How can i add "competing casuse lines" in a Kaplan- Meier plot

[#1]

Hi,

I am trying to do a Kaplan Meier plot eg comparison of survival across 2 treatments (Placebo vs drug) ( i am using <JMP PRO 16>). Once i get my "failing" plot, i would like to control for a covariate eg age or sex. I would like to have in the same Kaplan Meier plot the additional 2 lines, of the corrected result for the covariate (eg sex). However, what JMP does is to create a second plot with these lines which is of difficult interpretation and does not help much. Is there any way to add these lines in the original Kaplan-meier plot?

Thank you in advance.

[#2]

Would you please elaborate? E.g. which platform (there are two platforms in JMP that have KME), how did you specify launch dialog, how did you configure report (steps that you have done), which plot (failure or survival) do you refer to? And what is the "cause" in this case?

[#3]

Hi, thanks for your interest.

I am not sure about the platform but i can explain what i am doing: from Analyse-->Reliability and Survival--> SUrvival.

In Y, time to event (my days up to the event), then Grouping: Treatment (Placebo vs Drug) and Censor the censor for the event (1, and 0). I then choose the Failing plot (but it would be similar the other way round for survival).

If you click on the red triangle (Product-Limit Survival Fit) you can choose competing causes which has been explained to me as "correcting for covariates". So i choose competing causes and then the Sex column of my subjects. I get a fitted Hazard (please see also Figure attaced but what i would actually like to have - if doable, is the first graph with 2 additional lines (which would be the "corrected" for sex failing plot- drug vs placebo.

[#4]

In your example, Sex, if it is what it is literally, is not a competing cause. An example of competing cause, in your example, might be cause of the event (let's say death). Died of a disease or an old age is a pair of competing causes.

I think what you need is the Proportional Hazards platform. And you should supply Treatment and Sex to Model Effects. And you can examine the effects of treatment, sex, etc.

[#5]

Indeed i was thinking to use a proportional hazard analyses. However even in that case i can get only the statistics (eg the difference between the 2 treatments on an outcome correcting for sex), but another graph as the one i showed before i cannot get.
By the way based on what you wrote about competing cause can that be both both nominal and continuous variable and would it always have to lead to the outcome within the period of observation? (eg death, or hospitalization?)
I am asking this because very broadly also sex is linked to increased cardiovascular mortality in men, but for instance cancer (yes or no) is an obvious cause of event.

[#6 · Accepted solution]

Proportional Hazards platform provides the analysis that you need if you are assuming survival functions of different gender share the same hazard function, and you want "corrected" survival function for either male or female. Survival platform does not make such assumption, so it does not suit your need. If you don't have such assumption, you may still use Survival, and supply a grouping variable with 4 categories: Male with Drug, Female with Drug, Male with Placebo, and Female with Placebo. Then you will compare survival functions of 4 distinct groups.

I don't understand your further questions regarding competing cause. I will try to explain what competing cause is again. Competing cause is associated with the event outcomes. For example, death is an event. The cause can be cancer, heart attack, or just old age, whichever occurs first. The causes are competing in that sense. Try replace the wording with Sex, does that give a funny feeling?

[#7]

Ok thank you!