Transcript No Slide Title

1
Levels of evidence
Ia
Ib
IIa
IIb
III
IV
Meta-analysis or systematic review of RCTs
Randomised trial
Controlled non-randomised study
Cohort study
Case-control study
Descriptive study
Non-systematic review
Consensus report
Leading article
2
Survey of 85 physicians (36% from internal medicine)
”Very important” in influencing my prescribing
Own training and experience
Scientific papers
Advice from colleagues
Detail men
Drug adds
Patient preference
88%
62%
48%
20%
4%
2%
(Avorn J. Am J Med 1982;73:4-8)
3
”Is impaired cerebral blood flow a
major cause of senile dementia?”
Yes
No
No opinion
71%
14%
15%
32% found cerebral vasodilators useful
in managing confused geriatric patients
4
Survey of 85 physicians (36% from internal medicine)
”Very important” in influencing my prescribing
Own training and experience
Scientific papers
Advice from colleagues
Detail men
Drug adds
Patient preference
88%
62%
48%
20%
4%
2%
(Avorn J. Am J Med 1982;73:4-8)
5
Drug adds
287 advertisements for anti-hypertensive or lipidlowering drugs
- 125 promotional claims with references
- 23 refs. unretrievable (data on file, monographs)
- 45 of 102 claims not supported by reference
(Lancet 2003;361:27)
6
Randomised trials
Unclear randomisation method:
- effect exaggerated by 30%, on average
No blinding:
- effect exaggerated by 14%, on average
(BMJ 2001;323:42-6)
7
Outcome reporting bias
102 RCTs approved by the Copenhagen &
Frederiksberg Ethics Review Committee 1994-95
and subsequently published
Incompletely reported outcomes for meta-analysis:
50% for efficacy, 65% for safety
8
Outcome reporting bias
Unreported outcomes
86% of trialists denied unreported outcomes
despite evidence in publications & protocols
Only half of the trialists responded to the question
9
Outcome reporting bias
Full outcome reporting is associated with p<0.05
Odds ratio 2.4 (1.4 - 4.0) for efficacy
Odds ratio 4.7 (1.8 - 12) for safety
10
Are primary outcomes consistent
between protocols and publications?
Discrepancy in primary outcomes
Proportion (%) of trials
with inconsistencies
Changes to protocol-defined outcome
53% (40/76)
New publication-defined outcome
33% (21/63)
Change in power calculation outcome
29% (10/38)
ANY INCONSISTENCY
63% (52/82)
Conclusions
 Trial
outcomes are often inadequately reported
for inclusion in meta-analyses
 Reporting
of outcomes is biased to favour
p<0.05
 Primary
outcomes are omitted, changed, or
newly-introduced in over 60% of trials
Protocols should be publicly available
13
‘Positive’ studies are more likely to
be published than ‘negative’ studies
 Hazard
ratio for 130 clinical trials
‘positive’ (P<0.05) vs ‘negative’ (P>0.10)
3.13 (1.76 to 5.58), P=0.0001
 Median time to publication 4.7 vs 8 years
14
Cochrane Reviews
Freely available from
www.cochrane.dk
15
Based on
Cochrane Reviews
when possible
16
NSAIDs
“Systematic reviews of RCTs have found no important
differences in effect between different NSAIDs or doses
but have found differences in toxicity related to increased
doses and possibly to the nature of the NSAID itself.”
“The only meta-analysis that found one drug to be more
effective than another was funded by the manufacturer”
Clinical Evidence 1999;2
17
Clinical Evidence
(NSAIDs)
We have favoured systematic reviews that have not been
sponsored or authored by industry ...
it is easy to seemingly follow the rules for systematic
reviews and yet adopt inclusion and exclusion criteria
that omit inconvenient studies. In fact, it is hard to find
a systematic review sponsored by, or co-authored by,
industry that concludes that the company’s product is
not better than those of its competitors.
18
Celecoxib
Conclusion, industry supported meta-analysis
Celecoxib ... has significantly improved gastrointestinal
safety and tolerability (BMJ Sept 2002)
Conclusion, Cochrane Review
For an individual with RA the potential benefits of celecoxib
need to be balanced against the uncertainty that the shortterm reduced incidence of upper GI complications are
maintained in the long-term and its increased cost in
comparison to traditional NSAIDs.
19
Industrisponsorerede forsøg med
lægemidler
Ofte problemer med:
• Design
• Data-analyse
• Afrapportering
• Konklusion
20
Forsøg med psykofarmaka
Gamle præparater:
- Alt for høj dosis, alt for hurtig dosisøgning.
Nye præparater tilsyneladende lige så gode,
med langt færre bivirkninger.
Men: udbredt manipulation med dosis, dataanalyse og afrapportering.
(J Nerv Ment Dis 2002;190:583)
(BMJ 2003;326:1171)
21
Behandling for skizofreni
Olanzapin
Haloperidol
17.000 kr/år
2.400 kr/år
Forbrug i primærsektoren
Olanzapin
216 mio kr i 2002
Haloperidol
4 mio kr i 2002
”the new drugs have no unequivocal advantages
for first line use”
(BMJ 2000;321:1371)
22
Industrisponsoreret forskning versus
offentligt sponsoreret forskning
Chancen for et positivt udfald 4 gange så
stor for forsøg eller systematiske oversigter
over flere forsøg.
(BMJ 2003;326:1167)
Chancen for en positiv konklusion 5 gange
så stor for forsøg, trods samme effekt.
(JAMA 2003;290:921)
23
Cochrane Reviews
24
Hvilken forskning mangler vi?
- Sammenligninger med gamle, billige præparater, og
på en fair måde (offentlig finansiering)
- Sammenligninger med andet end lægemidler
Forebyggelse af ikke-insulinkrævende
sukkersyge:
- metformin:
31% effekt
- motion og vægttab: 56% effekt
(N Engl J Med 2002;346:393)
- Forskning i skadevirkninger
25
26