Confidence intervals for a random-effects meta-analysis based on Bartlett-type corrections

被引:62
作者
Noma, Hisashi [1 ]
机构
[1] Kyoto Univ, Sch Publ Hlth, Dept Biostat, Sakyo Ku, Kyoto 6068501, Japan
基金
日本学术振兴会;
关键词
meta-analysis; random-effects model; coverage properties; Bartlett correction; efficient score; ACUTE MYOCARDIAL-INFARCTION; SMALL SAMPLE INFERENCE; COX REGRESSION-MODEL; MIXED LINEAR-MODEL; PUBLICATION BIAS; INTRAVENOUS MAGNESIUM; CLINICAL RESEARCH; LIKELIHOOD; TRIALS; REEVALUATION;
D O I
10.1002/sim.4350
中图分类号
Q [生物科学];
学科分类号
07 ; 0710 ; 09 ;
摘要
In medical meta-analysis, the DerSimonian-Laird confidence interval for the average treatment effect has been widely adopted in practice. However, it is well known that its coverage probability (the probability that the interval actually includes the true value) can be substantially below the target level. One particular reason is that the validity of the confidence interval depends on the assumption that the number of synthesized studies is sufficiently large. In typical medical meta-analyses, the number of studies is fewer than 20. In this article, we developed three confidence intervals for improving coverage properties, based on (i) the Bartlett corrected likelihood ratio statistic, (ii) the efficient score statistic, and (iii) the Bartlett-type adjusted efficient score statistic. The Bartlett and Bartlett-type corrections improve the large sample approximations for the likelihood ratio and efficient score statistics. Through numerical evaluations by simulations, these confidence intervals demonstrated better coverage properties than the existing methods. In particular, with a moderate number of synthesized studies, the Bartlett and Bartlett-type corrected confidence intervals performed well. An application to a metaanalysis of the treatment for myocardial infarction with intravenous magnesium is presented. Copyright (C) 2011 John Wiley & Sons, Ltd.
引用
收藏
页码:3304 / 3312
页数:9
相关论文
共 39 条
[1]  
[Anonymous], 2000, Methods for meta-analysis in medical research
[2]  
[Anonymous], 2015, Cengage learning
[3]  
BARNDORFFNIELSEN OE, 1988, BIOMETRIKA, V75, P374, DOI 10.2307/2336188
[4]   Properties of sufficiency and statistical tests [J].
Bartlett, MS .
PROCEEDINGS OF THE ROYAL SOCIETY OF LONDON SERIES A-MATHEMATICAL AND PHYSICAL SCIENCES, 1937, 160 (A901) :0268-0282
[5]   PUBLICATION BIAS AND DISSEMINATION OF CLINICAL RESEARCH [J].
BEGG, CB ;
BERLIN, JA .
JOURNAL OF THE NATIONAL CANCER INSTITUTE, 1989, 81 (02) :107-115
[6]   A simple method for inference on an overall effect in meta-analysis [J].
Brockwell, Sarah E. ;
Gordon, Ian R. .
STATISTICS IN MEDICINE, 2007, 26 (25) :4531-4543
[7]   A comparison of statistical methods for meta-analysis [J].
Brockwell, SE ;
Gordon, IR .
STATISTICS IN MEDICINE, 2001, 20 (06) :825-840
[8]   BARTLETT-TYPE MODIFICATION FOR RAO EFFICIENT SCORE STATISTIC [J].
CHANDRA, TK ;
MUKERJEE, R .
JOURNAL OF MULTIVARIATE ANALYSIS, 1991, 36 (01) :103-112
[9]   BARTLETT CORRECTIONS TO LIKELIHOOD RATIO TESTS [J].
CHESHER, A ;
SMITH, RJ .
BIOMETRIKA, 1995, 82 (02) :433-436
[10]  
COLLINS R, 1995, LANCET, V345, P669