A multicenter retrospective study of 239 patients aged over 70 years with diffuse large B-cell lymphoma in China

Diffuse large B-cell lymphoma (DLBCL) is the most common aggressive lymphoma subtype worldwide and frequently occurs in the elderly population. However, there was a lack of large series of data regarding the clinical proles of over 70 years in DLBCL patients in China. We aim to summarize this population’s clinical features and survival outcomes in the real world. The multicenter retrospective study was conducted in Western China from Jan 2012 to July 2020 to investigate clinical features and survival outcomes of 239 DLBCL patients over 70 years. All patients underwent pre-treatment evaluations, treatment, and follow-up at local hospitals. The primary endpoints of this study were progression-free survival (PFS) rate and overall survival (OS) rate at two years. The second endpoints included median PFS and OS, 5-year PFS and OS rates, and adverse events during treatment. dose should be administered for elderly DLBCL patients under the balance of benet and risk.


Introduction
Diffuse large B-cell lymphoma (DLBCL) is the most common aggressive lymphoma subtype worldwide, accounting for 30 ~ 40% of non-Hodgkin lymphoma [1] . From the SEER (Surveillance, Epidemiology, and End Results Program) database, the incidence of DLBCL increases with age; the median diagnosis age was 66 years [2] . Compared to DLBCL patients younger than 60-70 years, older patients had a worse prognosis [3][4][5] . Moreover, most clinical trials excluded these patients, leading the optimal treatment options and the survival outcome in the real-world remain to be elucidated, especially for over 70 years. R-CHOP is the standard regimen for the rst-line DLBCL treatment [6,7] , reaching a 60% cure rate for the entire cohort [8] . Whereas most elderly patients had poor physical state and were frequently accompanied by other diseases, which limited the usage of standard dose of the regimen and further had a worse impact on the survival. Thus, dose reduce regimens or no-anthracycline-containing treatment were explored in clinical trials, such as R-miniCHOP, R-GemOx, R2 [9][10][11] , which resulted in a balance of e cacy and tolerance in selected elderly DLBCL patients. In the United States, the 5-year survival rate was 78.4% in the DLBCL population of those younger than 55 years, while 54.3% was for people older than 65 years [2] . In China, it reported that 5-year OS rate 51.9% for DLBCL patients aged over 60 years between 2006 to 2012 in Beijing of 349 cases, and 42.8% for aged over 70 years of 145 patients [12] . To date, there has been a limited number of studies regarding the clinical pro les of aged over 70 years in DLBCL patients. Furthermore, as the anti-CD20 antibody was widely applied in clinical practice in Western China, and better administration of treatment related adverse effects in recent years than ever, indicating that the survival outcome of Chinese elderly DLBCL needs to be investigated.
In this study, we collected 239 DLBCL patients aged over 70 years from ve university hospitals in Western China to summarize the clinical features, analyze survival outcomes in the real world, and examine the factors that in uenced the prognosis.

Methods
We retrospectively reviewed the record of all newly diagnosed DLBCL patients who were aged over 70 years at ve hospitals located in southwest China from Jan 2012 to July 2020, including West China Hospital, Sichuan University; Sichuan Cancer Hospital & Institute, Sichuan Provincial People's Hospital;

A liated Hospital of North Sichuan Medical College and A liated Hospitals of Southwest Medical
University. In all cases, the pathology diagnosis was con rmed by expert hematopathologists at these ve hospitals per the World Heathy Organization classi cation of hematopoietic and lymphoid tumors [13] . Critical exclusion criteria were age less than 70, central nervous system lymphoma involvement at diagnosis.
All patients underwent pre-treatment evaluations, treatment, and follow-up at the local hospital mentioned above. Evaluations included patient demographics, blood test, serum LDH, bone marrow detection. The staging was determined per the Ann Arbor staging system, and the treatment e cacy was evaluated by the enhanced computed tomography scans (CT) or positron emission tomography scans (PET-CT). Han's algorithm was applied to divide patients into germinal center B-cell (GCB) and non-GCB subtypes. The bulky disease was de ned as any mass diameter exceeding 7 cm. The investigators also reviewed the therapy options, regimens, cycles, and adverse events. Reduced dose chemotherapy in this study was explained as a decrease of 20% -50% of standard dose.
The primary endpoints of this study were PFS rate and OS rate at two years. The second endpoints included median PFS and OS, 5-year PFS and OS rate, adverse events during treatment. The PFS was calculated from the diagnosis to disease progression, recurrence, or any cause of death, and the OS was from the diagnosis to death for any reason or last follow-up. The study was done under regulatory requirements and approved by the Ethics Committee of West China Hospital of Sichuan University.
The statistical methods of comparing the difference between the groups were the χ2 test. We summarized PFS and OS using the Kaplan-Meier method, compared the difference between groups on PFS and OS with the log-rank test. Univariate and multivariate analyses related to survival were through the Kaplan-Meier method, the Cox proportional hazard models and the log-rank test. All data were analyzed with SPSS 25.0 (IBM Corp., Armonk, NY), and a two-sided p < 0.05 was considered a statistically signi cant difference.

Patient characteristics
We identi ed 316 patients newly diagnosed with DLBCL aged over 70 years from Jan 2012 to July 2020.
Of these screened patients, 22 cases had other malignant tumors concurrent, and 55 patients had only pathological diagnosis were excluded; 239 patients were nally included in the analyses.
Clinical characteristics are listed in Table1. The median age at diagnosis was 75.4 years (range, 70-91), and the ratio of male to female was 1.5:1. Staged III/IV patients were accounted for 59.8%, and 27.2% for extranodal involved more than one site. There were 200 patients had subtypes via Han's algorithm, 21.8% for GCB, 61.9% for non-GCB, and 16.3% of patients remained unknown. According to the IPI score, 25.5% of cases were evaluated as a high-risk group. It was 85.8% of patients accompanied by other diseases.
From the subgroups analysis data, between patients younger and older than 80 years, we observed no signi cant difference at baseline on clinical features.

Treatment and response
As presented in Table2, 239 cases were received initial treatment, including 64.4% for chemotherapy, two instances for radiotherapy, four for surgery, and 28.5% for combinatory therapy options. Overall, there were 219 patients administered with chemotherapy alone or with rituximab (Table3), 1029 treatment cycles were achieved, and the median cycles for a patient were 5 (range, 1-11). The most common chemotherapy regimen applied in this study was CHOP, accounting for 84.5%, other regimens comprising CVP, BR or oral vp16, and CTX. Due to the heavy nancial burden of the family before 2017, when the price decreased, 34.7% of patients were treated without rituximab in this cohort. It was similar between the age under 80 years and over 80 years on the usage of rituximab; however, in the age over 80 years group, dose reduction was occurred in more patients than those aged 70-79 years (45.1% vs. 20 The impact of rituximab on PFS and OS was also evaluated in our study. We found rituximab involved therapy could pronounced improve the OS and PFS for the entire cohort, the 2-y OS rate was 72.9% (95% CI: 65.9-80.1) in the rituximab group and 56.8% (95% CI: 44.5-69.2) in the no-rituximab group, and 5-y OS rate was 63.6% (95% CI: 55.2 -72.0) vs. 50.1% (37.6-62.6) (Figure2), p 0.008. The 2-y PFS rate was 59.5% (95% CI: 51.7-67.3) in rituximab-containing treatment group and 44.2% (95% CI: 31.9 -56.5), and 5-y PFS rate was 50.5% (95% CI: 42.1-58.9) vs 40.7% (28.4-53.0) (Figure2), p 0.031. In addition, strati ed by age, patients older than 80 could bene t from rituximab both on PFS and OS. In comparison, for younger than 80 years patients, we observed the improvement in PFS and OS from rituximab, but the p values were not signi cant (Supplementary gure2).
Reduced chemotherapy dose of regimens was frequently delivered in clinical practice for elderly lymphoma patients; in this study, we detected its role on survival outcomes. As shown in Figure3, there was no statistical difference between standard dose and reduced dose groups overall; however, in the younger than 80 years group, reduced dose was associated with remarked shorter OS, the 2-y OS was 74.4% (95% CI: 67.0-81.8) in full dose group in comparison with 67.1% (95% CI: 51.8-82.4) in decreased dose group, p 0.044. Meanwhile, we did not observe the remarked deterioration on PFS or OS in the older than 80 years group with dose reduction (Supplementary gure3).
Additionally, we observed that treatment cycles (chemotherapy ± rituximab) were associated with survival outcomes. As Figure4 presented, patients treated with six cycles or more obtained a superior OS and PFS than that with less than six cycles, the 2-y PFS were 66.2% vs. 44.5%, p 0.000, and 2-y OS were 77.6% vs. 59.5%, p 0.000. Further subgroup analysis showed that ≥ 6 cycles of R-CHOP/CVP treatment were achieved better PFS and OS than less than six cycles in the cohort, 68.7% and 49.0% for 2-y PFS, p 0.001, 81.0%, and 63.8% for 2-y OS, p 0.000 (Supplementary gure4).
IPI score was frequently applied to predict the prognosis of DLBCL patients. Here, we used IPI score to divide patients into IPI ≤ 3 or IPI > three groups; as presented in the Figure5, patients with IPI ≤ 3 had longer PFS and OS than those with IPI > 3, the 2-y PFS were 58.7% vs. 35.7%, p 0.005; and 2-y OS were 71.5% vs. 47.3%, p 0.001, individually. In younger than 80 years population, the 2-y PFS rate in the group of IPI ≤ 3 and of IPI > 3 was 61.2%, 31.8%, p 0.001, the 2-y OS rate in the group of IPI ≤ 3 and of IPI > 3 was 74.9%, 44.7%, p 0.000. In ≥ 80 years population, the 2-y PFS rate in group of IPI ≤ 3 and of IPI > 3 was 43.6%, 49.0%, p 0.741, the 2-y OS rate in group of IPI ≤ 3 and of IPI > 3 was 51.7%, 56.3%, p 0.888 (Supplementary gure5). However, further analysis in patients with the IPI = 1, 2, 3, 4, 5, we only found the OS difference on the IPI = three vs. IPI = 4 and IPI =4 vs. IPI =5 groups (Supplementary gure6 and supplementary table1), yet PFS bene t was negative.

Toxicity
The data of 187 patients were available to perform safety analysis; the most common adverse effects in this study were hematological toxicities, which accounted for 90.4%, and leukopenia (< lower limit of normal) was the most frequently happened of ≥ grade 3 events ( Table 5). The severity of treatmentemergent adverse events was similar between groups (Table 6). A higher incidence of fatigue, hypoalbumin were reported in older group, 63.9% vs. 93.1%; 53.8% vs. 89.7%. Whereas thrombocytopenia (any grade) was more common in the younger group, 34.8% vs. 13.8%.

Discussion
To our knowledge, this is the largest series that has analyzed clinical features and survival outcomes in aged ≥70 years DLBCL patients in China. The prognosis of this elderly DLBCL population with a 2-y OS rate of 65.5% was better than that reported in China with 42.8% [12] . Considering two signi cant reasons that may contribute to the disparity, one was a higher proportion of rituximab usage in our study, 65.3% vs. 53.8%, which signi cantly impacted the overall survival for this population, including the very elderly DLBCL patients (≥80 years). The other was that a larger series of 239 elderly DLBCL patients aged ≥ 70 years in our cohort, whereas 145 cases were included in that study, which may lead to the results bias.
Rituximab could improve the EFS and OS for 60-80 years DLBCL patients [14,15] . In this cohort, we found similar results: rituximab increased 2-y OS rate against only CHOP, 72.9% vs. 56.8%, p 0.008; and 2-y PFS 59.5% vs. 44.2%, p 0.031. Further analysis showed this result could continue to be made in the≥80 years subgroup. In the younger than 80 years group, we could see the improvement on the PFS and OS, but the p values were not signi cant, which may be related to the small sample size. Currently, rituximab is delivered with 375 mg/m 2 with an interval of every three weeks for 6-8 cycles in the standard regimen; additionally, some studies investigated whether increased dose of rituximab could improve treatment e cacy in elderly patients, showing that 500 mg/m 2 rituximab enhanced the PFS and OS of elderly male cases compared to the normal dose [16][17][18] . Meanwhile, a study from Netherland presented that rituximab intensi cation during the rst four cycles of R-CHOP14 had no improvement in the complete response rate, 3-y PFS rate, and 3-y OS rate compared with standard R-CHOP14 and increased the toxicities in the patients between ages 66-80 years [19] . There were no patients administered with this maximum rituximab dose in this study. In clinical practice, it is recommended that eight cycles of R plus six cycles of CHOP21 or six cycles R-miniCHOP in treating elderly DLBCL according to some critical studies results [15,[20][21][22][23] . The median cycles applied with a patient in our cohort were 5, of which 51.5% were administered with less than six cycles of treatment, while 48.5% with six cycles or more. In contrast to usage with ≥ 6 cycles of therapy in our data, patients with less than six cycles had inferior PFS and OS, suggesting the 6 or 8 cycles of treatment is emphasized in clinical practice.
The treatment strategy for elderly DLBCL patients has not been well established. In comparison with R-CHOP14 applied in the 60-80-years old DLBCL patients, R-CHOP21 could attain similar e cacy and decreased frequency and severity of hematological adverse effects and red-blood-cell transfusion, leading the R-CHOP21 as the frequently administered regimen [21,24] . In our study, 74.4% of patients were administered with full-dose for those aged 70-79 years and 67.1% for those older than 80 years. R-miniCHOP is recommended in ≥ 80y cases [7,9,25] . Peyrade F reported that ≥80y European patients treated with R-miniCHOP obtained a 2-y OS rate of 59%, 2-y PFS rate 47% [9] . From our data, we observed similar results in the Chinese population with a 2-y OS rate of 52.9% and a 2-y PFS rate of 45.3% treated with full dose or reduced dose therapy. Additionally, in our subgroup analysis, we found that reduced chemotherapy dose was not correlated with the 2-y OS rate compared with the standard dose in older than 80 years population, which was comparable to results of some studies ever reported [9,25] , indicating that reduced chemotherapy dose may balance the e cacy and side effects for this population. However, in patients aged 70-79 years, chemotherapy dose reduction (≥20%) could have a worse impact on the OS than those with the full dose, 2-y OS rate 67.1% vs. 74.4%, p 0.044. In line with a previous study result [26] , it reported that 70% of standard dose would affect the OS and EFS in patients aged over 70 years. Taken together, our data suggested that reduced chemotherapy dose exerted survival deterioration for all aged ≥ 70 years patients, except for over 80 years.
DLBCL is divided into GCB, ABC and unclassi ed subtypes with a gene expressing pro le (GEP) [27] . It also reported that Han's algorithm had prognosis value with the cell of origin (COO) determination [28] . In our study, we applied the Han's algorithm to divide patients into GCB and non-GCB subtypes. Additionally, Mareschal S et al. reported that ABC subgroup was more common in elderly patients [29] . In agreement, a similar result was observed in our study with 21.8% of GCB subtype and 61.9% of non-GCB subtype. Lenz reported that GCB subtype has superior PFS and OS rates than ABC subtype with R-CHOP [30] . Our cohort found that GCB subtype with 2-y PFS rate 63.5% and 2-y OS rate 72.2%, whereas non-GCB subtype with 2y PFS rate 55.8% and 2-y OS rate 66.5% (data not shown); however, the difference was not signi cant. Moreover, in the univariate analyses, patients with GCB or non-GCB had similar OS and PFS, associated with the non-GCB dominance in this cohort. IPI score is a prognostic model widely performed in DLBCL patients. Recently, a study reported that NCCN-IPI had the greatest absolute difference in OS estimates between the highest-and lowest-risk groups and best-discriminated OS compared to the IPI score and the aaIPI for the entire DLBCL cohort [31] . In our study, NCCN-IPI and aaIPI were not t for evaluating the prognosis as the included patients were all aged ≥ 70 years. According to the IPI score, there were signi cant discriminations in PFS and OS between the IPI ≤ 3 and IPI > 3 groups, which remained in the subgroups analysis of younger or older than 80 years patients, suggesting that IPI score could distinguish the prognosis in elderly patients with a highrisk group or not. Moreover, we observed inferior survival outcomes in cases with IPI=5 than with IPI=4. However, it presented no discriminated PFS and OS in low-risk, low-intermediate, high-intermediate, indicating a prognostic model to evaluate in elderly DLBCL patients is an urgent need. Recently, a study from Italy reported that combined the comprehensive geriatric assessment (CGA) with IPI score could identify three risk groups with the remarked difference in terms of OS [32,33] ; however, we did not perform this method on our cohort as the CGA lack.
Nonetheless, this study had several limitations. First, our study was retrospective, selection and information bias could not have been avoided. Second, in our research, 34.7% of patients were treated without rituximab, due to the heavy nancial burden of the family before 2017, which may exert a deteriorating impact on the whole cohort survival. Third, patients in our study had not done the CGA, which could predict OS and may assist clinicians in determining the treatment with full-dose chemotherapy or not [34,35] .

Conclusion
This study presented that rituximab improved the PFS and OS for the older than 70 years patients with DLBCL. Additionally, reduced chemotherapy dose had a deteriorated in uence on the OS for patients 70-79 years; however, it did not affect the survival outcome of older than 80 years. Furthermore, less than six cycles were related to the inferior OS and PFS. Collectively, rituximab, enough treatment cycles, full chemotherapy dose, and more e cient regimens with lower toxicities should be administered for elderly DLBCL patients under the balance of bene t and risk. To this end, a simpli ed predictive model to optimize the strati ed therapy needs to be explored in the future. The protocol of this retrospective study, involving individuals' clinical data collection, was approved and the need for informed consent was waived by the Ethics Committee of West China Hospital of Sichuan University.

Consent for publication
Not applicable Availability of data and material (ADM) All data generated or analyzed during the present study are included in this article. The authors declare that materials described in the manuscript, including all relevant raw data, will be freely available to any scientist wishing to use them for non-commercial purposes, without breaching participant con dentiality. The authors con rmed that all methods were performed in accordance with the relevant guidelines and regulations

Competing interests
The authors declare no competing interests.    (n=219). Kaplan-Meier survival curve of PFS and OS for elderly patients with DLBCL strati ed by rituximab usage. The 2-y PFS rate in rituximab-containing treatment group and without rituximab group was 59.5%, 44.2%, p 0.031; the 2-y OS rate in rituximab-containing treatment group and without rituximab group was 72.9%, 56.8%, p 0.008.  (n=219). Kaplan-Meier survival curve of PFS and OS for elderly patients with DLBCL strati ed by chemotherapy cycles. The 2-y PFS rate in group of < 6 chemotherapy cycles and in group of ≥ 6 cycles was 44.5%, 66.2%, p 0.000, the 2-y OS rate in group of in group of < 6 chemotherapy cycles and in group of ≥ 6 cycles was 59.5%, 77.6%, p 0.000. Figure 5 (n=239