Hi, welcome back to the systematic review posts. In the last blog post, we discussed the risk of bias assessment of included studies in a review (https://scienceiq.blog/2019/04/18/how-to-do-risk-of-bias-assessment-in-a-systematic-review/). Today, we will move one step further and learn how to do the meta-analysis of the included studies. This post is a step by step tutorial so you can follow each step and perform the analysis as you read. Before we begin, I would like to give a brief introduction about meta-analysis and when it should be done.
Meta-analysis is a pooled analysis of all the included studies. While you may be analyzing your single study, in the meta-analysis, you will have to pool the results of all the studies and then analyze to get the final result. Now, you may be wondering why we should do a meta-analysis.In literature, few studies may be against a particular intervention, and few may be for it. Then you do meta-analysis and take a final decision about the effectiveness of an intervention based on your results.
It may sound easy, but there are certain criteria which should be followed before considering meta-analysis. Firstly, you should have a homogeneous data, which means that all the studies should have similar outcomes with similar units of measurement. E.g. For pooling data and analyzing gait speed, all the studies should have measured gait speed. If few have measured gait cadence, few have measured balance, and one study has measured gait speed then you won’t be able to do a meta-analysis for gait speed. Secondly, it is advisable to consult a statistician or your supervisors for deciding which outcomes you want to analyze and how you should go about it. There are innumerable ways to analyze the data depending on your objectives.
I have used Review Manager (RevMan) software for the analysis and attached screenshots for each step. Let us see how:
- Extract all mean values from the included studies in an excel sheet as shown in the picture below.
2. Open RevMan software and choose your review
3. After opening the file, click on the study and references option as shown below and add study under characteristics of included studies
4. Enter all the included studies as shown using first author name and year of publication
5. After adding all the studies, click on add outcome – under data and analyses option
6. Following which, a new comparison wizard opens where you have to click on – add an outcome under new comparison
7. Go to continue and then click on continuous data. You can change it based on the type of your data.
8. Click on next and enter the outcome measure which you are analyzing. Eg. Gait speed
9. Click on next. Depending on the type of analysis you wish to perform, click on either fixed effect or random effect; mean difference or standard mean difference.
10. Go to next and choose your percentage of confidence interval as shown
11. Go to next and label the axis for experimental and control group depending on your outcome measure, whether it is increasing or decreasing.
12. Click on next and click – add study data for the new outcome
13. Continue and choose the studies from which you will add the data. These chosen studies should have the homogeneous data.
14. After adding the studies, click on finish and you will see the following table.
15. Now manually add the post intervention mean and SD values as well as total number of participants for the experimental and control group as shown in the picture
16. You will get the analysis as soon as you enter the data. Click on the forest-plot icon to generate the forest plot
17. This is your final forest plot which you can copy in your text. While reporting, include p-value, confidence interval, heterogeneity and effect size
That’s it! Hope you enjoyed doing the meta-analysis on your own. Feel free to comment if you did not understand any of the steps.