Agreement among and within groups of pathologists in the classification of rhabdomyosarcoma and related childhood sarcomas. Report of an international study of four pathology classifications

Abstract

Background. An International Pathology study was conducted to measure the agreement demonstrated among and within groups of pathologists involved in the categorization of childhood rhabdomyosarcoma according to four pathology classifications. Data concerning agreement and survival experience according to pathonew subtypes were used as a basis for selection of a proposed new pathologic classification. Methods. A random sample of 800 eligible patients was chosen from the Intergroup Rhabdomyosarcoma Study II (IRS‐II) and was reviewed by pathologists representing eight institutions. A 20% sample of the 800 patients was then rereviewed by the pathologists to determine the level of agreement with their original classification. In each instance the patients were classified according to four pathology systems: the conventional system, the International Society for Pediatric Oncology system (SIOP), the National Cancer Institute (NCI) system, and the cytohistologic system. Results. Among the groups of pathologists, the highest measure of agreement was a Kappa value of K = 0,451 for the conventional system, followed by K = 0,406 for the SIOP system, K = 0,384 for the NCI system, and K = 0,328 for the cytohistologic system. For reproducibility within the groups of pathologists, the highest measure of agreement was K = 0.605 for the conventional system, followed by K = 0.579 for the NCI system, K = 0.573 for the SIOP system, and K = 0.508 for the cytohistologic system. Conclusions. There was a general similarity between the agreement reached within the modified conventional, SIOP, and NCI systems, with the modified conventional system having the highest Kappa values, and thus the highest measure of agreement, both among and within the groups of pathologists. Also, the subtypes of the conventional system demonstrated a highly significant relationship to survival time. Hence, based on criteria of reproducibility and prognostic significance, the proposed classification will essentially be a modification of the conventional system with elements of the SIOP and NCI systems. Copyright © 1994 American Cancer Society