Hello,
I have plotted survival curves for three outcomes (cardiovascular mortality, non-cardiovascular mortality, and all-cause mortality) and a continuous variable that has been split into tertiles ("rankvar"). However, I noticed that the lifetable of number at risk is identical for each of the outcomes. Can you help me understand what went wrong here? The curves appear very different so it seems that the analysis started out correctly...
In this case my "death" variable is 0=alive, 1=deceased.
Thanks!
Sophie
Here is the table I get with each outcome:
proc lifetest data=working plots=survival(test atrisk(maxlen=50 outside) nocensor); strata rankvar / order=internal test=(logrank); time length_yr*death(0); run;
Hello @sophiec,
Good news! I think nothing went wrong here. A patient can die only once, either from a cardiovascular event or from a non-cardiovascular event. So, at any point in time:
Hence, the numbers must be identical.
Hello @sophiec,
Good news! I think nothing went wrong here. A patient can die only once, either from a cardiovascular event or from a non-cardiovascular event. So, at any point in time:
Hence, the numbers must be identical.
Thank you so much for the explanation! It makes a lot more sense to me now and I appreciate the thorough explanation. Because I'm still quite new at these analyses, I would like to make sure that my coding of the outcome variables wouldn't impact the interpretation of the lifetable.
- I started with a categorical variable listing 5+ causes of death.
- I then created 3 dummy variables coded as 0=alive and 1=event occurred: all-cause, cardiovascular mortality, and non-cardiovascular mortality.
- I used the dummy variables in the proc lifetest command (where "death" was included in the example I provided previously)
If I'm understanding correctly, the lifetable provides the population at risk regardless of cause of death, meaning that it does not specifically relate to the dummy variable used in the proc lifetest command. Is that correct, or have I bungled something in creating these dummy variables?
Thank you again for your thoughtful and patient response!
Sophie
You're welcome.
I would say that the population at risk is the correct one for each of the three outcomes. It's just in the nature of these particular outcomes that the populations are identical. The situation would be different if non-fatal events were among the outcomes. For example, a patient who experienced a transient ischemic attack (TIA; hypothetical "outcome 1") would still be at risk for myocardial infarction (outcome 2) and death (outcome 3), i.e., they would drop out of only one of the three risk sets due to the TIA (assuming that the first such event is the outcome of interest).
The description "0=alive" might be misleading, though. For a patient who died from a non-cardiovascular disease such as cancer, the value 0 of the dummy variable associated with cardiovascular mortality must be interpreted as "censored."
As you wrote yesterday, the Kaplan-Meier curves for the three outcomes "appear very different," which (in conjunction with the strata differences) is the important result, not the three populations at risk being identical at each point in time for logical reasons.
Are you ready for the spotlight? We're accepting content ideas for SAS Innovate 2025 to be held May 6-9 in Orlando, FL. The call is open until September 16. Read more here about why you should contribute and what is in it for you!
Learn how use the CAT functions in SAS to join values from multiple variables into a single value.
Find more tutorials on the SAS Users YouTube channel.
Select SAS Training centers are offering in-person courses. View upcoming courses for: