- Email: [email protected]
Comparative Effectiveness Analysis of Monotherapy With Cytotoxic Agents in Triple-negative Metastatic Breast Cancer in a Community Setting
Clinical Therapeutics/Volume 37, Number 1, 2015
Comparative Effectiveness Analysis of Monotherapy With Cytotoxic Agents in Triple-negative Metastatic Breast Cancer in a Community Setting George Dranitsaris, BPharm, PhD1; Stefan Gluck, MD2; Claudio Faria, Pharm D, MPH3; David Cox, PhD3; and Hope Rugo, MD4 1
Augmentium Pharma Consulting Inc, Toronto, Ontario, Canada; 2Sylvester Comprehensive Care Center, Leonard M. Miller School of Medicine, University of Miami, Miami, Florida; 3Eisai Inc, Woodcliff Lake, New Jersey; and 4Helen Diller Family Comprehensive Cancer Center, University of California, San Francisco, California ABSTRACT Purpose: There has been considerable progress in the treatment of metastatic breast cancer. However, the identification of optimal cytotoxic agents in patients with triple-negative breast cancer (TNBC) (negative for hormone receptors and human epidermal growth factor receptor 2) remains a therapeutic challenge. We conducted a comparative effectiveness analysis of 4 cytotoxic agents in patients with TNBC. Methods: We retrospectively identified patients who received single-agent chemotherapy with eribulin, capecitabine, gemcitabine, or vinorelbine from 19 community oncology clinics across the United States. Data collection included baseline patient and disease characteristics, prior therapies, performance status, duration of current therapy, growth-factor use and other supportive care, and dose-limiting toxicities and associated dose reductions or delays or skipped doses. Time to treatment failure (TTF) was measured from the first cycle of chemotherapy until disease progression, discontinuation due to toxicity, or death. TTF was estimated using the Kaplan-Meier method and Cox proportional hazards modeling adjusted for clustering on the practice site. To control for selection bias, which is inherent in observational studies, a propensity score–weighted TTF analysis was also conducted. Findings: Data from 225 patients were included in the analysis (eribulin, 47 patients; capecitabine, 69; gemcitabine, 56; and vinorelbine, 53). The median age of each group was o60 years, with the exception of the gemcitabine group (63 years). The 4 groups were comparable with respect to age, performance status, duration of disease-free survival, presence of
134
comorbidities, and hemoglobin level before the start of chemotherapy. Median lines of therapy of eribulin, capecitabine, gemcitabine, and vinorelbine and were 4th, 2nd, 3rd, and 3rd, respectively. The median durations of treatment were 2 months with eribulin, capecitabine, and gemcitabine compared with 1.6 months with vinorelbine. Using eribulin as the reference drug, and with adjustment for line of therapy and associated prognostic factors, the propensity score– weighted Cox regression analysis did not identify significant between-treatment differences in TTF (hazard ratios [95% CI] vs eribulin: capecitabine, 1.15 [0.75–1.76]; gemcitabine, 0.62 [0.34–1.13]; and vinorelbine, 1.0 [0.67–1.67]). Implications: In this assessment of patients with TNBC treated in a community oncology setting, eribulin was utilized in later lines compared with the other agents. However, comparable drug activity was reported among the 4 agents. (Clin Ther. 2015;37:134–144) & 2015 Elsevier HS Journals, Inc. All rights reserved. Key words: eribulin, capecitabine, gemcitabine, metastatic breast cancer, triple negative, vinorelbine.
INTRODUCTION Despite recent advances in its early detection and treatment, breast cancer remains one of the leading causes of death in women. In 2013 the United States Accepted for publication October 29, 2014. http://dx.doi.org/10.1016/j.clinthera.2014.10.023 0149-2918/$ - see front matter & 2015 Elsevier HS Journals, Inc. All rights reserved.
Volume 37 Number 1
G. Dranitsaris et al. alone, there were 232,340 new cases of breast cancer diagnosed and 39,620 deaths from the disease.1 However, 5-year survival rates after a diagnosis of early stage breast cancer have increased, from 75% in the 1970s to approximately 90% in 2013.1 This increase has been largely due to improved options focused on specific tumor biology, with availability and widespread use of hormonal and cytotoxic therapies, as well as targeted agents (TAs), with treatment benefit extending into the metastatic setting.2–4 On closer inspection of the data, gains in survival have primarily been realized in patients whose tumors express hormone receptors or human epidermal growth factor receptor (HER)-2. In contrast, options are more limited for patients whose tumors are negative for hormone receptors and HER-2 (triple-negative breast cancer [TNBC]), in whom hormone and HER-2 TAs are ineffective. This heterogeneous group encompasses 15% of all new diagnoses.5 One of the distinctive characteristics of TNBC is the rapid development of, or de novo, resistance to chemotherapy, leading to shorter disease-free and overall survival (OS).5,6 In a cohort study in 1601 patients with breast cancer who were followed up for a median of 8.1 years, patients with TNBC were at a higher risk for distant recurrences (hazard ratio [HR] = 2.6; P o 0.05) and death (HR = 3.2; P o 0.05) within 5 years of the initial diagnosis.7 After a diagnosis of metastatic disease, patients in the TNBC cohort had a significantly shorter OS compared with the reference population (9 vs 22 months; P o 0.05).7 Clearly, the treatment of TNBC represents a major therapeutic challenge and is an active area of clinical research. Bevacizumab is a monoclonal antibody that targets vascular endothelial growth factor. There is some experience with bevacizumab in combination with chemotherapy in TNBC. In a subgroup analysis of data from RIBBON-2 (A Randomized, Double-Blind, Placebo-Controlled, Phase III Trial Evaluating the Efficacy and Safety of Bevacizumab in Combination With Chemotherapy for Second-Line Treatment of Human Epidermal Growth Factor Receptor 2Negative Metastatic Breast Cancer),8 which evaluated various cytotoxic agents with and without bevacizumab in metastatic disease, patients with TNBC had a statistically significant improvement in progression-free survival (6.0 vs 2.7 months; P o 0.001) and a numerically greater but statistically similar OS. However, these results have yet to be
January 2015
confirmed in an adequately powered Phase III randomized comparative trial. Therefore, it is unlikely that bevacizumab will address the current unmet need that is TNBC. Other agents being investigated in clinical trials include anti–vascular endothelial growth factor receptor tyrosine kinase inhibitors, anti–epidermal growth factor receptor therapies, poly (adenosine diphosphate ribose) polymerase inhibitors, Src tyrosine kinase inhibitors, and oral mammalian target of rapamycin inhibitors.5,9 The most advanced of these agents is iniparib, with a randomized Phase II trial reporting statistically significant improvements in objective response, progression-free survival, and most importantly OS (7.7 vs 12.3 months; HR = 0.57; P = 0.01).10 Iniparib was then taken into Phase III development in a trial that mimicked the Phase II study. In that randomized, controlled trial (RCT), the drug failed to meet its progression-free survival and OS end points, and further development of iniparib was halted.11 Because there are no specifically targeted chemotherapeutic drugs with proven clinical benefit in TNBC, patients typically are treated with the same agents used in other breast cancer subgroups. Patients with metastatic TNBC would be offered single-agent or combination chemotherapy containing an anthracycline or taxane agents, capecitabine, gemcitabine, vinorelbine, or carbo- or cisplatin.6,9 However, some results with eribulin were reported in a subset analysis of data from the Phase III EMBRACE (Eribulin Monotherapy Versus Treatment of Physician’s Choice in Patients With Metastatic Breast Cancer) trial.12 In that global study, which enrolled 762 patients into an eribulin arm or a physician’s-choice comparator arm, the experimental therapy was reported to have had an OS benefit (HR ¼ 0.81; P ¼ 0.041); 74% of enrolled patients were HER-2 negative, and 19% had TNBC. Eribulin was more effective in both hormone-negative and TNBC patients than in the control group. However, these intriguing findings have yet to be confirmed in an RCT. Although ongoing and planned trials are evaluating a variety of TAs for the treatment of patients with metastatic TNBC, clinicians need immediate information that will help to guide medical decision making. Comparative effectiveness studies, through observational data, are a reasonable approach to assessing treatment effects in a clinical practice setting.13,14 Several examples illustrate the concordance between
135
Clinical Therapeutics observational studies and the findings from subsequent randomized trials.15,16 In the present study, observational data collected from US community oncology practices were used for measuring differences in drug toxicity, health care resource utilization, and Time to treatment failure (TTF) between 4 antineoplastic agents in a sample of patients with metastatic TNBC.
PATIENTS AND METHODS Inclusion and Exclusion Criteria This retrospective, observational study was conducted in patients with metastatic TNBC treated with eribulin, capecitabine, gemcitabine, or vinorelbine between November 1, 2010, and August 31, 2013. All patients were treated in 1 of 19 community oncology practices that are a part of the Cancer Clinics of Excellence network. To be entered into the study, patients must have received 1 of the 4 agents for the treatment of advanced disease as a part of routine clinical practice and according to institutional administration guidelines. Each drug must have been given as single-agent chemotherapy; the addition of a TA was permitted. Eligible patients were at least 18 years of age and had a diagnosis of Stage IV breast cancer, with chemotherapy for advanced disease initiated after November 1, 2010. Patients were excluded if they were enrolled in a clinical trial of an investigational anticancer agent, but those enrolled in a clinical trial evaluating a supportive care agent were eligible.
Data Collection Baseline data collection consisted of demographic and disease characteristics, existing comorbidity as assessed by the Charlson comorbidity index,17 duration of treatment with a previous line(s) of chemotherapy, number and site(s) of metastases, Eastern Cooperative Oncology Group performance status before the start of chemotherapy, and the addition of TAs. From the first cycle of single-agent chemotherapy, clinical data collection consisted of baseline hemoglobin concentration; white blood cell, neutrophil, and platelet counts; the use of colonystimulating factors or antiemetic agents; total number of cycles delivered; and severe dose-limiting toxicities. Severe toxicities were events severe enough to require an intervention such as pharmacotherapy, hospitalization, a visit to an emergency department, or an unscheduled clinic visit.
136
Health care resource data collection included the starting dose of each anticancer drug administered; number of doses; total number of cycles delivered; number of premature discontinuations of chemotherapy; standard biochemistry and diagnostic test utilization; the use of prophylactic antibiotic agents, growth factors, serotonin antiemetic agents, and/or bone TAs; total number of red blood cell units and platelets administered; as well as hospitalizations, visits to an emergency department, or unscheduled clinic visits as a result of a treatment-related toxicity. The probable or possible cause of each clinically relevant event (eg, febrile neutropenia) was also recorded.
End Points The primary end point was TTF as a surrogate for disease progression. TTF was measured as the time from first dose of chemotherapy until the start of the next line of active anticancer treatment, the permanent discontinuation of therapy due to toxicity or death from any cause. Secondary end points included health care resource utilization, treatment-related toxicities, and factors related to treatment choice.
Statistical Analysis All outcomes data are presented descriptively as means, medians, or percentages. Given the observational nature of the study, no formal sample size or power analysis was undertaken. Therefore, the most relevant outcomes from the statistical analysis are the 95% confidence intervals (CIs). P values should be viewed as being of secondary importance. One of the challenges associated with observational studies (unlike randomized trials) is patient-selection bias. Systematic factors such as self-selection, physician selection, or geographic selection can lead to patients’ receiving one therapy over another. Such bias can affect the results of any comparative analysis in an observational study. To overcome this challenge, patient groups need to be adjusted for the potential impact of selection bias. In the present study, selection bias was addressed by a series of reevaluations characterized by propensity score analysis.18 Propensity score analysis is a 2-phase technique used for estimating a treatment effect in comparative groups selected by nonrandom means. In the first phase of a propensity score analysis, variables that influence selection to group assignment are used for
Volume 37 Number 1
G. Dranitsaris et al. modeling the probability of receiving a treatment (or of being in a reference group—in this case, the eribulin group).19,20 The resulting probability is called the propensity score. In the second phase, the propensity score is used for adjusting for preexisting group differences in the analysis of the relevant outcomes. There are several ways to use propensity scores, such as stratification variables, matching patients based on their propensity score, or their use as a weighting or adjustment variable in multivariate analysis.19,20 Assuming that all relevant covariates are included in the propensity score model, the group effect observed in a propensity score analysis represents an unbiased estimate of the true treatment effect.18,19 In the present study, any baseline covariate with a marginal association with treatment group (P o 0.10) was considered for inclusion in the final propensity score model. Because there were 4 groups, a maineffects multinomial logistic regression model was developed, using variables identified in the previous step, and retained after a backward elimination process with P o 0.10. The final propensity scores derived for each patient from the multinomial logistic regression model were used in the subsequent Kaplan-Meier survival analyses for adjustment for potential selection bias. The initial Kaplan-Meier analysis was a nonweighted comparison between the 4 groups. The propensity scores determined for each patient were then used as a weighting factor in the subsequent reevaluation. Survival curves for TTF were weighted by the estimated propensity scores, generated by the method of Kaplan-Meier and compared with the log-rank test. In a supporting analysis, a propensity score–weighted Cox proportional hazards regression analysis was also conducted to assess the relative risk for treatment failure in patients who received capecitabine, gemcitabine, or vinorelbine relative to eribulin. Variables evaluated in the Cox regression analysis included age, sex, performance status, time to diagnosis of advanced disease, prior adjuvant chemotherapy, existing comorbidities, site(s) of metastatic disease, the addition of TAs, and hematologic parameters at baseline. Independent baseline prognostic variables with a P value o0.05 were retained in the final regression model via backward elimination. The final regression model was also adjusted for clustering on the practice site to avoid violation of the independence assumption. As a final analysis, multinomial logistic regression analysis was applied to identify factors that were associated
January 2015
with the selection of capecitabine, gemcitabine, or vinorelbine over eribulin. All of the statistical analyses were performed using Stata release 11.0 (StataCorp LP, College Station, Texas).
RESULTS Study Population We identified 225 patients who met the inclusion criteria (eribulin, 47 patients; capecitabine, 69; gemcitabine, 56; and vinorelbine, 53). In this observational study, there were expected imbalances in some demographic and clinical characteristics (Table I). The median age of each group was o60 years, with the exception of the gemcitabine group (63 years). Eribulin was administered as 4th-line therapy compared with 2nd, 3rd, and 3rd for capecitabine, gemcitabine, and vinorelbine, respectively. The median duration of treatment was 2 months with eribulin, capecitabine, and gemcitabine compared with 1.6 months with vinorelbine. In addition, the gemcitabine group had a lower proportion of white patients and a longer median time from initial diagnosis to the development of metastatic disease. In general, hematologic parameters before the start of chemotherapy were comparable and within normal limits. The median number of prior chemotherapies varied between the treatment groups. The eribulin group received a median of 3 prior lines of chemotherapy compared with 1, 2, and 2 in the capecitabine, gemcitabine, and vinorelbine groups, respectively. There were also differences in the proportions of patients who received a TA in combination with single-agent chemotherapy—2.1% of patients received a TA with eribulin compared with 13.0%, 16.1%, and 9.4% with capecitabine, gemcitabine, and vinorelbine. The most commonly used TA was bevacizumab; a total of 5 patients received trastuzumab (Table I). The median numbers of treatment cycles were similar between the treatment groups, with 4 cycles of eribulin and 3 cycles of capecitabine, gemcitabine, and vinorelbine. There were numeric between-group differences in the percentages of patients with dose reductions or delays or missed doses, but these differences failed to reach statistical significance. At the completion of the course of therapy, patients treated with gemcitabine were significantly less likely to have received a next line of therapy (p o 0.05).
137
Clinical Therapeutics Table I. Demographic and clinical characteristics of the patients* in this comparative study of the effectiveness of monotherapy with cytotoxic agents in triple-negative metastatic breast cancer in a community setting. Characteristic Demographic characteristics Age, median (range), y Race, % White Black Other Clinical characteristics Weight, mean (range), lbs BSA, mean (range) Biochemistry, mean (SD) Hb, g/dL WBC, cells 109/L ANC, cells 109/L Platelets, cells 109/L Creatinine, mg/dL Disease characteristics ECOG PS, % 0 or 1 2 or 3 Not documented Charlson comorbidity score, median (range)† Time from diagnosis to metastatic disease, median (range), y Metastatic site, % Lung Liver Bone Brain Treatment history No. of treatments (inclusive), median (range) Chemotherapy Hormone therapy Treatments received, % Taxane agents Adjuvant therapy Anthracycline agent Capecitabine Current regimen‡ Single agent TA agent added§ No. of cycles, median (range) Treatment duration, median (range), mo Dose reductions, delays, or skipped doses, % Delays, median (range)
Eribulin (n ¼ 47)
Capecitabine (n ¼ 69)
Gemcitabine (n ¼ 56)
Vinorelbine (n ¼ 53)
54 (33–72)
59 (29–85)
63 (29–86)
59 (36–85)
69.6 24.6 5.8
68.1 28.0 4.3
56.1 29.3 14.6
81.0 16.7 2.3
68.9 (39–124) 1.8 (1.3–2.3)
70.8 (36.3–134) 1.8 (1.3–2.3)
72.1 (38.1–134) 1.8 (1.3–2.3)
72.1 (39.9–134) 1.7 (1.4–2.2)
11.0 7.6 5.6 246 0.80
11.6 6.2 4.6 253 0.83
11.5 6.9 5.1 256 0.96
11.4 8.2 5.6 266 0.84
(1.4) (3.8) (3.3) (112) (0.33)
68.1 23.4 8.5 6 (6–8) 1.8 (0–17.3)
(1.3) (2.8) (3.1) (100) (0.36)
63.8 23.2 13.0 6 (6–13) 1.1 (0–11.5)
(1.5) (4.4) (3.9) (124) (0.76)
64.3 19.6 16.1 6 (6–12) 2 (0–30.2)
(1.6) (7.9) (4.0) (106) (0.37)
54.7 30.2 15.6 6 (6–10) 1.7 (0–17.3)
40.4 29.8 25.5 10.6
40.6 24.6 40.6 4.4
46.4 26.8 32.1 8.9
32.1 18.9 26.4 1.9
4 (1–7) 3 (0–6) 0 (0–1)
2 (1–6) 1 (0–5) 0 (0–1)
3 (1–8) 2 (0–7) 0 (0–2)
3 (1–6) 2 (0–5) 0 (0–1)
91.5 78.7 59.5 57.4
66.7 69.6 14.5 0
85.7 71.4 39.3 34.0
86.8 81.1 34.0 45.2
97.9 2.1 4 (1–14) 2.1 (0–9.8) 38.3 0 ( 0–2)
87.0 13.0 3 (1–18) 2.0 (0.2–15.8) 34.8 0 (0–3)
83.9 16.1 3 (1–14) 2.1 (0–9.0) 42.9 0 (0–2)
90.6 9.4 3 (1–21) 1.6 (0–13.5) 45.3 0.5 (0–13) (continued)
138
Volume 37 Number 1
G. Dranitsaris et al. Table I. (continued).
Characteristic Reductions, median (range) Skipped, median (range) Received a next line of anticancer therapy, %
Eribulin (n ¼ 47) 0 (0–2) 0 (0–4) 66.0
Capecitabine (n ¼ 69) 1 (0–2) 0 (0–7) 63.8
Gemcitabine (n ¼ 56)
Vinorelbine (n ¼ 53)
0.5 (0–6) 1 (0–7) 44.6
0 (0–1) 1 (0–7) 73.6
ANC ¼ absolute neutrophil count; BSA ¼ body surface area; ECOG PS ¼ Eastern Oncology Cooperative Group performance status; Hb ¼ hemoglobin; TA ¼ targeted agents; WBC ¼ white blood cells. * All patients had HER-2– and hormone receptor–negative disease. † The weighted comorbidity classifications were as follows: 0 ¼ low; 1 or 2 ¼ intermediate; 3 or 4 ¼ high; and Z5 ¼ very high. ‡ Capecitabine: 1 g/m2 PO BID on days 1–14 q21d, 36.8%; 1.5 g/m2 PO BID on days 1–14 q21d, 12.1%; and other, 51.1%. Eribulin: 1.4 mg/m2 on days 1 and 8 q21d, 91.5%; 1.1 mg/m2 on days 1 and 8 q21d, 4.2%; and other, 4.2%. Gemcitabine: 1 g/m2 on days 1, 8, and 15 q28d, 22.0%; 1 g/m2 on days d1 and d8 q21d, 24.4%, and other, 53.6%. Vinorelbine regimens: 25 mg/m2 on day 1 q7d, 23.8%; 25 mg/m2 on days 1, 8, and d 15 q28d, 23.8%; and other, 52.4%. § In capecitabine patients, 6 patients received bevacizumab and 2 patients received trastuzumab. In eribulin and gemcitabine patients, one and nine patients received bevacizumab respectively. With vinorelbine, 3 patients received bevacizumab and two received trastuzumab.
Approximately two thirds of patients treated with eribulin went on to receive a 5th line of therapy (Table I).
Time to Treatment Failure The initial Kaplan-Meier survival analysis was not weighted by patients’ individual propensity scores (Figure 1). Using eribulin as the reference group, there were no significant differences between groups. Overall, the median TTF was 3 to 4 months with eribulin, capecitabine, and gemcitabine, with vinorelbine being 2.7 months. Eribulin, despite being given in as a later line of therapy (4th vs 2nd/3rd), provided comparable TTF benefits (hazard ratios [95% CI] vs eribulin: capecitabine, 1.15 [0.75–1.76]; gemcitabine, 0.62 [0.34–1.13]; and vinorelbine, 1.0 [0.67–1.67]). Given the potential for selection bias with observational data, the TTF analysis was then weighted by individual patients’ propensity scores. The final propensity scores incorporated factors such as age, prior therapies, race, and line of therapy. The findings of the weighted analysis did not alter the initial observations. There were no significant differences in TTF with capecitabine, gemcitabine, or vinorelbine relative to eribulin (Figure 2). These findings were further supplemented with a propensity score–weighted Cox proportional hazards analysis of TTF. Statistically
January 2015
significant covariates that were retained in the final multivariate model included the presence of bone metastases, baseline absolute neutrophil count, and creatinine concentration (Table II). Even with the inclusion of these covariates in the weighted model, the group differences relative to eribulin were not statistically significant.
Supportive Care and Treatment-Related Toxicity Relative to other agents, significantly fewer patients treated with capecitabine received myeloid growth factor support and serotonin antiemetic agents (P o 0.01). In contrast, there were no significant differences in the proportions of patients receiving erythropoietinstimulating agents or bone TAs (Table III). Significantly greater percentages of patients in the eribulin and vinorelbine groups reported Z1 toxicity event compared with those treated with capecitabine (both, P o 0.05). The most common dose-limiting toxicities with eribulin use were anemia (36.2%), neutropenia (31.2%), and neuropathy (19.4%). Similarly, the events related to myelosuppression were also the toxicities most commonly reported with vinorelbine and gemcitabine use. Patients treated with capecitabine were more likely to have developed hand/ foot skin reaction (37.7%) and/or diarrhea (29.4%) (Table III).
139
Clinical Therapeutics
Table II. Propensity score weighted Cox proportional hazards analysis of TTF in this comparative study of the effectiveness of monotherapy with cytotoxic agents in triple-negative metastatic breast cancer in a community setting.* Variable CAP vs ERB GEM vs ERB VIN vs ERB Bone metastases Baseline ANC Baseline creatinine
HR (95% CI) 1.15 0.62 1.0 1.49 1.07 1.79
(0.75–1.76) (0.34–1.13) (0.60–1.67) (1.20–1.84) (1.03–1.12) (1.08–2.97)
P
Impact TTF
0.53 0.12 0.99 o0.001 0.002 0.024
NS NS NS ↑ Risk by 49% ↑ Risk, ↑ ANC ↑ Risk, ↑ creatinine
ANC ¼ absolute neutrophil count; CAP ¼ capecitabine; ERB ¼ eribulin; GEM ¼ gemcitabine; HR ¼ hazard ratio; NS = not significant; TTF ¼ time to treatment failure; VIN ¼ vinorelbine. * Analysis was also adjusted for clustering on cancer clinic.
Similar percentages of patients receiving eribulin or gemcitabine required a medical intervention to treat an adverse event (23.4% and 17.8%, respectively). These interventions consisted of blood product transfusions, an unplanned visit to a clinic or an emergency department visit, and/or hospital admission. Conversely, 11.6% of the capecitabine group received a medical intervention for an adverse event, although the number of treatment discontinuations due to toxicity was highest with capecitabine, at 11.6%. The lowest rate of discontinuation due to an adverse event was 2.1% (eribulin). With all 4 drugs, the most common reason for the discontinuation of therapy was disease progression (Table III).
Factors Associated With Treatment Choice
The final investigation performed in the current study was multinomial logistic regression analysis to identify factors associated with the selection of the 3 chemotherapeutic agents compared with eribulin. Capecitabine was more likely to have been prescribed than was eribulin in older patients, in those with bone metastases, and in patients with higher baseline hemoglobin levels. In contrast, capecitabine was less likely to have been prescribed in patients with longer intervals of disease-free survival and as a later line of therapy (Table IV). Similarly, gemcitabine and vinorelbine were more likely to have been prescribed in older patients and as earlier lines of therapy. In summary, the multinomial analysis
140
identified patient age and line of therapy to be the 2 most important factors in the selection of capecitabine, gemcitabine, or vinorelbine over eribulin. In general, the likelihood of using eribulin over the other the agents decreased by 4 to 5% with each additional year of patient age. Hence, eribulin was more likely to have been prescribed in younger patients with TNBC and as a later line of therapy.
DISCUSSION In chemotherapy decision making, RCTs remain the “gold standard” for providing data on the efficacy and tolerability of new treatments. The major advantage of RCTs is their ability to balance both known and unknown confounding variables between the experimental and control groups and to incorporate double-blinding into the assessment of outcomes. As a result, most of the known biases that can affect treatment outcomes are either eliminated or minimized in a well-designed RCT. Nonetheless, RCTs are both resource and time intensive to undertake because they often require large sample sizes in highly select patient populations. This is particularly relevant in oncology, in which it is not uncommon for trial sample sizes to exceed 1000 patients, and often in less prevalent patient subgroups, such as TNBC. Hence, the results from RCTs may not be available until several years after the trial has been initiated. In other situations, a drug may be off-patent (or soon to go offpatent), so there would be no commercial incentive for
Volume 37 Number 1
G. Dranitsaris et al.
1.00 0.75 0.50 0.25 0.00 0
10
20 30 Time (mo) Eribulin Gemcitabine
40
50
Capecitabine Vinorelbine
Figure 1. Kaplan-Meier curve of time to treatment failure (TTF) from the start of chemotherapy in this comparative study of the effectiveness of monotherapy with cytotoxic agents in triplenegative metastatic breast cancer in a community setting. Median (interquartile range) TTF values (in months): eribulin, 3.2 (2.5–5.3); capecitabine, 3.4 (2.1–5.3); gemcitabine, 4.1 (3.0–8.4); and vinorelbine, 2.7 (1.8–4.6). Univariate hazard ratios (95% CI; P): capecitabine versus eribulin, 1.00 (0.70–1.51; 0.88); gemcitabine versus eribulin, 0.64 (0.38– 1.05; 0.08); and vinorelbine versus eribulin, 1.10 (0.69–1.75; 0.69) (with no baseline covariate adjustment).
the brand manufacturer to initiate a costly RCT. Therefore, alternative methods of data generation that provide robust information for medical decision making need to be identified. Comparative effectiveness research is a relatively new field, in which existing health care interventions are evaluated to determine which provide optimal benefits for patients, with the least harm. A main objective of comparative effectiveness research is to determine which treatment works best, in which patients, and in what circumstances in a clinical setting.21 In the present study, a comparative effectiveness analysis was undertaken to compare drug toxicity, health care resource utilization, and TTF between 4 cytotoxic agents widely used in community practice in patients with metastatic TNBC.
January 2015
The findings suggest comparative effectiveness, in terms of TTF, between eribulin, capecitabine, gemcitabine, and vinorelbine, despite the former agent being used in a later line of therapy. These findings were consistent after the data were adjusted by propensity score–weighted Cox multivariate analysis. Eribulin, even as a 4th line of therapy, offered a TTF benefit comparable to that of capecitabine, gemcitabine, and vinorelbine in patients with TNBC in this community-based dataset. In addition, a lesser percentage of patients treated with eribulin had permanent treatment discontinuation because of toxicity. The multinomial logistic regression analysis also yielded some results on key factors associated with the
1.00 0.75 0.50 0.25 0.00 0
10
20 30 Time (mo) Eribulin Gemcitabine
40
50
Capecitabine Vinorelbine
Figure 2. Propensity score–weighted time to treatment failure (TTF) from the start of chemotherapy in this comparative study of the effectiveness of monotherapy with cytotoxic agents in triple-negative metastatic breast cancer in a community setting. Median (interquartile range) TTF values (in months): eribulin ¼ 3.2 (2.5–5.3); capecitabine ¼ 2.9 (1.9–4.1); gemcitabine ¼ 4.1 (3.2–7.0); and vinorelbine ¼ 2.7 (1.8–4.4). Univariate hazard ratios (95% CI; P): capecitabine versus eribulin, 1.32 (0.90–1.95; 0.16); gemcitabine versus eribulin, 0.69 (0.43–1.15; 0.16); and vinorelbine versus eribulin, 1.16 (0.71–1.88; 0.55) (with no baseline covariate adjustment).
141
Clinical Therapeutics
Table III. Supportive care and tolerability during current therapy in this comparative study of the effectiveness of monotherapy with cytotoxic agents in triple-negative metastatic breast cancer in a community setting. Parameter Supportive care, % G-CSF 5-HT3 antiemetic Bone targeted agent ESA Prophylactic antibiotic Tolerability Patients with Z1 AE, no. (%) Treatment-emergent AE (any grade), %* Anemia Neutropenia Neuropathy Emesis Edema Thrombocytopenia Elevated liver enzymes Stomatitis Rash Febrile neutropenia Bleeding events Hand/foot skin reaction Diarrhea Medical interventions for toxicity, no. (%) No. of interventions, by type† Blood product transfusion Unplanned clinic visit Hospital admission Emergency department visit Supportive care medication prescribed Second medical intervention for another toxicity Reason therapy was stopped, % Disease progression Went to hospice Disease stabilization Other Toxicity Patient wish Death Disease response
Eribulin (n ¼ 47)
Capecitabine (n ¼ 69)
Gemcitabine (n ¼ 56)
Vinorelbine (n ¼ 53)
55.3 53.8 25.5 14.9 12.5
4.3 7.2 37.7 15.9 0
26.8 71.4 25.0 17.8 0
34.0 58.5 15.1 26.4 5.7
33 (70.2)
35 (55.1)
34 (60.7)
39 (73.6)
36.2 31.2 19.4 10.6 7.1 6.4 6.4 6.3 2.1 2.1 2.1 0 0 9 (23.4)
15.9 4.3 2.9 4.3 4.3 7.2 1.4 10.1 5.9 0 0 37.7 29.4 8 (11.6)
32.1 32.1 1.8 1.8 0 23.2 5.4 3.6 1.8 1.8 1.8 0 1.8 10 (17.8)
45.3 45.3 3.8 7.5 3.8 3.8 5.7 0 1.9 1.9 0 0 3.8 4 (7.1)
9 1 1 1 0 4
2 5 2 1 3 0
7 3 2 1 3 2
2 1 2 0 1 2
74.5 8.5 6.4 4.2 2.1 2.1 2.1 0
73.9 4.3 0 1.4 11.6 2.9 1.4 4.3
46.4 3.6 3.6 25.0 7.1 10.7 3.6 0
73.6 5.7 0 9.3 5.7 1.9 1.9 1.9
5-HT3 ¼ serotonin; AE ¼ adverse event; ESA ¼ erythropoietin-stimulating agents; G-CSF ¼ granulocyte colony-stimulating factor. * Some patients experienced 41 event. † Multiple types of interventions were given for a single event.
use of the alternatives to eribulin. Capecitabine, gemcitabine, and vinorelbine were more likely to have been prescribed in older patients and in earlier lines of therapy. The latter finding is likely related to the fact that the pivotal trial of eribulin evaluated patients
142
who were heavily pretreated with an anthracycline and/or a taxane agent.12 However, the observation that eribulin was preferentially used in younger TNBC patients was unexpected and requires a qualitative investigation.
Volume 37 Number 1
A final comment has to be made about the relatively modest TTF benefits of only 3 to 4 months identified in the present study. This finding suggests that the benefit of cytotoxic therapy in later lines of treatment remains modest and that more effective therapy needs to be identified to treat this patient population.
2.0 0.10 NS 0.75 0.009 25% Less likely per line 1.23 0.19 NS 3.19 0.002 3-Fold more likely 0.51 o0.001 49% Less likely per line 1.43 0.03 ↑ Likelihood with ↑ Hb
Study Limitations
CAP ¼ capecitabine; ERB = eribulin; GEM ¼ gemcitabine; Hb ¼ hemoglobin; OR ¼ odds ratio; VIN ¼ vinorelbine. * Analysis was adjusted for clustering on cancer clinic.
1.83 0.12 NS 0.77 0.0051 23% Less likely 1.33 0.092 NS
NS 0.93 0.88 68% Less likely NS 0.47 0.74
0.32 0.009
1.04 0.036 ↑ Likelihood with ↑ age 0.96 0.52 NS ↑ Likelihood with ↑ age NS 0.023 ↑ Likelihood with ↑ age 0.003 23% Less likely per year 1.05 0.77
Age Time to metastatic disease Nodal metastases present Bone metastases present Line of therapy Baseline Hb
P OR
1.05 0.005 0.90 0.14
Likelihood of Receiving the Regimen
January 2015
Variable
Likelihood of Receiving the Regimen
OR
P
Likelihood of Receiving the Regimen
OR
P
VIN vs ERB GEM vs ERB CAP vs ERB
Table IV. Multinomial logistic regression analysis of factors associated with the selection of alternatives to eribulin (ERB) in this comparative study of the effectiveness of monotherapy with cytotoxic agents in triple-negative metastatic breast cancer in a community setting.*
G. Dranitsaris et al.
Because this study was retrospective and observational and not a randomized trial, there were between-group imbalances in several important prognostic factors, such as age, performance status, site(s) of metastatic disease, and recurrence-free survival. There was also an imbalance in prior therapies, which may have affected treatment selection secondary to drug toxicity. To address these biases, a propensity score– weighted evaluation and a multivariate analysis were undertaken to support the initial Kaplan-Meier survival analyses. Notwithstanding, these statistical techniques do not entirely eliminate all of the selection biases that are inherent in observational studies. The total sample size was small (n ¼ 225), and all regions of the United States as well as ethnic groups were not equally represented, which could compromise the generalizability of the results to other jurisdictions. There were differences in the concurrent use of TA, namely bevacizumab, between the 4 cytotoxic-agent groups. However, in the multivariate analysis, the addition of a TA was not retained in the final Cox regression model. Because the drugs were not given in the same line of therapy, OS could not be properly evaluated. As a result, we were able to measure differences only in TTF between the 4 drugs.
CONCLUSIONS In this assessment of the use of 4 cytotoxic agents in patients with metastatic TNBC treated in a community oncology setting, eribulin was utilized as a later line of therapy relative to capecitabine, gemcitabine, and vinorelbine. However, eribulin was reported to have had at least comparable drug activity and tolerability, even when used in more heavily pretreated patients. These findings warrant further analyses in a larger population with an evaluation of biologic heterogeneity.
ACKNOWLEDGMENTS
Eisai Inc. provided financial support to conduct this study. The corresponding author had full access to the data in the study and had the final responsibility for
143
Clinical Therapeutics the decision to submit the paper. George Dranitsaris, Stefan Gluck and Hope Rugo have acted as consultants to the sponsor.
CONFLICTS OF INTEREST This study and its publication were funded by Eisai Inc, the manufacturers of eribulin. Drs. Faria and Cox are employees of, and may hold stock options in, Eisai Inc. The authors have indicated that they have no other conflicts of interest with regard to the content of this article.
REFERENCES 1. Surveillance Epidemiology and End Results (SEER). National Cancer Institute 2011. Washington, USA, 2013 (http://seer.cancer.gov/). Accessed January 31, 2014. 2. Slamon DJ, Leyland-Jones B, Shak S, et al. Use of chemotherapy plus a monoclonal antibody against HER2 for metastatic breast cancer that overexpresses HER2. N Engl J Med. 2001;344:783–792. 3. Mouridsen H, Gershanovich, Sun Y, et al. Superior efficacy of letrozole versus tamoxifen as first line therapy for postmenopausal women with advanced breast cancer: Results of a phase III study of the International Letrozole Breast Cancer Group. J Clin Oncol. 2001;19:2596-06. 4. Verma S, Miles D, Gianni L, et al. Trastuzumab emtansine for HER2-positive advanced breast cancer. N Engl J Med. 2012;367:1783–1791. 5. Bayraktar S, Glück S. Molecularly targeted therapies for metastatic triple-negative breast cancer. Breast Cancer Res Treat. 2013;138:21–35. 6. André F, Zielinski CC. Optimal strategies for the treatment of metastatic triple-negative breast cancer with currently approved agents. Ann Oncol. 2012;23(Suppl 6):vi46–vi51. 7. Dent R, Trudeau M, Pritchard KI, et al. Triple negative breast cancer: clinical features and patterns of recurrence. Clin Cancer Res. 2007;13:4429–4434. 8. Brufsky A, Valero V, Tiangco B, et al. Second-line bevacizumab-containing therapy in patients with triplenegative breast cancer: subgroup analysis of the RIBBON-2 trial. Breast Cancer Res Treat. 2012;133:1067–1075.
9. Crown J, O’Shaughnessy J, Gullo G. Emerging targeted therapies in triple-negative breast cancer. Ann Oncol. 2012;23 (Suppl 6):vi56–vi65. 10. O’Shaughnessy J, Osborne C, Pippen JE, et al. Iniparib plus chemotherapy in metastatic triple-negative breast cancer. N Engl J Med. 2011;364:205–214. 11. O’Shaughnessy J, Schwartzberg LS, Danso MA, et al. A randomized phase III study of iniparib (BSI-201) in combination with gemcitabine/carboplatin (G/C) in metastatic triple negative breast cancer. J Clin Oncol. 2011;29:15s. Abstr 1007. 12. Cortes J, O’Shaughnessy J, Loesch D, et al. Eribulin monotherapy versus treatment of physician’s choice in patients with metastatic breast cancer (EMBRACE): a phase 3 open-label randomised study. Lancet. 2011;377: 914–923. 13. Carroll J. Payers step in with ‘real-world’ comparative effectiveness research. Manag Care. 2011;20:23–26. 14. Teutsch SM, Fielding JE. Comparative effectiveness–looking under the lamppost. JAMA. 2011;305:2225–2226. 15. Byers T, Sedjo RL. Does intentional weight loss reduce cancer risk? Diabetes Obes Metab. 2011;13:1063–1072. 16. Cooper K, Squires H, Carroll C, et al. Chemoprevention of colorectal cancer: systematic review and economic evaluation. Health Technol Assess. 2010;14:1–206. 17. Charlson ME, Pompei P, Ales KL, MacKenzie CR. A new method of classifying prognostic comorbidity in longitudinal studies: development and validation. J Chronic Dis. 1987;40:373–383. 18. Rosenbaum PR, Rubin DB. The central role of the propensity score in observational studies for causal effects. Biometrika. 1983;70:41–55. 19. Guo S, Fraser MW. Propensity score analysis: Statistical methods and applications. Sage University Press. 2011. 20. Weitzen S, Lapane KL, Toledano AY, et al. Principles for modeling propensity scores in medical research: a systematic literature review. Pharmacoepidemiol Drug Saf. 2004;13: 841–853. 21. Fleurence RL, Naci H, Jansen JP. The critical role of observational evidence in comparative effectiveness research. Health Aff (Millwood). 2010;29:1826–1833.
Address correspondence to: George Dranitsaris, BPharm, PhD, 283 Danforth Avenue Suite 448, Toronto M4K 1N2, Ontario, Canada. E-mail: [email protected]
144
Volume 37 Number 1
Comparative Effectiveness Analysis of Monotherapy With Cytotoxic Agents in Triple-negative Metastatic Breast Cancer in a Community Setting George Dranitsaris, BPharm, PhD1; Stefan Gluck, MD2; Claudio Faria, Pharm D, MPH3; David Cox, PhD3; and Hope Rugo, MD4 1
Augmentium Pharma Consulting Inc, Toronto, Ontario, Canada; 2Sylvester Comprehensive Care Center, Leonard M. Miller School of Medicine, University of Miami, Miami, Florida; 3Eisai Inc, Woodcliff Lake, New Jersey; and 4Helen Diller Family Comprehensive Cancer Center, University of California, San Francisco, California ABSTRACT Purpose: There has been considerable progress in the treatment of metastatic breast cancer. However, the identification of optimal cytotoxic agents in patients with triple-negative breast cancer (TNBC) (negative for hormone receptors and human epidermal growth factor receptor 2) remains a therapeutic challenge. We conducted a comparative effectiveness analysis of 4 cytotoxic agents in patients with TNBC. Methods: We retrospectively identified patients who received single-agent chemotherapy with eribulin, capecitabine, gemcitabine, or vinorelbine from 19 community oncology clinics across the United States. Data collection included baseline patient and disease characteristics, prior therapies, performance status, duration of current therapy, growth-factor use and other supportive care, and dose-limiting toxicities and associated dose reductions or delays or skipped doses. Time to treatment failure (TTF) was measured from the first cycle of chemotherapy until disease progression, discontinuation due to toxicity, or death. TTF was estimated using the Kaplan-Meier method and Cox proportional hazards modeling adjusted for clustering on the practice site. To control for selection bias, which is inherent in observational studies, a propensity score–weighted TTF analysis was also conducted. Findings: Data from 225 patients were included in the analysis (eribulin, 47 patients; capecitabine, 69; gemcitabine, 56; and vinorelbine, 53). The median age of each group was o60 years, with the exception of the gemcitabine group (63 years). The 4 groups were comparable with respect to age, performance status, duration of disease-free survival, presence of
134
comorbidities, and hemoglobin level before the start of chemotherapy. Median lines of therapy of eribulin, capecitabine, gemcitabine, and vinorelbine and were 4th, 2nd, 3rd, and 3rd, respectively. The median durations of treatment were 2 months with eribulin, capecitabine, and gemcitabine compared with 1.6 months with vinorelbine. Using eribulin as the reference drug, and with adjustment for line of therapy and associated prognostic factors, the propensity score– weighted Cox regression analysis did not identify significant between-treatment differences in TTF (hazard ratios [95% CI] vs eribulin: capecitabine, 1.15 [0.75–1.76]; gemcitabine, 0.62 [0.34–1.13]; and vinorelbine, 1.0 [0.67–1.67]). Implications: In this assessment of patients with TNBC treated in a community oncology setting, eribulin was utilized in later lines compared with the other agents. However, comparable drug activity was reported among the 4 agents. (Clin Ther. 2015;37:134–144) & 2015 Elsevier HS Journals, Inc. All rights reserved. Key words: eribulin, capecitabine, gemcitabine, metastatic breast cancer, triple negative, vinorelbine.
INTRODUCTION Despite recent advances in its early detection and treatment, breast cancer remains one of the leading causes of death in women. In 2013 the United States Accepted for publication October 29, 2014. http://dx.doi.org/10.1016/j.clinthera.2014.10.023 0149-2918/$ - see front matter & 2015 Elsevier HS Journals, Inc. All rights reserved.
Volume 37 Number 1
G. Dranitsaris et al. alone, there were 232,340 new cases of breast cancer diagnosed and 39,620 deaths from the disease.1 However, 5-year survival rates after a diagnosis of early stage breast cancer have increased, from 75% in the 1970s to approximately 90% in 2013.1 This increase has been largely due to improved options focused on specific tumor biology, with availability and widespread use of hormonal and cytotoxic therapies, as well as targeted agents (TAs), with treatment benefit extending into the metastatic setting.2–4 On closer inspection of the data, gains in survival have primarily been realized in patients whose tumors express hormone receptors or human epidermal growth factor receptor (HER)-2. In contrast, options are more limited for patients whose tumors are negative for hormone receptors and HER-2 (triple-negative breast cancer [TNBC]), in whom hormone and HER-2 TAs are ineffective. This heterogeneous group encompasses 15% of all new diagnoses.5 One of the distinctive characteristics of TNBC is the rapid development of, or de novo, resistance to chemotherapy, leading to shorter disease-free and overall survival (OS).5,6 In a cohort study in 1601 patients with breast cancer who were followed up for a median of 8.1 years, patients with TNBC were at a higher risk for distant recurrences (hazard ratio [HR] = 2.6; P o 0.05) and death (HR = 3.2; P o 0.05) within 5 years of the initial diagnosis.7 After a diagnosis of metastatic disease, patients in the TNBC cohort had a significantly shorter OS compared with the reference population (9 vs 22 months; P o 0.05).7 Clearly, the treatment of TNBC represents a major therapeutic challenge and is an active area of clinical research. Bevacizumab is a monoclonal antibody that targets vascular endothelial growth factor. There is some experience with bevacizumab in combination with chemotherapy in TNBC. In a subgroup analysis of data from RIBBON-2 (A Randomized, Double-Blind, Placebo-Controlled, Phase III Trial Evaluating the Efficacy and Safety of Bevacizumab in Combination With Chemotherapy for Second-Line Treatment of Human Epidermal Growth Factor Receptor 2Negative Metastatic Breast Cancer),8 which evaluated various cytotoxic agents with and without bevacizumab in metastatic disease, patients with TNBC had a statistically significant improvement in progression-free survival (6.0 vs 2.7 months; P o 0.001) and a numerically greater but statistically similar OS. However, these results have yet to be
January 2015
confirmed in an adequately powered Phase III randomized comparative trial. Therefore, it is unlikely that bevacizumab will address the current unmet need that is TNBC. Other agents being investigated in clinical trials include anti–vascular endothelial growth factor receptor tyrosine kinase inhibitors, anti–epidermal growth factor receptor therapies, poly (adenosine diphosphate ribose) polymerase inhibitors, Src tyrosine kinase inhibitors, and oral mammalian target of rapamycin inhibitors.5,9 The most advanced of these agents is iniparib, with a randomized Phase II trial reporting statistically significant improvements in objective response, progression-free survival, and most importantly OS (7.7 vs 12.3 months; HR = 0.57; P = 0.01).10 Iniparib was then taken into Phase III development in a trial that mimicked the Phase II study. In that randomized, controlled trial (RCT), the drug failed to meet its progression-free survival and OS end points, and further development of iniparib was halted.11 Because there are no specifically targeted chemotherapeutic drugs with proven clinical benefit in TNBC, patients typically are treated with the same agents used in other breast cancer subgroups. Patients with metastatic TNBC would be offered single-agent or combination chemotherapy containing an anthracycline or taxane agents, capecitabine, gemcitabine, vinorelbine, or carbo- or cisplatin.6,9 However, some results with eribulin were reported in a subset analysis of data from the Phase III EMBRACE (Eribulin Monotherapy Versus Treatment of Physician’s Choice in Patients With Metastatic Breast Cancer) trial.12 In that global study, which enrolled 762 patients into an eribulin arm or a physician’s-choice comparator arm, the experimental therapy was reported to have had an OS benefit (HR ¼ 0.81; P ¼ 0.041); 74% of enrolled patients were HER-2 negative, and 19% had TNBC. Eribulin was more effective in both hormone-negative and TNBC patients than in the control group. However, these intriguing findings have yet to be confirmed in an RCT. Although ongoing and planned trials are evaluating a variety of TAs for the treatment of patients with metastatic TNBC, clinicians need immediate information that will help to guide medical decision making. Comparative effectiveness studies, through observational data, are a reasonable approach to assessing treatment effects in a clinical practice setting.13,14 Several examples illustrate the concordance between
135
Clinical Therapeutics observational studies and the findings from subsequent randomized trials.15,16 In the present study, observational data collected from US community oncology practices were used for measuring differences in drug toxicity, health care resource utilization, and Time to treatment failure (TTF) between 4 antineoplastic agents in a sample of patients with metastatic TNBC.
PATIENTS AND METHODS Inclusion and Exclusion Criteria This retrospective, observational study was conducted in patients with metastatic TNBC treated with eribulin, capecitabine, gemcitabine, or vinorelbine between November 1, 2010, and August 31, 2013. All patients were treated in 1 of 19 community oncology practices that are a part of the Cancer Clinics of Excellence network. To be entered into the study, patients must have received 1 of the 4 agents for the treatment of advanced disease as a part of routine clinical practice and according to institutional administration guidelines. Each drug must have been given as single-agent chemotherapy; the addition of a TA was permitted. Eligible patients were at least 18 years of age and had a diagnosis of Stage IV breast cancer, with chemotherapy for advanced disease initiated after November 1, 2010. Patients were excluded if they were enrolled in a clinical trial of an investigational anticancer agent, but those enrolled in a clinical trial evaluating a supportive care agent were eligible.
Data Collection Baseline data collection consisted of demographic and disease characteristics, existing comorbidity as assessed by the Charlson comorbidity index,17 duration of treatment with a previous line(s) of chemotherapy, number and site(s) of metastases, Eastern Cooperative Oncology Group performance status before the start of chemotherapy, and the addition of TAs. From the first cycle of single-agent chemotherapy, clinical data collection consisted of baseline hemoglobin concentration; white blood cell, neutrophil, and platelet counts; the use of colonystimulating factors or antiemetic agents; total number of cycles delivered; and severe dose-limiting toxicities. Severe toxicities were events severe enough to require an intervention such as pharmacotherapy, hospitalization, a visit to an emergency department, or an unscheduled clinic visit.
136
Health care resource data collection included the starting dose of each anticancer drug administered; number of doses; total number of cycles delivered; number of premature discontinuations of chemotherapy; standard biochemistry and diagnostic test utilization; the use of prophylactic antibiotic agents, growth factors, serotonin antiemetic agents, and/or bone TAs; total number of red blood cell units and platelets administered; as well as hospitalizations, visits to an emergency department, or unscheduled clinic visits as a result of a treatment-related toxicity. The probable or possible cause of each clinically relevant event (eg, febrile neutropenia) was also recorded.
End Points The primary end point was TTF as a surrogate for disease progression. TTF was measured as the time from first dose of chemotherapy until the start of the next line of active anticancer treatment, the permanent discontinuation of therapy due to toxicity or death from any cause. Secondary end points included health care resource utilization, treatment-related toxicities, and factors related to treatment choice.
Statistical Analysis All outcomes data are presented descriptively as means, medians, or percentages. Given the observational nature of the study, no formal sample size or power analysis was undertaken. Therefore, the most relevant outcomes from the statistical analysis are the 95% confidence intervals (CIs). P values should be viewed as being of secondary importance. One of the challenges associated with observational studies (unlike randomized trials) is patient-selection bias. Systematic factors such as self-selection, physician selection, or geographic selection can lead to patients’ receiving one therapy over another. Such bias can affect the results of any comparative analysis in an observational study. To overcome this challenge, patient groups need to be adjusted for the potential impact of selection bias. In the present study, selection bias was addressed by a series of reevaluations characterized by propensity score analysis.18 Propensity score analysis is a 2-phase technique used for estimating a treatment effect in comparative groups selected by nonrandom means. In the first phase of a propensity score analysis, variables that influence selection to group assignment are used for
Volume 37 Number 1
G. Dranitsaris et al. modeling the probability of receiving a treatment (or of being in a reference group—in this case, the eribulin group).19,20 The resulting probability is called the propensity score. In the second phase, the propensity score is used for adjusting for preexisting group differences in the analysis of the relevant outcomes. There are several ways to use propensity scores, such as stratification variables, matching patients based on their propensity score, or their use as a weighting or adjustment variable in multivariate analysis.19,20 Assuming that all relevant covariates are included in the propensity score model, the group effect observed in a propensity score analysis represents an unbiased estimate of the true treatment effect.18,19 In the present study, any baseline covariate with a marginal association with treatment group (P o 0.10) was considered for inclusion in the final propensity score model. Because there were 4 groups, a maineffects multinomial logistic regression model was developed, using variables identified in the previous step, and retained after a backward elimination process with P o 0.10. The final propensity scores derived for each patient from the multinomial logistic regression model were used in the subsequent Kaplan-Meier survival analyses for adjustment for potential selection bias. The initial Kaplan-Meier analysis was a nonweighted comparison between the 4 groups. The propensity scores determined for each patient were then used as a weighting factor in the subsequent reevaluation. Survival curves for TTF were weighted by the estimated propensity scores, generated by the method of Kaplan-Meier and compared with the log-rank test. In a supporting analysis, a propensity score–weighted Cox proportional hazards regression analysis was also conducted to assess the relative risk for treatment failure in patients who received capecitabine, gemcitabine, or vinorelbine relative to eribulin. Variables evaluated in the Cox regression analysis included age, sex, performance status, time to diagnosis of advanced disease, prior adjuvant chemotherapy, existing comorbidities, site(s) of metastatic disease, the addition of TAs, and hematologic parameters at baseline. Independent baseline prognostic variables with a P value o0.05 were retained in the final regression model via backward elimination. The final regression model was also adjusted for clustering on the practice site to avoid violation of the independence assumption. As a final analysis, multinomial logistic regression analysis was applied to identify factors that were associated
January 2015
with the selection of capecitabine, gemcitabine, or vinorelbine over eribulin. All of the statistical analyses were performed using Stata release 11.0 (StataCorp LP, College Station, Texas).
RESULTS Study Population We identified 225 patients who met the inclusion criteria (eribulin, 47 patients; capecitabine, 69; gemcitabine, 56; and vinorelbine, 53). In this observational study, there were expected imbalances in some demographic and clinical characteristics (Table I). The median age of each group was o60 years, with the exception of the gemcitabine group (63 years). Eribulin was administered as 4th-line therapy compared with 2nd, 3rd, and 3rd for capecitabine, gemcitabine, and vinorelbine, respectively. The median duration of treatment was 2 months with eribulin, capecitabine, and gemcitabine compared with 1.6 months with vinorelbine. In addition, the gemcitabine group had a lower proportion of white patients and a longer median time from initial diagnosis to the development of metastatic disease. In general, hematologic parameters before the start of chemotherapy were comparable and within normal limits. The median number of prior chemotherapies varied between the treatment groups. The eribulin group received a median of 3 prior lines of chemotherapy compared with 1, 2, and 2 in the capecitabine, gemcitabine, and vinorelbine groups, respectively. There were also differences in the proportions of patients who received a TA in combination with single-agent chemotherapy—2.1% of patients received a TA with eribulin compared with 13.0%, 16.1%, and 9.4% with capecitabine, gemcitabine, and vinorelbine. The most commonly used TA was bevacizumab; a total of 5 patients received trastuzumab (Table I). The median numbers of treatment cycles were similar between the treatment groups, with 4 cycles of eribulin and 3 cycles of capecitabine, gemcitabine, and vinorelbine. There were numeric between-group differences in the percentages of patients with dose reductions or delays or missed doses, but these differences failed to reach statistical significance. At the completion of the course of therapy, patients treated with gemcitabine were significantly less likely to have received a next line of therapy (p o 0.05).
137
Clinical Therapeutics Table I. Demographic and clinical characteristics of the patients* in this comparative study of the effectiveness of monotherapy with cytotoxic agents in triple-negative metastatic breast cancer in a community setting. Characteristic Demographic characteristics Age, median (range), y Race, % White Black Other Clinical characteristics Weight, mean (range), lbs BSA, mean (range) Biochemistry, mean (SD) Hb, g/dL WBC, cells 109/L ANC, cells 109/L Platelets, cells 109/L Creatinine, mg/dL Disease characteristics ECOG PS, % 0 or 1 2 or 3 Not documented Charlson comorbidity score, median (range)† Time from diagnosis to metastatic disease, median (range), y Metastatic site, % Lung Liver Bone Brain Treatment history No. of treatments (inclusive), median (range) Chemotherapy Hormone therapy Treatments received, % Taxane agents Adjuvant therapy Anthracycline agent Capecitabine Current regimen‡ Single agent TA agent added§ No. of cycles, median (range) Treatment duration, median (range), mo Dose reductions, delays, or skipped doses, % Delays, median (range)
Eribulin (n ¼ 47)
Capecitabine (n ¼ 69)
Gemcitabine (n ¼ 56)
Vinorelbine (n ¼ 53)
54 (33–72)
59 (29–85)
63 (29–86)
59 (36–85)
69.6 24.6 5.8
68.1 28.0 4.3
56.1 29.3 14.6
81.0 16.7 2.3
68.9 (39–124) 1.8 (1.3–2.3)
70.8 (36.3–134) 1.8 (1.3–2.3)
72.1 (38.1–134) 1.8 (1.3–2.3)
72.1 (39.9–134) 1.7 (1.4–2.2)
11.0 7.6 5.6 246 0.80
11.6 6.2 4.6 253 0.83
11.5 6.9 5.1 256 0.96
11.4 8.2 5.6 266 0.84
(1.4) (3.8) (3.3) (112) (0.33)
68.1 23.4 8.5 6 (6–8) 1.8 (0–17.3)
(1.3) (2.8) (3.1) (100) (0.36)
63.8 23.2 13.0 6 (6–13) 1.1 (0–11.5)
(1.5) (4.4) (3.9) (124) (0.76)
64.3 19.6 16.1 6 (6–12) 2 (0–30.2)
(1.6) (7.9) (4.0) (106) (0.37)
54.7 30.2 15.6 6 (6–10) 1.7 (0–17.3)
40.4 29.8 25.5 10.6
40.6 24.6 40.6 4.4
46.4 26.8 32.1 8.9
32.1 18.9 26.4 1.9
4 (1–7) 3 (0–6) 0 (0–1)
2 (1–6) 1 (0–5) 0 (0–1)
3 (1–8) 2 (0–7) 0 (0–2)
3 (1–6) 2 (0–5) 0 (0–1)
91.5 78.7 59.5 57.4
66.7 69.6 14.5 0
85.7 71.4 39.3 34.0
86.8 81.1 34.0 45.2
97.9 2.1 4 (1–14) 2.1 (0–9.8) 38.3 0 ( 0–2)
87.0 13.0 3 (1–18) 2.0 (0.2–15.8) 34.8 0 (0–3)
83.9 16.1 3 (1–14) 2.1 (0–9.0) 42.9 0 (0–2)
90.6 9.4 3 (1–21) 1.6 (0–13.5) 45.3 0.5 (0–13) (continued)
138
Volume 37 Number 1
G. Dranitsaris et al. Table I. (continued).
Characteristic Reductions, median (range) Skipped, median (range) Received a next line of anticancer therapy, %
Eribulin (n ¼ 47) 0 (0–2) 0 (0–4) 66.0
Capecitabine (n ¼ 69) 1 (0–2) 0 (0–7) 63.8
Gemcitabine (n ¼ 56)
Vinorelbine (n ¼ 53)
0.5 (0–6) 1 (0–7) 44.6
0 (0–1) 1 (0–7) 73.6
ANC ¼ absolute neutrophil count; BSA ¼ body surface area; ECOG PS ¼ Eastern Oncology Cooperative Group performance status; Hb ¼ hemoglobin; TA ¼ targeted agents; WBC ¼ white blood cells. * All patients had HER-2– and hormone receptor–negative disease. † The weighted comorbidity classifications were as follows: 0 ¼ low; 1 or 2 ¼ intermediate; 3 or 4 ¼ high; and Z5 ¼ very high. ‡ Capecitabine: 1 g/m2 PO BID on days 1–14 q21d, 36.8%; 1.5 g/m2 PO BID on days 1–14 q21d, 12.1%; and other, 51.1%. Eribulin: 1.4 mg/m2 on days 1 and 8 q21d, 91.5%; 1.1 mg/m2 on days 1 and 8 q21d, 4.2%; and other, 4.2%. Gemcitabine: 1 g/m2 on days 1, 8, and 15 q28d, 22.0%; 1 g/m2 on days d1 and d8 q21d, 24.4%, and other, 53.6%. Vinorelbine regimens: 25 mg/m2 on day 1 q7d, 23.8%; 25 mg/m2 on days 1, 8, and d 15 q28d, 23.8%; and other, 52.4%. § In capecitabine patients, 6 patients received bevacizumab and 2 patients received trastuzumab. In eribulin and gemcitabine patients, one and nine patients received bevacizumab respectively. With vinorelbine, 3 patients received bevacizumab and two received trastuzumab.
Approximately two thirds of patients treated with eribulin went on to receive a 5th line of therapy (Table I).
Time to Treatment Failure The initial Kaplan-Meier survival analysis was not weighted by patients’ individual propensity scores (Figure 1). Using eribulin as the reference group, there were no significant differences between groups. Overall, the median TTF was 3 to 4 months with eribulin, capecitabine, and gemcitabine, with vinorelbine being 2.7 months. Eribulin, despite being given in as a later line of therapy (4th vs 2nd/3rd), provided comparable TTF benefits (hazard ratios [95% CI] vs eribulin: capecitabine, 1.15 [0.75–1.76]; gemcitabine, 0.62 [0.34–1.13]; and vinorelbine, 1.0 [0.67–1.67]). Given the potential for selection bias with observational data, the TTF analysis was then weighted by individual patients’ propensity scores. The final propensity scores incorporated factors such as age, prior therapies, race, and line of therapy. The findings of the weighted analysis did not alter the initial observations. There were no significant differences in TTF with capecitabine, gemcitabine, or vinorelbine relative to eribulin (Figure 2). These findings were further supplemented with a propensity score–weighted Cox proportional hazards analysis of TTF. Statistically
January 2015
significant covariates that were retained in the final multivariate model included the presence of bone metastases, baseline absolute neutrophil count, and creatinine concentration (Table II). Even with the inclusion of these covariates in the weighted model, the group differences relative to eribulin were not statistically significant.
Supportive Care and Treatment-Related Toxicity Relative to other agents, significantly fewer patients treated with capecitabine received myeloid growth factor support and serotonin antiemetic agents (P o 0.01). In contrast, there were no significant differences in the proportions of patients receiving erythropoietinstimulating agents or bone TAs (Table III). Significantly greater percentages of patients in the eribulin and vinorelbine groups reported Z1 toxicity event compared with those treated with capecitabine (both, P o 0.05). The most common dose-limiting toxicities with eribulin use were anemia (36.2%), neutropenia (31.2%), and neuropathy (19.4%). Similarly, the events related to myelosuppression were also the toxicities most commonly reported with vinorelbine and gemcitabine use. Patients treated with capecitabine were more likely to have developed hand/ foot skin reaction (37.7%) and/or diarrhea (29.4%) (Table III).
139
Clinical Therapeutics
Table II. Propensity score weighted Cox proportional hazards analysis of TTF in this comparative study of the effectiveness of monotherapy with cytotoxic agents in triple-negative metastatic breast cancer in a community setting.* Variable CAP vs ERB GEM vs ERB VIN vs ERB Bone metastases Baseline ANC Baseline creatinine
HR (95% CI) 1.15 0.62 1.0 1.49 1.07 1.79
(0.75–1.76) (0.34–1.13) (0.60–1.67) (1.20–1.84) (1.03–1.12) (1.08–2.97)
P
Impact TTF
0.53 0.12 0.99 o0.001 0.002 0.024
NS NS NS ↑ Risk by 49% ↑ Risk, ↑ ANC ↑ Risk, ↑ creatinine
ANC ¼ absolute neutrophil count; CAP ¼ capecitabine; ERB ¼ eribulin; GEM ¼ gemcitabine; HR ¼ hazard ratio; NS = not significant; TTF ¼ time to treatment failure; VIN ¼ vinorelbine. * Analysis was also adjusted for clustering on cancer clinic.
Similar percentages of patients receiving eribulin or gemcitabine required a medical intervention to treat an adverse event (23.4% and 17.8%, respectively). These interventions consisted of blood product transfusions, an unplanned visit to a clinic or an emergency department visit, and/or hospital admission. Conversely, 11.6% of the capecitabine group received a medical intervention for an adverse event, although the number of treatment discontinuations due to toxicity was highest with capecitabine, at 11.6%. The lowest rate of discontinuation due to an adverse event was 2.1% (eribulin). With all 4 drugs, the most common reason for the discontinuation of therapy was disease progression (Table III).
Factors Associated With Treatment Choice
The final investigation performed in the current study was multinomial logistic regression analysis to identify factors associated with the selection of the 3 chemotherapeutic agents compared with eribulin. Capecitabine was more likely to have been prescribed than was eribulin in older patients, in those with bone metastases, and in patients with higher baseline hemoglobin levels. In contrast, capecitabine was less likely to have been prescribed in patients with longer intervals of disease-free survival and as a later line of therapy (Table IV). Similarly, gemcitabine and vinorelbine were more likely to have been prescribed in older patients and as earlier lines of therapy. In summary, the multinomial analysis
140
identified patient age and line of therapy to be the 2 most important factors in the selection of capecitabine, gemcitabine, or vinorelbine over eribulin. In general, the likelihood of using eribulin over the other the agents decreased by 4 to 5% with each additional year of patient age. Hence, eribulin was more likely to have been prescribed in younger patients with TNBC and as a later line of therapy.
DISCUSSION In chemotherapy decision making, RCTs remain the “gold standard” for providing data on the efficacy and tolerability of new treatments. The major advantage of RCTs is their ability to balance both known and unknown confounding variables between the experimental and control groups and to incorporate double-blinding into the assessment of outcomes. As a result, most of the known biases that can affect treatment outcomes are either eliminated or minimized in a well-designed RCT. Nonetheless, RCTs are both resource and time intensive to undertake because they often require large sample sizes in highly select patient populations. This is particularly relevant in oncology, in which it is not uncommon for trial sample sizes to exceed 1000 patients, and often in less prevalent patient subgroups, such as TNBC. Hence, the results from RCTs may not be available until several years after the trial has been initiated. In other situations, a drug may be off-patent (or soon to go offpatent), so there would be no commercial incentive for
Volume 37 Number 1
G. Dranitsaris et al.
1.00 0.75 0.50 0.25 0.00 0
10
20 30 Time (mo) Eribulin Gemcitabine
40
50
Capecitabine Vinorelbine
Figure 1. Kaplan-Meier curve of time to treatment failure (TTF) from the start of chemotherapy in this comparative study of the effectiveness of monotherapy with cytotoxic agents in triplenegative metastatic breast cancer in a community setting. Median (interquartile range) TTF values (in months): eribulin, 3.2 (2.5–5.3); capecitabine, 3.4 (2.1–5.3); gemcitabine, 4.1 (3.0–8.4); and vinorelbine, 2.7 (1.8–4.6). Univariate hazard ratios (95% CI; P): capecitabine versus eribulin, 1.00 (0.70–1.51; 0.88); gemcitabine versus eribulin, 0.64 (0.38– 1.05; 0.08); and vinorelbine versus eribulin, 1.10 (0.69–1.75; 0.69) (with no baseline covariate adjustment).
the brand manufacturer to initiate a costly RCT. Therefore, alternative methods of data generation that provide robust information for medical decision making need to be identified. Comparative effectiveness research is a relatively new field, in which existing health care interventions are evaluated to determine which provide optimal benefits for patients, with the least harm. A main objective of comparative effectiveness research is to determine which treatment works best, in which patients, and in what circumstances in a clinical setting.21 In the present study, a comparative effectiveness analysis was undertaken to compare drug toxicity, health care resource utilization, and TTF between 4 cytotoxic agents widely used in community practice in patients with metastatic TNBC.
January 2015
The findings suggest comparative effectiveness, in terms of TTF, between eribulin, capecitabine, gemcitabine, and vinorelbine, despite the former agent being used in a later line of therapy. These findings were consistent after the data were adjusted by propensity score–weighted Cox multivariate analysis. Eribulin, even as a 4th line of therapy, offered a TTF benefit comparable to that of capecitabine, gemcitabine, and vinorelbine in patients with TNBC in this community-based dataset. In addition, a lesser percentage of patients treated with eribulin had permanent treatment discontinuation because of toxicity. The multinomial logistic regression analysis also yielded some results on key factors associated with the
1.00 0.75 0.50 0.25 0.00 0
10
20 30 Time (mo) Eribulin Gemcitabine
40
50
Capecitabine Vinorelbine
Figure 2. Propensity score–weighted time to treatment failure (TTF) from the start of chemotherapy in this comparative study of the effectiveness of monotherapy with cytotoxic agents in triple-negative metastatic breast cancer in a community setting. Median (interquartile range) TTF values (in months): eribulin ¼ 3.2 (2.5–5.3); capecitabine ¼ 2.9 (1.9–4.1); gemcitabine ¼ 4.1 (3.2–7.0); and vinorelbine ¼ 2.7 (1.8–4.4). Univariate hazard ratios (95% CI; P): capecitabine versus eribulin, 1.32 (0.90–1.95; 0.16); gemcitabine versus eribulin, 0.69 (0.43–1.15; 0.16); and vinorelbine versus eribulin, 1.16 (0.71–1.88; 0.55) (with no baseline covariate adjustment).
141
Clinical Therapeutics
Table III. Supportive care and tolerability during current therapy in this comparative study of the effectiveness of monotherapy with cytotoxic agents in triple-negative metastatic breast cancer in a community setting. Parameter Supportive care, % G-CSF 5-HT3 antiemetic Bone targeted agent ESA Prophylactic antibiotic Tolerability Patients with Z1 AE, no. (%) Treatment-emergent AE (any grade), %* Anemia Neutropenia Neuropathy Emesis Edema Thrombocytopenia Elevated liver enzymes Stomatitis Rash Febrile neutropenia Bleeding events Hand/foot skin reaction Diarrhea Medical interventions for toxicity, no. (%) No. of interventions, by type† Blood product transfusion Unplanned clinic visit Hospital admission Emergency department visit Supportive care medication prescribed Second medical intervention for another toxicity Reason therapy was stopped, % Disease progression Went to hospice Disease stabilization Other Toxicity Patient wish Death Disease response
Eribulin (n ¼ 47)
Capecitabine (n ¼ 69)
Gemcitabine (n ¼ 56)
Vinorelbine (n ¼ 53)
55.3 53.8 25.5 14.9 12.5
4.3 7.2 37.7 15.9 0
26.8 71.4 25.0 17.8 0
34.0 58.5 15.1 26.4 5.7
33 (70.2)
35 (55.1)
34 (60.7)
39 (73.6)
36.2 31.2 19.4 10.6 7.1 6.4 6.4 6.3 2.1 2.1 2.1 0 0 9 (23.4)
15.9 4.3 2.9 4.3 4.3 7.2 1.4 10.1 5.9 0 0 37.7 29.4 8 (11.6)
32.1 32.1 1.8 1.8 0 23.2 5.4 3.6 1.8 1.8 1.8 0 1.8 10 (17.8)
45.3 45.3 3.8 7.5 3.8 3.8 5.7 0 1.9 1.9 0 0 3.8 4 (7.1)
9 1 1 1 0 4
2 5 2 1 3 0
7 3 2 1 3 2
2 1 2 0 1 2
74.5 8.5 6.4 4.2 2.1 2.1 2.1 0
73.9 4.3 0 1.4 11.6 2.9 1.4 4.3
46.4 3.6 3.6 25.0 7.1 10.7 3.6 0
73.6 5.7 0 9.3 5.7 1.9 1.9 1.9
5-HT3 ¼ serotonin; AE ¼ adverse event; ESA ¼ erythropoietin-stimulating agents; G-CSF ¼ granulocyte colony-stimulating factor. * Some patients experienced 41 event. † Multiple types of interventions were given for a single event.
use of the alternatives to eribulin. Capecitabine, gemcitabine, and vinorelbine were more likely to have been prescribed in older patients and in earlier lines of therapy. The latter finding is likely related to the fact that the pivotal trial of eribulin evaluated patients
142
who were heavily pretreated with an anthracycline and/or a taxane agent.12 However, the observation that eribulin was preferentially used in younger TNBC patients was unexpected and requires a qualitative investigation.
Volume 37 Number 1
A final comment has to be made about the relatively modest TTF benefits of only 3 to 4 months identified in the present study. This finding suggests that the benefit of cytotoxic therapy in later lines of treatment remains modest and that more effective therapy needs to be identified to treat this patient population.
2.0 0.10 NS 0.75 0.009 25% Less likely per line 1.23 0.19 NS 3.19 0.002 3-Fold more likely 0.51 o0.001 49% Less likely per line 1.43 0.03 ↑ Likelihood with ↑ Hb
Study Limitations
CAP ¼ capecitabine; ERB = eribulin; GEM ¼ gemcitabine; Hb ¼ hemoglobin; OR ¼ odds ratio; VIN ¼ vinorelbine. * Analysis was adjusted for clustering on cancer clinic.
1.83 0.12 NS 0.77 0.0051 23% Less likely 1.33 0.092 NS
NS 0.93 0.88 68% Less likely NS 0.47 0.74
0.32 0.009
1.04 0.036 ↑ Likelihood with ↑ age 0.96 0.52 NS ↑ Likelihood with ↑ age NS 0.023 ↑ Likelihood with ↑ age 0.003 23% Less likely per year 1.05 0.77
Age Time to metastatic disease Nodal metastases present Bone metastases present Line of therapy Baseline Hb
P OR
1.05 0.005 0.90 0.14
Likelihood of Receiving the Regimen
January 2015
Variable
Likelihood of Receiving the Regimen
OR
P
Likelihood of Receiving the Regimen
OR
P
VIN vs ERB GEM vs ERB CAP vs ERB
Table IV. Multinomial logistic regression analysis of factors associated with the selection of alternatives to eribulin (ERB) in this comparative study of the effectiveness of monotherapy with cytotoxic agents in triple-negative metastatic breast cancer in a community setting.*
G. Dranitsaris et al.
Because this study was retrospective and observational and not a randomized trial, there were between-group imbalances in several important prognostic factors, such as age, performance status, site(s) of metastatic disease, and recurrence-free survival. There was also an imbalance in prior therapies, which may have affected treatment selection secondary to drug toxicity. To address these biases, a propensity score– weighted evaluation and a multivariate analysis were undertaken to support the initial Kaplan-Meier survival analyses. Notwithstanding, these statistical techniques do not entirely eliminate all of the selection biases that are inherent in observational studies. The total sample size was small (n ¼ 225), and all regions of the United States as well as ethnic groups were not equally represented, which could compromise the generalizability of the results to other jurisdictions. There were differences in the concurrent use of TA, namely bevacizumab, between the 4 cytotoxic-agent groups. However, in the multivariate analysis, the addition of a TA was not retained in the final Cox regression model. Because the drugs were not given in the same line of therapy, OS could not be properly evaluated. As a result, we were able to measure differences only in TTF between the 4 drugs.
CONCLUSIONS In this assessment of the use of 4 cytotoxic agents in patients with metastatic TNBC treated in a community oncology setting, eribulin was utilized as a later line of therapy relative to capecitabine, gemcitabine, and vinorelbine. However, eribulin was reported to have had at least comparable drug activity and tolerability, even when used in more heavily pretreated patients. These findings warrant further analyses in a larger population with an evaluation of biologic heterogeneity.
ACKNOWLEDGMENTS
Eisai Inc. provided financial support to conduct this study. The corresponding author had full access to the data in the study and had the final responsibility for
143
Clinical Therapeutics the decision to submit the paper. George Dranitsaris, Stefan Gluck and Hope Rugo have acted as consultants to the sponsor.
CONFLICTS OF INTEREST This study and its publication were funded by Eisai Inc, the manufacturers of eribulin. Drs. Faria and Cox are employees of, and may hold stock options in, Eisai Inc. The authors have indicated that they have no other conflicts of interest with regard to the content of this article.
REFERENCES 1. Surveillance Epidemiology and End Results (SEER). National Cancer Institute 2011. Washington, USA, 2013 (http://seer.cancer.gov/). Accessed January 31, 2014. 2. Slamon DJ, Leyland-Jones B, Shak S, et al. Use of chemotherapy plus a monoclonal antibody against HER2 for metastatic breast cancer that overexpresses HER2. N Engl J Med. 2001;344:783–792. 3. Mouridsen H, Gershanovich, Sun Y, et al. Superior efficacy of letrozole versus tamoxifen as first line therapy for postmenopausal women with advanced breast cancer: Results of a phase III study of the International Letrozole Breast Cancer Group. J Clin Oncol. 2001;19:2596-06. 4. Verma S, Miles D, Gianni L, et al. Trastuzumab emtansine for HER2-positive advanced breast cancer. N Engl J Med. 2012;367:1783–1791. 5. Bayraktar S, Glück S. Molecularly targeted therapies for metastatic triple-negative breast cancer. Breast Cancer Res Treat. 2013;138:21–35. 6. André F, Zielinski CC. Optimal strategies for the treatment of metastatic triple-negative breast cancer with currently approved agents. Ann Oncol. 2012;23(Suppl 6):vi46–vi51. 7. Dent R, Trudeau M, Pritchard KI, et al. Triple negative breast cancer: clinical features and patterns of recurrence. Clin Cancer Res. 2007;13:4429–4434. 8. Brufsky A, Valero V, Tiangco B, et al. Second-line bevacizumab-containing therapy in patients with triplenegative breast cancer: subgroup analysis of the RIBBON-2 trial. Breast Cancer Res Treat. 2012;133:1067–1075.
9. Crown J, O’Shaughnessy J, Gullo G. Emerging targeted therapies in triple-negative breast cancer. Ann Oncol. 2012;23 (Suppl 6):vi56–vi65. 10. O’Shaughnessy J, Osborne C, Pippen JE, et al. Iniparib plus chemotherapy in metastatic triple-negative breast cancer. N Engl J Med. 2011;364:205–214. 11. O’Shaughnessy J, Schwartzberg LS, Danso MA, et al. A randomized phase III study of iniparib (BSI-201) in combination with gemcitabine/carboplatin (G/C) in metastatic triple negative breast cancer. J Clin Oncol. 2011;29:15s. Abstr 1007. 12. Cortes J, O’Shaughnessy J, Loesch D, et al. Eribulin monotherapy versus treatment of physician’s choice in patients with metastatic breast cancer (EMBRACE): a phase 3 open-label randomised study. Lancet. 2011;377: 914–923. 13. Carroll J. Payers step in with ‘real-world’ comparative effectiveness research. Manag Care. 2011;20:23–26. 14. Teutsch SM, Fielding JE. Comparative effectiveness–looking under the lamppost. JAMA. 2011;305:2225–2226. 15. Byers T, Sedjo RL. Does intentional weight loss reduce cancer risk? Diabetes Obes Metab. 2011;13:1063–1072. 16. Cooper K, Squires H, Carroll C, et al. Chemoprevention of colorectal cancer: systematic review and economic evaluation. Health Technol Assess. 2010;14:1–206. 17. Charlson ME, Pompei P, Ales KL, MacKenzie CR. A new method of classifying prognostic comorbidity in longitudinal studies: development and validation. J Chronic Dis. 1987;40:373–383. 18. Rosenbaum PR, Rubin DB. The central role of the propensity score in observational studies for causal effects. Biometrika. 1983;70:41–55. 19. Guo S, Fraser MW. Propensity score analysis: Statistical methods and applications. Sage University Press. 2011. 20. Weitzen S, Lapane KL, Toledano AY, et al. Principles for modeling propensity scores in medical research: a systematic literature review. Pharmacoepidemiol Drug Saf. 2004;13: 841–853. 21. Fleurence RL, Naci H, Jansen JP. The critical role of observational evidence in comparative effectiveness research. Health Aff (Millwood). 2010;29:1826–1833.
Address correspondence to: George Dranitsaris, BPharm, PhD, 283 Danforth Avenue Suite 448, Toronto M4K 1N2, Ontario, Canada. E-mail: [email protected]
144
Volume 37 Number 1