Findings in meta-analysis are integrated by the means of effect sizes also known as the currency of meta-analysis. They span the group of ‘indices’. An index is a statistical measure which is a compound representation of the several observations in the study. It gives a general dimension to the phenomenon. Therefore, indices are used to evaluate the size or magnitude of the treatment effect (Becker 2000).
There is a wide variety of available effect sizes which can be calculated for the treatment data and could be potentially confusing. They have been explored in this article. Understanding the type of study, group and type of data is crucial for conducting meta-analysis, especially in Comprehensive Meta Analysis (CMA). Hence, this article provides an insight into the concept of effect sizes and different types of data which can lead to these effect sizes. The figure below shows the steps for identifying and understanding data for meta-analysis in CMA.
Step 1: Understanding the data
As the figure above shows, data can be of three types:
- rate or computed effect sizes.
This category provides the option of entering data available in the form of dichotomous variables. This means, the data has only two categories, wherein the variables take up discreet quantitative values. The table below shows an example of dichotomous data. Other such examples can be success or failure of treatment or presence or absence of any adverse events. CMA software designates the discreet quantitative data as dichotomous data. Dichotomous data can be used to compare two groups or to analyze the events or means for a single group only.
Table 1: Discreet dichotomous data
The ‘continuous’ category consists of another type of quantitative data. Here, the variables are measured over a scale, and the dataset consists of a range of values. The table below shows an example of continuous data. Such data could be for assessing functional mobility of disabled patients, measures of quality of life or even evaluating psychiatric symptoms of patients. Continuous data values can be statistically analyzed to obtain mean and standard deviation values for the group of patients studied. The aim is to obtain the effect size. The table below shows a sample set of continuous data.
Table 2 : Continuous data consisting of mean, standard deviation and sample size
CMA also provides the user with the option to enter the data available as correlations and rates. In correlation, the user can enter the data pertaining to correlation, and different measures such as Fisher’s Z, t-value, p-value.
This category allows the user to enter data available for the rates of occurrence of events. Also, the software allows to enter raw data and data for computed effect sizes.
Both the dichotomous and continuous data constitute ‘outcomes’, also referred to as events or endpoints in clinical studies. Analysis of these outcomes helps assess the impact of an intervention or exposure on the study group’s health. These outcomes can be classified as ‘primary’ or ‘secondary’ outcomes. The primary outcomes are those which hold the highest degree of relevance in answering the issue at hand. They pertain to the primary research question of the study. On the other hand, the secondary outcomes prove to be important in interpretation of the results obtained from primary outcomes.
Step 2: Selecting appropriate data-entry formats for meta-analysis
CMA software provides users with a list of multiple data entry formats, where the selection of appropriate format is important. This selection depends upon three factors which influence the selection decision i.e. study design, data type and groups. The figure below shows the steps to for the selection process.
The section below provides a detailed explanation to each of these steps.
Selecting appropriate types of study
Clinical research covers a wide range of studies using different research designs, depending upon the research problem and objectives. The figure below gives a review of these types of studies. Among all the study designs, Randomised Control Trials (RCT) are the most common and significant in establishing the evidentiary value of an intervention or exposure. Hence, a large number of meta-analyses involve the assessment of effect sizes based on results of RCT. However, it should be noted that the data from other study designs which lack a control group or are observational in nature also contribute to scientific literature and could be used for meta-analysis.
Types of study groups
Different studies include different types of study groups, selected by the researcher on the basis of study design. There are mainly three types of groups:
- matched groups,
- unmatched groups and
- single group studies.
The figure below shows these groups.
Determining type of data
In discussing types of data, it is essential to differentiate it into two broad categories:
- categorical or classification data and
- quantitative data.
Both the data types involve the recording of information with respect to certain variables. A ‘variable’ refers to specific characteristic of a subject that could assume one or more different values. With respect to the present discussion, the data entry options include dichotomous or continuous data.
Step 3: Obtaining effect sizes
Besides the primary and secondary outcome data of raw nature (non-standardized), some studies may also report computed effect sizes. It is the pre-estimated size of particular treatment or exposure effect as evaluated by the specific studies’ researcher. For instance, female subjects having an odds ratio of 12.32, with respect to prevalence of Post Traumatic Disorder (PTSD) among older adolescents. These pre-calculated effect sizes can also be used for performing meta-analysis. The nature of data reported by the study determines the type of effect size being calculated for the assessment of different studies and generation of an overall summary effect size.
The present article establishes the procedure for conducting meta-analysis in the CMA software. The next article discusses different data entry formats in CMA software.
- Becker, L.A., 2000. Effect Size. , pp.155–159. Available at: http://www.uv.es/~friasnav/EffectSizeBecker.pdf [Accessed January 1, 2017].
- Ip, S. et al., 2013. Role of single group studies in agency for healthcare research and quality comparative effectiveness reviews.
- Kumar, R., Khan, A.M. & Chatterjee, P., 2014. Types of observational studies in medical research. Astrocyte, 1, pp.154–159.