Evaluation of diagnostic tests using relative operating characteristic (ROC) curves and the differential positive rate. An example using the total serum bile acid concentration and the alanine aminotransferase activity in the diagnosis of canine hepatobiliary diseases.

The value of a diagnostic test depends on most cases on its ability to discriminate between patients with and without a certain disease. One way of evaluating a diagnostic test is to use the relative operating characteristic curve (ROC curve) and the differential positive rate (DPR). The ROC curve displays the relationship between the true positive ratio and the false positive ratio for a range of cutoff values and it can be used to compare various diagnostic tests under equivalent conditions (equal true positive ratios or false positive ratios) and over the entire range of cutoff values. The DPR is the difference between the true positive ratio and the false positive ratio at various cutoff values and it can be used to obtain the cutoff value associated with the highest sensitivity and specificity. The purpose of this study was to describe the evaluation and comparison of diagnostic tests using ROC curves and DPR. Eventually, the positive and negative predictive values were used to assess the differences between the sensitivity and specificity obtained when the upper limit of the reference interval, or the optimal cutoff value indicated by the DPR, was used as cutoff value. To illustrate the methods, the 2 h post-prandial total serum bile acid concentration (PSBA) and the alanine aminotransferase activity (ALAT) in the diagnosis of primary or secondary hepatobiliary diseases in dogs were used. The ROC curves showed, as expected from previous studies, that PSBA was superior to ALAT in diagnosing dogs with hepatobiliary diseases. Using DPR, the optimal cutoff value for PSBA was suggested to be 15.48 mumol/l. Compared to the traditionally used cutoff value of 22.24 mumol/l, no decisive difference in the positive predictive values were observed. However, the cutoff value of 15.48 mumol/l appeared to produce higher negative predictive values compared to a cutoff value of 22.24 mumol/l. Seemingly, ROC curves and DPR are simple methods useful to the evaluation of diagnostic tests and due to the simplicity, there seems to be a great potential for these methods in the evaluation of diagnostic tests in veterinary medicine.