What Is Meta-Analysis?

Meta-analysis Of epidemiological exploration and evaluation, replacing the individual as the analytical entity with the findings of the original research. The main reason for the meta-analysis is that for multiple independent studies, many observation groups are too small to produce any clear opinions.

Meta-analysis

This entry lacks an overview map . Supplementing related content makes the entry more complete and can be upgraded quickly. Come on!

Meta-analysis Of epidemiological exploration and evaluation, replacing the individual as the analytical entity with the findings of the original research. The main reason for the meta-analysis is that for multiple independent studies, many observation groups are too small to produce any clear opinions.

Chinese name

Meta-analysis

Types of

concept

Presenter

Light and Smith

Presentation time

1971

The concept of meta-analysis was first introduced by Light and Smith in 1971. At that time, in response to the issue of a large number of published scientific papers, the same research yielded very different results, they proposed that small samples of various therapies for a disease, the results of a single clinical trial should be collected worldwide. It conducts systematic evaluation and statistical analysis, and provides the real scientific conclusions as soon as possible to the society and clinicians, so as to promote the promotion of truly effective treatment methods, and abandon ineffective or even harmful methods that have no basis.

Glass first named the concept Meta-analysis in 1976 and defined it as a method of collecting, combining, and statistically analyzing different research results. This method has gradually developed into a new subject-the main content and research method of "Evidence-based Medicine". The main purpose of meta-analysis is to reflect the results of previous research more objectively and comprehensively. Researchers do not conduct original research, but comprehensively analyze the results obtained by the research.

A literature review under a general concept is a list, simple description, and preliminary discussion of the content or results of the relevant literature, while a meta-analysis takes it to the next level. It can be divided into three categories according to the basis or data source on which the meta-analysis is based: Meta-analysis based on literature (MAL); Meta-analysis based on summary data (MAS) ; Meta-analysis based on indlidual patient data, MAP or IPD Meta-analysis. The difference is that: MAL's literature search is limited to published studies, and then the results of these studies are combined for analysis; MAS must not only obtain relevant published literature, but also a summary of relevant statistical data conducted by the author; In addition to searching all published relevant literature, the IPD meta-analysis also looks for unpublished related research that exists in various scientific communities, a step further on the basis of MAS. All clinical trials, whether or not they have been published, must be able to obtain from the investigator individual patient raw data as well as various effector indicators. This is important for analysis of tumor etiology or efficacy. Because most of the phase III clinical trials on the prognosis of cancer patients, the main research indicators are mostly survival time or survival rate, or time to disease progression, etc. In most cases, the information obtained in different publications is insufficient for An analysis of the entire time of the occurrence of a real event (such as tumor death). This makes MAL and MAS based on published literature more difficult. At the same time, considering that statistically significant positive results are easier to publish than negative results, etc. can cause

Pros and cons of meta-analysis

The application of meta-analysis avoids the limitations of a single small sample clinical trial, makes the results of the analysis more comprehensive and reliable, and provides a good basis for medical decisions. However, many artificial factors may affect the results of the analysis. Such as the choice of trials, the determination of research endpoints, and the degree of acceptance of trial homogeneity. To overcome these disadvantages, we should strictly follow the relevant regulations of the meta-analysis and gradually form relatively fixed standards, such as using the survival index as the prescribed research effect index. At the same time, it must be recognized that meta-analysis is not a good cure for all diseases, it cannot replace large single randomized clinical trials, and it should not be used as an excuse for conducting small, not so convincing, meaningless clinical trials. It and the large-scale randomized clinical trials should complement each other and take their own strengths. We should not treat meta-analysis as just a statistical analysis tool, but combine it with clinical observations or a critical review of data to help us evaluate the quality of some clinical trials and study the efficacy of different trials. The differences and causes, as well as providing direction and evidence for further research. ?

The future of meta-analysis

The meta-analysis has developed rapidly in the past 20 years. The number of such articles published each year has ranged from dozens of articles in the 1980s to nearly 500 in 2000. Some people think that the IPD meta-analysis should be carried out systematically and updated regularly. The World Cochrane Collaboration is currently undertaking a huge effort to register all randomized clinical trials around the world, followed by a corresponding meta-analysis. At the same time, with the development of information technology and the continuous improvement of large-scale clinical trial databases, there is currently a tendency to switch from retrospective meta-analysis to forward-looking meta-analysis, because the latter can better avoid publication bias and get analysis earlier. Results and the basis for medical decisions. This, I believe, should better promote the further development of meta-analysis. Meta-Analysis-Examples of Meta-analysis of Jian select for non-small cell lung cancer At the just-concluded 10th World Conference on Lung Cancer (Canada, March 2003?), Professor Le Chevalier T and colleagues reported a report on Jian select The results of a meta-analysis of survival data for treatment of advanced non-small cell lung cancer (NSCLC) aroused great interest from participants.

The purpose of this meta-analysis study is clear, comparing the efficacy of Jianjian combined with other drugs in the treatment of advanced non-small cell lung cancer with other chemotherapy regimens, and see if Jianjian / platinum can be slightly improved compared to other platinum-based regimens. The patient's overall survival rate and TTP, so the observation indicators include overall survival rate and progression-free survival rate. A comprehensive review of all relevant literature published before December 2002, searching of major tumor databases, registered clinical trial databases, conference abstracts, and published reviews. The design characteristics of each independent study were defined head-to-head (Head to head)

Planning stage

First determine the subject of the study, and then clarify the purpose of the program, the inclusion and exclusion criteria of the trial, the planned analysis indicators, and the statistical methods to be applied. A meta-analysis is best to study only one main problem, but after the main purpose of the research is clear, other secondary problems can be studied at the same time.

Finding and selecting clinical trials

Ideally, all literature relevant to the research topic should be included in the meta-analysis, whether or not it has been published. Bias in article publication, language, and citations must be taken into account. In general, trials showing statistical significance are more likely to be published in some journals. These papers have shorter publication cycles and they have the highest click index. They are usually published in English, and they are referenced and cited more often than trials that do not show statistical significance. Searching by computer alone is not enough, although it is more convenient. Because even large databases such as MEDLINE or EMBASE contain documents published in various journals, and documents from 1966 and 1974 respectively. It is also necessary to manually search some conference materials or directly contact researchers and pharmaceutical companies to ensure the comprehensiveness of the literature.

Test quality

Ensuring the quality of each trial is critical because it affects the quality of the entire meta-analysis. Inadequate randomization, excluding patients after randomization, non-parallel follow-up between treatment groups, and subjective evaluation of study endpoints will bias the results of the trial. The quality of independent research is therefore different. When conducting a meta-analysis, the results of each study should not be treated equally, but should be treated differently according to the quality of each independent study. For example, the quality of a single randomized controlled clinical trial is scored, the score is included in the selection criteria of the meta-analysis, or it is used as the weight in the combined test.

Describe the test

Before a meta-analysis of the results of each trial, each trial must be recorded and described. Including evaluation of trial design, characteristics of comparison between treatment groups, characteristics of patient population, evaluation of trial quality, and quantitative summary of trial results. This process allows researchers to find similar trials and combine them to understand the types of patients enrolled and the reliability of the evaluation data. Trials that were excluded from the meta-analysis and why they were excluded should also be described.

analysis

The homogeneity test is an important part of the meta-analysis, the purpose is to check whether the results of each test are consistent. In general, differences in results between trials simply due to sampling errors will not affect the reliability of the meta-analysis results. However, if it is found that the cause of the inconsistency is due to a special factor, such as too many cases of lost follow-up in a study, the results of this test should not be included in a meta-analysis. The most commonly used homogeneity test methods are 2 or Q test.

Summarizing and merging the data results is the essence of meta-analysis, and there are currently many statistical methods applied to this. Such as random effects model, Cochrane method, Glass method and Fisher-Z conversion method. In the field of tumor survival or curative effect research, analysis of the survival rate or the risk-to-risk ratio (HR) is more common. Generally, statistical calculations are performed on the data of each selected trial to obtain the main three values: O = number of events observed in the experimental group; E = assuming that the probability of events in the experimental group and the control group is the same, the expected test The number of events in the group; Var (OE) = (OE) Variance, which is used to measure the accuracy of treatment difference estimates in the trial. The formula for calculating the hazard ratio of a single test is: HR = Exp [(OE) / Var (OE)]. The calculation of the total hazard ratio for multiple experiments is: HRc = Exp [ (OE) / Var (OE)]. The 95% confidence interval for the risk ratio can be calculated by the following formula: upper limit = Exp [(OE) / Var (OE) + 1.96 / Var (OE) 1/2]; lower limit = Exp [(OE) / Var (OE) -1.96 / Var (OE) 11/2]. The data obtained from the analysis can finally be visually displayed using a forest plot.

The difference in absolute survival rate or disease-free survival between the two treatment groups is the main analysis indicator in the efficacy study. By using certain statistical methods, a combined estimate of the survival rate at a certain point in time (such as 5 years) can be obtained, The combined estimates of the difference in survival rates between the two groups can be found in the relevant literature.