topic Re: The difference of median survival time in Statistical Procedures

The difference of median survival time

Faker — Mon, 16 Sep 2024 03:08:54 GMT

Hello,

I am conducting a survival analysis using proc lifetest to compare two treatments. I need to obtain the difference in median survival times between the two treatments and the corresponding 95% confidence interval. However, the standard output seems to only provide the medians and 95% confidence intervals for each treatment, but not the difference between the treatments. Is there a way to achieve this in SAS?

Thank you very much

Re: The difference of median survival time

Ksharp — Mon, 16 Sep 2024 03:37:40 GMT

Re: The difference of median survival time

Faker — Mon, 16 Sep 2024 06:39:22 GMT

Hi Ksharp,

Thank you for your reply.

I saw this example in the document (as shown in the figure below). Since the code performs a log transformation, the estimated coefficient after taking exp should be a ratio rather than a difference.

I would like to ask further how to obtain the median difference and its 95% CI in this procedure .

Alternatively, I have thought of two other ways to handle this. I'm not sure if they are appropriate:

1. Using the formula provided in the example, first calculate the 50% survival time for both groups and also calculate their 95% CI. The median difference and its 95% CI can be obtained by subtracting the 50% survival times of the two groups.

2. Remove the log transformation from the code. The estimated value obtained will be the difference between the two groups, which would be the median difference.

Thank you!

Re: The difference of median survival time

Ksharp — Mon, 16 Sep 2024 07:12:14 GMT

I don't agree with 1).

But agree with 2) .

Here is an example:

data have;
set sashelp.bmt;
if Group ne 'ALL';
run;


proc quantlife data=have  seed=1268;
class group;
model T*Status(0) = group / quantile=(0.25 0.5);
run;

Using PROC LIFETEST to check the median of these two groups:


proc lifetest data=have;
strata group;
time T*Status(0);
run;

Here 183-2204=-2021 . is the same as PROC QUANTLIFE .