Meta-analysis made easy: 6 Essential statistical considerations to keep in mind
As the volume of published literature in the biomedical sciences increases, literature syntheses and meta-analyses become increasingly important because they enable researchers and practitioners to quickly and efficiently keep abreast of the overall state of evidence on a particular intervention, clinical outcome, or condition.
What is meta-analysis?
Meta-analysis is a statistical technique used to combine evidence provided by multiple studies, thereby improving the precision of existing knowledge, uncovering answers to questions that are not solved by individual studies, and resolving conflicting scientific claims or data. The Preferred Reporting Items for Systematic reviews and Meta-Analyses (PRISMA) checklist (updated in 2020) is a valuable resource when reporting a meta-analysis in a research manuscript.
The statistical component of a meta-analysis usually involves (1) calculating a summary statistic for each of the studies included and (2) calculating a combined (i.e., summarized) intervention effect estimate. Meta-analyses are usually illustrated using a forest plot, a visual depiction of effect estimates and confidence intervals for both individual studies and the overall meta-analysis.
Statistical Considerations During Meta-Analysis
Because meta-analyses can have a substantial impact on the decisions made by clinicians, researchers, policymakers, etc., they need to be conducted carefully and rigorously. Below, we list some important steps you need to follow while conducting a meta-analysis
1. Research Question
Formulate a strong and precise research question/objective. This allows you to determine what data you need to extract and which variables you will be paying attention to. For example, “interventions for diabetes management” is a very broad-ranging topic; it’s advisable to narrow it down to something like “exercise interventions to improve glycemic control” or “vitamin D supplementation to improve glucose metabolism”.
2. Defining Outcomes
List and define all the outcomes for which you are seeking data. Two studies on the same topic may collect data on different outcomes (e.g., one study on a weight-loss intervention may measure percentage of weight changed while another study on the same intervention may measure change in body mass index).
3. Handling Missing Data
Follow a defined method for handling missing summary statistics, data conversions, etc. These methods include multiple imputation, inverse-probability weighing, and complete-case analysis. You will also need to report the methods you have used in your research manuscript under the Methods section.
4. Heterogeneity
Always calculate and report statistical heterogeneity. Heterogeneity (widely calculated using Cochran’s Q and the I2 statistic) gives your readers an idea about the variation in observed outcomes or intervention effects between studies. A certain amount of heterogeneity is expected between the studies included in your meta-analysis, but it must be quantified and reported.
5. Risk of Bias
Always assess and report risk of bias for each included study. Because the quality of each study included in the meta-analysis may vary, it is important to evaluate every study's design, conduct, and reporting to identify any factors that could introduce bias. Assessing risk of bias allows you to determine the quality of evidence provided by different studies and to judge the reliability of the overall effect size estimate. There numerous tools available to assess risk of bias, such as the Cochrane RoB tool and RoB 2.0 tool, the Effective Practice and Organisation of Care (EPOC) RoB tool, or the NIH quality assessment tools.
6. Sensitivity
Conduct and report the results of sensitivity analysis. Sensitivity analysis can help readers determine how robust the overall findings of your meta-analysis are, by indicating the impact of individual studies on the overall outcomes. Sensitivity analysis can also help you determine which studies are outliers (i.e., those with unusually large or small effects compared to other similar studies).
Looking for expert advice from a biostatistician in conducting your next meta-analysis or research project? Editage’s Statistical Analysis & Review service can help. Book your consultation today.
Comments
You're looking to give wings to your academic career and publication journey. We like that!
Why don't we give you complete access! Create a free account and get unlimited access to all resources & a vibrant researcher community.
Subscribe to Conducting Research
