Effect of Rosiglitazone on the Risk of Myocardial Infarction And Death from Cardiovascular Causes Al

1
Effect of Rosiglitazone on the Risk of Myocardial
Infarction And Death from Cardiovascular Causes
Alternative Interpretations of the Evidence
George A. Diamond, MD Sanjay Kaul, MD Division
of Cardiology Cedars-Sinai Medical Center Los
Angeles, California No conflicts to disclose
2
(No Transcript)
3
Rosiglitazone and Cardiovascular Events
Myocardial Infarction
27,833 Patients 158 Events 42 Trials
15,470
12,205
No Event
86
72
MI
Rosiglitazone Control
0.59 0.55
Event Rate
4
Rosiglitazone and Cardiovascular Events
Myocardial Infarction
5
Rosiglitazone and Cardiovascular Events
Cardiovascular Death
6
Rosiglitazone and Cardiovascular Events
Peto Meta-Analysis
Myocardial Infarction
Cardiovascular Death
1
Odds Ratio

Odds Ratio

1.43 (1.03-1.98) p0.03 N38
1.64 (0.98-2.14) p0.06 N23
7
Rosiglitazone and Cardiovascular Events
Myocardial Infarction
8
Rosiglitazone and Cardiovascular Events
Cardiovascular Death
9
Rosiglitazone and Cardiovascular Events
Impact of Zero Events on Petos Odds Ratio
10
Rosiglitazone and Cardiovascular Events
Cardiovascular Death
11
Rosiglitazone and Cardiovascular Events
Continuity Correction
Sweeting et al, What to add to nothing? Stat Med
2006231351-75.
12
Rosiglitazone and Cardiovascular Events
Meta-Analytic Sensitivity
Myocardial Infarction
Cardiovascular Death
Peto ( - ) Inverse variance 1/N ( - ) Inverse
variance 1/2 ( - ) Mantel-Haenszel 1/N ( -
) Mantel-Haenszel 1/2 ( - ) Mantel-Haenszel 1/N
() Mantel-Haenszel 1/2 () Uniform Bayes 1/N
() Uniform Bayes 1/2 ()
0.5 1.0 1.5 2.0 2.5 3.0
0.5 1.0 1.5 2.0 2.5 3.0
Odds Ratio
Odds Ratio
13
Rosiglitazone and Cardiovascular Events
Magnitude of Harm
Myocardial Infarction
Cardiovascular Death
Uncorrected
Uncorrected
Probability of Harm
Corrected
Corrected
Relative Risk Threshold
Relative Risk Threshold
14
Rosiglitazone and Cardiovascular Events
Limitations of the Published Meta-Analysis

• Not designed to assess outcomes
• No central adjudication of events
• No standardized definitions of events
• Limited sample size
• Short term duration
• No patient level data
• No sensitivity analysis
• No continuity correction

15
Effect of Rosiglitazone on the Risk of Myocardial
Infarction And Death from Cardiovascular Causes
Alternative Interpretations of the Evidence
Sanjay Kaul, MD George A. Diamond, MD Division
of Cardiology Cedars-Sinai Medical Center Los
Angeles, California No conflicts to disclose
16
Rosiglitazone and Cardiovascular Events Key
Questions Regarding the Published Meta-Analysis

• Are the risk estimates robust?
• Is there heterogeneity?
• - What is the impact of continuity corrections
  on clinically relevant subgroups?
• Are the risk estimates consistent with other
  studies?

17
Rosiglitazone and Cardiovascular Events Is There
Heterogeneity?

• Pooling justified due to lack of statistical
  heterogeneity
• Cochrans Q test of heterogeneity Limited
  ability to detect variability across studies with
  sparse data (low statistical power)

Even if studies are statistically homogeneous
there may be clinical heterogeneity in study
design and population
18
Rosiglitazone and Cardiovascular Events Clinical
Heterogeneity in Patient Populations
19
Rosiglitazone and Cardiovascular Events Clinical
Heterogeneity in Trial Design
20
Rosiglitazone and Cardiovascular Events Clinical
Heterogeneity in Treatment Groups
21
Rosiglitazone and Cardiovascular Events Is There
Heterogeneity?
Absence of statistical heterogeneity does not
imply absence of clinical heterogeneity
22
Rosiglitazone and Cardiovascular
Events Meta-Analytic Subgroups
Myocardial Infarction
23
Rosiglitazone and Cardiovascular
Events Meta-Analytic Subgroups
Cardiovascular Death
24
Rosiglitazone and Cardiovascular
Events Meta-Analytic Subgroups
Myocardial Infarction
25
Rosiglitazone and Cardiovascular
Events Meta-Analytic Subgroups
Cardiovascular Death
26
Rosiglitazone and Cardiovascular Events Are the
Risk Estimates Consistent?
Myocardial Infarction/Ischemia
GSK ICT analysis (N42 trials)
RECORD (N4407)
Balanced Cohort Study (N33363)
0
1
2
3
Rate ratio
Nonsignificantly increased odds ratio
Cochrane Review (N18 trials)
27
Rosiglitazone and Cardiovascular
Events Conclusions

• Sensitive to meta-analytic method
• Sensitive to continuity correction
• Sensitive to subgroup analysis
• If present, magnitude of harm is small

We need more data!
28
(No Transcript)