Purpose: To demonstrate that decisions based on expected incremental net benefit (INB) will be inappropriate when agencies cannot commission or mandate additional research. To establish the opportunity costs of decision uncertainty and illustrate their impact on appropriate decision rules in cost-effectiveness analysis.

Background: The decision to adopt a technology should be based on expected INB and the uncertainty around that estimate should inform the simultaneous decision to demand or commission further research. However, institutions with the remit for making adoption decisions often lack the remit to demand that further research is conducted and are separated from those responsible for commissioning research. Such decision-makers do not directly control future research, so the adoption decision is their only policy instrument.

Methods: The prospects of acquiring further evidence about a technology are not independent of the adoption decision. In the face of a positive adoption decision, incentives for conducting future research can be reduced and further experimental research may be regarded as unethical. Hence, the opportunity cost of adoption should include the forgone expected value of information (EVI) for current and future patients. The opportunity costs of rejecting an apparently cost-effective technology will be the INB forgone by current and future patients. Using decision theory and value of information analysis, it is possible to establish appropriate decision rules which take account of these opportunity costs and an assessment of the uncertainty surrounding when new evidence may become available.

Results: The expected net benefits of immediate adoption vary between a maximum gain equal to INB of the optimal strategy based on current evidence and a maximum loss equal to EVI. We show that the expected net benefits of adoption fall as the incremental cost-effectiveness ratio (ICER) approaches the threshold and decision uncertainty increases. This introduces incentives on sponsors to provide more evidence or reduce the price of technologies. By integrating over the uncertainty surrounding the arrival of further evidence, it is possible to determine decision rules in terms of expected opportunity loss.

Conclusions: Decisions based on expected INB do not account for the full opportunity costs of a decision to adopt a technology. Failure to account for these opportunity costs results in decision rules that are too permissive and will undermine the evidence base for clinical practice.