ABSTRACT
18Defining a hypothesis and determining research questions must be specific and clear about what you really focus in it. Analytical meta-analysis is considered to invade the points of estimation, and exploratory meta-analysis is mainly focused on investigating potential source of heterogeneity and may reveal important effect modifiers. There are two basic approaches to combininge evidence across studies in meta-analysis. One approach involves testing the statistical significance of combined results of the collection of studies. The other approach involves estimating an average treatment effect. The sources of search for literature in meta-analysis include the published literature, unpublished literature, uncompleted research reports, and work in progress. Reliance on only published reports leads to publication bias—the bias resulting from the tendency to publish results that are statistically significant. Given a vast quantity of heterogeneous literature, suitable studies have to be selected for a meta-analysis. Meta-analysis is a two-stage process. In the first stage, the effect sizes are collected from each primary study. Methods of quality assessment provide a systematic approach to describe primary studies and explain heterogeneity. A formal approach to decide the ultimate inclusion criteria of a study may be undertaken by using panel of judges/experts. The internal validity of a study is the extent to which systematic error (bias) is minimized. The external validity is the extent to which the results of the study provide a correct basis for applicability to other circumstances. The first step in meta-analysis is to prepare a master sheet (data-points table). The first column in the master sheet consists of the list of selected studies according to their chronological order of publication. In more complex situations to understand heterogeneity and its sources, several graphs and diagrams such as Forest plot, Funnel plot, etc. have been established to use in meta-analysis. The methods used to pool end-points explain a weighted averages of the end-points in which the larger studies generally have more influence than the smaller ones. The methods are based on the assumptions such as fixed effects and random effects models. It will be advantageous to extend meta-analysis by applying several additional meta-analysis techniques such as sensitivity analysis techniques, influence meta-analysis technique, subgroup meta-analysis technique, and cumulative meta-analysis technique. In reporting meta-analysis, various implications of the results such as research implications, clinical implications, economic implications, and implications for policy making have to be specified.
