We conducted a multi-institutional population-based analysis of the survival and toxicity associated with the addition of rituximab to chemotherapy for patients with diffuse large B-cell lymphoma (DLBCL), including patients aged ≥80 years, who were excluded from published randomized trials. Using population-based registries in Ontario, we identified 4021 patients who received chemotherapy with or without rituximab (R-CHOP [rituximab with cyclophosphamide, doxorubicin, vincristine and prednisone] or CHOP) for DLBCL between 1996 and 2007, including 397 patients aged ≥80 years. After propensity score matching, the overall survival (OS) and significant toxicities for R-CHOP and CHOP treatment groups were compared. R-CHOP was associated with a significant increase in 5-year OS compared to CHOP alone (62% vs. 54%; hazard of death = 0·78, P = 0·0004). Survival benefit was seen in all age groups, including those aged ≥80 years. Patients treated with rituximab did not have a significant increase in 1-year hospitalization rates for cardiac, pulmonary, gastrointestinal or neurological diagnoses compared to those treated with CHOP alone. The addition of rituximab to CHOP improves survival in the general population of patients with DLBCL and produces early survival benefit for very elderly patients, without any significant increase in the risk of serious toxicity.
Four randomized trials have demonstrated that the addition of rituximab to combination chemotherapy significantly improves the overall survival (OS) of patients undergoing primary treatment for diffuse large B-cell lymphoma (DLBCL) (Coiffier et al, 2002; Habermann et al, 2006; Pfreundschuh et al, 2006, 2008). Among elderly patients (i.e. aged >60 years), the combination of rituximab with cyclophosphamide, doxorubicin, vincristine and prednisone (R-CHOP) produced a 10–13% improvement in 3- to 5-year OS compared to treatment with CHOP (Coiffier et al, 2002, 2010; Feugier et al, 2005; Habermann et al, 2006; Pfreundschuh et al, 2008). Similarly, a 9% absolute improvement in 3-year OS has been reported among a younger (aged ≤60 years), good-prognosis population (Pfreundschuh et al, 2006).
However, the patients enrolled on these clinical trials may not be representative of the general population of patients with aggressive but potentially curable lymphomas. For example, in all but one of the aforementioned trials, patients over 80 years of age were explicitly excluded, and in the one trial that included patients age older than 80 years, only nine such patients were treated with rituximab (Habermann et al, 2006). Elderly patients tend to have more comorbidities (Coebergh et al, 1999), which may impact outcomes directly as a competing cause of death or indirectly by reducing treatment tolerability. Further, post-marketing reports on rituximab have cited increased rates of rare, but serious complications including hepatitis reactivation (Ennishi et al, 2010; Niitsu et al, 2010), interstitial lung disease (Katsuya et al, 2009), bowel perforation or obstruction, and progressive multifocal encephalopathy (PML) (Carson et al, 2009; Seruga et al, 2011). A limitation of these reports is that the number of patients treated (i.e. the total at risk for complications) is not clear, making it difficult to quantify the risk of these significant adverse outcomes. As a result of these issues, post-approval surveillance has been recognized as an important feature of optimizing the outcome of patients receiving novel cancer treatments (Committee on the Assessment of the US Drug Safety System, 2006).
Little is known about the outcome of unselected patients receiving R-CHOP chemotherapy, particularly the very elderly. One study of 292 patients reported a significant improvement in OS associated with R-CHOP, however, the outcome among very elderly patients, and toxicity were not reported (Sehn et al, 2005).
We conducted a population-based analysis to evaluate the impact of rituximab on the long-term survival and toxicity outcomes of patients with newly diagnosed DLBCL, with specific evaluation of different age groups including patients aged ≥80 years at diagnosis.
Design and methods
This is a population-based retrospective analysis of all individuals in Ontario with a new diagnosis of DLBCL, who received at least one cycle of CHOP-based chemotherapy (either with or without rituximab) between January 1, 1997 and December 31, 2007. The last follow up date was March 31, 2009.
Patients receiving rituximab for DLBCL were identified from the New Drug Funding Program (NDFP) of Cancer Care Ontario (CCO). The NDFP provided reimbursement to all hospitals in Ontario for rituximab in combination with chemotherapy for patients with newly diagnosed DLBCL beginning January 10, 2001 (for those aged 60–80 years) and expanded funding was applied from April 2, 2001 (for those aged >80 years), and July 1, 2004 (for all other age groups). Anonymized individual-level records for all patients receiving rituximab for this indication between January 2001 and December 2007 were obtained from the NDFP. CCO reimbursement policy required clinicians to confirm that patients receiving rituximab had a pathological diagnosis of DLBCL. To identify patients receiving CHOP alone, patients with a new diagnosis of DLBCL were identified from the Ontario Cancer Registry (OCR) from January 1997 to age-specific date of rituximab approval. OCR histology codes for CHOP patients were reviewed to exclude pathology that was incompatible with DLBCL or was inconclusive. A 10% sample of all pathology reports were reviewed in detail and confirmed that >90% of cases were compatible with DLBCL. Patients receiving CHOP chemotherapy were identified using CCO Activity Level Reporting (ALR) data, which includes individual-level treatment data regarding radiation therapy and chemotherapy drugs prescribed in 41 Regional Cancer Centers and general hospitals in Ontario. Patients were categorized as having CHOP alone if they received all of doxorubicin, cyclophosphamide, and vincristine within a span of 7 d, and had no evidence of having rituximab based on ALR or NDFP records. All patients in this study received at least one cycle of chemotherapy. Patients were excluded if histological diagnosis, Ontario Health Insurance Number, or treatment information were missing. We also excluded patients with human immunodeficiency virus infection or a previous diagnosis of lymphoma.
Patient age, gender, socioeconomic status (based on neighbourhood income quintile according to residential postal code), and vital status were obtained by deterministic linkage of cancer records to the Ontario Registered Persons Database, which contains demographic data on Ontario's 13-million population. In addition, records were linked to the Canadian Institute for Health Information (CIHI) hospital discharge database, and Ontario's outpatient physician claims database to identify comorbidities, hospital admissions, outpatient physician visits, and treatment information (billings for radiotherapy [RT] and chemotherapy, dates of drug administration). Comorbidity was measured using Adjusted Clinical Group (ACG) score, which is a population/patient case-mix adjustment system that provides a relative measure of morbidity and expected consumption of health services (Starfield et al, 1991). Stage and International Prognostic Index (IPI) score were not available. As a proxy we included three levels of treatment intensity: (i) 3–4 cycles of chemotherapy followed by RT within 60 d, (ii) ≥4 cycles of chemotherapy +/− RT and (iii) unclassifiable (if two or fewer cycles were administered or if an individual received three or four cycles without RT).
The CIHI hospitalization registry codes the major diagnosis accounting for each hospitalization using the International Classification of Disease (ICD) diagnosis codes (ICD-9 or ICD-10). ICD codes were used to identify post-diagnosis hospitalization for potentially toxic complications including infections (febrile neutropenia, other infections); cardiac problems (congestive heart failure, angina or ischaemic heart disease, arrhythmias); respiratory problems (pneumonia, interstitial lung disease); neurological problems (PML, cognitive impairment or delirium); or gastrointestinal problems (mucocutaneous reactions, hepatitis reactivation, bowel perforation or obstruction).
Patients receiving CHOP and R-CHOP were matched for age at diagnosis (within age groups: <19 years, 20–59 years, 60–69 years, 70–79 years, and ≥80 years) and treatment intensity (described above). Propensity to receive rituximab in addition to CHOP was then estimated using a multivariate logistic regression model using the following variables: gender, neighborhood income quintile, and histological diagnosis code. After hard matching on age group and treatment intensity, R-CHOP versus CHOP patients were then 1:1 matched based on propensity scores. Nearest neighbour matching using calliper width of 0·2 standard deviations of the logit of the propensity score was used (Austin, 2009). Matching was performed without replacement and all unmatched patients were removed from further analysis. Standardized differences were used to assess balance in patient and treatment characteristics between the treatment groups. Standardized differences compared the difference in prevalence within each row of the categorical variables, and the difference in means for continuous variables, measured in units of the pooled standard deviation (Austin, 2009). A standardized difference of <10% represents good balance in a given covariate between treatment groups (Austin, 2009).
OS was calculated based on time from diagnosis to date of death from any cause or date of last known follow-up, using the Kaplan-Meier method. OS was compared using the Klein test within the matched population (Klein & Moeschberger, 1997). The hazard of death (HRdeath) associated with R-CHOP versus CHOP was estimated using a univariate Cox model of survival in the matched cohorts. The Wilcoxon signed rank test was used to compare number of hospitalizations, and number of outpatient visits within 1 year of lymphoma diagnosis.
Research ethics approvals were obtained from St Michael's Hospital and Sunnybrook and Women's Health Science Centre, which acts as the board of record for research conducted at the Institute for Clinical Evaluative Sciences.
A total of 4021 patients treated at 53 different hospitals for DLBCL were identified: 2825 were treated with R-CHOP and 1196 received CHOP (Table 1). There were 1403 young, 2221 elderly, and 397 very elderly patients. Of these patients, 725, 1766, and 334 patients in each respective age group were treated with R-CHOP. Given that rituximab was initially approved only for those ≥60 years (January 2001–July 2004), the patients in the unmatched R-CHOP group were significantly older than in the CHOP group (P < 0·001), and had higher comorbidity scores (P < 0·001). The median follow up in the unmatched groups was 3·5 years (R-CHOP) and 9·7 years (CHOP), respectively. There was a significant difference in the distribution of histology codes (P < 0·001), probably related to time-dependent changes in lymphoma classification and in ICD codes that occurred during the decade span of our study (Clarke et al, 2006).
Following matching by age, treatment intensity and propensity score, there were 1099 patients in each of the R-CHOP and CHOP treatment groups. The median follow-up was 3·5 and 9·7 years, respectively. After matching, there were 586 matched young patients, 450 matched elderly patients and 63 matched very elderly patients in each of the R-CHOP and CHOP treatment groups for the age-based analysis. Treatment groups were well balanced with respect to the covariates included in the propensity score model in the overall cohort and absolute standardized differences were <10% in all categories (Table 1).
For all 2825 patients receiving R-CHOP, 3-year OS was 64% and 5-year OS was 57%. When examined by age group, the respective 3-year and 5 year-OS rates for those treated with R-CHOP were 77% and 76% in the young patients, 62% and 55% in the elderly patients and 44% and 33% in the very elderly patients (Fig 1).
Among matched patients, R-CHOP was associated with a significant increase in OS of 9% at 3 years compared to CHOP alone (68% vs. 59%; HRdeath = 0·78, P = 0·0004), and 8% at 5 years (62% vs. 54%; HRdeath = 0·78, P = 0·0004). The survival advantage associated with the use of rituximab in different age groups is summarized in Table 2 and illustrated in Fig 2. Among young patients, rituximab was associated with an 8% better 3-year OS (77% vs. 69%; HRdeath = 0·70, P = 0·002), which was maintained at 5 years (76% vs. 67%; HRdeath = 0·70, P = 0·002). Similar results were found for patients aged 60–79 years, with rituximab-treated patients having superior OS at 3 years (58% vs. 50%; HRdeath = 0·84, P = 0·09) and 5 years (52% vs. 41%; HRdeath = 0·81, P = 0·02). In the very elderly, R-CHOP was also associated with an increase in 3-year survival (47% vs. 32%; HR death = 0·73, P = 0·18), which had attenuated by 5 years (32% vs. 27%; HRdeath = 0·82, P = 0·36). This was not statistically significant although the absolute improvement in survival and relative reduction in mortality were comparable in magnitude to the results seen among younger patients.
Table 2. Overall survival outcomes in the matched treatment cohorts
The overall 1-year hospitalization rate was significantly greater for patients treated with CHOP alone than R-CHOP (67·2% vs. 61·2%, P < 0·001) while the number of outpatient physician visits was greater for patients receiving R-CHOP (median 43 vs. 39 within first year after diagnosis, P < 0·001, Table 3). The risk of hospitalization for interstitial lung disease was slightly increased among patients receiving rituximab (0·9% vs. 0·3%, P = 0·09), while there was no significant increase in the risk of hospitalization for hepatitis reactivation, febrile neutropenia, or other infectious, cardiac, pulmonary, or gastrointestinal diagnoses with the addition of rituximab. Comparisons of hospitalization rates within age strata of matched patients (<60, 60–79, ≥80 years) similarly did not reveal any significant increase in the risk of hospitalization for specific diagnoses among patients treated with rituximab (data not shown). The greatest difference was in the risk of admission for congestive heart failure among patients ≥80 years (R-CHOP: 4·8%, CHOP: 1·6%, P = 0·67).
Table 3. Toxicity outcomes according to matched treatment cohorts
CHOP, cyclophosphamide, doxorubicin, vincristine, prednisone; R-CHOP, CHOP + rituximab; N/A, not applicable.
P-values for rate differences among age groups: Infections: P = 0·07; Cardiac: P < 0·001; Respiratory: P = 0·013; Central Nervous System: P < 0·001; Gastrointestinal: P = 0·91; All-cause hospitalizations: P < 0·001.
One-year hospitalization rate (%)
Congestive Heart Failure
Interstitial Lung Disease
Central Nervous System
Hospitalizations within 1 year
% Patients Hospitalized
Outpatient visits within 1 year
Among all patients treated with R-CHOP (i.e. the entire unmatched R-CHOP cohort), the median number of hospitalizations was 2, and 64·8% required at least one hospital admission. Increasing age was associated with a significantly greater risk of hospitalization for infections (P = 0·07), as cardiac (P < 0·001), respiratory (P = 0·013), and central nervous system (P < 0·001) diagnoses, and all causes combined (P < 0·001). Only the admission rate for gastrointestinal diagnoses was not significantly higher among older patients (P = 0·91).
Based on the results of randomized trials, R-CHOP has been widely adopted as the standard of care for patients with DLBCL. However, results obtained in clinical trials may not be generalizable to the population at large, because trial patients are selected based on defined age ranges and those with significant comorbidities are usually excluded (Townsley et al, 2005). To our knowledge this is the largest population-based study to describe the outcome of patients with DLBCL and the benefit of rituximab outside of the context of a clinical trial, including patients over 80 years old at diagnosis, who were largely excluded from the landmark rituximab trials. The results from this study are important in providing realistic estimates of the true benefit of rituximab within an unselected population.
The 3-year OS rate among young patients receiving R-CHOP in this study was 16% lower than that reported by the MabThera International Trial (MInT) Group (77% vs. 93%, respectively), probably due to selection of patients with 0–1 IPI risk factors in the trial (Pfreundschuh et al, 2006). Among elderly patients, our results were consistent with the findings of the Eastern Cooperative Oncology Group 4494/Cancer and Leukemia Group B 9793 Intergroup (Habermann et al, 2006) and Groupe d' Etude des Lymphomes de l' Adulte non-Hodgkin lymphoma 98-5 (Feugier et al, 2005)trials: our 3-year and 5-year OS rates were 62% and 55%, respectively, compared to the 3-year OS of 67% (Habermann et al, 2006) and 5-year OS of 58% (Feugier et al, 2005) reported in these trials. The RICOVER-60 (cyclophosphamide, doxorubicin, vincristine, and prednisone plus rituximab [R-CHOP] for patients older than age 60 years) trial reported substantially better early survival among elderly patients, with a 3-year OS of 78% (Pfreundschuh et al, 2008), but R-CHOP was administered every 2 weeks compared to the 3-weekly schedule in the other trials. These findings illustrate the substantial variability that can occur among studies due to difference in case mix and in treatment administration.
The 8% increase in absolute 5-year OS associated with the addition of rituximab to CHOP in the whole cohort was comparable across ages, and was similar to that described in clinical trials (Coiffier et al, 2002, 2010; Feugier et al, 2005; Habermann et al, 2006; Pfreundschuh et al, 2006, 2008). Treatment with rituximab led to an 8% improvement in 3-year OS among our young, unselected patients, which was comparable to 9% absolute difference demonstrated by the MInT Group (Pfreundschuh et al, 2006). Among elderly patients, the improvements in 3-year OS (8%) and 5-year OS (11%) associated with rituximab were very similar to those reported in the Intergroup and GELA trials (Feugier et al, 2005; Habermann et al, 2006). Among the very elderly, the early survival benefit of R-CHOP at 3-years (15%) had reduced to 5% by 5 years, probably due to competing causes of death. The benefit of rituximab for very elderly patients both in terms of the HRdeath and the absolute 3-year OS improvement were comparable to those seen in younger patients. However, the exclusion of patients in the process of matching CHOP and R-CHOP patients rendered this improvement in outcome in the very elderly receiving rituximab not statistically significant. Although this study provides some insight into outcomes for the very elderly, we note that only patients who were able to tolerate at least one cycle of curative-intent chemotherapy were included, which is a select group amongst all patents aged ≥80 years diagnosed with DLBCL.
In a previous population-based study of 292 DLBCL patients (Sehn et al, 2005), OS and progression-free survival (PFS) were significantly better among the 152 patients treated in the post-rituximab era in British Columbia. While the magnitude of improvement in the 2-year PFS results was comparable to that seen in the GELA trial (Coiffier et al, 2002), the magnitude of the OS improvement following adoption of rituximab for primary therapy was 26% (31% among the 170 patients aged ≥60 years), over twice the survival benefit reported in clinical trials (Feugier et al, 2005; Pfreundschuh et al, 2006, 2008). The median follow up for the post-rituximab group in this study was two years, and longer follow-up of the British Columbia cohort is needed to determine whether this dramatic survival improvement is maintained.
In clinical trials, the addition of rituximab to CHOP chemotherapy was not associated with significant toxicities other than a trend toward increased infection (Feugier et al, 2005). Subsequent post-marketing surveillance, however, has noted cases of interstitial lung disease (Liu et al, 2008; Katsuya et al, 2009), hepatitis reactivation (Ennishi et al, 2010; Niitsu et al, 2010), bowel obstruction, and PML (Carson et al, 2009; Ram et al, 2009), leading to revisions of the product monograph for rituximab by the manufacturer and advisory warnings by regulatory authorities (http://www.hc-sc.gc.ca/dhp-mps/medeff/advisories-avis/public/index-eng.php; http://www.mhra.gov.uk/index.htm#page=DynamicListMedicines). However, reports are largely from case series, retrospective analyses, or small prospective single-centre studies, making it very difficult to establish the true rate of these adverse events. We found that, with the exception of febrile neutropenia, the rate of hospitalization for potentially toxic complications among all 2825 patients treated with R-CHOP was low and that the addition of rituximab did not result in significant increases hospital admissions for infectious, cardiac, respiratory, neurological, or gastrointestinal diagnoses. As expected, increasing age was significantly associated with a greater risk of hospitalization.
This study has limitations that warrant consideration. Firstly, despite matching using age, treatment intensity, and propensity scores to balance treatment cohorts, we did not have full prognostic data and consequently, were unable to match based on IPI score. As a proxy for disease extent, we used treatment intensity, but treatment practices may differ for localized disease and could lead to misclassification, although it is not clear that this would bias the apparent benefit of rituximab. Secondly, we used hospitalization diagnoses to identify toxicities, and consequently significant toxicities managed on an outpatient basis were not captured. Thirdly, patients treated with R-CHOP were diagnosed and treated more recently than those receiving CHOP, therefore other (i.e. non-rituximab) time-related improvements in management could be associated with superior outcomes in the more recently treated R-CHOP patients. Use of granulocyte-colony stimulating factor was probably more frequent in the recent era, but it is unlikely to have influenced survival outcomes, as we found no significant difference in infectious complications between the two treatment groups. Furthermore, with the exception of rituximab, randomized trials have shown that different chemotherapy regimens do not impact OS in DLBCL, so it is unlikely that the survival differences found in this study could be attributed to other systemic treatments.
While we did detect a greater frequency of outpatient physician visits among R-CHOP patients, the cause of these visits is not known and some may have been related to the administration of rituximab and CHOP on separate days, due to the length of rituximab infusion. There has been a widespread effort in Ontario over the last decade to limit elective hospitalizations, and the greater number of outpatient visits and fewer inpatient admissions could reflect the trend towards managing side effects in the outpatient rather than in-hospital setting. Even so, our results represent valid estimates of the risk of serious R-CHOP-related complications requiring hospitalization in the context of contemporary management.
As standards of care change, it is important to systematically evaluate the incremental impact of new cancer treatments on survival, toxicity and cost at a societal level. It is recognized that as new drugs or procedures diffuse into general use, the benefits and risks seen in clinical trials need to be re-evaluated (Committee on the Assessment of the US Drug Safety System, 2006). Our study provides evidence that, for patients with DLBCL, the magnitude of the survival benefit seen with rituximab in clinical trials is reproduced when applied to the general population of patients treated with systemic therapy, without an apparent increase in significant toxicity.
This work was supported by Cancer Care Ontario and the Ontario Ministry of Health and Long Term Care Drug Innovation Fund grant. Dr. Hodgson is supported by a Cancer Care Ontario Research Chair. All authors were involved in the design of the study, in analysing and discussing the data, and have read and approved the final version of the manuscript.