Rituximab/Bendamustine or Rituximab/Ifosfamide/Carboplatin/Etoposide as Second-Line Therapy for Relapsed/Refractory Diffuse Large B-Cell Lymphoma: A Real-World Analysis

Rituximab/bendamustine (RB) and rituximab/ifosfamide/carboplatin/etoposide (R-ICE) are commonly used to treat relapsed/refractory diffuse large B-cell lymphoma (DLBCL), although their effectiveness has not been compared in real-world practice. This study evaluated data from DLBCL patients who relapsed after or were refractory to rst-line therapy in an electronic health record (EHR)-derived database between 2011 and 2018. One hundred thirty-seven patients using RB and 270 patients using R-ICE were included for analysis. Transplantation after second-line therapy was considered a censored event in the time-to-event analyses. Patients in the RB group were older and had poorer performance status while there were no signicant differences in stage, cell of origin, and double-/triple-hit subtypes. Relative to the R-ICE group, the RB group had signicantly longer time-to-next-treatment (TTNT) and overall survival (OS). Subgroup analyses revealed that patients who were <70 years or had better performance status consistently had better TTNT and OS if they had received RB. Patients who had disease progression within 12 months after induction chemotherapy had a signicantly inferior prognosis, regardless of the salvage treatments. Multivariable analysis revealed that RB treatment independently predicted better TTNT and OS. These data indicate that RB may be an alternative to R-ICE as second-line therapy for selected DLBCL patients.


Introduction
During the last two decades, the development of combination immunochemotherapeutic regimens that incorporate rituximab (an anti-CD20 monoclonal antibody) has improved the response rates to induction therapy among patients with diffuse large B-cell lymphoma (DLBCL). Although R-CHOP treatment (rituximab, cyclophosphamide, doxorubicin, vincristine, and prednisone) initially achieves complete response (CR) in 70-80% of patients, 20% of patients have primary refractory disease and another 20-30% of patients eventually experience disease recurrence. [1][2][3][4][5][6] Thus, salvage therapies are needed to improve survival outcomes. [7][8][9][10][11][12] The CORAL study evaluated 191 patients who received second-line R-ICE therapy (rituximab, ifosfamide, etoposide, and carboplatin) and revealed a response rate of 64%, with 3-year rates of 31% for progressionfree survival (PFS) and 47% for overall survival (OS). 9 However, only 51% of the patients in the study underwent autologous stem cell transplantation (ASCT), which was primarily related to an insu cient response to the second-line treatment. Treatment using rituximab plus bendamustine (RB) is a popular alternative for patients who are not eligible for intensive treatment because of advanced age or comorbidities. [13][14][15] Various studies have revealed response rates of approximately 40-60% to RB treatment among patients with relapsed/refractory DLBCL, including among patients who were 5-10 years older than the patients in the CORAL study. [16][17][18][19] However, to the best of our knowledge, no studies have compared the effectiveness of RB and R-ICE as second-line treatments for relapsed/refractory DLBCL. Therefore, this study involved a real-world analysis to evaluate the effectiveness of these regimens, as well as the clinical parameters that might in uence the long-term survival outcomes.

Results
The Flatiron Health database included records for 5,787 patients who were diagnosed with DLBCL during 2011-2018, including 5,356 patients (92.6%) who received induction chemotherapy for DLBCL. The most commonly used regimens were R-CHOP-like regimens (78% , Table S1). Among patients who received induction therapy, 135 patients (2.5%) underwent upfront SCT after induction chemotherapy, 1,365 patients (25.5%) underwent second-line treatment for DLBCL, 2,888 patients (53.9%) did not receive further treatment until the data lock date (December 31, 2019), and 968 patients (18.1%) died during or after induction therapy. Therefore, the nal study population included 137 patients in the RB group and 270 patients in the R-ICE group (Figure 1). The most commonly used second-line treatments were the RB and R-ICE regimens (Table S2).
Comparing the clinical characteristics of patients who received RB or R-ICE treatment Table 1 shows that patients in the RB group were older than patients in the R-ICE group (median age: 77 years vs. 60 years). The RB group was also more likely to have an ECOG performance status of 2-3 and less likely to have received R-CHOP-like regimens as induction therapy. There were no signi cant differences in terms of sex, race, calendar year of diagnosis, stage at diagnosis, cell of origin (GCB vs. non-GCB), Ki-67 index, proportion of the double-/triple-hit subtypes, number of extranodal sites, rwPOD, or the time from the end of induction therapy to the start of second-line treatment.

Analyses of TTNT and OS
Relative to the R-ICE group, the group that received second-line RB had signi cantly longer median TTNT (5.7 months vs. 9.2 months, aHR: 1.448, 95% CI: 1.04-2.01, p=0.0276) and signi cantly longer median OS (9.2 months vs. 15.9 months, aHR: 1.589, 95% CI: 1.116-2.263, p=0.0102) ( Figure 2). Subgroup analyses revealed that, relative to patients with rwPOD at >12 months and regardless of the second-line treatment, patients with rwPOD at ≤12 months had signi cantly shorter median TTNT (16.1 months vs. 4 (Figure 3). Furthermore, among patients who were <70 years old and relative to the R-ICE subgroup, the RB subgroup had signi cantly longer TTNT and OS, while there was no signi cant difference between the RB and R-ICE subgroups among patients who were ≥70 years old ( Figure S1). Among patients with an ECOG performance status of 0-1 and relative to the R-ICE subgroup, the RB subgroup had longer TTNT and OS, while there was no signi cant difference between the RB and R-ICE subgroups among patients with an ECOG performance status of 2-3 ( Figure S2).

Discussion
Previous studies have revealed poor outcomes among patients with relapsed/refractory DLBCL, 1,2,6 although there is no consensus regarding the optimal second-line treatment at this time. 9-12, 17,19-21 To the best of our knowledge, this is the rst study to compare RB and R-ICE as second-line treatments for relapsed/refractory DLBCL. The de-identi ed EHR-derived database was used to perform real-world analysis of patients who were treated outside of clinical trials, which revealed that the RB group had signi cantly longer TTNT and OS than the R-ICE group, despite patients in the RB group tending to be older and have poorer ECOG performance statuses. The multivariable analysis revealed that second-line RB treatment appeared to reduce the risk of relapse or death, regardless of age, rwPOD, ECOG performance status, stage at diagnosis, double-/triple-hit subtype, and number of extranodal sites.
The CORAL study and other studies of non-Hodgkin lymphoma revealed that early progression predicted poor OS, without any signi cant differences in the effects of the second-line regimens. 9,22−24 In the present study, the median time from diagnosis to the start of second-line treatment (which we interpreted as rwPOD) was ≤12 months for the RB and R-ICE groups, which re ects the aggressive nature of relapsed/refractory DLBCL. As expected, patients in our study with rwPOD at ≤12 months had poorer outcomes than patients with rwPOD at >12 months, and patients with early relapse did not exhibit any treatment-speci c differences in their TTNT and OS outcomes.
The treatment outcomes varied between the RB and R-ICE groups, as death was observed for 91 patients in the RB group (66%) and 103 patients in the R-ICE group (39%). Furthermore, ASCT was only performed for two patients in the RB group (1.5%), while ASCT was performed for 124 patients in the R-ICE group (46%). These differences re ect real-world practice, where younger and more t patients tend to receive R-ICE treatment, while more frail patients tend to receive RB treatment. Nevertheless, >50% of the patients in the R-ICE group experienced second-line treatment failure and were not able to undergo ASCT, which highlights the di culty in identifying patients who can ultimately undergo ASCT after intensive chemotherapy. These results highlight the need for better therapeutic and diagnostic strategies. Unfortunately, despite the majority of patients not responding to even intensive chemotherapies, our understanding of the biomolecular basis for chemoresistance in relapsed/refractory DLBCL remains limited. Furthermore, there are few prognostic biomarkers that can be used to guide the selection of patients who are likely to respond to intensive chemotherapy and undergo ASCT or patients who are more suitable for alternative or novel treatment options. For example, some early-phase studies have provided promising data regarding the use of targeted therapy in combination regimens for treating relapsed/refractory DLBCL. [26][27][28][29][30][31][32][33] Nevertheless, long-term follow-up data are needed to evaluate the real-world outcomes of targeted therapy in this setting.
This study provided real-world results regarding second-line treatments for patients with relapsed/refractory DLBCL. However, the retrospective design is subject to various confounding factors, and it would be di cult to assess the potential contributions of these factors. In addition, we substituted TTNT for PFS (a more commonly used endpoint), as the EHRs did not contain su cient data to calculate PFS. Nevertheless, TTNT may be a clinically meaningful endpoint and has become increasingly used in real-world practice. [34][35][36][37][38][39] Lastly, this study included patients who were treated during a prolonged study period, and changes in treatment strategy may be a relevant factor, although we con rmed that the calendar years of diagnosis were fairly evenly distributed in both groups (Table 1).
In summary, we believe that this is the rst real-world analysis to compare the effectiveness of second-line

Patients and treatments
The EHR data were collected from January 1, 2011 to December 31, 2019, and we evaluated patients who were diagnosed with DLBCL between January 1, 2011 and December 31, 2018. Patients who received RB as second-line treatment for DLBCL were assigned to the "RB group" and patients who received R-ICE were assigned to the "R-ICE group". The index date was de ned as the date the patient started RB or R-ICE treatment. Oncologist-de ned and rule-based lines of treatment were considered, although drugs administered within 28 days were generally considered within the same line of treatment, while a subsequent line of treatment was considered when the patient received any additional immunochemotherapy agents or had a treatment gap of >120 days.

Endpoints
The present study used the time to next treatment (TTNT) and OS as real-world endpoints. The TTNT outcome was used as a surrogate for PFS, as the real-world database did not contain data regarding the timing of disease progression/relapse. The TTNT interval was de ned as the time from the index date to the rst instance of the next treatment line or death. The OS interval was de ned as the time from the index date to the time of death because of any cause. Patients were censored on the dates of receiving autologous or allogeneic stem cell transplantation (SCT), a disease course-modifying therapy, or the last follow-up. The time from diagnosis to the start of second-line treatment was interpreted as a substitute for real-world progression of disease (rwPOD).

Statistical analysis
Continuous variables were reported as median (range) and were compared using the Mann-Whitney U   Figure 1 Flow chart for patient identi cation.  at ≤12 months or >12 months. Patients with rwPOD at ≤12 months had shorter TTNT and OS than patients with rwPOD at >12 months, regardless of the second-line treatment they received. However, patients with rwPOD at >12 months had signi cantly longer TTNT and OS if they received rituximab/bendamustine (RB), relative to the other regimens.

Supplementary Files
This is a list of supplementary les associated with this preprint. Click to download. Supplementarymaterial.pdf