A prognostic immune risk score for diffuse large B-cell lymphoma

We constructed a prognostic score for persons with diffuse large B-cell lymphoma (DLBCL) based on infiltrating immune cells. Data of 956 consecutive subjects were retrieved from the Gene Expression Omnibus database and assigned to training (GSE10846, n = 305) or validation (GSE87371 n = 206 and GSE117556 n = 445 combined) cohorts. Proportions of non-lymphoma cells in the sample were inferred using the ESTIMATE algorithm. An immune risk score was constructed comprised of eight types of non-lymphoma immune cells calculated using the CIBERSORT algorithm. Five-year survival of subjects with an immune risk score ≤ 0·45 in the training cohort was better than that of subjects with a score > 0·45 (hazard ratio [HR] = 3·99; 95% confidence interval [CI] = 2·74, 5·82; P < 0·001). HR in the validation cohort was HR = 2·17 (1·47, 3·21; P < 0·001). Enrichment analyses indicated correlations with genes controlling immune-related biological processes and pathways. A nomogram comprised of the immune risk score and most covariates including age, lactate dehydrogenase concentration (LDH), lymphoma-type (germinal centre B cell [GCB] versus non-GCB), Eastern Cooperative Oncology Group performance status (ECOG-PS) and rituximab therapy had a C-statistic of 0·76 compared with C-statistics of 0·69 and 0·69 for the International Prognostic Index (IPI) and Revised International Prognostic Index (R-IPI). These data indicate the immune risk score is an accurate, independent survival predictor in persons with DLBCL.