Association between analytic strategy and estimates of treatment outcomes in meta-analyses
- 1Centre de Recherche Epidémiologie et Statistique, INSERM U1153, Paris, France2Centre d'Épidémiologie Clinique, Hôpital Hôtel Dieu, Assistance Publique des Hôpitaux de Paris, Paris, France3Faculté de Médecine, Université Paris Descartes, Sorbonne Paris Cit.
- 2Centre for Statistics in Medicine, Oxford, United Kingdom.
- 3Centre de Recherche Epidémiologie et Statistique, INSERM U1153, Paris, France5Department of Epidemiology, Mailman School of Public Health, Columbia University, New York, New York.
- 0Centre de Recherche Epidémiologie et Statistique, INSERM U1153, Paris, France2Centre d'Épidémiologie Clinique, Hôpital Hôtel Dieu, Assistance Publique des Hôpitaux de Paris, Paris, France3Faculté de Médecine, Université Paris Descartes, Sorbonne Paris Cit.
Related Experiment Videos
Contact us if these videos are not relevant.
Contact us if these videos are not relevant.
View abstract on PubMed
Summary
This summary is machine-generated.Including all trials in meta-analyses, especially for subjective outcomes, often yields larger treatment effects. This highlights the importance of sensitivity analyses in meta-analysis to ensure robust conclusions.
Area Of Science
- Biostatistics
- Clinical Epidemiology
- Evidence Synthesis
Background
- Meta-analyses are crucial for synthesizing research findings.
- A key challenge is deciding whether to include all available trials or use selective strategies.
Purpose Of The Study
- To compare treatment outcomes from meta-analyses including all trials versus alternative strategies.
- Investigate the impact of different inclusion criteria on meta-analysis results.
Main Methods
- Analyzed 163 meta-analyses (1240 randomized clinical trials) from high-impact journals and the Cochrane Database.
- Compared meta-analysis of all trials against strategies like the single most precise trial, largest trials, limit meta-analysis, and low risk of bias trials.
- Used ratio of odds ratios (ROR) to quantify differences, with ROR > 1 indicating larger outcomes with all-trial inclusion.
Main Results
- Meta-analysis of all trials showed larger treatment outcomes for both subjective (ROR 1.13) and objective (ROR 1.03) outcomes compared to the single most precise trial.
- Substantial differences were observed in 51% of subjective and 39% of objective outcome meta-analyses.
- Varying results were found when comparing all-trial inclusion with other strategies, emphasizing outcome instability.
Conclusions
- Treatment outcome estimations in meta-analyses vary significantly based on the analytic strategy employed.
- This variability can lead to major shifts in conclusions, underscoring the necessity of systematic sensitivity analyses.
Related Experiment Videos
Contact us if these videos are not relevant.
Contact us if these videos are not relevant.