Complete Systematic Reviews and Meta-Analyses Faster

The video shows a real review, step by step. The narration is spoken by a synthetic voice; this is its full text.

1. Step one: define the question. Type the population, the intervention, the comparison and the outcome. Here that is adults undergoing cardiac surgery, colchicine versus placebo, for postoperative atrial fibrillation. Then generate the search strategy, and start the search.
2. Step two: search. The site queries PubMed, Europe PMC, Cochrane and ClinicalTrials dot gov at once, removes duplicates, and shows the total for every database, so every record is accounted for.
3. Step three: screen. Filter by study type, for example randomised trials, read the titles and abstracts, and click Include on the studies you want. Here, four colchicine trials are included. Then continue to data extraction.
4. Step four: extract. Each included study becomes a row. Extract All Visible reads each article, and fills in the country, design, population, sample size, and the number of events in each arm, whenever the text reports them. Every value has to be quoted from the paper, and nothing is invented. If a full text is not free, you can attach the PDF yourself.
5. Here are the extracted event counts for each arm, next to the arm sizes.
6. Step five: choose the primary outcome that the analysis should use, and appraise the risk of bias. Here I fill in the Newcastle-Ottawa scale with example scores for each study. In your own review, enter your own assessment, and use the tool that suits your study designs.
7. Step six: the RevMan workbench. Open the table to see the counts for the chosen outcome, with the risk ratio and confidence interval for each study. You can export the combined table as a CSV file for RevMan.
8. Step seven: run the analysis. A random-effects meta-analysis pools the studies. The forest plot shows each trial's risk ratio, and the diamond is the pooled result with its confidence interval. When heterogeneity is high, interpret it with care. The funnel plot helps check for small-study effects.
9. The results text is generated from your data. Regenerate all to refresh it after any change, and edit any paragraph by hand.
10. The discussion is drafted from the same results, so its wording always matches the numbers.
11. Step eight: the manuscript. The site assembles a full paper from your data, with a title, an abstract, methods and results. Anything the site cannot know is marked in square brackets, so you complete those details yourself.
12. The results report the PRISMA numbers, the included studies and the pooled estimate. Then download the manuscript as Word, PDF, HTML or a journal-style version, together with the reporting checklist. Always verify extracted values against the papers before you publish.