DETERMINATION OF THE OPTIMUM CUTOFF VALUE OF A DIAGNOSTIC-TEST

Many diagnostic tests require the determination of cut-off values that discriminate uninfected from infected individuals. The cut-off values of a serodiagnostic test were calculated to minimize the total cost of misdiagnoses. Three different assumptions were made about the density distributions of the uninfected and infected populations to which the test was to be applied: (1) the population distributions were identical to the sample distributions; (2) both populations were normally distributed; (3) the population distributions may not be normal but may follow other density distributions that are used commonly. The cut-off values calculated using each assumption varied considerably. This indicates that accurate estimations of the density distributions of the populations are required to determine the optimum cut-off values. There is often insufficient information to estimate population density distribution adequately. Decisions regarding cut-off values of diagnostic tests should be reviewed as more data become available.