Determining transition probabilities based on PFS and OS curves

I am new to TreeAge and as part of a collaboration am trying to create a cost-effectiveness analysis of 4 first-line treatments for a specific cancer.

It seems PartSA can take the difference between PFS and OS to generate the Markov transition probabilities of Progression-free --> Death, Progression-free --> Progression, and Progression --> Death. 

However, there some event-based elements I need to include like severe AEs and I don't believe the continuous accumulation of cost is accurate to my model as patients typically get treatments every 3 weeks rather than daily. So it makes more sense to me to have the model run as 3-week cycles.

Does anyone have a reference for how one can use only the data published in Phase III clinical trials to generate the necessary transition probabilities for the Markov model?

The issue I see with a previous answer is that it cites a NICE technology evaluation and the probability of death during progression free survival was calculated by actually counting the patients using the raw data.

Any guidance would be greatly appreciated.



1 comment
  • There are no event probabilities in a Partitioned Survival Model, so you probably will be better off with a Markov model. Markov models provide more flexibility for individual events with probabilities, costs, etc.

    To do this in a Partitioned Survival Model, you would have to average out the cost of events over time, which I do not recommend.

    Comment actions Permalink

Please sign in to leave a comment.

Didn't find what you were looking for?

New post