Bias in assignment to clinical trialsBMJ 1994; 308 doi: https://doi.org/10.1136/bmj.308.6933.920c (Published 02 April 1994) Cite this as: BMJ 1994;308:920
- J Galende,
- J Soto,
- J A Sacristan
- General Direction for Drugs and Sanitary Products, Ministry of Health, 28014 Madrid
- Clinical Research Department, Abbott Laboratories, Madrid
- Clinical Research Department, Lilly SA, Madrid.
EDITOR, - Despite the many biases to which clinical trials are subject, trials are still considered the best way to show the efficacy of a new therapeutic or prophylactic measure.1 Perhaps the most important biases are those of selection and measurement, as they most determine the validity of a study. The respective techniques of random assignment to study treatment groups and of blind experimentation are attempts to remove these biases.
We have detected a type of selection bias that does not affect the assignment of subjects in a single study to different groups but rather the assignment of patients to different clinical trials. In other words, one patient meets the criteria for inclusion in parallel trials being carried out in the same department or by the same researcher. Assignment to either study is not randomised but depends on conscious or unconscious factors that are not in any way controlled and therefore may affect the final results.
Clinical trials concerning diseases such as cancer and AIDS for which there is no curative treatment and which confound help oriented and experimental practices are most likely to present this type of bias.
This phenomenon would deserve no further mention were it not for the fact that clinical trials of several different drugs are being performed by the same researcher or research group, each trial having similar inclusion and exclusion criteria. We will not analyse how this arises but we would like to point out some of its consequences. The researcher might think that some of the drugs in one study are more effective than those being tested in others, which means that he or she might include patients whose condition is most critical in that study. Or, he or she may find that the drugs in one study are more appropriate for treating patients with certain characteristics and only include patient with them. In both cases the external validity of the study is compromised as the study group is not representative of the population to which the conclusions were intended to apply.
One possible solution to the problem would be random assignment of patients to different studies. This would create considerable practical difficulties, however, as studies are not usually proposed or initiated by researchers all at once. Perhaps the only feasible alternative for researchers is to promote awareness of the importance of this fact and to comply with standards of good clinical practice, which state that researchers are responsible for providing the sponsor with a list of the human and technical resources at their disposal, recruitment possibilities, commitments obtained, etc.2 In addition, the possibility of being audited and inspected by health authorities, as the Food and Drug Administration in the United States has been doing,3 is another inducement worth considering.