On the probability of cost-effectiveness using data from randomized clinical trials