|Home | About | Journals | Submit | Contact Us | Français|
Women of African ancestry (AA) have lower WBC counts and are more likely to have treatment delays and discontinue adjuvant breast cancer therapy early compared with white women. We assessed the association between race and treatment discontinuation/delay, WBC counts, and survival in women enrolled onto breast cancer clinical trials.
AA and white women from Southwest Oncology Group adjuvant breast cancer trials (S8814/S8897) were matched by age and protocol. Only the treatment arms in which patients were scheduled to receive six cycles of chemotherapy were analyzed.
A total of 317 pairs of patients (n = 634) were analyzed. At baseline, AA women had higher body-surface area (P < .0001) and lower WBC (P = .0009). AA women were more likely to have tumors that were ≥ 2 cm (P = .01) and hormone receptor negative (P < .0001). AA women, versus white women, were marginally more likely to discontinue treatment early (11% v 7%, respectively; P = .07) or have one or more treatment delays (85% v 79%, respectively; P = .07) and were significantly more likely to experience the combined end point (discontinuation/delay; 87% v 81%, respectively; P = .04). The mean relative dose-intensity (RDI) was similar for both groups (87% in AA women v 86% in white women); however, overall, 43% had an RDI of less than 85%. After adjusting for baseline WBC and prognostic factors in a multivariate model, AA women had worse disease-free survival (hazard ratio [HR] = 1.56; 95% CI, 1.15 to 2.11; P = .005) and overall survival (HR = 1.95; 95% CI, 1.36 to 2.78; P = .0002). The inclusion of RDI and treatment delivery/quality in the regression had little impact on the results.
On cooperative group breast cancer trials, AA and white women had similar RDIs, but AA women were more likely to experience early discontinuation or treatment delay. Despite correcting for these factors and known predictors of outcome, AA women still had worse survival.
Despite lower breast cancer incidence, women of African ancestry (AA) have significantly worse survival from breast cancer than white women.1 A prior analysis from the Southwest Oncology Group (SWOG) reporting outcomes from several adjuvant breast cancer trials demonstrated that AA patients fared worse than other patients with respect to disease-free survival (DFS), overall survival (OS), and estimated cause-specific survival after adjusting for age, receptor status, number of positive lymph nodes (LNs), tumor size, socioeconomic status, and body mass index (BMI).2 Several other large studies have confirmed differences in OS3,4 and breast cancer–specific survival.5 However, controversy still exists regarding the etiology of this disparity. The individual contributions of socioeconomic factors, tumor biology, non–cancer-related mortality, and chemotherapy treatment quality (eg, dose-intensity, chemotherapy dose reductions, and delays between treatments) have been difficult to elucidate. Although the first three have received significant attention, less is known about the impact of treatment delivered and treatment quality on long-term outcome.
Recently, large clinical trials established that dose reductions, delays, or interruptions of chemotherapy for breast cancer can reduce its benefit.6–8 Nonetheless, these types of changes in planned treatment are common. In a nationwide retrospective study of dose-intensity among 20,799 patients in community oncology practices, Lyman et al9 found that 36.5% of patients received less than 85% of their planned cumulative dose; 24.9% experienced treatment delays longer than 7 days; and an aggregate of 55% failed to receive the threshold 85% relative dose-intensity (RDI). RDI is defined as the ratio of the delivered dose/time to planned dose/time.
AA women with breast cancer are more likely to experience delays in initial cancer diagnosis and subsequent treatment.10 Furthermore, disparities exist in the use and dose-intensity of chemotherapy, which in turn may contribute to differences in outcome.11–13 It has been postulated that the lower baseline WBC count observed in AA women may contribute to this because a low baseline WBC count is one of the strongest predictors of reduced dose-intensity.14 Other studies report that black race, in addition to older age and socioeconomic status, may be associated with nonadherence to tamoxifen therapy15; however, the literature evaluating compliance to oral therapies is sparse. Disparities in treatment quality among women enrolled onto cooperative group clinical trials have not been explored.
Our objective was to assess whether AA and white women with breast cancer differ in frequency of early discontinuation of treatment, treatment delay, and dose-intensity. We analyzed patients enrolled onto SWOG adjuvant breast cancer clinical trials where the treatment prescribed was homogeneous for all women and eligibility criteria for trial entry were standardized. Furthermore, we assessed the relationship between these indicators of chemotherapy treatment quality and their impact on disparities in survival.
Breast cancer patients from two SWOG historical phase III trials, S8814 (LN positive) and S8897 (LN negative), were analyzed. The treatment arms of interest included those in which patients were scheduled to receive six cycles of chemotherapy; the treatment arms were as follows: cyclophosphamide, doxorubicin, and fluorouracil for six cycles; cyclophosphamide, methotrexate, and fluorouracil for six cycles; cyclophosphamide, doxorubicin, and fluorouracil for six cycles followed by tamoxifen for 5 years; or cyclophosphamide, methotrexate, and fluorouracil for six cycles followed by tamoxifen for 5 years. (Note that treatment arms in which tamoxifen was administered concurrently with chemotherapy were excluded to avoid confounding with respect to treatment delay and early discontinuation of treatment.) All AA women enrolled and a random sample of white women matched by age (< 45 v 45 to 55 v 55 to 65 v > 65 years) and study protocol (S8814 v S8897) were selected for a detailed treatment chart review. A strategy of 1:1 matching was used as a result of feasibility (cost) issues for this time-intensive, retrospective data abstraction. Race (black/white) and ethnicity were defined by self-report in accordance with the National Cancer Institute reporting criteria at the time of enrollment. All comparisons discussed hereafter are between AA and white patients.
In a univariate analysis, baseline tumor characteristics, body-surface area (BSA), WBC, and absolute neutrophil count (ANC) were compared between AA and white women, as was RDI for each chemotherapeutic agent and for the combined regimen. RDI, adjusted for BSA (directly reported from the institutions), was calculated on a percentage scale (0% to 100%) and was defined as the ratio of the delivered dose/time to planned dose/time. In addition, rates of neutropenic fever and granulocyte colony-stimulating factor use were compared between the two groups.
The primary outcomes of interest were early discontinuation of treatment, treatment delay, and survival. Early discontinuation was defined as not completing the full six scheduled treatments. The proportions of patients with early discontinuation were compared by race. The complementary set of patients were those who received the complete six cycles of chemotherapy for this group; the proportion of patients with at least one treatment delay were compared by race, as were reasons for treatment delay. Finally, the proportions of patients experiencing the combined end point (either early discontinuation or ≥ one treatment delay) were compared by race.
The reasons for treatment delay were broadly categorized as treatment toxicity, low blood counts (WBC, RBC, or platelet), and missed appointment for any reason. The category of missed appointment was analyzed as a single item comprising logistics, patient refusal, and other conflicts. The composite score was created because the reasons are often difficult to separate (for instance, patients who did not arrive for their appointments may have had transportation issues [logistics] or may simply have refused without stating why).
OS was measured from the time of registration until death as a result of any cause. DFS was defined as the time from registration to local recurrence, distant recurrence, new breast primary, or death as a result of any cause. Ten-year survival rates by race were estimated using the Kaplan-Meier method.16 Covariate adjustment was performed via Cox regression modeling17 and included the major disease prognostic factors of hormone receptor status (positive v negative), tumor size (< 2 v ≥ 2 cm), axillary nodal status (positive v negative), and menopausal status (as defined by National Cancer Institute criteria), along with baseline ANC (continuous variable). The Cox regression models were stratified by chemotherapy regimen, so the results represent the overall mean effect between the two treatment regimens. An exploratory Cox regression analysis that included RDI (as a time-dependent covariate) and treatment delay as a result of missed appointments, along with the major disease prognostic factors, was also performed. Post hoc power calculations for a study of this size, follow-up, and survival indicate adequate power (≥ 80%) to detect a hazard ratio (HR) of 1.57 for DFS (approximate 10-year DFS rate of 75%) and adequate power to detect an HR of 1.67 for OS (approximate 10-year DFS of 80%).
A total of 317 matched pairs (634 total patients) were analyzed. All patients were enrolled at SWOG sites from July 28, 1989, through February 22, 1995. Age, study, and treatment type distributions are listed in Table 1. Age and study proportions are the same by race in accordance with the matching. Treatment arm was not a matching factor but was reasonably well balanced by race. AA patients had higher baseline BSA (P < .0001), lower baseline ANC (P = .0002), and lower baseline WBC (P = .0009) than white patients. AA patients were also more likely to have tumor size ≥ 2 cm (P = .01) and were less likely to be hormone receptor positive (P < .0001) but were similar to white patients with respect to nodal status and menopausal status.
Fifty-seven (9%) of the total of 634 patients had early discontinuation of treatment (received < six cycles). AA patients were marginally more likely to discontinue treatment early compared with white patients (35 patients, 11% v 22 patients, 7%, respectively; P = .07).
Patients without early discontinuation of treatment (ie, received all six cycles) were analyzed to assess the number of treatment delays by race. AA patients were marginally more likely than white patients to have had at least one treatment delay while on treatment (85% v 79%, respectively; P = .07). AA patients more often experienced either early discontinuation of treatment and/or at least one treatment delay compared with white patients (87% v 81%, respectively; P = .04).
There were no differences in RDI for any of the specific chemotherapy types (Table 2). Total RDI, which was calculated as the unweighted average of the RDI measures for each chemotherapy type, was also similar by race. Overall, 43% of patients had an RDI < 85%, with no differences by race (P = .57).
Similar proportions of patients by race experienced neutropenic fever, had at least one course of granulocyte colony-stimulating factor, or had at least one treatment delay as a result of toxicity or insufficient laboratory counts (Table 3). However, AA patients were significantly more likely than white patients to miss an appointment for any reason (19% v 9%, respectively; P = .0002).
DFS and OS were analyzed by race as indicated in Figures 1 and and2,2, respectively. Median follow-up time among patients still alive is 13.7 years (maximum, 16.6 years). The Kaplan-Meier estimate of 10-year DFS was 71% for AA patients and 78% for white patients; for OS, the 10-year estimates were 86% and 76%, respectively. In a univariate Cox regression analysis by race, AA patients had worse DFS than white patients (HR = 1.44; 95% CI, 1.08 to 1.90; P = .01) and worse OS than white patients (HR = 1.73; 95% CI, 1.25 to 2.40; P = .0009). After adjusting for receptor status (DFS, P = .24; OS, P = .28), tumor size (DFS, P = .16; OS, P = .03), menopausal status (DFS, P = .005; OS, P < .0001), nodal status (DFS, P = .19; OS, P = .09), and baseline ANC (DFS, P = .06; OS, P = .004) in a multivariate Cox regression model, AA patients still had worse DFS than white patients (HR = 1.56; 95% CI, 1.15 to 2.11; P = .005) and worse OS than white patients (HR = 1.95; 95% CI, 1.36 to 2.78; P = .0002). No evidence of deviation from proportional hazards was found for either DFS (P = .75) or OS (P = .96), supporting the use of Cox regression. No evidence of an interaction between race and tumor receptor status was observed either for DFS (P = .48) or OS (P = .36), suggesting a consistent effect of race within receptor strata, although, given the sample size and the high survival rate at 10 years, a negative finding in this context should not be overinterpreted. The inclusion of RDI as a time-dependent variable in the multivariate model had little impact on the estimates for racial disparities in DFS (HR = 1.56; 95% CI, 1.15 to 2.12; P = .004) or OS (HR = 1.96; 95% CI, 1.37 to 2.80; P = .0002). Finally, the addition of baseline BSA (or, alternatively, baseline BMI), baseline WBC, or any of the treatment quality variables to the multivariate survival models had minimal impact on survival estimates.
We demonstrated that AA patients enrolled onto cooperative group adjuvant breast cancer clinical trials had lower baseline WBC than white patients, but AA and white women had similar RDIs. AA women were more likely to experience early discontinuation of treatment and/or treatment delay, mostly because of an increase in missed appointments. The mean RDI was similar for both groups, with 43% of all women having an RDI of less than 85%. After adjusting for the components of chemotherapy treatment quality such as dose reductions and dose delays, BSA (or, alternatively, BMI), baseline WBC, and other known predictors of survival, AA women still had worse DFS and OS than white women.
We found that AA women with breast cancer had significantly lower baseline ANC and WBC counts than white women. This was similar to the results from other studies that found that WBC counts are 25% to 40% lower on average among AA patients than white patients.14,18 Racial differences in baseline WBC count may contribute to black patients' risk of receiving dose reductions or delays because low baseline WBC is one of the strongest predictors of an RDI of less than 85%. Again, however, despite the lower WBC and ANC counts in AA women, we did not find that AA women received lower total treatment doses compared with white women.
In our study and others, delays in treatment were common; however, we saw no association between race and dose delay, as noted earlier, or between dose delay and all-cause mortality (P = .18 for DFS and P = .23 for OS). Cumulative dose delays longer than 30 days have been found to increase the risk of relapse.7 In our sample, the delays may not have been long enough to affect survival; our study may have been underpowered to detect their effects; or under conditions that have resulted in delay in other studies, the patients in our sample may have been more likely to discontinue treatment entirely than to delay completing it. A better understanding of the barriers to complete and timely adjuvant treatment may lead to cost-effective interventions to target those at high risk for missed appointments and treatment delays and improve outcome for all women with breast cancer.
The relationship between race and dose-intensity in the treatment of breast cancer was assessed in a prior retrospective chart review study.11 AA patients received less chemotherapy than white patients (P = .03), and overweight status (BMI > 25 mg/kg2) seemed to modify the association of treatment with race. In a multivariable model, compared with normal- or low-weight white patients, normal- or low-weight AA patients were nearly four times as likely to receive lower dose treatment, and overweight AA patients were nearly 20 times as likely to receive lower dose treatment.11,19 Although we found that AA women had a higher baseline BSA (which is highly correlated with BMI), after adjusting for BSA, we did not find an association between race and RDI. However, higher BSA may contribute to recurrence risk by other mechanisms because studies have suggested that diet, exercise, and BMI may be associated with outcome in women with breast cancer.20,21
Despite the fact that the mean RDI of the cohort was 88%, 43% of all the women evaluated had an RDI of less than 85%. This is similar to a prior community-based study that found that only 45% of women received at least the threshold 85% RDI.9 The earliest adjuvant clinical trials showed that RDI was a predictor of survival, with patients receiving an RDI of less than 85% achieving substantially less benefit.22,23 Despite this, RDI is not routinely reported in clinical trials. Furthermore, randomized trials demonstrated that increasing the dose density of chemotherapy improved both DFS and OS.8,24 Few studies have evaluated the impact of RDI on survival with more modern chemotherapy regimens, and few trials report the percentage of patients who achieved more than 85% RDI; therefore, it is hard to know how these results compare with other clinical trials. A reassessment of this RDI cutoff may be warranted.
Despite controlling for treatment-related factors, known prognostic factors, and chemotherapy delivery, disparities in both DFS and OS were still observed. The reasons for this remain elusive. One possible explanation may involve differential rates of adherence to hormonal therapy because AA women may, on average, be less likely to complete a full 5-year course of adjuvant hormonal therapy and clinical trials do not usually measure hormonal therapy adherence.15 In addition, racial or ethnic differences in genes responsible for the metabolism of either chemotherapeutic agents or hormonal treatments may contribute to these findings, and this variability may affect both toxicity and effectiveness of the treatment.25,26 It is known that AA women with breast cancer, especially those who are premenopausal, seem to have a higher incidence of biologically more aggressive cancers that are basal-like or triple negative.27–29 This may be contributing to some of the disparity that we observed; however, one limitation to our study was the we were unable to look at this factor because of incomplete information on HER-2/neu status. Survival differences by race were also observed for postmenopausal women and women with hormone receptor–positive tumors in our study, as well as others.2 Thus, our study refutes claims that the reason for racial disparities in adjuvant therapy outcomes is solely a result of frequency of triple-negative disease in the cohort analyzed.
Our study is unique in that patients were treated in a uniform fashion, detailed information on tumor characteristics was available, prospective evaluation of outcome was recorded, and the maximum follow-up time exceeded 15 years. Selection bias may limit the generalizability of the findings because patients who agree to participate in clinical trials are often more compliant and may be observed more closely than patients treated in the community and may be more adherent to tamoxifen therapy as well.
In summary, we have shown that AA women with breast cancer have lower WBC counts at baseline than white women. This did not result in a difference in RDI between the groups. However, AA women were more likely to experience early discontinuation or treatment delay, mostly as a result of missed appointments as opposed to toxicity or insufficient blood counts. Because we were unable to demonstrate that any factor related to treatment quality or delivery contributed to the racial difference in survival between the groups, further investigation into the etiology of this disparity is justified.
Supported in part by Public Health Service Cooperative Agreement Grants No. CA32102, CA38926, CA13612, CA68183, CA37981, and CA46282 awarded by the National Cancer Institute, Department of Health and Human Services. This research was also supported in part by a grant from the HOPE Foundation (D.L.H.), an American Society of Clinical Oncology Advanced Clinical Research Award (D.L.H.), and Grant No. RSGT-08-009-01-CPHPS from the American Cancer Society.
Presented in part at the 29th Annual San Antonio Breast Cancer Symposium, December 14-17, 2006, San Antonio, TX.
Authors' disclosures of potential conflicts of interest and author contributions are found at the end of this article.
The author(s) indicated no potential conflicts of interest.
Conception and design: Dawn L. Hershman, Joseph M. Unger, Robert B. Livingston, Kathy S. Albain
Financial support: Dawn L. Hershman
Administrative support: Dawn L. Hershman
Provision of study materials or patients: Laura F. Hutchins, C. Kent Osborne, Kathy S. Albain
Collection and assembly of data: Dawn L. Hershman, Joseph M. Unger, Laura F. Hutchins, Kathy S. Albain
Data analysis and interpretation: Dawn L. Hershman, Joseph M. Unger, William E. Barlow, Silvana Martino, Robert B. Livingston, Kathy S. Albain
Manuscript writing: Dawn L. Hershman, Joseph M. Unger, William E. Barlow, Kathy S. Albain
Final approval of manuscript: Dawn L. Hershman, Joseph M. Unger, William E. Barlow, Laura F. Hutchins, Silvana Martino, C. Kent Osborne, Robert B. Livingston, Kathy S. Albain