Controlled trials: allocation concealment, random allocation, and blindingBMJ 2015; 350 doi: https://doi.org/10.1136/bmj.h2633 (Published 15 May 2015) Cite this as: BMJ 2015;350:h2633
- Philip Sedgwick, reader in medical statistics and medical education
The efficacy of nicotine patches in pregnant women who smoke was investigated using a randomised placebo controlled trial. The intervention was the administration of 16 hour nicotine patches until the time of delivery. Participants were 402 pregnant women older than 18 years who smoked at least five cigarettes a day and whose babies were between 12 and 20 weeks’ gestation. Women were recruited from 23 maternity wards throughout France. To ensure allocation concealment, the allocation sequence was randomly generated by computer by someone independent of the trial. After recruitment, participants were allocated to intervention patches (n=203) or placebo patches (n=199) according to the allocation sequence.1
The primary outcome measures were achievement of complete abstinence until delivery and birth weight. The proportion of women who achieved complete abstinence was higher in the nicotine patch group than in the placebo group, although the difference was not significant (11 (5.5%) v 10 (5.1%); odds ratio 1.08, 95% confidence interval 0.45 to 2.60). The mean birth weight was higher in the nicotine patch group, although the difference was not significant (3065 g (standard error 44 g) v 3015 g (44 g); difference 50 g, −71.1 to 172.3). It was concluded that the nicotine patch did not increase smoking cessation rates or birth weights.
Which of the following statements, if any, are true?
a) Allocation concealment ensured that the allocation sequence was not known by women before recruitment to the trial
b) Allocation concealment reduced the potential for selection bias
c) The random allocation of women minimised allocation bias
d) Double blinding minimised the potential for resentful demoralisation
Statements a, b, c, and d are all true.
The purpose of the trial was to investigate the efficacy of nicotine patches in pregnant women who smoked. Women were allocated to the intervention of 16 hour nicotine patches until the time of delivery or placebo patches. Allocation concealment, random allocation, and double blinding were used to minimise biases that result from trial participants or personnel involved in the study being aware of the treatment allocation.
Allocation concealment involved not disclosing to women and to those recruiting the trial participants the allocation sequence before recruitment to the trial (a is true). The allocation sequence—the order in which participants would be allocated to treatment once recruited—was generated randomly before the trial started. It was important that the allocation sequence was concealed because otherwise it might have been possible to influence the recruitment of women and the treatment they were allocated. If the researchers and personnel involved in recruitment knew the allocation sequence it may have influenced, unconsciously or otherwise, which women they recruited or the order in which this was done. For example, they may have believed that some women would not have accepted or would have been unsuitable for the next treatment in the sequence. Furthermore, if women were aware of the allocation sequence, knowledge of the subsequent treatment in the sequence may have influenced whether they participated in the trial. Therefore, allocation concealment reduced the potential for selection bias (b is true). Selection bias would have occurred in the above trial if there was a systematic difference between those women who were recruited and those who were not, resulting in a sample that was not representative of the patient population. Selection bias has been discussed in more detail in a previous question.2
The allocation sequence was generated randomly. Therefore, each woman had an equal probability of being allocated to the intervention patch group or the placebo patch group. Random allocation, sometimes referred to as randomisation, minimised allocation bias—that is, a systematic difference between participants in how they were allocated to the treatment group (c is true). Allocation bias has been discussed in more detail in a previous question.2 The main objective of random allocation was to minimise confounding by achieving treatment groups that were similar in baseline characteristics. Confounding is a difference between treatment groups at baseline in those factors that affect treatment and outcome measures. These include demographic characteristics, prognostic factors, and other characteristics that may influence participation or withdrawal from the trial. If confounding is minimised, differences between treatment groups in outcome at the end of the study are more likely to result from differences in treatments received and not an imbalance in baseline characteristics. In general, as sample size increases a greater balance in numbers between treatment groups is achieved and confounding will tend to be reduced.
The use of placebo nicotine patches in the above trial meant that it was possible to make the trial double blind, a general term used to indicate that neither the trial participants nor those assessing the outcome measures were aware of the treatment allocation after randomisation.3 Blinding of the participants and the researchers minimised ascertainment bias. Ascertainment bias would have occurred if the assessment of outcome measures had been systematically distorted because the investigators or participants were aware of the treatment allocation. Ascertainment bias is sometimes referred to as detection bias.
Ascertainment bias would have occurred, for example, if the researchers favoured the nicotine patches and wanted to show that they were effective in helping women abstain from smoking during pregnancy. If the investigators were aware of the treatment allocation, they could have been biased in their assessment of the outcomes, subconsciously or otherwise, towards the intervention treatment. Ascertainment bias would also have occurred, for example, if women knew their treatment allocation; women may have had a preference for the intervention and if allocated the nicotine patches they may have been biased, subconsciously or otherwise, towards the intervention in their reporting of the outcomes. If ascertainment bias occurs it can result in an exaggerated difference between the treatment groups in outcome. When ascertainment bias occurs on behalf of the investigators it is called assessor bias, and when it occurs on behalf of the participants it is called response bias.
Double blinding in the above trial minimised the potential for resentful demoralisation (d is true). If women knew their treatment allocation they might have been disappointed, for example, if they had been allocated to the control group because they might have felt less cared for as they were not receiving the active intervention. Women in the control group might therefore have been less likely to complete treatment and might have withdrawn from the trial.
Allocation concealment, random allocation, plus the blinding of participants and all research personnel are seen as essential components in a clinical trial. The use of such methodology in trial design minimises biases that result from trial participants or research personnel being aware of the treatment allocation. When these methods are used, differences between treatment groups in outcome at the end of a study are likely to be the result of differences in treatment received, thereby permitting the inference of causation due to treatment to be made. Allocation concealment and random allocation are essential if blinding of the participants and research personnel is to be achieved. The success of random allocation depends on consecutively recruited patients being allocated to treatment groups according to a randomly generated allocation sequence. However, blinding cannot always be achieved in trials owing to the nature of some intervention and control treatments.4
Cite this as: BMJ 2015;350:h2633
Competing interests: None declared.