html xmlns=”http://www.w3.org/1999/xhtml” xmlns:mml=”http://www.w3.org/1998/Math/MathML” xmlns:epub=”http://www.idpf.org/2007/ops”>
Many measurements in clinical studies have a subjective component that may be influenced by knowledge of the participant’s treatment. Blinding, also called masking, is a process that is specified in the study protocol to keep such information secret from the participants themselves, the physicians and other clinical assessors, the laboratory personnel, and even the statisticians. Blinding may occur on several levels, from none (an open-label or unblinded study), to single-blind (only the participant or the assessors are blinded), to double-blind (both the participants and the assessors are blinded), to studies in which everyone working with the participants or with the data is blinded (completely blind). The double-blind study is the standard for interventional studies. In this chapter we discuss different levels of blinding in interventional studies and how they may be part of a study design. In the next chapter we give specific methods for implementing and maintaining the blind. Although blinding is most commonly associated with randomized studies, it should also be used in observational studies, as assessor’s knowledge of a participant’s clinical status or prior exposures can bias results. Blinding in observational studies is discussed in Chapter 28.
22.1 Why Blinding Is Used
In the ideal world of scientific research, all outcomes would be measured by precise and reproducible criteria and judgments would be completely objective. However, we do not live in an ideal world. Many measurements in clinical studies have a subjective component, such as an assessor’s assessment of a participant’s mood, the participant’s assessment of her well-being, or a relative’s assessment of the participant’s level of functioning. For brevity, we will use the term “assessors” to refer to all those who evaluate or interact with the participant. This may include examining physicians, nurses, psychologists, nutritionists, exercise therapists, and laboratory personnel.
Even objective tests, such as muscle strength, depend on how hard the participant tries to do the test. It is difficult for anyone, either the assessors or the participants, to be completely objective when they have strong ideas or hopes about what the effects of a treatment will be. Similarly, a participant may not be able to assess his physical or psychological responses objectively if he is aware of what treatment he is getting. Thus, the results of a study can be biased if either the participant or the assessors know what treatment the participant is getting. To prevent this, we use masking so that the information is not available to bias the participant or the assessor. Blinding is one of the most important ways to avoid bias in any comparative study. Even if you are convinced that you and your associates can be completely objective in a study, if you choose not to use masking whenever possible, you will have difficulty convincing reviewers, referees, and readers of the validity of your study.
22.2 The Hierarchy of Blinding
Often it is not possible to mask everyone in a study. There is a natural hierarchy of blinding, from the open-label study with no blinding to the completely blinded study in which, in addition to the participants and all assessors, the data management staff and biostatisticians are also blinded, even though they do not do any participant assessments.
In interventional studies the intervention is masked. In the following sections we begin with the double-blind study, which is the standard for interventional studies and should be used whenever possible. The completely blinded study is an extension of the double-blind study. We follow with examples of single-blind studies and finally open-label studies that have no masking at all. These are presented with the understanding that you must deal with the limitations of these approaches, but we emphasize that efforts to increase masking in a study will greatly benefit the reliability and credibility of the results. Methods for implementing blinding are discussed in the next chapter.
22.2.1 Double-Blind Studies
In a double-blind study neither the participants nor the assessors know the treatment the participant is receiving. This is the standard for an interventional study, since it minimizes both potential participant biases and potential assessor biases. It should be used whenever possible, which is whenever it is ethically permissible to mask the participants.
In a classic double-blind study comparing a new oral treatment for Type II diabetes to an existing treatment, the treatments are both administered as identical white tablets. Participants are in the study for 1 year. The primary outcomes are monthly average morning glucose levels and periodic HbA1c measurements. Participants measure and record their own morning glucose levels. Blood samples are drawn for insulin, glucose levels, and HbA1c as well as safety measures at monthly visits. The participants are also assessed for ophthalmologic problems at the beginning, at six months and at the end of the study, and have complete physical examinations at the beginning and last visit. All the laboratory tests use standard methods, and the samples are identified only by a study code. The persons assessing the ophthalmologic status of the individual and the physical examination are also masked to the participant’s treatment.
Sometimes one or more of the assessors cannot be masked if the treatment would be obvious from some of the laboratory tests or if treatment knowledge is needed to ensure appropriate handling of side effects. The study may still be considered double-blind if these unmasked individuals are not involved in assessing the endpoints of the study.
In a placebo-controlled study comparing two treatments for depression, both participants and assessors were blinded to treatment. It was necessary to monitor drug levels to make sure the participant was achieving therapeutic levels and to adjust drug dosage accordingly, and it was clear from the magnitude of the levels what the medication was. The samples of treatment drugs were given a separate drug sample number that was linked to the participant’s study number. The drug levels were assayed by an outside commercial laboratory and the results were sent to one individual to review. The person who reviewed the drug levels was only aware of the participant’s drug sample number and did not interact with participants at all or perform any assessments. The pharmacist was given all changes only by the drug sample number that could be linked to the study number. Thus the study was double-blind.
In Example 22B, as in many studies, the pharmacist had no contact with the participants or their data, knew only the drug sample number, and was not an assessor. Therefore, it was not necessary that the pharmacist be masked to the medication.
Assay results must be checked for values beyond known limits or unreasonable differences between replicates. The discrepancies may be flagged by a computer system, but the report must be reviewed by a human. Discrepant values can be accepted, deleted, or recommended to be assayed again. All laboratories have standard quality control procedures to handle such decisions for routine tests. For specialized assays done by members of the study team, the review (like the assays themselves) needs to be done by someone blind to the treatment given to each participant. Assays should be identified by a laboratory number that is different from the participant’s study number. If the information in the assay values does not reflect the treatment, then a member of the study team could do the review, otherwise an individual not associated with the study should be the one to make decisions about the laboratory data to maintain the blind.
22.2.2 Complete Blinding
A study is completely blinded when all personnel who evaluate the participants or analyze the data are masked to the participant’s treatment. This is an expansion of the double-blind study described in the previous section. The additional masking would refer to personnel who do not see the participant or create data, such as the data management and statistical staff. Although data management is probably the most objective activity in a study, blinding these staff members prevents their inadvertently disclosing the participant’s status to another staff member. In addition, it ensures that data from all participants receive equal scrutiny. The study plan sometimes requires that statisticians make decisions about the direction of the analysis after the data is collected, and thus the statisticians should be masked to ensure that the analysis is not directed toward a preferred conclusion.
Statisticians usually have to be able to identify participants as members of different groups (e.g., group A, B, C, etc.), but they do not need to know which treatment is associated with each letter. In practice, at some point, if the treatment is effective, it may become clear which group is which, but this should not be used to modify the data analysis.
Stay updated, free articles. Join our Telegram channel
Full access? Get Clinical Tree
