|
Advertisement |
|
|
||
 |
|
|||||
|
|||||
|
 |
||||||
|
||||||
|
||||||
|
Originally published as JCO Early Release 10.1200/JCO.2008.17.2544 on July 28 2008 © 2008 American Society of Clinical Oncology. Effect of Molecular Disease Subsets on Disease-Free Survival in Randomized Adjuvant Chemotherapy Trials for Estrogen Receptor–Positive Breast Cancer
From the Departments of Breast Medical Oncology and Pathology and Division of Quantitative Sciences, The University of Texas M. D. Anderson Cancer Center, Houston, TX; and the Translational Research Unit, Institut Gustave Roussy, Villejuif, France Corresponding author: Lajos Pusztai, MD, DPhil, Department of Breast Medical Oncology, The University of Texas M. D. Anderson Cancer Center, PO Box 301439, Houston, TX 77230-1439; e-mail: lpusztai{at}mdanderson.org
Purpose The majority of estrogen receptor (ER)–positive cancers are sensitive to endocrine therapy and may not derive much further benefit from chemotherapy, but a subset are potentially chemotherapy sensitive. Molecular diagnostic tests allow the identification of these various subsets with some accuracy. The goal of the current analysis was to examine how the proportion of cases in the various risk (recurrence score [RS]) categories of a commercially available multigene assay influences the power of randomized trials to show benefit from adjuvant chemotherapy. Methods We modeled 10-year disease-free survival (DFS) for hypothetical, two-arm clinical trials that randomly assigned patients with ER-positive breast cancer to endocrine therapy alone or endocrine therapy plus chemotherapy. We varied the proportion of patients in low, intermediate, and high RS categories and used DFS estimates for each risk group based on results from the Southwest Oncology Group 8814 study. Results The probability of observing significant improvement in DFS as a result of chemotherapy decreases as the proportion of patients in the low RS category increases. For example, if a trial is designed with 80% power and the actual proportion of low RS patients accrued to the study increases from 40% to 60%, the power drops to 63%. Conclusion Variable accrual of low RS patients into different randomized adjuvant chemotherapy trials may partly explain contradictory results in the literature. Studies can be underpowered to detect improvement with chemotherapy as a result of inclusion of too many patients with low RS. Future adjuvant studies for ER-positive breast cancer will need to consider stratifying patients by molecular subtype.
Results of several neoadjuvant clinical trials indicate that estrogen receptor (ER)–negative cancers are more sensitive to chemotherapy than ER-positive tumors. Pathologic complete response rates are two- to four-fold higher in ER-negative compared with ER-positive cancers.1-4 Randomized adjuvant chemotherapy trials also demonstrated greater relative improvements in survival in ER-negative cancers compared with ER-positive cancers, which is consistent with the greater chemotherapy sensitivity of these tumors. A recent combined analysis of three different adjuvant chemotherapy studies (Cancer and Leukemia Group B 8541, 9344, and 9741) including a total of more than 6,000 patients showed that the hazard reduction owing to chemotherapy was twice as high in ER-negative than in ER-positive breast cancer.5 A series of meta-analyses by the Early Breast Cancer Trialists’ Collaborative Group also indicated greater benefit from adjuvant chemotherapy in ER-poor disease.6 The 5-year absolute reductions in recurrence resulting from poly-chemotherapy were 13.2% and 7.6% in ER-poor and ER-positive cancers, respectively, in woman younger than 50 years of age (P < .00001). In the older than 50 years age group, the absolute gains from chemotherapy were smaller (9.6% for the ER-poor and 4.9% for the ER-positive cancers) but remained statistically significant in both receptor groups. Conversely, several other randomized trials failed to detect significantly different efficacy of adjuvant chemotherapy according to ER status. A recent report showed that weekly paclitaxel after doxorubicin and cyclophosphamide (AC) adjuvant chemotherapy improved survival to the same extent in both patients with ER-negative and ER-positive disease compared with every 3 weeks paclitaxel.7 The National Surgical Adjuvant Bowel and Breast Project (NSABP) B28 trial that compared four courses of AC with AC followed by four additional courses of paclitaxel (every 3 weeks) also failed to demonstrate a statistically significant interaction between hormone receptor status and benefit from paclitaxel, although the reduction in recurrence rate was numerically smaller in patients with ER-positive (12%) than in those with ER-negative disease (17%).8 A recent combined analysis of the Breast Cancer International Research Group-001 and the PACS-01 studies also did not show differential efficacy of docetaxel adjuvant chemotherapy by hormone receptor status.9 There are several variables that may affect the power of a clinical trial to detect an interaction between benefit from adjuvant chemotherapy and hormone receptor status. These include the number of patients in the different receptor categories, the length of follow-up, the efficacy of treatment, and the reliability of ER measurements. To further complicate matters, adjuvant cytotoxic therapy can induce early menopause in younger women, and this endocrine effect may exaggerate the magnitude of what is assumed to be chemotherapy benefit in premenopausal women with ER-positive disease. However, an equally important and frequently ignored variable that affects outcome may be the biologic heterogeneity of ER-positive breast cancer. Some ER-positive cancers are clearly more sensitive to chemotherapy than others. These chemotherapy-sensitive, ER-positive tumors have distinct molecular features. They tend to have HER2 amplification and high grade and belong to the luminal B molecular class.10-12 Oncotype DX (Genomic Health, Redwood, CA) is a commercially available multigene assay that simultaneously measures the expression of ER, HER-2, and several ER-regulated transcripts, as well as proliferation related and other genes. An important advantage of Oncotype DX is that it combines these measurements into an easily interpretable recurrence score (RS) that can be used to estimate the probability of recurrence with 5 years of adjuvant tamoxifen therapy.13 The ability of the RS to predict benefit from adjuvant cyclophosphamide, methotrexate, and fluorouracil (CMF or MF) chemotherapy in ER-positive, node-negative breast cancers was also examined in the NSABP-B20 randomized clinical trial, which compared adjuvant tamoxifen with tamoxifen plus CMF (or MF). The study showed that a higher RS was associated with greater benefit from adjuvant chemotherapy (and higher risk of recurrence with endocrine therapy alone).14 Similar correlation between RS and survival after adjuvant chemotherapy was observed in three other studies.15-17 A plausible inference from these observations is that the proportion of patients with ER-positive disease in the various RS categories may have substantial impact on the overall survival results of adjuvant trials. When large proportions of patients are in the low RS category, few patients in the entire study will benefit from the addition of chemotherapy, and conversely, when many patients are in the high RS category, the benefit from chemotherapy will be more readily detectable. There are significant variations in the clinical characteristics of patients with ER-positive disease (ie, median age, menopausal status, grade, and nodal status) who are accrued to various randomized adjuvant chemotherapy trials. It is also likely that the proportion of patients in different RS categories may vary from study to study. The purpose of this article was to examine the potential effect of including varying proportions of patients in low, intermediate, and high RS categories on the power of adjuvant chemotherapy trials for ER-positive breast cancer. To model this effect we used disease-free survival (DFS) results from the Southwest Oncology Group (SWOG) 8814 study that was presented at the 2007 San Antonio Breast Cancer Symposium. In the SWOG 8814 trial, postmenopausal patients with ER-positive and lymph node-positive tumors were randomly assigned to receive tamoxifen alone or tamoxifen plus cyclophosphamide, doxorubicin, and fluorouracil chemotherapy. The Oncotype DX assay was applied to a subset of patients (n = 367) retrospectively. Forty percent of the tested patients were in the low RS category, 28% were in the intermediate RS category, and 32% were in the high RS category. The 10-year DFS rates were approximately 64%, 63%, and 55% versus 60%, 49%, and 43% in the low, intermediate, and high RS categories with and without chemotherapy, respectively.17 We used these DFS numbers as the basis for our simulations.
We modeled DFS for hypothetical, two-arm clinical trials that randomly assigned patients with ER-positive disease equally into one of two treatment arms: arm A, chemotherapy plus endocrine therapy, or arm B, endocrine therapy alone. Patients were assigned to low, intermediate, or high RS category based on the uniform (0,1) distribution and sequentially assigned to treatment arms A or B. The base trial assumed that 40%, 30%, and 30% of patients accrued to the study were in the low, intermediate, and high RS categories, respectively, as observed in the SWOG 8814 trial. In arm A, the 10-year DFS rates were assumed to be 65%, 65%, and 55% for the low, intermediate, and high RS patient groups, respectively. In arm B, the rates were 60%, 50%, and 45% for the low, intermediate, and low RS patient groups, respectively. Each simulated trial included 2,000 patients with ER-positive disease. For each simulated patient, we generated both a DFS time and a follow-up time. DFS times were generated from exponential distributions with means equal to the hazard rates corresponding to the assumed 10-year DFS rates. This is a simplification because the estimated hazard functions are probably not exponential, treatments effects could vary not only by RS category but also over time. Follow-up time was generated from an exponential distribution and chosen to provide four different levels of power for the base study: 80%, 85%, 90%, and 95%. Median follow-up time was between 3.5 and 6.0 years in these simulated studies. Patients whose DFS time was shorter than their follow-up time were considered to have events and patients with a shorter follow-up time than their DFS time were considered censored. The goal of the simulations was to examine how the calculated power of the base study changes as the proportion of patients in various RS categories changes. We varied the proportion of patients in the different RS groups from 20% to 80%. As the proportion of patients in a given RS category changed, the ratio of the other two categories was kept at 1:1 for simplicity. DFS was compared between treatment arms with the log-rank test, and P values less than .05 were considered statistically significant. We report the median hazard ratio for DFS for the whole study population and the probability of finding a statistically significant difference in DFS between the two treatment arms. Results are based on 5,000 individual trial simulations performed in R 2.4.1 (R Development Core Team, Vienna, Austria).18
Figure 1 illustrates how the probability of observing statistically significant improvement in DFS resulting from adjuvant chemotherapy decreases as the proportion of patients in the low RS category increases. For example, if it is assumed that 40% of patients have low RS and the trial is designed with 80% power (ie, base study modeled after SWOG 8814), but the actual proportion of low RS patients accrued to the study increases to 60%, the power of the trial drops to 63%. Even if the original study has a 90% power, the same increase in the proportion of low RS patients (from 40% to 60%) can result in a drop of power to 84%. At more extreme increases in the proportion of low RS patients, these shifts in power are more pronounced.
Varying the proportion of patients in the intermediate RS category while keeping the ratio of the remaining two categories equal has a lesser but positive effect on power (Fig 2). As the proportion of patients with intermediate RS increases from 30% to 60%, the power of the study increases from 83% to 95%. Changing the proportion of high RS patients has minimal effect on power (Fig 3).
It is important to note that the hazard ratios remained more than 1.0 in all simulations, indicating a trend at least for improved survival with chemotherapy in each study. Figure 4 illustrates how the hazard ratios change as the proportion of patients in the various RS categories change. The base study had a hazard ratio of 1.33 (and 80% power to detect this difference as significant), which decreased to 1.27 (power of 63%) as the proportion of low RS cases increased from 40% to 60%. These results indicate that many of our simulated, and probably some real studies, have modest power to show statistically significant improvement in DFS from adjuvant chemotherapy as a result of accruing relatively high proportions of patients with low RS. Power calculations may be unreliable without considering the proportion of ER-positive patients in the various RS categories.
It has been suggested previously that variable accrual of patients with different degrees of chemotherapy sensitivity into clinical trials can influence the outcome of a study.19,20 Advances in molecular diagnostics made it possible to model this impact more realistically. For example, Oncotype DX was applied to archived specimens from four different randomized adjuvant clinical trials and also to a population-based cohort of patients. In these studies, the proportion of patients in the low RS category varied from 40% to 56%. The proportion in the intermediate RS category varied from 19% to 31%, and in the high RS category, the proportion varied from 20% to 32%.13-17 This indicates that studies that were designed to address similar clinical questions may have accrued substantially different ER-positive patient populations, which may partly explain the variable and sometimes contradictory results from these trials. In this analysis, we used data from a recently presented study to model the impact of Oncotype DX subsets on the outcome of randomized adjuvant chemotherapy clinical trials in ER-positive breast cancer. Our results indicate that as the number of patients in the low RS category increases, the overall power of the study to detect a significant benefit from adjuvant chemotherapy decreases. For studies with smaller sample size (ie, < 2,000 patients with ER-positive disease) and shorter follow-up, this effect could be even more pronounced. Unfortunately, there is no information on the distribution of molecular subtypes in any of the other published adjuvant studies that compared endocrine therapy with endocrine therapy plus chemotherapy (or one chemotherapy regimen v another) in patients with ER-positive disease except the SWOG 8814 trial and the NSABP-B20 trial that was used to develop the Oncotype DX assay. As a further example, we performed simple calculations to illustrate how relatively small shifts in the proportion of patients in low RS category could have affected the outcome of the NSABP-B20 trial. That study included 54% of patients in the low RS, 21% in the intermediate RS, and 25% in the high RS categories. The 10-year distant metastasis-free survival of patients in these three categories were 97%, 91%, and 60%, respectively, when treated with tamoxifen alone, and they were 96%, 89%, and 88% when treated with both tamoxifen and CMF (or MF) adjuvant chemotherapy.14 On the basis of these numbers, the overall survival difference at 10 years was 6% between the two treatment arms in the original study. A 12% increase in the proportion of low RS patients from 54% to 66% (and corresponding decrease in the other categories to 17% each) would have reduced the overall survival difference to 3.7%. (Calculated survival in the original study: tamoxifen arm = [0.54 x 97] + [0.21 x 91] + [0.25 x 60] = 86.49%; tamoxifen + chemotherapy = [0.54 x 96] + [0.21 x 89] + [0.25 x 88] = 92.53%. After changing the proportion of molecular subsets, tamoxifen arm = [0.66 x 97] + [0.17 x 91] + [0.17 x 60] = 89.63%; tamoxifen + chemotherapy = [0.66 x 96] + [0.17 x 89] + [0.17 x 88] = 93.45%.) Conversely, a 14% reduction in the proportion of low RS patients to 40% would have increased the survival benefit to 7.4%. Patient referral bias may explain variable accrual of different molecular subsets of ER-positive patients into randomized clinical trials.21,22 Referral bias can operate at several levels. Studies at large academic centers accrue significantly younger patients than community practice–based studies, and younger patients with ER-positive disease tend to have more chemotherapy-sensitive cancers in general. Competing studies that are simultaneously open for accrual may result in entering higher-risk patients to the study with the treatment regimen that is perceived to be more aggressive (eg, more chemotherapy v less chemotherapy) rather than to a study that is less aggressive (ie, endocrine therapy v endocrine therapy plus chemotherapy). Shifts in standard practice over time can also affect accrual. For example, patients with ER-positive and HER-2 amplified cancers now routinely receive trastuzumab or are referred to studies with HER-2 targeted drugs, which removes this potentially most chemotherapy-sensitive, ER-positive patient population from other adjuvant chemotherapy trials.23 However, it is important to recognize that age and HER-2 or progesterone receptor status alone are less than ideal surrogates for RS categories; therefore, using these clinical variables when trying to compare molecular heterogeneity among the ER-positive patients between different studies is suboptimal and will likely underestimate heterogeneity. We attempted to perform as realistic trial simulations as possible; however, simulation results are susceptible to a priori assumptions used in the model. We used 10-year DFS estimates that were similar to those observed in a subset analysis of a randomized clinical trial for low, intermediate, and high RS patients treated with tamoxifen alone or with sequential tamoxifen plus cyclophosphamide, doxorubicin, and fluorouracil chemotherapy. Changing these DFS estimates by increasing or decreasing the efficacy of therapy or the baseline risk of recurrence in the various RS categories (ie, multiple node-positive population) could alter the power calculations, but the main trends would remain the same. We did not simultaneously vary the proportion of all three risk categories; as one category increased, the other two were kept equal. It is unlikely that large deviations form 1:1 proportion would occur, and because of the large number of possible combinations in three-way proportions, we did not examine this effect. However, if complex three-way changes in proportions affect study power, this would further underline our point that variable accrual of patients with different molecular subsets of breast cancer represents an important unaccounted for variable in clinical trials. In the modeling process, we simulated DFS as exponential distribution; this is a simplification, but it is unlikely to alter the main conclusions from the analysis. The implication of our results is that future adjuvant studies for ER-positive breast cancer will need to consider the various molecular subsets to generate the most informative results. At the minimum, studies will need to stratify patients by their molecular subtype. Currently, Oncotype DX represents a standardized and commercially available molecular test that could be used for this purpose, but several other assays are under development and may become available in the future.24-26
The author(s) indicated no potential conflicts of interest.
Conception and design: Lajos Pusztai, Kristine Broglio, Kenneth R. Hess Financial support: Lajos Pusztai, Gabriel N. Hortobagyi Administrative support: Lajos Pusztai, Gabriel N. Hortobagyi Provision of study materials or patients: Lajos Pusztai Collection and assembly of data: Lajos Pusztai Data analysis and interpretation: Lajos Pusztai, Kristine Broglio, Fabrice Andre, W. Fraser Symmans, Kenneth R. Hess Manuscript writing: Lajos Pusztai, Kristine Broglio Final approval of manuscript: Lajos Pusztai, Kristine Broglio, Fabrice Andre, W. Fraser Symmans, Kenneth R. Hess, Gabriel N. Hortobagyi
published online ahead of print at www.jco.org on July 28, 2008. Authors’ disclosures of potential conflicts of interest and author contributions are found at the end of this article.
1. Liedtke C, Mazouni C, Hess KR, et al: Response to neoadjuvant therapy and long-term survival in patients with triple-negative breast cancer. J Clin Oncol 26:1275-1281, 2008 2. Ellis P, Smith I, Ashley S, et al: Clinical prognostic and predictive factors for primary chemotherapy in operable breast cancer. J Clin Oncol 16:107-114, 1998 3. Andre F, Mazouni C, Liedtke C, et al: HER2 expression and efficacy of preoperative paclitaxel/FAC chemotherapy in breast cancer. Breast Cancer Res Treat 108:183-190, 2008[CrossRef][Medline] 4. Bear HD, Anderson S, Brown A, et al: The effect on tumor response of adding sequential preoperative docetaxel to preoperative doxorubicin and cyclophosphamide: Preliminary results from National Surgical Adjuvant Breast and Bowel Project Protocol B-27. J Clin Oncol 21:4165-4174, 2003 5. Berry DA, Cirrincione C, Henderson IC, et al: Estrogen-receptor status and outcomes of modern chemotherapy for patients with node-positive breast cancer. JAMA 295:1658-1667, 2006 6. Early Breast Cancer Trialists’ Collaborative Group (EBCTCG): Effects of chemotherapy and hormonal therapy for early breast cancer on recurrence and 15-year survival: An overview of the randomised trials. Lancet 365:1687-1717, 2005[CrossRef][Medline] 7. Sparano JA, Wang M, Martino S, et al: Weekly paclitaxel in the adjuvant treatment of breast cancer. N Engl J Med 358:1663-1671, 2008 8. Mamounas EP, Bryant J, Lembersky B, et al: Paclitaxel after doxorubicin plus cyclophosphamide as adjuvant chemotherapy for node-positive breast cancer: Results from NSABP B-28. J Clin Oncol 23:3686-3696, 2005 9. Andre F, Broglio K, Roche H, et al: Estrogen receptor expression and efficacy of docetaxel containing adjuvant chemotherapy in patients with node-positive breast cancer: Results from a pooled analysis. J Clin Oncol 26:2636-2643, 2008 10. Hayes DF, Thor AD, Dressler LG, et al: HER2 and response to paclitaxel in node-positive breast cancer. N Engl J Med 357:1496-1506, 2007 11. Mazouni C, Kau SW, Frye D, et al: Inclusion of taxanes, particularly weekly paclitaxel, in preoperative chemotherapy improves pathologic complete response rate in estrogen receptor-positive breast cancers. Ann Oncol 18:874-880, 2007 12. Rouzier R, Perou CM, Symmans WF, et al: Breast cancer molecular subtypes respond differently to preoperative chemotherapy. Clin Cancer Res 11:5678-5685, 2005 13. Paik S, Shak S, Tang G, et al: A multigene assay to predict recurrence of tamoxifen-treated, node-negative breast cancer. N Engl J Med 351:2817-2826, 2004 14. Paik S, Tang G, Shak S, et al: Gene expression and benefit of chemotherapy in women with node-negative, estrogen receptor-positive breast cancer. J Clin Oncol 24:3726-3734, 2006 15. Habel LA, Shak S, Jacobs MK, et al: A population-based study of tumor gene expression and risk of breast cancer death among lymph node-negative patients. Breast Cancer Res 8:R25, 2006[CrossRef][Medline] 16. Goldstein LJ, Gray R, Childs B, et al: Prognostic utility of the 21-gene assay in Hormone Receptor (HR) positive operable breast cancer and 0-3 positive axillary nodes treated with adjuvant chemo-hormonal therapy: An analysis of Intergoup trial E2197. J Clin Oncol 25:9s, 2007 (suppl; abstr 526)[CrossRef] 17. Albain K, Barlow B, Shak S, et al: Prognostic and predictive value of the 21-gene recurrence score assay in postmenopausal, node-positive, ER-positive breast cancer. Beast Cancer Res Treat 106:8, 2007 (suppl 1; abstr 10) 18. R Development Core Team R: A language and environment for statistical computing. Vienna, Austria, R Foundation for Statistical Computing, 2006 19. Jiang W, Freidlin B, Simon R: Biomarker-adaptive threshold design: A procedure for evaluating treatment with possible biomarker-defined subset effect. J Natl Cancer Inst 99:1036-1043, 2007 20. Dowsett M: Designing the future shape of breast cancer diagnosis, prognosis and treatment. Breast Cancer Res Treat 87:S27-S29, 2004 (suppl 1)[CrossRef][Medline] 21. Sorenson EJ, Mandrekar J, Crum B, et al: Effect of referral bias on assessing survival in ALS. Neurology 68:600-602, 2007 22. Gibbons RJ: Referral bias: A difficult but important topic. J Nucl Cardiol 11:107-109, 2004[CrossRef][Medline] 23. Burstein HJ, Winer EP: HER2 or not HER2: That is the question. J Clin Oncol 23:3656-3659, 2005 24. Ross DT, Kim C, Tang G, et al: Chemosensitivity and stratification by a five monoclonal antibody IHC test in the NSABP B20 trial. J Clin Oncol 25:10s, 2007 (suppl; abstr 529)[CrossRef] 25. Symmans WF, Hatzis C, Sotiriou C, et al: Ability of a 200-gene endocrine sensitivity index (SET) to predict survival for patients who receive adjuvant endocrine therapy or for untreated patients. J Clin Oncol 25:72, 2007 (abstr 25) 26. Ma XJ, Hilsenbeck SG, Wang W, et al: The HOXB13:IL17BR expression index is a prognostic factor in early-stage breast cancer. J Clin Oncol 24:4611-4619, 2006 Submitted March 24, 2008; accepted May 27, 2008.
This article has been cited by other articles:
|
||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
|
|
|||||||||||
|
|||||||||||
|
|
Copyright © 2008 by the American Society of Clinical Oncology, Online ISSN: 1527-7755. Print ISSN: 0732-183X
|
TOP
ABSTRACT
INTRODUCTION
