1 / 141

Meta -分析的统计学基础

Meta -分析的统计学基础. 王洪源 北京大学公共卫生学院 流行病学与卫生统计学系 why_w@sohu.com. 一、 概 述. 什么是循证医学 Evidence-based medicine (EBM) 循证医学 是在对个体病人制定临床决策时应基于当前最佳的科学研究成果。 循证医学 是最佳的证据、临床经验和病人价值的有机结合。. 用 X 线进行乳腺癌筛查 Mammography for breast cancer is an established screening method.

jaden
Télécharger la présentation

Meta -分析的统计学基础

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. Meta-分析的统计学基础 王洪源 北京大学公共卫生学院 流行病学与卫生统计学系why_w@sohu.com

  2. 一、 概 述

  3. 什么是循证医学 • Evidence-based medicine (EBM) • 循证医学是在对个体病人制定临床决策时应基于当前最佳的科学研究成果。 • 循证医学是最佳的证据、临床经验和病人价值的有机结合。

  4. 用X线进行乳腺癌筛查Mammography for breast cancer is an established screening method • Is screening with mammography justifiable? • Gotzsche & Olsen [Nordic Cochrane Centre] conducted a systematic review in 2000 and updated it in 2001. • They identified 8 large RCTs on this topic, with over 182,000 women randomized

  5. The authors found that no trial data were of high quality • Two were of medium quality, and the rest were poor quality or flawed. • When the results of the two medium quality trials were combined, the risk ratio was 1.00 (95% CI 0.96, 1.05) • They concluded that “screening for breast cancer with mammography is unjustified”

  6. The US Preventive Services Task Force reviewed the same set of trials Recently, a 2001 Cochrane Collaboration review of the same trials concluded that six of the eight trials were "flawed" or of "poor quality" and that the pooled results from the remaining two better trials did not support a benefit from mammography.

  7. The meta-analysis performed for the USPSTF on the most current published data found that the pooled effect size of the combined trials was sizable and statistically significant: • the summary relative risk (RR) of breast cancer death among women randomized to screening in seven trials that included women older than 50 was 0.77 (95 percent CI, 0.67-0.89).

  8. The USPSTF recommends screening mammography, with or without clinical breast examination, every 1-2 years for women aged 40 and older.

  9. 被动吸烟的危害 • A topic of great debate and controversy for many years • First few epidemiologic studies were published in 1918

  10. Hackshaw et al. conducted a very comprehensive systematic review in 1997 • They identified 37 published studies that reported risk of lung cancer among lifelong non-smoking women according to the husband’s smoking status

  11. Their meta-analysis revealed that the overall risk of lung cancer among lifelong non-smoking women was 1.24 times higher when their husbands smoked, as compared to those women whose husbands did not smoke.

  12. 什么是循证医学 利用发表的文献证据解决临床问题对证据进行严格分级,提供指导对医学文献评价分级的实践指南,并充分考虑病人的需求和意愿解决具体临床问题。 但这绝非否认从古至今传统决策中亦在部分使用上述方法。

  13. 循证方法的两个关键方面 • 证据应从最强(如随机试验及随机试验的系统评价)到最弱(如:运用生理学指标且非系统的临床观察)分级。 • 医疗决策仅靠证据远远不够,还须考虑病人的价值取向和意愿。

  14. 循证医学证据的分级 • 一级:所有RCT的系统评价/Meta-分析。 • 二级:单个样本量足够大的RCT。 • 三级:设有对照组但未用随机方法分组的研究。 • 四级:无对照的系列病例观察。 • 五级: 专家意见。

  15. 系统性综述 • Systematic reviews(系统评价) • 是循证医学重要的手段。 • 是根据某一具体的临床问题,采用系统、明确的方法收集、选择和评估相关的临床原始研究,筛选出合格者并从中提取和分析数据,为疾病的诊治提供科学的依据。

  16. 循证医学与传统医疗实践的四个重要区别(1992 JAMA) • 系统收集的证据优于非系统的临床观察 • 以病人终点结局为判效指标的试验优于仅根据生理学原理制定指标的试验 • 解释医学文献对医生是一项重要技能,有必要正规学习一些证据的相关通则,以达到熟练解释的程度 • 医生对病人的个体化评价优于专家意见。

  17. Meta - analysis • 是在系统性综述时为了合并多个独立的研究结果,所使用的统计方法。 • 可以将针对同一问题的,多个独立的研究结果进行定量分析。 • 目前,国外文献常常将系统评价与Meta-分析交叉使用。 • All systematic reviews are not meta-analyses!

  18. Systematic reviews/meta-analyses indexed in PubMed – 10 years

  19. How to read a systematic review? • 研究显示 • Not all SRs are truly systematic (Petticrew 1999; Jadad 1998) • The quality of SRs are highly variable • (Mulrow 1987; McAlister 1999) • Cochrane reviews, on average, may be more rigorous and better reported than journal reviews • (Mulrow 1987; McAlister 1999) • However, recent studies show that even Cochrane reviews have methodological problems • (Olsen 2001; Shea 2002)

  20. Evaluation of quality of primary studies sets systematic reviews apart from traditional reviews • Empiric research shows that not all SRs assess study quality: • 240 SRs from journals: 48% assessed quality (Moher 1999) • 480 SRs in DARE: 52% assessed quality (Petticrew 1999) • 50 SRs on asthma: 28% reported validity assessment criteria (Jadad 2000)

  21. Berkeley, 2002 on quality of SRs on HIV, published during 2001 • Quality assessment done in 56% of reviews • Testing for heterogeneity done in 56% of reviews • Not all SRs with significant heterogeneity explored reasons for it • Many reviews did not state the rationale for choice of models used for combining data • Only 12% of reviews evaluated publication bias

  22. 近年Meta分析文献中存在的 主要问题(2002年到2004年5月) ㈠提供Meta分析流程图 ㈡文献搜索范围 ㈢发表偏倚的说明 ㈣异质性检验执行情况 ㈤敏感度分析

  23. Meta分析流程图 • 仅有2篇文献有进行流程图的描述,这种情况的发生可能与国内期刊并不要求等因素有关。

  24. 异质性检验执行情况 • 本次调查搜集所得文献中有88.0%的文献进行了异质性评价。但是其中30.2%的文献仅仅是提及了异质性评价,说明异质性评价的重要性并未引起国内研究者的普遍重视。 • 只有22.0%的国内研究者尝试对异质性来源进行了解释,相对的大部分的作者都未进行这个项目。

  25. 发表偏倚的说明及评估 • 仅有11.8%的文献提及发表偏倚及偏倚的评估方法,其中有0.9%的文献采用失安全系数发进行评价。 • 发表偏倚还有其他的评估方法,如漏斗图估计、线性回归法估计等,但是执行情况和失安全系数几乎相同。

  26. 因此当你找到所需的系统性综述时,你有责任去评价它的质量。因此当你找到所需的系统性综述时,你有责任去评价它的质量。

  27. 系统性综述与Meta-analyses的质量评价标准 • QUOROM评价标准(the quality of reporting of Meta-analyses)。 • 适用于随机对照试验性研究Meta分析 • MOOSE评价标准 • 适用于观察性研究Meta分析

  28. 二、 Meta analysis原理和基本思想

  29. 在用样本信息推断总体参数时,是存在抽样误差的,并且抽样误差的大小与样本量的大小有关。 • 统计学用抽样分布的理论来描述样本统计量的变化规律。

  30. 从一个均数为1.5,标准差为0.7的正态总体中进行随机抽样,样本量分别为20,50,100,200,300,500,1000,不同的样本量均进行20次抽样,共得到140个样本。从一个均数为1.5,标准差为0.7的正态总体中进行随机抽样,样本量分别为20,50,100,200,300,500,1000,不同的样本量均进行20次抽样,共得到140个样本。 分别计算每个样本的均数,标准差和标准误。 以样本的均数为横坐标,以样本量为纵坐标作散点图

  31. 从均数为1.5,标准差为0.7的正态总体的140次随机抽样结果从均数为1.5,标准差为0.7的正态总体的140次随机抽样结果

  32. 由于抽样过程存在抽样误差,样本量较大时抽样误差较小。由于抽样过程存在抽样误差,样本量较大时抽样误差较小。 • 所以从图中可以看出所有的点以样本量较大时的均数为轴,左右是基本对称的。 • 用方差分析的方法对不同的样本所对应的总体均数是否相等进行检验,方差分析的结果为F=0.862,P=0.878。 • 说明140个样本对应的总体均数是相同的。 • 用这140个样本的信息来估计总体的均数和标准差,=1.501,=0.699。这样做的结果是提高了估计的精度。

More Related