Febrile neutropenia (FN) is one of the most important adverse events of chemotherapy and was reported to be associated with increased mortality.1-3 Because elderly patients are vulnerable to FN, it is important to provide FN prophylaxis for elderly patients receiving chemotherapy.4, 5 To prevent FN, appropriate use of granulocyte colony-stimulating factor (G-CSF) is often recommended. We previously analysed the clinical impact of G-CSF for FN prophylaxis in day-to-day practice.6 The results showed that appropriate use of G-CSF reduced FN susceptibility and provided health-economic benefits. We further analysed the prophylactic effect of G-CSF according to age groups and found that the efficacy may be limited in very elderly patients aged ≥80 years.7 These results raised the possibility that very elderly patients had unique sensitivity to FN and suggested that different management strategies may be required for these patients. However, no comprehensive analyses have evaluated the FN risk in very elderly patients. Therefore, we conducted a retrospective study in Japan to determine the risk factors for FN in very elderly patients aged ≥80 years receiving chemotherapy.

This retrospective cohort study used data from the Diagnosis Procedure Combination database, a nationwide inpatient database in Japan. The database includes discharge abstracts and administrative claims data for >1200 acute-care hospitals and covers approximately 90% of all tertiary-care emergency hospitals in Japan. The study was conducted in accordance with the principles of the Declaration of Helsinki and was approved by the Ethics Review Board of The University of Tokyo. We identified patients aged ≥80 years who were hospitalized for previously untreated diffuse large B-cell lymphoma (DLBCL) between July 2010 and March 2017. Patients with DLBCL were identified by International Statistical Classification of Diseases and Related Health Problems (ICD)-10 code C83.3. We selected the initial hospitalization during which each patient received rituximab–cyclophosphamide–doxorubicin–vincristine–prednisolone (R-CHOP) therapy, because the first cycle of R-CHOP is generally performed in hospitals in Japan. The initial dose ratio (IDR; %) was calculated as the average ratio of the actual dose to the 100% dose of cyclophosphamide (750 mg/m2) and doxorubicin (50 mg/m2) in the first cycle of R-CHOP. Patients who received >120% or <40% IDR were excluded from the study. IDR was divided into three categories: 80% to 120%, 60% to <80% and 40% to <60%. Unreduced IDR was defined as IDR ≥80%. Primary prophylaxis with G-CSF was defined as non-therapeutic initiation of G-CSF between days 2 and 5 of the R-CHOP regimen. The primary outcome was occurrence of FN during the first cycle of R-CHOP. Patients who developed FN were defined as those who started treatment with antibiotics on the same day as a blood culture test between days 1 and 21 of the R-CHOP regimen. The antibiotics used to define FN were limited to cefepime, piperacillin-tazobactam and meropenem because these are the only antibiotics approved for FN treatment in Japan. Our previous validation study indicated that this definition was generally accurate for identification of FN patients.8 To exclude non-primary prophylactic use of antimicrobials and G-CSF, patients with active bacterial infections at admission were excluded from the study. Onset of FN was defined as the day of performance of a blood culture test. To account for clustering of patients within hospitals, we employed a multivariable logistic regression model with a generalized estimating equation and the Wald test to evaluate FN susceptibility using the following factors: primary antimicrobial prophylaxis, primary prophylaxis with G-CSF, IDR, sex, body mass index, Barthel index,9 Charlson comorbidity index,10 cancer stage, complications of fungal or viral infections at admission (ICD-10 codes: A08, A85–A90, B00, B02, B07, B09, B15–B27, B30, B33–B48, B97), surgery before chemotherapy, and central venous catheterization. To estimate the secular effect on FN prevention, we employed the Cochran–Armitage trend test. Primary antimicrobial prophylaxis was defined as use of non-therapeutic fluoroquinolones between days 1 and 5 of the R-CHOP regimen. Activities of daily living were evaluated using the Barthel index, with the maximum score of the index defined as fit. Statistical analyses were performed using IBM SPSS Statistics for Windows, version 25.0 (IBM Corp.). A two-tailed significance level of 0.05 was used in all analyses.

We identified 1 819 patients aged ≥80 years with newly-diagnosed DLBCL who received their first cycle of R-CHOP. The median length of hospital stay was 18 days (interquartile range: 13–23). The median age was 83 years (interquartile range: 82–85). The patient baseline characteristics are shown in Table 1. Overall, 73 patients (4%) received primary antimicrobial prophylaxis and 375 (21%) received unreduced IDR of R-CHOP. A total of 270 (15%) patients received primary prophylaxis with G-CSF. Pegfilgrastim was administered at a median of three days (interquartile range: 3–4) after chemotherapy. Daily G-CSF was initiated at a median of three days (interquartile range: 3–5) after chemotherapy and continued for a median of nine days (interquartile range: 7–10). FN occurred in 115 patients (6.3%) at a median of 10 days (interquartile range: 8–12) after administration of cyclophosphamide. To evaluate the secular effect on FN prevention, we assessed the occurrence of FN according to diagnostic year (Table S1). Although there was no significant difference, the occurrence of FN tended to be lower in recent diagnostic years. The results of the multivariable logistic regression analysis for occurrence of FN in the very elderly patients are shown in Table 2. Primary antimicrobial prophylaxis, primary prophylaxis with daily G-CSF and pegfilgrastim were not significantly associated with lower occurrence of FN [odds ratio: 0.69 (95% confidence interval: 0.30–2.21), 1.39 (0.73–2.64) and 1.43 (0.73–2.81) respectively]. Reduced IDR (40% to <60%) was significantly associated with lower occurrence of FN [0.54 (0.32–0.92)]. Advanced lymphoma stage (>stage 2) and central venous catheterization were significantly associated with FN occurrence [1.59 (1.02–2.49) and 3.19 (1.81–5.62) respectively].

TABLE 1. Baseline characteristics of the study population Characteristics Patients aged ≥80 years n = 1819 Antimicrobial prophylaxis 73 (4%) Primary prophylaxis with G-CSF None

1549 (85)

Pegfilgrastim 114 (6) Daily G-CSF 156 (9) Initial dose ratio (%) 80% to 120% 375 (21) 60% to <80% 697 (38) 40% to <60% 747 (41) Sex Male 915 (50) Female 904 (50) Body mass index (kg/m2) 18.5–24.9 1301 (67) <18.5 344 (18) 25.0–29.9 268 (14) ≥30.0 21 (1) Activities of daily living Fit 1192 (66) Frail 464 (26) No data available 163 (9) Charlson comorbidity index 2 1371 (75) >2 448 (25)

Cancer stage

1 or 2 652 (36) 3 or 4 976 (54) No data available 191 (11) Fungal or viral infections at admission 455 (25) Surgery before chemotherapy 31 (2) Central venous catheter 97 (5) Abbreviation: G-CSF, granulocyte colony-stimulating factor. TABLE 2. Multivariable logistic regression analysis for the occurrence of febrile neutropenia in the very elderly patients Factors

Patients aged ≥80 years

n = 1819

Odds ratio 95% confidence interval P Antimicrobial prophylaxis 0.69 0.30–2.21 0.685 Primary prophylaxis with G-CSF None Reference 0.367a Pegfilgrastim 1.43 0.73–2.81 Daily G-CSF 1.39 0.73–2.64 Initial dose ratio (%) 80% to 120% Reference 0.023a 60% to <80% 0.92 0.55–1.54 40% to <60% 0.54 0.32–0.92 Sex Male Reference 0.782 Female 0.95 0.65–1.39 Body mass index (kg/m2) 18.5–24.9 Reference 0.913a 18.5–24.9 1.06 0.66–1.72 25.0–29.9 1.06 0.64–1.75 ≥30.0 0.82 0.12–5.45 Activities of daily living Fit Reference 0.052 Frail 1.50 0.96–2.34 Charlson comorbidity index 2 Reference 0.842 >2 1.05 0.66–1.66 Cancer stage 1 or 2 Reference 0.042 3 or 4 1.59 1.02–2.49 Fungal or viral infections at admission 0.84 0.55–1.28 0.414 Surgery before chemotherapy 1.57 0.49–5.03 0.448 Central venous catheter 3.19 1.81–5.62 <0.001 Abbreviation: G-CSF, granulocyte colony-stimulating factor.

This nationwide analysis revealed the features of FN among very elderly patients with DLBCL, which remained to be well established. The proportion of very elderly patients who developed FN tended to be lower than the proportions in previous studies involving younger patients.6, 11, 12 There are several possible reasons for the lower incidence of FN in the present study compared with the previous studies, despite the older age of the patients and the lower percentage of prophylactic G-CSF use. One reason may be the high proportion of patients with reduced chemotherapeutic dose in the present study. Specifically, approximately 80% of patients received a reduced dose of R-CHOP therapy. This reduction in chemotherapeutic dose may have decreased the risk of FN even in our very elderly patients. Patients with DLBCL receiving R-CHOP were reported to be at high risk for infections compared with population-matched controls, and higher dose intensity was a major risk factor for infections including FN.13 Therefore, the occurrence of FN may be reduced in our patients aged ≥80 years who frequently received lower dose intensity of R-CHOP. In fact, a previous study that investigated a reduced-intensity R-CHOP regimen in patients aged ≥80 years showed a similar low frequency of FN (7.4%).14 In the present study, the incidence of FN with R-CHOP therapy in very elderly patients was low, despite the real-world clinical setting that includes patients with various comorbidities and poor activities of daily living. Another reason may be the exclusion of high-risk patients for FN because they had complications of active bacterial infections at admission. However, because such high-risk patients were also excluded in a previous prospective study,12 the impact of this exclusion is considered limited.

Adequate management for FN prevention in very elderly patients has not been fully investigated. One reason may be the difficulty in including very elderly patients in prospective trials because of complications or poor performance status. Another reason may be the limited number of very elderly patients with DLBCL. Regarding this point, Japanese data may be suitable for investigation of very elderly patients because Japan has one of the highest-aged populations. Our study highlighted a requirement for chemotherapeutic dose reduction in very elderly patients to prevent FN. The efficacy of pegfilgrastim was limited in the very elderly patients in the present study, consistent with our previous study.7 In the very elderly patients, only reduction in chemotherapeutic dose was significantly associated with lower occurrence of FN. Although reduction in chemotherapeutic dose was reported to be a risk factor for impaired treatment effect,15, 16 the risk may be minimized in very elderly patients.14 A systematic review also showed that survival was not affected by reduced dose intensity of R-CHOP in patients aged ≥80 years.17 Because dose reduction in very elderly patients has already been attempted in real-world clinical settings, the present study may not suggest a new practice. However, the findings may have importance in providing new evidence for current practice. Furthermore, the results showed that reduction in chemotherapeutic dose to 40% to <60% was associated with significantly lower occurrence of FN, while reduction to 60% to <80% was not. This may indicate validity for the current practice of reducing the dose by half in very elderly patients aged ≥80 years.

Some limitations of the present study should be considered. First, the database does not contain laboratory data and physical findings (e.g., neutrophil counts, body temperatures). Thus, patients with FN were defined as those who underwent a blood culture test and started treatment with antibiotics during chemotherapy. Our definition of FN may have included a certain number of patients without neutropenia because neutrophil counts were not included in the database. However, our validation study showed that the definition used in this study had a positive predictive value of 80%.8 Thus, we consider that patients who developed FN were identified appropriately. Second, the study only evaluated the incidence of FN in the first cycle of R-CHOP. We did not evaluate all cycles of R-CHOP. Finally, the study was restricted by its retrospective nature. Despite these limitations, this study using uniformly registered data for a large number of patients at the nationwide level suggested that very elderly patients, who are difficult to target in clinical trials, had a unique risk of FN.

In conclusion, this nationwide analysis showed that reduction in chemotherapeutic dose was commonly practiced and incidence of FN was not high in very elderly patients. Typical strategies with G-CSF for FN prevention may be less effective in very elderly patients.

Conflicts of interest

KM received lecture fees from Kyowa Kirin and Ono Pharmaceutical. TJ received consigned research funding from Tsumura and belongs to the laboratory of a joint programme with Tsumura. TJ received lecture fees from Tsumura, AstraZeneca, Sanofi and Boehringer Ingelheim. AS received lecture fees from Eisai. AH received lecture fees from Otsuka Pharmaceutical, Ono Pharmaceutical, Nippon Shinyaku, Takeda Pharmaceutical, Chugai Pharmaceutical and Jansen Pharmaceutical. KT received lecture fees from Ono Pharmaceutical, Chugai Pharmaceutical and Nippon Shinyaku. YM received lecture fees from Janssen Pharmaceutical, Otsuka Pharmaceutical, SymBio Pharmaceutical, AbbVie, Nippon Shinyaku, MSD, Eisai, Bristol-Myers Squibb, Ono Pharmaceutical, Chugai Pharmaceutical and Takeda Pharmaceutical. KF received a scholarship donation from Chugai Pharmaceutical. HY received lecture fees from Pfizer, Novartis, Boehringer Ingelheim, Chugai Pharmaceutical and Tsumura and advisory fees from Pfizer and Novartis. MK received research funding from Pfizer, Otsuka Pharmaceutical, Chugai Pharmaceutical, Astellas, Kyowa Kirin, Takeda Pharmaceutical, Teijin, Eisai, Sumitomo Dainippon Pharma, Nippon Shinyaku, AbbVie, Daiichi Sankyo and Ono Pharmaceutical; advisory fees from Kyowa Kirin, Celgene, Chugai Pharmaceutical and MSD; and lecture fees from MSD, Astellas, Otsuka Pharmaceutical, Ono Pharmaceutical, Celgene, Daiichi Sankyo, Sumitomo Dainippon Pharma, Takeda Pharmaceutical, Chugai Pharmaceutical, Janssen Pharmaceutical, Kyowa Kirin, AbbVie, Pfizer, AstraZeneca, Bristol-Myers Squibb, Amgen, Sanwa Kagaku, Sanofi, SymBio Pharmaceutical and Nippon Shinyaku. None of these are related to the current study.

Author contributions

KM and TJ designed the research; AS, AH, KT, YM, HY and MK advised on the research design and analyses; HY, HM and KF collected the patient data; KM and TJ analysed the data; KM wrote the paper. All authors revised the manuscript and approved the final manuscript.