Objective: The aim of this study was to evaluate the inter-laboratory reproducibility for atypical glandular cells (AGC) (The Bethesda System (TBS) 2001) of the laboratories involved in the screening programmes in Italy.
Methods: A set of 35 selected slides were circulated among 167 laboratories involved in local population-based cervical screening programmes. Each laboratory provided one single diagnosis per smear. The smears were read blind to the original diagnosis and to the diagnoses provided by other laboratories. A 'majority' diagnosis was defined for each case and assumed as the reference standard. The diagnosis provided from each laboratory was compared with the majority diagnosis.
Results: According to the majority report the 35 slides in the set were classified as negative in nine cases, AGC in eight, adenocarcinoma in eight, and squamous lesion or squamous + glandular lesion in 10. The crude agreement between all pairs of laboratories was 49.43%. K-values were 0.46, 0.21, 0.34, 0.36 and 0.32 for negative, AGC/AIS (adenocarcinoma in situ of endocervix), AdenoCa, Sq/Sq + Gl and all reporting categories respectively. Concordance according to overall K was moderate to substantial in 77% of the participating laboratories.
Conclusions: The present study shows that the AGC category is not easily reproducible. The data confirmed the importance, in a screening scenario, of AGC/AIS diagnoses, but also presented difficulties in differentiating between the two diagnoses. In addition to the results obtained from the circulation of the slides, laboratories which had annually a low number of cervical smears were able to gain experience focused on particular morphological pictures.