Histologic estimation of coronary artery stenoses: reproducibility and the effect of training.

Histologic estimation of coronary artery stenoses (CAS) provides the 'gold-standard' for clinicopathologic correlations and medicolegal investigations, yet little evidence supports histology as a reproducible diagnostic measure, and none addresses the effect of training on its use. To study these questions, 20 randomly selected Movat-stained coronary artery cross-sections were reviewed 3 times, at 3-month intervals, by six clinical pathologists (CPs), six pathology residents (Res), seven anatomic pathologists (APs), and two cardiovascular pathologists (CVPs). Before the third iteration, a guide to CAS assessment with illustrations was provided. Inter- and intraobserver reproducibility were determined using interclass correlation coefficients (ICC) (0.40-0.75 = fair-good; > or = 0.76 = excellent agreement beyond chance). Surprisingly, all study groups had excellent interobserver reproducibility. Before training, at Time 1, the scores were CPs, 0.77; Res, 0.89; APs, 0.93; and CVPs, 0.93. After training, at Time 3, the results were CPs, 0.81; Res, 0.91; APs, 0.86; and CVPs, 0.88. Intraobserver reproducibility for CPs overall was good (ICC, 0.74), and excellent for Res, APs, and CVPs (0.89, 0.94, and 0.97, respectively). In conclusion, statistical analysis failed to demonstrate any significant effect of training or experience on observer reproducibility.