- Email: [email protected]
Bevacizumab and breast cancer: the E2100 outlier
Reflection and Reaction
syndromes might be more common in individuals of specific ethnic origins. For example, individuals of Ashkenazi Jewish decent are more likely to inherit a BRCA1 or BRCA2 gene mutation, than are individuals of another ancestry.14 These hallmarks are considered when genetic risk-assessment models and referral guidelines are created. In a family with hereditary breast and ovarian cancer syndrome caused by a BRCA1 or BRCA2 gene mutation, we expect to see many women, on the same side of the family (maternal or paternal), who have been diagnosed with early-onset breast cancer or ovarian cancer. Because these are the highest cancer risks, referral and genetic testing guidelines for hereditary breast and ovarian cancer syndrome focus on the existence of early breast cancer and ovarian cancer; however, we might also see other cancer diagnoses, including early-onset prostate cancer, melanoma, and pancreatic cancer. With an increased awareness by health-care providers of the potential paternal transmission of BRCA1 and BRCA2 gene mutations, together with their ability to provide accurate risk assessments, fewer opportunities for cancer prevention will be missed.
Metabolic Genetics, The Hospital for Sick Children, Toronto, Canada (CS, DC); Department of Molecular Genetics (JMM, CS, DC, SRA); Department of Obstetrics and Gynecology (KJM, BR) University of Toronto, Toronto, Canada; Department of Molecular Genetics, Mount Sinai Hospital, Toronto, Canada (DC) [email protected] The authors declared no conflicts of interest. 1 2
3
4
5
6 7
8
9
10
Jeanna M McCuaig*, Celia Greenwood, Cheryl Shuman, David Chitayat, K Joan Murphy, Barry Rosen, Susan Randall Armel Division of Gynecologic Oncology, Princess Margaret HospitalUniversity Health Network, Toronto, Canada (JMM, KJM, BR, SRA); Lady Davis Institute for Medical Research, Montreal, Canada (CG); Department of Oncology; Department of Epidemiology, Biostatistics, and Occupational Health, McGill University, Montreal, Canada (CG); Division of Clinical and
11
12
13 14
GLOBOCAN 2008. Fast stats. http://globocan.iarc.fr/factsheets/ populations/factsheet.asp?uno=901 (accessed Sept 30, 2010). Slattery ML, Kerber RA. A comprehensive evaluation of family history of breast cancer risk. The Utah Population Database. JAMA 1993; 270: 1563–68. Lichtenstein P, Holm NV, Verkasalo PK, et al. Environmental and heritable factors in the causation of cancer–analyses of cohorts of twins from Sweden, Denmark, and Finland. N Engl J Med 2000; 343: 78–85. Peto J, Collins N, Barfoot R, et al. Prevalence of BRCA1 and BRCA2 gene mutations in patients with early-onset breast cancer. J Natl Cancer Inst 1999; 91: 943–49. Miki Y, Swensen J, Shattuck-Eidens, D, et al. A strong candidate for the breast and ovarian cancer susceptibility gene BRCA1. Science 1994; 266: 66–71. Wooster R, Bignell G, Lancaster J, et al. Identification of the breast cancer susceptibility gene BRCA2. Nature 1995; 378: 789–92. Ford D, Easton DF, Stratton M, et al.Genetic heterogeneity and penetrance analysis of the BRCA1 and BRCA2 genes in breast cancer families. The Breast Cancer Linkage Consortium. Am J Hum Gen 1998; 62: 676–89. Wideroff L, Vadaparampil ST, Greene MH, Taplin S, Olson L, Freedman AN. Hereditary breast/ovarian and colorectal cancer genetics knowledge in a national sample of US physicians. J Med Gen 2005; 42: 749–55. Yong MC, Zhou XJ, Lee SC. The importance of paternal family history in hereditary breast cancer is underappreciated by health care professionals. Oncology 2003; 64: 220–26. Burke W, Culver J, Pinsky L, et al. Genetic assessment of breast cancer risk in primary care practice. Am J Med Genet A 2009; 149A: 349–56. Katapodi MC, Aouizerat BE. Do women in the community recognize hereditary and sporadic breast cancer risk factors? Oncol Nurs Forum 2005; 32: 617–23. O’Neill SM, Peters JA, Vogel VG, Feingold E, Rubinstein WS. Referral to cancer genetic counseling: are there stages of readiness? Am J Med Genet C Semin Med Genet 2006; 142C: 221–31. Huang GJ, Penson DF. Internet health resources and the cancer patient. Cancer Invest 2008; 26: 202–07. Fodor FH, Weston A, Bleiweiss IJ, et al. Frequency and carrier risk associated with common BRCA1 and BRCA2 mutations in Ashkenazi Jewish breast cancer patients. Am J Hum Gen 1998; 63: 45–51.
Bevacizumab and breast cancer: the E2100 outlier The need to improve treatment for patients with metastatic breast cancer, and the heightened expectations for a drug that targets a key component of angiogenesis, led to trials combining bevacizumab with chemotherapy in metastatic breast cancer. This enthusiasm was accelerated with publication of the E2100 study.1 Designed to compare paclitaxel alone with paclitaxel plus bevacizumab, the study garnered attention when it initially reported a doubling of progression-free survival (PFS) from 5·9 to 11·8 months, albeit with no advantage in overall survival (OS) www.thelancet.com/oncology Vol 11 December 2010
(25·2 vs 26·7 months). Other combinations of bevacizumab also achieved statistical improvements in PFS, although these were less impressive than in E2100, and no combination showed a benefit in OS.2–5 Despite this lack of advantage in OS, the use of bevacizumab in metastatic breast cancer quickly gained wide acceptance in the community. However, the FDA’s Oncologic Drugs Advisory Committee have recently voted strongly against the use of bevacizumab in this setting on the basis of the later trial data,6 and here we discuss the E2100 data in the context of other trials. 1117
Dr Torsten Wittmann/SPL
Reflection and Reaction
See Online for webappendix
1118
Despite increased enthusiasm for using PFS as a surrogate for OS, debate about the wisdom of PFS as a surrogate has continued.7,8 We have noted a similar quantitative difference in PFS and OS between control and experimental groups9 and would argue that this similarity occurs because treatment does not change the biology of metastatic cancer, but only modifies it during treatment, and consequently, all benefit accrues during treatment administration.9 Provided that discontinuation of the experimental treatment does not accelerate tumour growth disproportionately, the on-treatment benefit—constrained by inaccuracies in the PFS estimation—is the benefit that accrues to OS. We have shown this benefit for 59 phase 3 trials, published since the mid-1990s, assessing chemotherapy in metastatic cancer, in which either the difference in PFS, or in OS, achieved statistical significance. The studies were identified using various search methods and references in other publications.9 Studies providing information for both PFS and OS were included. The slope of the regression line (1·21) noted in this analysis when the PFS difference versus the OS difference were plotted, emphasises the similarity between these differences for the two endpoints.9 However, in E2100, the 5·9-month advantage in the PFS that was initially reported, far exceeded the 1·5-month gain in OS. How could this be? The possibility that tumour growth accelerated disproportionately after discontinuation of bevacizumab cannot be discounted. However, the more likely explanation is that either the E2100 PFS or the E2100 OS finding is incorrect, and in view of the greater reliability of OS as an endpoint, one suspects the error is in the PFS result. With rare exceptions, measures of drug efficacy should be correlated. Therapies that shrink tumours slow tumour growth and delay time to progression; thus, a correlation between overall response rate (ORR) and PFS is expected. Such a correlation is not surprising because calculations of ORR and PFS use the same measures of tumour quantity. ORR represents maximum shrinkage and PFS is scored when the nadir value increases by 20%. A high correlation between ORR (but not stable disease) and PFS can be found in phase 2 studies with cytotoxic and targeted therapies.10 The same correlation can also be found in phase 3 studies (webfigure A). A similar correlation between PFS and ORR can also be seen in metastatic breast cancer, with E2100 identified
by a circle as an influential observation (webfigure B). That the removal of E2100 increases the correlation suggests that E2100 represents an anomalous result (red line, webfigure B). Finally, if this analysis is confined to studies assessing bevacizumab plus chemotherapy in metastatic breast cancer, a weak correlation between ORR and PFS, again with E2100 as an outlier, is noted (webfigure C). However, the correlation increases markedly when the E2100 data for bevacizumab are removed (webfigure D). For E2100, two points bear further scrutiny: (1) that a 36·9% ORR for bevacizumab plus paclitaxel achieved an 11·8-month PFS; and (2) how a 15·7% higher ORR (36·9% for bevacizumab plus paclitaxel vs 21·2% for paclitaxel) could double PFS. Imagine the excitement of investigators undertaking the AVADO trial when docetaxel plus 7·5 mg/kg or 15 mg/kg bevacizumab achieved ORRs of 55·2% and 64·1%, respectively, and their disappointment when the PFS findings were reported as 9 months and 10 months.2 How could this be when bevacizumab with paclitaxel achieved an 11·8-month PFS with a 36·9% response rate? Again, the simplest explanation is that the E2100 PFS finding is incorrect. Indeed, one would be hard pressed to show how increasing the ORR by 15·7% can double the PFS of the population from 5·9 to 11·8 months. For example, looking at studies of breast cancer included in the phase 3 trials described, differences in the ORR and PFS are unsurprisingly tightly linked (R²=0·649; p=0·0009). With a slope of 1·12 months for each 10% ORR difference a 1·76-month difference in PFS would have been predicted—much less than the 5·9 months reported. Note that stable disease does not affect PFS or OS and would be unlikely to contribute greatly to prolonging PFS.10 How then can we reconcile these results? PFS has to be assessed at fixed intervals, whereas OS is measured daily, therefore, the 12-week assessment interval in E2100 could have inflated the PFS advantage. Alternately, the apparent discrepancy might be explained by the unexpected observation that in the paclitaxel plus bevacizumab group, the median duration of paclitaxel treatment was 7·1 months, but patients did not progress until 4·7 months later (11·8 minus 7·1), even though 78·7% of the patients also stopped bevacizumab. Patients with metastatic breast cancer not receiving therapy but not experiencing disease progression for nearly 5 months suggests the trial group were unusual. www.thelancet.com/oncology Vol 11 December 2010
Reflection and Reaction
If PFS was overestimated for the bevacizumab group in E2100, then what might have been the true PFS for paclitaxel plus bevacizumab? Prediction of the PFS from the measured ORR of 36·9% (see regression line; webfigure D) yields a PFS estimate of 7·51 months. This estimate more likely reflects the true PFS. Although we recognise the theoretical nature of this analysis, we contend that the 11·8-month PFS emerges as an outlier when E2100 results are compared with other trials of metastatic breast cancer, and is probably inaccurate. A PFS difference of 1·61 months (7·51 minus 5·9 for paclitaxel alone, and similar to the 1·76-month estimate above) approximates the 1·5-month advantage in OS and aligns well with data showing concordance between actual differences between PFS and OS in phase 3 trials.9 Finally, we would address two questions. First, do patients benefit from prolonged time to progression (PFS) if OS is unchanged? Many argue that prolonged PFS is valuable; however, there is scant evidence for this view. One cannot argue that benefit accrues from keeping a patient unaware of disease progression longer if OS is not measurably improved. Furthermore, PFS should not be equated with time of onset of symptoms that might affect quality of life. Most women were probably asymptomatic when their disease was scored as progressive, and many might have been surprised to hear this news. The second question is whether marginal differences in OS should be interpreted as benefit for all patients treated? One should worry that marginal benefits occur because some patients benefit while others are harmed. In this regard, a study assessing vascular endothelial growth factor (VEGF) polymorphisms and outcome in E2100 is of interest.11 Data correlating VEGF genotype and OS in the bevacizumab group suggested a statistical benefit relative to the overall population was only achieved in the 7·6% of patients who presented with an AA/AA polymorphism at VEGF-2578/-1154. By contrast, statistical suggestion of harm was noted in the 21·5% of patients with CA/GG polymorphisms at these loci because their survival was less than the entire cohort. Although any conclusion must be tempered, because patients in the paclitaxel-only group were not genotyped and these analyses were exploratory, that PFS of patients with different genotypes were statistically similar begs for more extensive www.thelancet.com/oncology Vol 11 December 2010
investigation (Wang M, personal communication). Further study is needed because the possibility exists that three times as many women (21·5% vs 7·6%) had their survival compromised as benefited from bevacizumab treatment. These data also seem to indicate that patients with the AA/AA polymorphism had their marked benefit accrue after therapy was discontinued, in view of the 49·7-month median OS versus the 13·8-month median PFS—an unusual observation that could be explained by a more indolent disease in this group than for the other genotypes. In conclusion, we believe that the E2100 trial overestimated the benefit of bevacizumab for metastatic breast cancer. Future studies on the importance of VEGF polymorphisms might help identify a small fraction of patients that derive true benefit from bevacizumab. Tito Fojo*, Julia Wilkerson Medical Oncology Branch, National Cancer Institute, National Institutes of Health, Bethesda, MD, USA [email protected] The authors declared no conflicts of interest. 1
2
3
4
5
6 7
8
9
10
11
Miller K, Wang M, Gralow J, et al. Paclitaxel plus bevacizumab versus paclitaxel alone for metastatic breast cancer. N Engl J Med 2007; 357: 2666–76. Miles DW, Chan A, Dirix LY, et al. Phase III study of bevacizumab plus docetaxel compared with placebo plus docetaxel for the first-line treatment of human epidermal growth factor receptor 2-negative metastatic breast cancer. J Clin Oncol 2010; 28: 3239–47. Robert N, Dieras V, Glaspy J, et al. RIBBON-1: randomized, double-blind, placebo-controlled, phase III trial of chemotherapy with or without bevacizumab (B) for first-line treatment of HER2-negative locally recurrent or metastatic breast cancer (MBC). San Antonio Breast Cancer Symposium; San Antonio, TX, USA; Dec 10–13, 2009. Abstract 6084. Chan A, Miles DW, Pivot X. Bevacizumab in combination with taxanes for the first-line treatment of metastatic breast cancer. Ann Oncol 2010; published online March 24, 2010. DOI:10.1093/annonc/mdq122. O’Shaughnessy J, Dieras V, Glaspy J, et al. Comparison of subgroup analyses of PFS from three phase III studies of bevacizumab in combination with chemotherapy in patients with HER2-negative metastatic breast cancer (MBC). San Antonio Breast Cancer Symposium; San Antonio, TX, USA; Dec 10–13, 2009. Absract 207. National Cancer Institute. July 27, 2010. http://www.cancer.gov/ ncicancerbulletin/072710/page2 (accessed Nov 10, 2010). Lebwohl D, Kay A, Berg W, Baladi JF, Zheng J. Progression-free survival: gaining on overall survival as a gold standard and accelerating drug development. Cancer J 2009; 15: 386–94. Zhuang SH, Xiu L, Elsayed YA. Overall survival: a gold standard in search of a surrogate: the value of progression-free survival and time to progression as end points of drug efficacy. Cancer J 2009; 15: 395–400. Wilkerson J, Fojo T. Progression-free survival is simply a measure of a drug’s effect while administered and is not a surrogate for overall survival. Cancer J 2009; 15: 379–85. Vidaurre T, Wilkerson J, Simon R, Bates SE, Fojo T. Stable disease is not preferentially observed with targeted therapies and as currently defined has limited value in drug development. Cancer J 2009; 15: 366–73. Schneider BP, Wang M, Radovich M, for the ECOG 2100. Association of vascular endothelial growth factor and vascular endothelial growth factor receptor-2 genetic polymorphisms with outcome in a trial of paclitaxel compared with paclitaxel plus bevacizumab in advanced breast cancer: ECOG 2100. J Clin Oncol 2008; 26: 4672–78.
1119
syndromes might be more common in individuals of specific ethnic origins. For example, individuals of Ashkenazi Jewish decent are more likely to inherit a BRCA1 or BRCA2 gene mutation, than are individuals of another ancestry.14 These hallmarks are considered when genetic risk-assessment models and referral guidelines are created. In a family with hereditary breast and ovarian cancer syndrome caused by a BRCA1 or BRCA2 gene mutation, we expect to see many women, on the same side of the family (maternal or paternal), who have been diagnosed with early-onset breast cancer or ovarian cancer. Because these are the highest cancer risks, referral and genetic testing guidelines for hereditary breast and ovarian cancer syndrome focus on the existence of early breast cancer and ovarian cancer; however, we might also see other cancer diagnoses, including early-onset prostate cancer, melanoma, and pancreatic cancer. With an increased awareness by health-care providers of the potential paternal transmission of BRCA1 and BRCA2 gene mutations, together with their ability to provide accurate risk assessments, fewer opportunities for cancer prevention will be missed.
Metabolic Genetics, The Hospital for Sick Children, Toronto, Canada (CS, DC); Department of Molecular Genetics (JMM, CS, DC, SRA); Department of Obstetrics and Gynecology (KJM, BR) University of Toronto, Toronto, Canada; Department of Molecular Genetics, Mount Sinai Hospital, Toronto, Canada (DC) [email protected] The authors declared no conflicts of interest. 1 2
3
4
5
6 7
8
9
10
Jeanna M McCuaig*, Celia Greenwood, Cheryl Shuman, David Chitayat, K Joan Murphy, Barry Rosen, Susan Randall Armel Division of Gynecologic Oncology, Princess Margaret HospitalUniversity Health Network, Toronto, Canada (JMM, KJM, BR, SRA); Lady Davis Institute for Medical Research, Montreal, Canada (CG); Department of Oncology; Department of Epidemiology, Biostatistics, and Occupational Health, McGill University, Montreal, Canada (CG); Division of Clinical and
11
12
13 14
GLOBOCAN 2008. Fast stats. http://globocan.iarc.fr/factsheets/ populations/factsheet.asp?uno=901 (accessed Sept 30, 2010). Slattery ML, Kerber RA. A comprehensive evaluation of family history of breast cancer risk. The Utah Population Database. JAMA 1993; 270: 1563–68. Lichtenstein P, Holm NV, Verkasalo PK, et al. Environmental and heritable factors in the causation of cancer–analyses of cohorts of twins from Sweden, Denmark, and Finland. N Engl J Med 2000; 343: 78–85. Peto J, Collins N, Barfoot R, et al. Prevalence of BRCA1 and BRCA2 gene mutations in patients with early-onset breast cancer. J Natl Cancer Inst 1999; 91: 943–49. Miki Y, Swensen J, Shattuck-Eidens, D, et al. A strong candidate for the breast and ovarian cancer susceptibility gene BRCA1. Science 1994; 266: 66–71. Wooster R, Bignell G, Lancaster J, et al. Identification of the breast cancer susceptibility gene BRCA2. Nature 1995; 378: 789–92. Ford D, Easton DF, Stratton M, et al.Genetic heterogeneity and penetrance analysis of the BRCA1 and BRCA2 genes in breast cancer families. The Breast Cancer Linkage Consortium. Am J Hum Gen 1998; 62: 676–89. Wideroff L, Vadaparampil ST, Greene MH, Taplin S, Olson L, Freedman AN. Hereditary breast/ovarian and colorectal cancer genetics knowledge in a national sample of US physicians. J Med Gen 2005; 42: 749–55. Yong MC, Zhou XJ, Lee SC. The importance of paternal family history in hereditary breast cancer is underappreciated by health care professionals. Oncology 2003; 64: 220–26. Burke W, Culver J, Pinsky L, et al. Genetic assessment of breast cancer risk in primary care practice. Am J Med Genet A 2009; 149A: 349–56. Katapodi MC, Aouizerat BE. Do women in the community recognize hereditary and sporadic breast cancer risk factors? Oncol Nurs Forum 2005; 32: 617–23. O’Neill SM, Peters JA, Vogel VG, Feingold E, Rubinstein WS. Referral to cancer genetic counseling: are there stages of readiness? Am J Med Genet C Semin Med Genet 2006; 142C: 221–31. Huang GJ, Penson DF. Internet health resources and the cancer patient. Cancer Invest 2008; 26: 202–07. Fodor FH, Weston A, Bleiweiss IJ, et al. Frequency and carrier risk associated with common BRCA1 and BRCA2 mutations in Ashkenazi Jewish breast cancer patients. Am J Hum Gen 1998; 63: 45–51.
Bevacizumab and breast cancer: the E2100 outlier The need to improve treatment for patients with metastatic breast cancer, and the heightened expectations for a drug that targets a key component of angiogenesis, led to trials combining bevacizumab with chemotherapy in metastatic breast cancer. This enthusiasm was accelerated with publication of the E2100 study.1 Designed to compare paclitaxel alone with paclitaxel plus bevacizumab, the study garnered attention when it initially reported a doubling of progression-free survival (PFS) from 5·9 to 11·8 months, albeit with no advantage in overall survival (OS) www.thelancet.com/oncology Vol 11 December 2010
(25·2 vs 26·7 months). Other combinations of bevacizumab also achieved statistical improvements in PFS, although these were less impressive than in E2100, and no combination showed a benefit in OS.2–5 Despite this lack of advantage in OS, the use of bevacizumab in metastatic breast cancer quickly gained wide acceptance in the community. However, the FDA’s Oncologic Drugs Advisory Committee have recently voted strongly against the use of bevacizumab in this setting on the basis of the later trial data,6 and here we discuss the E2100 data in the context of other trials. 1117
Dr Torsten Wittmann/SPL
Reflection and Reaction
See Online for webappendix
1118
Despite increased enthusiasm for using PFS as a surrogate for OS, debate about the wisdom of PFS as a surrogate has continued.7,8 We have noted a similar quantitative difference in PFS and OS between control and experimental groups9 and would argue that this similarity occurs because treatment does not change the biology of metastatic cancer, but only modifies it during treatment, and consequently, all benefit accrues during treatment administration.9 Provided that discontinuation of the experimental treatment does not accelerate tumour growth disproportionately, the on-treatment benefit—constrained by inaccuracies in the PFS estimation—is the benefit that accrues to OS. We have shown this benefit for 59 phase 3 trials, published since the mid-1990s, assessing chemotherapy in metastatic cancer, in which either the difference in PFS, or in OS, achieved statistical significance. The studies were identified using various search methods and references in other publications.9 Studies providing information for both PFS and OS were included. The slope of the regression line (1·21) noted in this analysis when the PFS difference versus the OS difference were plotted, emphasises the similarity between these differences for the two endpoints.9 However, in E2100, the 5·9-month advantage in the PFS that was initially reported, far exceeded the 1·5-month gain in OS. How could this be? The possibility that tumour growth accelerated disproportionately after discontinuation of bevacizumab cannot be discounted. However, the more likely explanation is that either the E2100 PFS or the E2100 OS finding is incorrect, and in view of the greater reliability of OS as an endpoint, one suspects the error is in the PFS result. With rare exceptions, measures of drug efficacy should be correlated. Therapies that shrink tumours slow tumour growth and delay time to progression; thus, a correlation between overall response rate (ORR) and PFS is expected. Such a correlation is not surprising because calculations of ORR and PFS use the same measures of tumour quantity. ORR represents maximum shrinkage and PFS is scored when the nadir value increases by 20%. A high correlation between ORR (but not stable disease) and PFS can be found in phase 2 studies with cytotoxic and targeted therapies.10 The same correlation can also be found in phase 3 studies (webfigure A). A similar correlation between PFS and ORR can also be seen in metastatic breast cancer, with E2100 identified
by a circle as an influential observation (webfigure B). That the removal of E2100 increases the correlation suggests that E2100 represents an anomalous result (red line, webfigure B). Finally, if this analysis is confined to studies assessing bevacizumab plus chemotherapy in metastatic breast cancer, a weak correlation between ORR and PFS, again with E2100 as an outlier, is noted (webfigure C). However, the correlation increases markedly when the E2100 data for bevacizumab are removed (webfigure D). For E2100, two points bear further scrutiny: (1) that a 36·9% ORR for bevacizumab plus paclitaxel achieved an 11·8-month PFS; and (2) how a 15·7% higher ORR (36·9% for bevacizumab plus paclitaxel vs 21·2% for paclitaxel) could double PFS. Imagine the excitement of investigators undertaking the AVADO trial when docetaxel plus 7·5 mg/kg or 15 mg/kg bevacizumab achieved ORRs of 55·2% and 64·1%, respectively, and their disappointment when the PFS findings were reported as 9 months and 10 months.2 How could this be when bevacizumab with paclitaxel achieved an 11·8-month PFS with a 36·9% response rate? Again, the simplest explanation is that the E2100 PFS finding is incorrect. Indeed, one would be hard pressed to show how increasing the ORR by 15·7% can double the PFS of the population from 5·9 to 11·8 months. For example, looking at studies of breast cancer included in the phase 3 trials described, differences in the ORR and PFS are unsurprisingly tightly linked (R²=0·649; p=0·0009). With a slope of 1·12 months for each 10% ORR difference a 1·76-month difference in PFS would have been predicted—much less than the 5·9 months reported. Note that stable disease does not affect PFS or OS and would be unlikely to contribute greatly to prolonging PFS.10 How then can we reconcile these results? PFS has to be assessed at fixed intervals, whereas OS is measured daily, therefore, the 12-week assessment interval in E2100 could have inflated the PFS advantage. Alternately, the apparent discrepancy might be explained by the unexpected observation that in the paclitaxel plus bevacizumab group, the median duration of paclitaxel treatment was 7·1 months, but patients did not progress until 4·7 months later (11·8 minus 7·1), even though 78·7% of the patients also stopped bevacizumab. Patients with metastatic breast cancer not receiving therapy but not experiencing disease progression for nearly 5 months suggests the trial group were unusual. www.thelancet.com/oncology Vol 11 December 2010
Reflection and Reaction
If PFS was overestimated for the bevacizumab group in E2100, then what might have been the true PFS for paclitaxel plus bevacizumab? Prediction of the PFS from the measured ORR of 36·9% (see regression line; webfigure D) yields a PFS estimate of 7·51 months. This estimate more likely reflects the true PFS. Although we recognise the theoretical nature of this analysis, we contend that the 11·8-month PFS emerges as an outlier when E2100 results are compared with other trials of metastatic breast cancer, and is probably inaccurate. A PFS difference of 1·61 months (7·51 minus 5·9 for paclitaxel alone, and similar to the 1·76-month estimate above) approximates the 1·5-month advantage in OS and aligns well with data showing concordance between actual differences between PFS and OS in phase 3 trials.9 Finally, we would address two questions. First, do patients benefit from prolonged time to progression (PFS) if OS is unchanged? Many argue that prolonged PFS is valuable; however, there is scant evidence for this view. One cannot argue that benefit accrues from keeping a patient unaware of disease progression longer if OS is not measurably improved. Furthermore, PFS should not be equated with time of onset of symptoms that might affect quality of life. Most women were probably asymptomatic when their disease was scored as progressive, and many might have been surprised to hear this news. The second question is whether marginal differences in OS should be interpreted as benefit for all patients treated? One should worry that marginal benefits occur because some patients benefit while others are harmed. In this regard, a study assessing vascular endothelial growth factor (VEGF) polymorphisms and outcome in E2100 is of interest.11 Data correlating VEGF genotype and OS in the bevacizumab group suggested a statistical benefit relative to the overall population was only achieved in the 7·6% of patients who presented with an AA/AA polymorphism at VEGF-2578/-1154. By contrast, statistical suggestion of harm was noted in the 21·5% of patients with CA/GG polymorphisms at these loci because their survival was less than the entire cohort. Although any conclusion must be tempered, because patients in the paclitaxel-only group were not genotyped and these analyses were exploratory, that PFS of patients with different genotypes were statistically similar begs for more extensive www.thelancet.com/oncology Vol 11 December 2010
investigation (Wang M, personal communication). Further study is needed because the possibility exists that three times as many women (21·5% vs 7·6%) had their survival compromised as benefited from bevacizumab treatment. These data also seem to indicate that patients with the AA/AA polymorphism had their marked benefit accrue after therapy was discontinued, in view of the 49·7-month median OS versus the 13·8-month median PFS—an unusual observation that could be explained by a more indolent disease in this group than for the other genotypes. In conclusion, we believe that the E2100 trial overestimated the benefit of bevacizumab for metastatic breast cancer. Future studies on the importance of VEGF polymorphisms might help identify a small fraction of patients that derive true benefit from bevacizumab. Tito Fojo*, Julia Wilkerson Medical Oncology Branch, National Cancer Institute, National Institutes of Health, Bethesda, MD, USA [email protected] The authors declared no conflicts of interest. 1
2
3
4
5
6 7
8
9
10
11
Miller K, Wang M, Gralow J, et al. Paclitaxel plus bevacizumab versus paclitaxel alone for metastatic breast cancer. N Engl J Med 2007; 357: 2666–76. Miles DW, Chan A, Dirix LY, et al. Phase III study of bevacizumab plus docetaxel compared with placebo plus docetaxel for the first-line treatment of human epidermal growth factor receptor 2-negative metastatic breast cancer. J Clin Oncol 2010; 28: 3239–47. Robert N, Dieras V, Glaspy J, et al. RIBBON-1: randomized, double-blind, placebo-controlled, phase III trial of chemotherapy with or without bevacizumab (B) for first-line treatment of HER2-negative locally recurrent or metastatic breast cancer (MBC). San Antonio Breast Cancer Symposium; San Antonio, TX, USA; Dec 10–13, 2009. Abstract 6084. Chan A, Miles DW, Pivot X. Bevacizumab in combination with taxanes for the first-line treatment of metastatic breast cancer. Ann Oncol 2010; published online March 24, 2010. DOI:10.1093/annonc/mdq122. O’Shaughnessy J, Dieras V, Glaspy J, et al. Comparison of subgroup analyses of PFS from three phase III studies of bevacizumab in combination with chemotherapy in patients with HER2-negative metastatic breast cancer (MBC). San Antonio Breast Cancer Symposium; San Antonio, TX, USA; Dec 10–13, 2009. Absract 207. National Cancer Institute. July 27, 2010. http://www.cancer.gov/ ncicancerbulletin/072710/page2 (accessed Nov 10, 2010). Lebwohl D, Kay A, Berg W, Baladi JF, Zheng J. Progression-free survival: gaining on overall survival as a gold standard and accelerating drug development. Cancer J 2009; 15: 386–94. Zhuang SH, Xiu L, Elsayed YA. Overall survival: a gold standard in search of a surrogate: the value of progression-free survival and time to progression as end points of drug efficacy. Cancer J 2009; 15: 395–400. Wilkerson J, Fojo T. Progression-free survival is simply a measure of a drug’s effect while administered and is not a surrogate for overall survival. Cancer J 2009; 15: 379–85. Vidaurre T, Wilkerson J, Simon R, Bates SE, Fojo T. Stable disease is not preferentially observed with targeted therapies and as currently defined has limited value in drug development. Cancer J 2009; 15: 366–73. Schneider BP, Wang M, Radovich M, for the ECOG 2100. Association of vascular endothelial growth factor and vascular endothelial growth factor receptor-2 genetic polymorphisms with outcome in a trial of paclitaxel compared with paclitaxel plus bevacizumab in advanced breast cancer: ECOG 2100. J Clin Oncol 2008; 26: 4672–78.
1119