Small intestinal stromal tumors : A clinicopathologic study of 20 cases with immunohistochemical assessment of cell differentiation and the prognostic role of proliferation antigens

Small intestinal stromal tumors (SISTs), similar to their gastric counterpart, are complex because of their divergent cellular differentiation and because of the difficulty in accurately predicting their clinical outcome. We studied a series of 22 SISTs from 20 patients to characterize lineage and investigate prognostic morphologic parameters and possible histologic and immunohistochemical differences from gastric stromal tumors (GSTs) and to determine the potential prognostic value of proliferation markers. Cases were categorized into the three following groups based on mitotic count (MC) per 50 high-power fields and tumor size: (1) benign, n = 6 (< 5 MC, < 5 cm); (2) borderline, n = 6 (< 5 MC, > or = 5 cm); and (3) malignant, n = 10 (> or = 5 MC, any size). For the formalin-fixed, paraffin-embedded tissue sections, an immunohistochemical panel was used to characterize differentiation toward myogenic cells (pan-muscle specific actin [HHF-35], alpha-smooth muscle actin, and desmin), Schwann cells (S-100 protein), enteric glial (glial fibrillary acidic protein), and nerve cells (neurofilament). Cellular proliferative activity was assessed immunohistochemically using monoclonal antibodies to proliferating cell nuclear antigen (PCNA) and Ki-67 antigen (MIB-1) and a tumor proliferation index (TPI) was obtained as the percentage of positive-staining tumor nuclei. Clinical follow-up revealed that none of the benign tumors progressed (mean follow-up, 96 months). Half of the patients with borderline tumors were dead of disease (mean, 50.7 months), while 8 of 9 patients with a malignant tumor died of disease (mean, 24.6 months). By Cox Proportional Hazard Regression analysis, mitotic count, tumor size, and cellularity significantly predicted survival. PCNA, MIB-1, tumor necrosis, and atypia were not significant predictors of survival. All tumors stained with vimentin; 17 (77%) and 13 (59%) of the tumors showed immunoreactivity with muscle-specific actin markers (HHF-35) and alpha-smooth muscle actin, respectively. Only 1 tumor stained with desmin, and none stained with S-100 protein, neurofilament, or glial fibrillary acidic protein. Immunophenotypic characteristics did not differ among the 3 groups. The TPI for PCNA and MIB-1 significantly differed between benign and malignant tumors and between borderline and malignant tumors, but it failed to separate the benign and borderline groups. Compared with 52 cases of GST previously reported by us using the same criteri