Prospective correlation of cervicovaginal cytologic and histologic specimens.

ABSTRACT

An effective, prospective, computer-guided method of correlation is reported. The mechanism for identification of cases, comparison of diagnoses, and reconciliation of discrepancies are explained. The results are similar to prior, retrospective, correlation studies. The benefits specific to this unique prospective approach include optimal capture of cases for correlation, minimization of errors before diagnoses are released to clinicians and patients, and internal standardization of diagnostic criteria. Three thousand four hundred and four consecutive paired cervicovaginal cytologies and biopsies were accessioned at the Pathology Department of Duke University Medical Center over a 43-month period. Of these, 481 paired cases (14%) had discordant diagnoses, defined as differing more than one degree of dysplasia or as dysplasia or carcinoma identified by only one modality. Additional evaluation reconciled the diagnostic differences in 35 cases. Eighteen initial diagnostic differences arose from cytologic screening errors, 16 from interpretive errors by staff pathologists, and one from superficial initial histologic sections. The remaining 446 discordances were attributed to sampling differences. The cytologic smear contained the diagnostic lesion in 40% of the cases and the biopsy the remainder, emphasizing the utility of pairing these sampling techniques in patients at risk for dysplasia.