0

I'm trying to compare survival between two groups. While overall survival shows to be not significantly different, survival at fixed time points proves to be significantly different at set time points in the first five years.

I was wondering if there is any way to determine the time point where the difference has a pvalue >= 0.05 (similar to the crosspoint() function from the ComparisonSurv package that just defines where the graphs cross. So I don't have to manually try out different time points with the fixed.point() function.

Thanks!

Kathi
  • 1
  • I don't think this is a simple, properly-defined question on a statistical basis. I think you _could_ serially conduct multiple interim tests but if you used 95% as your confidence level at all of the time points, you would be violating the assumptions of the testing. There's an entire literature on sequential testing and the "multiple comparisons problem" in the survival analysis literature. That theory is used in designing interim test that do not violate the statistical assumptions. You should read up on sequential testing in clinical trials. My close vote is on the basis of no [MCVE]. – IRTFM Jul 01 '21 at 18:36
  • Here's a paper that might be useful in learning about sequential testing issues and approaches to their resolution on better statistical foundations: https://web.archive.org/web/20150408112729/https://www.niaid.nih.gov/about/organization/dcr/BRB/Documents/bvaluejcgs.pdf – IRTFM Jul 01 '21 at 18:54
  • And there is an R package that accopanies that paper: https://cran.r-project.org/package=MChtest – IRTFM Jul 01 '21 at 19:05

0 Answers0