Purpose: For diseases with a largely unknown underlying disease process or progression, the cumulative incidence of disease, over e.g. 10 years time, is often reported. For use in a decision model this is typically converted into a constant or average annual risk. However, this conversion ignoring uncertainty on disease progression may lead to biased results of subsequent cost-effectiveness analyses. We investigated this bias, for different screening strategies, by comparing outcomes associated with time-varying risk profiles, with outcomes for constant risks.

Methods: We simulated a large set of time-varying risk profiles, based on different variations of the risk over time (i.e. the occurrences and shapes of peaks in the risk profile). Screening strategies were defined by their screening interval and their ability to detect disease. Treatment effect was defined by a fixed period directly following treatment in which all individuals were assumed to be free of disease. For illustrative purposes we performed cost-effectiveness analyses for various risk profiles corresponding to the growth pattern of intracranial aneurysms and the risk of subarachnoid hemorrhage.

Results: The risk reduction achieved by screening for intracranial aneurysms ranged from to 85% to 95%, for a constant annual risk, while for time-varying risks this reduction ranged from 0% to 90%. Although smaller screening intervals resulted in increased risk reduction, the ICER of screening versus not screening also increased rapidly for all strategies, unless treatment was assumed to result in an unrealistically long-term period free of disease.

Conclusions: When disease progression is uncertain, simply converting a cumulative risk into a time-independent, annual risk leads to biased, too optimistic results of cost-effectiveness analyses. Slight irregularities in the risk of disease result in new, optimal screening strategies, with smaller screening intervals and higher ICERs (compared to not screening). For low cost highly accurate screening techniques this new ICER may still be acceptable. Large irregularities in the risk of disease always result in a dramatic deterioration of screening efficacy, and unacceptable ICERs. Evidence-based assumptions on the irregularity of growth rates of intracranial aneurysms yield unacceptable ICERs, for all possible screening strategies. We can therefore question the use of existing screening strategies, aimed at the prevention of subarachnoid hemorrhage from intracranial aneurysms.