Beyond classical meta-analysis: can inadequately reported studies be included?

Roberson, Chris and Nik Idris, Nik Ruzni and Boyle, Peter (2004) Beyond classical meta-analysis: can inadequately reported studies be included? Drug Discovery Today, 9 (21). pp. 924-931. ISSN 1359-6446

PDF (An empirical comparison of meta analysis models for continuous data with missing standard deviations) - Published Version Restricted to Repository staff only Download (156kB) | Request a copy

Abstract

Classical meta-analysis requires the same data from each clinical trial, thus data-reporting must be of a high-quality. Imputation methods are used to include studies that provide incomplete information on variability and the fixed and random effects of a drug. Regression models can be used to include studies other than randomized placebo-controlled studies. In the example outlined here, the use of non-randomized single-arm studies and studies against comparator treatments has little influence on the estimation of the treatment effect in comparison with placebo, an effect that is based on the randomized placebo-controlled studies. The inclusion of other studies serves to increase the precision of the effect of the treatment compared with baseline. Although multiple imputation techniques enable a larger number of studies to be included, which will typically increase the precision of the estimated effect, a careful sensitivity analysis is also required

Item Type:	Article (Journal)
Additional Information:	4059/5556
Subjects:	H Social Sciences > HA Statistics
Kulliyyahs/Centres/Divisions/Institutes (Can select more than one option. Press CONTROL button):	Kulliyyah of Science > Department of Computational and Theoretical Sciences
Depositing User:	Dr Nik Ruzni Nik Idris
Date Deposited:	30 Oct 2011 18:44
Last Modified:	22 Nov 2011 07:14
URI:	http://irep.iium.edu.my/id/eprint/5556

Actions (login required)

View Item