In this prospective study, for the first time, the authors compared the accuracy of reading urine specimens using the ThinPrep Imager System (TIS) with the accuracy of conventional screening for the detection of abnormal urine cells.
ThinPrep slides were made from 1455 urine specimens and were read with conventional screening and with TIS. Findings were categorized as "unsatisfactory or failure to read the slide," "benign," or "abnormal." The Cohen κ coefficient was calculated to determine inter-rater agreement between both methods. Urine samples that were followed by biopsies were used to compare the sensitivity, specificity, positive predictive value, and negative predictive value of both methods. From 22 urine specimens, the screening times were measured and compared.
There was substantial agreement between both methods (κ score, 0.77). Of 175 urine specimens that were followed by bladder biopsies, for conventional screening, the sensitivity was 51.3%, specificity was 68.4%, the positive predictive value was 77.2%, and the negative predictive value was 40.2%; for TIS screening, the respective values were 54.6%, 68.4%, 78.3%, and 41.9%. The average time was 5.2 minutes for conventional screening and 3.9 minutes for TIS.
With a κ score of 0.77, the current study demonstrated a good correlation between reading urine specimens with conventional screening and with TIS. Using TIS resulted in slightly increased sensitivity, positive predictive value, and negative predictive value compared with conventional screening and had the same specificity. These results indicate that reading urine specimens using TIS is equally reliable as conventional cytology.