Читать книгу Clinical Pharmacology and Therapeutics - Группа авторов - Страница 150

As has been discussed, pharmacoeconomics is not an exact science and the assessment of cost‐effectiveness requires a number of assumptions and/or extrapolations to be made. A good analysis should test the effect of varying these assumptions across a range of likely possibilities, showing which assumptions are the main drivers of the ICER – this is known as sensitivity analysis.

In a simple analysis, this can be done by varying assumptions one at a time (univariate analysis). For example, if the estimated survival with treatment A is only 10 years (i.e. we take away the survival gain), then the QALY gain falls to 2 QALYs (10 × 0.7 = 7; 10 × 0.5 = 5, QALY gain = 7 − 5) – the ICER is now £5 000 (£10 000/2) which remains within usual cost‐effectiveness limits. The survival gain, which may be based on extrapolation from the clinical trial, is not essential for the drug to be considered cost‐effective. The utility values could be varied similarly.

In a more complex model involving multiple health states and more complex costs and benefits, it may be more appropriate to vary all the parameters at once across a range of likely values, giving a large number of possible ICERs reflecting different combinations of assumptions. This multivariate approach is often called probabilistic sensitivity analysis as it shows a range of possible ICERs with an estimate of the likelihood of the actual ICER being below any specified threshold value.