Preface
Meta analysis is a type of statistical method that integrates many research results of the same subject with specific conditions. The most important thing to do a meta-analysis is to choose the topic. The topic selection determines the workload. The novelty and clinical applicability of the topic determine the value of the research and the difficulty of article submission. After selecting the research direction, the specific research steps include seven parts.
1 1. Develop a search strategy and determine the literature to be included in the study
Specify the search strategy in accordance with the topic selection, comprehensively and extensively collect randomized controlled trials to determine the inclusion and exclusion criteria, and eliminate the selection and extraction of documents that do not meet the requirements, including the quality evaluation of the original text result data, graphs and other trials and statistical processing of feature descriptions Interpretation of results, draw conclusions, evaluate, maintain and update data.
1 Second, the quality of research and evaluation
The quality of a Meta analysis depends on the quality of a single study. There are many ways to evaluate the quality of articles. The evaluation method is determined according to the type of research.
The Cochrane risk bias assessment tool was selected for randomized controlled studies, and The Newcastle-Ottawa Scale was used for non-randomized controlled studies.
1 3. Research quality and evaluation
Before data extraction, design the data extraction form. If you include more questions, rework is a very depressing thing. We extract the required literature data into the table. example:
1 4. Research quality and evaluation
Data analysis is to put the data in Revman, STATA and other software, and the rest of the work software will help you complete. Let's take Revman as an example to briefly introduce the process.
(1) Add research
The next step is to repeat the work. Just fill in all the research in the form. After the research is entered, it is Addcomparison.
(Two) add a comparison
(3) Add ending indicators
(4) Generate a forest map
1 V. Heterogeneity Evaluation
There is no exactly the same research in the world, and the literature we included must be different. The analysis of heterogeneity is to detect this difference. Heterogeneity evaluation can be displayed through forest diagrams and funnel diagrams, evaluated with numerical Q statistics and I2 statistics. If the heterogeneity is large, data processing methods include subgroup analysis, sensitivity analysis and meta regression. We will introduce them in detail in the following report.
1 6. Identification of publication bias
In a Meta analysis, there are often more positive results than negative results, mainly because positive results are easier to publish, so there may be many negative results that have not been published. If you consider this part of the data, the analysis results may be different. In order to control the bias, the main method is to collect the literature as comprehensively as possible, and the commonly used evaluation method is to observe whether the funnel chart is symmetrical. The commonly used method is the Begg and Eggers test. The Trim method can also be used to deal with bias.
1 Seven, sensitivity analysis
The commonly used method of sensitivity analysis is mainly to analyze the impact of a single study, delete a study, and reanalyze it to observe whether it affects the results. Or the cut-and-compensation method identifies and corrects the funnel asymmetry caused by bias.
references:
1.http://ebm.dxy.cn/bbs/topic/27847710
2. Concise tutorial on meta analysis of knife and pencil