Scientific and ethical consequences of disease prediction.

HLA-typing, gene analysis, anti-islet cell antibody testing and metabolic studies can identify people at high risk for developing Type 1 (insulin-dependent) diabetes mellitus prior to the onset of clinical disease. The positive predictive value of these tests is high in first degree relatives of patients with Type 1 diabetes, but six times less so in the general population, where disease incidence is much lower but where 90% of new cases occur. Multiple testing improves sensitivity but decreases specificity. Intervention strategies are being designed with the aim of delaying or preventing progression to clinical disease. The more invasive the intervention, the greater is the specificity required. The practical and ethical implications of identifying high risk of diabetes in healthy individuals are complex and require further research, but some lessons can be learned from the experience of other disease prediction programmes.