OBJECTIVES The aim of the study was to assess the delayed-time effect on survival of the immunotherapy with a cancer vaccine by comparing the conventional logrank test vs a weighted logrank in presence of non-proportional hazards. METHODS Data from a multicenter, open-label and randomized phase III clinical trial with an EGF-based cancer vaccine in advanced NSCLC. The diagnosis of the delayed-time effect was done and the time-delay was estimated.  The non-proportional hazards were also confirmed and tested delayed effect on survival of the treatment.  Weighted logrank tests was applied and the results were compared with those obtained using the conventional logrank test. The R software was used in the analysis. RESULTS The time-delay was estimated in 28 months and a significant effect after this time was verified.   The proportional hazard assumption was not satisfied. The median survival for the vaccinated arm was 10.37 months vs.  8.93 months for non-vaccinated arm. The difference was statistically significant by weighted logrank (p=0.04) and the conventional logrank test does not detect this difference. CONCLUSIONS Weighted logrank is substantially more efficient than the conventional logrank statistic in those situations in which non-proportional hazards are foreseen. This analysis is recommended for immunotherapies where the appearance of a late biological effect is displayed several months after randomization.