This work addresses the problem that common measures of uncertainty of cost-effectiveness, ie, cost-effectiveness acceptability curves and the expected value of perfect information (EVPI), may be biased by overestimating clinical effectiveness and underestimating uncertainty. Reasons are small randomized controlled trials (RCTs) as the underlying source of effectiveness data and the overoptimistic, albeit implicit, assumption that the prior probability of the null hypothesis being false is 50%. If clinical evidence is based on small RCTs with sensitivity and specificity of 65% and 70%, respectively (LeLorier 1997), the maximum probability of cost-effectiveness decreases to 68%, irrespective of the willingness to pay. If, in addition, a 10% prior probability of effectiveness is assumed (Sterne 2001), the maximum probability of cost-effectiveness drops to 19%. Similarly, the EVPI increases 8-fold if low sensitivity and specificity of small RCTs as well as a 10% prior probability of effectiveness are considered. Therefore, traditional CEACs and EVPI analyses based on small RCTs and an implicit 50% prior probability of the null hypothesis being false should be reassessed.