Understanding and interpreting funnel plots for the clinician

Abstract

Funnel plots are an increasingly common graphical tool which are widely used in the literature. They were first introduced by Light and Pillemer in 1984 . In scientific literature, funnel plots are used to identify the probability of bias in meta-analyses and compare institutional performance. The ability to identify variation is better with graphical than tabular display. In addition, the way data are presented can directly influence the interpretation of results. This was demonstrated by Marshall et al (2004) , who presented institutional mortality data in both a league table and control chart format. This study illustrated that when displayed as a league table, a greater number of units were identified for investigation than were actually required. The use of control charts or funnel plots may therefore show benefit in reducing the number of inappropriately labelled outliers. This article explains how clinicians should read and interpret funnel plots, and discusses their considerations and limitations.

Prevalence of chronic kidney disease stages 3–5 in low- and middle-income countries in Asia: A systematic review and meta-analysis

“…The pooled prevalence identified in this study was estimated from studies with moderate to high quality. No discovered publication bias distorted the result of the estimated prevalence of CKD by the funnel plot and Egger's test [59].…”

Orotracheal tube versus supraglottic devices in biological, chemical and radiological disasters: meta-analysis in manikin-based studies

“…Among the reasons, it was observed a variety of healthcare providers as participants in the selected primary studies, as shown in Chart 1. Thus, in order to reduce heterogeneity, a subgroup analysis was performed (29) . Firstly, subgroups were generated according to the setting of activity of the study participants: emergency and surgery room.…”

“…Another important analysis was the creation of subgroups with studies that assessed more than one supraglottic device compared to orotracheal tube, thus conferring a reproductive characteristic of the methodology. The methodological variation of each study is known to have a direct impact on the heterogeneity of data for meta-analyses (29,33) . Low heterogeneity (0 < I2 < 12%) and statistically significant results were obtained in the methodological reproduction subgroups, corroborating the previous analyses.…”

“…Regarding assessment of publication bias, it is known that methodological rigor and quality influence the asymmetry of the funnel plot (29) . For this reason, it should be emphasized that Figure 2 shows a low risk of publication bias among the studies of this review.…”

Impact of new-onset versus pre-existing atrial fibrillation on outcomes after transcatheter aortic valve replacement/implantation

“…A heterogeneity score > 60% was used to investigate the variability of the reported results. Publication bias in the outcomes of included studies was assessed by the construction of funnel plots [12] . The pairwise assessment of the study outcomes guided the analysis of their clinical correlation with pre-AF/NOAF versus sinus rhythm.…”