Immune Cell and Stromal Signature Associated With Progression-Free Survival of Patients With Resected Pancreatic Ductal Adenocarcinoma

BACKGROUND & AIMS: Changes to the microenvironment of pancreatic ductal adenocarcinomas (PDACs) have been associated with poor outcomes of patients. We studied the associations between composition of the pancreatic stroma (fibrogenic, inert, dormant, or fibrolytic stroma) and infiltration by inflammatory cells and times of progression-free survival (PFS) of patients with PDACs after resection. METHODS: We obtained 1824 tissue microarray specimens from 385 patients included in the European Study Group for Pancreatic Cancer trial 1 and 3 and performed immunohistochemistry to detect alpha smooth muscle actin, type 1 collagen, CD3, CD4, CD8, CD68, CD206, and neutrophils. Tumors that expressed high and low levels of these markers were compared with patient outcomes using Kaplan-Meier curves and multivariable recursive partitioning for discrete-time survival tree analysis. Prognostic index was delineated by a multivariable Cox proportional hazards model of immune cell and stromal markers and PFS. Findings were validated using 279 tissue microarray specimens from 93 patients in a separate cohort. RESULTS: Levels of CD3, CD4, CD8, CD68, and CD206 were independently associated with tumor recurrence. Recursive partitioning for discrete-time survival tree analysis identified a high level of CD3 as the strongest independent predictor for longer PFS. Tumors with levels of CD3 and high levels of CD206 associated with a median PFS time of 16.6 months and a median prognostic index of -0.32 (95% confidence interval [CI] -0.35 to -0.31), whereas tumors with low level of CD3 cell and low level of CD8 and high level of CD68 associated with a median PFS time of 7.9 months and a prognostic index of 0.32 (95% CI 0.050-0.32); we called these patterns histologic signatures. Stroma composition, when unassociated with inflammatory cell markers, did not associate significantly with PFS. In the validation cohort, the histologic signature resulted in an error matrix accuracy of predicted response of 0.75 (95% CI 0.64-0.83; accuracy P<.001). CONCLUSIONS: In an analysis of PDAC tissue microarray specimens, we identified and validated a histologic signature, based on leukocyte and stromal factors, that associates with PFS times of patients with resected PDACs. Immune cells might affect the composition of the pancreatic stroma to affect progression of PDAC. These findings provide new insights into the immune response to PDAC.