Publication bias is best detected in a meta-analysis using which of the following graphical forms?
A. Funnel plotAnswer A
Funnel plots graphically display a measure of study precision (vertical axis) against treatment effects for each of the studies included in the meta-analysis. The resultant symmetry or indeed asymmetry can give an indication of the presence of publication bias. Publication bias describes the phenomenon whereby systematic reviews and meta-analysis do not necessarily include all available studies of interest; smaller studies without statistically significant results are less likely to be published. In the absence of publication bias, a symmetrical inverted funnel (hence the name funnel plot) is generated because the treatment effect estimates from smaller studies scatter more widely at the base of the graph with those of larger studies narrowing towards the apex. If a meta-analysis is over-represented by larger studies with more significant results, and therefore, smaller studies are not included, asymmetry of the ‘funnel’ occurs. Funnel plots are commonly used by the Cochrane collaboration reviews. Receiver operator characteristic (ROC) curve is a plot of a test’s (with a binary outcome) sensitivity (y-axis) against 1 – specificity (x-axis) using different cut-off values. By calculating the area under the curve, the accuracy (discriminatory ability) of the test can be determined. Forest plots are commonly used in a meta-analysis of randomised controlled trials and is a way of presenting the outcome data for each RCT included in the meta-analysis (including the confidence intervals and weight of each RCT in the meta-analysis) and provides a summary estimate. Histograms are used to illustrate frequency distributions of data.