Despite the use of modern immunochemotherapy (R-CHOP) regimens, almost 50% of patients with diffuse large-B-cell lymphoma (DLBCL) will relapse. Current prognostic models, most notably the International Prognostic Index, are comprised of patient and tumor characteristics and are unable to identify patients with less than a 50% chance of long-term survival. However, recent observations demonstrate that factors related to host adaptive immunity and the tumor microenvironment are powerful prognostic variables in non-Hodgkin lymphoma
We retrospectively examined the absolute neutrophil count (ANC), monocyte count (AMC) and lymphocyte count (ALC), obtained from an automated complete blood count with differential, as prognostic variables in a cohort of 255 consecutive DLBCL patients that were uniformly treated with R-CHOP between 2000 and 2007 at a single institution. The primary study objective was to assess if ANC, AMC, and ALC at diagnosis were predictors of overall survival (OS) in DLBCL.
At diagnosis, the median ANC was 4720/uL (range 1190–17690), the median AMC was 610/uL (range 30–4040), and the median ALC was 1220/uL (range 140–5410). The median follow-up for these patients was 48 months. In the univariate analysis, each of these variables predicted OS as continuous variables. As dichotomized variables, an elevated ANC (≥5500/μL; hazard ratio 1.75, 95% confidence interval 1.14–2.60, p=0.01) and AMC (≥610/μL; hazard ratio 3.36, 95% confidence interval 2.10–5.59, p<0.0001) were each associated with inferior OS. In contrast, the presence of lymphopenia, defined as an ALC ≤1000/uL, was associated with inferior OS (hazard ratio 2.21, 95% confidence interval 1.43–3.39, p=0.0004). When components of the IPI were included on multivariate analysis only the AMC and ALC were independently significant prognostic factors for OS, with hazard ratios of 3.37 (95% confidence interval 2.05–5.74, p<0.0001) and 2.19 (95% confidence interval 1.38–3.44, p=0.0009), respectively. The dichotomized AMC and ALC generated the AMC/ALC prognostic index (PI) and stratified patients into 3 risk groups: very good (AMC <610/uL and ALC >1000/uL), good (AMC ≥610/uL or ALC ≤1000/uL), and poor-risk (AMC ≥610/uL and ALC ≤1000/uL) populations. For both the very good (n=79) and good-risk (n=134) groups median OS has not been reached with estimated 5-year overall survival of 88% and 69%, respectively. Median OS for poor-risk (n=42) patients was 1.7 years (95% confidence interval 1.1–2.7 years) with an estimated 5-year overall survival of 28% (p<0.0001). By comparison, the R-IPI was unable to identify a group of patients with a median survival less than 8 years. The estimated 5-year OS was 93%, 71% and 53% for very good, good and poor-risk patients, respectively. We sought to determine whether the AMC/ALC PI may provide additional prognostic information when combined with the R-IPI. To test this possibility, the 171 very good/good risk and 84 poor risk patients identified by the R-IPI were subsequently risk stratified using the AMC/ALC PI. Among R-IPI very good/good risk patients a subset of poor risk patients (n=21) with a median OS of 2.2 years (95% confidence interval 1.1–6.6 years) and 35% 5-year OS could be identified with the AMC/ALC PI. In contrast, 5-year OS ranged from 75%-88% among very good and good risk patients. Similarly, stratification of R-IPI poor risk patients by the AMC/ALC PI identified subsets of very good (n=19) and good risk (n=44) patients with median OS that had not been reached and 86% and 55% 5-year OS, respectively. High risk (n=21) patients had a median OS of 1.4 years (95% confidence interval 0.9–2.2 years) and an estimated 5-year OS of less than 25%.
Measurement of AMC and ALC at diagnosis is widely applicable, cost effective, predicts OS, and identifies high-risk patients with DLBCL.
No relevant conflicts of interest to declare.
Asterisk with author names denotes non-ASH members.