A meta-analysis is a method of combining data from multiple independent studies that are investigating the same research question. Essentially, a meta-analysis is a statistical test that assesses whether findings from individual studies remain true across a whole body of research.
By its very nature, scientific research is a collaborative effort. Lots of scientists are investigating closely related questions using different methods. When combined, these efforts help to build a bigger picture of our understanding of biological processes.
However, it is when we analyze multiple findings together that we can make more general statements about how biological systems work and understand how meaningful the data is. This is where the meta-analysis comes in.
Examples of Meta-analysis in Biology
In biology, meta-analyses are most often found in clinical research. For example, testing how effective antipsychotic drugs are in treating schizophrenia, or determining how often two diseases – such as psoriasis and inflammatory bowel disease – are associated with one another.
One study might suggest, for instance, that people are more likely to have psoriasis if they also have inflammatory bowel disease. This data could indicate that there is a relationship between the pathogenesis or genetics of these two diseases in the general population. If ten studies across a wide range of populations all come to the same conclusion, then we are more confident that this finding applies to general populations.
Often, clinical research and clinical trials are carried out on small sample sizes. It can be tricky and expensive to carry out such studies on a large population. Additionally, clinical research is usually less flexible than biological research, so it is much more likely that the parameters between different study designs are similar. These features mean clinical research is particularly well suited to meta-analysis.
Not all biology research is clinical. Scientists working in the field of biology investigate cells, animals, plants, ecology, biochemistry, and many more research areas. Meta-analyses are less common in such ‘basic’ research. This is, in part, a reflection of the fact that basic research is often more exploratory and fluid, with less clear cut outcomes and definitive ‘yes/no’ answers.
However, there are some areas of biology research in which meta-analysis is frequently used. One such case is in genetics, particularly in genome-wide association studies (GWAS). In these studies, scientists are attempting to find variation in the human genome that is associated with specific traits. For example, researchers have identified various genes in which changes to the sequence at a specific region are more common in people who have type I diabetes.
A meta-analysis of several GWAS studies greatly increases the power of the data by vastly increasing the sample size and likelihood of finding rare significant variants. For example, a meta-analysis of four Alzheimer’s Disease GWAS datasets identified an additional 11 genes linked with an increased risk of developing the disorder.
Systematic Review vs. Meta-analysis
The terms systemic review and meta-analysis are sometimes (wrongly!) used interchangeably. While a meta-analysis is a statistical method to combine the data from multiple studies, a systemic review is a procedure in which literature surrounding a particular research question is gathered, evaluated, and summarized in a single report.
A meta-analysis is often an important part of performing a systematic review. A meta-analysis can be easily incorporated into a systematic review when the data being analyzed is numerical, and the studies that are being investigated are similar.
How is a Meta-analysis Performed?
The necessary steps that should be followed when carrying out a meta-analysis are listed below.
- Identify the research question
First, researchers must decide precisely what they want to gain from their meta-analysis. This question should be based on a hypothesis and any pre-existing evidence that exists to support the theory. The research question should be specific and unambiguous but concise. This step is essential to keep the analysis focussed and direct.
- Decide on the methodology for conducting statistical testing
Before commencing any research, the investigators should decide upon and write their methodologies. This step limits potential bias during the research process. The methods should be clearly outlined and transparent, and someone who wanted to repeat the research should be able to follow these methods and attain the same results. If possible, researchers should pre-register their research. This is the gold standard for transparent study design and research process.
- Determine inclusion and exclusion criteria
Before carrying out literature or data searching, researchers should decide upon how they are going to select the study. Some inclusion criteria could be, for example, that the study must be in English, have been performed in the last ten years, be published in a reliable peer-reviewed journal, and study participants must have no illnesses other than the one tested.
- Develop a search procedure
After deciding what studies they will include in their study, researchers must scour the available literature and data to ensure they identify all the studies that meet their criteria. A reliable searching strategy will be transparent and involve searching through multiple databases. This limits the chance that they can be accused of bias in their meta-analysis by excluding an eligible study.
- Evaluate each study
When researchers have a collection of studies for their meta-analysis, they must objectively determine the quality of the findings and determine how reliable the data is. This will involve a pre-defined set of standards, such as the GRADE criteria.
- Perform statistical testing
Now the investigators are ready to perform their statistical tests according to the previously established protocols.
- There are various types of software that can help researchers perform such analysis, such as comprehensive and open-source packages available in R Studio. This could include a variety of features, such as:
- Determining how heterogeneous the data is within and between studies
- Estimating effect sizes reported by each study weighting them according to their sample size
- Evaluating any publication biases
- Analyzing different subsets of data to determine how robust the findings are. For example, separating out studies that studied different age groups or sexes. This is called sensitivity analysis.
- Interpret the findings
Finally, researchers must interpret the findings of their research and effectively summarise them. They should now be able to answer their research question, which should form the basis for their interpretations.
While meta-analyses have many advantages in biology research, there are also some considerations that must be made when considering them.
One limitation is that there are many opportunities for researcher bias. If the steps are above are followed, these biases can be limited. Taking steps to ensure transparency throughout the study design can limit researcher bias.
Another criticism of a meta-analysis is that it is like mixing ‘apples and oranges’ because it combines research carried out by different people on a different population.
Finally, an important consideration when performing – and interpreting data from – a meta-analysis is the potential for publication bias. Journals are much more likely to publish positive results. That is, there are far more reports in the literature showing results such as ‘drug A reduces the symptoms in disease A’ than ‘drug A has no effect on disease A’. This strongly influences the pool of data from which a meta-analysis is performed.