Results
Out of 122 patients diagnosed with DLBCL during the pre-specified period, 96 (74.8%) had a SEP before any chemotherapy, and among them, 12 (12.5%; 8 males; median age: 68 [55-89] years) had L-TGL. Demographical and biological data for both the L-TGL (n=12; TGL≤5.5 g/L) and H-TGL (n=84; TGL>5.5 g/L) subgroups are summarized in Table 1.
Ann Arbor staging and International prognosis index were comparable between L-TGL and H-TGL groups. However no Richter syndrome was reported in the L-TGL group, but 2 were described in the H-TGL group.

IgG level was lower in L-TGL regardless of albumin and inflammatory status

Immunoglobulin isotype levels were available only in two-thirds (8/12) of L-TGL and in approximately half (46/84; 55%) of H-TGL patients, respectively: the median levels of IgG were significantly lower (p<0.001) in the L-TGL subgroup than in the H-TGL subgroup and in parallel with TGL levels, whereas the IgA and IgM levels were not significantly different (p=0.31 and 0.32, respectively). No patient had detectable viral load for Epstein Barr virus or Cytomegalovirus or active bacterial infection at diagnosis.
The serum total protein level was lower (p<0.01) and AG ratio was higher (p<0.01) in L-TGL than in H-TGL patients. Even though higher serum ferritin levels were found in L-TGL patients than in H-TGL patients (p=0.02), which could indicate higher biological inflammatory status or specific organ involvement, the levels of CRP were comparable in both subgroups.
Lymphocyte level was significantly lower in L-TGL patients (p=0.02), but was comparable between L-TGL and H-TGL deceased patients (Table 1).

Death rate and infection related death are higher in L-TGL subgroup

The mortality rate was higher in L-TGL (mortality rate: 10/12 [83%] versus 22/84 [26.2%]; p=0.03) and median follow-up duration was shorter (follow-up duration: 15.2 months versus 55.53 months; p<0.001) than in H-TGL subgroup (Table 1). Similarly, the rate of death caused by an infection was significantly higher in L-TGL than in H-TGL patients (10/10 [100%] versus 6/22 [27.3%]; p<0.001), as shown in Table 1. In H-TGL patients, deaths were caused by DLBCL progression for most patients (54.5%; 12/22 deceased patients), whereas the remaining patients (18.2%; 4/22 deceased patients) died from other causes that were independent of the background disease or related treatments.
Regarding infection-related deaths, no opportunistic infection was identified. All 10 deaths in L-TGL subgroup and 3 out of 6 infection-related deaths in H-TGL subgroups were related to pleuro-pneumopathy, either associated or not with ear, nose and throat infections and/or Streptococcus pneumoniae . The remaining 3 H-TGL infection-related deaths were caused by septic shock-complicated pyelonephritis, staphylococcus bacteremia and cutaneous cellulitis of the diabetic foot.
As seen on Kaplan Meier curves shown in Figure 1 , the survival of L-TGL patients was significantly lower (p<0.001) than all the other groups considered in the whole DLBCL patients cohorts, including H-TGL patients (n=84), all patients who had a SEP (i.e. L-TGL and H-TGL patients considered together; n=96) and patients who didn’t get a SEP at diagnosis (n= 26).

No concordance between TGL and albumin levels in L-TGL and H-TGL subgroups

In addition to the comparisons of demographics, clinical and biological characteristics in both L-TGL and H-TGL subgroups, shown in Table 1; both subgroups were similar for all clinical and usual biological parameters including albumin level, except for the lower median TGL (p<0.01) and IgG levels (p<0.01) in L-TGL patients than in H-TGL patients, as expected. Moreover the AG ratio was higher in L-TGL than in H-TGL patients (p<0.01) indicating that TGL decrease was more pronounced than that of albumin, suggesting that albumin serum level was unrelated to TGL in our DLBCL patients subgroups.

Chemotherapy regimen was comparable between L-TGL and H-TGL subgroups

The distribution of chemotherapy regimens is shown in Table 2 between deceased and living patients in both the L-TGL and H-TGL subgroups. First-line chemotherapy was the R-CHOP regimen (21-day interval; rituximab, cyclophosphamide, hydroxyadriamycin, vincristine, and oral prednisone) for 85 (88.5%) patients in total, whereas 3 others, all in the H-TGL subgroup, received ACBVP (two-week interval; adriamycin, cyclophosphamide, bleomycin, vindesine, and oral prednisone) followed by sequential consolidation therapy consisting of two cycles of methotrexate because of their younger age; 7 others (2 and 5 in the L-TGL and H-TGL subgroups, respectively) received the RMPV regimen (rituximab, methotrexate, procarbazine, and vincristine) because of cerebral involvement, and finally, the remaining patient was an elderly patient in the H-TGL subgroup who refused any chemotherapy.
A second-line chemotherapy regimen was chosen for 37 of the 96 patients, including 9 and 28 patients in the L-TGL and H-TGL subgroups respectively, whereas third- and fourth-line chemotherapy regimens were proposed for 10 patients and 2 patients, respectively. As shown in Table 2, the characteristics of patients who had died and those who remained alive at last follow-up in each subgroup appeared comparable, except for lymphoma stage: stage I lymphomas were significantly overrepresented in H-TGL patients who remained alive (p=0.02; Table 2). Of note, stage IV lymphoma and central nervous system involvement were not significantly associated with death among H-TGL patients, although the low number of events precluded any firm conclusion (Table 2).

Hypogammaglobulinemia at diagnosis is associated with higher risk of death

Table 3 represents hazard ratio for the different variables studied using a cox proportional hazards model. Ann Arbor stage IV was positively associated with the occurrence of death (HR 3.47 (1.61-7.47), p < 0.01). More interestingly, L-TGL was also found to be independently associated with a higher risk of death (HR 12.8 [4.93-34.31], p <0.001) while H-TGL was associated with a lower risk of death (HR 0.21 [0.09-0.46], p <0.01).