Article type
Year
Abstract
Background:
Over the last 20 years, there has been a significant rise in the number of systematic reviews and meta-analyses being published. These types of studies are highly valued as they are believed to offer the highest level of evidence if carried out correctly and can offer clear and trustworthy summaries for clinical decision-making. However, researchers of meta-analyses may encounter the problem of missing mean and standard deviation (SD), often resulting in only median, interquartile range, or range data being available. Unfortunately, these data cannot be directly used for meta-analysis. Some estimation and conversion methods have been proposed, but there is currently no user-friendly tool available based on multiple scenarios of missing SD.
Objectives:
The objective of this study was to develop a useful tool that can find sample mean and SD for common scenarios.
Methods:
Two researchers extensively reviewed relevant methodologic references such as textbooks and systematic reviews. Based on the work, they compiled a list of possible scenarios in which there were inadequate data for meta-analysis and sought out formulas for finding sample mean or SD in each case.
Results:
The working group double-checked the scenarios and corresponding solutions, and the identified 10 primary scenarios can be categorized into two broad categories. The first category involves circumstances in which descriptive statistics of a single group are available (within-group circumstances). The second category involves situations in which effect estimates of two treatment groups are available. Then, the working group developed an Excel tool for estimating and converting the data in each scenario. The solutions in the tool are primarily based on the Cochrane Handbook and a methodology study.
Conclusions:
Researchers need to understand the circumstances and solutions of missing data before conducting meta-analysis, and appropriate data processing could improve precision and quality of evidence. Conversely, excluding certain irregular outcome reporting in a meta-analysis may lead to imprecision and biased estimates. Therefore, familiarity with the different scenarios and the tools available to address them is crucial for researchers to conduct a high-quality meta-analysis.
Over the last 20 years, there has been a significant rise in the number of systematic reviews and meta-analyses being published. These types of studies are highly valued as they are believed to offer the highest level of evidence if carried out correctly and can offer clear and trustworthy summaries for clinical decision-making. However, researchers of meta-analyses may encounter the problem of missing mean and standard deviation (SD), often resulting in only median, interquartile range, or range data being available. Unfortunately, these data cannot be directly used for meta-analysis. Some estimation and conversion methods have been proposed, but there is currently no user-friendly tool available based on multiple scenarios of missing SD.
Objectives:
The objective of this study was to develop a useful tool that can find sample mean and SD for common scenarios.
Methods:
Two researchers extensively reviewed relevant methodologic references such as textbooks and systematic reviews. Based on the work, they compiled a list of possible scenarios in which there were inadequate data for meta-analysis and sought out formulas for finding sample mean or SD in each case.
Results:
The working group double-checked the scenarios and corresponding solutions, and the identified 10 primary scenarios can be categorized into two broad categories. The first category involves circumstances in which descriptive statistics of a single group are available (within-group circumstances). The second category involves situations in which effect estimates of two treatment groups are available. Then, the working group developed an Excel tool for estimating and converting the data in each scenario. The solutions in the tool are primarily based on the Cochrane Handbook and a methodology study.
Conclusions:
Researchers need to understand the circumstances and solutions of missing data before conducting meta-analysis, and appropriate data processing could improve precision and quality of evidence. Conversely, excluding certain irregular outcome reporting in a meta-analysis may lead to imprecision and biased estimates. Therefore, familiarity with the different scenarios and the tools available to address them is crucial for researchers to conduct a high-quality meta-analysis.