[Agreement between colposcopic diagnosis with 2011 international terminology of colposcopy and cervical pathology in cervical lesions].

Objective: To evaluate the agreement between colposcopic diagnosis with 2011 colposcopic terminology of the International Federation for Cervical Pathology and Colposcopy (IFCPC) and cervical pathology in cervical lesions.
Methods: A retrospective cohort study was performed, which included 376 patients who underwent colposcopy with 2011 international terminology of colposcopy at Obstetrics and Gynecology Hospital of Fudan University from September 2014 to November 2014. With conization or cervical biopsy pathology as the gold standard, the agreement between colposcopic diagnosis and pathologic diagnosis was calculated and correlations between variables were analyzed.
Results: With 2011 international terminology of colposcopy, agreement of colposcopic diagnosis and cervical pathology was 60.9% (229/376) perfectly matched, and the strength of agreement with weighted Kappa statistic was 0.401 (P < 0.01), and agreement within one grade was 97.6% (367/376), which were improved compared with traditional methods. Colposcopic diagnosis were 19.9% (75/376) overestimated and 19.1% (72/376) underestimated. There were no significant difference between agreements in various grade lesions (χ(2) = 1.996, P = 0.573). Positive predictive value of high grade colposcopy or more was 84.4%, the negative predictive value of low grade colposcopy or less was 88.8%, whereas false positives were 3.5% and false negatives were 39.3%. A linear trend among three types of transformation zone and patient ages was found (χ(2) = 45.910, P < 0.01), whereas lesion sizes were not linearly correlated with lesion degrees (χ(2) = 0.690, P = 0.406). In grade 1, grade 2 and nonspecific findings, thin acetowhite epithelium, dense acetowhite epithelium and Lugol's non-staining were most frequent, the Youden indexes of each were 0.170, 0.373 and 0.145, 0.069 respectively. Positive predictive value of some other findings (such as fine mosaic) and two new signs (inner border sign and ridge sign) were 100.0%. There were no significant difference between agreements in examiners with different levels of experience (χ(2) = 1.197, P = 0.550).
Conclusions: Compared to traditional methods (such as Reid index), 2011 international terminology of colposcopy could improve the agreement between colposcopic diagnosis and pathologic diagnosis, without significant differences by the severity of lesion and the level of examiners' experience. Common findings were classified reasonably, and some signs were highly reliable, which is important for guiding biopsy. However, the reproducibility of transformation zone types and the implication of lesion size need to be further discussed.