Reproducibility of asbestos body counts in digestions of autopsy and surgical lung tissue

Abstract

Asbestos body (AB) counting by a single observer is the most commonly used objective technique to demonstrate asbestos deposition in the lung. In this study, the accuracy and reliability of this procedure is analyzed by evaluating the degree of agreement between two experienced readers. Lung tissue specimens from 66 individuals, most of whom had not been exposed to asbestos, were studied: 35 were obtained in postmortem studies (upper, middle, and lower lung) and 31 were from patients who underwent surgery for lung cancer. Overall, 167 samples were analyzed. Lung tissue sections weighing 0.5 g were obtained prospectively and processed, and the inorganic residue was analyzed by light microscopy at 400× magnification by two experienced readers. Results were expressed as AB/g of dry lung tissue. Interobserver variability was analyzed using the Spearman correlation coefficient and agreement was evaluated by the Bland-Altman method and the kappa index. The interobserver correlation was 0.8975: 0.8029 for autopsy samples and 0.9592 for biopsy samples. Bland-Altman plots showed that most values were grouped around the 95% limits of agreement. The kappa index was 0.87 for all samples, and 0.79, 0.65, and 0.54 for upper, middle, and lower lung specimens, respectively. Asbestos body counting by a single reader is a reliable method, especially at low concentrations of asbestos bodies in lung tissue. Double reading may be indicated in borderline cases with asbestos body levels close to levels of 1 000 AB/g.