There were two key points in this Nature Review paper:
1. Meta-analysis should be used for the original purpose of the
reviewed studies, not other purpose such as the adverse effects.
2. There are many biases in meta-analysis, including the selection of
the studies, the exclusion of studies of zero outcomes, and the method
of inverse variance.
My defense of meta-analysis is as following:
1. Although certain outcomes were not the purpose of the study,
investigators did collect the data without bias. Such as the adverse
outcomes of medication were always collected in clinical trials. Those
data were as valid as the major outcomes. I see no reason why those
data can not be used for meta-analysis.
2. There are also many biases in RCT, such as selection bias,
performance bias, detection bias, and attrition bias. However, we
admire the credibility of RCT because there are methods to avoid those
biases.
There are many biases in meta-analysis as well, such as the bias of
selection of the studies, publication bias, language bias. There are
also methods to avoid those biases, such as duplicated reviews, funnel
plots and comprehensive registration of the trials.
It is not the question whether these are biases in meta-analysis, but
the question whether the quality of the individual systemic review is
good enough to avoid biases.
I also add some comments about the critics of Nature Review on
Nissen¡¦s NEJM paper of Avandia. It is true Nissen¡¦s paper was not a
perfect meta-analysis paper. There were no duplicated reviewers to
select the trials, there was not a funnel plot to examine the
publication bias.
However, he did some correct things. He used Peto method, instead of
the inverse variance method mentioned in Nature review. Peto method is
more robust for outcomes with very low incidence, such as the
myocardial infarction in Avadia¡¦s studies. He examined the
heterogeneity of the trials and found no heterogeneity among trails,
therefore the fixed effect model was suitable for the analysis.
I would say he did a not perfect but good job. Most of the critics I
learned on this paper were not credible, including this paper in
Nature Review.
I would like to listen to your comments on my opinions
--
Guo, Fred M.D.
Deaprtment of Family Medicine
National Taiwan University Hospital
¤Þz "Dr. Carlos Cuello" <[log in to unmask]>:
> I would like to hear your opinions on this article published in Nature
> Reviews. It begins in page 2
>
> --
> Carlos A. Cuello-Garc'ia, MD, Doctoral Candidate
> Director, Centre for Evidence-Based Practice-Tecnologico de Monterrey
> Cochrane-ITESM coordinator. Professor of Paediatrics and Clinical Research
> Avda. Morones Prieto 3000 pte. Col. Doctores. CITES 3er. piso,Monterrey NL,
> M'exico. CP64710 Phone. +52 (81) 88882154 & 2141
> www.cmbe.net
>
|