The sensitivity of rapid (partial) review of cervical smears

Rapid review involves a daily rapid (e.g. 30 s) review of all smears not normally double-screened. It has been suggested that the method may increase the sensitivity of cervical cytology by identifying abnormalities not reported on initial screening, true false negatives (TFN). Rapid screening is reported to have high sensitivity for cervical neoplasia when used as a preview tool. To be effective, however, in a review mode it must be able to detect TFN. Several studies have found that many TFN result from factors such as low numbers of abnormal cells or subtle expression of diagnostic criteria. Studies on the sensitivity of rapid screening for detecting TFN would therefore provide a more reliable estimate of its value as a review tool. The sensitivity of rapid re-screening was evaluated using a test set of 200 cases. Each of 15 screeners rapidly reviewed (30 s partial screen) the set over a 2-week period. The set consisted of 129 normal, 28 low-grade squamous lesions (CIN I), 37 high-grade lesions (CIN II, III and adenocarcinoma in situ (AIS)) and six invasive carcinomas. The abnormals included 20 TFN cases. The median sensitivity for abnormalities was 62%. Rapid review was more sensitive for CIN II and CIN III (67%) and invasive carcinoma (66.7%) than for CIN I (53%). Great variation was apparent in the sensitivity for individual screeners, with a range of 41-86% for all abnormalities. The sensitivity for TFN cases varied even more (10-75%, median 35%) and for most screeners was significantly (P < 0.05) lower than for cases which were detected on initial screen (53-90%, median 70.6%). Following this trial rapid review was used routinely for a period of 3 months. In this time 11 413 cases were rapidly reviewed. This led to the full review of 415 slides (3.5%) and the identification of 16 cases of undetected CIN (12 CIN I, three CIN II, one CIN III). Based on current estimates of our laboratory false-negative rate this represents between a quarter and half of the TFN cases of CIN that probably occurred in this period. In conclusion, rapid screening is likely to be significantly less sensitive when used in a review rather than a preview mode. In routine practice the method requires a daily commitment of screener time, but does provide a higher yield of TFN smears than does random review, and allows amendment of these results prior to reporting.