Tumor-infiltrating immune cells signature predicts recurrence free survival after complete resection of localized primary gastrointestinal stromal tumors

BackgroundGrowing evidence has proposed prognostic value of immune infiltration in gastrointestinal stromal tumors (GISTs). Therefore, we aimed to develop a novel immune-based prognostic classifier (IPC) to help better stratify and predict prognosis of GISTs. MethodsThe gene expression profiles of 22 immune features of GISTs were detected from GEO dataset. The IPC was constructed using the LASSO Cox regression model and validated in a cohort including 54 patients with complete resection of localized primary GISTs via immunohistochemistry process. The performance assessment of the IPC was estimated, then compared with conventional risk prognostic criteria. ResultsThe IPC was established based on 4 features: CD8, CD8/CD3, CD68, CD163/CD68 and validated to be an independent predictor of RFS for GISTs (HR 5.2, 95%CI 1.99-13.65). Significant differences were found between low- and high-IPC group in 5-year RFS (92.6% vs 48.1%, p < 0.001). Using the IPC, the high-risk group of the Modified NIH classification was split into two groups in 5-year RFS (low-IPC vs high-IPC, 85.7% vs 30.0%, p < 0.001). The IPC showed a higher net benefit than both "treat all" or "treat none" methods for the threshold probability within a range of 0-0.62 and exhibited a performance (AUC 0.842) superior to modified NIH classification (AUC 0.763). ConclusionThe IPC was effective to predict RFS after complete resection of localized primary GISTs, adding prognostic value to the routine clinical prognostic criteria. Prospective studies are needed to further validate the analytical accuracy and practicability of the IPC in estimating prognosis of GISTs.