+ Reply to Thread
Results 1 to 2 of 2

Thread: Reconciling differences seen in survival (log-rank vs logistic regression)

  1. #1
    Points: 7, Level: 1
    Level completed: 13%, Points required for next Level: 43

    Posts
    1
    Thanks
    0
    Thanked 0 Times in 0 Posts

    Reconciling differences seen in survival (log-rank vs logistic regression)




    Doing survival analysis on an intervention that has been shown to significantly improve overall survival (OS) in previous studies (time and again).

    When I assess OS using log-rank in univariate analysis, the intervention is not associated with overall survival. However, if I stratify my cohort by 5-year survivors and perform bivariate logistic regression, I find that the intervention is significantly associated with 5y survivors.

    I know that the tests are inherently different, so that is probably part of it, if not all. But when I look at the data in detail, I found several patients who did not receive the intervention that lived a long time. Can that skew my log-rank analysis for OS? What else should I look into to explain this?

    Thanks.

  2. #2
    Point Mass at Zero
    Points: 13,828, Level: 76
    Level completed: 45%, Points required for next Level: 222
    ledzep's Avatar
    Location
    Berks,UK
    Posts
    684
    Thanks
    188
    Thanked 143 Times in 139 Posts

    Re: Reconciling differences seen in survival (log-rank vs logistic regression)


    1) The univariate log-rank tests are all great. BUT, there are many factors that will affect your survival. Ideally, you should compare the differences between the intervention (or comparator) using a multivariate cox regression model adjusting for the possible confounder factors (e.g. Age, weight, dose etc...disease specific)

    2) Its interesting that you observe difference in 5-year survival estimates but not on overall survival. It could be possible that standard log-rank test might not be appropriate. For example, if your survival lines crosses early on, then you might have to adjust for this (e.g. using Gehan's test when the survival curves crosses)

    Hope this helps
    Oh Thou Perelman! Poincare's was for you and Riemann's is for me.

+ Reply to Thread

           




Posting Permissions

  • You may not post new threads
  • You may not post replies
  • You may not post attachments
  • You may not edit your posts






Advertise on Talk Stats