Reproducibility of the histological diagnosis of cervical dysplasia among pathologists from 4 continents.

The reliable histological diagnosis of cervical squamous intraepithelial lesions (SILs), especially low-grade SIL, is known to be problematic. Poor diagnostic reproducibility can complicate studies addressing its appropriate management. As part of an international study comparing expectant management of histologically proven low-grade SIL with immediate loop electrocautery excisional procedure, this study was carried out to assess interobserver agreement on the histological diagnosis of SILs among a group of 22 pathologists from 5 countries and the intraobserver reliability among a subset of 7 Canadian pathologists. Fifty-six histological slides from colposcopically obtained cervical biopsies were circulated to each of the 22 pathologists. To assess intraobserver reliability, 7 Canadian pathologists assessed 40 of the slides once and 16 of the slides twice. Kappa values were used to measure interobserver agreement with an overall kappa value of 0.61 (95% confidence interval, 0.60-0.62) corresponding to moderate reliability. The weighted kappa values for interobserver agreement ranged from 0.46 to 0.88 (median, 0.79). The intraobserver reliability of 7 Canadian pathologists ranged from substantial to excellent based upon the weighted kappa values ranging from 0.62 to 0.94 (median, 0.72). This degree of reliability is comparable to that found in similar studies. In an individual case, there can be considerable disparity in diagnosis that can result in disparate management strategies. This adds a layer of complexity to any trial that attempts to assess optimal treatment strategies or the natural history of this disease.