10.2217/fon-2015-0085 © 2016 Future Medicine Ltd At the end of the year, it is time to take stock. What has been the revolutionary innovation in oncology in 2015? Key steps in translational research are certainly strengthening the foundations of cancer knowledge. As genome sequencing has allowed outlining the mutational landscape of several tumor types [1–3], the historical classification based on mere tumor histology is giving way to a molecular cancer definition. In short, we can differentiate tumors relying on a specific dominantmutated oncogene for cell proliferation and survival (i.e., EGFR-mutated or ALKrearranged lung cancer, BRAF-mutated melanoma, HER2-positive breast cancer), from cancers with high genetic instability (i.e., melanoma or lung cancer wild-type for driver mutations, renal cell carcinoma [RCC], prostate cancer). Targeted agents directed against gatekeeper mutations currently represent the standard of care for oncogene-addicted tumors; immunotherapy has ousted chemotherapy for malignancies with high mutation loads. What remains the main goal to achieve when testing a new cancer treatment? The answer is prolonging overall survival (OS). For years, we used surrogate end points for OS (i.e., progression-free survival [PFS], disease-free survival, time to progression), supposing that early achievement of the study’s end points would primarily result in positive outcomes for the patients (due to the quicker approval of new agents), independently from the economic logic of the big pharmaceutical companies. This approach convinced us of the great benefit of prolonging PFS, inconveniently forgetting (or conveniently?) the unproven correlation with OS. Indeed, validation studies of surrogate end points often failed in demonstrating strong correlation with survival, highlighting the limited evidence supporting the use of surrogate end points in oncology [4]. Can PFS be considered a surrogate end point of OS? Theoretically, a significant advantage in PFS should result in prolonged overall survival. Actually, the PFS/OS correlation model appears to closely depend on the mechanism of action of the drug, aside from the tumor type. Curves comparing a targeted-therapy in oncogene-addicted tumors separate early and markedly, but then may converge. Clinically, this corresponds to an early and significant benefit of the targeted agent in a molecularly selected population, and to the succeeding eventual development of resistance. In this case, PFS not always correlates with OS. On the contrary, with COMMENTARY