Model analysis is an extremely powerful tool for estimating cost-effectiveness over time spans that exceed the duration of clinical trials, but it also presents problems. The most significant is the handling of the uncertainty surrounding the parameters applied. In dice, for example, there is no uncertainty about the probability of throwing a "1" (1/6). This is not the case with the efficacy rate for antihypertensive pharmaceuticals. Let us assume that one clinical trial yields a figure of 85 percent as the efficacy rate, but also that the 95-percent confidence interval is 75 ~ 95 percent. In such a case, the true efficacy rate would fall within this interval with 95-percent reliability, but there is no telling exactly where. This is to say that some uncertainty is associated with the 85-percent efficacy rate.

In order to confirm the influence of such uncertainty, sensitivity analyses are conducted in pharmacoeconomic analyses using models. A sensitivity analysis is a way of examining any change in results due to changes in the parameter values applied within a certain scope. In the case of the aforementioned 85-percent efficacy rate, for example, it would proceed by varying this rate in the 95-percent confidence interval (75 ~ 95 percent) and checking the robustness of the conclusion that the pharmaceutical is cost-effective. Such variation in the parameter values naturally leads to a change in the numeric output, but the conclusion that the pharmaceutical is cost-effective will be supported if the ICER, nevertheless, falls within the ceiling value (e.g., 6 million yen). A sensitivity analysis using only one parameter is termed a one-way sensitivity analysis (see Figure 9), while one that simultaneously varies two parameters is a two-way sensitivity analysis (see Figure 10).

Figure 9 One-way Sensitivity Analysis

Figure 10 Two-way Sensitivity Analysis

More sophisticated types of sensitivity analysis treat parameter uncertainty as a probability distribution and check the results in terms of probability. These are called "probabilistic sensitivity analyses". The major type is the Monte Carlo simulation (Figure 11). Probabilistic sensitivity analysis enables determination of the probability of the ICER falling under the ceiling value. A graph with the ICER ceiling value on the horizontal axis and the probability of the ICER falling under this value on the vertical axis produces a curve for cost-effectiveness acceptability. This curve often appears in analyses these days. From Figure 12, for example, it can be seen that the probability of the ICER for the drug under consideration not exceeding six million yen is about 60 percent.

Figure 11 Results of Monte Carlo Simulation

Figure 12 Cost-effectiveness Acceptability Curve