Screening for Melanoma and Options For-Its Evaluation

Abstract

A review of the published evidence presented here argues that screening for melanoma is recommended and practised at present, but with wide diversity of opinions about its value; there is evidence that screening has considerable potential for benefit, but the evidence of actual benefit is limited; and there are substantial costs and potential hazards from screening.

On this basis the evaluation of screening procedures for melanoma is important, and options for this are discussed. The ideal study design to assess the efficacy of melanoma screening in reducing mortality is a large scale randomised trial. This may need a well coordinated proposal involving several centres in one or more countries, and the cost would be substantial. Without such a trial, however, it is most likely that increasing resources will be put into poorly designed screening programmes of unknown value.

The simplest and strongest designs use individual randomisation, but group randomisation designs may have practical advantages, though they require a greater sample size. Designs based on general population screening, and on screening only high risk groups, are both considered. They answer different questions. In countries with high incidence the value of general population screening is probably the more critical. Not enough is known to specify the type and frequency of screening precisely; both screening by doctors and self screening require evaluation, and annual screening should probably be tested. The age range at risk will depend on the local incidence, but is likely to be quite wide — for example, 45–69, and both sexes need inclusion. Thus a suggested design for a moderate to high incidence area would be a trial, randomised by individual or group, assessing at least two annual rounds of both screening by doctor and self screening (ideally by a factorial design), for adults aged 45–69, with mortality over several years' follow up as the critical outcome. In an area with good data systems such a study could compare screening offered to some 260.000 subjects with 10 times that number of controls passively followed up, with 90% power to detect a one third reduction in mortality. A general assessment of costs over five years gave estimates of $8.3 million for the screening programme and $2.4 million for the evaluation.

The much weaker designs, area based cohort studies using individual data or a simpler ecological comparison, and case-control studies, are also considered. If well designed with attention to their methodological limitations they may be valuable but are unlikely to be as definitive as a randomised trial.