+ Reply to Thread
Results 1 to 2 of 2

Thread: ROC analysis using survival data

  1. #1
    Points: 5,396, Level: 47
    Level completed: 23%, Points required for next Level: 154

    Posts
    45
    Thanks
    29
    Thanked 1 Time in 1 Post

    ROC analysis using survival data




    Dear readers,
    I would like to create two groups of patients depending on value of numerical parameter (like age), i.e. younger and older than 60 years... and check how they compare in survival analysis (log-rank test).
    I did arbitrary division and I noticed significant difference in survival between two groups. But I was suggested to use ROC analysis to find optimal cut off point.

    Can someone please suggest how to perform ROC analysis with one numerical variable A and survival data (follow up time and censoring status)? I tried several times but I can't figure it out. (as far as I understand, one would need tested variable A and second "classification" yes-no variable). Thank you in advance

  2. #2
    Omega Contributor
    Points: 38,374, Level: 100
    Level completed: 0%, Points required for next Level: 0
    hlsmith's Avatar
    Location
    Not Ames, IA
    Posts
    6,998
    Thanks
    398
    Thanked 1,186 Times in 1,147 Posts

    Re: ROC analysis using survival data


    Interesting question. I have experience with ROC and some with survival analysis. Your issue is that AUC will change throughout study period (time).


    I am sure someone has come up with a clever solution. It would be labor intensive, but you could come up with a potential cut-off just using logistic regression, then run PHREG (survival) and break the sample down into say deciles based on n-value minus the number of events and censored subjects. Then you could calculate AUC for each component of the decile playing around with different ages, initially guided by the logistic results. At the end you will try to find the age value with the best overall AUC.


    This is given that you are just trying to maximize AUC, you could also try to maximize a component of it, say SEN or SPEC.


    Keep searching and update us on what the literature is recommending!
    Stop cowardice, ban guns!

+ 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