TEN WAYS TO ENHANCE THE ESTIMATION, REPORTING AND INTERPRETATION OF THE COST-EFFECTIVENESS OF CANCER SCREENING INTERVENTIONS

Purpose: To explain simple steps that can enhance the usefulness of cost-effectiveness analyses of cancer screening interventions to policy makers.

Method(s): We use examples from the cancer screening literature to show how the reporting and interpretation of cost-effectiveness estimates can fail to lead to optimal policy choices and how this can be avoided. The examples chosen are primarily from a recent systematic review of the cost-effectiveness of cervical screening, but also include examples found in the breast and colorectal screening literature.

Result(s): The ten recommendations are: (i) report costs and effects, rather than just incremental cost-effectiveness ratios (ICERs) or a cost-effectiveness plane; (ii) present a cost-effectiveness plane; (iii) report cost and effects for all strategies, not just those on the efficient frontier; (iv) do not report ICERs for dominated strategies; (v) do not report multiple ICERs for each strategy based on comparisons with multiple comparators (vi) report costs and effects to sufficient significant figures to permit at least approximate replication of the reported ICERs; (vii) include all relevant comparators considered in the basecase analysis; (viii) do not report ICERs for strategies for which it is anticipated the inclusion of additional strategies would lead to significant changes in the estimated ratio; (ix) when there are multiple factors to vary in a screening programme, only vary these factors one at a time when creating alternative strategies; (x) if possible, attempt to include a sufficiently large variety of screening strategies such that the analysis yields an efficient frontier with ICERs that increase gradually from below the cost-effectiveness threshold to above it.

Conclusion(s): The cost-effectiveness estimates from simulation models are dependent on the choices taken by analysts regarding both the modelling of alternatives and the interpretation of the resulting estimates. This is particularly so in the case of cancer screening interventions, as multiple intervention intensities are possible due to variation in screening intervals, start ages, stop ages and testing technologies. Our analysis shows that there are a number of flaws in economic appraisals of cancer screening interventions, which although may seem obvious, occur with surprising frequency in the literature. The ten-item list presented here will facilitate easy-to-implement improvements that should yield more reliable and relevant evidence for policy makers.