OBJECTIVES:  To demonstrate the limitations of the incremental cost-effectiveness ratio (ICER) as a measure of improved cost-effectiveness in the context of periodic screening and to propose a new measure based on net benefit that quantifies the proportional gain in cost-effectiveness relative to the status quo. METHODS:  We use published cost-effectiveness estimates of cervical cancer screening to show why the ICER is not an appropriate measure of improvement in cost-effectiveness when screening performance improves, either through enhanced test characteristics or better risk stratification. We then propose a new metric based on the net health benefit measure previously devised by Stinnett and Mullahy. RESULTS: It is shown how an improvement in screening may enhance cost-effectiveness, represented by an outward shift of the efficient frontier in the cost-effectiveness plane, but that this improvement is not necessarily reflected in the ICER. This is because the whole efficient frontier may shift when all strategies are affected by a common technological change, and so ICERs on the frontier can be insensitive to this improvement. It is also shown that the ratio of costs to effects of a given strategy before and after a technological improvement also does not provide a useful measure of improved cost-effectiveness. The alternative measure of the proportional increase in net health benefit following the adoption of a new technology is then demonstrated. This metric quantifies the increase in net health benefit resulting from the new technology over a range of threshold values. CONCLUSIONS: ICERs will remain important for the identification of optimal intervention strategies. However, they are not suited for all circumstances. The net benefit alternative proposed here is a simple quantification of improved cost-effectiveness. The results presented here most readily apply to screening interventions, but also have application in other cases in which a technological development enhances multiple strategies simultaneously.