Reproducibility of major diagnoses in a binational study of lung cancer in uranium miners and atomic bomb survivors.

A binational panel of four Japanese and four American pathologists examined 208 pulmonary neoplasms, according to the World Health Organization (WHO) recommendations, second edition, for the histologic typing of lung tumors. The study design included independent evaluations by pathologists working alone, followed by group reviews. The individual evaluations, and their implications for reproducibility of the WHO recommendations, are reported. Consensus (agreement by six or more pathologists) with respect to major (ie, first digit) diagnosis was obtained for 76.4% of the cases. Consensus was obtained for 72.5% of the cases with any major diagnosis of small cell cancer; the comparable figures for adenocarcinoma and squamous cell carcinoma were 56% and 48%, respectively. American pathologists were twice as likely as Japanese pathologists to diagnose large-cell cancer, the only significant national difference. Consensus was far less frequent with the minor (ie, second digit) diagnosis categories. This study shows that lung cancers continue to be difficult to classify reproducibly.