Send to:

Choose Destination
See comment in PubMed Commons below
Stat Med. 1996 Jun 30;15(12):1299-306; discussion 1307-11.

The role of meta-analysis in monitoring clinical trials.

Author information

  • 1Department of Epidemiology and Biostatistics, Memorial Sloan-Kettering Cancer Center, New York, N.Y. 10021, USA.


The randomized clinical trial is the gold standard methodology for evaluating treatment interventions. The randomized trial is a prospective experimental study, the embodiment of the scientific method applied to the clinic. Both organizational and ethical considerations encourage the use of formal criteria for termination. Such criteria, typically based on an overall 5 per cent significance test, are frequently portrayed as guidelines, but none the less usually succeed in focusing the debate on termination decisions and thus facilitate decision making. The use of the 5 per cent criterion is entirely arbitrary and is not based in any way on optimality considerations. No credible model for optimizing the decision to terminate trials has been developed. Meta-analysis is an evolving technique for evaluating the totality of the evidence on general treatment strategies. It has enjoyed considerable success in influencing medical opinion and treatment practice in selected areas. Its role is in establishing proof of concept of general treatment strategies, and is thus a consensus builder, making use of data from trials which may be heterogeneous in the details of treatment administration and patient selection, but similar in conceptual intent. The level of evidence required for consensus is not well understood and cannot be defined precisely. The retrospective nature of meta-analysis and the risks of biased data acquisition mandate a conservative approach to data analysis and interpretation. Randomized trials and meta-analyses have distinct but complementary goals. Meta-analysis can be used productively in planning new clinical trials, and in supplying updated information to study monitors in the course of a trial. However, it is not advisable to employ meta-analysis in formal statistical procedures for terminating clinical trials.

[PubMed - indexed for MEDLINE]
PubMed Commons home

PubMed Commons

How to join PubMed Commons

    Supplemental Content

    Loading ...
    Write to the Help Desk