### Comprehensive **Meta-Analysis**

Comprehensive Meta-Analysis is a powerful computer program for meta-analysis. The program combines ease of use with a wide array of computational options and sophisticated graphics.

The mechanics of data entry are much simpler in CMA – you work with a spreadsheet interface, and can copy-and paste-data as easily as you could in Excel. By contrast, the data entry process in Revman requires the user to set up tables and comparisons before starting data entry.

In every meta-analysis you start with the published summary data for each study and compute the treatment effect (or effect size). For example, if a study reports the number of events in each group you might compute the odds ratio. Or, if a study reports means and standard deviations you might compute the standardized mean difference.

Revman will accept summary data in only two formats – events and sample size, or means and standard deviations. If any studies provide data in another format (such as odds ratio and confidence intervals) you would need to compute the effect sizes and variances manually for those studies. By contrast, CMA allows you to enter data in more than 100 formats, and will compute the effect size and variance for all of these formats. Equally important, Revman requires that data for all studies be entered using the same format. By contrast, with CMA you can enter data for each study in its own format, and use as many formats as needed in the same analysis. CMA also supports a much wider range of effect sizes than Revman.

The forest plot in Revman offers few options for customization. By contrast, CMA allows the user full control over all elements in the forest plot, will create scalable plots (that print at the highest resolution possible for the printer or journal), and allows the user to control the color for every element on the plot. CMA also allows one-click export to other programs such as PowerPoint™ and Word™.

CMA allows you to assess the impact of moderator variables. Use analysis of variance to compare the treatment effect across groups (“Is the treatment more effective for acute patients than for chronic patients?”). Use meta-regression to assess the impact of continuous moderators (“Does the treatment effect increase with dosage?”). To assess the potential impact of publication bias CMA includes an array of functions including a funnel plot, where Revman includes only the funnel plot. CMA will run a cumulative meta-analysis to show how the evidence has shifted over time. It will also run a one-study removed analysis to show the impact of each study on the combined effect.

Yes. The development team for CMA includes some of the same people responsible for the development of Revman. CMA includes all of the same computational formulas, has been validated against Revman and provides exactly the same results (see documentation). CMA offers additional options as well, but includes a button to "Use the same options as Revman," which sets all options to match Revman.

NEXT: StataCMA is a program developed specifically for meta-analysis. As such, it includes functions to automatically compute effect sizes, to perform basic and advanced meta-analyses, and to create publication quality graphics.

Stata is a general purpose statistical package. While Stata has no intrinsic support for meta-analysis, various experts have written macros for meta-analysis, which can be downloaded from the Stata web site and incorporated into Stata. These macros include procedures for basic analysis, for cumulative analysis, for meta-regression, for publication bias, and more. Since the functionality of Stata and CMA are comparable, the main difference is in ease of use and in the options for customizing the output.

Stata is a command-driven language, which means that you type commands, or use a dialog box to create commands, which are then submitted to the program. CMA is a menu-driven program, similar to Excel™.

Stata will accept summary data in only three formats – events and sample size, means and standard deviations, or (in some cases) point estimate and confidence interval. If any studies provide data in another format you would need to compute the effect sizes and variances manually or by writing code. By contrast, CMA allows you to enter data in more than 100 formats, and will compute the effect size and variance for all of these formats. Equally important, Stata requires that data for all studies be entered using the same format. By contrast, with CMA you can enter data for each study in its own format, and use as many formats as needed in the same analysis.

Stata computes values and sends these to a DOS-like window for viewing. In CMA the analysis screen is interactive – you can use the screen interactively to explore the impact of different studies, the effect of alternate weighting schemes, and so on.

Stata’s forest plot offers few options for customization. It includes a column for the study name and a symbol representing the point estimate and confidence interval. By contrast, the forest plot in CMA can be customized extensively.

Yes. The development team for CMA includes some of the same people who developed the Stata macros. CMA includes all of the same computational formulas, was validated against Stata and provides exactly the same results (see documentation). CMA offers additional options as well, but includes a button to "Use the same options as Stata," which sets all options to match Stata.

Yes. You can open the data sheet in Stata, copy the data onto the Windows clipboard, and then paste it into CMA. Then you tell CMA what kind of data is located in each column. The process takes only moments to complete.

PREVIOUS: Revman NEXT: SpssCMA is a program developed specifically for meta-analysis. As such, it includes functions to automatically compute effect sizes, to perform basic and advanced meta-analyses, and to create publication quality graphics.

SPSS is a general purpose statistical package with no intrinsic support for meta-analysis. However, David Wilson has written macros that can be incorporated into SPSS and will run a basic meta-analysis, an analysis of variance, and meta-regression.

In every meta-analysis you start with the published summary data for each study and compute the treatment effect (or effect size). For example, if a study reports the number of events in each group you might compute the odds ratio. Or, if a study reports means and standard deviations you might compute the standardized mean difference. Additionally, you need to compute the variance for each effect size.

Wilson’s macros require the user to compute an effect size and variance for each study, and then provide these values to the program. Therefore, the user must either compute these values separately and then enter them as data, or write code to compute these values within SPSS.

By contrast, with CMA you enter the data directly in almost any format(s), and the program computes the effect size and variance automatically.

Wilson’s macros will report all relevant statistics, but will not create graphics such as a forest plot. CMA is able to create a forest plot, which can play a key role in helping the researcher to interpret the data and to convey it to others.

SPSS itself does not include any support for meta-analysis. Wilson’s macros use the same formulas as CMA (CMA offers additional options as well), and so will yield identical results. This assumes, of course, that the user has used the same formulas to compute effect sizes and variance for each study.

Yes. You can open the data sheet in SPSS, copy the data onto the Windows clipboard, and then paste it into CMA. Then you tell CMA what kind of data is located in each column.

PREVIOUS: Stata NEXT: SasCMA is a program developed specifically for meta-analysis. As such, it includes functions to automatically compute effect sizes, to perform basic and advanced meta-analyses, and to create publication quality graphics.

SAS is a general purpose statistical package with no intrinsic support for meta-analysis. David Wilson has written macros that can be incorporated into SAS and will run a basic meta-analysis, an analysis of variance, and meta-regression.

In every meta-analysis you start with the published summary data for each study and compute the treatment effect (or effect size). For example, if a study reports the number of events in each group you might compute the odds ratio. Or, if a study reports means and standard deviations you might compute the standardized mean difference. Additionally, you need to compute the variance for each effect size.

Wilson’s macros require the user to compute an effect size and variance for each study, and then provide these values to the program. Therefore, the user must either compute these values separately and then enter them as data, or write code to compute these values within SAS.

By contrast, with CMA you enter the data directly in almost any format(s), and the program computes the effect size and variance automatically.

Wilson’s macros will report all relevant statistics, but will not create graphics such as a forest plot. CMA is able to create a forest plot, which can play a key role in helping the researcher to interpret the data and to convey it to others.

SAS itself does not include any support for meta-analysis. Wilson’s macros use the same formulas as CMA (CMA offers additional options as well), and so will yield identical results. This assumes, of course, that the user has used the same formulas to compute effect sizes and variance for each study.

Yes. Since SAS does not support copy-and-paste, it is not possible to simply copy the data to CMA. However, SAS is able to export data to a file, which can then be imported to CMA.

PREVIOUS: SPSS NEXT: ExcelExcel is a spreadsheet program with no intrinsic support for meta-analysis. While it is possible to program all meta-analysis formulas in Excel, this requires knowledge of the formulas and a substantial investment of time for development and testing. Also, there is no mechanism in Excel that can be used to create a forest plot.

By contrast, all of this functionality is built into CMA.

PREVIOUS: SAS NEXT: MetawinCMA and Metawin are both dedicated meta-analysis programs but CMA incorporates a much wider and more fully developed set of options.

In every meta-analysis you start with the published summary data for each study and compute the treatment effect (or effect size). For example, if a study reports the number of events in each group you might compute the odds ratio. Or, if a study reports means and standard deviations you might compute the standardized mean difference. Additionally, you need to compute the variance for each effect size.

Both Metawin and CMA will allow you to enter summary data and will compute the effect size from that data. However, CMA will work with a much wider array of data formats. Metawin will accept data in a few formats where CMA can accept more than 100. Metawin requires that all studies provide data in the same format while CMA allows you to enter data for each study in its own format. Metawin can work with a few indices of treatment effect (or effect size) where CMA includes more than 15.

Metawin will run the analysis and show the computed values. CMA displays all values as part of a scrollable grid which makes the analysis transparent – you can see which studies are included in the analysis, how the studies were weighted, and so on. CMA also includes a much more extensive set of computational options.

In CMA you can fully customize the plot, to ensure that each study stands out clearly, so that the plot is proportioned properly on the page, include all relevant columns, and so on. In Metawin the plot is very basic, allowing one column for study names but little control over formatting.

PREVIOUS: ExcelComprehensive Meta-Analysis is a powerful computer program for meta-analysis. The program combines ease of use with a wide array of computational options and sophisticated graphics.