Frontline treatment options for patients with follicular lymphoma (FL) include chemotherapy plus rituximab [1]. Randomized clinical trials have demonstrated that rituximab added to frontline CHOP (cyclophosphamide [C], doxorubicin, vincristine [V], and prednisone [P]) or CVP results in improved overall survival in patients with advanced disease [2, 3]. However, the impact of rituximab has not been evaluated in routine clinical practice where differences in the treated population and treatment practices could produce differences between trial efficacy and “real-world” effectiveness. In this study, we used data from the Surveillance, Epidemiology, and End Results (SEER)—Medicare database to identify a cohort of 1,117 elderly patients (>66) who received frontline CHOP or CVP, with or without rituximab. The median age was 73, compared to between 52 and 57 in the clinical trials [2, 3] depending on the treatment group and trial, and 38% had Stage I/II disease, an exclusion criterion in the trials. In multivariate analysis, we found chemotherapy regimens that included rituximab were associated with lower overall mortality and non-Hodgkin's lymphoma (NHL)—specific mortality, but not mortality due to other causes. Our findings indicate that the survival benefits of rituximab observed in clinical trials translate into benefits for elderly patients in routine clinical practice.

FL is the second most common subtype of NHL, with an estimated 14,500 new cases in the United States in 2008 [4, 5]. The incidence of FL increases with age, and in the US the median age at diagnosis is approximately 64 years [6]. Frontline treatment of FL depends on histologic grade and extent of disease. FL is classified into three histological grades according to the number of centroblasts, and presently the National Comprehensive Cancer Network recommends that patients with Grade 3 FL be managed according to the diffuse large B-cell lymphoma guidelines with rituximab plus CHOP (R-CHOP) [1]. Recommended frontline treatment options for advanced (Stage III/IV) Grade 1 or 2 disease also include chemotherapy with or without rituximab [1].

The efficacy of rituximab as frontline therapy for advanced FL was established in two randomized trials. The German Low Grade Lymphoma Study Group compared CHOP (n = 205) to R-CHOP (n = 223) in a cohort of patients with untreated, advanced-Stage III/IV FL. The study showed a statistically significant improvement in overall survival, with six deaths in the R-CHOP group compared to 17 deaths in the CHOP group within the first three years [2]. A second Phase III study compared CVP alone (n = 162) to R-CVP (n = 159) in a similar cohort of untreated patients with FL [3]. In that study, the percent of patients surviving four years was significantly higher in the R-CVP (83%) versus the CVP alone group (77%) (hazard ratio [HR] = 0.60; 95% confidence interval [CI]: 0.38–0.96). The median age of patients enrolled in these trials was 52–57 depending on the treatment group and trial, which is considerably younger than the general population of patients with FL; one trial [2] did not include patients with Grade 3 disease.

Evidence from clinical trials indicates rituximab added to frontline chemotherapy for FL improves survival. However, to our knowledge this finding has not been confirmed in routine clinical practice in an elderly population. Therefore, we used SEER-Medicare data from 1999 through 2007 to examine the association between rituximab and survival in a cohort of patients receiving frontline chemotherapy for FL in routine clinical practice. The inclusion of only elderly patients in the SEER-Medicare database provided a unique opportunity to explore whether the benefits of rituximab seen in clinical trials also extend to the population of elderly patients with FL commonly encountered in clinical practice.

The study cohort included 1,117 patients who began chemo-immunotherapy (C-I therapy) consisting of CHOP, R-CHOP, CVP, or R-CVP within 90 days following diagnosis of FL. The median age at diagnosis was 73 years, 54% were female, 38% were diagnosed with Stage I or II disease, and 68% (754) received CHOP with (543 [72%]) or without rituximab (Table I). Patients receiving rituximab were similar to those who received chemotherapy alone with respect to age, gender, race, stage at diagnosis, lymph node involvement (extranodal versus nodal), and National Cancer Institute Comorbidity Index score. Patients receiving rituximab were diagnosed in later years, had higher grade histology, were more likely to receive CHOP, and were less likely to receive radiation.

Table I. Patient Characteristics
CharacteristicOverall (n = 1,117)Rituximab plus chemotherapy (n = 750)Chemotherapy alone (n = 367)P valuea
  • a

    Based on Chi-Square analysis.

  • b

    Cells in this row intentionally left blank as one or more cells contains n < 11.

  • c

    Percents are of all cause in that column.

Year of diagnosis
 1999b      <0.0001
 Grade 118816.8%11214.9%7620.7%0.02
 Grade 227424.5%17523.3%9927.0%
 Grade 334530.9%24833.1%9726.4%
Lymph node involvement
NCI Comorbidity Index score
Treatment type
Cause of death
 All cause38834.7%19225.6%19653.4% 
 Other than NHLc22457.8%12565.1%9950.5% 

In multivariate logistic regression analysis of factors associated with receiving rituximab alongside chemotherapy (Table II), patients receiving CVP were less likely to receive rituximab than those receiving CHOP. Patients diagnosed after 2003 were more likely to receive rituximab. In multivariate survival analysis (Table III), rituximab plus chemotherapy was associated with lower all-cause mortality (HR = 0.62; P < 0.0001) and with lower NHL mortality (HR = 0.46; P < 0.001), but not with mortality due to causes other than NHL (HR = 0.80; P = 0.17). In the model of NHL mortality, diagnosis after 2003 also was associated with lower mortality, whereas older age and later stage were associated with higher mortality. In the model of other-cause mortality, older age, male gender, higher NCI Comorbidity Index, and diagnosis after 2003 all were associated with higher mortality.

Table II. Multivariate Analysis of Factors Associated with Receiving Rituximab
 ORaP valueb95% CI
  • a

    Odds ratio.

  • b

    Based on Chi-square analysis.

 Grade 21.060.800.671.69
 Grade 31.360.190.862.14
NCI Comorbidity Index Score
Year of diagnosis
 2003 or beforeRef.   
 After 200315.42<0.00019.9323.94
% age 25+ w/ some college
% living in poverty
Lymph node involvement
Table III. Multivariate Survival Analysis
VariableAll cause mortalityNHL mortalityOther cause mortality
HRaP value95% CIHRP value95% CIHRP value95% CI
  • a

    Hazard ratio.

 Chemotherapy aloneRef.   Ref.   Ref.   
 Rituximab plus chemotherapy0.62<0.00010.490.780.46<0.00010.320.650.800.170.591.10
 66–<70Ref.   Ref.   Ref.   
 FemaleRef.   Ref.   Ref.   
 Non-whiteRef.   Ref.   Ref.   
 Grade 1Ref.   Ref.   Ref.   
 Grade 20.990.940.721.361.000.990.611.661.010.980.661.54
 IRef.   Ref.   Ref.   
NCI Comorbidity Index Score
 0Ref.   Ref.   Ref.
Year of diagnosis
 2003 or beforeRef.   Ref.   Ref.   
 After 20030.970.850.731.290.460.010.280.761.73<
% age 25+ w/some college
 0–24%Ref.   Ref.   Ref.   
% living in poverty
 0–4%Ref.   Ref.   Ref.   
Lymph node involvement
 NodalRef.   Ref.   Ref.   

We performed a study using data from routine clinical practice to examine the association between rituximab and survival in patients receiving frontline chemotherapy for FL. Our study selection criteria were developed to include patients who were similar to those enrolled in the clinical trials of rituximab in advanced FL [2, 3]. Specifically, we included only patients who received C-I therapy consistent with the clinical trials in advanced FL, [2, 3] provided that CHOP or CVP with or without rituximab was the first infused C-I therapy regimen and that it began within 90 days after diagnosis. The selection criteria also were developed to exclude patients who, by virtue of their initial treatment regimen (for example, radiation therapy alone or rituximab monotherapy), may have differed in FL severity and underlying comorbidity from those who received frontline therapy with a multidrug C-I therapy regimen. In doing so, we sought to minimize the potential for selection bias inherent in all observational data analysis. Further, since data on oral therapy were not available in SEER-Medicare at the time our study was performed, we sought to minimize the chance of misclassifying as treatment-naïve those patients who received oral chlorambucil as frontline therapy, or misclassifying as rituximab monotherapy those who received rituximab plus chlorambucil.

Despite the selection criteria, our study population was quite different from those enrolled in clinical trials that first demonstrated the efficacy of rituximab in advanced FL. Our population was considerably older, which was in part an artifact of the dataset and selection criteria we used. Most Medicare beneficiaries qualify for coverage because they are at least 65 years old, and in our study we required at least one year of Medicare coverage prior to FL diagnosis to ensure we could calculate an NCI Comorbidity Index score for each patient. Also, our population included patients with Stage I/II and/or Grade 3 FL, whereas these patients were excluded in one or both of the Phase III trials [2, 3].

It is important to note several limitations of our study. First, it is possible that part or all of the observed difference in NHL survival between the treatment groups is due to unobserved differences in the characteristics of patients in these groups, or in the treatment practices between the groups. For instance, Medicare claims do not contain laboratory data so we were unable to construct a Follicular Lymphoma International Prognostic Index (FLIPI) score for patients [7]. Also, we could not account for the relative dose intensity of C-I therapy. Moreover, we elected not to classify patients according to the number of cycles in the course of C-I therapy, as this would have required moving the start of the observation period in the survival analysis back to the end of the initial course to avoid immortal time bias [8]. Second, any errors in coding the cause of death in SEER-Medicare may have resulted in under- or overestimating the impacts on NHL and other cause mortality. For instance, if death due to NHL had been under coded—perhaps a more likely scenario—it may have resulted in underestimating the impact on NHL mortality and overestimating the impact on other cause mortality. Third, we chose to use the ICD-O-3 histology codes for Grade 1, 2, and 3 FL diagnosed between 1999 and 2005 even though SEER did not begin to use ICD-O-3 until 2001. Since ICD-O-3 codes differ slightly from ICD-O-2, this could have resulted in a slightly different classification of Grade before versus after the change in codes. Finally, we did not consider other uses of rituximab in our analyses, such as second-line C-I therapy or frontline monotherapy. As noted earlier, patients receiving these regimens may differ considerably from our study population, and our findings may not be generalizable to other types of patients with FL.

These differences notwithstanding, our findings indicate that rituximab was associated with improved NHL and overall survival in this cohort of elderly patients with FL treated in routine clinical practice. Since rituximab was not associated with death due to causes other than NHL, its impact on overall survival appears to be due entirely to NHL survival. These findings support those from the trials, and may be reassuring to physicians who often are required to make treatment decisions for individual patients based on clinical trials data that are usually not population-based or particularly large.


Data source

The source of data for this study was the National Cancer Institute's (NCI) Surveillance, Epidemiology, and End Results (SEER) cancer registry linked to Medicare enrollment and claims data (SEER-Medicare data). This database has been described in detail elsewhere [9]. Briefly, as of 2010, SEER collects and publishes cancer incidence and survival data from 18 population-based cancer registries throughout the United States covering approximately 26% of the U.S. population [10]. The registries routinely collect data on patient demographics, primary tumor site, tumor morphology and stage at diagnosis, first course of treatment, and follow-up for vital status. In the SEER-Medicare data, for persons age 65 years or older, 97% are eligible for Medicare and 93% of patients in the SEER files are matched to the Medicare enrollment file [11]. At the time this study was performed, the SEER-Medicare linkage included all Medicare eligible persons appearing in the SEER data through 2005 and their Medicare claims for Part A (inpatient) and Part B (outpatient and physician services) through 2007.

Patient eligibility

Patients were included in this study if they were diagnosed with FL between January 1, 1999 and December 31, 2005, FL was the first primary cancer diagnosed, they began systemic chemo-immunotherapy (C-I therapy) consisting of CHOP, R-CHOP, CVP, or R-CVP within 90 days following diagnosis, and they survived at least 60 days after the beginning of their C-I therapy. Identification of FL was made using four World Health Organization (WHO) ICD-O-3 histology codes: 9695 (FL Grade 1), 9691 (FL Grade 2), 9698 (FL Grade 3), and 9690 (FL not otherwise specified) [12, 13]. Since, for confidentiality reasons, SEER provides only the calendar month in which cancer is diagnosed, the date of FL diagnosis was defined as the first day of the calendar month in which patients were diagnosed. Medicare Part A and B claims containing ICD-9-CM procedure codes [14] and Healthcare Common Procedure Coding System (HCPCS) codes [15] were used to identify the beginning date and type of C-I therapy. The use of prednisone was assumed when the other agents were present because oral medications are not available in the dataset. The first 60 days following the beginning of C-I therapy was used to establish the type of regimen.

Patients were excluded for the following reasons: FL diagnosed before age 65; diagnosis made by death certificate or autopsy; death within the first month following diagnosis; or Medicare enrollment less than 12 months before diagnosis. In addition, to ensure complete claims history, patients had to have been enrolled in both Medicare Parts A and B, with no health maintenance organization (HMO) coverage for 12 months prior to diagnosis.

Mortality and censoring

The date of death was assigned by using the Medicare date, if available, even in cases where the SEER date was also available. The Medicare date was preferred because it is more current than the SEER date [16]. In cases where date of death was available in SEER but missing for Medicare, the SEER date was used. The cause of death was classified as non-Hodgkin's lymphoma (NHL) or other cause based on ICD-9-CM codes. All other patients were assumed to be alive at the end of the analysis period (December 31, 2007), although they may have been censored earlier for other reasons, such as switching to HMO coverage.

Patient characteristics

Patients were described according to their demographic, clinical, and socioeconomic characteristics. Patient age at diagnosis was stratified into four groups: 66–69; 70–74; 75–79; and ≥80. Requiring eligible patients to have at least one year of Medicare enrollment prior to diagnosis ensured that the minimum age in the cohort was 66 years. Race/ethnicity was defined using the SEER recoded race variable as white, black, Hispanic, and other (which consists predominantly of American Indian/Native Alaskan, Native Hawaiian or Other Pacific Islander, and Asian) [17].

Summary staging is the approach SEER uses to categorize how far a cancer has spread from its point of origin [18]. FL is classified as Stage I–IV according to the number of lymph node regions, the location of those regions, involvement of the spleen, and involvement of extralymphatic organs/sites and/or bone marrow involvement. With the exception of Stage IV disease, in which patients have either multifocal involvement or bone marrow involvement, patients classified as Stage I–III may or may not have extranodal involvement. Consequently, patients also were classified according to whether their disease was confined to one or more lymph node regions (nodal), or involved the spleen or an extralymphatic organ or site (extranodal). Finally, they were classified according to the Grade of FL.

We used the Medicare inpatient (Part A) and physician (Part B) claims to calculate an NCI Comorbidity Index for each patient [19]. This approach [20, 21] entails first removing claims that are considered to have unreliable diagnosis coding, such as those for testing procedures used to rule out conditions. Then, remaining diagnosis and procedure codes are used to identify the 15 noncancer comorbidities in the Charlson Comorbidity Index (CCI) [22]. The algorithms used to identify these conditions reflect the Deyo [23] adaptation of the CCI, and include several procedure codes from the Romano [24] adaptation. A weight is assigned to each condition, and the weights are summed to obtain the Index for each patient.

Socioeconomic information for individual patients is not available. Instead, the SEER-Medicare dataset contains information from the 2000 Census, reported at the tract level in which the patient lives, for median income, percent of the population living in poverty, and percent of those age 25 years or older with some college. We used these as indicators of the socioeconomic status of individual patients in the FL cohort. SEER registry and the assigned metropolitan statistical area were used as geographic indicators.

Statistical analysis

Logistic regression analysis was used to identify factors associated with receiving a C-I therapy regimen that included rituximab compared to one that did not include rituximab. Cox proportional hazards models were used to identify factors associated with death due to NHL, and separately those associated with death due to other causes. To avoid immortal time bias [8], a period of observation during which an event cannot occur, we began the survival analysis on day 61 following the start of C-I therapy.


This study used the linked SEER-Medicare database. The interpretation and reporting of these data are the sole responsibility of the authors. We acknowledge the efforts of the Applied Research Program, NCI; the Office of Research, Development and Information, CMS; Information Management Services (IMS); and the Surveillance, Epidemiology, and End Results (SEER) Program tumor registries in the creation of the SEER-Medicare database. We gratefully acknowledge the editorial assistance of Jen Deuson. Dr. Reyes is an employee of Genentech. Drs. Griffiths, Gleeson, Knopf, and Danese are consultants to Genentech.


  1. Top of page
  • 1
    National Comprehensive Cancer Network. NCCN Clinical Practice Guidelines in Oncology™ Non-Hodgkin's Lymphomas Version 1. 2010. (accessed 27 August 2010).
  • 2
    Hiddemann W,Kneba M,Dreyling M, et al. Frontline therapy with rituximab added to the combination of cyclophosphamide, doxorubicin, vincristine, and prednisone (CHOP) significantly improves the outcome for patients with advanced-stage follicular lymphoma compared with therapy with CHOP alone: Results of a prospective randomized study of the German Low-Grade Lymphoma Study Group. Blood 2005; 106: 37253732.
  • 3
    Marcus R,Imrie K,Solal-Celigny P, et al. Phase III study of R-CVP compared with cyclophosphamide, vincristine, and prednisone alone in patients with previously untreated advanced follicular lymphoma. J Clin Oncol 2008; 26: 45794586.
  • 4
    American Cancer Society. Cancer Facts and Figures. 2008. (accessed 10 July 2010).
  • 5
    Armitage JO,Weisenburger DD. New approach to classifying non-Hodgkin's lymphomas: clinical features of the major histologic subtypes. Non-Hodgkin's Lymphoma Classification Project. J Clin Oncol 1998; 16: 27802795.
  • 6
  • 7
    Solal-Celigny P,Roy P,Colombat P, et al. Follicular lymphoma international prognostic index. Blood 2004; 104: 12581265.
  • 8
    Suissa S. Immortal time bias in pharmaco-epidemiology. Am J Epidemiol 2008; 167: 492499.
  • 9
    Warren JL,Klabunde CN,Schrag D, et al. Overview of the SEER-Medicare data: Content, research applications, and generalizability to the United States elderly population. Med Care 2002; 40: IV-18.
  • 10
    National Cancer Institute. Overview of the SEER Program [Internet]. Bethesda: National Cancer Institute; 2010. (accessed 27 August 2010).
  • 11
    National Cancer Institute. Overview of the SEER Program [Internet]. Bethesda: National Cancer Institute; 2010. (accessed 27 August 2010).
  • 12
    North American Association of Central Cancer Registries: NAACCR ICD-0-3 Implementation Work Group. Guidelines for ICD-0-3 Implementation. (accessed 10 July 2010).
  • 13
    World Health Organization. International Classification of Diseases for Oncology, 3rd ed. World Health Organization; 2000.
  • 14
    Practice Management Information Corporation. ICD-9-CM, 6th ed. Los Angeles: Practice Management Information Corporation; 2005.
  • 15
    Practice Management Information Corporation. HCPCS. Los Angeles: Practice Management Information Corporation; 2005.
  • 16
    ResDAC. Research Data Assistance Center. CMS 301: Using SEER/Medicare Data for Research. ResDAC Minneapolis, MN, 15 May–16 May, 2008.
  • 17
    FritzA, RiesL, editors. SEER Program Code Manual, 3rd ed. Bethesda: Cancer Statistics Branch, Surveillance Program, Division of Cancer Control and Pop Sciences, National Cancer Institute, National Institutes of Health, Public Health Service, U.S. Dept of Health and Human Services; 1998.
  • 18
    Young JLJr,Roffers SD,Ries LAG, et al. SEER Summary Staging Manual—2000: Codes and Coding Instructions. NIH Pub. No. 01-4969. National Cancer Institute; 2001.
  • 19
    Klabunde CN,Potosky AL,Legler JM,Warren JL. Development of a comorbidity index using physician claims data. J Clin Epidemiol 2000; 53: 12581267.
  • 20
    National Cancer Institute. Overview of the SEER Program [Internet]. Bethesda: National Cancer Institute; 2010. (accessed 27 August 2010).
  • 21
    National Cancer Institute. Overview of the SEER Program [Internet]. Bethesda: National Cancer Institute; 2010. (accessed 27 August 2010).
  • 22
    Charlson ME,Pompei P,Ales KL,MacKenzie CR. A new method of classifying prognostic comorbidity in longitudinal studies: Development and validation. J Chronic Dis 1987; 40: 373383.
  • 23
    Deyo RA,Cherkin DC,Ciol MA. Adapting a clinical comorbidity index for use with ICD-9-CM administrative databases. J Clin Epidemiol 1992; 45: 613619.
  • 24
    Romano PS,Roos LL,Luft HS, et al. A comparison of administrative versus clinical data: Coronary artery bypass surgery as an example. Ischemic Heart Disease Patient Outcomes Research Team. J Clin Epidemiol 1994; 47: 249260.

Robert Griffiths*, Michelle Gleeson*, Carolina Reyes†, Kevin Knopf‡, Mark Danese*, * Outcomes Insights, Inc., Westlake Village, California, † Genentech, Inc., South San Francisco, California, ‡ California Pacific Medical Center, San Francisco, California.