Please help me with the calculation of the overall survival rate.

Dear Forum Members,

I have a question about the calculation of the overall survival rate for both control and treatment group. Suppose the one-year progression-free survival rate of the control group is 60% and that of the treatment group is 80%. Is the overall one-year progression-free survival rate for both groups 70% (suppose the ratio of subject numbers between control and treatment is 1:1)?

Would you please give me some idea on how to calculate the overall survival rate (or maybe more accurately, the overall proportion) for both arms in a trial?

Thank you.

Best regards,


Less is more. Stay pure. Stay poor.
Can you just run an empty survival model with not predictors. I believe that will provide the solution you are seeking.


Less is more. Stay pure. Stay poor.
I am sure there could be some tedious way to get at it using weights, but you have to take into account it is a time to event metric so it wouldn't be as easy as say collapsing two odds into one when not controlling for time (e.g., contingency table).

You could run the empty model to get a target estimate, then play around with weighting, but I am unsure of a direct calculation given the model is optimizing something (e.g., maximum likelihood). Keep us updated if you come up with something.

@Miner - any suggestions since you seen PHREG savvy?


TS Contributor
My experience with survival analysis is product related, so things are often done slightly different than in other disciplines. We usually have data that contains a mixture of failure modes (similar to combining the control and treatment groups above) and need to separate the mixture. Also, failure rates are misleading, because we are often more interested in the infant mortality (decreasing failure rate) or wear out (increasing failure rate) phases of life. The constant failure rate zone is misleading if you used it to predict overall life. After all, the failure rate for a white male in the USA would predict a life expectancy of over 900 years because it ignores wear out (old age). It is only useful to predict the likelihood of dying in an accident.

Having said that, take a look at this article on multimodal analysis.