Sources of bias in randomised controlled trials IIBMJ 2010; 341 doi: https://doi.org/10.1136/bmj.c7053 (Published 22 December 2010) Cite this as: BMJ 2010;341:c7053
- Philip Sedgwick, senior lecturer in medical statistics
- 1Section of Medical and Healthcare Education, St George’s, University of London, Tooting, London, UK
Previous questions described a cluster randomised controlled trial that assessed the long term effects of an obesity prevention programme in schools.1 2 The intervention was a focused education programme delivered over one school year that promoted a healthy diet and discouraged consumption of carbonated drinks. The control treatment was no intervention.3
The clusters were classes of children aged 7-11 years within schools. Classes were allocated to a treatment group by using a random sequence generated by a researcher not involved in recruitment. A second researcher who was unaware of the sequence recruited 29 classes from six schools. As classes were recruited they were allocated to the next treatment in the sequence. Fifteen classes totalling 325 children were randomised to the active intervention and 14 classes totalling 319 children to the control.
The main outcome measures were change in anthropometric measures such as height, weight, and waist circumference. Three years after baseline, outcome measures were obtained for 219 (67.4%) children allocated to the active intervention and 215 (67.4%) children assigned to the control.
Which of the following statements, if any, are true?
a) Allocation concealment was achieved
b) Allocation bias was minimised
c) Attrition bias was minimised
d) The study was a pragmatic trial
Answers a, b, c, and d are all true.
In this cluster randomised controlled trial, classes of children were randomly allocated to either the active intervention or the control (no active intervention). Within each class, all children undertook the allocated treatment. Allocation of classes to active intervention or control was achieved using a random sequence generated by a researcher not involved in recruitment. A second researcher unaware of the allocation sequence recruited the classes. After a class had been recruited it was allocated to the next treatment in the sequence. Treatment allocation was subsequently revealed. Given that allocation was performed blind to the recruitment of schools and classes, allocation concealment was achieved (a is true).
Allocation concealment has been described in a previous question.4 Allocation concealment was essential in this study in order to minimise selection bias. If the recruiting researcher knew the allocation sequence, he or she might have influenced, unconsciously or otherwise, those classes recruited to the trial. The researcher may have believed that some classes would not have agreed to participate, or could have been unsuitable for their allocated treatment, and might, therefore, simply not recruited these classes.
Allocation concealment would have also minimised allocation bias (b is true). The recruiting researcher was unaware of the random order of the sequence and, therefore, could not systematically influence the allocation of classes. Allocation bias would have occurred if the researcher allocated the recruited classes to active intervention or control in order to favour one of the treatment arms.
In most trials participants drop out or are lost to follow-up. Unless attrition is random in each group, attrition bias will be observed. Attrition bias occurs if the rates at which participants leave the trial, or their characteristics or reasons for leaving, differ between treatment arms. Although attrition bias can be reduced or minimised by diligent follow-up, it is rarely eliminated. The proportion of participants that provided outcome measures at three years was the same for both the active intervention group and the control group (67.4% in each group). Given that the proportion of children lost to follow-up was the same in each group, attrition bias was minimised in this trial (c is true). The reasons why children were lost to follow-up were reported to be similar for both treatment groups, although the characteristics of these children were not provided.
Clinical trials are described as explanatory or pragmatic. This study was a pragmatic trial (d is true). In pragmatic trials participants may not actually adhere to, or complete, their allocated intervention or treatment. The analysis of a pragmatic trial will always include intention to treat analyses, with participants analysed in the group to which they were originally allocated regardless of whether or not they completed or received that treatment.5 Therefore, pragmatic trials tend to assess effectiveness—that is, the benefit of the active intervention as would be observed in routine clinical practice (d is true). An explanatory trial measures the benefits of a treatment or intervention in ideal conditions, with participants remaining on their assigned treatment or intervention throughout the entire study period. The explanatory approach aims to measure the efficacy or true effect of a treatment.
Cite this as: BMJ 2010;341:c7053
Competing interests: The author has completed the Unified Competing Interest form at www.icmje.org/coi_disclosure.pdf (available on request from the corresponding author) and declares: no support from any organisation for the submitted work; no financial relationships with any organisations that might have an interest in the submitted work in the previous three years; no other relationships or activities that could appear to have influenced the submitted work.
Log in using your username and password
Log in through your institution
Register for a free trial to thebmj.com to receive unlimited access to all content on thebmj.com for 14 days.
Sign up for a free trial