Retrospective evaluation of gynecologic cytodiagnosis. II. Interlaboratory reproducibility as shown in rescreening large consecutive samples of reported cases.

Abstract

Two laboratories exchanged and rescreened a large sample of cases with cervicovaginal smears they had consecutively accessioned to examine the reproducibility of gynecologic cytodiagnosis under optimum conditions. At least a "working agreement" (diagnoses within +/- 1 category on a ten-category scale) was achieved in diagnoses of normal, benign reaction and squamous abnormality (from minimal dysplasia though invasive cancer) in 18,859 cases (96.8%), of endometrial abnormality in 21 cases (42%) and of "unsatisfactory" in 99 cases (20.7%). Larger differences occurred in greater than or equal to 30% of cases except in the categories of "normal" and "benign reaction," reaching a maximum of 82% for moderate dysplasia. Reexamining 382 cases decreased disagreement by category to the 20% to 65% range only in the five categories of dysplasia plus carcinoma in situ. Agreement was not predicated on the presence of endocervical cells or squamous metaplasia; the basis for "unsatisfactory" calls was not uniform. Comparison of the laboratories' diagnoses with referee diagnoses or, on 178 cases, with tissue diagnoses also demonstrated differences in diagnostic criteria.