Format

Send to

Choose Destination
See comment in PubMed Commons below
Rev Recent Clin Trials. 2006 Jan;1(1):81-6.

A review of methods for ensuring the comparability of comparison groups in randomized clinical trials.

Author information

1
National Cancer Institute, Biometry Research Group, Bethesda, MD 20892-7354, USA. bergerv@mail.nih.gov

Abstract

While different design features of medical studies ostensibly serve different functions, many fall under the umbrella of methods aimed at ensuring the comparability of the comparison groups. Randomization rightly occupies the top spot in the hierarchy of design types, as it eliminates some biases (that is, systematic differences in comparison groups) that no other design can claim to eliminate. It is often assumed, and sometimes even asserted explicitly, that randomization by itself suffices to ensure that the comparison groups are sufficiently comparable that they would differ only randomly, but two points need to be made in this context. First, the assertion is not true. Second, even if it were true, it would still not be a cause for complacency, because even random baseline imbalances can wreck havoc on the valid interpretation of randomized clinical trials. Additional methods, beyond randomization, are therefore seen to be essential to the design of a good randomized clinical trial. Such methods include masking, allocation concealment, restrictions on the randomization, adjustment for prognostic variables, and the intent-to-treat approach to data analysis. Masking aims to ensure that those individuals in any one group formed by randomization are treated as similarly as possible subsequent to randomization as those in any other group formed by randomization. In contrast, allocation concealment and restricted randomization aim to create groups that start off as comparable. Adjustment for prognostic variables aims to change the comparison groups themselves to make them comparable. For example, one might find gender to be both predictive of outcome and unbalanced across treatment groups, and so one would compare the treatment groups not overall but rather first only among females and second only among males. The intent-to-treat approach aims to keep similar groups similar by not allowing for patient selection based on post-randomization outcomes (including failure to comply with the protocol). The key to understanding masking, allocation concealment, and randomization is to recognize that none of them are binary phenomena, even though they are often incorrectly understood to be. So one must question how these methods are actually carried out, rather than contenting oneself with the vague statement that these methods were performed. This review will shed light on the distinction between the process and the outcome of each of these methods (masking, allocation concealment, and randomization), and will also consider issues related to adjustment for prognostic covariates.

PMID:
18393784
[Indexed for MEDLINE]
PubMed Commons home

PubMed Commons

0 comments
How to join PubMed Commons

    Supplemental Content

    Full text links

    Icon for Bentham Science Publishers Ltd.
    Loading ...
    Support Center