BMJ No 7109 Volume 315
Papers Saturday 13 September 1997
Bias in meta-analysis detected by a simple, graphical test
Matthias Egger, George Davey Smith, Martin Schneider, Christoph Minder
| Table 1-Characteristics of nine pairs of meta-analyses and corresponding large trials | |||
|---|---|---|---|
| Topic | Typical end point | Large trial (year of publication) | Meta-analysis (year of publication) |
| Concordant pairs | |||
| Beta Blockers in myocardial infarction | Mortality in hospital | ISIS-1 (1986)20 | Yusuf et al (1985)19 |
| Streptokinase in myocardial infarction | Mortality in hospital | GISSI-1 (1986)18 | Yusuf et al (1985)17 |
| Angiotensin converting enzyme inhibitors in heart failure | Mortality at 3 months | SOLVD (1991)15 | Mulrow et al (1988)26 |
| Intensive therapy in insulin-dependent diabetes mellitus | Progression of retinopathy over several years | DCCT (1993)22 | Wang et al (1993)21 |
| Discordant pairs | |||
| Magnesium in myocardial infarction | Mortality in hospital | ISIS-4 (1995)28 | Teo et al (1993)27 |
| Nitrates in myocardial infarction | Mortality in hospital | GISSI-3 (1994)25 | Yusuf et al (1988)24 |
| Inpatient geriatric consultation service | Mortality at 6 months | HMO (1995)23 | Stuck et al (1993)14 |
| Aspirin for preventing pre-eclampsia | Development of pre-eclampsia | CLASP (1994)30 | Imperiale and Stollenwerk (1991)29 |
| GISSI=Gruppo Italiano per lo Studio della Streptochinasi nell'Infarto miocardico; ISIS=International Study of Infarct Survival; DCCT=Diabetes Control and Complications Trial; HMO=Health Maintenance Organisation study; SOLVD=Studies of Left Ventricular Dysfunction; CLASP, Collaborative Low-dose Aspirin Study in Pregnancy.. | |||
| Table 2-Analysis of funnel plot asymmetry | |||
| Meta-analysis |
|
|
|
| Intercept (90% CI) | P value | ||
| Results concordant with single large trial | |||
| Beta Blockers in myocardial infarction (19) | 26 | 0.44 (-0.11 to 1.00) | 0.19 |
| Streptokinase in myocardial infarction (17) | 20 | 0.59 (-1.30 to 2.48) | 0.59 |
| Angiotensin converting enzyme inhibitors in heart failure (26) | 13 | -0.14 (-0.44 to 0.16) | 0.43 |
| Intensive treatment in insulin-dependent diabetes mellitus (21) | 6 | -0.75 (-2.53 to 1.03) | 0.44 |
| Results discordant with single large trial | |||
| Magnesium in myocardial infarction (27 | 10 | -1.19 (-2.26 to -0.12) | 0.068 |
| Nitrates in myocardial infarction (24) | 10 | -1.84 (-3.25 to -0.43) | 0.043 |
| Inpatient geriatric consultation service (14) | 8 | -2.60 (-4.84 to -0.37) | 0.069 |
| Aspirin for preventing pre-eclampsia (29) | 6 | 0.37 (-1.84 to 2.59) | 0.75 |
Fig 1- Results from four concordant and four discordant pairs of
meta-analysis and large scale randomised controlled trial.
Fig 2- Funnel plots and single large trials. Points indicate odds ratios
from trials included in meta-analysis; squares with horizontal
lines show odds ratio from large trial with 95% confidence interval.
See table 1 for abbreviations of trial names.
Fig 3- Funnel plot of trials of low dose aspirin in the prevention of
pre-eclampsia. Trials included in Imperiale and Stollenwerk's
1991 meta-analysis (closed circles),(29) trials published in subsequent years (1990 to 1993, open circles)
and the large 1994 CLASP (collaborative low-dose aspirin study
in pregnancy) trial (square with horizontal line indicating 95%
confidence interval) (30)
Fig 4 - Distribution of intercepts from regression analysis of funnel
plot asymmetry for 38 meta-analyses from the Cochrane Database of Systematic
Bias in meta-analysis detected by a simple, graphical test
Home | Current contents | Past issues | Classified ads | Career Focus | Feedback
Collections | About this site | About the BMJ | BMA | Medline