Problem in probability sensitivity analysis (PSA) for a partitioned survival (PartSA) model

Dear community,

I recently encoutered a problem when doing the PSA for my PartSA model.

I used literature reported numbers to calculate the parameters of exponential distribution for the progression free survival (PFS) and the overall survival (OS) of a cancer treatment, and then assign normal distributions to those parameters.

But as PFS and OS curves only seprate mildly and the confidence interval for survival time is large, when doing Monte Carlo simulation, there are considerable possibilities of drawing numbers that makes the survival curve cross.

Do you know if there is a way to make sure that the number drawed from OS would always be greater than the number drawed from PFS so that the survival curve wound not cross please? Thank you so much!




1 comment
  • Sibo

    From your question, you have exponential distributions you are using for PFS and OS curves - making the assumptions they have a fixed rate of change.

    It then seems as though you are doing Probabilistic Sensitivity Analysis on the model. Because of the nature of sampling, there is no way to ensure that the parameter for your exponential distribution for OS is always greater than for your PFS curve.

    You could probably set up your model to ensure this is the case, but that would likely involve some conditional sampling, or setting the the ranges to that the parameters for the Exponential Distributions are always in the correct order as such.

    You can submit your question to if you have an active license to discuss further.


    Comment actions Permalink

Please sign in to leave a comment.

Didn't find what you were looking for?

New post