Candidate for the Lee B. Lusted Student Prize Competition
Purpose: To show that cost-effectiveness acceptability curves (CEACs) are contingent on the number of alternative treatment strategies compared and explain how this compromises their use as an objective measure of uncertainty in cost-effectiveness analysis (CEA).
Methods: CEAs typically compare a finite number of treatment alternatives. However, in many cases the actual number of treatment alternatives is very large or infinite. In such cases the cost-effectiveness frontier is not composed of a discrete number of points in the cost-effectiveness plane, but can be continuous. We use the example of a hypothetical intervention with a continuous dose-response relationship to show how increasing the number of treatment alternatives compared influences the shape of CEACs. How these curves change depends on the correlation of uncertainty between treatment alternatives in the probabilistic sensitivity analysis used to derive the CEACs. We compare the cases of perfect and zero correlation and an intermediate case, in which the correlation between alternatives increases with their proximity in the cost-effectiveness plane.
Results: In the case of zero correlation, increasing the number of treatment alternatives causes the CEACs to fall towards a probability of zero (Figures 1 & 2). With perfect correlation, the curves lie at probability of zero and jump to a probability of one over the range of the cost-effectiveness threshold where the given intervention has the highest net benefit. As the number of alternatives included grows large, the portion of the CEAC lying at probability of one converges to a single spike. In the intermediate case, the CEACs may initially lie at probability of zero or one, as in the case of perfect correlation, but eventually fall towards zero as the number of alternatives grows large (Figures 3 & 4).
Conclusions: This analysis shows that CEACs are contingent on the number of treatment alternatives compared. Without an objective basis to choose the number alternatives or the increments between them, the resulting CEACs seem arbitrary in part. Consequently, the usefulness of CEACs as an objective measure of uncertainty is questionable when many treatment alternatives are possible.
See more of: The 33rd Annual Meeting of the Society for Medical Decision Making