Is there a better way to assess performance in breast cytology?

Is there a better way to assess performance in breast cytology? Results generated from a series of breast fine needle aspirates (BrFNA) are semi-qualitative and therefore sensitivity, specificity and accuracy are not necessarily the best measures of effectiveness. This paper considers the likelihood ratio of cancer (LR+), which can provide the probabilities for cancer being present or absent for cancer across the range of diagnostic categories used in the National Health Service Breast Screening Programme (NHSBSP), i.e. C1 to C5. Using LR+, receiver operating characteristic (ROC) curves can be generated and used to compare performance between pathologists, laboratories or years. These are illustrated using test results from 1997 to 1999 for the Derbyshire Royal Infirmary. LR + and ROC are meaningful measures. They could replace the 12 calculations currently used for the evaluation of effectiveness of BrFNA in the National Health Service Breast Screening Programme (NHSBSP) and for validation of pathologists' performance.