Contents
Academy · Research Data, Analytics & AI · Step-by-Step Construction
Construction of Funnel Plots in Meta-Analysis: A Stepwise Procedure
Constructing a funnel plot follows a defined sequence: selection of one effect measure for every study, calculation of the standard error of each study, plotting of effect size against that standard error, and addition of a pooled-effect reference line. A formal asymmetry test is considered only after these steps, and only if sufficient studies are available.
Constructing a funnel plot is a specific sequence of methodological decisions, not a single chart-type selection in a software menu. Each decision (which effect measure to calculate, which precision statistic to plot against it, whether to add reference lines, and whether a formal test is appropriate given the number of studies) affects whether the resulting plot can support any interpretation at all. Egger and colleagues' foundational 1997 paper and the methodological literature that followed it describe this sequence in enough procedural detail that it can be followed as a build process, from extracting each study's data through to deciding what, if anything, the finished plot can responsibly be used to say.
Learning objectives
- The sequence of steps involved in constructing a funnel plot from extracted study data
- Why the effect measure must be selected and applied consistently before any plotting
- The point in the construction sequence at which a formal asymmetry test becomes appropriate
Common misconceptions
- “Constructing a funnel plot requires only the selection of a chart type in meta-analysis software.” The methodological work of Egger and Sterne indicates that the validity of the plot depends on decisions made before any chart is drawn, namely the selection of one consistent effect measure, correct calculation of the standard error for that measure and confirmation of an adequate number of studies, rather than on the default settings of the software.
Step one: selection of a single effect measure for all studies
The first construction step is selecting a single effect measure (a mean difference, a risk ratio, an odds ratio or a standardised mean difference) and applying that same measure to every study being plotted, since mixing measures across studies undermines the mathematical relationship a funnel plot depends on1. For ratio measures such as odds ratios, this step also typically requires a log transformation before plotting, because untransformed ratio measures are not symmetrically distributed around the true effect, while their logarithms generally are, a detail Egger and Sterne's 2001 analysis identifies as essential to obtaining a symmetrical plot under the no-bias scenario2.
This step also requires deciding how each study's outcome will be coded if a study reported its results in a form that does not map directly onto the chosen effect measure, for example a study reporting only a p-value and sample size rather than a usable effect estimate and standard error. Lau and colleagues' 2006 editorial found that inconsistent outcome coding across studies was one of several technical problems undermining published funnel plots, which makes this an area where careful, documented decisions at the extraction stage matter as much as anything done later in the construction sequence6.
Step two: calculation of precision and plotting against effect size
Once a consistent effect measure has been selected, the second step is calculating each study's standard error for that specific measure, which Egger and Sterne's analysis establishes as the preferred vertical-axis variable over raw sample size, because standard error maintains a consistent mathematical relationship to the effect estimate across different measures and study designs2. For a mean difference, this typically involves the pooled standard deviation and the two group sizes; for a log odds ratio, it typically involves the cell counts of a two-by-two table. Meta-analysis software packages generally calculate this figure automatically once raw data are entered correctly, but the underlying formula differs by effect measure, which is partly why the first step, choosing the measure, must precede it.
The plot itself is then built with each study's effect estimate on the horizontal axis and its standard error on the vertical axis, conventionally inverted so that more precise, lower-standard-error studies appear near the top of the chart and less precise studies appear toward the bottom, producing the recognisable inverted-funnel orientation3. Each study becomes a single point on this scatter, and the resulting shape, before any further additions, is what the rest of the construction process will annotate and, where appropriate, test formally.
Step three: addition of reference lines and selection of a formal test
The next construction step adds a vertical reference line at the pooled effect estimate from the meta-analysis itself, giving the plot a clear visual centre against which symmetry can be judged, and, where the standard-error axis has been used correctly, diagonal lines marking a 95% pseudo-confidence interval around that pooled estimate2. Peters and colleagues' 2006 JAMA paper describes a further, increasingly standard addition at this stage: shading or contour lines marking conventional statistical-significance thresholds, which help distinguish a gap plausibly caused by missing non-significant studies from a gap with some other explanation7.
Only after these visual elements are in place does the construction sequence reach the question of a formal statistical test. Sterne and colleagues' 2011 BMJ guidance specifies that Egger's regression-based test is generally appropriate for continuous outcomes with at least around ten included studies, while dichotomous outcomes may call for the Harbord, Peters, or Rücker arcsine tests depending on the amount of between-study heterogeneity present, and that testing strategy should be decided in advance and reported in full regardless of the result3. Duval and Tweedie's trim-and-fill method can be layered on top of this sequence as a further, clearly labelled step, imputing a number of hypothetically missing studies and showing how the pooled estimate would shift if they were included, which Begg and Mazumdar's and Sterne, Gavaghan, and Egger's separate analyses of test power both suggest should be read as a sensitivity check rather than a corrected final result, given the well-documented power limitations of these methods in meta-analyses with few studies548. The Cochrane Handbook's chapter on reporting biases presents this same sequence (measure, precision, plot, reference lines, test) as the standard workflow review authors should follow and document in full9.
Documenting the construction for subsequent verification
A construction sequence is only as useful as the record left behind of how it was followed, which is why PRISMA 2020's reporting checklist requires systematic reviews to describe, in the methods section itself, exactly how any assessment of bias from missing results was carried out, rather than leaving readers to infer the construction choices from the finished figure alone10. In practice, this means recording the effect measure and any transformation used, the formula applied to calculate standard error, the software or package used to generate the plot, and the specific asymmetry test selected along with the reasoning for choosing it, ideally in a supplementary methods note rather than only in the software's output file.
This documentation step matters because several of the construction decisions described above (the effect-measure transformation, the precision formula, the test choice relative to study count) are precisely the decisions later found to have been handled inconsistently across published reviews. Lau and colleagues' audit of funnel-plot use in BMJ systematic reviews found technical inconsistencies of this kind across a substantial share of the reviews that used a funnel plot at all, which is a strong practical argument for treating the construction sequence as something to document step by step rather than as a single chart-generation action accepted without verification6. A review that documents its construction choices in full also gives a future reader, or a future update of the same review, a clear basis for reproducing or re-checking the plot years later.
Documentation also matters for a more immediate, practical reason: it enables a peer reviewer or co-author to detect a construction error before publication rather than after. A methods note that states the chosen effect measure, the standard-error formula, and the asymmetry test used, with its minimum study-count threshold checked, gives a second reader everything needed to verify each step independently, rather than having to reverse-engineer the construction sequence from the finished chart alone. Given how many of the documented problems with funnel plots in the published literature trace back to exactly these construction-stage decisions rather than to the underlying data, this final documentation step is not a formality; it renders every earlier step verifiable. Teams that build this record into their standard systematic-review workflow, rather than reconstructing it retrospectively at a reviewer’s request, generally find that the additional effort is modest relative to the time saved if the plot’s methodology is questioned during peer review or a later update of the same review. The same record also makes it considerably easier to extend or re-run the analysis when new studies become available, which is most important when a review is updated years later by a different team that has only the written record, rather than the original analyst’s memory, on which to rely.
Key points
Key points
A trustworthy funnel plot is constructed in a fixed sequence: one consistent effect measure across all studies, correctly calculated standard error on the vertical axis, a pooled-effect reference line and, only then, a formal asymmetry test selected according to the outcome type and number of studies. Each step should be documented rather than left to software defaults.
Practical application
Practical application
- The effect measure and any transformation should be fixed before data extraction. The choice of mean difference, risk ratio or odds ratio (log-transformed) should be made in advance so that every study is coded consistently from the outset.
- Authors should verify that the standard-error calculation in their software corresponds to the chosen effect measure. Different effect measures require different standard-error formulas; this should be checked rather than a generic default assumed to be correct.
- The strategy for asymmetry testing should be decided before the plot is constructed, not after its shape is observed. Pre-specifying the test to be used, and the minimum number of studies required, avoids selecting a test on the basis of the expected result.
Frequently asked questions
The general sequence (effect measure, standard error, plot, reference lines) is consistent across major packages, but defaults for axis choice and available tests vary, so a specific package's documentation should be checked against current methodological guidance.
Egger and Sterne's guidance treats a consistent effect measure as a precondition for a meaningful funnel plot; studies that genuinely cannot be converted to a common scale generally should not be plotted together on the same chart.
It is the final step only when enough studies are included and heterogeneity is accounted for; Sterne and colleagues' 2011 guidance recommends against adding a formal test at all below roughly ten studies, in which case the plot itself, with its reference lines, remains the end product.
References
- Egger M, Davey Smith G, Schneider M, Minder C. Bias in meta-analysis detected by a simple, graphical test. BMJ. 1997;315(7109):629-634. read 10 October 2026. View on The BMJ
- Egger M, Sterne JAC. Funnel plots for detecting bias in meta-analysis: guidelines on choice of axis. Journal of Clinical Epidemiology. 2001;54(10):1046-1055. read 10 October 2026. View on Journal of Clinical Epidemiology
- Sterne JAC, Sutton AJ, Ioannidis JPA, et al. Recommendations for examining and interpreting funnel plot asymmetry in meta-analyses of randomised controlled trials. BMJ. 2011;342:d4002. read 10 October 2026. View on The BMJ
- Sterne JAC, Gavaghan D, Egger M. Publication and related bias in meta-analysis: power of statistical tests and prevalence in the literature. Journal of Clinical Epidemiology. 2000;53(11):1119-1129. read 10 October 2026. View on Journal of Clinical Epidemiology
- Duval S, Tweedie R. Trim and fill: a simple funnel-plot-based method of testing and adjusting for publication bias in meta-analysis. Biometrics. 2000;56(2):455-463. read 10 October 2026. View on Biometrics
- Lau J, Ioannidis JPA, Terrin N, Schmid CH, Olkin I. The case of the misleading funnel plot. BMJ. 2006;333(7568):597-600. read 10 October 2026. View on The BMJ
- Peters JL, Sutton AJ, Jones DR, Abrams KR, Rushton L. Comparison of two methods to detect publication bias in meta-analysis. JAMA. 2006;295(6):676-680. read 10 October 2026. View on JAMA
- Begg CB, Mazumdar M. Operating characteristics of a rank correlation test for publication bias. Biometrics. 1994;50(4):1088-1101. read 10 October 2026. View on PubMed
- Sterne JAC, Egger M, Moher D (editors). Chapter 10: Addressing reporting biases. In: Higgins JPT, Green S, editors. Cochrane Handbook for Systematic Reviews of Interventions, Version 5.1.0. The Cochrane Collaboration; 2011. read 10 October 2026. View on Cochrane Training
- Page MJ, McKenzie JE, Bossuyt PM, et al. The PRISMA 2020 statement: an updated guideline for reporting systematic reviews. BMJ. 2021;372:n71. read 10 October 2026. View on The BMJ
Every source above is a peer-reviewed article, a publisher’s record of one, or a standard issued by the body that owns it. Each carries the date it was last read. Where a standard moves, this article is revised rather than left up with a note.
How to cite this article
Pick your referencing style and copy it. Your university or target journal may vary the punctuation — check its guide if it has one.
Pubrica Academy. (2026, October 10). Construction of Funnel Plots in Meta-Analysis: A Stepwise Procedure. Pubrica. https://pubrica.com/academy/funnel-chart-construction-in-meta-analysis/
Pubrica Academy. “Construction of Funnel Plots in Meta-Analysis: A Stepwise Procedure.” Pubrica, 10 Oct. 2026, pubrica.com/academy/funnel-chart-construction-in-meta-analysis/.
Pubrica Academy (2026) Construction of Funnel Plots in Meta-Analysis: A Stepwise Procedure. Pubrica. Available at: https://pubrica.com/academy/funnel-chart-construction-in-meta-analysis/ (Accessed: 15 October 2026).
Pubrica Academy. “Construction of Funnel Plots in Meta-Analysis: A Stepwise Procedure.” Pubrica. October 10, 2026. https://pubrica.com/academy/funnel-chart-construction-in-meta-analysis/
Pubrica Academy, ‘Construction of Funnel Plots in Meta-Analysis: A Stepwise Procedure’ (Pubrica, 10 October 2026) <https://pubrica.com/academy/funnel-chart-construction-in-meta-analysis/> accessed 15 October 2026.
Pubrica Academy. Construction of Funnel Plots in Meta-Analysis: A Stepwise Procedure. [Internet]. Pubrica; 2026 Oct 10 [cited 2026 Oct 15]. Available from: https://pubrica.com/academy/funnel-chart-construction-in-meta-analysis/
Pubrica Academy, “Construction of Funnel Plots in Meta-Analysis: A Stepwise Procedure,” Pubrica, Oct. 10, 2026. [Online]. Available: https://pubrica.com/academy/funnel-chart-construction-in-meta-analysis/
@misc{/la_funnel_chart_construction_meta_analysis_2026,
author = {{Pubrica Academy}},
title = {Construction of Funnel Plots in Meta-Analysis: A Stepwise Procedure},
howpublished = {Pubrica},
year = {2026},
month = {oct},
url = {https://pubrica.com/academy/funnel-chart-construction-in-meta-analysis/},
note = {Accessed: <span data-cite-acc="iso">2026-10-15</span>}
}This article is written to be learned from. Cite it if you are referring to what it says; if you are using the evidence behind it, cite the sources listed above instead.
Funnel Plot Construction Sequence
This order should be followed when a funnel plot is built from extracted study data.
[ ] Chose one effect measure and applied it (with log transformation if a ratio measure) to every study [ ] Calculated standard error correctly for that specific effect measure [ ] Plotted effect size against standard error with the precision axis inverted [ ] Added a pooled-effect reference line and pseudo-95% confidence interval lines [ ] Added contour/significance shading if appropriate [ ] Decided on a formal asymmetry test only if ~10+ studies are included, matched to outcome type
Pubrica AcademyWritten and maintained by Pubrica’s subject matter experts. Articles are revised when the underlying standards change, and each article states the date on which it was last reviewed.
How Pubrica can help
Pubrica’s meta-analysis team follows this full construction sequence for every funnel plot it prepares, documenting each methodological decision so that the finished chart and any accompanying test are prepared for scrutiny in peer review.