Big Data, Better Treatment

download report

Transcript Big Data, Better Treatment

“Big Data, Better Treatment”: The work of the
Early Breast Cancer Trialists’ Collaborative Group
Rory Collins
BHF Professor of Medicine
& Epidemiology
Nuffield Department of Population Health
University of Oxford
Oxford, UK
What is the Early Breast Cancer Trialists’
Collaborative Group (EBCTCG)?
AIM: To assess the effects of various treatments for early
breast cancer on LONG-TERM survival more reliably than
can be done by any individual randomised controlled trial
METHOD: Collaboration between hundreds of trial groups
sharing individual patient data for “meta-analyses”, often
involving 10-20,000 randomized women in each question
BENEFITS: Really reliable identification of treatments that
can each MODERATELY improve breast cancer survival
and, in combination, improve survival SUBSTANTIALLY
Requirements for the RELIABLE assessment
of plausibly MODERATE treatment effects
SYSTEMATIC ERRORS (i.e. biases) and
RANDOM ERRORS (i.e. play of chance)
must BOTH be small compared with the difference
between any MODERATE effects and NO effect
This is achieved with large-scale evidence based
on meta-analyses of randomised controlled trials
(but not with non-randomised database studies)
Several MODERATE treatment effects
have almost HALVED breast cancer mortality
at ages 35-69 since 1990
EBCTCG helped reduce national breast cancer mortality rates
by demonstrating the moderate effects on long-term survival of
endocrine therapy (if ER+), chemotherapy, and radiotherapy
UK & USA: Breast cancer mortality 1950-2015 at ages 35-69
Probability of a
35-year-old dying
of breast cancer
aged 35-69
Risks halved, 1990-2015
UK: 2.4% down to 1.2%
US: 1.9% down to 1.1%
(& risks are still falling)
Several MODERATE treatment effects
have almost HALVED breast cancer mortality
at ages 35-69 since 1990
EBCTCG helped reduce national breast cancer mortality rates
by demonstrating the moderate effects on long-term survival of
endocrine therapy (if ER+), chemotherapy, and radiotherapy
It has also fostered respect for the value of conducting much
larger randomised trials than was customary and of combining
their results in systematic meta-analyses of all related trials
Demonstrating the need for “Big (randomised) Data”
to assess the effects of treatment properly
Selected EBCTCG NEJM/Lancet reports
History: 1985-2010
1985-88: Chemo & endocrine therapy additively affect 5-year survival
1990-92: Chemo & endocrine therapy additively affect 10-year survival
1995-98: Polychemotherapy benefits many different subgroups
1995-98: Tamoxifen benefits by treatment duration in many subgroups
2000-05: Chemo & endocrine therapy additively affect 15-year survival
2000-05: Radiotherapy after mastectomy in N+ affects 15-year survival
2005-10: Aromatase inhibitor (AI) vs tamoxifen delays recurrence
Past 5 years:15,000
2011-15citations: EBCTCG findings have
2011: Radiotherapy
breast conservation
improves
15-year survival
changedafter
guidelines
and worldwide
practice
2011: 5y of tamoxifen improves 15-year survival in all ER+ subgroups
~5 years of tamoxifen vs no tamoxifen in ER+ disease:
value of long-term follow-up, particularly for mortality
(Highly significant mortality gains in years 0-4, then in 5-9, then in 10-14)
Recurrence
Breast cancer mortality
EBCTCG, Lancet 2011; 378: 771
RARE EXAMPLE of a real difference in efficacy in different
patients: ER status is predictive of response to tamoxifen
ER negative
ER positive
ER POSITIVE
EBCTCG, Lancet 2011; 378: 771
Selected EBCTCG NEJM/Lancet reports
History: 1985-2010
1985-88: Chemo & endocrine therapy additively affect 5-year survival
1990-92: Chemo & endocrine therapy additively affect 10-year survival
1995-98: Polychemotherapy benefits many different subgroups
1995-98: Tamoxifen benefits by treatment duration in many subgroups
2000-05: Chemo & endocrine therapy additively affect 15-year survival
2000-05: Radiotherapy after mastectomy in N+ affects 15-year survival
2005-10: Aromatase inhibitor (AI) vs tamoxifen delays recurrence
Past 5 years: 2011-15
2011: Radiotherapy after breast conservation improves 15-year survival
2011: 5y of tamoxifen improves 15-year survival in all ER+ subgroups
2012: Chemotherapy improves survival in many types of ER+ disease
Chemotherapy (any anthracycline-based regimen) vs no
adjuvant chemotherapy: benefit in ER- & ER+ disease
ER-Negative
ER-Positive
EBCTCG, Lancet 2012; 379: 432
Chemo-endocrine* vs same endocrine alone: ADDITIVE and
similar benefits for younger and older women (ER+ disease)
ER+ age <55
ER+ age 55-69
*chemo=standard anthracycline-based regimen or CMF; endocrine is for 5 years
EBCTCG, Lancet 2012; 379: 432
Modern chemotherapy vs older chemotherapy:
moderate further benefit on recurrence and mortality
(anthracycline-based regimen ± a taxane)
Recurrence
Breast cancer mortality
EBCTCG, Lancet 2012; 379: 432
10-year effects of modern chemotherapy
Proportional reduction of about 1/3 in breast cancer mortality,
even in postmenopausal women with endocrine-treated ER+
disease (which is contrary to strongly-held previous beliefs)
Selected EBCTCG NEJM/Lancet reports
History: 1985-2010
1985-88: Chemo & endocrine therapy additively affect 5-year survival
1990-92: Chemo & endocrine therapy additively affect 10-year survival
1995-98: Polychemotherapy benefits many different subgroups
1995-98: Tamoxifen benefits by treatment duration in many subgroups
2000-05: Chemo & endocrine therapy additively affect 15-year survival
2000-05: Radiotherapy after mastectomy in N+ affects 15-year survival
2005-10: Aromatase inhibitor (AI) vs tamoxifen delays recurrence
Past 5 years: 2011-15
2011: Radiotherapy after breast conservation improves 15-year survival
2011: 5y of tamoxifen improves 15-year survival in all ER+ subgroups
2012: Chemotherapy improves survival in many types of ER+ disease
2015: AI vs tamoxifen improves survival; bisphosphonates add benefit
2015: Current smoking critically increases long-term RT side-effects
Side-effects of modern radiotherapy regimens
in today’s patients, according to smoking habits
Lung cancer mortality
(if the excess risk is half
that in the old trials)
Cardiac mortality
(if mean heart dose is 2 Gy;
1/3 that in old trials)
Lancet, submitted Oct 2015
(confidential until published)
Finding many MODERATE (but then additive)
treatment effects on LONG-TERM survival
Having ALL the main trials in the EBCTCG meta-analyses
increases the numbers of outcomes (so is more precise), and
avoids undue emphasis on positive studies (so no bias)
Meta-analyses involving “big data” help distinguish between
“true negatives” (e.g. endocrine therapy in ER- disease) and
“false negatives” (e.g. chemotherapy in ER+ disease)
More examples from the 2011-15
EBCTCG Lancet publications
• Radiotherapy after surgery: 10,000 women
• Late side-effects of radiotherapy: 40,000 women
Effects of giving radiotherapy (RT) to a
conserved breast: 10,000 randomised
Recurrence
Breast cancer mortality
EBCTCG, Lancet 2011; 378: 1707
More examples from the 2011-15
EBCTCG Lancet publications
• Radiotherapy after surgery: 10,000 women
• Late side-effects of radiotherapy: 40,000 women
• Chemotherapy: 10,000 women per question
• Endocrine therapy: 10,000 women per question
• Bisphosphonates: 10,000 women per question
Editorial: Postmenopausal breast cancer
– a best endocrine strategy?
“……For decades, the Early Breast Cancer
Trialists’ Collaborative Group meta-analyses
have informed clinical practice in early stage
breast cancer, and the present report in The
Lancet on over 30,000 women treated with
adjuvant tamoxifen, aromatase inhibitors, or
a sequence of those agents continues that
tradition.”
Erica Mayer & Harold Burstein
Dana-Farber Cancer Institute
Lancet; July 2015
5 years of aromatase inhibitor (AI) vs 5 years tamoxifen in
ER+ early breast cancer trials: Effects on 10-year outcome
(AIs are for postmenopausal only; & side-effects include ~50% increase in fracture rates)
Recurrence
Breast cancer mortality
EBCTCG, Lancet; online July 24, 2015
2-5 years of a bisphosphonate vs none: Effects on 10-year
breast cancer mortality in post-menopausal subgroup only
(for whom bisphosphonate can protect those at risk of AI-induced bone fracture)
EBCTCG, Lancet; online July 24, 2015
1985-2015: The EBCTCG meta-analyses have formed
the basis for clinical practice guidelines worldwide:
This has been a major contributor to the almost halving
of national death rates in many countries over this period
EBCTCG meta-analyses have also resulted in a better
understanding of who does and does not derive
worthwhile benefit from breast cancer treatment:
This has been especially important for ER+ disease (which
is 75-80% of US/UK breast cancer), showing the long-term
gains from the combination of endocrine and chemotherapy
2011-2015: EBCTCG meta-analyses continue
to resolve major clinical uncertainties
For example:
• Bisphosphonates improve breast cancer survival,
although they are not currently recommended
• Starting with AI is better than starting with tamoxifen,
whereas either option is currently recommended
• Radiotherapy after mastectomy improves survival for
women with 1-3 positive lymph nodes, although it is
not currently recommended
2015 onwards: EBCTCG’s large-scale randomized
evidence will help identify extra MODERATE benefits
•
•
•
•
•
•
•
•
•
10 vs 5 years of tamoxifen
Aromatase inhibitor use beyond year 5
Dose-dense chemotherapy
Sequencing of taxanes
Herceptin (trastuzumab) vs not
Duration of herceptin treatment
Avastin (bevacizumab) vs not
Radiotherapy to internal mammary nodes
Mammographic screening
as well as help identify any side-effects that
may emerge during long-term follow-up
2015 onwards: EBCTCG’s large-scale randomized
evidence will help investigate who does and does not
derive worthwhile benefit from different treatments
• Studies of prognostic factors to identify low risk of recurrence
(and, hence, small potential benefit from adjuvant treatment)
• Studies of tumour and patient characteristics that may predict
response to particular treatments, hence avoiding ineffective
treatment:
◦
◦
◦
◦
Chemotherapy efficacy by tumour genotype
Chemotherapy efficacy in older women
Herceptin efficacy by HER-2 level
Influence of other molecular markers
UK & USA: Breast cancer mortality 1950-2015 at ages 35-69
Probability of a
35-year-old dying
of breast cancer
aged 35-69
Risks halved, 1990-2015
UK: 2.4% down to 1.2%
US: 1.9% down to 1.1%
(& risks are still falling)