Continuous data entry in CMA involving comparison between groups
The previous articles discussed two different ways of meta-analyzing the Cholecystitis and Gallbladder Cancer problem using two different effect size data formats. Acquaintance with different methods of entering data in the software makes the task simpler. The manual data entry included entering the raw data for the dichotomous category for the study investigating one group. On the other hand the demonstration for data loading used the computed effect sizes data for the dichotomous category for the study investigating 2 groups. The present article explains how to perform a meta-analysis on the dataset classified as continuous data, including the comparison of two groups. For the same, the second case study will focus on the treatment of Pediatric Obsessive- Compulsive Disorder (OCD).
Case study 2: Pediatric Obsessive- Compulsive Order (OCD)
OCD is an anxiety disorder showing recurring and persistent, often distressing, compulsions or obsessions, which significantly interfere with normal functioning. OCD is one of the most common psychiatric illnesses occurring in children, affecting 1-3% of children and adolescents globally. Initially a psychiatric condition, the disease now shows an extreme refractory tendency to treatment. Thus, a wide range of research conducted over the time has brought into light different methods of treating the condition, such as cognitive behavioral therapy, hypnotherapy, bibliotherapy, pharmacotherapy and others. The present case study focuses on evaluating the efficacy of pharmacotherapy on childhood OCD as demonstrated by different researchers.
Inclusion and exclusion criteria
Data Collection Procedure
For the purpose of completing the systematic review for the present case study, the NCBI, EMBASE, PSYCHINFO databases were searched, using the keywords ‘Cholecystitis’, ‘gallbladder cancer’, ‘gallbladder stones’, ‘gallstones’ and ‘gallbladder carcinoma’.
The data extraction procedures lead to the collection of two types of data:
- Data pertaining to the mean and standard deviation (continuous data), for unmatched groups, (treatment and placebo groups), and
- Data pertaining to the mean and standard deviation (continuous data), for patients receiving the pharmacotherapy treatment.
The present article will introduce the user to different formats for entering the two group data, to compare the primary outcomes. This comparison is on the basis of the type of treatment. The two groups include the pharmacotherapy treatment group and the group given the placebo treatment.
Entering the continuous data
The data collected in the present study pertains to both pre-treatment and post-treatment CY-BOCS scores, for each group, thus evaluating the efficacy of pharmacotherapy treatment. The meta-analysis between drug and placebo groups will involve entering the data in appropriate data entry format. The appropriate data entry format selection will involve identification of required information following the methodology, given in the previous article. Using the information as the guideline selecting appropriate data entry format is possible.
Following the above-shown identification steps and available data, two data entry formats directories for unmatched groups appear feasible. The image below shows the respective directories: ‘Unmatched groups, post data only’ and ‘Unmatched groups, pre and post data’ containing suitable data format options.
Unmatched groups, post data only
The image below show multiple data entry formats (offered by the Professional version) for the comparison of two groups. For the purpose of demonstration, first format offering fields for ‘Mean, standard deviation and sample size in each group’ is selected. However, other formats as per the available data can also be selected.
As shown below, Comprehensive Meta Analysis (CMA) application allows customization of the field names, with the changes before and after entering the group names. The software automatically generates the columns for relevant measures of effect size, as per the continuous data. Thus, any of the effect size measures can be selected as the primary index.
‘Difference in means’ will be suitable for data in raw units and ‘St. diff in means’ will be suitable for comparing data recorded over different scales. ‘Hedges’s g’ is yet another standardized effect size measure, suitable for working with small sample sizes. The figure below shows the complete spreadsheet, with the post-treatment values for ‘Drug’ and ‘Placebo’ groups. As all the studies use the same measure of assessment, the effect direction is set to ‘Auto’.
Note: As per the researcher’s knowledge of the scale used for measure outcome, the effect direction is set. In order to determine the effect size in simple data entry format, such as the one used presently, CMA software calculates the effect size by subtracting the second mean from first. However, for pre-post designs, CMA will calculate effect size by subtracting pre scores from the post. So, the user can select the appropriate direction, to favor the desired results.
Unmatched groups, pre and post data
The image below shows multiple data entry formats (offered by the Professional version) for ‘Unmatched groups, Pre and Post data’. For the purpose of demonstration, first format offering fields for ‘Mean, standard deviation pre and post, N in each group, Pre or Post Corr’ is selected. The ‘Pre or Post Corr’ field refers to the correlation values between the pre and post scores within groups.
Note: The correlation values may not be readily available always. In such a scenario of missing correlation values, the values could be either calculated or imputed (Fu et al. 2013). Furthermore, in the present analysis, the correlation values were missing and have been imputed as 0.5. This is because there is evidence indicating the correlation between baseline and post scores to be often greater than 0.5 (Balk et al. 2012). The image below shows the interface before and after customizing the field names, for the selected data entry format.
The figure below shows the complete spreadsheet, with the post – & pre-treatment values for ‘Drug’ and ‘Placebo’ groups. To direct the effect estimates towards drug treatment, the effect direction is set as ‘Negative’.
Proceed with the analysis following the continuous data entry in the software. Furthermore, the next article will guide through the analysis and interpretation of results for the present dataset.
- Balk, E.M. et al., 2012. Empirical assessment of within-arm correlation imputation in trials of continuous outcomes., Rockville.
- Fu, R. et al., 2013. Handling Continuous Outcomes in Quantitative Synthesis. Methods Guide for Comparative Effectiveness Reviews., Rockville.