Interobserver and intraobserver variability using the Fuhrman grading system for renal cell carcinoma.

CONTEXT: Histologic grading of renal cell carcinoma has been shown to be second to staging in prognostic significance. A 4-tier grading scheme proposed by Fuhrman et al and based on nuclear features is the system used most frequently in North America. There are, however, very few studies in the literature assessing the interobserver variability for this system, and to our knowledge, none addressing intraobserver variability. OBJECTIVE: To assess the interobserver and intraobserver agreement among 4 pathologists using the Fuhrman nuclear grading scheme for renal cell carcinoma. DESIGN: Representative hematoxylin-eosin-stained slides of 99 consecutive primary renal cell carcinoma cases diagnosed between 1994 and 1999 at St Joseph's Hospital, Hamilton, Ontario, were independently graded by 4 pathologists on 2 occasions with a minimum period of 3 months separating the 2 readings. RESULTS: Intraobserver kappa values ranged from 0.29 to 0.62 (mean = 0.45), and interobserver kappa values ranged from 0.19 to 0.38 and from 0.09 to 0.44 for the first and second rounds, respectively (combined mean kappa value = 0.29). When combining Fuhrman grades 1 and 2 as low-grade tumors and grades 3 and 4 as high-grade tumors, the intraobserver kappa values ranged from 0.4 to 0.64 (mean = 0.53) and interobserver kappa values ranged from 0.28 to 0.59 and from 0.26 to 0.58 for the first and second rounds, respectively (combined mean kappa value = 0.45). The admixture of 2 grades in the same tumor was observed in 53% of cases. CONCLUSIONS: We found only moderate intraobserver and interobserver agreement using the 4-grade Fuhrman scheme. After collapsing the diagnostic grades to 2, the intraobserver agreement changed from moderate to substantial. The collapsing of the 4-category Fuhrman grades into 2 categories is useful in improving intraobserver agreement.

Telepathology for routine light microscopic and frozen section diagnosis.

Telepathology (TP) uses telecommunication linkages to electronically capture, store, retrieve, and transmit images to distant sites. We assessed the feasibility of a dynamic real-time TP system for light microscopic (LM) diagnosis of anatomic pathology specimens, including frozen sections. Six pathologists, in 2 separate periods, read a set of 160 retrospectively retrieved slides (80 of which were frozen sections) by TP and LM. Reading times were recorded. Diagnoses were compared with the reference diagnosis (established by a group of 5 independent pathologists) and graded on a scale of 0 to 2 (2, correct; 1, incorrect but no clinical impact; 0, incorrect with clinical impact). Overall, LM was more accurate than TP compared with the reference diagnosis (score, 1.68 vs 1.54). There was no difference in accuracy between frozen section and paraffin-embedded tissue. Intraobserver agreement ranged from 82.5% to 88.2%. The average reading time was 6.0 minutes for TP and 1.4 minutes for LM. During the study, reading time decreased for TP but not for LM. These results show that despite marginally lower accuracy and longer reading times, TP isfeasible for routine light microscopic diagnosis, including frozen sections.