Exploring racial differences in outcome and treatment for metastatic colorectal cancer

Results from a large prospective observational cohort study (BRiTE)

Authors


Abstract

BACKGROUND:

African Americans are more likely to be diagnosed with metastatic colorectal cancer than whites and have shorter survival once they are diagnosed. In this analysis, the authors examined racial differences in clinical outcomes among patients with metastatic colorectal cancer (mCRC) who received bevacizumab.

METHODS:

The study cohort consisted of 1589 white patients (81.4%) and 227 African American patients (11.6%) with mCRC who received front-line bevacizumab therapy and who were enrolled in a large, predominantly community-based, prospective, observational cohort study. Differences in time-to-event endpoints and response rates were examined by race. Differences in the incidence of baseline and treatment-related toxicities associated with bevacizumab also were examined. Finally, differences in patterns of care by race were explored.

RESULTS:

The median overall survival was 22.6 months for African Americans and 22.9 months for whites, and the median progression-free survival was 9.5 months for African Americans and 9.8 months for whites. Response rates (complete responses plus partial responses) were 37.5% for African Americans and 46.3% for whites (adjusted odds ratio, 0.67; 95% confidence interval, 0.50-0.90). African Americans had higher rates of baseline diabetes (18.9% vs 11%; P = .002), higher rates of hypertension (52.9% vs 41.4%; P = .001), and worsening hypertension while on therapy (13.7% vs 8.9%; P = .02), but no differences in on-treatment arterial thromboembolic events were observed.

CONCLUSIONS:

This large observational cohort study of patients with mCRC demonstrated that, when treated in a similar fashion with modern chemotherapy, African Americans and whites had equivalent cancer outcomes. No significant differences in bevacizumab-related toxicity or patterns of care were observed between African Americans and whites. The lower response rate among African Americans deserves further study. Cancer 2012;. © 2011 American Cancer Society.

INTRODUCTION

It is well established that African Americans with colon and rectal cancer have both a higher incidence and poorer overall survival compared with whites and that this survival disparity actually is growing.1-3 Much of this disparity is attributed to differences in stage at diagnosis4; however, even when analyses are controlled for disease stage at diagnosis, survival disparities persist and are worsening over time.

The driving factors for this increased stage-specific survival disparity are unclear. There are several potential explanations, including lower rates or quality of treatment for African American patients during a time when the availability of curative therapy options was increasing; higher rates of baseline comorbidities, which lead to poorer overall survival; or differences in underlying tumor biology, which lead to lower survival rates for African American patients.5-12 Evaluations of the clinical trial data consistently indicate no differences in cancer-specific survival by race but, in certain trials, do reveal overall survival differences, suggesting that baseline comorbidities may be important.5, 13-15

The baseline comorbidity differences between African Americans and whites may be particularly important with bevacizumab, which is known to cause increases in the rates of hypertension and arterial thromboembolic events.16 Given the demonstrated efficacy of bevacizumab in the front-line treatment of metastatic colorectal cancer,17 it will be particularly important to explore racial differences in outcomes among patients who receive bevacizumab in the front-line metastatic setting given the higher rates of cardiovascular risk factors among African Americans.18-21 In this report, we present results from the Bevacizumab Regimens: Investigation of Treatment Effects and Safety (BRiTE) study, an observational study that, as described below, provides data that overcome many of the limitations described above.

MATERIALS AND METHODS

Study Details

Details of the BRiTE registry have been published previously; however, in brief, BRiTE is a large, US-based, prospective, observational cohort study that was conducted at 248 study sites in 49 states.22 Also, 98% of patients in the study were from community-based versus academic medical practices. Patients were able to participate if they met the following criteria: previously untreated metastatic colorectal cancer, treatment with bevacizumab as first-line therapy, and signed informed consent. There were no specific exclusion criteria. Patients were considered on study until death, withdrawal of consent, or loss to follow-up. There were no protocol-specified treatments or assessments. All aspects of patients' treatments over time, including specific chemotherapy agents and/or combinations and the dose, schedule, and duration of bevacizumab treatment, were determined by the patient's physician. Occurrences of adverse events were reported by each patient's physician. Data were collected at baseline and every 3 months through a web-based electronic data-collection system. Total dropout or loss to follow-up was 12% (235 patients) and was similar between African Americans and whites (10.6% vs 11.6%, respectively).

Clinical Outcomes and Measures

Measures of clinical outcomes were based on physician determination and included the time to progression from the date first-line treatment was initiated (progression-free survival [PFS]); survival from the date of initiation of first-line treatment to death (overall survival [OS]); and survival beyond first progression (SBP), which was measured only for patients who had reported first progression (progressive disease [PD]) and was calculated from the date of first PD to death. Patients who had not died were censored at study termination (because of loss to follow-up, patient's decision, investigator's decision, or sponsor's decision) or at the study end (July 30, 2008). The best first-line tumor response and PFS were analyzed according to investigator determination, including the imaging method and interval.

Collaboration of data on baseline comorbidities and on-treatment adverse events in BRiTE were focused on previously described, bevacizumab-related adverse events.17 Baseline comorbidities included arterial disease history (history of stroke, transient ischemic attack, myocardial infarction, angina, and peripheral artery disease), diabetes requiring medication, and hypertension requiring medication. On-study adverse events included arterial thromboembolic events, venous thromboembolic events, new or worsening hypertension (defined as an increase in dose of or the addition of new hypertensive medications for those already on therapy or the initiation of hypertensive medication for those without a previous history), gastrointestinal (GI) perforation, and grade 3 or 4 bleeding.

Statistical Analyses

PFS, OS, and SBP were analyzed using the Kaplan-Meier method, and comparisons by race (based on patient self-identification) were analyzed using the log-rank test. A Cox proportional hazards model was used to assess the independent effects of race on time-to-event endpoints, adjusting for the following factors in the multivariate models: age, sex, Eastern Cooperative Oncology Group performance status, any prior adjuvant therapy, hypertension requiring medication at baseline, arterial disease history at baseline, diabetes mellitus requiring medication at baseline, and disease-free interval (measured from the date of initial colorectal cancer diagnosis). Response rates were compared using the chi-square test and a logistic regression model with the inclusion of the same factors specified above for the multivariate analysis and also with the addition of the type of front-line chemotherapy received. PFS and response rates were analyzed further using a random-effects Cox model and a mixed logistic regression model, respectively, to adjust for site effects (physician information was not available) on patient disease status assessment. All other comparisons by race were analyzed in fashion similar to that used for response rate assessment described above.

RESULTS

Baseline Characteristics

The study population consisted of 1589 white patients (81%) and 227 African American patients (11.6%) (Table 1). The median follow-up was 20 months. Overall, the study populations were similar with the exception that, compared with white patients, African American patients were more likely to be women (54% vs 43%; P = .003), to have an Eastern Cooperative Oncology Group performance status of 0 (47% vs 43%; P = .052), to have hypertension requiring medication (53% vs 41%; P = .001), and have to diabetes requiring medication (19% vs 11%; P = .002). There were no differences in the frequency of de novo metastatic disease (42% vs 41%), the interval from primary colon cancer diagnosis to metastatic disease for those with a previous history of colon cancer, previous receipt of adjuvant chemotherapy for colon cancer (35% vs 36%), or location of metastatic disease sites.

Table 1. Baseline Characteristics by Race
CharacteristicWhite (N = 1589)AA (N = 227)P
  • Abbreviations: AA, African American; ECOG PS, Eastern Cooperative Oncology Group performance status; FOLFIRI, 5-fluorouracil, leucovorin, and irinotecan; FOLFOX, oxaliplatin, 5-fluorouracil, and folinic acid; XELOX, capecitabine and oxaliplatin.

  • a

    Primarily irinotecan plus 5-fluorouracil/leucovorin, bolus and infusional 5-fluorouracil, and single-agent capecitabine.

Median age at baseline, y63.961.8.02
Age group, %   
 <50 y1418.5 
 50 to <65y3941 
 ≥65 y47.140.5 
Women, %43.554.2.003
Baseline ECOG PS, %   
 042.747.1 
 141.843.6 
 ≥27.15.7 
 Unknown8.43.5 
Site of primary tumor, %   
 Colon78.783.3 
 Rectum21.116.7 
 Missing information0.20 
Metastatic disease status, %   
 Local/regional only19.314.5 
 Hepatic only45.346.7 
 Lung only10.19.7 
 Local/regional and hepatic4.95.3 
 Other20.423.8 
Received prior adjuvant therapy, %37.236.1 
First-line chemotherapy, %   
 FOLFOX, FOLFIRI, or XELOX75.371.8 
 Othera24.728.2 
Hypertension requiring medication at baseline, %41.452.9.001
Diabetes mellitus requiring medication at baseline, %1118.9.002
Stroke history at baseline, %2.12.2 
Transient ischemic attack history at baseline, %10.4 
Myocardial infarction history at baseline, %3.72.2 
Angina history at baseline, %1.30.9 
Peripheral artery disease history at baseline, %1.50.4 
Median albumin, g/dL3.83.8 
Median alkaline phosphatase, U/L101108 

Treatment Variables

African Americans were slightly less likely to receive front-line oxaliplatin, fluorouracil, and folinic acid (FOLFOX), capecitabine plus oxaliplatin (XELOX), or fluorouracil, leucovorin and irinotecan (FOLFIRI) with bevacizumab (72% vs 75%), but this difference was not statistically significant (P = .25) (Table 2). African American patients were as likely as white patients to receive therapy postprogression, including an equal likelihood of receiving all 3 types of active cytotoxic chemotherapy agents (fluropyrimidines, oxaliplatin, or irinotecan), epidermal growth factor receptor-based therapy postprogression, and bevacizumab postprogression. African Americans and whites had similar median duration of total chemotherapy/epidermal growth factor receptor inhibitor use (11.2 months vs 11.1 months, respectively; P = .91). African Americans also were as likely as whites to undergo hepatic metastatectomy after enrollment in the study (5.7% vs 5.5%, respectively).

Table 2. Treatment Differences by Race
VariableAA, %White, %Unadjusted OR/HR (AA/White)95% CIAdjusted OR/HR (AA/White)a95% CI
  • Abbreviations: AA, African American; CI, confidence interval; EGFR, epidermal growth factor receptor; FOLFIRI, 5-fluorouracil, leucovorin, and irinotecan; FOLFOX, oxaliplatin, 5-fluorouracil, and folinic acid; HR, hazard ratio; OR, odds ratio; PD, progressive disease; XELOX, capecitabine and oxaliplatin.

  • a

    Adjusted for age, sex, Eastern Cooperative Oncology Group performance status, any prior adjuvant therapy, and diabetes mellitus requiring medication at baseline.

  • b

    The denominator is the number of patients with disease progression.

  • c

    Capecitabine or 5-fluorouracil, irinotecan, oxaliplatin.

Receipt of first-line FOLFOX, XELOX, or FOLFIRI71.875.30.830.61-1.140.780.57-1.08
Receipt of any treatment postprogression84.8b85b0.980.64-1.510.930.60-1.44
Exposure to 3 active chemotherapy agentsc52.948.71.180.89-1.561.150.87-1.54
Exposure to EGFR inhibitors after first PD50b47.1b1.120.82-1.531.110.81-1.53
Use of bevacizumab beyond progression48.4b49.8b0.940.69-1.290.960.70-1.32
Postbaseline hepatic metastatectomy5.75.51.040.57-1.891.020.56-1.88

Clinical Outcomes

There were no differences between African Americans and whites in OS (22.6 months vs 22.9 months; P = .83), PFS (9.3 months vs 9.7 months; P = .15), or SBP (13.2 months vs 12 months; P = .63) (Table 3; Figs. 1 and 2). Unadjusted hazard ratios for OS and SBP were near 1.0; whereas, for PFS, there was a statistically nonsignificant 11% greater failure risk for African Americans (Table 3). These similar outcomes persisted when other patient and disease characteristics were adjusted for in the multivariate hazard models (Table 3). Furthermore, a random-effects Cox model that adjusted for the correlation of patient PFS assessments within a site yielded the same PFS results.

Figure 1.

This Kaplan-Meier analysis illustrates the effect of race on overall survival among patients who were enrolled in the Bevacizumab Regimens: Investigation of Treatment Effects and Safety (BRiTE) registry.

Figure 2.

This Kaplan-Meier analysis illustrates the effect of race on progression-free survival among patients who were enrolled in the Bevacizumab Regimens: Investigation of Treatment Effects and Safety (BRiTE) registry.

Table 3. Effect of Race on Response, Overall Survival, Time to Progression, and Survival Beyond First Progression
VariableAAWhiteUnadjusted OR/HR (AA/White)95% CIAdjusted OR/HR (AA/White)1a95% CI
  • Abbreviations: AA, African American; CI, confidence interval; CR, complete response; HR, hazard ratio; OR, odds ratio; OS, overall survival; PFS, progression-free survival; PR, partial response; SBP, survival beyond first progression.

  • a

    Adjusted for age, sex, Eastern Cooperative Oncology Group performance status, any prior adjuvant therapy, hypertension requiring medication at baseline, arterial disease history at baseline, diabetes mellitus requiring medication at baseline, and disease-free interval. Chemotherapeutic agents also were adjusted in the response model.

  • b

    Patients who had the best first-line response but were not evaluable or had no assessment done (93 white patients and 10 AA patients) were excluded from the response analysis.

  • c

    Median (Kaplan-Meier estimate).

Response (CR plus PR), %b39.249.20.660.50-0.890.670.50-0.90
OS, mo22.6c22.9c1.020.86-1.211.030.87-1.22
PFS, mo9.3c9.7c1.110.96-1.291.090.94-1.29
SBP, mo13.2c12.0c0.960.80-1.150.970.81-1.17

African Americans did have a lower objective response rate to first-line therapy (39.2% vs 49.2%; P = .043), which remained statistically significant after covariate adjustment that included the type of primary chemotherapy (odds ratio [OR], 0.67; 95% confidence interval [CI], 0.50-0.90). Similar results (OR, 0.65; 95% CI, 0.47-0.91) were obtained from a mixed logistic regression model that adjusted for the correlation among patient response assessments within a site.

On-Therapy Toxicity

Overall, African Americans and whites had similar rates of bevacizumab-associated adverse events, including development of new hypertension, arterial thromboembolic events, GI perforation, and bleeding (Table 4). However, among those patients who had hypertension at baseline, African Americans were more likely than whites to develop worsening hypertension while on therapy (13.7% vs 8.9%; OR, 1.61; 95% CI, 1.06-2.47).

Table 4. Toxicity by Race
Toxicity VariablesAA, %White, %Unadjusted OR (AA/White)95% CIAdjusted OR (AA/White)a95% CI
  • Abbreviations: AA, African American; ATEs, arterial thromboembolic events; CI, confidence interval; GI, gastrointestinal; HTN, hypertension; OR, odds ratio; VTEs, venous thromboembolic events.

  • a

    Adjusted for age, sex, Eastern Cooperative Oncology Group performance status, and surgical resection of primary disease.

  • b

    The adjusted ORs for ATE and GI perforation are not provided because of the small number of events in AA patients, which yielded invalid adjusted ORs.

ATEsb1.82.30.770.27-2.19
VTEs5.34.81.110.59-2.081.070.57-2.01
Grade 3-4 bleeding2.621.320.55-3.201.370.56-3.35
GI perforationb0.420.220.03-1.640.03-1.50
New HTN13.212.71.050.69-1.581.030.68-1.56
Worsening HTN13.78.91.621.07-2.461.611.06-2.47

DISCUSSION

Consistent with data from the colorectal clinical trial literature, the current results indicate that, when similar therapy is given, racial differences in colorectal cancer survival are minimal to absent.5, 13-15, 23, 24 The criticism of studies based on clinical trial cohorts is that they represent a highly selected cancer population that may not be representative of the general cancer population treated in the community; their advantage is that treatment and follow-up data are standardized and rigorously collected. The findings from clinical trial cohorts have differed from those reported from many of the national and state registry database studies in which racial differences in survival persist even after controlling for known patient and tumor-related factors.1-3, 25-32 Such registry-based studies have the advantage of being population-based but suffer from the lack of detailed, patient-specific treatment data. The strength of the current analysis from BRiTE is that it likely includes a somewhat less highly selected population than a clinical trial study as a result of its minimal eligibility criteria and physician-determined treatment regimen while, at the same time, prospectively collecting more detailed patient treatment and toxicity information than can be obtained in national and state registries. The current analysis also has the added advantage of including more detailed information on relevant baseline comorbidities and follow-up treatment information than is typically included in either clinical trial or registry studies. This last point is particularly relevant for metastatic colorectal cancer, because it has been demonstrated that exposure to all active chemotherapeutic agents is associated significantly with OS.33

This also is the first study to our knowledge that has explicitly explored the toxicity of bevacizumab among African American patients. This is highly relevant given the higher prevalence of cardiovascular risk factors, such as diabetes and hypertension, among African Americans compared with whites.18-21 African American patients could be at higher risk for arterial events, such as myocardial infarction and stroke, which have known associations with bevacizumab therapy.16 The absence of higher rates of arterial thromboembolic events in the current study among African Americans compared with whites despite a higher prevalence of baseline hypertension and diabetes among African American patients in this study is encouraging. It also should be noted that the prevalence of diabetes and hypertension in our cohort is consistent with that observed in the National Health and Nutrition Examination Survey (NHANES) data, suggesting a lack of significant selection bias in this cohort observational study.19, 20 The cardiovascular event of developing worsening hypertension observed during our study period was higher among African Americans. We explored the possibility that African Americans received bevacizumab for a shorter duration, but no differences in the median duration of first-line or total bevacizumab doses were observed (data not shown). Whether this worsening hypertension among African Americans compared with whites was directly related to bevacizumab therapy or would have occurred even in the absence of such therapy cannot be answered by this study, because there was no control group of patients who did not receive bevacizumab.

Although it was not designed as a study to evaluate the use of different treatment modalities in the community, given the considerable discretion afforded a patient's oncologist in terms of the backbone chemotherapy regimen used and the choice of whether to proceed with therapy beyond first progression, the current study does afford the opportunity to determine whether there are differences in treatment by race among community oncologists. Again, it is encouraging to note that there do not appear to be any significant differences in treatment. An important caveat is that this database did not allow for direct calculation of dose intensity or density by race, and these variables reportedly differ in the breast cancer literature.34 We are comforted by the finding that, if there was under dosing by race (lower dose density or intensity for African Americans), then this probably would bias the results away from the findings in this study of no difference, and controlling for dose intensity would have made our findings even more robust.

Although no significant racial differences in OS or PFS were observed in this study, African Americans did have a lower response rate to therapy. This is consistent with the results reported by Sanoff et al in the N9741 analysis.5 A second finding noted in the N9741 analysis and in a previously published analysis of a GI Intergroup adjuvant stage II and III study (INT 0089) was that African Americans had lower rates of GI toxicity than whites.23 This raises the question of whether a certain subset of African American patients actually may benefit from higher doses of chemotherapy. Our current state of knowledge regarding pharmacogenomic differences by race for the chemotherapeutic agents used in colorectal cancer is limited. One study has suggested that African Americans have higher rates of dihydropyrimidine dehydrogenase deficiency compared with whites,35 which may lead to more frequent dose reductions of 5-fluorouracil for African American patients. Another study has suggested possible differences between African Americans and whites in alleles of unknown significance in the enhancer region that controls thymidylate synthase expression, which theoretically may lead to tumor resistance to 5-fluorouracil.36 Several other pharmacogenomic variants by race for 5-fluorouracil, irinotecan, and oxaliplatin were noted in the N9741 analysis discussed above.5 It will be important to explore whether this difference in the response rate is confirmed in ongoing randomized trials for metastatic colorectal cancer, because this may have implications for the treatment of patients with suspected micrometastatic disease in the adjuvant setting, in which complete eradication of these micrometastases is the goal.

An important caveat to our findings regarding response rates and PFS is that tumor response and progression were physician determined, were not required to be assessed according to Response Evaluation Criteria in Solid Tumors, and did not undergo central review, as is often done in a clinical trial. We attempted to compensate for this in our models by controlling for the practice site where a patient was treated; however, if there were differences within a site in when and how response rates and progression were characterized for African American versus white patients, then the modeling used would not compensate for these differences.

There are several additional limitations to the current study. First, although eligibility criteria and treatment requirements were very flexible for this observational cohort study, a basic requirement was that all patients had to be started on front-line bevacizumab therapy. Oncologists, who are well aware of the cardiovascular risks associated with bevacizumab, likely would not start such therapy on patients deemed at high risk for such events. It is also possible that oncologists were even more selective in placing their African American patients on bevacizumab given the higher prevalence of cardiovascular risk factors in this group. This introduces a selection bias that limits the external validity of the study. This concern is somewhat countered by noting that the prevalence of diabetes and hypertension in this study were in line with national prevalence estimates by race. The median age of 63.5 years for the patients in this study also was younger than the general population of patients with metastatic colon cancer, especially among whites, in which 60% of patients with metastatic colorectal cancer are aged ≥65 years. The median age of 61.7 years for African Americans was more in line with registry data, which reflect that 50% of patients are aged <65 years.2

A second limitation of this study is that no socioeconomic data, such as insurance status, were collected. To the extent that oncologists are more likely to initiate high-cost therapies like bevacizumab only in patients who have adequate insurance, the current analysis will not be representative of the general colon cancer population and could introduce a more significant bias for African Americans, who have higher rates of noninsurance coverage than whites.37 The significance that this potential bias has on the findings of this study is not clear. In a meta-analysis of studies with adjustments for socioeconomic status that examined differences in cancer-specific survival and OS between African Americans and whites with colorectal cancer, Du et al observed that African Americans continued to have a 14% increase in the hazard of all-cause mortality and a 13% increase in colon cancer mortality.32 In the National Cancer Institute's Black and White Cancer Survival Study, which included in-person interviews to collect detailed sociodemographic information on each patient, African Americans were 50% more likely to die from colon cancer than whites.4 When stage was included in the model, the relative mortality excess decreased to 20%. However, the inclusion of individual sociodemographic variables (education, poverty index, marital status, occupation, source of care, type of insurance) in the model did not alter the mortality excess, which remained 20%. This suggests that the sociodemographic effect may be mediated through disease stage at diagnosis but does not exert additional influence on survival once stage is taken into consideration.

In conclusion, this large observational cohort study of patients with metastatic colorectal cancer who received bevacizumab plus chemotherapy in the first-line setting demonstrates that, when patients are treated in a similar fashion with modern chemotherapy, African Americans and whites have equivalent clinical outcomes. This suggests that differences in treatment received may be driving the disparities observed in population-based epidemiologic data. It is encouraging that we observed no significant differences in bevacizumab-related toxicity or patterns of care between African Americans and whites in this primarily community-based study. The lower response rate among African Americans despite similar therapy confirms previous findings and deserves further study.

FUNDING SOURCES

This work was supported by funding from Genentech Inc.

CONFLICT OF INTEREST DISCLOSURES

Dr. Sing is an employee of Genentech. Dr. Sargent is a consultant to and has received honoraria from Genentech. Dr. Grothey is an unpaid consultant to Genentech. Dr. Berlin has received research support from and is a consultant to Genentech. Dr. Kozloff is a consultant to and has received honoraria from Genentech. Dr. Feng is an employee of Genentech.

Ancillary