Article type
Year
Abstract
Background:
Meta-analysis and network meta-analyses are the methods of choice in systematic reviews to summarize the effect estimates of the included studies. Frequently, random-effects meta-analyses and network meta-analyses are applied, which require the estimation of the heterogeneity parameter tau. However, in the case of very few studies, the heterogeneity parameter cannot be reliably estimated leading to broad confidence intervals (Bender et al., 2018). In such situations, the application of Bayesian methods with informative prior distributions is an option (Friede et al., 2017; Bender et al., 2018). Different choices for prior distributions for tau are possible according to several proposals given in the literature (e.g., Turner et al., 2015; Friede et al., 2017; Rhodes et al., 2018).
Objectives:
The goal of the talk is to explore the empirical distribution of tau from IQWiG reports in order to inform future Bayesian (network) meta-analysis in the case of few studies.
Methods:
We collected all published meta-analyses from IQWiG reports for the years 2005 to 2019 and recalculated the estimates of tau by applying random-effects meta-analyses and the Paule-Mandel method. Sensitivity and subgroup analyses were not taken into account. We used the effect measures as used in the original meta-analysis. In the case of binary data, we calculated the risk ratio and the odds ratio for the same data. We summarized the empirical distributions of tau in various settings (comparison, endpoint category, effect measure) and compared these distributions with the proposals for prior distributions in the literature.
Results:
Different empirical distributions of tau can be derived from IQWiG reports in various settings. Descriptive analyses of the various distributions will be reported at the Colloquium.
Conclusions:
It should be discussed in which situations prior distributions for Bayesian meta-analyses and network meta-analyses in the framework of health technology assessment can be derived from the empirical distributions of tau from IQWiG reports in various settings.
Patient or healthcare consumer involvement:
Not applicable
References
Bender, R. et Al. (2018): Methods for evidence synthesis in the case of very few studies. Res. Syn. Methods 9, 382–392.
Friede, T. et al. (2017): Meta-analysis of few small studies in orphan diseases. Res. Syn. Methods 8, 79-91.
Rhodes, K.M. et al. (2018): Between-trial heterogeneity in meta-analyses may be partially explained by reported design characteristics. J. Clin. Epidemiol. 95, 45-54.
Turner, R.M. et al. (2015): Predictive distributions for between-study heterogeneity and simple methods for their application in Bayesian meta-analysis. Stat. Med. 34, 984-998.
Meta-analysis and network meta-analyses are the methods of choice in systematic reviews to summarize the effect estimates of the included studies. Frequently, random-effects meta-analyses and network meta-analyses are applied, which require the estimation of the heterogeneity parameter tau. However, in the case of very few studies, the heterogeneity parameter cannot be reliably estimated leading to broad confidence intervals (Bender et al., 2018). In such situations, the application of Bayesian methods with informative prior distributions is an option (Friede et al., 2017; Bender et al., 2018). Different choices for prior distributions for tau are possible according to several proposals given in the literature (e.g., Turner et al., 2015; Friede et al., 2017; Rhodes et al., 2018).
Objectives:
The goal of the talk is to explore the empirical distribution of tau from IQWiG reports in order to inform future Bayesian (network) meta-analysis in the case of few studies.
Methods:
We collected all published meta-analyses from IQWiG reports for the years 2005 to 2019 and recalculated the estimates of tau by applying random-effects meta-analyses and the Paule-Mandel method. Sensitivity and subgroup analyses were not taken into account. We used the effect measures as used in the original meta-analysis. In the case of binary data, we calculated the risk ratio and the odds ratio for the same data. We summarized the empirical distributions of tau in various settings (comparison, endpoint category, effect measure) and compared these distributions with the proposals for prior distributions in the literature.
Results:
Different empirical distributions of tau can be derived from IQWiG reports in various settings. Descriptive analyses of the various distributions will be reported at the Colloquium.
Conclusions:
It should be discussed in which situations prior distributions for Bayesian meta-analyses and network meta-analyses in the framework of health technology assessment can be derived from the empirical distributions of tau from IQWiG reports in various settings.
Patient or healthcare consumer involvement:
Not applicable
References
Bender, R. et Al. (2018): Methods for evidence synthesis in the case of very few studies. Res. Syn. Methods 9, 382–392.
Friede, T. et al. (2017): Meta-analysis of few small studies in orphan diseases. Res. Syn. Methods 8, 79-91.
Rhodes, K.M. et al. (2018): Between-trial heterogeneity in meta-analyses may be partially explained by reported design characteristics. J. Clin. Epidemiol. 95, 45-54.
Turner, R.M. et al. (2015): Predictive distributions for between-study heterogeneity and simple methods for their application in Bayesian meta-analysis. Stat. Med. 34, 984-998.