Background
Diffuse large B-cell lymphoma (DLBCL) is the most common non-Hodgkin lymphoma, with a median age at diagnosis of 70 years.1 Older adult patients are at risk for frailty, a multidimensional syndrome of loss of reserves (such as energy, physical ability, cognition, or health) that gives rise to vulnerability.2 Recent work has shown the importance of frailty as a predictor of survival in patients with DLBCL,3–6 although studies have been limited by sample size and have been performed mainly in tertiary care settings. Furthermore, the mechanism by which frailty impacts survival remains unclear: are patients undertreated, do they have poor treatment tolerance, or do they inherently have higher-risk lymphoma?
Using a population-based sample of patients, the objectives of this study were to determine the association between frailty and 1-year mortality in patients newly diagnosed with DLBCL receiving curative-intent therapy and to describe the healthcare utilization patterns of frail versus nonfrail patients during treatment.
Methods
A retrospective cohort study was conducted using linked population-based health administrative data in Ontario, Canada, housed at ICES (formerly known as the Institute for Clinical Evaluative Sciences), an independent, nonprofit research institute whose legal status under Ontario’s health information privacy law allows it to collect and analyze healthcare and demographic data, without consent, for health system evaluation and improvement. Anonymized, individual-level records are available for all 14 million residents of Ontario from approximately 1990 onward. Use of the data in this project is authorized under section 45 of Ontario’s Personal Health Information Protection Act and does not require review by a research ethics board.
Cohort Creation
The study cohort included patients aged ≥66 years living in Ontario who were newly diagnosed with DLBCL (including transformed follicular lymphoma) and received first-line curative-intent rituximab-containing chemoimmunotherapy between January 1, 2006, and December 31, 2017. We excluded patients who had a primary diagnosis of central nervous system DLBCL and HIV-associated lymphoma and those who had a competing cancer diagnosis in the year before cohort entry. Patients were followed until March 31, 2019. Supplemental eAppendix 1 provides further details on cohort creation (available with this article at JNCCN.org).
Exposure
A modified version of the generalizable frailty index (FI) derived and validated for use in Ontario health administrative data by McIsaac et al7 was used. This index contains 30 multidimensional variables indicative of frailty (eg, assisted living, socioeconomic status, mobility, healthcare use, comorbidities; see supplemental eTable 1 for full FI). The FI is calculated as the sum of all deficits divided by the total number of deficits measured. In the analysis, the exposure was categorized into binary (frail >0.21; nonfrail ≤0.21) and quartile variables as previously described.7
Covariates
We collected demographic information on patient age, sex, socioeconomic status (based on neighborhood income quintile according to residential postal code), and vital status by deterministic linkage of cancer records to the Ontario Registered Persons Database. Other accessed databases are detailed in supplemental eAppendix 1 and were used to identify comorbidities and determine comorbidity burden, calculate the frailty score, and identify emergency department visits and hospital admissions. Information on cause of death was obtained from the Ontario Registrar General dataset on deaths and was available for all deaths occurring up to December 2017. Census data from 2006 were used for characterizing baseline neighborhood-level educational attainment status. All datasets were linked using unique encoded identifiers and were analyzed at ICES.
Healthcare utilization was used as a surrogate for severe treatment-related toxicity and was defined as the number of emergency department visits not resulting in hospitalization, and hospitalizations not resulting in death. We also performed a sensitivity analysis in which we did include hospitalizations resulting in death. We looked at 2 time periods: (1) from the start of treatment to 3 weeks after the second cycle of treatment, and (2) from the start to the end of treatment. Length of stay was calculated as the total number of days in the hospital during each time period. Using the Canadian Institute for Health Information Discharge Abstract Database, we identified the primary diagnosis accounting for each hospitalization using ICD-9 or ICD-10 diagnosis codes.
Outcomes
The primary outcome was 1-year mortality from the index date, identified using vital status records from the Ontario Registered Persons Database. Secondary outcomes included overall survival, cancer-related death, death during active cancer treatment (defined by receipt of chemotherapy and/or radiotherapy during the 2 weeks preceding date of death), chemoimmunotherapy cycles received, and healthcare utilization during treatment.
Statistical Analysis
Descriptive statistics were reported using means and standard deviations, medians and interquartile ranges (IQRs), or frequency and percentages where appropriate. The Student t test was used to compare baseline characteristics, frailty scores, and chemoimmunotherapy cycles received between frail and nonfrail patients. The chi-square test was used to compare categorical variables. In the primary analysis, we used the Kaplan-Meier method to estimate the 1-year survival probability for frail and nonfrail patients, and for patients in each quartile of the frailty score. Cox proportional hazard models were used to generate the unadjusted and adjusted hazard ratios (HRs) comparing frail patients to nonfrail patients. All covariates were adjusted in multivariable Cox regression, and healthcare utilization was modeled as a time-varying covariate in the analyses. Sensitivity analyses were performed including cancer stage, given that more than half the records had missing stage data; year of treatment; or healthcare utilization during treatment, including hospitalizations that resulted in death. There were no missing outcomes data.
An exploratory univariable landmark analysis was conducted assessing whether hospitalization during the first 2 cycles of treatment, in patients who survived at least 2 cycles of treatment, was associated with 1-year mortality (index date unchanged) (supplemental eAppendix 2). Two cycles was chosen as a short time window after which clinicians might be able to reassess treatment tolerability. For this exploratory analysis, hospitalizations did not include patients who were already admitted to the hospital when starting their treatment. Exploratory analyses were also performed on frail patients, stratifying by those who survived ≤1 year compared with those who survived >1 year. All analyses were conducted using SAS version 9.4 (SAS Institute Inc.) (supplemental eAppendix 2).
Results
The study cohort included 5,527 patients aged ≥66 years diagnosed with DLBCL between January 1, 2006, and December 31, 2017. Of these patients, 5,216 (94%) had de novo DLBCL and 311 (6%) had transformed follicular lymphoma. Median age was 75 years (IQR, 70–80 years), and 48% (n=2,672) were female (Table 1).
Baseline Characteristics
The FI was scored using all 30 variables in 5,516 patients and using 29 variables in 11 patients who had missing neighborhood income quintile data (see supplemental eTable 2 for the distribution of frailty scores in frail vs nonfrail patients).
Using the FI, 2,699 (49%) patients were classified as frail. Frail patients tended to be older than nonfrail patients (median age, 76 [IQR, 71–81] years vs 74 [IQR, 70–79] years), more likely to live in a rural location (15.3% vs 12.7%), and more likely to be diagnosed with DLBCL as an inpatient (28.0% vs 15.9%) (Table 1). Although stage data were not available for all patients, frail patients seemed more likely to have advanced-stage disease than nonfrail patients (Table 1). As expected, based on the FI composition, patients who were frail also had a greater comorbid disease burden than nonfrail patients (aggregated diagnosis group [ADG] score, 13 [IQR, 12–15] vs 10 [IQR, 8–12]) and tended to reside in areas with lower neighborhood income quintiles (Table 1; see FI in supplemental eTable 1).
Overall Survival
The median duration of follow-up was 324 days (IQR, 103–1,028 days), during which 2,964 (54%) patients died (1,626 [60%] of frail and 1,338 [47%] of nonfrail). Within 90 days of initiation of first-line rituximab, the survival probability ± standard error among frail patients was 86% ± 0.67% versus 93% ± 0.47% for nonfrail patients (P<.0001). Within 1 year of first-line treatment, it was 68% ± 0.90% versus 81% ± 0.75%, respectively (unadjusted HR, 1.8; 95% CI, 1.6–2.0; P<.0001) (Figure 1A). The relationship between frailty and survival remained consistent when measured in quartiles (unadjusted HR compared with Q1: 1.6 [95% CI, 1.3–1.9] for Q2; 2.0 [95% CI, 1.7–2.4] for Q3; and 2.7 [95% CI, 2.3–3.2] for Q4; P<.0001) (Figure 1B).
Survival probability (n=5,527). (A) Time to 1-year mortality among overall cohort, stratified by frailty index (binary). (B) Time to 1-year mortality among overall cohort, stratified by frailty index (quartiles). (C) Overall survival among entire cohort, stratified by frailty score (binary).
Citation: Journal of the National Comprehensive Cancer Network 20, 6; 10.6004/jnccn.2022.7014
All covariates included in the multivariable model were significantly associated with 1-year mortality in univariable modeling (supplemental eTable 3). In multivariable modeling controlling for age, number of comorbidities, diagnosis with DLBCL as an inpatient, number of chemoimmunotherapy cycles received (time-varying covariate), and healthcare utilization during treatment (time-varying covariate), frailty as a binary exposure remained independently associated with 1-year mortality (adjusted HR, 1.50; 95% CI, 1.32–1.70; P<.0001; supplemental eTable 4). The association remained consistent when modeling frailty in quartiles (supplemental eTable 4).
The average age at time of death did not differ significantly between groups (Table 2). Frail patients were more likely to die during active cancer treatment (20.6% vs 17.7%; P=.0472) but were not more likely to die of lymphoma compared with nonfrail patients (70.8% vs 69.6%; P=.506).
Clinical Characteristics of Frail Versus Nonfrail Patients
Treatment Exposure
Patients who were frail tended to receive fewer curative-intent treatment cycles on average than patients who were not frail (Table 2). We found that 14% of frail patients received only 1 cycle of chemoimmunotherapy compared with 7% of nonfrail patients (P<.001), and that 48% of frail patients completed at least 6 cycles of chemoimmunotherapy compared with 61% of nonfrail patients (P<.001). Among patients who died in the first year after starting treatment, 34% of frail patients received 1 cycle of chemoimmunotherapy compared with 24% of nonfrail patients (P<.0001), and 22% of frail patients completed at least 6 cycles of chemoimmunotherapy compared with 34% of nonfrail patients (P<.0001).
A moderate number of patients in both groups experienced at least one dose delay, although this was more common in frail patients (Table 2).
Healthcare Utilization During Treatment
During treatment, 2,531 patients had a hospitalization, of whom 192 died in the hospital. Frail patients were significantly more likely to visit the emergency department, be admitted to the hospital, and be admitted to the ICU than nonfrail patients during both the first 2 cycles and the entire course of treatment (Table 2). Frail patients also spent a significantly longer number of days in the hospital compared with nonfrail patients. During both time frames, the top 3 reasons for admission among both frail and nonfrail patients were fever/infection/neutropenia, lymphoma, and symptom management/palliative care/difficulty managing at home. Frail patients were also more likely to be admitted for exacerbations of chronic conditions than nonfrail patients.
Exploratory Analyses
See supplemental eTable 5 and eFigure 1 for details of exploratory analyses, including the association of early hospitalization with death and characteristics of frail patients who survived ≤1 year versus >1 year.
Sensitivity Analyses
Sensitivity analyses on the primary multivariable model including stage, year of treatment, or healthcare utilization including hospitalizations resulting in death are shown in supplemental eTable 6. Frailty remained a significant predictor for death even when including these additional predictors.
Discussion
This large population-based study provides evidence to support the association of frailty with survival in patients with DLBCL receiving curative-intent therapy. We studied >5,500 older adult patients with newly diagnosed DLBCL who were deemed fit enough to tolerate curative-intent treatment by their treating physician. Despite this diagnosis, half of patients were classified as frail based on a validated FI, and their relative rate of death in the first year after starting treatment was 50% higher than that of nonfrail patients, even when controlling for other important variables such as age, stage, diagnosis as an inpatient, comorbidities, healthcare utilization (time-varying covariate), and number of chemoimmunotherapy cycles received (time-varying covariate).
Our work builds upon that of previous groups who laid the groundwork regarding frailty measurement in lymphoma. Tucci et al8 assessed 84 patients with DLBCL in 2009 using a comprehensive geriatrics assessment and showed that it was more effective than clinical judgment in the classification of frailty. Merli et al3 created and validated the Elderly Prognostic Index in 2021 for the prediction of outcomes in 1,163 patients with DLBCL; however, only 109 patients classified as frail and 271 classified as unfit received full-dose or reduced-dose therapy (vs palliative therapy). Isaksen et al9 studied 784 patients with DLBCL aged ≥70 years to develop a simplified frailty score that predicted 2-year overall survival. Fit patients benefited from full-dose R-CHOP (rituximab/cyclophosphamide/doxorubicin/vincristine/prednisone), whereas unfit and frail patients did not benefit from full-dose compared with attenuated R-CHOP (R-miniCHOP), although R-miniCHOP was still better than an anthracycline-free regimen. Although this was an important initial study of the association between treatment intensity and frailty, only 63 frail patients and 192 unfit patients had received R-CHOP (vs palliative therapy). Our study adds to this literature with its large sample size of >2,500 frail patients treated with curative-intent therapy and its real-world nature, including patients treated both in the community and at academic centers. We also are the first to measure healthcare utilization during treatment, as a surrogate for treatment-related toxicity, in frail versus nonfrail patients. Our finding of a 22% admission rate during treatment as a result of infection, with higher rates in frail patients, mirrors the findings of a recently published study deriving a risk score for infection-related death in older adult patients with DLBCL.10
In our study, frailty was measured using a modified version of the generalizable FI, developed for use with health administrative data in Ontario. Although this is a different method of measuring frailty compared with what other studies have used, results from the FI are known to be generally reproducible as long as they have sufficient multidimensional variables.11 Furthermore, frailty indices have high agreement with a clinical frailty phenotype, although FIs are considered to discriminate better at the lower end of the frailty spectrum.12 Thus, cross-study comparison across other studies measuring frailty in different ways is reasonable.
In previous studies of general medical populations, frailty was shown to be significantly associated with healthcare utilization.13 We similarly found that frail patients were significantly more likely to present to the emergency department, be admitted to the hospital, and/or be admitted to the ICU during treatment, with most admissions (60%) related to infection or lymphoma. A further 8% of admissions in frail patients were for the sole purpose of symptom management, palliative care planning, or functional decline at home. Length of stay was also significantly longer in frail patients than in nonfrail patients. These findings suggest that healthcare utilization during treatment could be considered a surrogate marker for treatment toxicity and/or symptomatic lymphoma.
What remains unclear is whether the underlying cause for mortality in frail patients is undertreatment or poor treatment tolerance. Although we found that frail patients tended to have more dose delays than nonfrail patients, a substantial number in both groups (41%) had at least one dose delay. However, frail patients also received fewer cycles of chemoimmunotherapy compared with nonfrail patients. It is unclear whether the differences in treatment characteristics were related to toxicity, progressive lymphoma because of undertreatment, progressive lymphoma resulting from disease biology (higher tumor burden or more chemo-refractoriness in frail patients), or a personal decision by the patient and provider. Because of data limitations, it is unknown what types of palliative regimens patients went on to receive.
It is important to note that the median overall survival of frail patients was 3.5 years, indicating that not all frail patients should be excluded from aggressive therapy. In an exploratory analysis, early hospitalization during the first 2 cycles of treatment predicted 1-year mortality in both frail and nonfrail patients. Early hospitalization, age, and comorbidity burden were the only factors that differentiated frail patients who survived ≤1 year compared with >1 year. Geriatric assessment may also help with identifying and managing these patients. Future work could focus on more precisely identifying the subset of frail patients who have worse survival, including treatment-related mortality.
During treatment, 19% of frail patients were admitted to an ICU. This was higher than the rate for nonfrail patients, suggesting that complications resulting in hospitalization may have enhanced severity in patients who are frail. Furthermore, frail patients were more likely to die during active systemic cancer treatment than nonfrail patients, suggesting that physicians may not be able to predict the trajectory toward death as easily in patients who are already frail from other conditions, leading them to continue therapies near the time of death. Understanding this trajectory may better aid patients in end-of-life planning and other discussions with their physicians.14 Finally, many frail patients were unable to complete even 1 or 2 cycles of chemoimmunotherapy, leading us to wonder whether predictive tools for poor tolerability of treatment may have influenced patient preferences and goals of care from the outset. On the other hand, could frailty have been modified to improve treatment tolerance? Currently, although we can assess the association of baseline frailty with overall survival, no studies have assessed whether frailty measurement and frailty modification impact treatment decision-making, quality of care, patient satisfaction, and quality of life.
Our study has limitations owing to the retrospective and population-based nature of the work. Because this was a retrospective study, we may have been limited by information bias resulting from incomplete or misclassified information. Given that it was a population-based study, we were missing granularity in relation to the DLBCL diagnosis; specifically, stage data were not available for all patients, and no information was available regarding the exact regimen that patients received, the dose of anthracycline given, or patients’ International Prognostic Index score, all of which are known to be associated with survival in patients with DLBCL.15–17 Controlling for these important factors in our analysis would have added to our certainty about the impact of frailty on mortality in these patients. In addition, in our multivariable analysis, we did not control for the severity of comorbidities or the number of hospital admissions. Finally, although we performed multiple statistical tests, we did not adjust for multiple comparisons in our analysis. However, we did not intend to test any specific null hypothesis, nor did we draw conclusions based on any specific P value, and the primary outcome analyzed using Cox regression shows the effect size and 95% confidence interval between both frail and nonfrail patients. Despite these drawbacks, the large sample size, population-based design, and inclusion of data on healthcare utilization in our study make it a unique contribution to the frailty literature in DLBCL.
Conclusions
In a large population-based study of 5,527 patients with DLBCL receiving curative-intent therapy, including 2,699 frail patients, we have confirmed that frailty is significantly associated with 1-year mortality independent of age, stage, diagnosis of lymphoma as an inpatient, comorbidities, number of chemoimmunotherapy cycles received, and healthcare utilization. Frail patients had poor treatment tolerability and high treatment-related toxicity, demonstrated by increased use of the emergency department, hospitalizations, and ICU admissions compared with nonfrail patients during treatment. This study highlights the importance of baseline frailty assessment in patients with DLBCL. Future work should focus on whether frail patients can be further stratified to identify those at highest risk of early death and whether knowledge of the impacts of frailty on treatment outcomes affect patient and provider treatment decisions and quality of life. In addition, future clinical trials should consider frailty status in their design and report outcomes stratified by frailty.
Acknowledgments
The authors appreciate the funding supports for this study, including from the Conquer Cancer Foundation of the ASCO Young Investigator Award, the Lymphoma Clinical Research Mentoring Program, the Hold ’Em for Life Oncology Clinician Scientist Award, and the Ontario Ministry of Health and Long Term Care Clinician Investigator Program. The authors also sincerely thank the Lymphoma Clinical Research Mentoring Program for mentorship and feedback regarding this project, particularly Drs. Christopher Flowers, Michael Williams, and Matthew Maurer. In addition, parts of this material are based on data and information provided by Cancer Care Ontario (CCO), the Canadian Institute for Health Information (CIHI), and the Ontario Registrar General (ORG) information on deaths, the original source of which is ServiceOntario. The opinions, results, view, and conclusions reported in this article are those of the authors and do not necessarily reflect those of CCO, CIHI, or ORG. No endorsement by CCO, CIHI, or ORG is intended or should be inferred.
References
- 1.↑
Smith A, Crouch S, Lax S, et al. Lymphoma incidence, survival and prevalence 2004-2014: sub-type analyses from the UK’s Haematological Malignancy Research Network. Br J Cancer 2015;112:1575–1584.
- 2.↑
Rockwood K, Song X, MacKnight C, et al. A global clinical measure of fitness and frailty in elderly people. CMAJ 2005;173:489–495.
- 3.↑
Merli F, Luminari S, Tucci A, et al. Simplified geriatric assessment in older patients with diffuse large B-cell lymphoma: the Prospective Elderly Project of the Fondazione Italiana Linfomi. J Clin Oncol 2021;39:1214–1222.
- 4.↑
Fama AMP, Allmer C, Casulo C, et al. Vulnerable Elders Survey-13 (VES-13) predicts 1-year mortality risk in newly diagnosed non-Hodgkin lymphoma (NHL) [abstract]. Blood 2019;134(Suppl 1):Abstract 69.
- 5.↑
Tucci A, Martelli M, Rigacci L, et al. Comprehensive geriatric assessment is an essential tool to support treatment decisions in elderly patients with diffuse large B-cell lymphoma: a prospective multicenter evaluation in 173 patients by the Lymphoma Italian Foundation (FIL). Leuk Lymphoma 2015;56:921–926.
- 6.↑
Merli F, Luminari S, Rossi G, et al. Outcome of frail elderly patients with diffuse large B-cell lymphoma prospectively identified by Comprehensive Geriatric Assessment: results from a study of the Fondazione Italiana Linfomi. Leuk Lymphoma 2014;55:38–43.
- 7.↑
McIsaac DI, Wong CA, Huang A, et al. Derivation and validation of a generalizable preoperative frailty index using population-based health administrative data. Ann Surg 2019;270:102–108.
- 8.↑
Tucci A, Ferrari S, Bottelli C, et al. A comprehensive geriatric assessment is more effective than clinical judgment to identify elderly diffuse large cell lymphoma patients who benefit from aggressive therapy. Cancer 2009;115:4547–4553.
- 9.↑
Isaksen KT, Mastroianni MA, Rinde M, et al. A simplified frailty score predicts survival and can aid treatment-intensity decisions in older patients with DLBCL. Blood Adv 2021;5:4771–4782.
- 10.↑
Eyre TA, Wilson W, Kirkwood AA, et al. Infection-related morbidity and mortality among older patients with DLBCL treated with full- or attenuated-dose R-CHOP. Blood Adv 2021;5:2229–2236.
- 11.↑
Searle SD, Mitnitski A, Gahbauer EA, et al. A standard procedure for creating a frailty index. BMC Geriatr 2008;8:24.
- 12.↑
Blodgett J, Theou O, Kirkland S, et al. Frailty in NHANES: comparing the frailty index and phenotype. Arch Gerontol Geriatr 2015;60:464–470.
- 13.↑
Zhang J, Xu L, Sun L, et al. Gender difference in the association of frailty and health care utilization among Chinese older adults: results from a population-based study. Aging Clin Exp Res 2020;32:1985–1991.
- 14.↑
Odejide OO, Uno H, Murillo A, et al. Goals of care discussions for patients with blood cancers: association of person, place, and time with end-of-life care utilization. Cancer 2020;126:515–522.
- 15.↑
Bataillard EJ, Cheah CY, Maurer MJ, et al. Impact of R-CHOP dose intensity on survival outcomes in diffuse large B-cell lymphoma: a systematic review. Blood Adv 2021;5:2426–2437.
- 16.↑
Bartlett NL. Approaches to aggressive B-cell lymphomas in less fit patients. Hematology Am Soc Hematol Educ Program 2020;2020:140–147.
- 17.↑
International Non-Hodgkin’s Lymphoma Prognostic Factors Project. A predictive model for aggressive non-Hodgkin’s lymphoma. N Engl J Med 1993;329:987–994.
