SEARCH

SEARCH BY CITATION

Keywords:

  • radiation therapy;
  • cardiotoxicity;
  • doxorubicin;
  • non-Hodgkin lymphoma;
  • elderly;
  • Surveillance;
  • Epidemiology;
  • End Results-Medicare data base

Abstract

  1. Top of page
  2. Abstract
  3. MATERIALS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. Acknowledgements
  7. REFERENCES

BACKGROUND.

In the past 25 years, clinical trials have demonstrated the benefits of chemotherapy for patients with aggressive non-Hodgkin lymphoma. The authors analyzed the predictors and outcomes of chemotherapy among elderly patients with lymphoma.

METHODS.

Patients age ≥65 years who were diagnosed with Stage III and IV diffuse large B-cell lymphoma [according to the SEER Summary Staging Manual, 2000] between 1991 and 1999 in the Surveillance, Epidemiology, and End Results-Medicare data base were categorized by treatment: no chemotherapy, a doxorubicin-containing regimen, a regimen without doxorubicin, or chemotherapy not otherwise specified. Among the patients who survived for >6 weeks after diagnosis and who had a chemotherapy regimen specified, logistic regression analysis was used to identify predictors of doxorubicin-based treatment, and Cox proportional-hazards regression was used to analyze outcomes.

RESULTS.

Less than 66% of patients received any chemotherapy in the 6 months after diagnosis, and 42% of untreated patients died within 6 weeks. Older age, congestive heart failure, and other comorbidities were strong predictors of treatment without doxorubicin. From 1991 to 1999, the proportion of patients who received doxorubicin increased from <20% to >50%. Patients who received doxorubicin survived more than twice as long (24.4 months) as patients who did not receive doxorubicin (11.2 months). Survival was no better among patients who received chemotherapy without doxorubicin than among patients who received no chemotherapy.

CONCLUSIONS.

By 1999, doxorubicin-based chemotherapy had gained general acceptance for use among the elderly, although nearly 50% of elderly patients still were not receiving it. Given the clinical trial-based evidence of its benefits, in the absence of specific contraindications, most patients, including the elderly, should be treated with regimens that include doxorubicin. Cancer 2006. © 2006 American Cancer Society.

Over the past 25 years, clinical research has improved dramatically the prognosis for patients who are diagnosed with aggressive non-Hodgkin lymphoma. Less than 50% of patients with diffuse large cell disease who were enrolled in recent Phase III, doxorubicin-based chemotherapy trials developed recurrent disease.1 The major changes in therapy began in the 1970s with the introduction of chemotherapy that combined cyclophosphamide, vincristine, and prednisone with doxorubicin (CHOP).2 In 1992 and 1993, it was demonstrated in randomized trials3, 4 that CHOP had as much efficacy as, and less toxicity than, the best of the other multidrug regimens that were in use at that time. For the next decade, CHOP represented the standard of care. In the most recent trials, it has been observed that the addition of rituximab to CHOP (R-CHOP) improved patient outcomes even more; now, R-CHOP is the latest standard of care.5, 6

Individuals age >60 years comprise a significant number of the patients who are diagnosed with diffuse large cell lymphoma in the U.S. each year.6, 7 Randomized trials have shown that patients in this age group respond to chemotherapy as well as younger patients, although to our knowledge little is known about the patterns of use and the effectiveness of chemotherapy among elderly patients in the general population.5 Recent studies among elderly patients have demonstrated that age bias may interfere with their receiving optimal cancer therapy.5, 8–10

In 1993, the National Cancer Institute's Surveillance, Epidemiology, and End Results (SEER) cancer registry data base11 was linked to the Medicare data base of the Center for Medicare and Medicaid Services.11 This linkage made it possible to analyze data on cancer treatment in relation to both demographic and clinical factors and cancer outcomes. Studies using the SEER-Medicare data base have shown that the use of state-of-the-art chemotherapy for several cancers may be associated with patient age, gender, race/ethnicity, or socioeconomic status.12–14 The objective of this study was to describe the predictors of treatment for patients with Stage III or IV [according to the SEER Summary Staging Manual, 2000], diffuse large B-cell lymphoma and the association of treatment with survival among elderly patients in the SEER-Medicare data base.

MATERIALS AND METHODS

  1. Top of page
  2. Abstract
  3. MATERIALS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. Acknowledgements
  7. REFERENCES

Data Source

We analyzed patient data from the SEER data base linked with diagnostic and procedural billing claims from Medicare.11 The SEER data base, based on the November 2004 submission,7 contains records of patients who were diagnosed with cancer in regions that represent approximately 14% of the U.S. population. SEER provides information on tumor histology, location, disease stage, treatment, and survival along with selected Census tract level demographic information. The Medicare data base includes Medicare A and B eligibility status, dates of enrollment in health maintenance organizations (HMOs), and billed claims, including inpatient and outpatient services, procedures, and diagnoses. We supplemented those data with details from the American Medical Association master file about the hospitals from which the Medicare inpatient and outpatient claims originated.

Sample Selection

We identified all individuals age ≥65 years who received a pathologically confirmed primary diagnosis of advanced, diffuse large B-cell lymphoma (International Classification of Diseases, 9th Revision [ICD-9] codes 9593, 9680, 9681, 9682, and 9683) from January 1, 1991 to December 31, 1999 (n = 2326 patients). We excluded patients who were enrolled in an HMO from 12 months before to 12 months after diagnosis and who were not covered by Medicare Parts A and B over the same period (n = 959 patients).

Patient Characteristics

The SEER data base includes data on age, race/ethnicity, gender, marital status, type of hospital, and area of residence. Age at diagnosis was broken down into pentads (5-year discrete numeric intervals) starting at 65 years. Because the sample included very few patients who were neither black nor white, we designated race as black, white, or other. For similar reasons, we recoded the SEER marital status variable into married, not married, and unknown. Using the SEER-American Medical Association-linked hospital file, we designated the hospitals where patients were diagnosed and/or treated as teaching or nonteaching hospitals. We generated an aggregate socioeconomic status variable with a range of values from 0 to 4 for each patients based on income data from the 2000 Census and an algorithm described by Bach et al.15 using the median income in the patient's Census tract of residence, the zip code median income, the Census tract per capita income, and the zip code per capita income. Patients for whom all of these values were missing were assigned to the lowest socioeconomic category.

Comorbid Disease

We computed a comorbidity score for each patient using the Klabunde adaptation of the Charlson comorbidity index.15, 16 Medicare inpatient and outpatient claims were searched from 365 days before to 120 days after the diagnosis of cancer for all ICD-9 Clinical Modification (ICD-9-CM) diagnostic codes corresponding to each of the following comorbid conditions: myocardial infarction, congestive heart failure, peripheral vascular disease, cerebrovascular disease, dementia, chronic pulmonary disease, connective tissue disease, peptic ulcer disease, mild-to-severe liver disease, diabetes with or without end-organ damage, hemiplegia, moderate or severe renal disease, or acquired immunodeficiency syndrome. Each disease category was weighted based on the Charlson index, and patients were assigned a score of 0, 1, or >1. We also generated an indicator variable for a history of congestive heart failure.

Treatment Characteristics

We extracted information on chemotherapy from the Medicare files by searching Level II Health Care Procedure Coding System (HCPCS) codes (J9XXX, Q0083-Q0085), Current Procedural Terminology codes (964XX, 965XX), ICD-9-CM diagnostic codes (V581, V662, V672, E9331, E9307), and procedure codes (9925), Diagnostic Related Grouping (DRG) code (410), and Revenue Center (CEN) codes (0331, 0332, 0335) from physician claims files (National Claims History), hospital outpatient claims files (Output), or Medicare provider review files (Medpar). For evidence of chemotherapy delivery during the first 6 months after diagnosis, we searched for HCPCS codes that corresponded to the following specific agents: doxorubicin (J9000), liposomal doxorubicin (J9001), mitoxantrone (J9293), cyclophosphamide (J8530, J9070-J9097), and vincristine (J9370, J9375, J9380). We also identified patients who had received other and unspecified chemotherapeutic drugs during the same 6-month period.

The validity of SEER-Medicare claims data for chemotherapy use has been described previously.17, 18 SEER does not provide valid data on chemotherapy, because such data often are unavailable to cancer registrars. Medicare is informative about diagnostic procedures and treatments for which hospitals and physicians bill, including, for cancer patients, intravenous and injectable chemotherapy. For our analysis, we grouped together patients who received any doxorubicin; we created a separate category for patients who received specific chemotherapeutic drugs but not doxorubicin. We classified patients as having received chemotherapy not otherwise specified if their records included a charge for administering chemotherapy with no designation of a particular medication. Some of these charges may represent supervision of oral medication.

Before 2006, Medicare did not cover oral medication, but most chemotherapy regimens for lymphoma include oral steroids. We assume that nearly all patients in our study received prednisone along with their other chemotherapeutic drugs; however, because we lacked specific data about treatment with prednisone, we did not identify our patients as having received CHOP or combined cyclophosphamide, vincristine, and prednisone.

We searched for data on radiation therapy in the SEER data base and for ICD-9-CM codes (V580, V661,V671, 9220-9229), HCPCS codes (77400-77490, 77500-77797), DRG code (409) and revenue center codes (0330, 0333, 0339) within 180 days after diagnosis in the Medicare records.

Mortality and Survival

Survival was calculated in months from the date of lymphoma diagnosis to the date of death according to the SEER data. Survival was censored as of the last Medicare coverage month when patients were known to be alive or as of December 31, 2000. Cause-specific mortality was assessed by using ICD8-10 codes for cause of death in the SEER data.

Statistical Analysis

Univariate associations between the type of chemotherapy and clinical and demographic factors were assessed by using chi-square tests. Among the patients we identified who had received specific chemotherapeutic drugs, we used multivariate logistic regression models to analyze the association of doxorubicin-based chemotherapy with age at diagnosis, year of diagnosis, race, urban/rural residence, treatment in a teaching or nonteaching hospital, and comorbidity score.

Because patients who are believed to be in imminent danger of death often are untreated, we modeled the association of death in the first 6 weeks after diagnosis with the same factors. We also examined mortality by treatment at 12 weeks and 26 weeks after diagnosis. Among patients who survived beyond 6 weeks, we developed Cox proportional hazards models to analyze the association of treatment and other covariates with cause-specific and overall mortality. We also calculated survival rates and generated survival curves for each of the treatment groups by using Kaplan–Meier analyses.

Propensity scoring by means of a logistic regression model was used to analyze the association of receiving doxorubicin-based chemotherapy with demographic and clinical factors. The propensity score was the probability of receiving treatment with doxorubicin given a patient's values for the variables in the logistic regression model. Patients were grouped into quintiles based on their propensity scores; 5 strata are considered adequate to control for >90% of the bias caused by each of the predictors.19–21 Within each stratum, patients theoretically are equivalent with respect to their propensity to have received treatment. The results of these analyses were similar to those of the standard proportional hazards analyses; we present the results of the latter, including the rate ratios for the individual covariates.

The proportionality of the model was validated by using Martingale residual plots. Cox regression was used to assess interactions between treatment and the other covariates in our models.

Sensitivity Analysis

In nonrandomized studies, an observed association between treatment and outcome may reflect the effects of unknown or unmeasured confounders. We conducted a sensitivity analysis of the effects of an unmeasured binary confounder on the estimated propensity score-adjusted, disease-specific mortality hazard ratio for doxorubicin treatment versus nondoxorubicin treatment.22, 23 In our analysis, we varied both the prevalence of the unknown confounder in the groups treated with and without chemotherapy and the mortality hazard ratio for the confounder. All statistical analyses were performed using SAS software (version 9.1; SAS Institute, Cary, NC) and R software (version 1.1.1; The R Foundation for Statistical Computing. Vienna, Austria).

RESULTS

  1. Top of page
  2. Abstract
  3. MATERIALS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. Acknowledgements
  7. REFERENCES

We identified 2326 patients who were diagnosed with Stage III or IV diffuse large B-cell lymphoma from 1991 through 1999. Of these patients, nearly 33% received doxorubicin-based chemotherapy within 6 months after their diagnosis, and slightly greater than 33% received no chemotherapy (Table 1). From 1991 to 1999, the proportion of patients who received treatment with doxorubicin increased from 19% to >50%, whereas the proportion of patients who received treatment with nondoxorubicin-based chemotherapy declined from 23.6% to 12.1%. The proportion of patients who did not receive chemotherapy declined only slightly, from 34.7% to 32.0%, but the proportion who received chemotherapy that was not otherwise specified fell from 23.6% to only 5.4% (Table 1).

Table 1. Demographic and Clinical Characteristics of Patients with Advanced Diffuse Large B-Cell Lymphoma by Chemotherapy Group
CharacteristicTreatment received
Specified chemotherapyChemotherapy NOSNo chemotherapyTotalP
With doxorubicinWithout doxorubicin
No.%No.%No.%No.%No.%
  • NOS indicates not otherwise specified.

  • *

    The comorbidity score was based on the Charlson comorbidity index.

  • Socioeconomic status was determined for each participant on a scale from 0 to 4 based on income data from the 2000 Census and an algorithm described by Bach et al., 2002.15

All patients76833.046820.126111.282935.72326100 
Age, y          <.001
 65–6917422.79019.24517.2779.338616.6 
 70–7425132.711023.56023.018822.760926.2 
 75–7920626.813829.57729.519623.661726.5 
 80–8410513.68718.64517.317120.640817.5 
 ≥85324.2439.23413.019723.830613.2 
Year of diagnosis455.85612.05420.7829.923710.2<.001
 1992729.46614.13613.810012.127411.8 
 1993729.46513.9249.210312.426411.4 
 1994759.85612.03312.68410.124810.7 
 1995769.96113.03613.89711.727011.6 
 19968410.95612.0218.19911.926011.2 
 199711314.7418.72710.38310.026411.3 
 199810213.3367.7166.19912.025310.8 
 199912916.8316.6145.4829.925611.0 
Male gender32542.322147.212046.037745.5104344.8.35
Race          .34
 White69290.142290.223188.572887.8207389.1 
 Black263.4163.4166.1425.11004.3 
 Other506.5306.4145.4597.11536.6 
Urban residence68188.740787.021783.173488.5203987.7.09
Marital status          <.001
 Married48062.528460.714455.239647.8130456.1 
 Not married27135.317637.611544.041650.297842.0 
 Unknown172.281.720.8172.0441.9 
Comorbidity score*           
 052268.027759.216663.643051.9139560.0 
 117522.812626.95320.320724.956124.1 
 > 1719.26513.94216.119223.237015.9 
Comorbidity score without congestive heart failure*          <.001
 058776.434172.920678.956667.5169472.8 
 114518.99319.94416.917721.445919.7 
 > 1364.7347.2114.29211.11737.4 
Congestive heart failure20626.815934.07930.325130.369529.9.06
Socioeconomic status          .01
 013217.28317.85621.419523.546620.0 
 114018.210322.05420.717320.947020.2 
 215720.49620.55019.214918.045219.4 
 319225.09019.25019.215018.148220.7 
 414719.19620.55119.516219.545619.6 
Teaching hospital14619.08017.13212.311714.137516.1.02
Radiation therapy15720.49921.27026.817120.649721.4.15
Rituximab202.620.420.800241.0< .001

Among all patients in our data base, 21% received radiation therapy. Radiation therapy was not associated with chemotherapy group. Only 24 patients (1%) received rituximab, and 20 of those patients also received doxorubicin.

Among patients who received no chemotherapy, 42% died within 6 weeks after diagnosis, compared with <3% of patients who received any chemotherapy and <1% of patients who received doxorubicin (Table 2). At 12 weeks, 4.7% of patients on doxorubicin died compared with 57.7% of patient who received no chemotherapy. Mortality within 6 weeks was associated with older age, especially older than 80 years, having a comorbidity score >1, and not receiving radiation therapy (Table 3).

Table 2. Cumulative Mortality by Chemotherapy Group and Number of Weeks after Diagnosis
WeekSpecified chemotherapyChemotherapy NOSNo chemotherapyP
With doxorubicinWithout doxorubicin
No.%No.%No.%No.%
  1. NOS indicates not otherwise specified.

100.000.010.4688.2< .001
200.010.241.513616.4 
300.020.4124.620725.0 
400.020.4249.226331.7 
520.351.13513.431137.5 
670.971.54818.434942.1 
12364.7347.38432.247857.7 
2611915.514531.012246.757267.8 
Table 3. Multivariate Adjusted Odds Ratios for the Association of Death in the First 6 Weeks after Diagnosis with Clinical and Demographic Characteristics
VariableOR95% CI
  • OR indicates odds ratio; 95% CI, 95% confidence interval.

  • *

    The comorbidity score was based on the Charlson comorbidity index.

  • Socioeconomic status was determined for each participant on a scale from 0 to 4 based on income data from the 2000 Census and an algorithm described by Bach et al., 2002.15

Age, y
 65–691.00Referent
 70–741.711.16–2.50
 75–791.581.07–2.32
 80–842.401.61–3.58
 ≥854.322.87–6.52
Year of diagnosis
 19911.00Referent
 19921.530.94–2.49
 19931.420.86–2.33
 19941.330.80–2.21
 19951.450.88–2.37
 19961.711.05–2.78
 19971.510.92–2.46
 19982.111.31–3.41
 19991.130.68–1.89
Gender
 Male1.00Referent
 Female0.840.67–1.06
Race
 White1.00Referent
 Black1.170.70–1.95
 Other1.250.83–1.88
Residence
 Urban1.00Referent
 Nonurban0.840.60–1.20
Marital status
 Married1.00Referent
 Not married1.250.99–1.59
 Unknown1.790.89–3.61
Comorbidity score without heart disease*
 01.00Referent
 11.371.06–1.78
 > 12.071.44–2.97
Socioeconomic status
 01.00Referent
 10.800.57–1.11
 20.770.55–1.10
 30.660.46–0.94
 40.840.59–1.19
Congestive heart failure
 No1.00Referent
 Yes0.870.69–1.10
Type of hospital
 Nonteaching1.00Referent
 Teaching1.030.76–1.39
Radiation therapy
 No1.00Referent
 Yes0.250.18–0.36

Among patients who were treated with specific chemotherapeutic drugs, our multivariable logistic regression analysis indicated that patients age >80 years, patients with congestive heart disease, or patients with comorbidity scores <1 were significantly less likely than younger and healthier patients to receive doxorubicin (Table 4). Treatment with doxorubicin was not associated with gender, marital status, socioeconomic status, or type or location of hospital. Patients who were diagnosed after 1995 were significantly more likely than those who were diagnosed earlier to receive doxorubicin.

Table 4. Multivariate Adjusted Odds Ratios for the Associations between Chemotherapy With (vs. Without) Doxorubicin and Clinical and Demographic Factors (n = 1236 Patients)
VariableOR95% CI
  • OR indicates odds ratio; 95% CI, 95% confidence interval.

  • *

    The comorbidity score was based on the Charlson comorbidity index.

  • Socioeconomic status was determined for each participant on a scale from 0 to 4 based on income data from the 2000 Census and an algorithm described by Bach et al., 2002.15

Age, y
 65–691.00Referent
 70–741.270.89–1.82
 75–790.730.51–1.04
 80–840.570.38–0.86
 ≥850.290.17–0.52
Year of diagnosis
 19911.00Referent
 19921.410.83–2.40
 19931.360.79–2.32
 19941.821.06–3.13
 19951.600.93–2.74
 19961.901.11–3.24
 19974.002.30–6.96
 19984.512.55–7.98
 19996.893.84–12.37
Gender
 Male1.00Referent
 Female1.240.95–1.61
Race
 White1.00Referent
 Black0.920.46–1.84
 Other0.980.59–1.62
Residence
 Urban1.00Referent
 Nonurban0.870.58–1.31
Marital status
 Married1.00Referent
 Not married0.980.74–1.30
 Unknown1.450.59–3.57
Comorbidity score without heart disease*
 01.00Referent
 10.840.61–1.16
 > 10.530.32–0.91
Socioeconomic status
 01.00Referent
 10.770.52–1.16
 20.950.62–1.45
 31.260.82–1.95
 40.870.57–1.34
Congestive heart failure
 No1.00Referent
 Yes0.650.50–0.86
Type of hospital
 Nonteaching1.00Referent
 Teaching0.970.69–1.35
Radiation therapy
 No1.00Referent
 Yes0.880.65–1.19

Adjusted disease-specific survival analyses (Fig. 1) showed that patients who received doxorubicin-based chemotherapy had a median survival more than double that in each of the other treatment groups (24.4 months compared with <12 months). The 3-year survival rate was 43.8% among patients in the doxorubicin-treated group but <30% in each of the other 3 groups; the 10-year survival rate was 34% in the doxorubicin-treated group and ≤ 17% in the other groups.

thumbnail image

Figure 1. (A) These Kaplan–Meier disease-specific survival curves were generated for each treatment group. (B) These Kaplan–Meier overall survival curves were generated for each treatment group.

Download figure to PowerPoint

Table 5 presents the results of the Cox proportional hazards regression analysis of the association of treatments and clinical and demographic characteristics with disease-specific and overall mortality. Age, comorbid conditions, and a history of congestive heart failure were associated with both disease-specific mortality and overall mortality. In both propensity score and multivariate Cox models, disease-specific and overall mortality rates were approximately 30% lower among patients who received doxorubicin than among patients who received other chemotherapeutic agents. No significant interactions were observed between treatment and the other covariates in our models.

Table 5. Mortality Hazard Ratios Associated with Treatment, Demographic, and Clinical Factors in a Cox Proportional Hazards Regression Model
VariableDisease-specific survivalOverall survival
HR95% CIHR95% CI
  • HR indicates hazards ratio; 95% CI, 95% confidence interval.

  • *

    The comorbidity score was based on the Charlson comorbidity index.

  • Socioeconomic status was determined for each participant on a scale from 0 to 4 based on income data from the 2000 Census and an algorithm described by Bach et al., 2002.15

Treatment
 Without doxorubicin1.00Referent1.00Referent
 With doxorubicin0.680.58–0.800.730.64–0.84
Age, y
 65–691.00Referent1.00Referent
 70–741.190.94–1.501.190.98–1.44
 75–791.431.13–1.801.431.18–1.74
 80–841.751.34–2.301.621.30–2.02
 ≥852.241.57–3.201.941.44–2.61
Year of diagnosis
 19911.00Referent1.00Referent
 19921.030.72–1.471.000.76–1.32
 19931.060.74–1.501.000.76–1.33
 19940.960.67–1.361.040.78–1.38
 19951.050.74–1.501.060.80–1.42
 19960.880.62–1.260.910.68–1.22
 19970.900.63–1.291.020.76–1.36
 19980.720.50–1.050.840.62–1.14
 19990.690.47–1.010.780.57–1.05
Gender
 Male1.00Referent1.00Referent
 Female0.920.780.860.75–0.99
Race
 White1.00Referent1.00Referent
 Black1.330.88–1.991.250.88–1.77
 Other0.850.61–1.190.850.65–1.12
Geography
 Urban1.00Referent1.00Referent
 Nonurban0.820.63–1.070.860.69–1.08
Marital status
 Married1.00Referent1.00Referent
 Not married1.080.91–1.291.100.95–1.26
 Unknown1.020.58–1.780.960.60–1.52
Comorbidity score without heart disease*
 01.00Referent1.00Referent
 11.150.94–1.401.120.95–1.33
 > 11.541.11–2.141.451.10–1.91
Socioeconomic status
 01.00Referent1.00Referent
 10.920.71–1.190.920.74–1.14
 20.920.70–1.200.910.73–1.14
 30.990.75–1.290.980.78–1.23
 41.000.76–1.320.920.73–1.15
Congestive heart failure
 No cardiac disease1.00Referent1.00Referent
 Cardiac disease1.271.07–1.511.301.13–1.50
Type of hospital
 Nonteaching1.00Referent1.00Referent
 Teaching0.790.63–0.990.860.72–1.02
Radiation therapy
 No1.00Referent1.00Referent
 Yes0.950.78–1.150.980.83–1.14

Table 6 presents the results of the sensitivity analysis using propensity scores for hazard ratios of disease-specific survival. Assuming that the prevalence of an uncontrolled confounder was 10% in the doxorubicin group, we observed that only if an unmeasured binary confounder were 7 times more prevalent in the nondoxorubicin group and increased mortality by ≥50% (hazard ratio, >1.5) would it attenuate the observed association of doxorubicin treatment with decreased mortality. If the unmeasured binary confounder (e.g., poor functional status) increased mortality by 25% (hazards ratio of 1.25), then, even if it were 9 times more prevalent among untreated patients than among treated patients, it would have little effect on the estimated hazard ratio for treatment. Only in extreme situations (in which the unmeasured confounder hazard is high or the imbalance of the unmeasured confounder in treated and nontreated is great) are our results sensitive to hidden bias (see Table 6, the rows with an asterisk, in which the upper bound of the 95% confidence interval crosses 1).

Table 6. Sensitivity Analysis of Uncontrolled Confounding of the Association of Doxorubicin with Disease-Specific Mortality
Prevalence of the uncontrolled confounder among patients who received chemotherapy (%)HR for unknown confounderHR for treatment, adjusted for unknown confounder (95%CI)
With doxorubicinWithout doxorubicin
  • HR indicates hazards ratio; 95% CI, 95% confidence interval.

  • *

    In these rows, the unknown confounder, if it had been included in the model, would be considered to account for the benefit that otherwise would be attributed to treatment.

10301.250.71 (0.61–0.83)
10601.250.76 (0.65–0.89)
10901.250.81 (0.69–0.94)
10301.500.74 (0.64, 0.87)
10501.500.81 (0.69, 0.94)
10701.500.87 (0.75–1.02)*
10301.750.77 (0.66–0.90)
10401.750.82 (0.70–0.95)
10501.750.87 (0.74–1.01)*
10202.000.74 (0.63–0.86)
10302.000.80 (0.69–0.93)
10402.000.87 (0.74–1.00)*
10203.000.79 (0.68–0.92)
10303.000.91 (0.77–1.05)*

DISCUSSION

  1. Top of page
  2. Abstract
  3. MATERIALS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. Acknowledgements
  7. REFERENCES

In our population-based study of elderly patients with diffuse large B-cell lymphoma who were followed for 10 years, doxorubicin-based chemotherapy had a significantly positive impact on both disease-specific and overall survival. Among patients who were treated with doxorubicin, the 3-year overall survival rate was 40%; whereas, in the randomized trials among similarly staged patients, the 3-year survival rate was approximately 50%.1, 4, 6, 24 Of 899 patients who enrolled in the Intergroup (Southwest Oncology Group and Eastern Cooperative Oncology Group) prospective, randomized Phase III trial published in 1993, 44% of those who received doxorubicin-based therapies remained alive without disease at 3 years.4 The overall survival rate was estimated to be 54% with CHOP and slightly lower for more complex drug combinations that included doxorubicin. Only 1% of the CHOP group had fatal reactions.4 Prior to publication of that milestone study, it was believed that the intensive second-generation and third-generation treatments were twice as effective as CHOP. After the Intergroup study was published, CHOP chemotherapy became the standard of care for intermediate-grade and high-grade lymphoma in the U.S.25

A small number of elderly patients in our sample received mitoxantrone, which is an anthracenedione derivative that has less cardiotoxicity than doxorubicin.26 Their survival was similar to that of other patients who were treated without doxorubicin.

In 2002, it was observed that patients who received R-CHOP in a randomized trial sponsored by Groupe d'Etude des Lymphomes de l'Adulte (GELA)5 had a 2-year survival rate of 70% compared with a rate of 57% for patients who received CHOP alone. With longer term follow-up, the overall survival rate at 5 years in that trial of healthy patients ages 60 to 75 years was 58% for R-CHOP and 45% for CHOP.27 That trial differed from our SEER-Medicare cohort; patients in the GELA study who had any serious concomitant disease or were considered unable to tolerate chemotherapy were excluded.

In earlier randomized Phase III trials, younger patients who received doxorubicin had long-term survival rates of up to 30%.28, 29 In our population-based study, patients who received doxorubicin had a disease-specific survival rate of 50% at 2 years and 34% at 10 years, whereas patients who received other chemotherapy regimens had a 34% disease-specific survival rate at 2 years.

We excluded from our main analysis patients who died in the first 6 weeks after diagnosis, because such patients did not have time to benefit from chemotherapy. Of the patients who died early, 88.8% had received no chemotherapy; and, of the patients who received no chemotherapy, 42% died early (Table 2). By 6 months, 67.8% of untreated patients had died, whereas only 15.5% of doxorubicin-treated patients had died. The association of being untreated with early death and with older age and comorbid conditions suggests that patients' health status rather than socioeconomic factors motivated decisions about their treatment.

Analyses of treatment outcomes in the SEER-Medicare data are limited by changes in pathologic classification and staging, uncertainties in the measurement of disease-free survival,30 the use of billing data to ascertain treatment, the time it takes to observe the outcomes of newly introduced therapies like rituximab,11, 24 the lack of information on functional status and quality of life among individual patients, and the use of demographic data from the U.S. Census to represent patients' demographic information.15 However, our sensitivity analysis indicated that an unmeasured confounder would have to have a large effect size and a highly unequal distribution between groups to account for the entire benefit associated with doxorubicin-based treatment.

The classification of lymphomas is evolving. In the last decade, increasing attention has been given to biologic predictors, such as lactate dehydrogenase, disease stage, performance status, and patients age, and to biomarkers, such as β-2 microglobulin.1 Today, microarray technology has made it possible to group subpopulations of lymphoma tumors genetically into prognostic and possibly therapeutic categories.31, 32 Soon, these newer biogenetic predictors may lead to more tailored therapies. Incorporating them into tumor registry data will enable health services investigators to assess their effectiveness.

Analyses of SEER-Medicare and other administrative data complement clinical trials. New therapies need to be used before their effectiveness can be monitored, and analyzing their effects on survival takes years. The SEER data base now covers 26% of the U.S. population. Linking SEER to Medicare has enabled us to study the real-world use of medications, such as doxorubicin, and their effectiveness over time.

Observational studies can provide important information regarding the diffusion of evidence-based treatments, the determinants of appropriate treatment, and the long-term overall and cancer-specific survival and adverse events that may accompany such therapy. The SEER-Medicare data base specifically provides that information about patients age >65 years, who generally have been under-represented in clinical trials but are a large proportion of the cancer patient population.

In the current study, we observed that, after publication of the randomized trial results in 1993, doxorubicin-based chemotherapy was used increasingly to treat elderly patients with aggressive B-cell lymphomas.4 Among patients with lymphoma,13, 33, 34 the major barriers to treatment among lymphoma patients were similar to those we observed among patients with ovarian cancer: comorbidities and age.35 In 1999, nearly 50% of elderly patients in the general population were still not receiving doxorubicin-based treatment. A nationwide cross-sectional study among 4522 elderly patients with aggressive lymphoma who were treated in the community also showed that the majority did not receive optimal treatment.8

State-of-the-art treatment is now R-CHOP,36 but doxorubicin most likely remains an important component of this therapy. In the absence of specific contraindications, our current findings justify giving doxorubicin-based chemotherapy, consistent with the clinical trial results,4, 27, 36 to elderly patients with advanced large B-cell lymphoma.

Acknowledgements

  1. Top of page
  2. Abstract
  3. MATERIALS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. Acknowledgements
  7. REFERENCES

We acknowledge the efforts of the Applied Research Branch, Division of Cancer Prevention and Population Science, NCI; the Office of Information Services, and the Office of Strategic Planning, Health Care Financing Administration; Information Management Services, Inc.; and the Surveillance, Epidemiology, and End Results (SEER) Program tumor registries in the creation of the SEER-Medicare data base

REFERENCES

  1. Top of page
  2. Abstract
  3. MATERIALS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. Acknowledgements
  7. REFERENCES
  • 1
    Fisher RI, Miller TP, O'Connor OA. Diffuse aggressive lymphoma. Hematology Am Soc Hematol Educ Program. 2004: 221236.
  • 2
    McKelvey EM, Gottlieb JA, Wilson HE, et al. Hydroxyldaunomycin (Adriamycin) combination chemotherapy in malignant lymphoma. Cancer. 1976; 38: 14841493.
  • 3
    Gordon LI, Harrington D, Andersen J, et al. Comparison of a second-generation combination chemotherapeutic regimen (m-BACOD) with a standard regimen (CHOP) for advanced diffuse non-Hodgkin's lymphoma. N Engl J Med. 1992; 327: 13421349.
  • 4
    Fisher RI, Gaynor ER, Dahlberg S, et al. Comparison of a standard regimen (CHOP) with three intensive chemotherapy regimens for advanced non-Hodgkin's lymphoma. N Engl J Med. 1993; 328: 10021006.
  • 5
    Coiffier B. Rituximab in combination with CHOP improves survival in elderly patients with aggressive non-Hodgkin's lymphoma. Semin Oncol. 2002; 29(2 Suppl 6 ): 1822.
  • 6
    Coiffier B, Lepage E, Briere J, et al. CHOP chemotherapy plus rituximab compared with CHOP alone in elderly patients with diffuse large-B-cell lymphoma. N Engl J Med. 2002; 346: 235242.
  • 7
    Ries L, Eisner M, Kosary C, et al. SEER cancer statistics review, 1975–2002, vol 2005. Bethesda: National Cancer Institute, 2005. Available at URL: http://seer.cancer.gov/csr/1975_2002/ [based on the November 2004 SEER data submission, posted to the SEER web site 2005]. Accessed December 14, 2005.
  • 8
    Lyman GH, Dale DC, Friedberg J, Crawford J, Fisher RI. Incidence and predictors of low chemotherapy dose-intensity in aggressive non-Hodgkin's lymphoma: a nationwide study. J Clin Oncol. 2004; 22: 43024311.
  • 9
    Bennett CL, Greenfield S, Aronow H, Ganz P, Vogelzang NJ, Elashoff RM. Patterns of care related to age of men with prostate cancer. Cancer. 1991; 67: 26332641.
  • 10
    Greenfield S, Blanco DM, Elashoff RM, Ganz PA. Patterns of care related to age of breast cancer patients. JAMA. 1987; 257: 27662770.
  • 11
    Potosky AL, Riley GF, Lubitz JD, Mentnech RM, Kessler LG. Potential for cancer related health services research using a linked Medicare-tumor registry database. Med Care. 1993; 31: 732748.
  • 12
    Sundararajan V, Mitra N, Jacobson JS, Grann VR, Heitjan DF, Neugut AI. Survival associated with 5-fluorouracil-based adjuvant chemotherapy among elderly patients with node-positive colon cancer. Ann Intern Med. 2002; 136: 349357.
  • 13
    Sundararajan V, Hershman D, Grann VR, Jacobson JS, Neugut AI. Variations in the use of chemotherapy for elderly patients with advanced ovarian cancer: a population-based study. J Clin Oncol. 2002; 20: 173178.
  • 14
    Krzyzanowska MK, Weeks JC, Earle CC. Treatment of locally advanced pancreatic cancer in the real world: population-based practices and effectiveness. J Clin Oncol. 2003; 21: 34093414.
  • 15
    Bach PB, Guadagnoli E, Schrag D, Schussler N, Warren JL. Patient demographic and socioeconomic characteristics in the SEER-Medicare database applications and limitations. Med Care. 2002; 40(8 Suppl ): IV-1925.
  • 16
    Charlson ME, Pompei P, Ales KL, MacKenzie CR. A new method of classifying prognostic comorbidity in longitudinal studies: development and validation. J Chron Dis. 1987; 40: 373383.
  • 17
    Warren JL, Harlan LC, Fahey A, et al. Utility of the SEER-Medicare data to identify chemotherapy use. Med Care. 2002; 40(8 Suppl ): IV-55IV-61.
  • 18
    Lamont EB, Lauderdale DS, Schilsky RL, Christakis NA. Construct validity of Medicare chemotherapy claims: the case of 5FU. Med Care. 2002; 40: 201211.
  • 19
    D'Agostino RBJr. Propensity score methods for bias reduction in the comparison of a treatment to a non-randomized control group. Stat Med. 1998; 17: 22652281.
  • 20
    Rubin DB. Estimating causal effects from large data sets using propensity scores. Ann Intern Med. 1997; 127(8 Pt 2): 757763.
  • 21
    Cochran WG. The effectiveness of adjustment by subclassification in removing bias in observational studies. Biometrics. 1968; 24: 295313.
  • 22
    Lin DY, Psaty BM, Kronmal RA. Assessing the sensitivity of regression results to unmeasured confounders in observational studies. Biometrics. 1998; 54: 948963.
  • 23
    Mitra N, Heitjan DF. Sensitivity of the Hazard Ratio to Non-Ignorable Treatment Assignment in an Observational Study. Upenn Biostatistics Working Papers. Paper 5, 2006. The Berkeley Electronic Press. Available at: http://biostats/bepress.com/upennbiostat/papers/arts [accessed May 24, 2006].
  • 24
    Cheson BD. CHOP plus rituximab—balancing facts and opinion. N Engl J Med. 2002; 346: 280292.
  • 25
    Grann A, Grann VR. The case for randomized trials in cancer treatment: new is not always better. JAMA. 2005; 293: 10011003.
  • 26
    Sonneveld P, de Ridder M, van der Lelie H, et al. Comparison of doxorubicin and mitoxantrone in the treatment of elderly patients with advanced diffuse non-Hodgkin's lymphoma using CHOP versus CNOP chemotherapy. J Clin Oncol. 1995; 13: 25302539.
  • 27
    Feugier P, Van Hoof A, Sebban C, et al. Long-term results of the R-CHOP study in the treatment of elderly patients with diffuse large B-cell lymphoma: a study by the Groupe d'Etude des Lymphomes de l'Adulte. J Clin Oncol. 2005; 23: 41174126.
  • 28
    Miller TP, Dana BW, Weick JK, et al. Southwest Oncology Group clinical trials for intermediate- and high-grade non-Hodgkin's lymphomas. Semin Hematol. 1988; 25(2 Suppl 2 ): 1722.
  • 29
    Fisher RI, Miller TP, Dana BW, Jones SE, Dahlberg S, Coltman CAJr. Southwest Oncology Group clinical trials for intermediate- and high-grade non-Hodgkin's lymphomas. Semin Hematol. 1987; 24(2 Suppl 1 ): 2125.
  • 30
    Earle CC, Nattinger AB, Potosky AL, et al. Identifying cancer relapse using SEER-Medicare data. Med Care. 2002; 40(8 Suppl ): IV-7581.
  • 31
    Rosenwald A, Wright G, Chan WC, et al. The use of molecular profiling to predict survival after chemotherapy for diffuse large-B-cell lymphoma. N Engl J Med. 2002; 346: 19371947.
  • 32
    Bea S, Zettl A, Wright G, et al. Diffuse large B-cell lymphoma subgroups have distinct genetic profiles that influence tumor biology and improve gene-expression-based survival prediction. Blood. 2005; 106: 31833190.
  • 33
    Steyerberg EW, Earle CC, Neville BA, Weeks JC. Racial differences in surgical evaluation, treatment, and outcome of locoregional esophageal cancer: a population-based analysis of elderly patients. J Clin Oncol. 2005; 23: 510517.
  • 34
    Neugut AI, Fleischauer AT, Sundararajan V, et al. Use of adjuvant chemotherapy and radiation therapy for rectal cancer among the elderly: a population-based study. J Clin Oncol. 2002; 20: 26432650.
  • 35
    Hershman D, Jacobson JS, McBride R, et al. Effectiveness of platinum-based chemotherapy among elderly patients with advanced ovarian cancer. Gynecol Oncol. 2004; 94: 540549.
  • 36
    National Comprehensive Cancer Network. Clinical Practice Guidelines in Oncology – v.2.2006 Non-Hodgkin's Lymphoma. Available at: http://www.nccn.org/professionals/physicians_gls/PDF/nhl.pdf [accessed May 24, 2006].