Weighted Kaplan Meier for a Case Control study

[Nancych]

Dear all,

We have a study with 10 patients treated. All of them were found the matches from another study which are natural history patients might using the ratio of 1:3, 1:2 or 1:1 (based on certain rules), thus we assigned those matched patients with weight of 1/3, 1/2, or 1 individually, and of course 1 for each treatment patients. To consider those weights, for KM plots and related percentiles, is there any SAS procedure you could suggest? The inverse probability was not applied to this case.  

Thank you very much!

Nancy

[SteveDenham]

There are others who have a better knowledge of PROC LIFETEST than I do ( @FreelanceReinh ), but it seems that the WEIGHT statement would provide what you are looking for. I think the values you give as weights should be inverted as the documentation says the weights should be assigned based on the inverse probability of belonging to the group.  So if the matching is 1:1 the weights would be 1 and 1, for 1:2 matching the weights would be 3 and 1.5, and for 1:3 the weights would be 4 and 1.33333.  So looking at those values and taking common denominators, it appears that the weight of the treated group would be the coefficient for matching of the historical group, such that 1:1 matching would have equal weights 1 and 1, 1:2 would have weights 2 and 1, and 1:3 would have weights 3 and 1.

But I could be really, really wrong.

SteveDenham