**The combination of single trials into one meta-analysis provides a more precise and stable estimate of the treatment efficacy than analysis provided by each single trial. Evidence-based medicine is one of the most efficient ways to prove efficacy in randomized clinical trials. CROS NT discusses why meta-analysis is becoming more relevant in clinical trial data analysis.**

Meta-analysis in clinical trials is becoming more relevant as a statistical methodology as more and more clinical data becomes available. Because a meta-analysis combines several clinical trial data, it increases the sample size and the power to study various effects. Meta-analysis is being used today also as a method of traceability for regulatory submissions.

Meta-analysis is a statistical procedure that combines the results of several independent studies which address the same set of research related hypotheses in order to integrate their findings. In order to effectively perform a meta-analysis, the following steps need to be considered:

- Define an objective
- Define inclusion and exclusion criteria
- Conduct comprehensive research of the relative trials
- Perform a critical assessment of the trials – the Jaded index could be helpful as it is used to assess the methodological quality of the clinical trial
- Pool the results; often by using graphical representation
- Interpret the results with regard to the issue of inter-trial heterogeneity

The following image demonstrates the common example of a graphical representation of a meta-analysis. The vertical line represents the equivalence of treatments. Point estimates of differences on the left show a better result in treated patients as opposed to controlled patients. This result is statisticially significant if the confidence interval does not exceed the equivalence line.

**Validity Criteria for Meta-Analysis**

The validity criteria of a meta-analysis require that the following aspects are met:

- The primary objective should be clearly stated and clinically relevant
- The trial research should be complete and comprehensive
- The inclusion and exclusion criteria for the trials have to be reasonable and coherent with respect to the primary industry
- A sufficient number of trials regarding the treatment of interest should exist and the number of patients involved in each trial should be large enough
- The analyzed trials should be similar in terms of patient characteristics, treatment conditions and end-point follow up

**Interpreting the Results**

The verification of the validity criteria is crucial for meta-analyses. If the analysis involves a small number of trials or a limited number of patients, the results are at risk of being contradicted by a mega-trial. Another risk is the presence of a stronger inter-trial heterogeneity. Graphically, this is easy to detect by the presence of treatment differences to left and right of the equivalence line which becomes more serious when the corresponding confidence intervals are not overlapping.

The most important conclusion of a meta-analysis is the investigation of all potential sources of heterogeneity which is the quantitative summary of the results associated with single clinical trials. It should be noted, however, that meta-analyses are controversial in nature. There is often a publication and search bias when selecting research and trials to be included in the meta-analysis which can be detected using a funnel plot to spot potentially missing data. In terms of heterogeneity, as the level of dissimilarities increases, the justification for an integrated result becomes difficult. Expert biostatisticians are able to draw the line on these dissimilarities before completing the meta-analysis.

**CROS NT and Meta-Analysis**

CROS NT biostatisticians have extensive experience in performing meta-analyses for a variety of therapeutic areas. We also have experience preparing graphical representations of meta-analyses and our Principal Biostatisticians can provide consultancy on implementing meta-analysis.