original articles
Annals of Oncology 9. Gnant M. The role of mammalian target of rapamycin (mTOR) inhibition in the treatment of advanced breast cancer. Curr Oncol Rep 2013; 15(1): 14–23. 10. Jerusalem G, Rorive A, Collignon J. Use of mTOR inhibitors in the treatment of breast cancer: an evaluation of factors that influence patient outcomes. Breast Cancer 2014; 6: 43–57. 11. Jerusalem G, Bachelot T, Neven P et al. A new era of improving progression-free survival with dual blockade in postmenopausal HR(+), HER2(−) advanced breast cancer. Cancer Treat Rev 2015; 4(2): 94–104. 12. Gnant M. Overcoming endocrine resistance in breast cancer: importance of mTOR inhibition. Expert Rev Anticancer Ther 2012; 12(12): 1579–1589. 13. Buolay A, Rudloff J, Ye J et al. Dual inhibition of mTOR and estrogen receptor signaling in vitro induces cell death in models of breast cancer. Clin Cancer Res 2005; 11(14): 5319–5328.
14. Yardley DA, Noguchi S, Burris HA et al. Everolimus plus exemestane in postmenopausal patients with HR(+) breast cancer: BOLERO-2 final progressionfree survival analysis. Adv Ther 2013; 30(10): 870–884. 15. Piccart M, Hortobagyi GN, Campone M et al. Everolimus plus exemestane for hormone-receptor-positive, human epidermal growth factor receptor-2-negative advanced breast cancer: overall survival results from BOLERO-2. Ann Oncol 2014; 25(12): 2357–2362. 16. Pritchard KI, Burris HA, Ito Y et al. Safety and efficacy of everolimus with exemestane vs. exemestane alone in elderly patients with HER2-negative, hormone receptor-positive breast cancer in BOLERO-2. Clin Breast Cancer. 2013; 13(6): 421–432.e8. 17. National Comprehensive Cancer Network. NCCN Clinical Practice Guidelines in Oncology: Breast Cancer. V1. 2015.
Annals of Oncology 27: 1725–1732, 2016 doi:10.1093/annonc/mdw260 Published online 19 July 2016
S. Delaloge1*, D. Pérol2, C. Courtinard3, E. Brain4, B. Asselain3, T. Bachelot2, M. Debled5, V. Dieras4, M. Campone6, C. Levy7, W. Jacot8, V. Lorgis9, C. Veyret10, F. Dalenc11, J. M. Ferrero12, L. Uwer13, P. Kerbrat14, A. Goncalves15, M. A. Mouret-Reynier16, T. Petit17, C. Jouannaud18, L. Vanlemmens19, G. Chenuc20, T. Guesmia3, M. Robain3 & C. Cailliot3 1
Department of Cancer Medicine, Institut Gustave Roussy, Villejuif; 2Department of Biostatistics, Centre Léon Bérard, Lyon; 3Department of Research and Development, R&D Unicancer, Paris; 4Department of Medical Oncology, Institut Curie, Paris and Saint-Cloud; 5Department of Medical Oncology, Institut Bergonié, Bordeaux; 6 Department of Medical Oncology, Institut de Cancérologie de l’Ouest, Nantes and Angers; 7Department of Medical Oncology, Centre François Baclesse, Caen; 8 Department of Medical Oncology, Institut du Cancer de Montpellier, Montpellier; 9Department of Medical Oncology, Centre Georges François Leclerc, Dijon; 10 Department of Medical Oncology, Centre Henri Becquerel, Rouen; 11Department of Medical Oncology, Institut Claudius Regaud, IUCT-Oncopole, Toulouse; 12 Department of Medical Oncology, Centre Antoine Lacassagne, Nice; 13Department of Medical Oncology, Institut de Cancérologie de Lorraine, Vandoeuvre-lès-Nancy; 14 Department of Medical Oncology, Centre Eugène Marquis, Rennes; 15Department of Medical Oncology, Institut Paoli-Calmettes, Marseille; 16Department of Medical Oncology, Centre Jean Perrin, Clermont Ferrand; 17Department of Medical Oncology, Centre Paul Strauss, Strasbourg; 18Department of Medical Oncology, Institut de Cancérologie Jean-Godinot, Reims; 19Department of Medical Oncology, Centre Oscar Lambret, Lille; 20Capionis, Paris, France
Received 19 April 2016; revised 15 June 2016; accepted 21 June 2016
Background: Bevacizumab combined with paclitaxel as first-line chemotherapy for patients with HER2-negative metastatic breast cancer (MBC) has led to mixed results in randomized trials, with an improvement in progression-free survival (PFS) but no statistically significant overall survival (OS) benefit. Real-life data could help in assessing the value of this combination. Patients and methods: This study aimed to describe the outcome following first-line paclitaxel with or without bevacizumab in the French Epidemiological Strategy and Medical Economics (ESME) database of MBC patients, established in 2014 by Unicancer. The primary and secondary end points were OS and PFS, respectively. Results: From 2008 to 2013, 14 014 MBC patient files were identified, including 10 605 patients with a HER2-negative status. Of these, 3426 received paclitaxel and bevacizumab (2127) or paclitaxel (1299) as first-line chemotherapy. OS adjusted for major prognostic factors was significantly longer in the paclitaxel and bevacizumab group compared with paclitaxel [hazard ratio (HR) 0.672, 95% confidence interval (CI) 0.601–0.752; median survival time 27.7 versus 19.8 months]. Results were consistent in all supportive analyses (using a propensity score for adjustment and as a matching factor for
*Correspondence to: Dr Suzette Delaloge, Department of Cancer Medicine, Institut Gustave Roussy, 114 rue Edouard Vaillant, 94805 Villejuif, France. Tel: +33-1-42-11-42-93; Fax: +33-1-42-11-52-74; E-mail:
[email protected]
© The Author 2016. Published by Oxford University Press on behalf of the European Society for Medical Oncology. All rights reserved. For permissions, please email:
[email protected].
Downloaded from http://annonc.oxfordjournals.org/ at EPFL Lausanne on November 14, 2016
Paclitaxel plus bevacizumab or paclitaxel as first-line treatment for HER2-negative metastatic breast cancer in a multicenter national observational study
original articles
Annals of Oncology
nested case–control analyses) and sensitivity analyses. Similar results were observed for the adjusted PFS, favoring the combination (HR 0.739, 95% CI 0.672–0.813; 8.1 versus 6.4 months). Conclusions: In this large-scale, real-life setting, patients with HER2-negative MBC who received paclitaxel plus bevacizumab as first-line chemotherapy had a significantly better OS and PFS than those receiving paclitaxel. Despite robust methodology, real-life data are exposed to important potential biases, and therefore, results need to be treated with caution. Our data cannot therefore support extension of current use of bevacizumab in MBC. Key words: metastatic, breast cancer, paclitaxel, bevacizumab, real-life
introduction
methods study design We conducted a non-interventional, retrospective, comparative study to describe the outcome of predefined MBC patients selected in the ESME MBC database. This database gathers individual data from all patients, male or
| Delaloge et al.
evaluation criteria The primary end point was OS defined as time (months) between index date (start of first-line chemotherapy) and date of death (any cause). The main secondary end point was PFS defined as time between index date and date of first disease progression or death. Disease progression was defined as appearance of new metastatic site, progression of existing metastasis, local or locoregional recurrence of the primary tumor, discontinuation of chemotherapy and/or targeted therapy due to metastatic progression ( judged by the reference physician), or death from any cause.
statistical analysis Descriptive statistics [including mean, standard deviation (SD)] were used to summarize patients’ initial characteristics at diagnosis of metastatic disease. These characteristics were compared between the two groups using Pearson’s χ 2 test or Student t-test, when appropriate; a P value of <0.05 was considered statistically significant. Both OS and PFS were estimated using the Kaplan– Meier method. Survival curves with associated log-rank tests were generated. Censored data (duration of follow-up and patient status at last contact) were descriptively summarized for the two groups. The reverse Kaplan–Meier method was used to estimate the median follow-up durations [10]. A minimum set of forced variables (prognostic factors) and unforced variables were selected by univariate analysis. The primary analysis was based on multivariate analysis using a Cox model adjusted and stratified for prognostic factors of survival and potential cofounders. Prognostic factors for which the proportional hazards assumption was violated (i.e. significant interaction of covariate with time) were introduced as stratification factors, and factors for which the proportional hazards assumption was verified were included as covariates. Adjusted survival curves were generated using the Breslow Estimator [11]. Power calculations based on the observed sample size for the primary analysis were carried out for OS using the hazard ratio (HR) for death observed between the two groups. PFS was analyzed in a similar way to OS. HRs are presented on a descriptive basis with 95% confidence intervals (CI). No formal statistical test was provided to compare OS or PFS between groups, considering the retrospective, non-randomized design of this observational
Volume 27 | No. 9 | September 2016
Downloaded from http://annonc.oxfordjournals.org/ at EPFL Lausanne on November 14, 2016
Despite significant advances, metastatic breast cancer (MBC) remains a non-curable disease, with a median overall survival (OS) of 24–36 months for human epidermal growth factor receptor 2 (HER2)-negative MBC patients. Systemic treatment includes chemo-, endocrine, and targeted therapies, according to factors such as hormone receptors, endocrine resistance, visceral crisis, previous adjuvant therapy, and relapse-free interval [1]. However, very few first-line therapies have demonstrated an OS benefit in prospective clinical trials, except for HER2-directed therapies [2]. Bevacizumab is a monoclonal antibody that inhibits VEGFmediated tumor angiogenesis [3]. Based on results from the first pivotal phase III study (E2100) [4], the US Food and Drug Administration (FDA) granted accelerated approval to bevacizumab in combination with paclitaxel as first-line treatment for HER2negative MBC in 2008. The combination showed a 6-month improvement in progression-free survival (PFS) but had no impact on OS. Later, two trials [5, 6] investigating the combination of bevacizumab with other cytotoxic agents (and a global meta-analysis of this strategy [7]) failed to show an OS improvement and suggested a non-favorable risk–benefit profile. Although the FDA revoked approval of bevacizumab for MBC in 2011, the combination remained approved in several countries, albeit with some restrictions. Real-world evidence studies are complementary tools for describing drug effectiveness and safety over the entire product life-cycle, and are increasingly important in light of the recent expansion of products granted conditional approval. Health authorities are also interested in real-world data for long-term benefit–risk assessment as well as medico-economic evaluation, requiring improvements in research methods and data quality. The Epidemiological Strategy and Medical Economics (ESME) program is an academic initiative led by Unicancer, the French network of cancer centers, to centralize real-life data on MBC. The program involves 18 specialized cancer centers managing together over one-third of all breast cancer cases nationwide. Academic or private organizations may query the ESME MBC database for predefined questions, without direct access to the data or involvement in the work program. The primary objective of this first analysis of the database was to describe OS in patients with HER2-negative MBC treated with first-line paclitaxel chemotherapy, with or without bevacizumab.
female, ≥18 years, having started an anti-cancer treatment for a metastatic breast carcinoma in 1 of 18 cancer centers participating in the ESME program, from 1 January 2008 to 31 December 2013. For the present study, patient selection focused on those with HER2negative MBC who started first-line paclitaxel-based chemotherapy (at least one infusion delivered), with (combination group) or without ( paclitaxel group) bevacizumab. Data were collected until the cut-off date (1 October 2014), death (if this occurred before the cut-off date), or date of last contact (if lost to follow-up). In compliance with French regulations, the ESME MBC database was authorized by the French data protection authority (authorization no. 1704113) and managed by R&D Unicancer in accordance with current best practice guidelines [8, 9]. The study was approved by an independent ethics committee which waived the requirement for informed consent. A detailed description of the database is available in the supplementary material, available at Annals of Oncology online.
original articles
Annals of Oncology study. Supportive analyses (including adjustment using propensity scores and related nested case–control analyses) were conducted on the primary findings (see supplementary material, available at Annals of Oncology online, for details). Statistical analyses were carried out by Capionis (Paris, France) using SAS® software (version 9.4). For quality purposes, a second analysis was carried out by an independent academic statistician (Centre Léon Bérard, Lyon).
results patient characteristics and follow-up
treatment exposure and first-line chemotherapy The median (±SD) treatment duration with paclitaxel-based chemotherapy was slightly longer in the combination group [5
overall survival In the multivariate analysis, OS adjusted on major prognostic factors was longer for the combination group than for the paclitaxel group, with the median survival times of 27.7 months (95% CI 25.7–29.0) and 19.8 months (18.3–21.0), respectively (HR 0.672; 95% CI 0.601–0.752) (Figure 2). Results were consistent in all supportive analyses (Figure 3). When the propensity score was considered in the adjusted analyses, OS was statistically significantly different between the two groups (HR 0.700, 95% CI 0.635–0.771). Regarding the analyses of patients case-matched for propensity score at 1% and 2%, HRs were 0.759 (95% CI 0.677–0.851) and 0.761 (0.679–0.854), respectively. All sensitivity analyses yielded similar results as well as analyses conducted in subgroups (Table 2 and supplementary material, available at Annals of Oncology online). Causes of death are described in supplementary Table S1, available at Annals of Oncology online, as reported in patients’ files.
progression-free survival In the multivariate analysis, adjusted PFS was significantly longer for the combination group than the paclitaxel group (HR 0.739, 95% CI 0.672–0.813). Respective adjusted median values were 8.1 versus 6.4 months (supplementary Figure S1, available
14 014 selected in ESME MBC database 2413 HER2 positive 996 undetermined HER2 status 10 605 HER2 negative 1747 did not receive chemotherapy 5062 received 1st line chemotherapy other than paclitaxel +/– bevacizumab 370 received 1st line chemotherapy after selection period (2008–2013) 3426 received paclitaxel as 1st line chemotherapy
1299 received paclitaxel
2127 received paclitaxel + bevacizumab
Figure 1. Study flow chart.
Volume 27 | No. 9 | September 2016
doi:10.1093/annonc/mdw260 |
Downloaded from http://annonc.oxfordjournals.org/ at EPFL Lausanne on November 14, 2016
Of 14 014 patient files identified in the ESME MBC database, 10 605 patients had HER2-negative breast cancer at the time of first MBC diagnosis. Of these, 3426 started a first-line paclitaxelbased chemotherapy: 2127 (62.1%) in the combination group ( paclitaxel with bevacizumab) and 1299 (37.9%) in the paclitaxel group ( paclitaxel without bevacizumab) (Figure 1). Patients’ characteristics at baseline (time of diagnosis of metastatic status) are shown in Table 1. Compared with the paclitaxel group, patients in the combination group were younger (median age, 55 versus 63 years; P < 0.001), had a lower prevalence of central nervous system metastases (5.3% versus 8.2%; P < 0.001) and lung/pleura metastases (36.8% versus 40.6%; P = 0.025), but had more liver metastases (40.9% versus 35.9%; P = 0.004). They also had fewer metastatic sites (≥3: 28.9% versus 33.6%; P = 0.004) and were less often diagnosed with breast cancer and upfront concomitant metastasis (16.3% versus 24.3%; P < 0.001). The median follow-up duration was longer in the combination group (33.8 months, 95% CI 31.5–35.4) than in the paclitaxel group (27.2 months, 25.8–29.7). The proportion of patients categorized as ‘lost to follow-up’ in the combination and paclitaxel groups was 9.4% and 10.2%, respectively (no data available within 9 months before cut-off date) (supplementary Table S1, available at Annals of Oncology online).
months (3.45–6.04); n = 2081] than in the paclitaxel group [3.7 months (1.87–5.35); n = 1231]. The median duration of bevacizumab treatment was 5.32 months (3.25–7.92). During first-line treatment, patients were exposed to other therapies, including endocrine therapy, radiotherapy, and chemotherapeutic or targeted agents other than bevacizumab or paclitaxel (Table 1), usually given as a maintenance strategy and more often in the paclitaxel group: capecitabine (8.7% versus 8.9%), cyclophosphamide (2.5% versus 9.1%), anthracyclines (1.9% versus 8.1%), docetaxel (1.6% versus 3.5%), and others (2.2% versus 6.2%). Chemotherapy was the first treatment received for MBC in 86.5% of patients.
original articles
Annals of Oncology
Table 1. Patient characteristics at diagnosis of metastatic disease, according to study group Study population (n = 3426)
at Annals of Oncology online). Results were consistent in all
| Delaloge et al.
Paclitaxel group (n = 1299)
P valuea
54.2 (11.2) 55.0 (46–63)
62.7 (12.5) 63.0 (54–73)
<0.001
17 (0.8%) 2110 (99.2%)
18 (1.4%) 1281 (98.6%)
0.103
626 (29.4%) 1388 (65.3%) 113 (5.3%)
375 (28.9%) 817 (62.9%) 107 (8.2%)
0.004
113 (5.3%) 1237 (58.2%) 783 (36.8%) 687 (32.3%) 869 (40.9%) 433 (20.4%)
107 (8.2%) 763 (58.7%) 528 (40.6%) 446 (34.3%) 466 (35.9%) 374 (28.8%)
<0.001 0.738 0.025 0.220 0.004 <0.001
2.0 (1.1) 2.0 (1–3) 1–8 1513 (71.1%) 614 (28.9%)
2.2 (1.3) 2.0 (1–3) 1–8 863 (66.4%) 436 (33.6%)
0.005
347 (16.3%) 1780 (83.7%)
316 (24.3%) 983 (75.7%)
<0.001
61.2 (65.7) 38.6 (17–86) 347 (16.3%) 378 (17.8%) 1402 (65.9%)
67.6 (79.4) 40.9 (9–97) 316 (24.3%) 152 (11.7%) 831 (64.0%)
0.749
665 (31.4%) 1454 (68.6%) 8
377 (29.1%) 917 (70.9%) 5
0.166
989 (50.5%) 971 (49.5%) 167 270 (12.7%)
525 (44.2%) 662 (55.8%) 112 103 (7.9%)
<0.001
260 (12.2%) 322 (15.1%) 317 (14.9%) 418 (19.7%) 414 (19.5%) 396 (18.6%)
135 (10.4%) 153 (11.8%) 178 (13.7%) 238 (18.3%) 281 (21.6%) 314 (24.2%)
<0.001
595 (28.0%) 463 (21.8%) 364 (17.1%) 542 (25.5%) 163 (7.7%)
361 (27.8%) 379 (29.2%) 193 (14.9%) 267 (20.6%) 99 (7.6%)
<0.001
0.004
<0.001
<0.001
Continued
Volume 27 | No. 9 | September 2016
Downloaded from http://annonc.oxfordjournals.org/ at EPFL Lausanne on November 14, 2016
Age (years) Mean (SD) 57.4 (12.4) Median (Q1–Q3) 58.0 (48–66) Sex Male 35 (1.0%) Female 3391 (99.0%) Type of metastases Non-visceral 1001 (29.2%) Visceral/non-cerebral 2205 (64.4%) Cerebral 220 (6.4%) Metastatic sites CNS/CSF 220 (6.4%) Bone 2000 (58.4%) Lung/pleura 1311 (38.3%) Metastatic nodes 1133 (33.1%) Liver 1335 (39.0%) Other 807 (23.6%) Number of metastatic sites Mean (SD) 2.1 (1.2) Median (Q1–Q3) 2.0 (1–3) Min–max 1–8 <3 2376 (69.4%) ≥3 1050 (30.6%) Initial diagnosis with upfront concomitant metastases Yes 663 (19.4%) No 2763 (80.6%) Time (months) between initial diagnosis and metastatic diagnosis Mean (SD) 63.6 (71.3) Median (Q1–Q3) 38.9 (15–90) <6 663 (19.4%) (6–24) 530 (15.5%) ≥24 2233 (65.2%) ER status Negative 1042 (30.5%) Positive 2371 (69.5%) Not determined 13 Histological grade III Yes 1514 (48.1%) No 1633 (51.9%) Missing 279 Inclusion in a clinical trial during first-line 373 (10.9%) treatment Period of care (year) 2008 395 (11.5%) 2009 475 (13.9%) 2010 495 (14.4%) 2011 656 (19.1%) 2012 695 (20.3%) 2013 710 (20.7%) Region of care Paris and Ile de France 956 (27.9%) North West 842 (24.6%) North East 557 (16.3%) South East 809 (23.6%) South West 262 (7.6%)
Combination group (n = 2127)
original articles
Annals of Oncology Table 1. Continued Combination group (n = 2127)
Paclitaxel group (n = 1299)
P valuea
561 (16.4%) 1093 (31.9%) 1772 (51.7%)
345 (16.2%) 718 (33.8%) 1064 (50.0%)
216 (16.6%) 375 (28.9%) 708 (54.5%)
0.010
2963 (86.5%) 1117 (32.6%) 726 (21.2%)
1878 (88.3%) 650 (30.6%) 409 (19.2%)
1085 (83.5%) 467 (36.0%) 317 (24.4%)
<0.001 0.001 <0.001
3.6 (7.1) 1.2 (1–2) 2397 (70.0%) 564 (16.5%) 465 (13.6%)
3.1 (6.2) 1.1 (1–2) 1563 (73.5%) 315 (14.8%) 249 (11.7%)
4.3 (8.2) 1.3 (1–3) 834 (64.2%) 249 (19.2%) 216 (16.6%)
<0.001 <0.001
a
P value is for the combination group versus the paclitaxel group. Number of selected patients treated in the center during the period 2008–2013. CNS, central nervous system; CSF, cerebrospinal fluid; ER, estrogen receptor; Q1–Q3, 25th and 75th percentile; SD, standard deviation. b
0.5 0.55 0.6 0.65 0.7 0.75 0.8 0.85 0.9 0.95 1
1.0 0.9
Adjusted overall survival
0.8 0.7
Univariate analysis 0.622 0.681 0.746
0.6 0.5 0.4 0.3 Primary analysis (multivariate adjusted model)
0.2
0.601 0.672
0.752
0.1 0.0 0
6
12
18
24
30
36
42
48
54
60
Duration of follow-up (months) since index date Combination group
Paclitaxel group
Adjusted median survival times: 27.7 months (combination); 19.8 months (paclitaxel)
Analysis adjusted for the propensity score
0.635
0.7
0.771
Adjustment for: group, time between metastatic diagnosis and index date, period of care, SBR grade III, age, triple-negative breast cancer status, type of metastases (visceral versus non visceral) and number of sites, time between initial diagnosis and metastatic diagnosis, initial management (adjuvant chemotherapy and/or adjuvant endocrine therapy)
Figure 2. Adjusted overall survival analyses.
Case-matched for propensity score at 1% (N = 1019)
0.677
0.759
0.851
Case-matched for propensity score at 2% (N = 1027)
0.679
0.761
0.854
supportive analyses (supplementary Figure S2, available at Annals of Oncology online) and sensitivity analyses (Table 2).
discussion The ESME program represents the first large-scale European real-life initiative on MBC, involving more than 14 000 patients with a mean follow-up time of 40.0 months (95% CI 39.2–40.7).
Volume 27 | No. 9 | September 2016
Figure 3. Analysis of overall survival: summary of hazard ratios.
doi:10.1093/annonc/mdw260 |
Downloaded from http://annonc.oxfordjournals.org/ at EPFL Lausanne on November 14, 2016
Activity level of cancer centerb <500 (500–800) >800 Therapy at/or following MBC diagnosis Chemotherapy Endocrine therapy Radiotherapy Time between selection and index dates (months) Mean (SD) Median (Q1–Q3) <2 (2–6) >6
Study population (n = 3426)
original articles
Annals of Oncology
Table 2. Hazard ratios (95% confidence intervals) for overall survival and progression-free survival in patient subgroups
Patients not enrolled in a clinical trial during period of care Triple-negative phenotypea ER-positive/HER2-negative phenotype Exclusive groupsb Restricted groupsc Patients with management of metastatic disease initiated between 2008 and 2010 Patients with management of metastatic disease initiated between 2011 and 2013 Patients not deceased within 3 months after MBC diagnosis Patients not deceased within 1 month after index date Histological grade III (yes, no, missing)d
n
OS HR (95% CI)
PFS HR (95% CI)
2817 993 2424 2806 2281 1365 2061 3266 3321 3426
0.594 (0.530; 0.666) 0.671 (0.563; 0.800) 0.622 (0.542; 0.712) 0.556 (0.496; 0.623) 0.589 (0.530; 0.654) 0.663 (0.573; 0.767) 0.562 (0.484; 0.654) 0.737 (0.656; 0.828) 0.706 (0.630; 0.792) 0.662 (0.595; 0.737)
0.727 (0.652; 0.811) 0.685 (0.564; 0.832) 0.753 (0.675; 0.841) 0.640 (0.576; 0.711) 0.705 (0.642; 0.774) 0.750 (0.636; 0.884) 0.703 (0.616; 0.801) 0.770 (0.696; 0.852) 0.752 (0.680; 0.831) 0.766 (0.699; 0.839)
a
ER < 10%, PR < 10%, HER2 score 0 or 1 or non-amplified. Defined as all patients having received at least one administration of bevacizumab AND paclitaxel (combination group) or paclitaxel only (paclitaxel group) without any other chemotherapy during the period from 12 weeks before or after MBC diagnosis up to 6 months after the index date (and before the date of first progression). c Defined as all patients having received at least one administration of bevacizumab AND paclitaxel (combination group) or paclitaxel without bevacizumab (paclitaxel group) during the period from 12 weeks before the MBC diagnosis up to 2 months after the index date (and before the date of first progression). d Data were missing for 167 patients in the combination group and 112 patients in the paclitaxel group. All analyses were based on multivariate analysis using a Cox model adjusted and stratified for prognostic factors of survival, without interaction of treatment with time. Index date was defined as the start of first-line treatment, and selection date was defined as the date of diagnosis of the metastatic disease. Cut-off date was the end of the follow-up period (01 October 2014). ER, estrogen receptor; CI, confidence interval; HER2, human epidermal growth factor receptor 2; HR, hormone receptor; OS, overall survival; PFS, progression-free survival; MBC, metastatic breast cancer; PR, progesterone receptor; SD, standard deviation. b
| Delaloge et al.
clinical trials, both longer median time to event and smaller relative size often decrease power for OS analysis compared with PFS, especially in the case of long post-progression survival [14]. In addition, well-known limitations lead to large uncertainty about OS as an indicator of treatment efficacy. A major drawback stems from a weakening of the potential benefits allocated by randomization through the use of maintenance therapies or subsequent lines of treatment following disease progression. Indeed, recent data suggest that maintenance treatment has a positive effect on OS, hindering conclusions [7, 15, 16]. Importantly, some recent oncology guidelines have been developed based on surrogates for survival in the absence of OS benefit [17]. As health authorities require real-life data to reassess therapeutic alternatives in pragmatic settings, carefully designed observational studies are important for bridging the gap between clinical research and clinical practice. Carefully built observational studies from real-world databases should be a useful complement of randomized trials in a near future, through the control of expected therapeutic activities of therapeutic agents, the identification of major medical needs and gaps, but also through improvement of our understanding of the choice of treatment strategies. As well, observational studies might help assessing real-world outcome of recently released agents, although this was not the case for safety with the present database. The present work has indeed several major strengths. First, rigorous standard screening procedures across all 18 cancer centers and statistical methodology were applied to minimize selection bias. In addition, propensity scores were used to adjust for baseline differences between groups and to reduce the
Volume 27 | No. 9 | September 2016
Downloaded from http://annonc.oxfordjournals.org/ at EPFL Lausanne on November 14, 2016
In the present analysis, patients who received the combination (paclitaxel plus bevacizumab) had a significantly higher OS compared with those who received paclitaxel (HR 0.672; median, 27.7 versus 19.8 months). PFS was also significantly higher with the combination group, with a more modest beneficial effect (HR 0.739; 8.1 versus 6.4 months). The results showed consistency across all supportive and sensitivity analyses. Our results on OS in the combination group are in line with those from phase III trials that investigated bevacizumab in combination with chemotherapy as first-line treatment for HER2negative MBC (E2100, AVADO, and RIBBON-1 [4–6]). However, while all three studies showed a benefit on PFS, none demonstrated an OS benefit, nor did a meta-analysis of the data [7]. E2100 is the only study to have investigated bevacizumab in combination with paclitaxel. Having enrolled 722 patients, it showed a higher 1-year OS with the combination compared with paclitaxel alone (81.2% and 73.4%, respectively; P = 0.01) but no significant impact on OS with longer follow-up (HR = 0.88, P = 0.16; median, 26.7 versus 25.2 months). In the present analysis, while OS was in line with that reported in the combination group in E2100, the paclitaxel group, however, had a lower OS (19.8 months) than that obtained in E2100. Regarding PFS, our results in the combination group are in line with those reported in the above-mentioned trials (stratified HR = 0.60, P < 0.001 in E2100 [4]) and in the MERiDiAN trial [12], and with the capecitabine plus bevacizumab combination [13]. Observational studies, with their larger sample size and longer follow-up, can provide complementary data to help improve the assessment of treatment value obtained within the strict framework of randomized, controlled trials. Indeed, in
original articles
Annals of Oncology
conclusions We observed that patients with HER2-negative MBC receiving first-line chemotherapy paclitaxel combined with bevacizumab in a real-life setting have an OS identical to that observed in clinical trials, while patients receiving paclitaxel have a significantly lower OS. Whether the observed difference is linked to the treatment itself, to prescription bias, or to a mixed effect of both cannot be ascertained. Based on the present study, data cannot support extension of current use of bevacizumab in MBC. Further analysis will be carried out to assess the impact of subsequent treatments on OS. The MBC database will be updated on a yearly basis.
Volume 27 | No. 9 | September 2016
acknowledgements We thank the 18 French Comprehensive Cancer Centers for providing the data and each ESME contact for coordinating the project at the local level. We pay tribute to the late Professor Josy Reiffers, former president of UNICANCER who was the first to support the launch of the ESME program. We acknowledge Dr Sarah Hopwood (Scinopsis, France) for medical writing assistance with this publication, paid by R&D UNICANCER. Finally, we thank the ESME Scientific and Ethical Committee members for their ongoing support.
ESME central coordinating staff Head of research and development: CCa. Program director: MR. Data managers: Irwin Piot and Olivier Payen. Project managers: CCo, TG, and Gaëtane Simon. Project assistant: Esméralda Perreira. Software designer: Alexandre Vanni.
ESME local coordinating staff Patrick Arveux, TB, Jean-Pierre Bleuse, Delphine Berchery, Mathias Breton, Didier Cauchois, Emmanuel Chamorey, Valérie Dejean, Véronique Diéras, Anne-Valérie Guizard, Anne Jaffré, Lilian Laborde, Agnès Loeb, Muriel Mons, Damien Parent, Geneviève Perrocheau, MAM-R, LU, and Michel Velten.
funding This work was supported by R&D UNICANCER. The ESME MBC database is supported by an industrial consortium (Roche, Pierre Fabre, and Pfizer). No grant numbers apply. Data collection, analyses and publications are totally independent of the industrial funding partners of R&D UNICANCERESME.
disclosure SD, CCo, TG, MR, and CCa all report grants from Pfizer, Pierre Fabre, and Roche (global grant to ESME database), during the conduct of the study. SD also reports personal fees and nonfinancial support from Roche/Genentech, grants, personal fees, and non-financial support from Pfizer, personal fees and nonfinancial support from Puma, grants, personal fees, and nonfinancial support from Astra Zeneca, grants, personal fees, and non-financial support from Novartis, personal fees and nonfinancial support from Amgen, outside the submitted work. DP reports personal fees from Roche, non-financial support from Roche, outside the submitted work. TB reports grants, personal fees, and non-financial support from Roche, grants, personal fees, and non-financial support from Novartis, grants and personal fees from AstraZeneca, outside the submitted work. VD reports personal fees from Roche/Genentech, during the conduct of the study. MC reports personal fees from Roche, during the conduct of the study; grants and personal fees from Novartis, personal fees from Astra Zeneca, personal fees from Menarini, outside the submitted work. WJ reports personal fees and non-financial support from Roche, outside the submitted work. CV reports personal fees from Roche, personal fees from Novartis, personal fees from GSK, non-financial support from Eisai, outside the
doi:10.1093/annonc/mdw260 |
Downloaded from http://annonc.oxfordjournals.org/ at EPFL Lausanne on November 14, 2016
impact of treatment selection bias (see supplementary material, available at Annals of Oncology online). Secondly, the analysis was based on a large number of well-documented cases, with a high level of quality control and exhaustive patient selection across participating centers. An important limitation is that it captures data from patients treated in academic comprehensive cancer centers, making it potentially hazardous to extrapolate results to other institutions. Other limitations are inherent to the retrospective and observational nature of the study design. Particularly, in a non-randomized setting, biases in treatment assignment linked to patient characteristics or underlying disease may be a major driver of different outcomes. In this study, patients’ baseline characteristics were carefully scrutinized for factors that might explain treatment assignment and poorer outcome observed in the paclitaxel group. Significant differences were observed, notably a younger age and shorter time between metastatic diagnosis and start of treatment in the combination group. Available data did not include performance status, a key factor for OS [18]; it was not routinely reported by physicians. Thus, the observed between-group OS difference might just reflect a lower performance status and higher level of co-morbidity in patients assigned to paclitaxel, deliberately chosen in order to minimize the risk of side-effects of the combination. However, this is not supported by the sensitivity analyses (see supplementary material, available at Annals of Oncology online). Of note, observational databases suggest that variability in real-world outcomes in cancer patients not only may be influenced by the selection of medications, but may also reflect the timely intersection of disease status, patient preference, and provider interpretation. Another limitation is that differences in patient monitoring and non-standardized retrospective data collection for disease progression may result in information bias, although a robust data management plan and quality control program were designed to limit this. As for randomized studies, we cannot clearly evaluate the impact of second-line treatments and there might be some imbalance between groups regarding them, although interpretation is very difficult (supplementary Table S2, available at Annals of Oncology online). Finally, data on comorbidities and safety were not systematically available and documented in the EMR. Nevertheless, we expected them to be potentially highly biased by centers’ and physicians’ habits. This precludes potential analyses such as those on competing mortality causes, risk/benefits, and, to some extent, propensity.
original articles submitted work. AG reports grants, personal fees, and non-financial support from Roche, during the conduct of the study. TP reports non-financial support and other support from Roche, outside the submitted work. All other authors declare no competing interests.
references
| Delaloge et al.
10. Schemper M, Smith TL. A note on quantifying follow-up in studies of failure time. Control Clin Trials 1996; 17: 343–346. 11. Cox DR. Regression models and life tables (with discussion). J Roy Stat Soc Ser B 1972; 74: 187–220. 12. Miles D, Cameron D, Bondarenko I et al. First results from the double-blind placebo (PL)-controlled randomised phase III MERiDiAN trial prospectively evaluating plasma ( p)VEGF-A in patients ( pts) receiving first-line paclitaxel (PAC) +/- bevacizumab (BV) for HER2-negative metastatic breast cancer (mBC). In The European Cancer Congress, 2015. Abstract 1866. 2015. 13. Zielinski CC, Lang I, Inbar M et al. Final results for overall survival (OS), the primary endpoint of the CECOG TURANDOT prospective randomised trial evaluating bevacizumab-paclitaxel (BEV-PAC) vs BEV-capecitabine (CAP) for HER2-negative locally recurrent/metastatic breast cancer (LR/mBC). In The European Cancer Congress, 2015. Abstract 1800. 2015. 14. Broglio KR, Berry DA. Detecting an overall survival benefit that is derived from progression-free survival. J Natl Cancer Inst 2009; 101: 1642–1649. 15. Gligorov J, Doval D, Bines J et al. Maintenance capecitabine and bevacizumab versus bevacizumab alone after initial first-line bevacizumab and docetaxel for patients with HER2-negative metastatic breast cancer (IMELDA): a randomised, open-label, phase 3 trial. Lancet Oncol 2014; 15: 1351–1360. 16. Park YH, Jung KH, Im SA et al. Phase III, multicenter, randomized trial of maintenance chemotherapy versus observation in patients with metastatic breast cancer after achieving disease control with six cycles of gemcitabine plus paclitaxel as first-line chemotherapy: KCSG-BR07-02. J Clin Oncol 2013; 31: 1732–1739. 17. Prasad V, Kim C, Burotto M, Vandross A. The strength of association between surrogate end points and survival in oncology: a systematic review of trial-level meta-analyses. JAMA Intern Med 2015; 175: 1389–1398. 18. Llombart-Cussac A, Pivot X, Biganzoli L et al. A prognostic factor index for overall survival in patients receiving first-line chemotherapy for HER2-negative advanced breast cancer: an analysis of the ATHENA trial. Breast 2014; 23: 656–662.
Volume 27 | No. 9 | September 2016
Downloaded from http://annonc.oxfordjournals.org/ at EPFL Lausanne on November 14, 2016
1. Cardoso F, Costa A, Norton L et al. ESO-ESMO 2nd international consensus guidelines for advanced breast cancer (ABC2). Ann Oncol 2014; 25: 1871–1888. 2. Swain SM, Baselga J, Kim SB et al. Pertuzumab, trastuzumab, and docetaxel in HER2-positive metastatic breast cancer. N Engl J Med 2015; 372: 724–734. 3. Ferrara N. Vascular endothelial growth factor as a target for anticancer therapy. Oncologist 2004; 9(Suppl 1): 2–10. 4. Miller K, Wang M, Gralow J et al. Paclitaxel plus bevacizumab versus paclitaxel alone for metastatic breast cancer. N Engl J Med 2007; 357: 2666–2676. 5. 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: 3239–3247. 6. Robert NJ, Dieras V, Glaspy J et al. RIBBON-1: randomized, double-blind, placebocontrolled, phase III trial of chemotherapy with or without bevacizumab for first-line treatment of human epidermal growth factor receptor 2-negative, locally recurrent or metastatic breast cancer. J Clin Oncol 2011; 29: 1252–1260. 7. Rossari JR, Metzger-Filho O, Paesmans M et al. Bevacizumab and breast cancer: a meta-analysis of first-line phase III studies and a critical reappraisal of available evidence. J Oncol 2012; 2012: 417673. 8. Guidelines for good pharmacoepidemiology practices (GPP). Pharmacoepidemiol Drug Saf 2008; 17: 200–208. 9. Association des Epidémiologistes de Langue Française (ADELF). Déontologie et bonnes pratiques en épidémiologie, recommendations révisées, 2007.
Annals of Oncology