Dose Finding using Bayesian methods: Continual Reassement Method (CRM)
We have an urgent probel:
We are analysing a dose-finding phase I clinical trial. The method used
is the CRM using a Bayesian approach (Chevret Stat Med,
1993-1093-1108):
6 doses are available. For each of them an a priori probability of
toxicity is chosen.
The mathematical model for dose-response is:
p(i)=(exp(3+alpha.x(i)))/(1+exp(3+alpha.x(i))) (logistic model)
We assign the first patient to the j dose. After observing the
dichotomous outcome of toxicity, the posterior distribution of alpha
would be calculated via Bayes theorem.
This process continues until we reach a predetermined fixed sample size.
We would like to calculate
-the intermediate a posteriori probability curves at each step
- the confidence interval of the final probabilities (it is our major
problem.
Do you think BUG can help we to solve the problem?
How can we proceed?
Thanks
Gabrielle Boissard
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|