Blinding (sometimes called masking) refers to the process by which study participants and personnel, including people assessing outcomes, are kept unaware of intervention allocations after inclusion of participants into the study. Blinding may reduce the risk that knowledge of which intervention was received, rather than the intervention itself, affects outcomes and assessments of outcomes.
Different types of participants and personnel can be blinded in a clinical trial (Gøtzsche 1996, Haahr 2006):
Participants (e.g. patients or healthy people);
Healthcare providers (e.g. the doctors or nurses responsible for care);
Outcome assessors, including primary data collectors (e.g. interview staff responsible for measurement or collection of outcome data) and any secondary assessors (e.g. external outcome adjudication committees);
Data analysts (e.g. statisticians);
Manuscript writers.
Lack of blinding of participants or healthcare providers could bias the results by affecting the actual outcomes of the participants in the trial. This may be due to a lack of expectations in a control group, or due to differential behaviours across intervention groups (for example, differential drop-out, differential cross-over to an alternative intervention, or differential administration of co-interventions). If participants, providers or outcome assessors are aware of assignments, bias could be introduced into assessments of outcome, depending on who measures the outcomes. If data analysts and writers are unblinded, reporting biases may be introduced. In assessing blinding in Cochrane reviews, the emphasis should be placed on participants, providers and outcome assessors. Given the overlapping considerations when participants or healthcare providers are also those assessing outcomes, we consider all types of participants jointly in assessing risk of bias.
In empirical studies, lack of blinding in randomized trials has been shown to be associated with more exaggerated estimated intervention effects, by 9% on average, measured as odds ratio (Pildal 2007). These studies have dealt with a variety of outcomes, some of which are objective. The estimated effect has been observed to be more biased, on average, in trials with more subjective outcomes (Wood 2008). Lack of blinding might also lead to bias caused by additional investigations or co-interventions regardless of the type of outcomes, if these occur differentially across intervention groups.
All outcome assessments can be influenced by lack of blinding, although there are particular risks of bias with more subjective outcomes (e.g. pain or number of days with a common cold). It is therefore important to consider how subjective or objective an outcome is when considering blinding. The importance of blinding and whether blinding is possible may differ across outcomes within a study. Seemingly objective assessments, e.g. doctors assessing the degree of psychological or physical impairment, can also be somewhat subjective (Noseworthy 1994).
Blinding can be impossible for at least some people (e.g. most patients receiving surgery). However, such studies can take other measures to reduce the risk of bias, such as treating patients according to a strict protocol to reduce the risk of differential behaviours by patients and healthcare providers.
An attempt to blind participants and personnel does not ensure successful blinding in practice. Blinding can be compromised for most interventions. For many blinded drug trials, the side effects of the drugs allows the possible detection of which intervention is being received for some participants, unless the study compares two rather similar interventions, e.g. drugs with similar side effects, or uses an active placebo (Boutron 2006).
In blinded studies, especially placebo-controlled trials, there may be concern about whether the participants truly were blinded (and sometimes also whether those caring for the patients were). Several groups have suggested that it would be sensible to ask trial participants at the end of the trial to guess which treatment they had been receiving (Fergusson 2004, Rees 2005), and some reviews of such reports have been published (Fergusson 2004, Hróbjartsson 2007). Evidence of correct guesses exceeding 50% would seem to suggest that blinding may have been broken, but in fact can simply reflect the patients’ experiences in the trial: a good outcome, or a marked side-effect, will tend to be more often attributed to an active treatment, and a poor outcome to a placebo (Sackett 2007). It follows that we would expect to see some successful ‘guessing’ when there is a difference in either efficacy or adverse effects, but none when the interventions have very similar effects, even when the blinding has been preserved. As a consequence, review authors should consider carefully whether to take any notice of the findings of such an exercise.