Cervical smears prepared by an automated device versus the conventional method. A comparative analysis.

An automated, fluid-based method for the preparation of cervical Papanicolaou smears/slides was compared to the conventional Papanicolaou smear (CPS) method used for the screening of neoplasia. We determined diagnostic agreement and sources of error for diagnostic disagreement. For 665 patients, one cervical sample was collected to make one CPS. The collection devices, a wooden Ayre spatula and endocervical brush, were rinsed into a vial with fluid medium to be processed in the automated device. All slides were distributed among five cytotechnologists in a blind fashion. Exact diagnostic agreement was 94.6%. The results were not statistically significant (P > or = .70, McNemar's test) but were clinically important, as evidenced by the detection of low grade lesions (LGL), during initial screening, on three slides prepared by the automated device but not on their matched-pair CPSs (0.5% of all specimens). After reevaluation, the three matched CPSs demonstrated LGL. Sources of diagnostic error on the CPSs were: air-drying artifact, obscuring blood/inflammation, crowding/overlapping of cells and/or absence of diagnostic cells. The only source of error in the automated-method smears was absence of diagnostic cells.