Download letter to the editor

letter to the editor
Everolimus, when combined with
exemestane, adds toxicity with
minimal benefit for women with
breast cancer
Annals of Oncology
has greater anti-tumour activity than exemestane alone, it leads
to only a modest improvement in survival in an unselected
population and it adds substantial toxicity’.
I. F. Tannock1* & G. R. Pond2
1
The lead editorial in the April 2014 issue of Annals of Oncology
sets appropriate standards for acceptance of manuscripts: they
should be reproducible, innovative, global and balanced. Yet in
this issue are two articles that comment on adverse events associated with everolimus and exemestane in the BOLERO-2 trial
[1, 2]. That trial was innovative and global, but these reports are
not balanced.
The original report of BOLERO-2 showed a substantial difference in progression-free survival (PFS) in favour of everolimus
and exemestane compared with exemestane alone; the hazard
ratio (HR) was 0.43 (P < 0.001) [3]. That analysis was subject to
bias because of informative censoring, which was not balanced
between the arms: almost 40% of locally assessed PFS events
were deemed non-events upon central review. Further, 24%
versus 6% of women on experimental versus control arms
withdrew from treatment before disease progression because of
toxicity or other reasons. Most patients who withdraw from
treatment are not doing well and many will refuse to continue
with scheduled radiologic assessments, resulting in a bias which
favours the more toxic therapy [4]. Finally, an improvement in
PFS is not an effective measure of benefit to patients, nor has it
been shown to be a valid surrogate for overall survival or its
quality. Analysis of overall survival of the BOLERO-2 trial was
presented at the 2014 European Breast Cancer Conference with
the finding of a modest (HR = 0.89) and non-significant difference between the arms of the study [5].
The articles by Aapro et al. [1] and by Rugo et al. [2] evaluating toxicity and its management show that there was greater
toxicity in the combined everolimus and exemestane arm of the
BOLERO-2 trial, including stomatitis, pneumonitis, hyperglycaemia and fatigue. Some women on the experimental arm
might have had improved or maintained quality of life (QoL),
and a previous analysis suggested slower decline in QoL, but this
was also confounded by unbalanced informative censoring [6].
Therefore, one can only conclude that the combination of everolimus and exemestane has greater biological activity than exemestane alone, but unfortunately, adds substantial toxicity with
only a small improvement in survival.
Then there is the question of balanced reporting. The final
sentence of one article [1] states: ‘Overall everolimus is an effective treatment option, with manageable toxicity, among patients
with HR+ advanced breast …’. The second article [2] ends with
the statement: ‘Understanding the time course of AEs … is particularly important given the clinical benefit obtained from
adding everolimus to exemestane and the differential toxicities
associated with this targeted agent’. In our lexicon, therapies
that increase toxicity with minimal improvement in survival are
not ‘effective treatment options’ and they do not provide ‘clinical
benefit’.
We suggest an alternative concluding sentence to these articles: ‘Although the combination of everolimus and exemestane
 | letters to the editor
Princess Margaret Cancer Centre and University of Toronto, Toronto,
Canada
2
Ontario Clinical Oncology Group, McMaster University, Hamilton, Canada
(*E-mail: [email protected])
disclosure
The authors have declared no conflicts of interest.
references
1. Aapro M, Andre F, Blackwell K et al. Adverse event management in patients with
advanced cancer receiving oral everolimus: focus on breast cancer. Ann Oncol
2014; 25: 763–773.
2. Rugo HS, Pritchard KI, Gnant M et al. Incidence and time course of everolimusrelated adverse events in postmenopausal women with hormone receptor-positive
advanced breast cancer: insights from BOLERO-2. Ann Oncol 2014; 25: 808–815.
3. Baselga J, Campone M, Piccart M et al. Everolimus in postmenopausal hormonereceptor-positive advanced breast cancer. N Engl J Med 2012; 366: 520–529.
4. Carroll KJ. Analysis of progression-free survival in oncology trials: some common
statistical issues. Pharm Stat 2007; 6: 99–113.
5. Piccart M, Hortobagyi GN, Campone M et al. Everolimus plus exemestane for hormone
receptor-positive (HR+), human epidermal growth factor receptor-2-negative (HER2−)
advanced breast cancer (BC): overall survival results from BOLERO-2. In: Proceeding of
the 9th European Breast Cancer Conference, Brussels, Belgium, 2014. Abstract LBA1.
6. Burris HA, III, Lebrun F, Rugo HS et al. Health-related quality of life of patients with
advanced breast cancer treated with everolimus plus exemestane versus placebo
plus exemestane in the phase 3, randomized, controlled, BOLERO-2 trial. Cancer
2013; 119: 1908–1915.
doi: 10.1093/annonc/mdu371
Published online 1 August 2014
Reply to the letter to the editor
‘Everolimus, when combined
with exemestane, adds toxicity
with minimal benefit for women
with breast cancer’ Tannock
and Pond
Thank you for your comments [1] regarding our papers [2, 3].
We appreciate the opportunity to respond in detail to your letter
that in our opinion contains statements that do not reflect the
actual Bolero-2 data. Our papers are balanced and accurately
present available data. One paper reviews in detail the incidence
and time course of toxicity from the combination of everolimus
(EVE) and exemestane (EXE) [2], and the other gives practical
guidance on how to manage these side-effects [3]. We believe
these papers provide an objective assessment of both efficacy
and toxicity of this novel therapy, and also represent an
Volume 25 | No. 10 | October 2014
Annals of Oncology
important educational service to our colleagues who are using
the combination in clinical practice.
In reply to your specific comments:
‘That analysis was subject to bias because of informative censoring, which was not balanced between the arms’ in reference to the
interim analysis of progression-free survival (PFS) [4].
Informative censoring refers to censoring which is related to
the occurrence of progression-free survival (PFS) events, for
example, censoring due to start of new antineoplastic therapy. In
contrast, noninformative censoring is independent of PFS. In
the BOLERO-2 trial, the potential bias of informative censoring
was carefully assessed. Several preplanned sensitivity analyses of
PFS were conducted (e.g. by not censoring patients after start
of a new anticancer therapy) and have confirmed the robustness
of the primary PFS results.
At the interim analysis (data cutoff 11 February 2011), a total
of 365 patients (50.4% of all randomized) were censored in the
primary analysis of PFS by investigator assessment [283/485
(58.4%) in the EVE arm and 82/239 (34.3%) in the placebo
(PBO) arm]. Among the censored patients, 77.5% [283 total;
77.7% (n = 220) EVE + EXE and 76.8% (n = 63) PBO + EXE]
were censored because they were event free at the time of analysis cutoff and still at risk of having a PFS event. These patients
continued to have all protocol-scheduled assessments. Clearly,
this type of censoring is noninformative. Other reasons for censoring include withdrawal of consent, loss to follow-up, start of
new anticancer therapy, or event documented after ≥2 missing
tumor assessments. At the interim analysis, censoring for
reasons other than ‘ongoing without event’ between the two
treatment arms was well balanced [22.3% (63/283) treated with
EVE + EXE and 23.2% (19/82) treated with PBO + EXE]. Of the
censored patients, only 10.2% in the EVE arm and 11.0% in the
PBO arm were censored due to new cancer therapy added.
A supportive PFS analysis by central assessment was conducted in BOLERO-2. Despite some discrepancies between
central and local assessment of PFS events, central analysis supported the primary (local) analysis of PFS, and demonstrated a
larger benefit from treatment with EVE and EXE compared with
EXE alone. The consistency between interim and final PFS analyses [4] at 18 months of median follow-up further supports the
robustness of the data from BOLERO-2.
Further, 24% vs. 6% of women on experimental vs. control arms
withdrew from treatment prior to disease progression because of
toxicity or other reasons. Most patients who withdraw from treatment are not doing well and many will refuse to continue with
scheduled radiologic assessments, resulting in a bias which favors
the more toxic therapy
Censoring patients who discontinue treatment because of
adverse events before having a PFS event may be considered ‘informative’ and could bias the PFS estimate because PFS prognosis for patients who discontinue due to adverse events (AE) may
be different from those continuing in the study at the same time
point. However, in the BOLERO-2 PFS analysis, treatment
discontinuation for AEs was not a censoring reason. On the contrary, patients who discontinued all study treatments for AEs
continued to be followed up for PFS as per protocol. Per
Volume 25 | No. 10 | October 2014
letters to the editor
protocol, patients were assessed as continuing on study without
any change in disease assessment schedule until disease progression/death/loss to follow-up/withdrawal of consent. In addition,
patients who discontinued one study treatment because of AEs
but continued to receive the second remained on-study and had
all scheduled assessments per protocol. Approximately 50% of
patients who discontinued due to AE had subsequent tumor assessment (see below).
At the interim analysis, 6.6% (32/485) of patients receiving
EVE + EXE and 2.5% (6/239) receiving PBO + EXE had permanently discontinued study treatment due to adverse events.
Of these, 46.9% (15/32) in the EVE + EXE arm and 33.3% (2/6)
in the PBO + EXE arm had subsequent tumor assessments.
These post-treatment evaluations led to the identification of 11
additional PFS events (2 deaths and 9 progressions) in the
EVE + EXE arm (i.e. in 73.3% of the 15 patients being followed)
and 2 additional progressions in the PBO + EXE arm (i.e. in
100% of the 2 patients being followed). Of the patients who discontinued study treatment due to AE, 43.8% (14/32) in the
EVE + EXE arm and 16.7% (1/6) in the PBO + EXE arm were
later censored due to new cancer therapy added and 9.4% (3/32)
in the EVE + EXE arm were lost to follow-up. Per protocol,
patients stayed on study if they discontinued one of the two
study treatments while continuing the other. Twenty-four
percent of patients in the EVE + EXE combination arm discontinued EVE due to an adverse event (19%) or withdrawal of
consent (5%). The 19% of patients who discontinued EVE due
to adverse event could still continue to receive EXE alone and
remain on study with no change in their tumor assessments,
hence did not introducing bias in the analysis of PFS.
Dr Tannock indicates that PFS is not a relevant endpoint in
advanced BC, stating ‘Finally, an improvement in PFS is not an
effective measure of benefit to patients, nor has it been shown to
be a valid surrogate for overall survival or its quality’.
Although PFS may not always be a surrogate for overall survival (OS), it is a clinically relevant end point for determining
the effectiveness of a therapeutic agent. In addition, PFS is not confounded by post-treatment therapies, and remains a standard end
point in the majority of clinical trials assessing treatment efficacy
in the metastatic setting. In contrast, OS is influenced by poststudy
treatment in an uncontrolled setting. For first- and second-line
treatments where survival prognosis is typically several years,
the dilution effects of poststudy treatment on OS are dramatic
and greatly inflate the study size needed to demonstrate survival
benefit.
Indeed, survival benefits have not been demonstrated for
effective and approved hormone therapies given following initial
progression.
Based on the non-significant difference in OS between arms in
BOLERO-2, Dr Tannock concludes that ‘Although the combination
of everolimus and exemestane has greater anti-tumour activity than
exemestane alone, it leads to only a modest improvement in survival
in an unselected population and it adds substantial toxicity’.
Sample size for the BOLERO-2 trial was driven by the primary
end point, PFS, and not by OS. Powering this trial for the observed
difference in OS would have required a much larger trial, with the
doi:10.1093/annonc/mdu373 | 
letter to the editor
disadvantages of significant time, cost, and patient lives. The
median OS in BOLERO-2 showed a nonsignificant 4.4-month
prolongation with EVE + EXE compared with PBO + EXE
(P = 0.1426) [5]. This difference is consistent with the benefit
observed for the primary end point of PFS (4.6-month increase in
median investigator-assessed PFS). Indeed, the median OS seen in
BOLERO-2, despite the use of additional treatments after progression, is one of the longest seen in phase III trials of hormone receptor-positive advanced breast cancer. No existing phase III trial has
shown a statistically significant OS benefit in this setting following
prior treatment with a nonsteroidal aromatase inhibitor (AI).
The maintenance in QoL in BOLERO-2 provides additional
support for the improved clinical efficacy outcomes in the
EVE + EXE arm; most AEs were grades 1 or 2. Of the toxicities
cited in the letter, the incidence of grade 3/4 events were: stomatitis (8%/0), hyperglycemia (4%/<1%), fatigue (3%/<1%), pneumonitis (3%/0). We remain confident that the addition of EVE
to an AI is a valid option for some postmenopausal patients
with disease progression after a nonsteroidal AI, since it significantly prolongs PFS by a clinically meaningful median interval of
4.6 months. On-going translational research efforts should allow
us to better define in which patients to use this combination.
To conclude, we have presented a careful and balanced analysis of side-effects and recommendations on how to mitigate them.
There can be different medical perspectives about what is important, but the patient is the final judge. The decision to treat any
patient with any drug or combination must take into account the
possible benefits and the relevant toxicities associated, and must be
made on a case-by-case basis. These comments will appear later
this year in the statements of the Advanced Breast Cancer −2
(ABC-2) consensus meeting.
 | letters to the editor
Annals of Oncology
H. S. Rugo1* & M. Aapro2
1
University of California San Francisco Helen Diller Family Comprehensive
Cancer Center, San Francisco, USA;
2
Institut Multidisciplinaire D’Oncologie, Genolier, Switzerland
(*E-mail: [email protected]).
disclosure
Both authors have participated in studies of Novartis and acted
as consultants for Novartis.
references
1. Tannock I, Pond G. Everolimus, when combined with exemestane, adds toxicity with
minimal benefit for women with breast cancer. Ann Oncol 2014; 25: 2096.
2. Rugo HS, Pritchard KI, Gnant M et al. Incidence and time course of everolimusrelated adverse events in postmenopausal women with hormone receptorpositive advanced breast cancer: insights from BOLERO-2. Ann Oncol 2014; 25:
808–815.
3. Aapro M, Andre F, Blackwell K et al. Adverse event management in patients with
advanced cancer receiving oral everolimus: focus on breast cancer. Ann Oncol
2014; 25: 763–773.
4. Baselga J, Campone M, Piccart M et al. Everolimus in postmenopausal
hormone-receptor-positive advanced breast cancer. N Engl J Med 2012; 366:
520–529.
5. Piccart M, Hortobagyi G, Campone M et al. Everolimus plus exemestane for
hormone receptor-positive (HR+), human epidermal growth factor receptor-2negative (HER2-) advanced breast cancer (BC): overall survival results from
BOLERO-2. Oral Presentation Abstract #LBA1. European Breast Cancer Conference
(EBCC-9), 2014, Glasgow, Scotland.
doi: 10.1093/annonc/mdu373
Published online 1 August 2014
Volume 25 | No. 10 | October 2014