Impact of relative dose intensity of standard regimens on survival in elderly patients aged 80 years and older with diffuse large B-cell lymphoma

Diffuse large B-cell lymphoma (DLBCL) is the most common type of non-Hodgkin lymphoma (NHL), and the incidence of DLBCL increases with age.1 In recent years, world global life expectancy has been increasing impressively and Japan has been challenged with an unprecedented aging society.2 Consequently, a continuing increase in the incidence of elderly patients with DLBCL is anticipated, and their management is an urgent issue. Although it is widely accepted that maintaining relative dose intensity (RDI) of chemotherapies for DLBCL is crucial for a better prognosis,3 few studies have focused on the elderly population, despite its importance.54 Two previous phase II study demonstrated the efficacy and safety of approximately 50% of RDI of CHOP (cyclophosphamide [CPA], adriamycin [ADR], vincristine [VCR], and prednisolone [PSL]) with rituximab [RTX),76 but little is known about the efficacy and safety when a higher RDI is provided in this vulnerable population. We report that, even in very elderly patients, there is a considerable number of patients who could achieve a better prognosis owing to maintaining the higher total average RDI (tARDI) of standard regimens, CHOP and THP-COP (CPA, tetrahydropyranyl adriamycin [THP], VCR, and PSL) combined with or without RTX (R-),98 in real-world practice using a Cox hazards model with restricted cubic spline (RCS). The factors affecting the physician’s decision to reduce tARDI are age, dementia, elevated lactate dehydrogenase (LDH), the Charlson comorbidity index (CCI), and the international prognostic index (IPI).

To fill the evident gap in the management of very elderly patients in real-world practice, we conducted a retrospective, multicenter investigation at three tertiary institutions in Japan (see the Online Supplementary Materials and Methods for details). The impact of the higher tARDI on survival in a pure population of >80 years of age with DLBCL were assessed by the multivariate visual graspable analysis model taking into account the confounding factors. From 2007 to 2017, 177 patients aged ≥80 years at diagnosis were newly diagnosed with DLBCL. A total of 50 patients were excluded due to the exclusion criteria, and the remaining 127 patients were enrolled in the present investigation (Online Supplementary Figure 1). The patients’ characteristics before the initial treatment are summarized in Table 1. The median ratio of the duration of hospitalization to total treatment duration in the whole study population was 82.9%. Patients in the lower tARDI group had a significant higher ratio of the duration of hospitalization to total treatment duration than in the higher tARDI group (98.7% vs. 71.9%, P=0.004).

Table 1.Patients’ characteristics at diagnosis.

The median follow-up period was 15.4 (range: 0.30-107.6) months, 64 patients (50.4%) died, 38 (29.9%) of whom died of lymphoma, and four (3.1%) of whom died of adverse events (AE). The estimated overall survival (OS) was significantly higher in the tARDI >50% group (Online Supplementary Figure 2). The 2-year survival rate was significantly higher in the tARDI >50% group (50.8% vs. 61.8%, P=0.029). Figure 1 shows the correlation between tARDI and OS. A Cox hazards model with RCS was used, because this non-linear model is more suitable to reflect real-world practice than the linear model. A nearly linear association was observed between tARDI and mortality (for non-linearity P=0.780, for effect of tARDI P=0.049). A gradual decrease of risk of mortality as tARDI increased was observed even in the very elderly population. There was no significant interaction between RTX and tARDI for OS (for interaction P=0.143). In the multivariate Cox proportional hazards model for OS, the significant predictors for OS were IPI (hazard ratio [HR] 1.973, 95% confidence interval [CI]: 1.427–2.727, P<0.001]) and tARDI (HR 0.888, 95% CI: 0.809–0.975, P=0.013) (Online Supplementary Table S1). The significance of tARDI also persisted in sensitivity analysis (Online Supplementary Table S2).

Figure 1.Association between total average relative dose intensity and overall survival in 127 patients by a covariate-adjusted restricted cubic spline hazards model with three knots. The solid line represents the log hazard ratio, and the shaded area is the 95% confidence interval. tARDI: total average relative dose intensity.

There was no significant difference in severe AE between the two groups according to tARDI (Online Supplementary Table S3). Most patients (115 of 127 patients, 90.6%) received primary prophylaxis with granulocyte colony-stimulating factor (G-CSF) including pegfilgrastim, 14 patients (11.0%) received prophylaxis with levofloxacin (LVFX), and 13 patients (10.2%) received prophylaxis with both G-CSF and LVFX. Approximately 20% of the patients (39 of 127) underwent the pre-phase treatment with PSL before conventional-dose chemotherapy. According to a multivariate logistic regression model for grade 3/4 non-hematological toxicity and/or febrile neutropenia, there was no significant association between tARDI and severe AE. The only significant factor associated with severe AE was IPI (odds ratio: 2.19, 95% CI: 1.38–3.49) (Online Supplementary Table S4).

To identify the reason for the reduction of tARDI, a multivariate logistic regression model for reduction of tARDI to <50% was performed (Online Supplementary Table S5), but there were no significant predictors, thus a random forest model was created. Figure 2 demonstrates that the variable importance in the random forest model for the reduction of tARDI to <50% were age, dementia, elevated LDH, CCI, and IPI.

Figure 2.Importance of each predictor for the reduction of the total average relative dose intensity ≤50% in the random forest algorithm. Variable importance is a measure scaled to have a maximum of 100. CCI: Charlson comorbidity index; ECOG PS: Eastern Cooperative Oncology Group performance status; IPI: International Prognostic Index; LDH: lactate dehydrogenase.

The present study demonstrates that providing the higher tARDI of standard regimens can result in better survival even in the elderly population (≥80 years). Of note, at least 80 patients (63%) could safely receive the higher tARDI >50% and achieve a better prognosis. The higher tARDI steadily decreased the mortality risk. The physician’s decision to adjust tARDI was affected by age, dementia, elevated LDH, CCI, and IPI.

Two previous phase II trials demonstrated the efficacy and safety of immunochemotherapy with mini-CHOP (approximately 50% RDI) for patients aged >80 years. The 2-year survival in these phase II trials was 59% and 64%, respectively.76 However, the evidence from these trials could not be well generalized to the entire very elderly population because the participants in the clinical trials usually had better conditions, such as a good performance status (PS) <2, than real-world practice. In contrast, half of the participants in our investigation had a PS >2. Despite the frailer population than the previous trials, 2-year OS was 61.8% in the tARDI >50% group. We cannot simply compare the data between a prospective and retrospective study, but this result indicated the efficacy of maintaining higher tARDI was non-inferior to the previous clinical trials. The potential therapeutic advantage of full-dose chemotherapy still remains controversial.54 A Cox hazards model with RCS led to the conclusion that a higher tARDI results in a better OS even in very elderly patients. In our investigation, at least 63% of patients receiving standard regimens could undergo the higher tARDI (>50%). Surprisingly, taking into account the non-linear representation of RCS, the risk of mortality decreased linearly, was not concave or plateaued, as tARDI increased. In two multivariate Cox proportional hazards models, a remaining independent predictor for OS was tARDI not CCI or the G8 industrialized nation status. This result showed that the impact of tARDI on survival was not simply due to the frailty of the patients.

Indeed, the treatment of very elderly patients can be complicated and result in difficulty maintaining RDI in real-world practice.1054 In our investigation, considerable number of patients could obtain a better prognosis owing to the maintaining the higher tARDI. This indicates that the physician’s judgement when selecting patients fit for the higher tARDI was appropriate. The occurrence of severe AE was not significantly associated with tARDI. A factor affecting the frequency of severe AE was IPI, perhaps because it includes PS, reflecting the frailty. In other words, the frailer patients are the more likely they experience severe AE regardless of tARDI. In the random forest model, age, dementia, CCI, and IPI, which reflect patients’ frailty, affected the physician’s decision on adjusting tARDI. Physicians are likely to reduce tARDI appropriately depending on their clinical assessment of the risk of severe AE based on the patient’s frailty. Treatment-related mortality of the patients aged >80 years through standard regimens for DLBCL was 8%-18%,64 a higher rate than in our investigation (3.1%). This difference can be explained by the fact that nearly all patients were treated as inpatients for a large part of the treatment period and intensive supportive care with prophylaxis with G-CSF including pegfilgrastim and/or LVFX which was supported by the Japanese universal health coverage.11 Generous support by the Japanese universal health coverage leads to a low socioeconomic inequality in medical care. Most patients who undergo the standard regimens for DLBCL are treated as inpatients at least during the initial cycle for safety sake and commonly elderly Japanese patients are hospitalized. Although some issues such as the medical costs still remain, careful and intensive managements, such as hospitalization, probably enables very elderly patients to receive sufficient RDI. We cannot simply apply the emphasis in our investigation to other countries because Japan is the most advanced “super-aged” society in the world. However, Japan is an epitome of the future of many other developed countries.

In conclusion, maintaining the higher tARDI can achieve a better outcome even in this vulnerable population in real-world practice. Age, dementia, elevated LDH, CCI and IPI affected the physicians’ decision-making to reduce tARDI. Surprisingly, at least 63% of patients could safely undergo the higher tARDI >50% and achieve a better prognosis. In the coming global “super-aged” society, a larger cohort is warrant to determine the optimal tARDI and fit-criteria for the higher tARDI.