In many documentation I have read that Kaplan-Meier curves followed by logrank test and/or cox models are the most recommended statistical methods to analyse Survival and test for different factors that may impact the Survival.

However, I have also heard that these are suitable if I have many "times" points in the data, but they may not be the best choice when I want to compare at one, two or three times only (let say at t=0 , t=7 days and t= 14days). In many examples with the methods above, they have at least 10 "times" or dates where the alive/dead information is available. Should I consider something else to enquire about the effect of a group or factor on the Survival ?

I was considering glms with a binomial distribution at only one target date (I remove day=0) and then compare the models (reduced vs non-reduced) by an anova, but I am not sure if this is appropriate for Survival and what assumption for such glm binomial family I should verify beforehand. The good point of glms would be the possibility to test for an interaction between two factors, which is not possible with the logrank test nor with cox models.

Many thanks if you have any good comments or advices to provide !