Adenocarcinoma in situ of the uterine cervix. Screening and diagnostic errors in Papanicolaou smears

Little attention has been given to the reasons for failure to detect adenocarcinoma in situ (AIS) of the uterine cervix in Papanicolaou (Pap) smears. In the current study, the authors examined a series of screening or diagnostic errors in cases in which the final histologic diagnosis was either AIS or AIS combined with a high-grade squamous intraepithelial lesion (AIS + HSIL). Smears obtained in the 3 years before histologically proven AIS or AIS + HSIL was diagnosed and within a specified 6-year period (1993-1998) were reviewed and reclassified. All were conventional Pap smears. The smears studied were those with a review diagnosis of possible or definite high-grade epithelial abnormality that initially were reported by a cytotechnologist to be negative (screening error) or that were reported by a pathologist to be negative, unsatisfactory, or indicative of a low-grade epithelial abnormality (diagnostic error). A semiquantitative, blinded assessment of the frequency of cytologic criteria for the diagnosis of AIS was made for smears with erroneous diagnoses compared with a series of smears that yielded true-positive findings. Sampling errors, which were defined as cases in which smears did not have sufficient evidence for a diagnosis of possible or definite AIS or HSIL on review, accounted for 35.1% and 36% of all smears from patients with a biopsy diagnosis of AIS and patients with a biopsy diagnosis of AIS + HSIL, respectively. With regard to AIS, there were 3 screening errors and 5 diagnostic errors, accounting for 10.4% of 77 smears. Minimal, poorly preserved material was evident in four smears, including three smears with only one sheet of abnormal glandular cells. In four other smears, there was a moderate amount of adequately preserved glandular material, mainly in large sheets, with varying degrees of crowding and hyperchromasia. With regard to AIS + HSIL, there were 6 screening errors and 6 diagnostic errors, accounting for 13.5% of 89 smears. In those smears, there generally was a moderate amount of abnormal material in the form of crowded groups of suboptimally preserved, hyperchromatic squamous cells. Only two of those smears yielded findings of possible abnormal glandular cells. Only 3 of 20 errors occurred in smears that were examined during the last 3 years of the study. In the semiquantitative assessment, smears with erroneous findings were shown to contain far less abnormal material than true-positive smears and to exhibit a corresponding