I am trying to figure out how many subjects and samples are needed in the following experiment.

Two groups of patients, treated and non-treated, will be monitored at a few (say 4) timepoints, via a panel of biomarkers. We intend (suggested by a statistician) to use GLM with 3 knotted splines to calculate probability of "Cured". Let's say we use 4 biomarkers to predict the outcome as "Cured" or "No Effect", setting Distribution to Binomial and Link Function to Logit.

We need to decide how many samples to collect, as they are rather expensive to analyze and the number of subjects is limited. How would one go about to do a Power calculation in this case?