Interpreting Information from Clinical Trials What are the Results

1
Interpreting Information from Clinical
TrialsWhat are the Results?

• lois.champion_at_lhsc.on.ca

2

• Lies, damn lies, and statistics.
• Benjamin Disraeli
• Corollary 46.7 of statistics are made up on
  the spot.

3
Objectives

• Primary and secondary outcomes
• Composite outcomes
• Surrogate outcomes
• Results ARR, RRR, NNT

4

• If you want to inspire confidence, give plenty of
  statistics. It does not matter that they should
  be accurate, or even intelligible, as long as
  there is enough of them.
• Lewis Carroll

5
Study Results Primary Outcome

• clinical trials done to answer specific question
• primary endpoint (outcome)
• must be clearly defined a priori
• conclusions must be based on primary endpoints

6
Secondary Outcomes

• other questions that are investigated as part of
  the trial
• also must be established a priori
• what if the primary outcome is not significant
  but secondary outcome is?

7
Secondary Outcomes

• an approach to secondary outcomes
• if primary outcome is negative then consider the
  secondary outcome exploratory or
  hypothesis-generating

8
Composite Outcomes

• size of trial (sample size required) depends on
  how likely an outcome will be
• higher event rates allow for smaller sample size
  or shorter follow-up or both
• avoid need to select a single outcome when
  several may reflect effect of therapy

9
Composite Outcomes

• identified before trial begins
• clinically meaningful
• similar importance to patients
• biologically plausible
• individual components of composite should also be
  reported separately (as secondary outcomes)

10
Composite Outcomes

• Possible problems with composite outcomes
• effect on each of components is not always the
  same
• individual outcomes might not be equally
  important
• reporting may imply results apply to all
  components of composite when it does not
• for example how do we interpret results of a
  study that shows a small increase in mortality
  and a decrease in hospitalizations and non-fatal
  MI?

11
Composite Outcomes

• PROactive Secondary prevention of macrovascular
  events in patients with type II diabetes. Lancet
  20053661279.
• placebo-controlled RCT of pioglitazone
• concealed allocation
• blinded (patients, clinicians, data collectors
  and outcome assessors)
• multicentre
• follow-up 99.96 with intention-to-treat analysis
• 5283 patients

12
PROactive Press Release

• The PROactive study is the first in the world to
  prospectively show that a specific oral glucose
  lowering drug, pioglitazone, can significantly
  improve cardiovascular outcomes by helping to
  delay or reduce heart attacks, strokes and death.
  This ground breaking study gives new hope to
  people with Type II diabetes.

13
Composite Outcomes

• PROactive Trial
• primary composite outcome
• all-cause mortality
• non fatal myocardial infarction (including silent
  MI)
• stroke
• major leg amputation
• acute coronary syndrome
• coronary artery bypass graft or angioplasty/stent
• leg revascularization

14
Composite Outcomes

• PROactive Trial
• primary composite outcome
• all-cause mortality
• non fatal myocardial infarction (including silent
  MI)
• stroke
• major leg amputation
• acute coronary syndrome
• coronary artery bypass graft or angioplasty/stent
• leg revascularization
• non significant difference 20 vs 22 p .10

15
Secondary Outcome

• PROactive trial
• reduces composite of all cause mortality,
  non-fatal myocardial infarction and stroke
• p .03
• what would you call this outcome?
• prespecified ? unclear

16
PROactive Press Release

• The PROactive study is the first in the world to
  prospectively show that a specific oral glucose
  lowering drug, pioglitazone, can significantly
  improve cardiovascular outcomes by helping to
  delay or reduce heart attacks, strokes and death.
  This ground breaking study gives new hope to
  people with Type II diabetes.

17
Quick Summary

• primary outcome (endpoint)
• secondary outcomes
• caution
• must be prespecified
• caution in interpretation if primary outcome
  negative
• composite outcomes
• increases event rate
• decreases sample size and/or length of study

18
Multiple Outcomes

• trials may use multiple primary outcomes
• these must be prespecified
• use of multiple endpoints require statistical
  adjustment

19
Surrogate Outcomes

• biological or imaging markers that are believed
  to be indirect measures of effect of treatment
• what are some advantages?
• can you think of an example?
• caution may be misleading

20
Surrogate Outcomes

• Criteria for valid surrogate outcome
• change in surrogate must predict relevant
  clinical outcome.
• surrogate must capture effect of intervention on
  clinical outcome.

21
Surrogate Outcomes

• some examples of surrogates that have misled us
• antiarrhythmic medications (CAST study)
• nitric oxide inhalation in ARDS
• improved oxygenation but no change in mortality
• nitric oxide antagonist in septic shock
• increased BP, but also increased mortality
• milrinone in heart failure
• increased cardiac contractility, increased
  mortality

22
Surrogate Outcomes - CAST

• PVCs precede fatal arrhythmia and sudden death
• use of antiarrhythmic medication (flecainide)
  associated with decrease of PVCs
• reasoning

23
Suppression of Arrhythmias
24
CAST Trial
25
Surrogate Outcomes

• be very careful in using surrogate outcome as
  evidence of benefit
• note surrogates may be used in marketing of
  pharmaceuticals

26
Quick Summary

• outcomes
• primary
• secondary
• composite
• surrogate outcomes

27
Clinical and Statistical Evidence

• statistically significant does not equal
  clinically significant

28
Outcome Reporting Bias

• retrospective review of trials
• prevalence of incompletely reported outcomes
• median of 11 outcomes/trial
• not all outcomes reported
• at least one unreported efficacy outcome in 33
  of trials
• BMJ 2006

29
Outcome Reporting Bias

• cohort study using protocols and published
  reports 1994-1995
• identified 102 trials
• overall 50 of efficacy and 65 of harm outcomes
  incompletely reported / trial
• statistically significant outcomes more likely to
  be reported (OR 2.4, CI 1.4 4.0)
• JAMA 2004

30
(No Transcript)
31
Summary To Avoid Being Misled

• read methods and results section
• read abstracted report in evidence-based
  secondary publication
• watch out for surrogate outcomes
• careful with composite outcomes

32
Study Results

• A therapy results in a 26 relative decrease in
  the incidence of fatal myocardial infarction.
• A therapy results in a 34 relative decrease in
  the incidence of fatal and nonfatal MI.
• A therapy results in a 1.4 decrease in the
  incidence of fatal and nonfatal MI. The decrease
  is statistically significant.
• A therapy results in a relative increase in
  all-cause mortality of 5.7 (not stat. sig.).
• A therapy results in a 0.1 decrease in the
  incidence of fatal MI.
• A therapy has the following characteristics 77
  persons must be treated for an average of over 5
  years to prevent 1 fatal or nonfatal MI.

33
Measures of Association

• relative
• relative risk reduction
• odds ratio
• absolute
• event rates
• absolute risk reduction
• number needed to treat

34
Event Rate
40
35
Event Rate
40
30
Event Rate Absolute Reduction
36
Event Rate
40
30
Event Rate Absolute Reduction 10 Relative
Reduction 25
37
Event Rate
40
30
Event Rate Absolute Reduction Relative
Reduction
4
3
38
Event Rate
40
30
Event Rate Absolute Reduction 1 Relative
Reduction 25
4
3
39
Event Rate

• number of people experiencing event as proportion
  of population
• control event rate is baseline risk
• experimental event rate is rate in experimental
  group

40
Absolute Risk Reduction

• ARR
• absolute difference between risk of event in
  control and experimental group
• ARR CER - EER

41
Relative Risk Reduction

• RRR
• reduction of adverse events in experimental group
  relative to proportion in control group
• may be more impressive than ARR
• RRR ARR /CER

42
Using a 2 x 2 Table
Control Event Rate (CER) c / (c d)
43
Using a 2 x 2 Table
Experimental Event Rate (CER) a / (a b)
44
Using a 2 x 2 Table
Absolute Risk Reduction (ARR) CER - EER
45
Using a 2 x 2 Table
Relative Risk Reduction (RRR) ARR / CER
46
Using a 2 x 2 Table PROactive Trial
Control Event Rate (CER) c / (c d) 0.136
47
Using a 2 x 2 Table PROactive Trial
Control Event Rate (CER) c / (c d)
0.136 Experimental Event Rate (EER) a / (a b)
0.116
48
Using a 2 x 2 Table PROactive Trial
Control Event Rate (CER) c / (c d) 0.136
13.6 Experimental Event Rate (EER) a / (a
b) 0.116 11.6 Absolute Risk Reduction (ARR)
2 Relative Risk Reduction (RRR) ARR/CER
14.7
49
Number Needed to Treat

• number of people needed to treat to prevent one
  adverse event over specific time
• NNT 1 / ARR
• rounded UP to nearest integer

50
PROactive Trial

• ARR 2
• NNT 1 / .02 50 (95 CI 27 407)
• therefore 50 patients would need to be treated
  with Pioglitazone for 3 years to prevent one
  event (composite outcome of mortality, non-fatal
  MI or stroke)

51
Number Needed to Harm

• calculated the same way as NNT
• NNH 1 / ARI

52
NNH PROactive Trial

• heart failure (diagnosis or hospitalization)
• absolute risk increase 3 (11 - 8)
• NNH 1 / .03 33. 3 34
• heart failure requiring hospitalization
• ARI 2 (6 - 4)
• NNH 1 / .02 50

53
Odds Ratio PROactive Trial

• OR (301 x 2275) / (2304 x 358)
• OR 0.83
• OR varies from 0 to infinity, OR of 1 is no
  effect

54
Quick Summary

• measures of association
• absolute ARR
• relative RRR
• NNT 1 / ARR
• relative risk tends to look more impressive

55
(No Transcript)
56
(No Transcript)
57
Thank you