How often do hepatobiliary randomised trials have low risk of bias and hence low risk of systematic

1
How often do hepato-biliary randomised trials
have low risk of bias- and hence low risk of
systematic error?

• Christian Gluud
• Cochrane Hepato-Bilary Group
• Copenhagen Trial Unit
• Denmark

2
(No Transcript)
3
Outline of the talk

• Previous CHBG studies on bias risks
• Are hepato-biliary trials worse?
• Recent reviews
• New publication on bias risks
• Conclusions

4
Outline of the talk

• Previous CHBG studies on bias risks
• Are hepato-biliary trials worse?
• Recent reviews
• New publication on bias risks
• Conclusions

5
Bias components

• Generation of the allocation sequence
• Allocation concealment
• Blinding
• Incomplete outcome data
• Selective outcome reporting
• Other bias components
• - Baseline imbalance
• - Early stopping
• - Invested interests

6
Adequate allocation sequence generation

• JHEP
  Hepat. Gas.
• Period 1985-1997 1981-1998 1964-2000
• Proportion 28 51
  42
• (95 CI) (22-36) (45-58)
  (37- 47)

7
Adequate allocation concealment

• JHEP
  Hepat. Gas.
• Period 1985-1997 1981-1998 1964-2000
  
• Proportion 13 34 39
  
• (95 CI) (9-20) (28- 41)
  (34- 44)

8
Adequate blinding

• JHEP Hepat.
  Gas.
• Period 1985-1997 1981-1998 1964-2000
  
• Proportion 30 34 62
  
• (95 CI) (23-38) (28- 41)
  (57- 67)

9
Adequate sample size calculation

• JHEP Hepat.
  Gas.
• Period 1985-1997 1981-1998 1964-2000
  
• Proportion 20 26 NR
  
• (95 CI) (14-27) (21-32)
  (NR-NR)

10
Outline of the talk

• Previous CHBG studies on bias risks
• Are hepato-biliary trials worse?
• Recent reviews
• New publication on bias risks
• Conclusions

11
616 randomised
519 randomised trials
hepato-biliary trials from all disease
areas (1985 to
1996) (December 2000) Proportion
with adequate - generation of the
48
21 allocation sequence - allocation concealment
38 18 -
blinding 34
38
12
616 randomised
519 randomised trials
hepato-biliary trials from all disease
areas (1985 to 1996)
(December 2000) Median number
of participants per intervention arm
(10th90th percentiles)
23 participants
32 participants
(7102 participants) (12159
participants)
13
Outline of the talk

• Previous CHBG studies on bias risks
• Are hepato-biliary trials worse?
• Recent reviews
• New publication on bias risks
• Conclusions

14
Review authors' judgements about each
methodological quality component presented as
percentages across all included trials
Gurusamy KS, Samraj K, Fusai G, Davidson BR.
Robot assistant for laparoscopic cholecystectomy.
Cochrane Database of Systematic Reviews 2009,
Issue 1.
15
Review authors' judgements about each
methodological quality component presented per
trial
Gurusamy KS, Samraj K, Fusai G, Davidson BR.
Robot assistant for laparoscopic cholecystectomy.
Cochrane Database of Systematic Reviews 2009,
Issue 1.
16
Outline of the talk

• Previous CHBG studies on bias risks
• Are hepato-biliary trials worse?
• Recent reviews
• New publication on bias risks
• Conclusions

17

• Bai et al, HEPATOLOGY 2009
• Assessed 105 trials published in 2006 in
• Journal of Hepatology
• Hepatology
• Gastroenterology
• Gut
• American Journal of Gastroenterology
• Clinical Gastroenterology and Hepatology

18
Adequate allocation sequence generation

• JHEP Hepat.
  Gas. Six j.
• Period 1985-1997 1981-1998 1964-2000
  2006
• Proportion 28 51 42
  81
• (95 CI) (22-36) (45-58)
  (37-47) (72-88)

19
Adequate allocation concealment

• JHEP Hepat.
  Gas. Six j.
• Period 1985-1997 1981-1998 1964-2000
  2006
• Proportion 13 34 39
  61
• (95 CI) (9-20) (28-41)
  (34-44) (51-70)

20
Adequate blinding

• JHEP Hepat.
  Gas. Six j.
• Period 1985-1997 1981-1998 1964-2000
  2006
• Proportion 30 34 62
  51
• (95 CI) (23-38) (28-41)
  (57-67) (42-61)

21
Adequate sample size calculation

• JHEP Hepat.
  Gas. Six j.
• Period 1985-1997 1981-1998 1964-2000
  2006
• Proportion 20 26 NR
  75
• (95 CI) (14-27) (21-32)
  (NR-NR) (66-83)

22
Outline of the talk

• Previous CHBG studies on bias risks
• Are hepato-biliary trials worse?
• Recent reviews
• New publication on bias risks
• Conclusions

23
Conclusions

• Most hepato-biliary trials carry risk of
• bias
• The majority even has several
• components associated with bias
• It is therefore difficult to estimate the
• true intervention effect

24
Conclusions

• Hepato-biliary trials does not seem to
• be
• - larger than
• - have substantially less bias risk than
• trials from other therapeutic fields

25
Conclusions

• There has recently been a significant
• improvement in trial reporting
• There is still ample room for further
• improvement
• All trial protocols must be published
• before randomisation of the first
• participant

26
Thank you