This correlation can create a false asymmetry in a funnel plot even when there is no bias. Artefactual: Certain effect estimates, such as odds ratios or standardised mean differences, are inherently correlated with their standard errors.These high-risk patients are more likely to be included in small, early trials, leading to asymmetry in the funnel plot. True heterogeneity : Sometimes a significant benefit of an intervention can only be observed in patients who are at high risk for the outcome targeted by the intervention.Poor methodological quality leading to exaggerated effects: Studies with inferior methods may show larger effect estimates of an intervention than would have been observed in a well-designed study.Non-reporting bias: Some studies, or specific results, are less likely to be published if they are not statistically significant, or if the effect size is very small or non-existent.Possible reasons for asymmetry in a funnel plot are: The greater the asymmetry, the greater the likelihood that the amount of bias in the meta-analysis will be significant. In this case, the meta-analysis summary estimate will tend to overestimate the intervention effect. Image A is shown again below alongside Image B, which depicts an asymmetrical funnel due to presence of bias (the points are now predominantly towards the left). In the absence of both bias and heterogeneity, 95% of studies would be expected to lie within the diagonal dotted ‘95% Confidence Interval’ lines, as shown in Figure A.Īs a rule of thumb, tests for funnel plot asymmetry should only be used when at least 10 studies are included in the meta-analysis, because the power of the tests is low when there are fewer studies.The scale for the x-axis can include risk ratios or odds ratios (which should be plotted on a logarithmic scale), or continuous measures such as mean difference or standardised mean difference. The x-axis displays the study estimated effect size for an outcome.Other measures such as the reciprocal of the standard error, the reciprocal of the sample size, or variance of the estimated effect can also be used as the y-axis. Larger studies with greater precision are displayed at the top and studies with lower precision at the bottom. The y-axis represents a measure of study precision, with standard error being commonly used. ![]() Each included study is represented as a dot.
0 Comments
Leave a Reply. |