Reporting of assignment methods in clinical trials

https://doi.org/10.1016/0197-2456(94)90045-0Get rights and content

Abstract

A stratified random sample of the reports of 200 clinical trials, conducted in the 1980s, listed in the February 1991 issue of Controlled Clinical Trials, was reviewed. The four strata were small single-center trials (SST), small multicenter trials (SMT), large single-center trials (LST), and large multicenter trials (LMT). The description of the assignment method in the published reports of the results of the trials was classified into four categories: good description (G), provides details; fair description (F), describes some details; poor description (P), only states that randomization was used; and no description (N). The percentage of good and fair descriptions among the four strata were:

SSTG F6 14
SMTG F6 22
LSTG F20 10
LMTG F4 22

The proportion of G or F trials is slightly higher in the LST because one third used systematic rather than random assignment, which is easier to describe. Good descriptions usually require only two or three lines of text, and yet the description of the randomization method for most clinical trials is so poor that the reader cannot tell if randomization has been applied properly.

References (6)

There are more references available in the full text version of this article.

Cited by (26)

  • Quantification methods were developed for selection bias by predictability of allocations with unequal block randomization

    2005, Journal of Clinical Epidemiology
    Citation Excerpt :

    In multicenter trials, contracts with individual investigator sites usually stipulate a fixed number of subjects to be enrolled (commonly 8 to 20), so that the study can be planned and managed properly. If the ratio of treatment arms is known (e.g., equal allocations to two treatment arms), and if the trial is not double-blind [1], or if the treatments do not resemble each other perfectly as to appearance [2] or effects, the eligibility and consent of patients may be biased by anticipating the next treatment allocation before the patients are enrolled [3–8]. For example, Peto [9] discussed the results of the Captopril Prevention project [10]: because some allocation envelopes had been opened in advance in an unknown proportion of cases, the outcome could be “distorted by foreknowledge of the next treatment allocation.”

  • When can a clinical trial be called 'randomized'?

    2003, Vaccine
    Citation Excerpt :

    With the consumer principle of randomization [9], patients may actually choose their own randomization probability from a given set of choices. So randomization can be quite broad, allowing for uneven allocation [8], adaptive probabilities [10], and even some degree of self-selection [9]. Yet not every allocation method is randomized.

  • Minimization in randomized clinical trials

    2023, Statistics in Medicine
View all citing articles on Scopus
View full text