We all are aware of the different types of publication documents, and meta-analysis is one of those documents with the highest level of evidence (Figure). Meta-analysis is a statistical analysis that combines or integrates the results of several independent clinical trials considered by the analyst to be combinable. It usually aims to resolve controversy over true effect, when results of individual studies are variable; and validate a statistically non-significant but clinically important result of small studies.
A meta-analysis usually considers the main outcome of the overall magnitude of the effect. The process of conducting a meta-analysis is often rigorous and well defined which leaves very less opportunities for bias to distort the results. While systematic reviews summarize the medical literature textually, meta-analyses statistically summarize results to obtain overall estimate of treatment effect.
Conducting a Meta-analysis
Over the years, the methodologies involved in conducting meta-analyses have changed. The Cochrane Collaboration has been the most important contributor to streamline and validate the procedures involved in conducting a meta-analysis. Major contributions of the Cochrane Handbook include development of protocols which describe literature search, and analytic and diagnostic methods for evaluating the output of meta-analyses. Additionally, the PreferredReporting Items for Systematic reviews and Meta-analyses(PRISMA) statement provides a more robust procedure in which meta-analyses can be conducted. Steps involved in a meta-analysis include:
A sound literature search is the key to achieving robust results. A clear definition of the hypotheses to be investigated can provide the framework for the overall process to be followed in a meta-analysis. The PRISMA statement recommends inclusion of PICOS (Participants, Interventions, Comparators, Outcomes, and Study Designs) explicitly in the research question. Inclusion of PICOS incorporates all aspects being considered for the selection of studies which further helps in searching for studies with specific information/results. Searching most electronic databases with relevant search terms is important to identify articles. However; identifying appropriate search terms is the first step to achieving this. According to the PRISMA statement, complete search strategy used for at least one electronic database must be reported.
The quality assessment of studies to be included is done by evaluating each study for the eligibility for inclusion, study bias, study quality, and reported findings. Often, two independent reviewers are involved in assessing the study quality of the included studies. This assessment basically provides insights to the degree to which the trial design, conduct, analysis, and presentation have minimized or avoided systematic biases.Several tools are available to assess the study quality of which JADAD, and QUADAS are a few to name.
Data extraction decides the result of the meta-analysis. Important data which requires to be collected includes study design, description of study groups, diagnostic information, treatments, length of follow-up evaluations, and outcome measures. Sometimes, data extraction may pose a challenge when studies use different outcome metrics. In these cases, the data must be converted to a uniform metric for easy pooling.
Measuring inter-study heterogeneity is very important to understand whether the data of the meta‑analysis has addressed the two most important questions:
- What is the overall relationship between the treatment/intervention/ exposure and the health outcomes?
- Is this association consistent across the studies that constitute the systematic review and meta-analysis?
Heterogeneity can be addressed by checking if the data is correct, analyzing variation in results of the study, further exploring heterogeneity by conducting sub-group analysis/ meta‑regression, using analysis procedures which ignore heterogeneity, change the effect measure, and finally exclude the studies which may create conflict.
The data analysis is very complex and involves several analysis techniques. This is usually done using the random effects model or the fixed effect model. The random effects model is used when there is considerable heterogeneity in the studies included while fixed effects model is used when the overall outcome is similar in all studies included. Meta-analyses may also include sensitivity analysis which is a repeat meta-analysis substituting alternative decisions and a meta-regression in which the outcome variable is predicted according to the values of one or more explanatory variables.
Interpretation of the analyzed results must provide answers which are relevant to the context of the current healthcare, state the methodological limitations of studies, consider size of effect in studies and review, their consistency and presence of dose-response relationship, consider interpreting results in context of temporal cumulative meta-analysis, make recommendations that are clear and practical, and finally propose future research age.
In conclusion, conducting a meta-analysis can prove beneficial as it summarizes the overall results in an area of research. However; it must be noted that a single study cannot provide definitive conclusions. In addition, larger randomized controlled trials may sometimes contradict to the results of a meta-analysis. Meta-analysis can summarize the results of studies with varying sample size, diverse populations across different ages which provide an opportunity to explore newer hypotheses. Having said that, meta-analysisstill remains the most important and efficient tool in adding value to the already available evidence. Turacoz Healthcare Solutions (THS) provides guidance in understanding the different attributes of a meta-analysis and its finer details.
Latest posts by Gurpreet Virya (see all)
- Benefits of Yoga in Diabetes - August 22, 2017
- Summing up Results of Research with Meta-Analysis: How it is Done and Why is it Important? - August 10, 2017