Comparative effectiveness and cost of adding rituximab to first-line chemotherapy for elderly patients diagnosed with diffuse large B-cell lymphoma

Authors


  • For this study, the linked Surveillance, Epidemiology, and End Results (SEER)-Medicare database was used. The interpretation and reporting of these data are the sole responsibility of the authors. The authors acknowledge the efforts of the Applied Research Program of the National Institute; the Office of Research, Development, and Information at the Centers for Medicare and Medicaid Services; Information Management Services, Inc.; and the SEER Program tumor registries in the creation of the SEER-Medicare database.

  • Editorial assistance on this article was provided by Kim Merjan and Suzanne Griffiths.

Abstract

BACKGROUND:

Clinical trials indicate that rituximab improves the survival of patients with diffuse large B-cell lymphoma (DLBCL). Economic models using multiple data sources, including clinical trials for survival outcomes, have projected cost offsets/savings and favorable cost-effectiveness associated with rituximab. In this study, the authors evaluated survival and cost impacts of adding rituximab to first-line chemotherapy for DLBCL using a single database that reflects routine clinical practice among elderly patients in the United States.

METHODS:

By using Surveillance, Epidemiology, and End Results (SEER) data linked to Medicare, the authors identified 5484 elderly patients who were diagnosed with DLBCL between January 1999 and December 2005 who had claims through December 2007. Included patients began chemotherapy with or without rituximab within 180 days of diagnosis. Multivariate analyses were conducted to estimate the impact of rituximab on mortality and costs to Medicare. The cost per life-year gained of rituximab was calculated using cost and survival estimates from the multivariate analyses.

RESULTS:

The mean patient age was 76 years, 43% of patients had stage III or IV disease, and 64% received rituximab. In a Cox regression model, rituximab resulted in lower 4-year all-cause mortality (hazard ratio [HR], 0.68; 95% confidence interval [CI], 0.61-0.74) and cancer mortality, and the incremental cumulative survival was 0.37 years. In least-squares regression, rituximab resulted in higher 4-year total costs ($23,097; 95% CI, $19,129-$27,298), immunochemotherapy costs ($12,069; 95% CI, $10,687-$13,634), other cancer costs ($7655; 95% CI, $5067-$10,489), and noncancer costs ($3461; 95% CI, $1319-$5650). The cost per life-year gained was $62,424.

CONCLUSIONS:

In routine clinical practice, rituximab was associated with survival benefits comparable to those observed in clinical trials. However, these benefits did not translate into the previously reported cost savings. Cancer 2012. © 2012 American Cancer Society.

INTRODUCTION

Diffuse large B-cell lymphoma (DLBCL) is the most common subtype of non-Hodgkin lymphoma (NHL), with an estimated prevalence of 129,500 and an incidence of 21,800 in the United States.1, 2 A disproportionate number of patients with DLBCL are aged ≥65 years, accounting for approximately 50% of prevalent cases and 59% of incident cases, and 58% of the incident cases are diagnosed in patients aged ≥75 years at diagnosis.1, 2 DLBCL is an aggressive subtype of NHL, and practice guidelines recommend induction chemotherapy plus the monoclonal antibody rituximab in most instances.3 It has been demonstrated in clinical trials of DLBCL4-7 that the addition of rituximab to first-line chemotherapy improves progression-free and overall survival in older patients, in which the reported risk ratios for all-cause mortality range from 0.53 (95% confidence interval [CI], 0.37-0.77)4 to 0.72 (95% CI, 0.52-1.00),5 and in younger patients.6 However, even in trials of older patients, the majority were aged <75 years4, 5; and, because older age is a prognostic factor for a poor outcome in aggressive lymphoma, the survival impact of rituximab in an older population in routine clinical practice reasonably may be expected to differ from that reported in those trials.

Like many countries, the United States is struggling with the financial burden associated with cancer and its treatment, and nowhere is this more apparent than within Medicare, where spending on Part B drugs, a category dominated by those used to treat cancer, rose 367% from $3 billion in 1997 to $11 billion in 2004.8 Rituximab also is approved for follicular lymphoma, chronic lymphocytic leukemia, and rheumatoid arthritis. In 2009, rituximab accounted for the largest percent of Medicare Part B expenditures (7.8%) of any drug administered in physicians' offices or furnished by suppliers.9 Efficacy notwithstanding, rituximab adds considerably to the cost of first-line therapy in NHL, with published estimates of $15,600 in follicular lymphoma10 and $17,200 in DLBCL.11 Given the added drug cost, there has been considerable interest in evaluating the net cost impact and cost-effectiveness of rituximab in DLBCL, taking into account its effects on progression-free and overall survival.11-15 Published estimates of the net cost impact range from −€700 in Italy,14 indicating that the cost savings associated with improved outcomes offset completely the treatment costs of rituximab, to +€15,900 in the Netherlands.12 The only study conducted in the United States projected a net cost impact of $12,374 over 6 years, which equated to cost offsets of $4491 because of lower salvage and end-of-life care.11 However, all of those projections were based on computer models that incorporated data from multiple sources, including clinical trials for survival data. Consequently, they may not reflect the cost impact of rituximab in routine clinical practice, in which practice patterns, patients, and survival outcomes may differ from those trial reports.

The objectives of the current study were to evaluate the survival, cost, and cost-effectiveness impacts of adding rituximab to first-line chemotherapy for DLBCL using a single source of data that reflects routine clinical practice in elderly patients and also the perspective of the largest single payer of medical care for DLBCL patients in the United States.

MATERIALS AND METHODS

Data Source

The source of data for this study was the National Cancer Institute (NCI) Surveillance, Epidemiology, and End Results (SEER) cancer registry linked to Medicare administrative and claims data.16 Currently, SEER contains cancer incidence and survival data from 17 population-based cancer registries throughout the United States that cover approximately 28% of the population.17 In the SEER-Medicare database, cancer registry data are linked to Medicare enrollment and claims data, which are available for 93% of individuals aged ≥65 years in the SEER registry.18 Medicare claims include hospital short-stay and long-stay, skilled nursing facility, physician/supplier, institutional outpatient, home health agency, and durable medical equipment.

Inclusion and Exclusion Criteria

Patients were included if they were diagnosed with DLBCL19 between January 1, 1999 and December 31, 2005, if DLBCL was the first primary cancer diagnosed, and if their first Medicare claim for immunochemotherapy was ≤180 days after diagnosis. To ensure complete claims history, patients had to be enrolled in a Medicare fee-for-service plan, with no health maintenance organization (HMO) coverage, for 12 months before DLBCL diagnosis. Patients were excluded for the following reasons: diagnosis before age 65 years, diagnosis made by death certificate or autopsy, death within the first month after diagnosis, or Medicare enrollment <12 months before diagnosis.

Patients and Variables

Patients were described according to their demographic, clinical, and socioeconomic characteristics. Requiring eligible patients to have at least 1 year of Medicare enrollment before diagnosis meant that the minimum age in the cohort was 66 years. Patients were classified by disease stage and extranodal involvement (yes/no).

Medicare claims do not contain the results of laboratory tests, including lactate dehydrogenase (LDH) level, which is an independent prognostic factor for overall survival in the International Prognostic Index (IPI) for DLBCL.20 Another prognostic factor for overall survival in the IPI that is not included in SEER-Medicare is Eastern Cooperative Oncology Group (ECOG) performance status.21 In the absence of performance status, we used Medicare claims to identify several predictors of poor performance status,22 including the use of oxygen and related respiratory therapy supplies, wheelchair and supplies, home health agency, and skilled nursing facility, all from 12 months before until 30 days after DLBCL diagnosis. Individual services were combined into a score of 0 (none) or 1 (use of any service). Medicare claims also were used to identify anemia23 and to calculate an NCI comorbidity index score for each patient.24, 25

SEER-Medicare contains information from the 2000 Census, reported at the tract level in which the patient lives, for the percentage of the population living in poverty and the percentage of those aged ≥25 years with some college. We used these as indicators of the socioeconomic status of individual patients in the cohort. The assigned metropolitan statistical area was used as a geographic indicator.

First-Line Therapy

We used the Medicare outpatient and physician/supplier files to identify all claims containing Healthcare Common Procedure Coding System (HCPCS) “J” codes for chemotherapy (J9000-J9999) or rituximab (J9310) within 30 days after the first such claim.26, 27 Patients were then classified as having received chemotherapy with or without rituximab as first-line therapy on the basis of these claims. In addition, we identified specific chemotherapy agents used, consisting of cyclophosphamide (J8530, J9070, J9080, J9090-J9097), doxorubicin (J9000, J9001), vincristine (J9370, J9375, J9380), and mitoxantrone (J9293), according to treatment guidelines.3 These claims were used to classify chemotherapy regimens; eg, cyclophosphamide, doxorubicin, vincristine, and prednisone (CHOP); cyclophosphamide, mitoxantrone, vincristine, and prednisone (CNOP); or cyclophosphamide, vincristine, and prednisone (CVP). Medicare did not begin covering oral therapies without an intravenous equivalent until January 2006. In the absence of data on oral therapy, the use of prednisone was assumed when the other agents were present.

Follow-Up and Survival

Patients were followed from the start of first-line therapy until death, the end of their claims (December 31, 2007), or until they switched to HMO coverage, whichever came first. Cause of death was classified as death caused by cancer of any type or noncancer causes. Noncancer causes included all other identified causes of death, such as heart disease or diabetes. It excluded missing or unspecified cause of death. These patients were censored at the time of death in both cancer and noncancer survival analyses.

Cost

Direct medical costs to Medicare were calculated using paid amounts from claims. Diagnosis and procedure codes within claims were used to identify 3 mutually exclusive categories of costs: immunochemotherapy,26, 27 other cancer care (any claim with an International Classification of Diseases, Ninth Revision, Clinical Modification [ICD-9-CM] diagnosis code from 140.xx to 208.xx [malignant neoplasms], but no code for immunochemotherapy), and noncancer care (any other claim). Paid amounts were inflated to 2009 US dollars.28

Statistical Analysis

Multivariate survival analysis was performed using Cox proportional hazards regression. Patients who received rituximab tended to be diagnosed in later years. Therefore, the primary survival analysis was limited to 4 years. However, additional survival analyses were conducted using the entire observation period, with cancer and noncancer mortality as the outcomes, and restricting the cohort to those who received CHOP or CNOP, because the intent of treating with CHOP may have differed from treating with CVP or other regimens. All survival analyses were then repeated using propensity score analysis,29 which entailed using propensity score quintile in lieu of all other independent variables except treatment in the survival analyses. Finally, we constructed an adjusted Kaplan-Meier curve of the difference in survival between the 2 first-line treatment groups using inverse probability of treatment weighting.30

We performed partitioned, inverse probability-weighted (IPW), least-squares regression analysis31, 32 to examine adjusted associations between cumulative costs over 4 years (48 partitions) and patient demographic, clinical, and treatment factors. Separate regression analysis was performed for each of the partitions. Coefficients for each patient factor were summed across the partitions to obtain the cumulative, incremental cost associated with that factor. Confidence intervals for the cumulative cost coefficients were calculated using a bootstrap approach,33 in which the process of performing each partitioned regression analysis and summing coefficients across partitions was repeated 1000 times using sampling with replacement from the original cohort. The cost analysis was limited to 4 years to match the time horizon for the primary survival analysis and because excessive differential censoring by treatment group resulting from comparing a newer to an older treatment can result in biased regression coefficients.

The cost per life-year gained from rituximab plus chemotherapy compared with chemotherapy alone was calculated by dividing the total incremental cumulative cost in the rituximab group by the total incremental cumulative survival at 48 months after the beginning of first-line treatment. In addition to performing analyses across the entire cohort, we then stratified the cohort by ages 66 to ≤80 years versus age >80 years at diagnosis, and we repeated the multivariate analyses of 4-year total cost and survival.

RESULTS

There were 5484 patients in the final cohort: 3531 (64%) received rituximab plus chemotherapy, and 1953 (36%) received chemotherapy alone (Table 1). The mean age at diagnosis was 76 years, 56% were aged >75 years, 43% had stage III/IV disease, 37% had extranodal involvement, and 22% had ≥1 indicator of poor performance status. Patients who received rituximab were diagnosed later in the study period and were less likely to have immunoblastic histology or ≥1 indicator of poor performance.

Table 1. Patient Characteristics
 First-Line Therapy 
 Rituximab Plus Chemotherapy, n = 3531Chemotherapy Alone, n = 1953 
VariableNo. of Patients%No. of Patients%P
  1. Abbreviations: DLBCL, diffuse large B-cell lymphoma; NCI, National Cancer Institute; SD, standard deviation.

Age at diagnosis, y     
 66-7060817.234317.6.87
 71-7596127.251326.3 
 76-8093426.552927.1 
 >80102829.156829.1 
 Mean ± SD75.9±6.2 76.1±6.2 .35
Sex     
 Men165146.889946.61
 Women188053.2105454 
Race/ethnicity     
 White304986.3169386.7.52
 Black1022.9623.2 
 Hispanic1955.51125.7 
 Other1855.2864.4 
Year of diagnosis     
 1999-2002110931.4168186.1<.0001
 2003-2005142268.627213.9 
Stage at DLBCL diagnosis
 I101628.862532.15
 II75021.238919.9 
 III58016.429815.3 
 IV97227.552727 
 Unknown21361145.8 
Histology     
 Centroblastic346298186995.7<.0001
 Immunoblastic692844.3 
Extranodal involvement     
 No131537.273437.6 
 Yes200356.7110556.6.94
 Unknown21361145.8 
NCI comorbidity index     
 0217261.5123263.1.42
 186324.447724.4 
 23128.81517.7 
 ≥31845.2934.8 
Performance status indicators
 0279079147575.5<.01
 ≥17412147824.5 

Treatment

Overall, 82% of patients (4510) received CHOP or CNOP as first-line chemotherapy, and the proportions were similar in the rituximab plus chemotherapy (83%) and chemotherapy alone (81%) groups. In the rituximab group, the mean number of rituximab administrations billed to Medicare during first-line therapy was 5.5 (median, 5 administrations; interquartile range [IQR], 4-7 administrations), and the average number of units (1 unit, 1 × 100 mg vial) per administration was 6.9 (median, 7 units; IQR, 6-8 units).

Survival

There were 2815 deaths, of which 1649 (58.6%) were classified as cancer, 1044 (37.1%) were classified as noncancer, and 122 (4.3%), were classified as missing or unknown. The median (IQR) survival was 66 months (IQR, 61-75 months) in the rituximab group, compared with 32 months (IQR, 28-37 months) for chemotherapy alone. Kaplan-Meier analysis indicated similar incremental survival benefits of rituximab for patients who received CHOP or CNOP and those who received other chemotherapy regimens, primarily CVP (Table 2, Fig. 1).

Table 2. Kaplan-Meier Analysis of Overall Survival
 No. at Risk
Chemotherapy Regimen1 Months6 Months12 Months18 Months24 Months30 Months36 Months42 Months48 Months
  1. Abbreviations: CHOP, cyclophosphamide, doxorubicin, vincristine, and prednisone; CNOP, cyclophosphamide, mitoxantrone, vincristine, and prednisone; R, rituximab.

CHOP/CNOP155913061101987888839779727677
R-CHOP/CNOP29022548223220431910159813121098859
Other3652672041631431231068978
R-other594485398351323258199147117
Figure 1.

This chart illustrates a Kaplan-Meier analysis of overall survival. CHOP indicates cyclophosphamide, doxorubicin, vincristine, and prednisone;. CNOP, cyclophosphamide, mitoxantrone, vincristine, and prednisone; R, rituximab.

In multivariate survival analysis, rituximab was associated with significantly lower all-cause mortality at 4 years (hazard ratio [HR], 0.68; 95% CI, 0.61-0.74; P < .0001) (Table 3). In the multivariate survival analysis stratified by age, rituximab was associated with lower all-cause mortality in patients aged ≤80 years (HR, 0.67; 95% CI, 0.60-0.75) and in patients aged >80 years (HR, 0.72; 95% CI, 0.62-0.84), with considerable overlap in the CIs. The cumulative incremental survival associated with rituximab was 0.37 years after 4 years. Women and patients with immunoblastic histology also were associated with lower mortality (P ≤ .05). Older age, stage III/IV disease, extranodal involvement, the presence of “B” symptoms, anemia, an NCI comorbidity index >0, and ≥1 indicator of poor performance were associated with higher mortality. In the analysis of cumulative survival stratified by age, rituximab was associated with an incremental survival benefit of 0.37 years in patients aged ≤80 and 0.41 years in patients aged >80 years, a difference of approximately 15 days between the 2 age groups.

Table 3. Multivariate Survival Analysis
VariableHR95% CIP
  1. Abbreviations: CI, confidence interval; DLBCL, diffuse large B-cell lymphoma; HR, hazard ratio; NCI, National Cancer Institute; Ref, reference category.

First-line therapy   
 Chemotherapy aloneRef  
 Rituximab plus chemotherapy0.680.61-0.74<.0001
Age at diagnosis, y   
 66-70Ref  
 71-751.421.23-1.64<.0001
 76-801.581.37-1.82<.0001
 >802.292.00-2.63<.0001
Sex   
 MenRef  
 Women0.820.76-0.89<.0001
Race/ethnicity   
 WhiteRef  
 Black1.040.83-1.31.75
 Hispanic0.930.77-1.11.42
 Other1.020.85-1.23.84
Year of diagnosis   
 1999-2002Ref  
 2003-20051.000.91-1.10.98
Stage at DLCBL diagnosis   
 I & IIRef  
 III & IV1.611.47-1.75<.0001
Histology   
 CentroblasticRef  
 Immunoblastic0.670.54-0.83<.001
Extranodal involvement   
 NoRef  
 Yes1.121.02-1.22.02
Presence of “B” symptoms   
 NoRef  
 Yes1.341.20-1.49<.0001
Anemia   
 NoRef  
 Yes1.211.08-1.36<.01
NCI comorbidity index   
 0Ref  
 11.141.04-1.260.01
 21.441.25-1.65<.0001
 ≥31.841.56-2.17<.0001
Performance status indicators   
 NoneRef  
 ≥11.451.32-1.60<.0001

Results from multivariate survival analyses using the entire observation period and using propensity score analysis (Fig. 2) were comparable to the primary analyses. Also, rituximab was associated with significantly lower cancer mortality, but not noncancer mortality, using both standard multivariate and propensity score analysis (Fig. 2). Results of the multivariate survival analysis that was restricted to patients who received CHOP or CNOP (n = 4510) were similar to those that included all patients (results not shown).

Figure 2.

Multivariate (MV) survival analyses are illustrated from a sensitivity analysis of hazard ratios for rituximab plus chemotherapy. These are results from 3 sets (all-cause mortality, cancer mortality, and noncancer mortality) of 4 multivariate survival analyses (2 for survival during the entire observation period and 2 for survival during the first 4 years of follow-up) that were designed to test the sensitivity of the findings reported in Table 3 to changes in the specification of the outcome variable as well as the approach to multivariate analysis. Standard multivariate survival analyses were performed with all individual patient variables included in the model. Propensity multivariate survival analyses were performed with propensity score quintile included in the model as a substitute for all patient variables except treatment. An asterisk indicates that the full results from this analysis are reported in Table 3. The y-axis indicates the hazard ratio for rituximab plus chemotherapy compared with chemotherapy alone. Triangles represent the estimated hazard ratio for rituximab plus chemotherapy compared with chemotherapy alone from the corresponding model on the x-axis. Bars around each triangle represent the upper and lower bounds of the 95% confidence interval for the hazard ratio. Confidence intervals that overlap the horizontal line at a hazard ratio of 1.0 indicate that the estimated hazard ratio for rituximab plus chemotherapy is not significant at P = .05.

Costs

The mean (±standard deviation) cost of first-line therapy was $24,431 ± $21,379 in the rituximab group and $13,230 ± $18,635 for the chemotherapy alone group. In multivariate analysis, rituximab was associated with significantly higher 4-year total costs ($23,097; 95% CI, $19,129-$27,298), immunochemotherapy costs ($12,069; 95% CI, $10,687-$13,634), other cancer costs ($7655; 95% CI, $5067-$10,489), and noncancer costs ($3461; 95% CI $1319-$5650), and immunochemotherapy (first-line plus subsequent) accounted for approximately 53% of the total. In the rituximab group, 65% ($14,918/$23,097) of the total incremental cost accrued during the first 6 months (Fig. 3). This was driven primarily by the cost of immunochemotherapy and, secondarily, by the costs of other cancer-related services. Noncancer costs in the rituximab group became statistically significantly higher only after month 35, consistent with improved survival.

Figure 3.

The incremental cost of rituximab plus chemotherapy is illustrated. This chart presents results from the multivariate analysis of cumulative incremental costs associated with rituximab plus chemotherapy compared with chemotherapy alone. CI indicates confidence interval.

Results from the multivariate analysis of total 4-year costs that included only patients who received CHOP or CNOP were similar to those from the analysis that included all patients (cost coefficient for rituximab plus chemotherapy, $24,451; 95% CI $20,015-$28,876). In the multivariate analysis stratified by age, rituximab was associated with higher total 4-year costs in patients aged ≤80 years ($24,109; 95% CI $19,163-$29,321) and in patients aged >80 years ($21,523; 95% CI $14,625-$28,029), with considerable overlap in the CIs.

Cost per Life-Year Gained

On the basis of the multivariate survival and cost analyses, the cost per life-year gained of rituximab plus chemotherapy compared with chemotherapy alone was $62,424 ($23,097 of 0.37 life-years) over the 4-year time horizon.

DISCUSSION

Clinical trials demonstrated that first-line rituximab plus chemotherapy resulted in improved survival for patients with DLBCL.4-7 In our study, the HR for all-cause mortality at 4 years was 0.68, compared with a range from 0.534 to 0.725 reported in clinical trials of older patients, although the patients in our study were considerably older than the patients in those trials. For example, in the phase 3 trial reported by Coiffier et al,4 only 20% of patients were aged ≥75 years, compared with >56% of the Medicare patients in our study. Our results were robust to changing the length of the observation period, restricting to patients who received CHOP or CNOP, and using propensity score techniques. Also, HRs were similar for those aged ≤80 years and those aged >80 years at diagnosis, indicating that the survival benefits observed in clinical trials also are realized in the very elderly patients in routine clinical practice. SEER does not include information on cancer relapse. Therefore, we did not examine relapse-free or progression-free survival.

Several economic analyses of first-line rituximab for DLBCL have projected cost offsets or outright savings related to improved clinical outcomes.11-15 Most were based on computer models that incorporated data from multiple sources, including survival data from clinical trials. On the basis of the published economic literature, in the rituximab group, we expected to observe higher immunochemotherapy costs during first-line therapy, followed by lower costs after first-line therapy because of lower rates of second-line treatment for relapse. In our study, immunochemotherapy costs did increase during the first 6 months, but to a lower level than expected based on the cost of rituximab itself. One possible explanation is that the mean number of rituximab administrations was lower (5.5 administrations) than expected6-8 based on clinical practice guidelines.3 From 7 months to 48 months, the incremental costs of immunochemotherapy were approximately zero, suggesting that any underlying differences in progression/relapse rates did not have an impact on the costs of second-line therapy. We also expected to observe lower incremental costs associated with other types of cancer care and noncancer care. Instead, we observed that the costs of other cancer care rose steeply during the initial treatment phase and then continued to rise, albeit at a slower rate, after the end of first-line therapy. A possible explanation for the early rise is that, because rituximab use was more common later in the study period, the rituximab coefficient picked up the costs of any supportive care (eg, granulocyte-colony–stimulating factor) that was more likely to be used later in the study period, although we adjusted for year of diagnosis in all the multivariate analyses. Indeed, during the first 6 months, the most frequent claims for other cancer services were for infused therapies, such as corticosteroids, granulocyte-colony–stimulating factors, erythropoietin, and antiemetics. Also, because the assignment of costs to these categories was based on diagnosis and procedure codes in claims, it is possible that the diagnosis of cancer was “carried forward” on claims for services unrelated to cancer. We observed a gradual increase in the incremental cost of noncancer care in the rituximab group, until the cumulative difference became statistically significant in year 4. This is consistent with the finding that patients who received rituximab survived longer; and, in an elderly population, cancer survivors would continue to incur medical costs related to other conditions. Incremental total costs of adding rituximab to chemotherapy were lower in those aged >80 years compared with those aged ≤80 years at diagnosis, although incremental survival benefits were similar. This may have been caused by differences in the rituximab regimen between the 2 groups and/or to the finding that overall expected survival is shorter in the very elderly.

Like any study that uses observational data to examine the outcomes of therapies or other interventions, in this study, the main threat to the validity of our findings is selection bias, in which unobserved factors influence both treatment selection and the outcome of interest.34 SEER-Medicare does not contain data on LDH or ECOG performance status—2 of the 5 prognostic factors for DLBCL overall survival that are included in the IPI.20 In lieu of ECOG performance status, we included a claims-based indicator of poor performance status.22 In the multivariate survival and cost analyses, we did observe that the presence of ≥1 indicator of poor performance was associated with a 45% higher risk of mortality and with $4774 higher total cost. Also, we observed significant differences in survival and/or cost according to disease stage, histology, extranodal involvement, the presence of anemia, and NCI comorbidity index. It is important to note that any variable constructed using claims for medical services, such as performance status and NCI comorbidity index, that is included in a multivariate cost analysis has a greater likelihood of being statistically significant simply by virtue of the finding that most claims have a cost attached.

In the current study, patients who were diagnosed in later years were more likely to receive rituximab. This may have introduced several possible sources of bias, including differential observation and censoring and secular trends in both the assignment of patients to different chemotherapy regimens and in the use of supportive care. In the survival and cost analyses, we limited the observation period to reduce differential censoring. We used IPW in the estimation of cumulative incremental survival. In the cost analysis, we used IPW and established separate weights for the 2 treatment groups to account for differential censoring between the 2 groups. Also, we included year of diagnosis as a covariate in all of the survival and cost models to adjust for temporal differences in patterns of care.

Most patients in both first-line treatment groups received CHOP chemotherapy. Therefore, we did not include type of chemotherapy in the primary analyses. When we restricted the cohort to those who received CHOP or CNOP, the survival and cost coefficients for the rituximab groups were similar to those in the primary analyses. One indicator of selection bias in survival analysis is treatment differences in noncancer survival that are as large as, or larger than, differences in cancer survival.34 In the analysis of cancer and noncancer mortality, we observed that the survival benefit was limited to cancer mortality.

In routine clinical practice, the comparative effectiveness of adding rituximab to first-line chemotherapy for elderly patients who are diagnosed with DLBCL is similar to its efficacy as measured in randomized trials. This applied to both elderly and very elderly patients. However, the survival benefits observed in this study did not translate into cost offsets or savings as previously reported by several studies in the peer-reviewed literature. Although these findings confirm the clinical benefits of rituximab, they suggest that it may not be realistic to expect effective new cancer therapies to attenuate the rising costs of cancer care in routine clinical practice, especially for populations in which multiple comorbidities are common.

FUNDING SOURCES

No specific funding was disclosed.

CONFLICT OF INTEREST DISCLOSURES

Robert Griffiths, Michelle Gleeson, and Mark Danese are employees of for Outcomes Insights, Inc. Outcomes Insights, Inc. has received funding from Amgen and Genentech to conduct research in non-Hodgkin lymphoma, including on rituximab.

Ancillary