. Median treatment-related mortality did not differ (0.4%, personalized vs. 1.0%, non-personalized, P=0.22). Conclusions: Personalized trials were associated with more pronounced clinical benefit, expressed by significantly higher RR and longer PFS and OS, despite the latter being confounded by more frequent crossover.