|
Advertisement |
|
|
||
 |
|
|||||
|
|||||
|
 |
||||||
|
||||||
|
||||||
|
© 2002 American Society for Clinical Oncology High-Dose Versus Standard Chemotherapy in Metastatic Breast Cancer: Comparison of Cancer and Leukemia Group B Trials With Data From the Autologous Blood and Marrow Transplant RegistryByFrom the University of Texas M.D. Anderson Cancer Center, Houston, and University of Texas, Health Science Center at San Antonio, San Antonio, TX; Duke University Medical Center, Durham, NC; Autologous Blood and Marrow Transplant Registry, Milwaukee, and Health Policy Institute, Medical College of Wisconsin, Milwaukee, WI; Columbia University, Memorial Sloan-Kettering Cancer Center, New York, and Roswell Park Cancer Institute, Buffalo, NY; The Cancer Institute of New Jersey, New Brunswick, and Progenitor Cell Therapy LLC, Saddle Brook, NJ; Cancer and Leukemia Group B, Chicago, and Lutheran General Hospital, Park Ridge, IL; University of Massachusetts Medical Center, Boston, MA; University of Pennsylvania Cancer Center, Philadelphia, PA; University of California at San Francisco, San Francisco, CA; Case Western Reserve University, Cleveland, OH; National Cancer Institute, Bethesda, MD; University of Missouri/Ellis Fischel Cancer Center, Columbia, MO; Prince of Wales Hospital, Randwick Sydney, New South Wales, Australia; and Cancer Center of the Carolinas, Greenville, SC. Address reprint requests to Donald A. Berry, PhD, Department of Biostatistics, University of Texas M.D. Anderson Cancer Center, 1515 Holcombe Blvd, Box 447, Houston, TX 77030-4009; email: dberry{at}odin.mdacc.tmc.edu
PURPOSE: To assess survival of patients with metastatic breast cancer treated with high-dose chemotherapy (HDC) versus standard-dose chemotherapy (SDC). PATIENTS AND METHODS: SDC in four Cancer and Leukemia Group B (CALGB) trials was compared with hematopoietic stem-cell support in patients from the Autologous Blood and Marrow Transplant Registry. Cox proportional hazard regression incorporated potentially confounding effects. A total of 1,509 women were enrolled onto CALGB trials, and 1,188 women received HDC. No significant survival differences existed by CALGB trial or HDC regimen. Consideration was restricted to candidates for both SDC and HDC. The resulting sample included 635 SDC and 441 HDC patients. The outcome of interest was overall survival. RESULTS: The HDC group displayed better performance status. The SDC group had slightly better survival in first year after treatment. The HDC group had lower hazard of death from years 1 to 4 and had somewhat higher probability of 5-year survival (adjusted probabilities [95% confidence intervals], 23% [17% to 29%] v 15% [11% to 19%], P = .03). CONCLUSION: After controlling for known prognostic factors in this nonrandomized analysis of two large independent data sets, women receiving HDC versus SDC for metastatic breast cancer have a similar short-term probability of survival, and might have a modestly higher long-term probability of survival.
THE ROLE OF HIGH-DOSE chemotherapy (HDC) with autologous bone marrow or peripheral blood progenitor stem-cell support (autotransplants) is controversial in breast cancer.1,2 Three randomized trials have compared HDC with standard-dose chemotherapy (SDC) in metastatic disease.3-5 The largest of the three included 553 patients, 199 of whom were randomized and 15 of these were deemed ineligible. An intent-to-treat analysis of the 184 eligible patients—including 20 who refused their assigned treatment—indicated no survival difference between the two groups.3 Median survival times were 2.00 years for women assigned to HDC (n = 101) and 2.33 years for those assigned to SDC (n = 83) (P = not significant). The other two trials suggested a benefit for HDC. One indicated a relatively large but not statistically significant survival advantage for HDC (median, 3.01 years; n = 32 v 1.31 years, n = 29; P = .08).4 The third study indicated a statistically significant survival advantage for HDC; however, it is currently being audited, so we cannot assess its reliability at this time.5
Conflicting results may derive from differences in trial design, drug regimens used, and patients studied. Perhaps more importantly, with a total of only 178 patients randomized to HDC and 153 patients randomized to SDC, the statistical power of these trials to detect clinically meaningful survival differences was limited. Power was especially limited for detecting a survival advantage of HDC that might accrue to a small proportion of women and therefore would not be evident until perhaps
To address the question of whether HDC improves survival of women with metastatic breast cancer, we compared two large databases. The first consisted of women participating in four Cancer and Leukemia Group B (CALGB) trials. The second consisted of women receiving autotransplants registered with the Autologous Blood and Marrow Transplant Registry (ABMTR). The SDC regimen depended on the CALGB trial, but HDC plus autotransplant was not a possible assignment in any trial. Because patients in the CALGB trials were restricted to those receiving their first course of chemotherapy for metastatic disease, we eliminated patients from the ABMTR who did not meet this criteria. Also, in both groups, we restricted consideration to women whose disease responded to conventional-dose chemotherapy and who were
Patients The SDC cohort included women accrued to four trials of chemotherapy for metastatic breast cancer performed from 1980 to 1992: CALGB 8081, 8281, 8642, and 9140 (Table 1). The HDC cohort included women receiving HDC and autotransplants between 1989 and 1995 and reported to the ABMTR (Table 2). These cohorts are described extensively elsewhere.6-10
Patients in the SDC group were compared by arm within each CALGB trial and also across the four CALGB trials (Fig 1). There were no significant survival differences. A prior study evaluated the impact of HDC regimen in the autotransplant cohort; no significant differences were found.7 One may expect a trend that indicates improved survival with time in the HDC group because peripheral blood stem cells were predominantly used in the more recently treated patients, and changes in supportive care may have favored improved outcome. However, on the basis of a Cox regression analysis (data not shown), there is no evidence of a trend in improved survival over time (P > .8) in the HDC group.
Statistical Methods There are several potential biases in comparing SDC to HDC in a nonrandomized study. First, induction therapy with standard doses of chemotherapy is used before HDC in most patients receiving autotransplants. Patients with a complete response (CR) or partial response (PR) to induction therapy are preferred candidates for transplantation.6 We addressed this bias by considering only those patients in both groups whose disease was chemosensitive. Specifically, in the ABMTR cohort, we restricted analyses to patients achieving CR or PR before HDC, and in the CALGB cohort, we restricted analyses to women achieving CR or PR on protocol therapy. Second, HDC is often used in women whose disease does not respond to other therapies, and the CALGB trials included only women with previously untreated metastatic breast cancer. We addressed this bias by excluding from the HDC cohort women who had received one or more prior treatment regimens in the metastatic setting. Table 3 lists the numbers of patients excluded according to each criterion.
Women receiving HDC and SDC may differ in other ways. For example, HDC is generally offered to younger women, women with good performance status, and women who are at high risk of experiencing unsuccessful conventional chemotherapy. To adjust for differences in potentially important patient and disease characteristics, we evaluated the association between survival and the following factors for each cohort, in univariate analyses by the log-rank test: patient age (in women no older than 65 years), performance status, sites of metastases, disease-free interval (time from first diagnosis of primary breast cancer to metastasis), hormone receptor status, and whether the patient had received prior adjuvant chemotherapy or tamoxifen. Variables with statistically significant (P < .05) associations with survival in univariate analysis were included in a multivariate Cox proportional hazards regression model comparing time to death for HDC versus SDC. The starting time for the SDC cohort in this model was the day of achieving CR or PR; the starting point in the HDC cohort was the day of autotransplant. The proportional hazard assumption was checked for all covariates. Variables with nonproportional hazards were handled by time-dependent covariates. First-order interactions between treatment (HDC v SDC) and other significant covariates were evaluated and were included in the model if they were statistically significant. Another potential source of bias is that patients in the ABMTR cohort had to live long enough to receive a transplant. Patients who might have been scheduled for transplantation after initial chemotherapy, but who died before they could receive it, are not included. This bias is partially (probably mostly) resolved by restricting analyses to women with chemosensitive disease because responding patients do not usually die early. Any unresolved portion of this bias probably favors the HDC group. We addressed the possibility of residual bias by left-truncated proportional hazards regression.11,12 At each time point in this model, the risk set in the SDC cohort consisted of all patients achieving CR or PR on CALGB trials, whereas the risk set in the HDC cohort included only those women with a waiting time to HDC (time between metastases and HDC) less than the current time point. The result was to give less weight to deaths in the SDC cohort that occur early—that is, before transplantation could have been performed, had one been intended, thus reducing time-to-treatment bias.13 Results of left-truncated and nontruncated models were similar, suggesting little effect of time to treatment. For simplicity, we present results for the nontruncated models only. Kaplan-Meier estimates of the two cohorts were compared by the Wilcoxon and log-rank tests.14,15 Relatively speaking, the former is more sensitive to early cohort differences, and the latter is more sensitive to late cohort differences. In addition, we indicate unadjusted annual hazards of death and their 95% confidence intervals (CIs) for the two groups. These are calculated for the first 3 months and in 6-month intervals thereafter. The hazard for any period is the (annualized) proportion of women who died during that period in comparison with the number alive at the beginning of the period. Therefore, each period represents an independent comparison of the two groups. Because the numbers of events within each interval are usually small, no formal statistical comparisons of these hazards are made. Adjusted probabilities of survival for women receiving HDC and SDC were calculated by a multivariate Cox model, weighted by the sample proportion value for each prognostic factor. This allows comparison of expected survival rates in two populations with a similar distribution of prognostic factors. The 95% CIs for adjusted survival probabilities and P values of pairwise comparisons were derived from pointwise estimates and calculated by standard techniques.16 Unadjusted Kaplan-Meier estimates of probability of survival for HDC and SDC were calculated for all patients and in subgroups defined by prognostic factors significant in the Cox model: women with estrogen receptor–positive versus –negative tumors; those with CR versus PR after conventional therapy; those who had received adjuvant chemotherapy versus those who had not; those with short (< 18 months) versus long disease-free interval to metastasis after primary treatment; and performance status (0 v > 0). In addition, unadjusted Kaplan-Meier estimates for HDC and SDC were compared for factors of potential interest to clinicians, even though none had strong prognostic significance in this data set: age (< 50 years v > 50 years old); number of sites of metastasis (one site v more than one site); and site of metastasis (bone only v soft tissue only v any visceral site). Statistical significance was taken to be P < .05. P values are given for the overall comparison of HDC and SDC and for prognostic factors within the multivariate model. However, in view of the problem of multiple comparisons, neither P values nor CIs for survival probabilities are given within patient subsets. No statistical inferences were made or should be inferred for the comparison of HDC and SDC within patient subsets.
The original data set contained 2,697 women with metastatic breast cancer, 1,509 treated with SDC enrolled onto CALGB trials and 1,188 receiving HDC and reported to the ABMTR (Table 4). Restricting the analysis to patients  65 years old who had achieved CR or PR after a single chemotherapy regimen reduces the sample size to 1,076 core patients: 635 receiving SDC and 441 receiving HDC (Table 4). These sample sizes provide 90% power to detect a 20% difference and 70% power to detect a 15% difference in median survival (assuming a two-sided significance level of .05).
Table 4 compares patient characteristics of the HDC and SDC groups and, within each of these groups, compares the core sample to the initial data set. Patients in the HDC group were significantly younger. They also had significantly better performance status. HDC patients more frequently had estrogen receptor–positive tumors (56% v 43%) and more frequently had received prior hormonal therapy (32% v 23%). The percentage of patients who had estrogen receptor–positive tumors and who had not yet received hormonal therapy was 52% for the HDC group and 64% for the SDC group. HDC patients were also more likely to have had adjuvant chemotherapy as part of their primary breast cancer treatment and to have disease restricted to bone or soft tissue. Figure 2 compares the survival of all core patients by treatment approach—that is, HDC versus SDC. This comparison is not adjusted for any differences in patient characteristics. Estimated survival is greater for SDC during the first 15 months; then it is greater for HDC. One-hundred-day mortality is 4.5% and 0.8% in the HDC and SDC arms, respectively. The 3- and 5-year probabilities of survival are 37% (95% CI, 32% to 42%) and 22% (95% CI, 18% to 27%) in the HDC cohort, and 27% (95% CI, 23% to 31%) and 13% (95% CI, 10% to 16%) in the SDC cohort. The medians are similar: 1.94 v 1.83 years. The area between the two survival curves is the estimated advantage in life expectancy for HDC, which is approximately 4 months (up to 7 years). Comparing the survival distributions by the Wilcoxon test (which gives more weight to early observations) gives a P value of .45 (not significant); comparing the distributions by the log-rank test (which gives more weight to late observations than does the Wilcoxon) gives a P value of .01.
Figure 3 compares hazards of death and their 95% CIs in the two groups. The risk of death is greater in the HDC cohort during the first year, but greater in the SDC cohort during each time interval between years 1 and 4. The CIs overlap because they are based on small numbers of events, and they become wider over time as the numbers of patients at risk get smaller. Over the 3.5-year period from month 9 to month 51, the estimated annual hazard of death with SDC was 50%, whereas with HDC it was 30%.
Table 5 presents the multivariate Cox proportional hazards models. Results are expressed as the relative risk (95% CI) of mortality for patients with the corresponding risk factor versus those without it. Initial analyses indicated that treatment (HDC v SDC) had nonproportional hazards such that the relative risk of mortality for women in the two cohorts differed over time. Consequently, relative risks are provided for the early (< 0.7 years) and late (  0.7 years) posttreatment time periods. Additionally, the Cox model suggested an interaction between treatment and the following variables: estrogen receptor status, use of prior adjuvant chemotherapy, disease-free interval after primary treatment, and response to conventional-dose chemotherapy. Consequently, relative risks of mortality with HDC versus SDC are provided separately for subgroups defined by these variables. Adjusting probabilities of survival by cohort for differences in prognostic factors gives essentially the same curves as the Kaplan-Meier curves in Fig 2. Table 6 presents adjusted probabilities of survival. Survival is similar for the first 2 years, but by 3 years, there is a statistically significant advantage for HDC of 39% v 31%.
Figure 4 illustrates unadjusted probabilities of survival in subgroups defined by various prognostic factors. This figure is important because it demonstrates that the comparison of HDC and SDC is essentially the same when separating the patients into subgroups. Exceptions are those suggested by the multivariate analysis presented in Table 5: there may be a differential relative effectiveness of HDC versus SDC depending on disease-free interval and whether the patient received adjuvant chemotherapy. The apparent interaction between therapy and CR versus PR is probably an artifact of the assessment process.
Progress in the treatment of metastatic breast cancer during the 1980s and early 1990s did little or nothing to prolong survival. The survival of patients treated in the four CALGB trials conducted over this period did not improve. By comparing the survival of patients in these trials with that of patients receiving high-dose therapy with autotransplants, we assessed whether such a strategy holds any promise for changing this trend.
Our retrospective study of 1,079 women with metastatic breast cancer, aged This study was a comparison that used two large databases of clinical information. It has the statistical power to detect a 20% difference in 3-year survival. It was not randomized. Randomization, which minimizes patient assignment bias,17 is the gold standard of medical research.18,19 An important limitation of randomization is the difficulty in accruing sufficient numbers of patients to assess interventions with adequate statistical power. Registries and databases contain valuable information regarding treatment efficacy, often in large numbers of patients. Comparisons that use such databases can have high power because of large sample sizes, and they are particularly useful in complementing data from randomized trials. The major concern about drawing inferences in nonrandomized settings is selection bias: patients’ treatments may be assigned for reasons related to their prognoses. Observed benefits may therefore result from prognosis rather than treatment. Comparing patients in databases is credible if patients assigned to different therapies have similar prognoses, if differences in prognosis can be assessed on the basis of measurable characteristics, or both. All relevant patient characteristics known to treating physicians must be available for analysis, as was the case in the present study. Importantly, disease staging in the treatment groups must be similar. On the basis of criteria for reporting and careful review of reported data, we were confident that all women in this study had chemosensitive disease. The randomized trials comparing HDC with SDC in metastatic breast cancer are small and have limited power. For example, the largest of these is the Philadelphia trial3 with a total sample size of 184, giving it only 20% power to detect a 20% improvement in median survival. The median follow-up of patients in the Philadelphia trial when reported was approximately 3 years; consequently, there is essentially no power to detect a difference in 3- to 5-year survival of the magnitude seen in our study. When survival after HDC in the ABMTR is compared with survival in the HDC arm of the Philadelphia trial, and when survival in the CALGB trials is compared with survival in the SDC arm in the Philadelphia trial, the results are similar. Median durations of survival in our study were 1.83 (SDC) and 1.94 (HDC) years, whereas in the Philadelphia trial they were 2.33 (SDC) and 2.00 (HDC) years. Therefore, the present study raises the question whether there may be a small survival advantage with HDC that is apparent only with larger numbers of patients and longer follow-up than provided for in the Philadelphia trial. Because of the vagaries of interpreting subset analyses, we have not included significance levels for Fig 4. However, we will discuss two of the observations from these figures. One is a potential problem in differentiating CR versus PR in the two cohorts. Response assessment (ie, tests conducted, definitions of response, amount of treatment before assessment) was rigidly dictated by protocol in the CALGB trials, but could potentially differ among transplantation centers. There is an apparent beneficial effect for SDC among complete responders (Fig 4e) and for HDC among partial responders (Fig 4f). This apparent interaction is probably an artifact of the assessment process; we do not believe that it is real. A second observation is the suggestion of increased benefit for HDC in patients who did not receive adjuvant chemotherapy (Fig 4l). A possible explanation is that patients whose tumors have not yet failed to respond to the best treatment regimens would be most likely to benefit from HDC. Tumor cells that had already survived standard regimens would be more likely to be resistant at the time of reinduction and HDC. However, any such explanations are post hoc and subject to the usual concerns about multiple comparisons and subset analyses.20 The results of randomized trials of chemotherapy in women with metastatic breast cancer conducted by the CALGB between 1980 and 1992 are disheartening. Not only was overall survival poor, but there were no significant differences by type and schedule of chemotherapy used, and there was no noticeable improvement in survival over this time period. Moreover, our analysis indicates that any improvement in survival associated with HDC was delayed by as much as 3 years. Clearly, metastatic breast cancer is relatively resistant to all the therapies considered in our study. Of the 97 women who received HDC after a PR to induction therapy, 42 women (43%) achieved a CR with HDC. Such a conversion from PR to CR may be a necessary (though not sufficient) requirement for prolonged disease-free survival. Addition of other novel agents to a state of minimal residual disease after HDC may provide further benefit. Of particular interest in this regard are immune therapies with antibodies (eg, trastuzumab), dendritic cells, or allogeneic cells. Better approaches are needed to improve the prognosis of women with metastatic breast cancer. The information available in the present study and in other studies of SDC and HDC should be used to guide future trials of new and, we hope, better treatment strategies.
Supported in part by grants from the National Cancer Institute through research grant no. CA31946 to the Cancer and Leukemia Group B, Richard L. Schilsky, chairman, by the Cancer and Leukemia Group B Statistical Office grant no. CA 33601, and by Public Health Service grant nos. P01-CA-40053 and U24-76518, from the National Cancer Institute, the National Institute of Allergy and Infectious Diseases, and the National Heart, Lung and Blood Institute of the United States Department of Health and Human Services; grant no. DAMD17-95-I-5002 from the Department of the United States Army Medical Research and Development Command; and by grants from Alpha Therapeutic Corp; Amgen, Inc; Anonymous; Baxter Fenwal; Berlex Laboratories; BioWhittaker, Inc; Blue Cross and Blue Shield Association; Bristol-Myers Squibb Co; Cell Therapeutics, Inc; Centeon; Center for Advanced Studies in Leukemia; Chimeric Therapies, Inc; Chiron Therapeutics; COBE BCT, Inc; the Eppley Foundation for Research; Fromstein Foundation; Fujisawa Healthcare, Inc; Genentech, Inc; Human Genome Sciences; Immunex Corp; the Kettering Family Foundation; Kirin Brewery Company; the Robert J. Kleberg Jr and Helen C. Kleberg Foundation; Mayer Ventures; the Milstein Family Foundation; Milwaukee Foundation/Elsa Schoeneich Research Fund; Nexell Therapeutics, Inc; NeXstar Pharmaceuticals, Inc; Novartis Pharmaceuticals; Orphan Medical; Ortho Biotech, Inc; Pharmacia and Upjohn; Pfizer, Inc; Roche Laboratories; SangStat Medical Corp; Schering AG; Schering Oncology; Searle; SmithKline Beecham Pharmaceutical; the Starr Foundation; SyStemix; TheraTechnologies; and United Resource Networks.
1. Zujewski J, Nelson A, Abrams J: Much ado about not . . . enough data: High-dose chemotherapy with autologous stem cell rescue for breast cancer. J Natl Cancer Inst 90: 200-209, 1998 2. Hudis CA, Münster PN: High-dose therapy for breast cancer. Semin Oncol 26: 35-47, 1999[Medline]
3.
Stadtmauer EA, O’Neill A, Goldstein LJ, et al: Conventional-dose chemotherapy compared with high-dose chemotherapy plus autologous hematopoietic stem-cell transplantation for metastatic breast cancer. N Engl J Med 342: 1069-1076, 2000 4. Lotz J-P, Curé H, Janvier M, et al: High-dose chemotherapy (HD-CT) with hematopoietic stem cell transplantation (HSCT) for metastatic breast cancer (MBC): Results of the French protocol PEGASE 04. Proc Am Soc Clin Oncol 18: 43a, 1999 (abstr 161) 5. Bezwoda WR: Primary high dose chemotherapy for metastatic breast cancer: Update and analysis of prognostic factors. Proc Am Soc Clin Oncol 17: 115a, 1998 (abstr 445)
6.
Antman KH, Rowlings PA, Vaughan WP, et al: High-dose chemotherapy with autologous hematopoietic stem-cell support for breast cancer in North America. J Clin Oncol 15: 1870-1879, 1997
7.
Rowlings PA, Williams SF, Antman KH, et al: Factors correlated with progression-free survival after high-dose therapy and hematopoietic stem cell transplantation for metastatic breast cancer. JAMA 282: 1335-1343, 1999
8.
Perloff M, Norton L, Korzun AH, et al: Post-surgical adjuvant chemotherapy of stage II breast carcinoma with or without crossover to a non–cross-resistant regimen: A Cancer and Leukemia Group B study. J Clin Oncol 14: 1589-1598, 1996 9. Aisner J, Cirrincione C, Perloff M, et al: Combination chemotherapy for metastatic or recurrent carcinoma of the breast: A randomized phase III trial comparing: cyclophosphamide, doxorubicin and 5-fluorouracil (CAF) versus vinblastine, doxorubicin, thiotepa and Halotestin (VATH) versus VATH alternating with cyclophosphamide, methotrexate, 5-fluorouracil, vinblastine and prednisone (CMFVP): Cancer and Leukemia Group B study 8281. J Clin Oncol 13: 1443-1452, 1995[Abstract]
10.
Costanza ME, Weiss RB, Henderson IC, et al: The safety and efficacy of using a single agent or phase II agent before instituting standard combination chemotherapy in previously untreated metastatic breast cancer patients: Report of a randomized study—CALGB 8642. J Clin Oncol 17: 1397-1406, 1999 11. Cox DR: Regression models and life-tables. J R Stat Soc Ser B 34: 187-220, 1972 (with discussion) 12. Keiding N: Independent delayed entry, in Klein JP, Goel P (eds): Survival Analysis: State of the Art. Boston, MA, Kluwer Academic Publishers, 1992, pp 309-326 13. Klein JP, Zhang MJ: Statistical challenges in comparing chemotherapy and bone marrow transplantation as a treatment for leukemia, in Jewell NP, Kimber AC, Lee MT, et al (eds): Lifetime Data: Models in Reliability and Survival Analysis. Norwell, MA, Kluwer Academic Press, 1996, pp 175-186 14. Gehan EA: A generalized Wilcoxon test for comparing K samples subject to unequal patterns of censorship. Biometrika 52: 203-223, 1995[CrossRef] 15. Peto R, Pike C, Armitage P: Design and analysis of randomized clinical trials requiring prolonged observation of each patient. Br J Cancer 35: 7-9, 1977 16. Klein JP, Moeschberger ML: Survival Analysis: Techniques of Censored and Truncated Data. New York, NY, Springer-Verlag, 1996 17. Byar DP, Simon RM, Friedewald WT, et al: Randomized clinical trials: Perspectives on some recent ideas. N Engl J Med 295: 74-80, 1976[Abstract] 18. Garcoa CR, Hidalgo M, Paz-Ares L, et al: Patient selection in high-dose chemotherapy trials: Relevance in high-risk breast cancer. J Clin Oncol 15: 3178-3184, 1997[Abstract] 19. Rahman ZU, Frye DK, Bazdar AU, et al: Impact of selection process on response rate and long-term survival of potential high-dose chemotherapy candidates treated with standard-dose doxorubicin-containing chemotherapy in patients with metastatic breast cancer. J Clin Oncol 15: 3171-3177, 1997[Abstract] 20. Berry DA: Subgroup analyses. Biometrics 47: 1227-1230, 1990 Submitted November 13, 2000; accepted September 7, 2001.
This article has been cited by other articles:
|
||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
|
|
|||||||||||
|
|||||||||||
|
|
Copyright © 2002 by the American Society of Clinical Oncology, Online ISSN: 1527-7755. Print ISSN: 0732-183X
|
TOP
ABSTRACT
INTRODUCTION
