Reproducibility in double scanning of cervical smears with the PAPNET system

To determine the reproducibility of the PAPNET scanning station (Neuromedical Systems, Inc., Suffern, New York, U.S.A.) in selecting cells from a cervical smear.

We compared the images of 196 smears that were scanned on two occasions by the PAPNET scanning station on two monitors simultaneously and compared the cellular contents and technical records provided by the scanning station. The sample consisted of 62 positive smears (mild dysplasia and more) and 134 negative smears.

Although differences were found in the technical information provided by the scanning station (kappa = .65, 95% confidence interval [CI] = .51-0.79) and in the reported percentages of air bubbles (kappa = .60, 95% CI = .51-.68), the detection of abnormal cases was not affected. Furthermore, the agreement on microscopic review was excellent (kappa = .92, 95% CI = .88-.96). In nine cases that did not differ in tech code or percentage of air bubbles, however, differences were found in the cellular content of the tiles that would have led to different advice for additional microscopic review. This would have had important clinical consequences in two cases because a serious abnormality would have been missed.

The consistency of the PAPNET scanning system is somewhat on the low side in providing technical information, although this did not affect the clinical outcome. In nine cases, for example, we found differences in the demonstration of cells in the tiles on the screen; two would have had clinical consequences. It is important to set goals for the performance of the machine and to incorporate them in the procedures to be used as standard practice. This is especially true when the scanning stations are going to be operated on site in the cytology laboratory.