Transcript Effective Presentations of Study Results

Effective Presentation of Study Results
How are RCTs presented in abstracts & publications?
and
Some things to consider in your own presentations
NCIC CTG New Investigators Course
October 2009
Christopher Booth MD FRCPC
Queen’s University Cancer Research Institute
The Cancer Research Institute at Queen’s University
Division of Cancer Care and Epidemiology
Outline: Part I
How are RCTs presented in abstracts & publications?
• Evolution of endpoints and perception of benefit
in oncology RCTs over time
• What results are clinically meaningful?
• RCTs closed early for benefit
• How are RCTs presented at conferences?
The Cancer Research Institute at Queen’s University
Division of Cancer Care and Epidemiology
Outline: Part II
Issues to consider when presenting your study
results…
• Audience
• Preparation
• Key messages
• Fancy PowerPoint
• Summary
The Cancer Research Institute at Queen’s University
Division of Cancer Care and Epidemiology
Part I
How are RCTs presented in
abstracts & publications?
The Cancer Research Institute at Queen’s University
Division of Cancer Care and Epidemiology
1. How have endpoints and perception of
benefit evolved in oncology RCTs?
Study Design
•
•
•
•
Overview of all RCTs systemic therapy in
breast, NSCLC, colorectal cancer
1975-2004
6 major journals: JCO, Cancer Treatment
Reports, JNCI, NEJM, Lancet, JAMA
Data abstraction using standardized forms and
methodology
The Cancer Research Institute at Queen’s University
Division of Cancer Care and Epidemiology
Temporal Trends
1975-84
1985-94
1995-2004
P value
(trend)
Total RCTs
47 (15%)
107 (33%)
167 (52%)
<0.0001
Breast RCTs
19 (40%)
53 (50%)
81 (49%)
0.475
NSCLC RCTs
23 (49%)
29 (27%)
39 (23%)
0.002
CRC RCTs
5 (11%)
25 (23%)
47 (28%)
0.017
321 RCTS over three decades
involving >170 000 patients
The Cancer Research Institute at Queen’s University
Division of Cancer Care and Epidemiology
Study Participants
1975-84
1985-94
1995-04
P (trend)
International 26%
Co-op group 28%
28%
56%
52%
43%
<0.0001
0.661
Sample size 100
Accrual time 30 mo
249
41 mo
446
33 mo
<0.0001
0.93
The Cancer Research Institute at Queen’s University
Division of Cancer Care and Epidemiology
Statistics
1975-84
1985-94
1995-04
P (trend)
1º endpoint
Time to event
39%
72%
78%
<0.0001
RR
ITT analysis
54%
25%
14%
<0.0001
Any
70%
87%
93%
<0.0001
33% of RCTs in 1995-2004 did not clearly identify primary endpoint
The Cancer Research Institute at Queen’s University
Division of Cancer Care and Epidemiology
Sponsorship
Government
Industry
1975-84
60%
1985-94
62%
1995-04
31%
P (trend)
<0.0001
4%
23%
57%
<0.0001
The Cancer Research Institute at Queen’s University
Division of Cancer Care and Epidemiology
Effect Size and Conclusions
1975-1984
1985-1994
1995-2004
Median HR (95%CI)
1.4 (1.0-2.3) 1.2 (1.0-2.4) 1.2 (1.1-1.3)
P<0.05 for primary EP
23%
30%
42%
Strong endorsement
31%
39%
49%
1. Effect size stable over time
2. Modern RCTs more likely to have p<0.05
3. Modern authors more likely to call their
trial “positive”
The Cancer Research Institute at Queen’s University
Division of Cancer Care and Epidemiology
The Cancer Research Institute at Queen’s University
Division of Cancer Care and Epidemiology
Key Findings
1.
2.
3.
4.
5.
6.
7.
Increase in number and size of RCTs
More international trials, faster accrual
Shift in primary endpoint from RR to survival EPs
Major shift towards for-profit sponsorship
Effect size has remained stable over time
Authors of modern RCTs more likely to endorse
experimental arm
For-profit sponsorship and p<0.05 are independently
associated with strong endorsement of experimental
arm
The Cancer Research Institute at Queen’s University
Division of Cancer Care and Epidemiology
2. Are these results clinically meaningful?
Clinically Relevant Endpoints
• Patients define a useful therapy as one that…
increases survival
OR
improves QOL/reduces symptoms of cancer
• Reassuring shift from RR to survival endpoints
• Increasing use and recognition of PROs
Gemcitabine
Pancreas
TAX 327
HRPC
The Cancer Research Institute at Queen’s University
Division of Cancer Care and Epidemiology
Clinically Relevant Endpoints
• Current standards for analyzing QOL and
symptom control need improvement
112 RCTs for advanced cancer
19% established a priori hypothesis
21% defined minimal differences in QOL scores that
were clinically meaningful
• Increasing use of surrogate endpoints (DFS, PFS)
Joly et al Ann Oncol 2007
Sargent et al JCO 2005
The Cancer Research Institute at Queen’s University
Division of Cancer Care and Epidemiology
The Cancer Research Institute at Queen’s University
Division of Cancer Care and Epidemiology
Clinically Relevant Endpoints
• Clinical benefit (CB) in trials of pancreas cancer
composite improvement in pain, weight, performance
status
• CB now widely (mis)used to describe PR/CR/SD
71 trials in JCO since 1997
28% used patient-centered definition
72% referred to objective tumor measurements
Burris et al JCO 1997
Ohorodnyk et al ASCO 2009
The Cancer Research Institute at Queen’s University
Division of Cancer Care and Epidemiology
3. RCTs Closed Early for Benefit
The Cancer Research Institute at Queen’s University
Division of Cancer Care and Epidemiology
Korn et al JCO 2009
Sargent JCO 2009
Korn et al JCO 2008
• 27 NCI co-operative group trials that were
closed early for benefit
• Of the 18 trials with follow-up available, initial
magnitudes of benefit were preserved in 17
(94%) trials
• “…the system is working”
Dan Sargent JCO 2008
The Cancer Research Institute at Queen’s University
Division of Cancer Care and Epidemiology
Booth, Meyer et al Under Review
• Literature search identified 62 RCTCEB
– Primary endpoint not explicitly stated in 19%
– ITT all randomized patients in only 66%
• Most trials open to accrual (45/62, 73%) at the
time of closure.
• Formal IA performed in 56 (90%) trials
– 75% (42/56) planned and 79% (44/56) reported
stopping rules.
• Trials on average accrued 73% of the planned
sample size.
• Follow-up reports for 18 (29%) RCTCEB show
that results and conclusions were maintained.
4. How are RCTs presented at conferences?
The Cancer Research Institute at Queen’s University
Division of Cancer Care and Epidemiology
Trial Reporting: NFAs
• 138 RCTs published 2000-2004
– 197 corresponding abstracts 1990-2004
• Results were stated or implied to be non-final
analyses in 86 abstracts (44%)
• 124 abstracts (63%) discordant with article
• Conclusions substantively different in 17 (10%)
Meeting abstracts often include NFAs and
are often discordant with mature publication
The Cancer Research Institute at Queen’s University
Division of Cancer Care and Epidemiology
Bias in Oncology RCTs
Publication Bias
• 510 RCTs presented at ASCO 1989-1998
26% were not published within 5 years
81% of trials with p<0.05 were published compared to
68% of trials with p>0.05
• This should improve with mandatory trial
registration
Krzyzanowska JAMA 2003
The Cancer Research Institute at Queen’s University
Division of Cancer Care and Epidemiology
Bias in Oncology RCTs
Sponsorship Bias
• Multiple studies have demonstrated that RCTs
sponsored by industry are more likely to be
“positive” than non-industry trials
Reasons are likely complex and multifactorial
Djulbegovic Lancet 2000
Booth JCO 2008
Peppercorn Cancer 2007
The Cancer Research Institute at Queen’s University
Division of Cancer Care and Epidemiology
ACS 2008 Statistics
Part I: SUMMARY
1. Major changes in cancer treatment/research since 1970s
Patient outcomes have improved
RCT methodology and reporting are improving
Trials are larger, complex, stronger correlative component
What constitutes a “positive trial” has changed
2. It is critical to keep in patient-centered outcomes in focus
at every step of the drug development pathway
3. Be critical in the way you interpret results of RCTs in
published form and at conferences
The Cancer Research Institute at Queen’s University
Division of Cancer Care and Epidemiology
Acknowledgements
Drs. Ralph Meyer and Bill Mackillop
NCIC Clinical Trials Group
Queen’s University Cancer Research Institute
Kingston, Ontario
Drs. Ian Tannock and Monika Krzyzanowska
Princess Margaret Hospital, Toronto, Ontario
The Cancer Research Institute at Queen’s University
Division of Cancer Care and Epidemiology
Part II:
Things to consider when presenting
your own study results
The Cancer Research Institute at Queen’s University
Division of Cancer Care and Epidemiology
Topics for Discussion
1.
2.
3.
4.
5.
6.
Audience
Preparation and timing
Key messages
Fancy PowerPoint (?)
Summary
Time for discussion
The Cancer Research Institute at Queen’s University
Division of Cancer Care and Epidemiology