if you find that a variable's p-value for difference between two groups is not significant can you assume it wouldn't be significant in a cox proportional hazards model looking at OS between the two groups (doesn't need to be adjusted for in the model)?