Inclusion criteria and definitions
We retrospectively extracted from the monocentric lymphoma database of Caen University Hospital all patients diagnosed with DLBCL between January 2015 and December 2016, counted those with SEP performed at diagnosis before receiving any specific treatments. Included patients were categorized into two groups according to whether they exhibited L-TGL or H-TGL on SEP at DLBCL diagnosis. Although CVID diagnosis requires a threshold < 5 g/L of IgG isotype in addition to other biological, clinical and anamnestic criteria(20–22), we chose serum total gamma-globulin level because it is an easily available biological test. We defined L-TGL as all TGL values ≤ 5.5 g/L on SEP as it is < 50% of lower range of our laboratory norm similarly to what was used in others studies in hematological malignancies(19) and because IgG represent 75% of TGL in the absence of monoclonal gammapathy, it would be unlikely to have more than 5g/L of IgG.

Exclusion criteria

As our hypothesis is based on L-TGL being a possible marker of PID or SID related to the DLBCL condition, we excluded all patients exhibiting an associated monoclonal peak on SEP; unrelated increased loss of serum protein such as exudative enteropathy, extended burns or nephrotic syndrome; previous treatment with cytotoxic drugs or immunosuppressant; and previous PID or SID diagnosed before DLBCL.

Studied parameters

For all patients, demographic information at diagnosis, disease status, chemotherapy regimen, Ann Arbor staging classification, central nervous system involvement (CNS) involvement, and international prognosis index (IPI) and outcome data were collected until the last follow-up date. For biological data, C-reactive protein (CRP), albumin, lymphocytes count, lactate deshydrogenase (LDH) and ferritin levels were collected at time of diagnosis in the absence of an infectious process, and when available, IgG, IgM, and IgA levels were implemented. We considered the albuminemia/gammaglobulinemia ratio (AG ratio) to discriminate possible causes for L-TGL : 1) For an unspecific decrease in TGL related to general state impairment, a decrease in albumin should be associated to decrease in TGL, with an unaltered AG ratio; 2) For a specific decrease in TGL directly related to our hypotheses, i.e., L-TGL due to either an undiagnosed PID or a SID developed with DLBCL, the albumin level should be close to normal value, resulting in an increase in the AG ratio.

Statistical analyses

Data were compiled as the median [range] or a number (%). Continuous variables were analyzed using the Mann-Whitney test, and Fisher’s test was used to compare categorical variables. A p value < 0.05 was considered statistically significant.
Predictors of death in DLBCL were evaluated using COX proportional hazard models. All the variables showing a univariate p-value less than 0.20 were entered into a multivariable logistic regression model. Results were expressed as hazard ratios (HRs) with 95% confidence intervals (95% CI). Two-sided P-values lower than 0.05 were considered significant.
Overall survival in DLBCL and subgroups L-TGL and H-TGL were analyzed using the Kaplan-Meier method, and these variables were compared using the log-rank test.
All statistical analyses were performed using GraphPad Prism software (7.0) and JMP 9.0.1 (SAS Institute Inc., Cary, NC, USA).