
226 Design & Analysis of Clinical Trials for Economic Evaluation
The chance of switching decision appears to plateau after 1000 patients
(Table8.8). Hence, any further data collected by way of clinical trials or obser-
vational studies are likely to result in ‘erroneous’ decisions not greater than
40. One may perform EVSI computations by taking into account uncertainty
in QoL or other parameters. The choice of the prior would be important, and
it might be preferable to simulate from correlated prior distributions of costs
and effects, which may be technically challenging.
A separate question is whether 33% is an acceptable risk to the decisi