Comparing survival curves while controlling for confounding factors

I am trying to compare three kaplan-meier survival curves. I am under the impression that a log-rank test would normally be used to accomplish this. However, I need to control for two confounding variables that vary between the three groups for whom I am measuring survival.

Would it be correct for me to do a MVA (cox proportional hazard)? I was thinking of doing a MVA in which I the independent variables are the two aforementioned confounders, as well as a third independent variable which accounts for the three groups I would like to compare.

Any help on this would be greatly appreciated.


TS Contributor
Yeah, I would definitely go with the COX model like you suggested.

Im vaguely aware of adjusting for covariates in a non parametric setting. I attached a paper I had meant to read but never did. Toward the end it talks about ANOVA and ANCOVA type models for KM curves. Pretty interesting but definetly put you to sleep.