Source: Medical Mathematics Thinking Cloud Classroom (ID:Datamedi)
Through the content of the last issue, we have a certain understanding of meta-analysis. But how to do meta-analysis and how to do it may still be at a loss. In this issue, Xiaoxing will introduce the general process of Meta-analysis. I hope you can have a certain understanding of the process of Meta-analysis.
Meta-analysis can be roughly divided into one step, as shown in the figure below.
0 1 put forward the problems that need to be evaluated.
The clinical questions raised should be scientific and can be answered through the results of research.
Good questions should be evidence or basis for clinical decision-making.
Appropriate questions should be put forward according to the actual clinical problems, and then further adjustments should be made according to the actual situation of preliminary retrieval.
If you want to evaluate the correlation between periodontal disease and bronchitis through Meta-analysis, and the preliminary search shows that there are few studies, then you can consider changing the evaluation question to evaluate periodontal disease and chronic obstructive pulmonary disease (note: chronic obstructive pulmonary disease mainly includes chronic bronchitis and emphysema).
02 to develop qualified standards
Eligibility criteria include inclusion criteria and exclusion criteria. It is suggested that the qualification standards should be formulated according to "Picos". ?
"P" refers to the participant, patient or target disease.
"I" means to interfere or expose.
"C" stands for control or comparison.
"O" indicates the result or end point.
"S" refers to the type of research design.
You can say "? Picos? "The information in also covers the qualification criteria for conducting research on the analyzed problems. Of course, the production team can also increase the relevant restrictions according to the actual situation, such as increasing the time limit, the restrictions on the research site (inpatients, community people, outpatients) and so on.
03
Formulate retrieval strategy
For example, search "Can patients with chronic kidney disease take long-chain polyunsaturated fatty acids (n-3 LCPULA) to prevent or delay the progression of kidney disease compared with placebo control group?"
Search term:
Various possible words (or) of the first group of polyunsaturated fatty acids
Various possible words (or) of the second group of chronic kidney disease
The third group of research methods (or)
These three groups are linked by "and"-extraction strategy.
Use c and o in PICO and research design types or research environments and settings to limit retrieval. A total of 242 related articles were obtained.
04? Time series prediction and analysis steps
Before the retrieval begins, producers should understand the establishment of PRISMA flow chart, and explain the identification, screening, qualified conditions, selection criteria and analysis process of the research.
05 time series prediction analysis steps
After the above steps are completed, you can consider registering on the registration platform.
Of course, scientific research registration is not a necessary link. If it is not registered, it is recommended to consult and read the manuscript contract when submitting the manuscript, and avoid those journals that need to be registered. Up meta now, only the meta-analysis of animal experiments has no platform registration.
06 research screening
After the retrieval, while recording the relevant retrieval results according to the PRISMA flow chart, the relevant research of retrieval should be imported into EndNote software, NoteExpress software, Microsoft Access software or other document management software for storage.
In previous issues, Xiaoxing introduced the use of EndNote software, which can help to record and check the repetition of research. The screening process is roughly as follows:
07 evaluation method quality
Cohort study/case control study: Newcastle-Ottawa Scale (NOS)
Cross-sectional study: (1) was recommended by American Health Care Research and Quality Inspection Agency (AHRQ), and the answers to 1 1 were "Yes", "No" and "Unclear" (2) was developed by Joanna Evidence-Based Nursing Center (JBI) in Australia.
Bias Risk Assessment Standard (JBI) in Current Situation Investigation
Experience Summary, Case Analysis and Literature Review Biased Risk Assessment Criteria (JBI)
Clinical trial evaluation: Jadad scale
08 data extraction
After screening out the qualified studies, relevant data can be extracted.
You can use Microsoft Excel software to extract data. Microsoft Excel is recommended? Access software. The extraction table can directly use the data extraction table of Cochrane cooperation network related groups, or can be modified; Or follow the recommendations of cochrane manual.
The contents of extracted data should generally include the first author, publication year, demographic data, trials to be evaluated and reference trials, methodological quality related information, etc.
At the same time, the basic characteristics of the research table are also formulated.
09? statistical analysis
Firstly, the heterogeneity of included studies should be evaluated. Chi-square test and test are still commonly used evaluation methods, and the p value of chi-square test is 0. 1.
Secondly, when investigating the sources of heterogeneity, it is very important to consider the true heterogeneity of a single study when the heterogeneity comes from accidental opportunities or errors caused by calculating diagnostic accuracy indicators. In addition, evaluating the methodological quality of a single study can be used as a useful tool to identify the sources of heterogeneity.
According to the heterogeneity, we can decide which effect model to use, and use related software to analyze related outcome indicators.
Generally speaking, if the heterogeneity is acceptable, the fixed effect model is adopted, otherwise the random effect model is adopted. Then according to the characteristics and clinical background of the study, the corresponding subgroup analysis, cumulative meta-analysis and publication bias analysis were carried out.
The content of statistical analysis is becoming more and more important. Xiaoxing will make a meta-statistical analysis and interpretation of the results alone. Don't miss it!
10? Interpretation of results
See you next time!
1 1? Meta-update
With the emergence of new research, meta-analysis needs to be constantly updated. Cochrane collaboration network suggests updating at least once every two years, and some CRGs suggest updating once a year. Of course, it can also be updated according to the actual clinical needs.
Ok, in this issue, Xiaoxing is here to share with you. I believe everyone has a certain understanding of the process of meta-analysis. The next issue of Xiaoxing will introduce you to Meta-statistical Analysis and the interpretation of the results, and then you can actually do meta-analysis! I hope to see you in every issue. See you next time! Click on the end of the article to read the full text to learn more!
Finally, I sincerely wish you all further and further on the road of scientific research.