The pathological features of 155 adult patients with soft-tissue sarcomas were studied retrospectively, in an attempt to set up a grading system for these tumors. As the first step, seven histological criteria (tumor differentiation, cellularity, importance of nuclear atypia, presence of malignant giant cells, mitosis count, pattern of tumor necrosis and presence of vascular emboli) were evaluated in a monofactorial analysis. Five of these (tumor differentiation, cellularity, mitosis count, tumor necrosis, and vascular emboli) were correlated with the advent of metastases and with survival. A multivariate analysis, using a Cox model, selected a minimal set of three factors (tumor differentiation, mitosis count, and tumor necrosis) the combination of which was necessary and sufficient to retain all the prognostic information. A grading system was elaborated, which turned out to be correlated with the advent of metastasis and with patients' survival. A second multivariate analysis introducing clinical prognostic features showed that the histological grade was the most important prognostic factor for soft-tissue sarcomas. Thus, this grading system appears to be highly interesting because of its prognostic value and the facility of its elaboration. However, its reproducibility should be tested.

Tumor grade has been proposed as an essential factor in the staging of patients with soft tissue sarcomas. In a previous study, a histopathologic grading system using the evaluation of tumor differentiation, mitosis count, and tumor necrosis was described. The current study was conducted to test its reproducibility. The pathologic sections of 25 soft tissue sarcomas were submitted to a study group composed of 15 pathologists who had not been involved in the development of the grading system. The results were compared with those of a panel group. The crude proportion in agreement observed between the study group and the panel group was 81% for the evaluation of tumor necrosis, 74% for tumor differentiation, and 73% for the mitosis count. The crude proportion in agreement for the tumor grade was 75%, which was significantly better than the crude agreement rate of 61% for the diagnosis of histologic type (P = 0.001). A kappa statistical analysis, to check the possibility of chance-related concordance, showed a proportion in agreement of 68%. A two-way variance analysis showed that the homogeneity of the evaluation of tumor grade is impaired by tumor-related and observer-related factors. However, an improvement may be obtained by better training of pathologists. We conclude that the tumor grading system developed inside the French Federation of Cancer Centers, although perfectible, already provides reliable prognostic information and its use in prospective clinical studies may provide more information about its clinical usefulness.

No abstract