Overview of meta‐analysis

Abstract

Meta‐analysis has been recognized as the best means to evaluate objectively and study the evidence for a particular issue. In order to give researchers a better understanding of the Meta process, we present an overall introduction to Meta‐analysis in terms of comprehensive assessment of the literature, goals, advantages, main steps, and article structure.

Introduction

With the development of evidence‐based medicine, Meta‐analysis has been recognized as the best means to evaluate objectively and study evidence for a specific problem, and is regarded as the highest level of evidence, which has become a good basis for evidence‐based decision‐making. Meta‐analysis is a kind of statistical method that synthesizes many research results of the same problem with specific conditions, which is the basis of clinical research (Zeng XT, et al., 2013). Compared with reviews and systematic reviews, the literature screening of Meta‐analysis is more rigorous, the data extraction of Meta‐analysis is more standardized and its refinement of results through statistical analysis makes the results more scientific and objective. Moreover, the purpose of Meta‐analysis is to increase the efficiency of statistical test, quantitatively estimate the average level of research effects, evaluate the inconsistency of research results, and find new hypotheses and research ideas.

Overview of meta‐analysis

The understanding of Meta‐analysis is mainly from the following aspects: (1) Meta‐analysis is a massive integration of literatures and plays a role of comprehensive evaluation. (2) The goal of Meta‐analysis is to evaluate the results of multiple research results of the same kind. (3) The method of Meta‐analysis is to collect literatures, extract specific problems from the literatures, then perform statistical analysis, and finally obtain the results. (4) The advantage of Meta‐analysis is to conduct an integrated analysis of published or unpublished research results and obtain convincing evidence. Compared with traditional reviews, Meta‐analysis is more comprehensive and accurate because it introduces statistical methods.

Mainsteps and key points

The steps of Meta‐analysis mainly include raising questions and determing the purpose of the research, literature retrieval and screening, literature quality evaluation, data extraction and entry, forest map and funnel plot drawing, heterogeneity test, bias and sensitivity analysis and results description (Figure  1 ). The overview is as follows.

Figure 1.

The main steps of Meta‐analysis

Topic selection

First of all, the topic selection of Meta‐analysis should be based on clinical significance. The principle of the topic selection mainly includes: the importance, controversial and innovative of the problem, as well as the need for appropriate original research and clear effect indicators (Zeng XT, et al., 2013). Topics can be selected from the relevant research, evaluation of intervention measures, evaluation of diagnostic methods, prognosis estimation, cost and benefit analysis, selection of suitable subjects, determination of reasonable indicators and making inclusion criteria.

Literature retrieval and screening

In the following literature retrieval, it is necessary to identify the search terms as comprehensively as possible in different databases (Zhan SY, et al., 2010). After retrieving the required documents, read and screen the retrieved documents with the steps of reading abstracts, full texts, and references.

Literature quality evaluation

Quality evaluation of literature is an important part of literature screening. In the quality evaluation, we should pay attention to whether the research design is clear;whether the method is random; whether the statistics are blind; whether the research background is similar; whether the effect evaluation is accurate; whether the research is adaptive and the results are fully described (Castaño‐RN, et al. 2017; Abt D, et al. 2017). Now commonly used quality assessment tools include Cochrane bias risk assessment tools, AMSTAR scale, OQAQ list, CASP list, SQAC scale, Chalmers scale (Munder T, et al., 2018). In the literature management, we can use Endnote and other softwares to record and manage the literature obtained by various ways, and easily generate a list of references to improve the efficiency of literature screening.

Data extraction and entry

Information extraction refers to the correct collection and recording of the included research results and all valuable information according to the inclusion criteria. It is a key step in Meta‐analysis, directly affects the accuracy of the results and connects the original research report and Meta‐analysis. The main function of information extraction is to use original data as the source of data analysis and to facilitate the verification of research data. In information extraction, we must first draw the information extraction table. The main contents should include: (1) Author, year, source; (2) Research design: method, grouping, blind method; (3) Object characteristics: sample size, region; (4) Intervention characteristics: method, blind method, dose; (5) Evaluation index: instrument, index, time; (6) Results (quality) rate: ratio, relative risk; It is worth noting that in the process of information extraction, two people are required to extract and cross‐check independently, while we need to constantly revise and improve the extraction form (Wang M, et al., 2017).

Meta‐software analysis‐forest map and funncl plot drawing

Meta‐analysis software provides an important guarantee for the implementation of various types of Meta‐analysis. Currently, there are many kinds of softwares, and the operating system and its functions are also different. At present, the commonly used softwares include Stata, R language (Foo YZ, et al., 2017), SAS, SPSS, etc. Among them, the statistical analysis of data mainly starts from the following aspects: (1) Calculate the effect value, variance and weight of each research; (2) Homogeneity test is needed for the effect values of each research result; (3) Calculate the combined effect value; The effect value and confidence interval of each research is classified; (4) Draw forest map.

In 2010, international evidence‐based medicine experts Sharon Straus and David Moher jointly issued a call that all systematic reviews/Meta‐analysis should be registered in order to reduce publication bias and promote transparency and cooperation in the production process. Registration can not only improve the quality of Meta‐analysis, but also avoid the waste of valuable manpower and material resources caused by repeated work (McClure GR, et al., 2016). At present, Meta‐analysis can be registered in Cochrane groups or PROSPERO. Compared with the two, Cochrane registration is more rigorous and cumbersome, and currently only randomized trials, diagnostic accuracy studies and methodological meta‐analysis are accepted. PROSPERO registration is relatively simple, and the audit is not very strict. At present, the Meta‐analysis registration of treatment, prevention, diagnosis, monitoring, risk factors and genetic association are accepted.

Heterogeneity analysis

Meta‐analysis is based on the original research, so its quality must be affected by the quality of the original research. According to the types of original research, the corresponding quality evaluation tools are also different. Similarly, the quality of the original research is divided into report quality and methodological quality. The report quality can be evaluated by the report specification, and the methodological quality needs special tools. Inevitably, all studies included in the same Meta‐analysis are different. We call the variations among different studies in Meta‐analysis heterogeneity (Qiu Y, et al. 2017; Abd EAMS, et al. 2017).

The heterogeneity of Meta‐analysis is divided into clinical heterogeneity, methodological heterogeneity and statistical heterogeneity. Clinical heterogeneity refers to the variation caused by different participants, different interventions and different endpoints of the study. Methodological heterogeneity refers to the variation caused by the differences in experimental design and quality, such as the hidden differences in the application and distribution of blind methods, or the inconsistency in the definition and measurement of outcome during the experiment. Statistical heterogeneity is the variation of estimated therapeutic effects between different trials, which is the direct result of clinical and methodological diversity between studies. In order to reduce the heterogeneity among studies, we should formulate strict and unified inclusion and exclusion criteria when conducting Meta‐analysis. Only studies with the same research purpose and high quality can be included in the analysis, and the consistency of research objects and treatment factors can be considered, which can ensure the clinical homogeneity of included studies to a certain extent, which is also a prerequisite for merging different studies. So as to ensure the homogeneity of methodology, it is necessary to conduct strict quality evaluation of the combined research, including random methods, blind implementation, random scheme concealment, intentional treatment analysis and baseline similarity. Only on the basis of certain homogeneity in clinical and methodological aspects, can we enter the statistical heterogeneity test between studies and the next merger.

As for the presentation and description of the results, we usually accurately display and highlight the results in the form of articles, fully discuss the interaction between the analysis results and clinical information, and finally draw appropriate conclusions to guide the future clinical operation and theoretical update.

Framework for writing meta‐analysis articles

A clear and smooth structure is especially important for the full and perfect presentation of meta results. At the end of the article, we introduce the structure of the Meta‐analysis, the content of the article mainly includes the title, abstract, background, data and methods, results, discussions or conclusions and references (Figure  2 ). The title of the article needs to contain the points in PICO (P: Population, I: Intervention, C: Comparison, O: Outcome) (Moher D, et al., 2009). Abstract writing is composed of purpose, methods, results and conclusions. In writing the background of the article, we need to explain the clinical problems and describe the burden and current situation of these clinical problems, then put forward the intervention and control measures in our study and clarify the purpose of the study. The description of materials and methods needs to explain the following issues in detail: (1) Search Strategies: What databases have we searched? What are the search terms? (2) The inclusion and exclusion criteria of research. (3) How do we conduct literature screening and data extraction? (4) Process of quality evaluation of methodologies included in the study. (5) What softwares are used for statistical analysis? The description of the results needs to include the literature screening process and results diagram, the baseline characteristics table of the included research, the methodological quality evaluation table of the included research and the bias risk map, the forest map to represent the analysis results, and the funnel plot to analyze and publish the bias results. The discussion part of this paper mainly summarizes the evidence, and compares it with other studies to describe the limitations of this study. Finally, it summarizes and prospects whether the results obtained by us have clinical value and what enlightenment they have.

Figure 2.

Article structure of Meta‐analysis

Outlook

In summary, Meta‐analysis is a development process that is constantly updated and improved with the demand, and there is still room for research and development and improvement with the deepening of practice. Meta‐analysis should be conducted according to the recommended reporting norms, and constantly follow up the progress of Meta‐analysis, timely learning, thinking and research.

Ethical statement

Not applicable.

Conflict of interest

No conflicts of interest were declared.

Funding

None.

Transparency statement

All the authors affirm that this manuscript is an honest, accurate, and transparent account of the study being reported; that no important aspects of the study have been omitted; and that any discrepancies from the study as planned (and, if relevant, registered) have been explained.

Authors' contribution

Xu Fang and Zhao‐Qiong Zhu contributed the central idea, analysed most of the data, made figures & tables and wrote the initial draft of the paper. Nan Zhao contributed to refining the ideas, carrying out additional analyses and finalizing this paper.