LAUSR

in the REMARC study had centralized prospective peripheral blood lymphocyte analyses by flow cytometry (FCM) either at diagnosis (n = 220), at randomization. (n = 188) or both (n = 73). Absolute counts of CD4+, CD8+ T cells, B cells, NK cells, and monocytes, were derived directly from the FCM data. According to CD16 and CD56 expression, NK were classified as cytotoxic (16+/56dim), cytokinic (16‐/56 bright) and anti‐tumoral natural cytotoxic (16‐/56 dim) NK cells. For each NK subpopulation, activating receptor expressions cited above were analyzed. Patients were grouped according to their NK subpopulation profile by hierarchical clustering. Associations between NK cell counts and respectively IPI, cell of origin (COO) by both Hans and nanotring, and PFS were assessed. Results: Similarly to entire REMARC cohort, patients analyzed had median age of 68, a low aaIPI for 43% and high for 56%. With median values of 388, 223, and 160 per μl, CD4+ CD8+ and NK cells were within the normal ranges in only half of the patients at diagnosis. Univariate analysis showed that NK cell count <100/μl at diagnosis was not associated with IPI or COO but was significantly associated with a shorter PFS [HR = 2.5 (1.6, 3.8) P < 0.0001]. In a Cox model for PFS including IPI, COO (Hans), randomization arm (LEN or observation), NK cell count <100/μl prognostic value was retained (HR = 2.9 (1.6, 5.1) P < 0.0005). Clustering of NK subpopulations showed that variation of total NK cell count was mostly associated with variation of the dominant CD16+/CD56dim population expressing activating receptors and identified a patient cluster with depleted count of all subpopulations and poor prognosis. Conclusion: Low NK cell count at diagnosis is associated with poor clinical outcomes in elderly patients with DLBCL treated with R‐CHOP independently of IPI, GCB/nonGCB subtypes and LEN maintenance. Clustering of NK subpopulations allows the identification of a small group of patients with a very unfavorable prognosis.