1. Missing outcome data in evidence synthesis with aggregate data
| In a systematic review with aggregate data, missing data may refer to missing statistics (e.g., standard deviation), missing studies or outcomes (publication bias, selective reporting) and missing outcome data (not having observed the outcome for all participants randomized to a study). In this lecture, we focused on missing outcome data, and presented the available methods in the literature for treating this problem at the meta-analysis level. We also presented a software for using these methods. | Presenters: Dimitris Mavridis, Associate professor in Statistics, Department of Primary Education, University of Ioannina, Ioannina, Greece Christos Christogiannis, PhD student, Department of Primary Education, University of Ioannina, Ioannina, Greece
|
Video lecture: |
2. Updating meta-analyses
| Systematic reviews and meta-analyse should be updated frequently if they are to remain up-to-date and scientifically relevant. However, repeated updating of meta-analyses can lead to problems with interpreting their results, particularly where there are few trials or substantial heterogeneity. In this lecture we examined some of the problems that can arise when updating meta-analyses, illustrated with data from Cochrane reviews. We presented some newer statistical methods that can avoid some problems when updating analyses, and discussed Cochrane’s current position on how to interpret findings as meta-analyses are updated. | Presenter: Mark Simmonds is a Senior Research Fellow at the Centre for Reviews and Dissemination in York, and a co-convenor of the Cochrane Statistical Methods Group. He has expertise in the statistical methodology of meta-analysis, covering a range of areas including individual participant data analysis, diagnostic test analysis, survival analysis and living reviews. His particular focus is on statistical methods for robustly updating and maintaining meta-analyses.
|
Video lecture: |
3. Including variants of randomised trials in meta-analysis
| Review authors frequently include variants of randomised trials such as crossover, cluster, factorial and multi-arm trials. In this lecture, we covered issues that arise in the analysis of these designs, the ramifications for meta-analysis, and discussed the biases particular to these designs. The lecture covered the content from Chapter 23 of the Cochrane Handbook for Systematic Reviews of Interventions. | Presenter: Joanne (Jo) McKenzie (Professor) is head of the Methods in Evidence Synthesis Unit within the School of Public Health and Preventive Medicine at Monash University, Melbourne, Australia. She is Co-convenor of the Cochrane Statistical Methods Group and co-chair of the Methods Executive.
|
Video lecture: |
4. Meta-analysis of time-to-event data
| Time-to-event data, such as time to death or time to first seizure, are commonly used outcomes in medical research to summarise the length of time taken for a particular event of interest to occur. Methods for analysis of time-to-event data need to account for both the time to event and any related but unobserved events. Meta-analysis of time-to-event outcomes can be undertaken using data extracted from trial reports using a variety of different methods. These methods are discussed in this lecture. | Presenter: Catrin Tudur Smith, Professor of Medical Statistics, Department of Health Data Science, University of Liverpool, UK; Co-convenor of Cochrane Statistical Methods Group.
|
Video lecture: |
5. Random-effects meta-analysis methods in RevMan
| Many alternative methods have been developed for calculating the summary effect, heterogeneity variance and their allied confidence intervals in a meta-analysis. The performance of the methods may vary depending on the characteristics of the meta-analysis (e.g., number and size of the included studies). Currently, the DerSimonian and Laird method is the only random-effects model available in Review Manager (RevMan). In this lecture, we discussed random-effects methods recommended by the Cochrane Statistical Methods Group to be implemented in RevMan. In particular, we focused on recommendations for the heterogeneity variance estimators and the confidence interval methods for the summary effect. | Presenter: Areti Angeliki (Argie) Veroniki, Co-convenor of the Cochrane Statistical Methods Group; Scientist, St. Michael’s Hospital, Unity Health Toronto; Assistant Professor, Institute of Health Policy, Management, and Evaluation, University of Toronto, Canada.
|
Video lecture: |
