[Analysis of the intralaboratory diagnostic variability in the Imola cervical screening program]

The value of the assessment of intralaboratory variability as a method for quality control of cervical screening is an ill-defined one. This model has been advocated and utilized by some Authors and only indirectly addressed by others. In Italy, only occasional studies have been published. The present analysis is part of a series of studies for the evaluation and quality control of the population cervical screening programme in the area of Imola. The agreement between the 5 cytologists was tested over a series of 150 smears selected from the routine material. The data were analyzed with the k statistic and the degree and direction of discrepancies were assessed by the percent distribution of paired diagnoses. The k values for agreement between the 10 possible pairs of cytologists varied between 0.25 and 0.49 (average value, 0.37). The indexes for agreement as based on three classes ('Negative'/ASCUS + LGSIL + HGSIL/'Unsatisfactory') varied from 0.35 to 0.56 (average value, 0.46). The cytologist-specific k values (as based on five classes) ranged from 0.29 to 0.43. For the observer with the lowest specific coefficient (designated as C4), the excess disagreement was restricted to the ASCUS class and the 'Unsatisfactory' notation. In particular, C4 diagnosed more often 'Negative' smears (74% vs. 44-55%) and less frequently ASCUS (11% vs. 22-31%) and 'Unsatisfactory' smears (3% vs. 6-11%). ASCUS was the least reproducible diagnosis (k = 0.22). The evaluation of the degree and direction of the diagnostic discrepancies between the five cytologists showed that the frequency of concordant diagnoses of ASCUS was as low as 39.7%. However, the discrepancies were almost always of one degree and were directed towards the 'Negative' class. Conversely, ASCUS was associated with considerable proportions of the 'Negative' diagnoses (17.3%) and those of LGSIL (31.8%). In the occasions when one observer interpreted a smear as 'Unsatisfactory' the paired diagnoses were almost evenly distributed among 'Unsatisfactory' (49.2%) and 'Negative' (43.0%). In absolute terms, the k coefficients in this study were compatible with moderate agreement. However, they were unfavourably influenced by the modalities for smear selection. The study specimens were selected by all cytologists from the routine material over a short time period and did not include smears interpretable as showing carcinoma. A comprehensive evaluation of the diagnostic performance of C4 showed that the disagreement with other cytologists was restricted to the diagnosis of ASCUS and the 'Unsatisfactory' notation and was accounted for by a lower frequency of such reports coupled with a greater frequency of the 'Negative' diagnoses. A conservative approach to lesions of minor significance is the most likely explanation for that pattern of disagreement. Although characterized by the poorest k value, the ASCUS diagnosis proved to be generally matched with reports of minor cytology changes. This was consistent with the formal definition for ASCUS in the original Bethesda System. The observation that the ASCUS diagnosis is applied to those changes is a prerequisite for a more conservative utilization of such report. In conclusion, the study revealed important correlates of the k coefficients and provided an encouraging picture of the cytologic substrates of the diagnosis of ASCUS.