Académique Documents
Professionnel Documents
Culture Documents
O R I G I N A L
R E P O R T
Purpose
Bevacizumab provides a survival benefit in first- and second-line metastatic colorectal cancer
(mCRC). In a large, observational, bevacizumab treatment study (Bevacizumab Regimens:
Investigation of Treatment Effects and Safety [BRiTE]) in patients who had mCRC, a longer-thanexpected overall survival (OS) of 25.1 months was reported. The association between various preand post-treatment factors (including the use of bevacizumab beyond first progression [BBP]) and
survival was examined.
Patients and Methods
The 1,445 of 1,953 previously untreated patients with mCRC who were enrolled in BRiTE and who
experienced disease progression (PD) were classified into three groups: no post-PD treatment
(n 253), post-PD treatment without bevacizumab (no BBP; n 531), and BBP (n 642).
Relevant baseline and on-study variables, including BBP, were analyzed with a Cox model with
respect to their independent effect on survival beyond first PD.
Results
Median OS was 25.1 months (95% CI, 23.4 to 27.5 months), and median progression-free survival
was 10.0 months in the overall BRiTE population. Baseline and postbaseline factors were well
balanced between the BBP and no-BBP groups. Median OS rates were 12.6, 19.9, and 31.8
months in the no post-PD treatment, no-BBP, and BBP groups, respectively. In multivariate
analyses, compared with no BBP, BBP was strongly and independently associated with improved
survival (HR, 0.48; P .001). Hypertension that required medication was the only bevacizumabrelated safety event that occurred more frequently in the BBP group (24.6% v 19.2%).
Conclusion
These results from a large, prospective, observational study suggest that continued vascular
endothelial growth factor inhibition with bevacizumab beyond initial PD could play an important
role improving the overall success of therapy for patients who have mCRC.
J Clin Oncol 26:5326-5334. 2008 by American Society of Clinical Oncology
DOI: 10.1200/JCO.2008.16.3212
INTRODUCTION
In the last decade, the development of novel therapies that target critical biologic pathways has
greatly expanded treatment options for patients
with metastatic colorectal cancer (mCRC) and has
shown substantial improvement in survival and
progression-free survival (PFS). Bevacizumab (Avastin; Genentech Inc, South San Francisco, CA), a
recombinant, humanized monoclonal antibody
that binds to and neutralizes vascular endothelial
growth factor (VEGF),1 has improved overall survival (OS) and PFS in bevacizumab-nave patients
who have mCRC when it is added to first- and
second-line chemotherapy regimens.2-5 In preclinical experiments, sustained VEGF inhibition has
5326
been shown to achieve and maintain tumor regression.6,7 However, the clinical benefit and risks associated with continued bevacizumab use after initial
progressive disease are currently unknown.
After the US Food and Drug Administration
approval of bevacizumab in 2004, the Bevacizumab Regimens: Investigation of Treatment Effects and Safety (BRiTE) prospective observational
cohort study (OCS)8-11 was initiated to evaluate the
safety and effectiveness of bevacizumab in combination with chemotherapy in a large, communitybased patient population with previously untreated
mCRC. Earlier reports from BRiTE showed a median PFS of 10.0 months (95% CI, 9.7 to 10.4
months) and a median OS of 25.1 months (95%
CI, 23.4 to 27.5 months).8,9 This PFS estimate is
Overall Survival
Time to
progression
Begin
bevacizumab
treatment
Patients
In BRiTE, disease progression (PD) was determined by the investigator
via clinical and/or radiographic assessment. For this analysis, patients with an
investigator-assessed PD (n 1,445) were classified on the basis of the reported post-PD therapy they received: no post-PD treatment; post-PD treatment without bevacizumab (no BBP); and BBP. Post-PD treatment was
defined as any systemic anticancer therapy, including cytotoxic and/or biologic agents. Patients who received bevacizumab alone post-PD were excluded
from the analysis (n 19), because this group was too small to consider
separately. Patients who discontinued bevacizumab within 28 days after first
PD were considered not exposed to BBP (n 75); the effect of this classification was explored in a sensitivity analysis. Gaps in bevacizumab or chemotherapy administration of less than 28 days were considered continuous treatment.
Data were collected at baseline and every 3 months via a Web-based
electronic data collection system, as previously described.8-11
Clinical Outcomes
Measures of clinical outcomes were based on physician determination
and included time to progression from the date of initiation of first-line
treatment to first PD; OS from the date of initiation of first-line treatment to
death; and survival beyond first progression (SBP), which was measured only
for patients who had reported first PD from the date of first PD to death (Fig
1A). Patients who were not dead were censored at study termination (because
of loss to follow-up, patients decision, investigators decision, or sponsors
decision) or at the data cutoff date (January 21, 2007). The primary end point
for the BBP analysis was SBP, because classification into post-PD therapy
groups only became relevant after first PD and because the purpose of the
analysis was to explore reasons for the longer OS observed in BRiTE despite the
equivalent time to first PD.
Safety outcomes focused on previously described bevacizumab-related
adverse events (AEs)2 and were classified according to the National Cancer
Institute Common Terminology Criteria for Adverse Events (NCI-CTCAE),
version 3.0.
Statistical Analyses
OS and SBP were analyzed using the Kaplan-Meier method, which
included medians, 95% CIs, 1-year landmark estimates, and survival curves.
www.jco.org
Death
First progression
Assessable patients
(N = 1,953)
First progression*
(n = 1,445)
No postprogression
treatment
(n = 253)
Postprogression
treatment without
bevacizumab
(No BBP)
(n = 531)
Postprogression
treatment with
bevacizumab
(BBP)
(n = 642)
The Cox proportional hazards model was used to assess the independent
effects of pre- and post-treatment patient-related factors on SBP. Prespecified
measured variables, including prognostic variables previously shown to be
associated with survival in mCRC, were included in the multivariate analyses.3,13,14 In the primary analysis, these included age at first PD, ethnicity,
baseline Eastern Cooperative Oncology Group performance status (ECOG
PS), baseline serum albumin, baseline serum alkaline phosphatase, site of
primary tumor, first-line chemotherapy regimen, first-line time to progression
(TTP), best first-line response, exposure to three active chemotherapy agents
(ie, FU or capecitabine, irinotecan, oxaliplatin), exposure to epidermal growth
factor receptor inhibitors, and BBP. To assess the extent of possible immortal
time bias15 introduced by the fact that patients who had a delayed initiation
of post-PD therapy must have lived long enough to receive additional
therapy distant from first PD, an alternative definition of BBP was used,
wherein all patients who initiated post-PD therapy (with or without bevacizumab) more than 2 months after first PD were excluded (n 249). The
end point for this analysis was survival from 2 months after first PD
until death.
Sensitivity Analyses
Multiple sensitivity analyses were performed to compare the BBP and
no-BBP groups. These included the use of BBP as a time-dependent
covariate, conduction of a stratified Cox regression model by treatment
site, and conduction of a frailty Cox model that included treatment site as
a random effect.16 An analysis was also performed, whereby BBP patients
were matched in a 1:1 ratio with patients in the no-BBP group on their
propensity score (a patient-level summary measure of the difference between the groups),17 to create groups that were more alike with respect to
prognostic factors.
An additional analysis was performed to assess the site propensity for
BBP use on the basis of the supposition that variations in BBP use by study site
could affect OS and that such variations in treatment would conceivably be
2008 by American Society of Clinical Oncology
5327
Grothey et al
All Patients
(N 1,953)
No Post-PD Treatment
(n 253)
No BBP
(n 531)
BBP
(n 642)
63.6
63.8
65.6
66.6
63.5
64.3
61.0
61.8
81.4
11.6
7.0
80.6
13.0
6.4
81.4
11.5
7.1
80.7
12.5
6.9
42.9
42.2
7.0
7.9
38.7
44.3
10.3
6.7
39.9
46.9
4.7
8.5
50.2
36.9
5.6
7.3
3.7
1.0-5.8
3.6
1.1-5.8
3.7
1.1-5.0
3.7
1.0-5.0
147.41
7.4-1,547.0
161.3
7.4-962.0
141.7
10.7-1,247.0
148.5
14.0-1,212.0
79.4
20.4
0.2
81.0
19.0
78.2
21.8
79.6
20.1
0.3
55.9
14.3
9.7
6.8
4.8
8.5
50.6
12.3
13.8
9.9
4.3
9.1
57.1
14.7
8.3
7.5
4.0
8.4
55.0
16.8
9.5
4.8
5.9
8.0
14.4
30.3
33.1
16.3
5.9
38.4
27.8
8.3
24.1
37.5
30.0
0.0
7.1
0.0
10.5
33.3
37.1
19.0
0.0
61.2
50.9
12.3
35.8
30.5
21.3
0.0
62.6
41.0
5.4
0.03-32.9
4.4
0.03-24.4
5.3
0.03-26.5
6.2
0.03-30.5
6.9
0.03-34.7
4.4
0.03-24.4
5.3
0.03-28.1
12.8
1.05-34.7
95.7
1.7
0.7
1.9
90.7
3.6
2.3
3.4
9.4
0.7-34.3
8.0
0.5-30.1
8.7
0.3-30.3
8.9
0.8-30.5
5.7
0.03-33.6
4.8
0.4-23.7
6.0
0.3-27.1
6.3
0.5-30.5
10.4
0.03-35.2
4.8
0.4-23.7
12.2
1.1-34.2
15.6
1.5-35.2
Abbreviations: PD, disease progression; BBP, bevacizumab use beyond progression; ECOG PS, Eastern Cooperative Oncology Group performance status; FOLFOX,
oxaliplatin infusional fluorouracil/leucovorin; FOLFIRI, irinotecan infusional fluorouracil/leucovorin; IFL, inirotecan bolus fluorouracil/leucovorin; XELOX,
capecitabine/oxaliplatin; CR, complete response; PR, partial response; SD, stable disease; NA, not assessable; EGFR, epidermal growth factor receptor; Q2W, every
2 weeks; Q3W, every 3 weeks.
5328
Survival Outcomes
The median OS for the overall BRiTE group was 25.1 months
(95% CI, 23.4 to 27.5 months). Univariate analyses showed that OS
was longer in patients with better ECOG PS at baseline, higher albumin levels, lower alkaline phosphatase levels, primary tumor of the
rectum, exposure to three active chemotherapy agents, longer firstline TTP, complete or partial response to first-line therapy, and exposure to any post-PD therapy (Table 2). The median OS was longer for
the BBP group compared with the no-BBP group (Table 2). Similarly,
SBP was longer for the BBP group (Fig 2B). Patients who received no
www.jco.org
Patient Characteristics
Between February 2004 and June 2005, 1,953 patients were enrolled on BRiTE. The median follow-up time was 19.6 months. As of
January 21, 2007, 97 patients (5.0%) were lost to follow-up, and 74
patients (3.8%) had withdrawn from the study; 850 (44%) remained
alive. Of the 1,953 patients, 1,445 (74.0%) had experienced first PD,
and the remaining 508 (26.0%) were either alive without documented
PD (n 320) or had died without having a PD assessment (n 188;
Fig 1B). Of patients with documented PD, 253 (17.5%) received no
post-PD treatment, 531 (36.7%) received no BBP, and 642 (44.4%)
received BBP (Fig 1B). Baseline characteristics for the post-PD groups
are listed in Table 1 and appeared similar between the BBP and noBBP groups, except for some subtle differencesmost notably a
higher proportion patients with baseline ECOG PS 1 in the no-BBP
group. Multivariate analyses adjusted for these differences between
groups. The Kaplan-Meier curves for first-line TTP for the three
post-PD treatment groups are presented in Figure 2A, which shows
that the median first-line TTP for the no-BBP and BBP groups were
similar (8.7 months v 8.9 months, respectively).
100
80
60
40
20
Postprogression therapy
No treatment
No BBP
BBP
10
15
20
25
B
100
RESULTS
80
60
40
20
Postprogression therapy
No treatment
No BBP
BBP
10
15
20
25
Number of deaths
Percent
1-year survival rate, %
95% CI
Median survival beyond first
progression, months
95% CI
All patients
(N = 1,953)
No postprogression
treatment
(n = 253)
No BBP
(n = 531)
BBP
(n = 642)
932
47.7
168
66.4
305
57.8
260
40.5
74.7
72.7 to 76.7
52.5
46.2 to 58.8
77.3
73.7 to 80.9
87.7
85.2 to 90.3
12.0
3.6
9.5
19.2
11.1 to 13.3
2.7 to 4.3
8.4 to 11.2
16.8 to 20.7
15
20
25
C
100
based on local practice standards, rather than on individual patient characteristics. A sites propensity to use BBP was calculated as the number of patients at
a site who were treated with BBP divided by the number of patients at the site
who had first PD. Sites then were grouped into quartiles, and patients outcomes at sites in the highest BBP-use quartile were compared with those at sites
in the lowest BBP-use quartile.
80
60
40
20
Postprogression therapy
No treatment
No BBP
BBP
10
5329
Grothey et al
Table 2. Median OS of Pre- and Post-Treatment Variables and Multivariate Analysis of the Variables on SBP
Evaluation Time Point
OS
Variable
Age at first progression (per 10 years)
Ethnicity
White (n 1,152)
Black (n 178)
Other (n 96)
Baseline ECOG PS
0 (n 632)
1 (n 598)
2 (n 87)
Baseline albumin, per g/dL
Baseline alkaline phosphatase, per 100 U/L
Site of primary tumor, %
Colon (n 1,131)
Rectum (n 293)
Exposure to EGFR inhibitors
No (n 884)
Yes (n 542)
Exposure to all three active chemotherapy
agents
No (n 681)
Yes (n 745)
Time to first progression, per month
Best first-line response
CR (n 156)
PR (n 468)
SD (n 488)
PD (n 314)
Postfirst progression therapy
No BBP (n 531)
BBP (n 642)
No postprogression treatment (n 253)
Median
Multivariate Analysis
SBP
95% CI
Median
95% CI
HR
95% CI
1.04
0.97 to 1.10
.284
0.88
0.77
1.0 (reference)
0.70 to 1.11
0.57 to 1.05
.268
.103
1.30
1.60
0.73
1.18
1.0 (reference)
1.10 to 1.53
1.21 to 2.12
0.64 to 0.82
1.13 to 1.23
.002
.001
.001
.001
22.9
24.2
27.5
21.4 to 25.1
20.7 to 31.3
20.3 to NA
11.6
13.7
13.2
10.4 to 13.0
12.0 to 15.8
10.0 to NA
29.5
19.9
12.8
27.2 to NA
17.8 to 21.6
10.9 to 16.3
16.5
9.6
6.2
15.0 to 19.2
8.4 to 11.1
3.9 to 8.5
21.9
29.2
20.4 to 24.1
26.2 to NA
11.3
17.5
10.2 to 12.6
12.8 to 22.1
0.79
1.0 (reference)
0.65 to 0.97
.020
22.4
24.8
19.9 to 26.6
22.2 to 27.0
11.0
13.4
9.0 to 13.0
12.1 to 15.7
0.97
1.0 (reference)
0.81 to 1.16
.738
20.2
26.6
17.5 to 22.6
24.7 to 28.4
8.7
14.4
7.5 to 11.1
12.7 to 15.9
0.94
0.91
1.0 (reference)
0.78 to 1.14
0.89 to 0.93
.539
.001
NA
27.2
22.9
11.6
29.2 to NA
24.7 to 33.2
20.2 to 25.2
10.3 to 13.0
NA
14.0
12.0
7.5
17.8 to NA
11.6 to 15.9
10.0 to 13.4
6.3 to 9.0
1.64
1.64
1.33
1.0 (reference)
1.15 to 2.35
1.15 to 2.35
0.89 to 1.99
.007
.007
.158
19.9
31.8
12.6
18.0 to 22.0
27.9 to NA
10.6 to 15.7
9.5
19.2
3.6
8.4 to 11.2
16.8 to 20.7
2.7 to 4.3
0.49
2.05
1.0 (reference)
0.41 to 0.58
1.64 to 2.56
.001
.001
Sample sizes (n) relate to patients who had an investigator-assessed disease response and were included in the multivariate model.
Unadjusted median overall survival and survival beyond first progression for patients who had investigator-assessed disease progression.
Multivariate result for survival beyond first progression performed only on patients who had an investigator-assessed disease response.
Unknown performance status score (n 109) was included in the model as a separate level.
Includes cetuximab and/or other epidermal growth factor receptor monoclonal antibodies.
Active chemotherapy agents include FU (or capecitabine), irinotecan, and oxaliplatin.
Multivariate Analyses
In multivariate analyses, baseline ECOG PS, baseline albumin
levels, baseline alkaline phosphatase levels, site of primary tumor
(measured at baseline), first-line TTP, best first-line response, and
post-PD therapy were all significantly and independently related to
SBP (Table 2). After analysis was adjusted for other important prognostic factors, BBP maintained a statistically significant effect on SBP
compared with no BBP (HR, 0.49; 95% CI, 0.41 to 0.58; P .001). The
multivariate (ie, adjusted) hazard ratios for BBP versus no BBP were
consistent across groups of patients stratified by pre- and posttreatment variables (Fig 3).
Sensitivity Analyses
A series of sensitivity analyses were performed to address the
contribution of observed variability in the treatment patterns of bevacizumab and chemotherapy to the effect of BBP on survival. Different
JOURNAL OF CLINICAL ONCOLOGY
Subgroup
Hazard Ratio
95% CI
0.47
0.51
0.37 to 0.60
0.40 to 0.65
0.51
0.32
0.48
0.42 to 0.62
0.18 to 0.57
0.21 to 1.08
0.53
0.45
0.37
0.40 to 0.71
0.35 to 0.59
0.18 to 0.76
0.49
0.49
0.40 to 0.59
0.32 to 0.75
0.47
0.38
0.62
0.37 to 0.59
0.24 to 0.59
0.46 to 0.85
0.52
0.42
0.41 to 0.66
0.32 to 0.54
0.51
0.42
0.39 to 0.68
0.33 to 0.52
0.48
0.52
0.37 to 0.62
0.41 to 0.65
0.44
0.50
0.33 to 0.59
0.41 to 0.63
0.49
0.41 to 0.58
Fig 3. Forest Plot that illustrates multivariate (ie, adjusted) hazard ratios for survival beyond progression (SBP) for the
comparison of bevacizumab use beyond
progression (BBP) v no BBP, according to
subgroups of pre- and post-treatment factors. The relative size of each square represents the number of patients, and larger
squares indicate a greater number. Solid
bar represents 95% confidence interval.
Hazard ratios less than 1 (dotted line)
represent a longer median SBP of patients
in the BBP than of those in the no-BBP
group. PD, progressive disease; ECOG,
Eastern Cooperative Oncology Group;
FOLFOX, infusional fluorouracil, leucovorin, and oxaliplatin; FOLFIRI, fluorouracil,
leucovorin, and irinotecan; EGFR, epidermal growth factor receptor; CT, chemotherapy; CR, complete response; PR,
partial response; SD, stable disease.
0.5
5331
Grothey et al
Safety
The incidence of most bevacizumab-associated AEs described in
the BRiTE study, including arterial thromboembolic events, grades 3
to 4 bleeding events, and GI perforation, were similar in the no
post-PD treatment, no-BBP, and BBP groups (Table 3). There was a
higher incidence of new or worsening hypertension in the BBP group
compared with the no-BBP group or with the overall BRiTE population (19.2% v 24.6%), which is expected, given the longer exposure to
bevacizumab in the BBP group.
DISCUSSION
mens, may result in continuing clinical benefit. The results of the BBP
survival analyses from BRiTE provide support for this hypothesis.
The multivariate analysis was prompted by the observation of a
median OS of 25 months, which was longer than expected, given that
the patient population was less selective compared with RCTs (ie, it
included a higher proportion of patients who were elderly and a higher
proportion of patients who had poor prognostic characteristics), in a
setting in which the first-line PFS was similar to the previous reports
from multiple RCTs. As expected, historically recognized baseline
prognostic variables and post-treatment variables, such as first-line
TTP and response to first-line therapy, were each significantly associated with survival beyond PD.13 The novel finding from this analysis
was the significant and independent association between BBP and
survival. Although it is not possible to estimate the absolute length of
survival improvement afforded by BBP in this type of analysis, the HR
for SBP between the BBP and no-BBP groups (HR, 0.49 after adjustment for other variables) suggests an appreciable survival benefit.
Inherent with all OCSs are limitations that are based largely on
the fact that patients are not randomly assigned to the treatment
groups being compared. To the extent possible, multivariate analyses
were used to adjust for possible confounding factors, and the effect of
BBP on survival was independent from historically recognized preand post-treatment prognostic factors and from other post-treatment
variables of unknown significance (eg, use of epidermal growth factor
receptorinhibiting agents). Although ECOG PS, serum albumin, and
serum alkaline phosphatase were collected only at baseline and were
not updated at the time of PD, these variables were significantly
related to SBP in these analyses. Some residual confounding as a
result of the timing of, or errors in, measurement of prognostic
variables is possible. Multiple sensitivity analyses also were performed,
the results of which uniformly supported the association between BBP
and survival in this study. In addition, the analysis of site propensity to
address unmeasured site-related variables that may have influenced
the decision to use BBP confirmed a strong and significant association
between BBP use at the patient level and survival. Another potential
Table 3. Incidence of Bevacizumab-Targeted Adverse Events in BRiTE Patient Groups Based on Treatment Beyond First Progression
Treatment Group
Adverse Event
New or worsened HTN (which required medication), %
ATE, %
Total
Prefirst PD
Postfirst PD
Bleeding event (grades 3 to 4), %
Total
Prefirst PD
Postfirst PD
GI perforation, %
Total
Prefirst PD
Postfirst PD
All Patients
(N 1,953)
No Post-PD Treatment
(n 253)
No BBP
(n 531)
BBP
(n 642)
19.4
19.0
19.2
24.6
1.8
1.4
0.4
2.0
2.0
0.0
1.3
1.1
0.2
1.2
0.3
0.9
2.4
1.8
0.5
2.4
1.6
0.8
2.1
1.9
0.2
1.9
0.6
0.9
1.8
1.4
0.5
2.4
2.0
0.8
1.5
1.1
0.4
1.6
0.8
0.6
Abbreviations: PD, disease progression; BBP, bevacizumab use beyond progression; HTN, hypertension; ATE, arterial thromboembolic event.
Date of onset was not collected and, therefore, could not be classified as pre- or post-PD.
Percentages of pre- and post-PD adverse events may not sum to the total because of missing onset dates or because a patient was counted in both the pre- and
post-progression periods if 1 adverse event was experienced; however, patients were counted only once in the total percentage.
5332
AUTHOR CONTRIBUTIONS
Conception and design: Axel Grothey, Mary M. Sugrue, Wei Dong, Eric
Hedrick, Mark Kozloff
Provision of study materials or patients: Mary M. Sugrue, Mark Kozloff
Collection and assembly of data: Axel Grothey, Mary M. Sugrue, David
M. Purdie
Data analysis and interpretation: Axel Grothey, Mary M. Sugrue, David
M. Purdie, Wei Dong, Daniel Sargent, Eric Hedrick, Mark Kozloff
Manuscript writing: Axel Grothey, Mary M. Sugrue, David M. Purdie,
Wei Dong, Daniel Sargent, Eric Hedrick, Mark Kozloff
Final approval of manuscript: Axel Grothey, Mary M. Sugrue, David M.
Purdie, Wei Dong, Daniel Sargent, Eric Hedrick, Mark Kozloff
5333
Grothey et al
with availability of fluorouracil, irinotecan, and oxaliplatin regardless of whether doublet or single-agent
therapy is used first line. J Clin Oncol 23:9441-9442,
2005
20. Lowery M, Power D, Behbehani A, et al:
Hypertension is a significant adverse effect of bevacizumab treatment. J Clin Oncol 25:622s, 2007
(suppl; abstr 14134)
Acknowledgment
The Acknowledgment is included in the full-text version of this article, available online at www.jco.org. It is not included in the PDF
version (via Adobe Reader).
Appendix
The Appendix is included in the full-text version of this article, available online at www.jco.org. It is not included in the PDF version
(via Adobe Reader).
5334