Intralaboratory reproducibility of cervical cytology diagnoses in the external quality assurance scheme of the Emilia-Romagna region of Italy

In an external quality assurance (EQA) scheme, 110 selected cervical smears were independently reported by seven cytology laboratories using the 1988 Bethesda System. In one of these, a random sample of 60 study smears was independently classified by five cytologists before undergoing joint examination according to EQA protocol. Internal agreement (i.e., among these cytologists) was compared with external agreement (between these cytologists and the other six laboratories) and with interlaboratory agreement (among these laboratories). The κ value for internal agreement on sample adequacy (0.57) was considerably greater than those for external (0.16) and interlaboratory agreement (0.10). For interpretation of epithelial abnormalities, internal κ was excellent (0.80) but not substantially greater compared with external κ (0.66) and interlaboratory κ (0.70). Calculation of class-specific κ values revealed (a) that superiority of internal agreement on sample adequacy was restricted to the notations of 'less than optimal' and of 'satisfactory,' and (b) that internal increase in agreement on cytologic abnormalities was substantial for the diagnosis of 'atypical cells of undetermined significance.'