cox regression v. kaplan-meier

#1
hi everyone! thanks in advance for your help!

i'm studying whether the presence of high v. low hormone levels (of several different hormones) affects survival - i assume that i am to use kaplan-meier to analyze the data rather than cox regression.

i am also not sure how to get SPSS to report whether the difference in survival is significant.

thanks!!!
 
#2
Hi cayetano,

Look for the log-rank test when creating the KM curves. That will test for a difference between hormone groups. You can also create a Cox proportional hazards model with hormone level as the only independent variable. If the hazard ratio is significantly different than 1, then you can say there is a difference in survival between groups. The hazard ratio will point to which group is more likely to have an event (if <1, then the reference group is more likely...>1, then the comparative group). I'm not familiar enough with SPSS to guide you in either method, but I would bet SPSS has some examples you can replicate.

Good luck!
 

Shavi

New Member
#3
By Definition: Whereas the Kaplan-Meier method with log-rank test is useful for comparing survival curves in two or more groups, Cox proportional-hazards regression allows analyzing the effect of several risk factors on survival.


You can use K-M statistics to find the association between hormones level and survival.

From the menus choose:

Analyze
Survival
Kaplan-Meier...

and follow the instructions. you can also select help in this window to see the details about data, assumptions and procedures.

Output: Select log rank test from "Compare factors" tab and survival plot from "options" tab. Log rank test p-value can define the association and you can see the pattern of survival for hormone levels from the survival plot.


Cox regression is next step if you want to see multiple independent (risk) factors collective effect on survival. Read some good literature or book chapters about K-M and Cox regression to clear your doubts and then proceed because understanding is more important here before application :)
 

Link

Ninja say what!?!
#5
Just throwing this out there:

Whilst the Cox model is great for analyzing multiple risk factors, you don't get this without giving something up. The Cox model is more prone to bias in the estimates of survival and hazards than the KM model.