Clear heads and bayesian tales: predictive value and the coin toss?

Abstract

In Reply .—Dr Mauro's remarks would be very pertinent if we had suggested in our reply that platelet antibodies should be detected as part of a screening program of ITP in the general population . This is not so, however. Such a screening test would be unjustifiable, not only because of its immense costs, but mainly because it would offer no therapeutic or preventive advantage to the patients detected in this manner, as compared with those whose thrombocytopenia has been discovered, as is usually the case, on clinical grounds. Thus, the controversy regarding the utility of platelet antibody testing is limited to the population of patients with thrombocytopenia and cannot be extrapolated to the general population. In this setting, our statement on "coin flipping" remains correct. Platelet antibody testing has a positive predictive value for ITP of 50%. In other words, the probability for a patient with thrombocytopenia of having ITP if the