我们将介绍对临床试验进行系统综述和元分析的方法，课程内容包括如何构建一个能够回答的研究问题，确定收入和排除条件，寻找证据，提取数据，评价临床试验中偏倚的风险，并进行元分析。 顺利完成课程后，同学们将能够： -描述进行系统综述的步骤 -采用“受试者，干预措施，比较方法，结果”(PICO)框架来提出能够回答的问题 -描述从临床试验报告中收集和提取数据的过程 -描述辩证地分析临床试验偏倚风险的方法 -描述和解释元分析的结果
Lecture 7A: Planning Your Meta-Analysis Section A
Loading...
來自 Johns Hopkins University 的課程
系统综述及元分析入门
1187 個評分
從本節課中
Planning the Meta-Analysis and Statistical Methods
This module will cover the planning of your meat-analysis and the statistical methods for meta-analysis.
與講師見面
Tianjing Li, MD, MHS, PHDAssistant Professor, Epidemiology
Bloomberg School of Public Health
Kay Dickersin, PhDProfessor, Epidemiology
Bloomberg School of Public Health
[MUSIC]
Hello everyone.
Today we're going to talk about planning your meta-analysis,
why planning your meta-analysis is important?
Because you have spent all the efforts in this course to identify studies for
your systematic review, and
now we're at the stage of quantitatively analyze the data from those studies.
And the learning objectives for today are: formulate a general framework for
quantitative analysis in a systematic review.
Quantitative analysis refers to meta-analysis.
I will define formally what is a meta-analysis and
talk about the potential advantages of doing meta-analysis.
And for any meta-analysis it's very important for you to understand the types
of data you have and the measure of effect or association we're going to use
to quantify the effect of intervention or to quantify the association.
By the end of the lecture, I hope you will take away with you the key messages.
The first essential element of analysis is a thoughtful approach to both its
qualitative and quantitative elements.
And you have heard about qualitative synthesis and
I will remind you in the next slide.
And you have to come up with an analytical framework
which include the specification of the comparison.
For example, are you comparing one intervention versus another one?
Or are you comparing people exposed to a risk factor versus
those who are not exposed to a risk factor.
And then you have to think about the types of the data you have.
Are the data continuous, are the data dichotomous or are they categorical?
And depending on the type of data you have,
you will choose the summary measures.
It could be risk ratio, odds ratio for dichotomous data, or mean difference for
continuous data.
And all these decisions, or
the choices you have made should address the study question.
And lastly, meta-analysis is really the statistical combination of the effect
estimates from two or more separate and independent studies.
Let's start by talking about planning your analysis.
Well, the results of a meta-analysis can be very misleading if you haven't paid
enough attention to the first few steps of a systematic review.
You have spent a lot of time in formulating the review question, and
the review question determines this eligibility criteria for
your systematic review.
Then you're going to follow your eligibility criteria and protocol to
identify, select, and critically appraise the studies for the systematic review.
And after you have decided on the studies to include,
you're going to collect appropriate data from those studies.
And lastly you're going to do the meta-analysis which is an optional
component of a systematic review, and
today we're going to focus on planning your analysis.
Systematic reviews contain analysis of the primary study, in other words the unit for
your analysis is the individual study.
And primary studies include analysis of their participants.
And there are really two components of any analysis.
The first one is qualitative analysis, or we refer to it as qualitative synthesis.
And the second part is the quantitative synthesis.
And I put qualitative synthesis before quantitative synthesis,
because it's the most important part of any systematic review.
It refers to a structured summary, description, and
discussion of the studies' characteristics that may affect the cumulative evidence.
For example, you have to look at the studies carefully and
ask yourself the question, are the studies similar or different?
Are the study participants comparable?
Are the interventions,
homogeneous across studies you're including in the systematic review.
Are the studies designed and conducted and reported properly.
Do you have enough information from those studies for your meta-analysis?
So please spend a lot of time and
put a lot of efforts in your qualitative synthesis.
And after you have done a thorough qualitative synthesis, you're at a stage
of doing quantitative analysis, which we refer to as meta-analysis.
And the general framework for
a synthesis include answering and thinking about the following questions.
What is the direction of effect or association?
Is the intervention preventive?
Is the intervention harmful?
And what is the size of effect?
So that refers to the point estimate from your estimate.
And then are the effects consistent across studies?
Are all the studies estimating a similar effect or
are they estimating a different effect?
If the results from studies are not consistent, why they're different, and
lastly, what is the strength of evidence for the effect?
The strength of evidence also relies on judgments based on the assessment of study
design and risk of bias, as well as statistical measures of uncertainty.
For example, you may have a small study that was poorly done and reported, but
it shows a very strong effect.
Then you don't want to put a lot of confidence in that particular study
because of your concerns about the risk of bias, weaning that study.
How to frame analysis plan?
Well, analysis plan should follow from the aims of the review.
Depending on the aim of the review, you will use different analysis techniques.
For example, if the goal is to obtain size of effect from similar studies estimating
the same effect, and you have only a pair of interventions you're comparing.
For example a mechanical verses manual chest compressions for cardiac arrest.
And then you will choose standard pair-wise meta-analysis, and
that's the type of analysis we will focus today.
You may have more than two interventions for the same condition.
For example, we have four different classes of drug for
primary open angle glaucoma.
And more than 16 different eye drops weaning those four classes.
And if the goal is to draw inference about the comparative effectiveness of
all these interventions for
primary open angle glaucoma, a pair wise meta-analysis won't be very helpful.
And in that situation you would use network meta-analysis, and
we will talk about network meta-analysis in a separate lecture.
You may also want to evaluate the relationship between the size of an effect
and some characteristics of the study.
For example, is the efficacy of the BCG vaccine, which is the vaccine used to
prevent tuberculosis, associated with the latitude?
And that means, if the study was conducted in the higher latitude,
the BCG vaccine may be more or
less effective, comparing to studies conducted in the lower latitude.
And in that type of situation,
you will use meta-regression to look at the association.
And again, we will talk about meta-regression in a separate lecture.
There could be other aims for your analysis, but overall,
remember that your analysis plan has to follow from the aim of the review.
And depending on the objective of the review you will choose different
meta-analytical technique.
And in the first section of this lecture we have talked about analysis plan and
why it's important to have a plan.
And to do qualitative synthesis before you move
onwards to the quantitative synthesis.
In the next section, we will talk about meta-analysis and
a brief introduction to it.
