The comparative diagnostic accuracy of conventional and liquid-based cytology in a colposcopic setting

Objective This study was conducted to compare the performance of liquid-based cytology (LBC) and conventional cytology (CS) in the high prevalence setting of colposcopy clinic. Design A split sample of matched ThinPrep (TP) and conventional smear from 563 patients were evaluated blindly. The performance of both techniques was compared with the gold standard of biopsy results or normal colposcopy examination in 441 cases. Setting Colposcopy clinic of an inner city hospital for women and children. Sample Five hundred and sixty-three women referred to colposcopy clinic over 14-month period. Methods Cervical smears were taken from 563 women referred for colposcopy. Using the split-sample technique, the material was spread on a conventional (CS) slide and the remaining material rinsed in a PreservCyt solution. A T2000 processor was used to prepare LBC preparations. All women underwent colposcopy/biopsy according to local protocol. Four hundred and forty-one women met the diagnostic standard criteria of the study, which was either a normal colposcopy or histopathology result. Sensitivity, specificity and positive and negative predictive values were calculated for both methods of cytology preparations. Main outcome measures Matched TP and conventional smears, detection of abnormality, matched biopsies, sensitivity, specificity, and positive and negative predictive values. Results Inadequate rates for CS and LBC (TP) were 4.3% and 0.68%, respectively. In 73% of cases, the CS and the LBC preparations showed exact agreement, whereas 77% agreement was seen when comparison was made for amalgamated low grade and high grade abnormalities. Low grade cytological abnormalities accounted for 44% of LBC slides versus 37% in CS slides. High grade cytological abnormalities accounted for 22% of LBC versus 17% seen in CS cases (P < 0.001). LBC showed increased sensitivity in the detection of CIN2 or worse than CS (92% and 83%, respectively) and CS showed greater specificity than LBC (82% and 76%, respectively). Conclusions In high prevalence setting, LBC performed at least as well as CS. The inadequate rate was significantly lower with LBC. The numbers are too small, however, to make confident comments about increased sensitivity and negative predictive value with LBC. Larger studies are required to verify these findings.