Should atypical squamous cells of undetermined significance (ASCUS) be subcategorized? Accuracy analysis of Papanicolaou smears using receiver operating characteristic curves and implications for the ASCUS/squamous intraepithelial lesion ratio.

We correlated all Papanicolaou test diagnoses over a 6-month period with biopsy results and determined accuracy using receiver operating characteristic curves and biopsy as the "gold standard." Accuracies were calculated using all atypical squamous cells of undetermined significance (ASCUS) cases or by eliminating subsets thereof. Retaining the ASCUS category resulted in significantly greater accuracy for the diagnosis of squamous intraepithelial lesion (SIL) on biopsy compared with eliminating it by diagnosing all such cases as negative. Subcategorization significantly improved the accuracy of the test only when all cases were included. The highest accuracy without subcategorization was achieved when ASCUS, favor reactive, cases were diagnosed as negative, but this threshold was significantly less sensitive than including all ASCUS cases. Increasing or decreasing the estimated ASCUS/SIL ratio from 2.4 without subcategorization significantly reduced accuracy. Similar results were obtained when high-grade SIL on biopsy was used as the gold standard. Use of the ASCUS category significantly improves the accuracy of the Papanicolaou test. Eliminating any subset of ASCUS reduces the ASCUS/SIL ratio but also significantly diminishes the sensitivity of the Papanicolaou test.