Context: The ThinPrep Imaging System (Imager) for cervical cytology is used in many US laboratories, but the ability of the system to identify classic changes of low-grade squamous intraepithelial lesion (LSIL) has not been independently reported.
Objective: To evaluate the accuracy of this system in detecting classic LSIL cells.
Design: A total of 114 imaged ThinPrep LSIL cases from April to June 2005 were reviewed to determine whether the most diagnostically relevant cells were present in the 22 fields selected by Imager. Those LSIL specimens from January to June 2005 that were initially screened by the Imager and classified as "negative" in the 22 fields were also reviewed. The average number of classic koilocytes per slide was compared among cases that had koilocytes within 22 fields with cases upgraded during full review.
Results: Twelve (10.5%) of 114 LSIL cases from April to June 2005 did not show diagnostic LSIL cells within the initial 22 fields. Full manual rescreening was performed on 1025 cases from January to June (20.4% of negative cases). Nine cases that were initially negative in the 22 fields were revised to LSIL (2.4% of all 381 LSIL cases reported from January to June 2005). An average of 10 to 11 koilocytes were present in these 2 groups, which was significantly lower than the average of 75 koilocytes in cases in which the 22 Imager-selected fields showed LSIL.
Conclusions: Although the ThinPrep Imager finds abnormal cells in most LSIL cases, the system may have limitations in detecting koilocytes in the 22 Imager-selected fields. When 10% quality control rescreening is performed as required by federal regulations, full manual rescreening will provide the most accurate results.