Purpose: For complex decision problems the use of decision analytic models has become an accepted standard. Given that the decision making process has to be carried out under conditions of uncertainty, it might be worthwhile to collect further data to base the decision on stronger evidence. Expected value of perfect information analysis (EVPI) that requires probabilistic decision models provides a framework to formally address the potential value of information that could be obtained through further research.

Drug eluting stents (DES) for patients undergoing revascularization procedures are about to replace bare metal stents. Although some studies on the cost-effectiveness of DES include probabilistic sensitivity analysis, none show estimates of EVPI. Within this study, EVPI and expected value of partial perfect information (EVPPI) were calculated.

Methods: A Markov model was developed and populated with data from a meta-analysis and other sources. The meta-analysis comprised 3-year follow-up data from large randomized trials. A 3-year time horizon and a US health care perspective were used. EVPI was calculated with a Monte Carlo simulation (10 000 iterations) and EVPPI with a two-level algorithm (400 inner and 40 outer loops).

Results: The paclitaxel-eluting stent is more costly (incremental costs $1256) and slightly less effective (incremental effect -0.0007 QALYs) than the bare metal stent. At a willingness to pay value of $100 000/QALY, the paclitaxel-eluting stent has a 29% probability of being cost-effective. At the same threshold EVPI is $14.5 million per 10 000 patients. EVPPI for the quality of life parameters is $26.8 million (95% CI $0 – $387.8 million) , $4.9 million (95% CI $0 – $34.8 million) for the cost parameters, $5.5 million (95% CI $0 – $33.2 million) for the relative risk parameters and $2.9 million (95% CI $0 – $38.9 million) for the baseline transition parameters.

Conclusion: EVPI calculations allow to identify the potential value of eliminating parameter uncertainty in relation to a particular decision problem. EVPPI further indicates which groups of parameters contribute the most to the decision uncertainty (here: quality of life parameters). The results of these calculations provide valuable input for decisions on future research.

However, the calculations are computationally intensive and require several days of computation time. Faster computers or better algorithms may be needed for EVPPI to become a standard part of a model based cost-effectiveness analysis.