Brotons Cuixart C, Permanyer Miralda G. [Meta-analysis, megatrials, and clinical practice in cardiology].
Rev Esp Cardiol 1999;
52:840-50. [PMID:
10563158 DOI:
10.1016/s0300-8932(99)75011-6]
[Citation(s) in RCA: 0] [Impact Index Per Article: 0] [Reference Citation Analysis] [Abstract] [MESH Headings] [Track Full Text] [Journal Information] [Subscribe] [Scholar Register] [Indexed: 11/18/2022]
Abstract
One of the factors that raises more scepticism and concern in physicians is the combination of data from different clinical trials. In order to combine the results of different trials three requisites must be followed through: similar treatment, the patients have the same disease, and the main outcome variables must be the same. Publication bias is one important limitation of meta-analysis, and it occurs when studies with negative results are not published, which causes the effect of the treatment to be overestimated. It might seem reasonable that the demonstrated effect in the entire study population with a large sample size could be easily analyzed in different subgroups of patients, getting closer to the prediction of effect in the individual patient. This apparently obvious observation is a fallacy and subgroup analysis is often problematic. Possibly the most important condition in the analysis of subgroups is the definition of the subgroups in the design stage of the study, which can permit an adequate presight of the necessary requisites for the validity of the study. The problem of generalization of the results has been traditionally related with of extrapolation of the results to certain groups of patients not included in the clinical trial. In these cases it is important to consider other nonexperimental epidemiological studies, its internal validity and the consistency of their results. The most important characteristics that differentiate conventional clinical trials from megatrials are the following: megatrials recruit a very large and heterogenous population, with few inclusion and exclusion criteria; the only outcome variable that can be better assessed in these heterogenous conditions is eventually mortality; due to their very large sample size, results are obtained with a high level of precision (very narrow confidence intervals).
Collapse