Accuracy of liquid‐based cytology

Abstract

In the New Technologies for Cervical Cancer Screening (NTCC) randomized controlled trial, no significant increase in the sensitivity of liquid-based cytology (LBC) was observed compared with conventional cytology. Both were interpreted by cytologists who had limited previous LBC experience. The objective of the current study was to assess whether different results could be expected with experienced LBC interpreters. A stratified, random sample of 818 LBC slides from the NTCC study was obtained. These slides were reviewed blindly and independently by 3 international experts who did not participate in the NTCC. The sensitivity and specificity of external experts were estimated for cervical intraepithelial neoplasia grade 2 or greater (CIN2+) and for CIN3+ histology, and the differences were compared with the sensitivity and specificity of the original cytologic interpretation using cutoffs of atypical squamous cells of undetermined significance (ASCUS) and low-grade squamous intraepithelial lesion (LSIL). With the endpoint of CIN2+ histology, the difference in sensitivity between external experts and the original interpretation was -5.3 (95% confidence interval [CI], -16.0 to 5.4) with ASCUS as the cutoff and 3.8 (95% CI, -8.2 to 15.8) with LSIL as the cutoff. External experts had slightly lower specificity using ASCUS as the cutoff (-3.4; 95% CI, -3.9 to -2.9) and LSIL as the cutoff (-0.7; 95% CI, -1.0 to -0.4). The accuracy of the external experts' interpretation was similar to that of the original interpretation. Therefore, the current results indicated that LBC is not expected to increase sensitivity even if it is used by interpreters who have extensive experience with this technique.