Series: Living Systematic ReviewLiving systematic reviews: 3. Statistical methods for updating meta-analyses
Section snippets
Background
The key intention of a living systematic review (LSR, see Box 1), which differentiates it from a standard systematic review, is that it will be updated frequently, ideally as soon as any new relevant study is published or identified [1], [2], [3]. Over time the information available to be included may increase, requiring the review to be updated to ensure it is presenting the best available evidence. In many updates, this will require updating one or more of the meta-analyses included in the
Analysis methods for repeated meta-analyses
Updating a meta-analysis has some similarities with interim analyses of clinical trials [9], [10], [11]. Interim analyses are often performed in trials so the trial can be stopped early if there is convincing evidence that the intervention is beneficial or harmful. Methods have been developed to avoid type I and II errors and produce robust conclusions for these trial sequential analyses. These methods have been adapted for the analysis of repeated meta-analyses and more recently for the
Methods for network meta-analysis
A multivariate extension of the alpha-spending boundaries method has been proposed for updating network meta-analysis under the assumption of consistency [25]. Despite the computational complexity in the presence of multiple interventions, the approach is essentially the same as in pairwise meta-analysis. Relative treatment effects between the compared treatments need to be set so as to satisfy the consistency assumptions. Then successively, monitoring boundaries for a predefined level of power
Commentary on the methods
The key properties of each method are outlined in Table 1. Most of the methods for handling repeated meta-analysis are based on an analogy between repeating meta-analysis and sequential analysis of a single clinical trial. While this analogy is generally reasonable, it has some limitations because meta-analyses are based on multiple studies and are not a single controlled trial. Heterogeneity between studies is an obvious key difference. In all methods, if a random-effects meta-analysis is
Conclusions and recommendations
The aim of an LSR is to provide the best available evidence to support decision-making by updating frequently, potentially as soon as a single relevant new study is identified. As with conventional approaches to updating, it is to be expected that the findings of the meta-analyses may change between updates and so reviewers should be suitably cautious when drawing conclusions from a meta-analysis in an LSR, particularly when considering if a result is statistically significant.
The methods
Acknowledgments
The authors would like to thank the members of the Living Systematic Review Network for their comments on drafts of this paper, particularly Philippe Ravaud, Andrew Maas, Kurinichi Gurusamy, Laura Martinez, Joerg Meerpohl and Stefania Mondello.
References (31)
- et al.
Living systematic reviews: 1. Introduction—the why, what, when and how
J Clin Epidemiol
(2017) - et al.
Living systematic reviews: 2. Combining human and machine effort
J Clin Epidemiol
(2017) - et al.
Recursive cumulative meta-analysis: a diagnostic for the evolution of total randomized evidence from group and individual patient data
J Clin Epidemiol
(1999) - et al.
Living systematic reviews: 4. Living guideline recommendations
J Clin Epidemiol
(2017) - et al.
Updating meta-analyses leads to larger type I errors than publication bias
J Clin Epidemiol
(2009) - et al.
Cumulating evidence from randomized trials: utilizing sequential monitoring boundaries for cumulative meta-analysis
Control Clin Trials
(1997) - et al.
Trial sequential analysis may establish when firm evidence is reached in cumulative meta-analysis
J Clin Epidemiol
(2008) - et al.
Graphical augmentations to the funnel plot assess the impact of additional evidence on a meta-analysis
J Clin Epidemiol
(2012) - et al.
Uncertainty of the time of first significance in random effects cumulative meta-analysis
Control Clin Trials
(1996) - et al.
Living systematic reviews: an emerging opportunity to narrow the evidence-practice gap
PLoS Med
(2017)
Evidence-based sample size calculations based upon updated meta-analysis
Stat Med
The impact of study size on meta-analyses: examination of underpowered studies in Cochrane reviews
PLoS One
Discrete sequential boundaries for clinical-trials
Biometrika
A unified theory for sequential clinical trials
Stat Med
Apparently conclusive meta-analyses may be inconclusive–trial sequential analysis adjustment of random error risk due to repetitive testing of accumulating data in apparently conclusive neonatal meta-analyses
Int J Epidemiol
Cited by (100)
Living systematic reviews: aims and standards
2024, Clinical Microbiology and InfectionMethods for living guidelines: early guidance based on practical experience. Paper 1: Introduction
2023, Journal of Clinical EpidemiologyFrom standard systematic reviews to living systematic reviews
2023, Zeitschrift fur Evidenz, Fortbildung und Qualitat im GesundheitswesenEditorial
2023, Zeitschrift fur Evidenz, Fortbildung und Qualitat im GesundheitswesenMethods for developing and reporting living evidence synthesis
2022, Journal of Clinical EpidemiologyCitation Excerpt :When ordinary volumes are anticipated [7], authors suggest integrating evidence every 6–12 months or when new evidence has the potential to impact review conclusions [9,11,19]. Simmonds et al. [3] reported the most appropriate statistical approach for integrating new evidence into an updated MA. Authors distinguish between the “continued integration” of new evidence, expected when the LE synthesis is aimed to inform health decision-making, and the “periodic integration” of evidence conducted to provide an up-to-date evidence summary.
Funding: All the authors of this paper were funded to produce this research by a grant from the Cochrane Methods Innovation Fund. The Living Systematic Review Network is supported by funding from Cochrane and the Australian National Health and Medical Research Council (Partnership Project grant APP1114605). Georgia Salanti is supported by a Marie Skłodowska-Curie fellowship (MSCA-IF-703254).