A quality control system involving peer review of abnormal cervical smears

An internal quality control system which is used in the centralized cytology laboratory of a population-based cervical cancer screening programme in Florence is described. It includes a peer review procedure. Abnormal cervical smears are circulated among all the cytologists and a consensus on the final diagnosis is reached. This daily procedure is designed to evaluate the performance of each cytologist and of the laboratory as a whole but can also be considered a valuable training opportunity. During an 18-month period 1197 smears were reviewed by 15 readers using a reporting form with six main categories of reporting (from 'negative' to 'invasive carcinoma'), plus an 'unsatisfactory' category. Overall the concordance between the 15 cytologists, assessed using the kappa statistic (range 0.46-0.71; median 0.60), was good. The level of agreement increased when a weighted kappa statistic (range 0.55-0.78; median 0.68) was used. Kappa values were also calculated for specific categories and suggested an increasing concordance with increasing severity of the lesions, the categories of 'severe dysplasia' and 'invasive carcinoma' showing the highest agreement. The poor results for the 'moderate dysplasia' confirmed the need for combining this group with the 'severe dysplasia', as proposed in the Bethesda system.