Egger Test

A statistical test for funnel plot asymmetry, and by extension potential publication bias, using regression of effect estimates against standard errors.

Source: Egger et al. 1997

The Egger test measures asymmetry in a funnel plot by regressing each study's effect estimate, scaled by its precision, against that precision. If small, imprecise studies systematically report larger effects than large ones, the regression line departs from what symmetry would give, and the test flags it. This asymmetry is often read as a sign of publication bias, though other causes can produce it. Quantifying the tilt of the funnel is its function. Egger and colleagues (1997) describe this test.

Source: Egger et al. 1997

The Egger test works by regressing a measure of the effect estimate, standardised by its precision, against a measure of precision, such as the inverse standard error, across the studies. The intercept of this regression captures asymmetry: a significant departure from zero indicates that smaller, less precise studies have systematically different effects, consistent with funnel plot asymmetry and possible publication bias. A non-significant intercept is consistent with symmetry. So the Egger test detects asymmetry by testing whether the regression intercept differs from zero, quantifying the association between effect size and precision that publication bias would produce.

Source: Egger et al. 1997

The Egger test detects asymmetry in a funnel plot, which can indicate publication bias, where smaller studies with unfavourable or non-significant results are less likely to be published, leaving a skewed set of studies. However, funnel plot asymmetry can also arise from causes other than publication bias, such as genuine heterogeneity, differences in study quality, or chance, so a significant Egger test signals possible bias rather than confirming it. So the Egger test detects the asymmetry that publication bias would cause, but its result is interpreted as suggestive, requiring consideration of alternative explanations for the asymmetry.

Source: Sterne, Gavaghan & Egger 2000

The limitations of the Egger test include low power when few studies are included, so it may miss real bias; its detection of asymmetry, which can arise from causes other than publication bias, such as heterogeneity or quality differences, so a positive result does not confirm bias; and potential unreliability with particular effect measures or when studies are very heterogeneous. These limitations mean the Egger test is interpreted cautiously, used alongside visual inspection of the funnel plot and other methods, and its result treated as indicating possible publication bias to be investigated rather than as definitive evidence of it.

Source: Egger et al. 1997

The Egger test relates to the funnel plot as a formal statistical assessment of the plot's symmetry: the funnel plot displays each study's effect against its precision, expected to be symmetric in the absence of bias, and the Egger test quantifies whether the plot is asymmetric through a regression, providing an objective complement to visual inspection. Asymmetry in the funnel plot corresponds to a significant Egger test. So the Egger test formalises the assessment of funnel plot asymmetry, giving a statistical test to accompany the visual examination, both aimed at detecting the asymmetry that may indicate publication bias.

Source: Egger et al. 1997