Statistical measurements of accuracy in cervical cytology.

Statistical measures that can be used to monitor the level of accuracy of cervical cytologic screening are examined: sensitivity, the false-negative rate, the interval cancer rate among women with negative reports and the predictive value of a positive test. Sensitivity and the false-negative rate are subject to biases and problems of determination that may make them less useful as measures of accuracy. The interval cancer rate and the positive predictive value may be better methods of assessing the frequency of serious abnormalities following negative cytologic reports and the accuracy of cytologic reports of serious abnormalities, respectively. It is important to recognize that no single measure of accuracy can adequately define a laboratory's performance. The use of statistical analyses can be invaluable in placing in context the criticisms of cervical cytology as an appropriate screening test for the prevention of cervical cancer; statistical analysis should be used as one component of quality control.