Context: Expert opinion is often used as a gold standard for gynecologic cytology in the evaluation of new technologies, in the legal setting, and in the validation of cases for use in educational programs and proficiency testing. However, the reliability of expert opinion alone in selecting slides of a specific cytodiagnosis that can be reproducibly and reliably identified by subsequent reviewers has not been determined.
Objective: To assess the ability of expert opinion to select slides that are validated in subsequent reviews.
Design: In the College of American Pathologists Interlaboratory Comparison Program in Cervicovaginal Cytology, each case in every cytodiagnostic category is accepted for circulation only after review by 3 expert cytopathologists. The percentage of these cases that could not be reliably and reproducibly identified by program participants for each cytodiagnostic category ("failed field validation") was determined during the duration of the program from 1989 to 2004.
Results: More than 10,000 conventional smears and ThinPrep cases were selected by the expert panel for circulation. Of these selected slides, 19% of conventional smears and 15% of ThinPrep specimens failed field validation. Compared with the overall slide performance, significantly higher percentages (P < .001) of conventional smears with reference diagnoses of unsatisfactory (51.7%), repair (58%), or low-grade intraepithelial lesion (31.8%) and of ThinPrep specimens with reference diagnoses of unsatisfactory (54.5%) and repair (76.9%) failed field validation. In contrast, significantly lower percentages of conventional smears with reference diagnoses of squamous cell carcinoma (4.5%), high-grade squamous intraepithelial lesion (9%), Trichomonas vaginalis infection (11.7%), or herpes (9.9%) and of ThinPrep specimens with reference diagnoses of adenocarcinoma (5.1%), herpes (2.1%), and fungal organism consistent with Candida (8.4%) failed field validation (P < .001 for all).
Conclusions: Between 15% and 19% of gynecologic cytologic cases that have been selected by expert cytopathologists as good examples of cytodiagnostic abnormalities fail field validation. The proportion of cases failing field validation varies with cytodiagnostic category, but it occurs in all cytodiagnostic entities.