Dose–response relationship, dose determination, and adjustment 281
0.5824 − 10 = 48.24. T he refore, the optimal strategy is to take the top path
and to use a phase II tr ial first. Based on the results, we can also calculate
p
23
as
> mean(reward*res[,2]-C3>0)
[1] 0.8737
and p
35
as
> mean(res[reward*res[,2]-C3>0,2])
[1] 0.7497676
as in Figure 9.3.
For simplicity we have assumed that the reward of success is the same
with or without the phase II trial. In practice, the time delay due to running
the phase II trial makes the reward with the phase II trial much lower than
those without it. A simple approach is to use a dis count rate (e.g., 3% each
year) to adjust for the delay. Although this approach is commonly used for
public health policy making, e.g., reimbursement, ...