Hi all,
In the review of a scientific article in which I made two type of survival analysis 1) Kaplan-Meier method with log-rank test and 2) Cox regression model, I received the following recommendation of the Editorial Board:
"If the author decided to use semiparametric approach, proportional hazard assumption should be tested first. If the assumption met, the author should report the result from Cox regression. In term of the survival curve, since the author adjusted for covariates, the survival curve can be produced based on the clinically meaningful definition of the covariate. For example if age is the covariate, it could be mean age, so that the survival curve will represent the survival of patient with average age. If the assumption not met, the author may consider to use non-parametric approach, but need to respect to the shape of the survival curve before choosing a test to test the difference between the curve (such as Log-Rank test). ".
The assumption has been met and I will report the result from Cox regression, but I have doubts about what I have to do with survival curves: ¿should I keep the same graph made with sts graph, failure by () without reporting the long rank test? or ¿should I create a new graph using sts graph, failure by () adjustedfor ()?
Thanks,
Germán
In the review of a scientific article in which I made two type of survival analysis 1) Kaplan-Meier method with log-rank test and 2) Cox regression model, I received the following recommendation of the Editorial Board:
"If the author decided to use semiparametric approach, proportional hazard assumption should be tested first. If the assumption met, the author should report the result from Cox regression. In term of the survival curve, since the author adjusted for covariates, the survival curve can be produced based on the clinically meaningful definition of the covariate. For example if age is the covariate, it could be mean age, so that the survival curve will represent the survival of patient with average age. If the assumption not met, the author may consider to use non-parametric approach, but need to respect to the shape of the survival curve before choosing a test to test the difference between the curve (such as Log-Rank test). ".
The assumption has been met and I will report the result from Cox regression, but I have doubts about what I have to do with survival curves: ¿should I keep the same graph made with sts graph, failure by () without reporting the long rank test? or ¿should I create a new graph using sts graph, failure by () adjustedfor ()?
Thanks,
Germán