Figure 1 presents the incremental costs and effects associated with the one-off and repeat-dose treatments (compared with routine treatment) in the incremental cost-effectiveness plane. Given the fixed, upfront, cost associated with the one-off treat- ment, there is very little uncertainty in the associated expected costs (essentially a horizontal line) although there is considerable uncertainty in the expected QALYs associated with treatment. For the repeat-dose therapy, there is a high level of uncertainty in both the expected costs and expected QALYs and the 2 are posi- tively correlated, reflecting the fact that the iterations where the treatment generates the most QALYs (ie, treatment is most effec- tive) involve longer treatment duration and thus greater costs.
As such, for the one-off treatment it is the iterations in which the treatment is least effective (ie, the duration of effect is the shortest) that are not cost-effective whereas the opposite holds for the repeat- dose treatment. Here, it is the iterations in which the treatment is most effective (ie, the duration of effect is the longest) that result in ICER values that are close to (above) the threshold. This is because in these iterations the costs of repeat-dose treatment are higher, and because of the assumption of waning effectiveness, the additional cost of producing each additional QALY increases in these iterations.
Figure 2 shows the cost-effectiveness acceptability curve (CEAC) and the expected value of perfect information (EVPI) associated with the one-off treatment and repeat-dose treatment, respectively, compared with routine treatment. The degree of decision uncertainty (as measured by the EVPI) associated with the one-off treatment is 4 times that of the repeat-dose treatment, with a maximum population EVPI of more than $160 million as compared with less than $40 million, and a probability of being cost-effective at the $50 000 threshold of 86% as compared with 100% for the repeat-dose treatment. Because treatment outcomes are the same in terms of QALY gains and the net present value of treatment cost is the same, and we have only a prevalence effect with no continuing incidence, the only difference between the 2 treatment profiles is the discontinuation effect, that is, the irreversibility of payment should the one-off treatment stop working. The effect of introducing an outcomes-based payment equivalent to the cost of the one-off treatment will, however change the profile of the one-off treatment to resemble the ongoing treatment (ie, the financial irreversibility is removed for the payer).