Blinding is a cornerstone of robust clinical trials. It is described in the ICH E9 ‘Statistical Principles for Clinical Trials’ as being one of the most important design techniques for avoiding bias in studies, along with randomisation. This is why randomised ‘double-blind’ trials are considered to be the gold standard for assessing the effectiveness of new treatments.

Despite the significance of blinding in trial design, achieving it successfully is not always straightforward. This is because it involves a complex chain, from the manufacture and distribution of investigational products (IPs), through their being assigned to participants, to the management of data collection and timing of study analyses. Certain decisions, such as whether to use a matching placebo, can introduce extra challenges. It is, therefore, essential for all those involved in the design, conduct and analysis of trials to develop a thorough understanding of the range of factors influencing blinding.

To start, it is imperative that blinding considerations are laid out in the protocol and it is best practice to include these details in the title. The planned level of blinding must be carefully described, particularly for studies with complex dose requirements or where there are plans to unblind staff during the course of the research. In line with Consolidated Standards of Reporting Trials (CONSORT) guidance, details should be given about who is going to be blinded and who will be unblinded, with proper justifications and timing specifications detailed for the latter. This not only helps the trial itself run smoothly, but also provides security for the interpretation of study findings.

“There is no point creating a matching placebo or masking a comparator if it’s going to be packaged differently from the IP. Products must be indistinguishable in every respect.”

In some cases, setting out blinding plans in the protocol might reveal a fundamental gap in the level of blinding that is possible within the study. For example, depending on the available tablet sizes, a nominally double-blind dose-range trial might only be partially blind to certain doses. The research team should thus consider whether describing the trial as ‘double-blind’ is misleading. Careful consideration must also be given about whether titration algorithms described in the protocol might inadvertently result in unblinding during the trial. Engaging with all parties involved in the study at an early stage is essential to prevent such issues from occurring.

Chasing objectivity

Procedures for maintaining blinding and data integrity should also be considered where blinding may be compromised by treatment-induced effects. A potential solution is proposed in ICH E9, which suggests that, “blinding may be improved by blinding investigators and relevant sponsor staff to certain test results – for example, selected laboratory measures”.

In this scenario, an independent committee could adjudicate on end points. One approach is to have it review only source data relevant to a single blinded end point.

The use of objective end points where possible also helps to prevent blinding failures. In these situations, the practice of independent adjudication is considered controversial, with some researchers arguing that it has little value or that a sampling approach might be sufficient. Nevertheless, the use of some form of adjudication does enhance the perceived integrity of a trial.

For a fully binded trial, the physical characteristics of the IP must be indistinguishable from the comparator or placebo in every aspect, including taste, texture and smell. In practice, however, absolute blinding is likely impossible as, by definition, the IP for each treatment group must be different.

Challenges can arise in studies were there is an active comparator arm sourced from a commercial formulation, where multiple tablet sizes are required for different dose levels, or where co-administration of dummy regimens is required when comparing different dosing schedules.

Although the latter is often necessary to maintain blinding, ICH E9 notes that this approach “may sometimes force an administration scheme that is sufficiently unusual to influence adversely the motivation and compliance of the subjects”. Implementing such strategies, therefore, needs to be carefully considered.

Under the blindfold

One of the major considerations is, of course, packaging. There is no point creating a matching placebo or masking a comparator if it’s going to be packaged differently from the IP. Products must be indistinguishable in every respect. ICH emphasises that modifying original packaging and completely repackaging products are the most efficient approaches for achieving blinding, but this does create a large amount of waste.

It is essential that packaging suppliers keep accurate records of the placebo or comparator drugs that can be quickly accessed if a medical emergency occurs. This data should be stored in an efficient system that can be updated by different individuals as necessary, without compromising the security or validity of the data.

Unplanned blinding is never desirable, but does not necessarily compromise a study. It depends on how many participants have been unblinded, to whom and the degree to which there was the possibility of bias in subsequent trial conduct or analysis.

There are a number of different situations that could lead to unplanned unblinding, such as when container assignment patterns reveal that participants have been assigned either the same or different treatments. Those in the treatment group may have experienced serious adverse events, requiring immediate reporting, and details may be transmitted to a variety of individuals involved in the study, including those who were blinded. There can also be errors in information flow where the safety oversight committee accidentally shares information with blinded researchers.

If an unplanned unblinding event is likely to impact the safety of a trial subject or the trial’s scientific integrity, it is essential to escalate the issue to the sponsor’s compliance or qualityassurance organisation, in order to identify potential solutions, including plans for reporting to regulators. Important compromises to study blinding should be included in the study report according to CONSORT guidance about how such events should be discussed in publications.

That said, it is important to note the mitigating factors if an unplanned unblinding event occurs. In large multicentre studies, there is a relatively low risk that an issue at one site would have a large impact on the overall trial results. In addition, applications for new drug approvals usually require multiple successful studies. Single pivotal study submissions must have compelling results, further compounding the damage of even one unblinding event.

It has been suggested that to assess the success of blinding, it is useful to ask investigators or participants to guess which treatment they thought they were randomised to after the study has finished. If results indicate preferences for a certain response higher than would be expected by chance alone, this can indicate blinding failure, undermining the integrity of trial findings. Some have even suggested that this should be common practice, regardless of the level of suspicion about potential unblinding. However, such a strategy is not the consensus opinion, as discussed in the CONSORT guidance.

Randomisation central

There are some special considerations for randomised open-label or single-blind studies, including for the avoidance of selection bias. This is explicitly described in ICH E9, which says, “it is particularly important that the investigator’s knowledge of the next treatment should not influence the decision to enter the subject; this decision should precede knowledge of the randomised treatment”.

This means using ‘central randomisation’ as much as possible for open-label and single-blind randomised trials. In this approach, the individual ‘blocks’ comprising the randomisation code are open for use by all sites, rather than assigning individual blocks to each site. The main advantage of this technique is that it eliminates the potential for selection bias through foreknowledge of the randomised treatment. There are disadvantages, however, such as a reduced ability to determine the exact amount of IP needed at each site and a greater potential for imbalanced recruitment numbers across treatment groups at each site. Both of these have to be taken into consideration when determining the statistical analysis to be performed.

The use of objective end points is another helpful strategy, particularly for open-label studies, as they are much less susceptible to bias than subjective measures. Having clinical assessments made by medical staff who are not involved in treating participants and are blinded to the treatment can also be useful.

Ensuring that data is only aggregated in accordance with the protocol is essential to preserve the integrity of the trial. It can be very difficult to prevent sponsor staff with legitimate access to trial data from being unblinded to participant-level data while an open-label trial is ongoing, as certain aspects of data collection are often inherently unblinding. If inappropriate aggregation does occur, this constitutes an unplanned interim analysis and must be disclosed as such in the clinical study report.

Although blinding appears to be as simple as closing one’s eyes, there are a number of challenges to navigate throughout the trial process. It is, therefore, critical to give the topic as much thought as other aspects of a study. This should include consideration of clinical, pharmaceutical, logistical and resource factors.


Key questions: the context of blinding in a clinical trial

There are many considerations to account for when planning blinded clinical trials, at both the design and operational levels.

  1. Will a matching placebo be developed centrally or sourced at investigator sites? The latter introduces considerable complexity.
  2. Within the protocol itself, are there aspects of titration regimens or descriptions of IP characteristics that are inadvertently unblinding, and does this have implications for the extent to which the trial can be described as ‘double-blind’?
  3. Are there other aspects of supply strategy that might unblind, and could such risks be enhanced or reduced depending on the randomisation strategy?
  4. Has the study team obtained robust estimates of patient enrolment and discontinuation rates, and if study conduct assumptions are found to be flawed, could these undermine supply strategy and introduce blinding risk?
  5. Are there aspects of anticipated clinical outcome, laboratory results or adverse experience that might compromise blinding?
  6. Are there aspects of the IP’s physical characteristics that might not be as blinded as might be ideal?

Source: Contemporary Clinical Trials Communications