Consistency of cytology diagnosis for cervical cancer between two laboratories

The principal approach to the prevention of cancer of the cervix uteri has been through screening programmes, using the cervical smear (Pap test) to detect precursor lesions. The sensitivity and specificity of Pap smears depend on the skill of the observer in recognizing and classifying a variety of cellular abnormalities. We have studied the reproducibility of cytological diagnosis, according the Bethesda classification, made by cytologists in Khon Kaen, north-east Thailand, and in Helsinki, Finland, on smears taken from rural women undergoing screening during 1994-2001. A total of 313 slides were reviewed. The prevalence of abnormalities was relatively high, since the series included smears judged abnormal in Khon Kaen or from women who developed cancer during follow-up, as well as a group whose smears were negative. In general, the reviewing cytologist in Finland evaluated more slides as abnormal than in the initial report. The level of agreement between the two observers was evaluated by calculating the coefficient of concordance (Kappa). The kappa score depended upon the degree of detail in the diagnosis; it was 0.43 for the presence or not of an epithelial abnormality (the General Categorization of the Bethesda system), and rather higher (0.5) for separating low grade from high grade (HSIL or worse) abnormalities or glandular lesions. Agreement was only fair (0.37) when the more detailed Bethesda categories (seven) were used. The reproducibility of cervical cytology evaluations is critical to the success of screening programmes, and in this programme in a moderate-high risk population of women in rural Thailand, we found that agreement between skilled observers, at the level of tests requiring diagnostic follow-up or not, was only moderate. The women in this study are being traced to evaluate the true sensitivity of screening in terms of the lesions found on histology, during a prolonged follow up of 4 or more years.