(This is a draft and truncated version - for final and full version, see
Concise Encyclopedia of Biostatistics for Medical Professionals)
blinding, masking and concealment of allocation
Blinding in health and medicine is withholding the information from those involved. This can be used in a variety of research setups such as blinded reviews of the articles submitted for publication so the reviewers do not know who has conducted the study and where, and can provide their unbiased view. In case–control studies, for example, blinding the observers, where feasible, can help in obtaining unbiased data. The purpose is to look fair to those who are being assessed and to provide unbiased data as much as possible. Our concern in this section is with blinding in a clinical trial setup where fairness and unbiased data have special significance. Related topics of masking and concealment of allocation are also discussed in this section.
In a clinical trial setup, sometimes the exposure or the accompanying baseline information and sometimes the outcome are not correctly assessed. This can occur due to bias of the observer, or of the recording clerk who may classify a subject into a particular category of interest supporting one’s individual hypothesis. Observer bias can also occur if the subjects with disease are evaluated more intensively and more carefully than those without disease. In addition, there is a tendency for the subjects to respond differently depending on whether they are in the treatment group or in the control group. They may show some psychological effect when actually none is present or may conceal a real effect. They may want to switch from one group to another depending upon which group is showing better results. This can disturb the trial. To control all these biases, three precautions are taken. The first is called blinding of the subjects and the assessors, the second is called masking of the regimen and the third is concealment of allocation. These are closely related but different procedures although often confused with one another. ... ...
Single, Double and Triple Blinding
Trials with no blinding are called open trials or open label trials. In this type of trial, everybody knows who is receiving which treatment. Experience suggests that the unblinded trials are more likely to falsely show benefit of the active regimen than the blinded trials. Some subjects may show improvement or deterioration unrelated to the true effect of the treatment if they know to which group they belong. Participants in the control group may feel discriminated against if the allocation is open. Also, patients who know that they are receiving a new regimen may either exhibit increased anxiety or may have favorable expectations. In addition, subjects are more likely to seek adjunct intervention in an open trial, and more likely to dropout. If the treating clinician is aware of the upcoming allocation, (s)he could exclude patients (s)he considers unsuitable for that treatment or include patients considered specially suitable. This can introduce bias. During the course of the trial, a clinician, who knows the group assignment, is more likely to administer co-intervention and more likely to adjust dose. All these potential sources of bias can be avoided if subjects and the observers are blinded.
Single blinding refers to subjects not told about the treatment allocation. This eliminates the possibility of participants psychologically changing their response when they know that they are in a particular group. Blinding works as an insulation against such biases as the subjects are generally more committed when told in advance that they can get any of the regimens under trial and give consent to participate. Single blinding occurs automatically if the subjects are unconscious or anesthetized when a treatment is given or a procedure undertaken such as using or implanting a device, or changing the policy of intensive care. The term single blinding can also be used when the assessors are blinded but not the subjects. When this is done, this is specified upfront else the usual single blinding is assumed that the subjects are blinded and not the assessors. Although any single blinding is far from ideal, it may be unavoidable if some features or effects of treatments are apparent to trial personnel and observers but not to the subjects, or vice-versa.
If the assessor knows what treatment a particular subject is receiving, this may affect the way the questions are asked, investigations done, or interpretations made. Thus it is desirable that the assessor also is kept blind. This removes possible bias of the physicians or nurses involved in patient assessment— at least mitigates any subconscious influence on the outcome assessment. Unblinded assessments can also be unbiased when done by credible people but blinding improves the confidence in the results without much additional cost. When observers, assessors and other study-related personal are blinded, in addition to the subjects this is called double blinding. Such precaution is an important criterion for validity of the results of a trial. Double-blind randomized controlled trial is considered a gold standard for assessing the efficacy and safety of a new regimen or for any other intervention.
Sometimes even the data analyst has preconceptions, expectations and prejudices. (S)He might be interested in particular findings and can gear the analysis and interpretation accordingly. Most such biases at the data management stage occur in coding of adverse events when manually done, or in interpretation of handwriting when the case sheets are hand written. If the data analyst is aware of the treatment allocation, (s)he could be selective in whether to seek clarification when something is not clear and may impose his or her own interpretation. To avoid this, the analyst can also be kept blind about the codes. The codes are broken only after the data analysis is complete. This makes the trial triple blind. However make sure blinding the data analyst is not jeopardizing the analysis and presentation. Ii is better to specify in advance what analysis would be done and what tables would be prepared.
In a blinded trial, other than during the phase of analysis, treatment allocations should not be available to anyone, even in the form of treatment-A and treatment-B. It is also important that nobody involved in the trial knows which patients are receiving the same treatment. Blinding can be implemented by involving a third party who keeps the record that a subject has received treatment or control. Rigid coding systems, such as code X for the treatment and code Y for the placebo, should be avoided because breaking the code for one patient breaks it for the rest of the trial. This can easily happen in a medical care setup where laboratory investigations or side-effects can identify that the subjects has received active regimen or placebo.
Details of how blinding is actually intended to be implemented should be fully specified in the protocol and how was actually implemented should be stated when reporting a trial. Merely stating that blinding was done is not enough. In fact, such details should be given in the protocol itself.
Difficulties in Blinding
Blinding can be difficult and sometimes not achievable. One problem is masking as described later in this section and the second problem is its feasibility. For assessing the outcomes such as quality of life, readmissions, and falls after hip surgery, blinding is just not possible if one maneuver is keeping the patients in hospital for a specified number of days, and the other is early discharge and home rehabilitation. In most surgical interventions, the control has to be another kind of surgery, and not a placebo. A sham surgery may be unethical many times because it exposes a patient to surgical risks. In either case, it is extremely difficult to enforce blinding in a surgical trial. The patient can be kept blind after proper consent but the surgeon definitely knows. However, a mechanism can possibly be developed wherein all assessments subsequent to the operations are done by another surgeon who does not know and cannot decipher whether the patient belongs to the test surgery or the control surgery.
If the effects of one of the treatments are such as to be apparent (e.g., marked facial flushing after taking a drug or developing different symptoms after different types of surgery), it is almost impossible to maintain the blindness. A similar problem arises if, for example, two different durations of hospitalizations are being compared.
Blinding is rarely feasible in studies requiring long follow-up. At some stage something happens that can break the code. Field trials are very difficult to be blinded except in some typical situations. Special efforts may be needed. For example, Lwegaba [1] reports a single-blinded field trial on educational material for tobacco prevention. The trial was conducted on school students and the schools were distantly located so that contact between students did not occur and the blinding could be maintained.
Some trials keep a provision of interim appraisals. This has the potential to unblind an otherwise blinded trial. Unblinding is necessary to assess whether the treatment arm is giving sufficient evidence of efficacy or of futility relative to the control or for sample reestimation. Efforts are made that this unblinding remain confined to an unconcerned group such as Data Safety and Monitoring Board and does not extend to the participants and the assessors and the blinding for them remains in effect till such time that the trial is completed.
Morality issues are attached to blinding because the information is withheld from the participants on the one hand who would be keen to know what they are getting in case an unusual side-effect appears or unusual recovery occurs, and on the other hand, i If the doctor does not know, he may not be able to take remedial measures if anything happens to the detriment of the participant. Nonetheless, this is considered acceptable so long as the subjects are fully informed about blindness and they provide consent. Also, breaking the code must be axiomatic if the care providers consider it necessary in the interest of the patients. ... ...
Masking
The term blinding applies to the human beings involved in the project and masking to the regimens and the procedures. An obvious prerequisite for maintaining blinding is that the treatments must not be distinguishable. Regimens and procedures in a trial must be disguised to look similar to the subjects and to the assessors as much as possible. This can be easily illustrated for drugs where the active drug under trial and the placebo should exactly look alike in appearance (color, size, weight, packaging, smell, etc.) and possibly in taste. Note that the treating physician would know the treatment because (s)he is giving the treatment unless the masking of the regimen is done and the regimen such as the tablets are in the envelopes for administration. For masking, both the regimens have to be administered equally often. If one is once a day and the other twice a day, the former has to be given a placebo second time each day to look like similar treatments. This looks easy in this example but consider one regimen requiring six tablets a day and the other just two. You can have four placebos but the subjects may find six too cumbersome to swallow and may decide to take three. All these may be placebos. This may be difficult to track. Thus masking is not always easy. Consider beforehand what is feasible. A classic example is two anesthetic agents, one of which is to be given 30 ml and the other 60 ml. Dilution in this case is not advisable. For masking in this situation, a placebo (say, saline) 60 ml to the first group and of 30 ml to the other groups of the same appearance is given. The order is randomized. This might look like strange as both the groups are getting placebo. This becomes necessary in this case to mask the regimen. This is called double dummy technique. ... ...
Concealment of Allocation
Allocation concealment is that the person allocating the treatments in a clinical trial does not know what treatment the next person is going to get. The allocation of the subjects to the treatments under trial is generally done with the help of opaque sealed envelopes that contain the allocation. They are opened only after the subject’s name and other identification is written on the envelope so that the treatment cannot be changed. Concealment means that the envelopes are in random sequence and the serial on the envelope is not able to reveal what treatment it contains. A third party keeps all the record of the random sequence. This prevents bias of the allocating person in choosing which subject will get what treatment. Using pharmacy as the third party is common for concealment of allocation in case of drug trials.
Do not confuse concealment of allocation with blinding. ... ...
For final and full version, see
Concise Encyclopedia of Biostatistics for Medical Professionals
Concise Encyclopedia of Biostatistics for Medical Professionals)
blinding, masking and concealment of allocation
Blinding in health and medicine is withholding the information from those involved. This can be used in a variety of research setups such as blinded reviews of the articles submitted for publication so the reviewers do not know who has conducted the study and where, and can provide their unbiased view. In case–control studies, for example, blinding the observers, where feasible, can help in obtaining unbiased data. The purpose is to look fair to those who are being assessed and to provide unbiased data as much as possible. Our concern in this section is with blinding in a clinical trial setup where fairness and unbiased data have special significance. Related topics of masking and concealment of allocation are also discussed in this section.
In a clinical trial setup, sometimes the exposure or the accompanying baseline information and sometimes the outcome are not correctly assessed. This can occur due to bias of the observer, or of the recording clerk who may classify a subject into a particular category of interest supporting one’s individual hypothesis. Observer bias can also occur if the subjects with disease are evaluated more intensively and more carefully than those without disease. In addition, there is a tendency for the subjects to respond differently depending on whether they are in the treatment group or in the control group. They may show some psychological effect when actually none is present or may conceal a real effect. They may want to switch from one group to another depending upon which group is showing better results. This can disturb the trial. To control all these biases, three precautions are taken. The first is called blinding of the subjects and the assessors, the second is called masking of the regimen and the third is concealment of allocation. These are closely related but different procedures although often confused with one another. ... ...
Single, Double and Triple Blinding
Trials with no blinding are called open trials or open label trials. In this type of trial, everybody knows who is receiving which treatment. Experience suggests that the unblinded trials are more likely to falsely show benefit of the active regimen than the blinded trials. Some subjects may show improvement or deterioration unrelated to the true effect of the treatment if they know to which group they belong. Participants in the control group may feel discriminated against if the allocation is open. Also, patients who know that they are receiving a new regimen may either exhibit increased anxiety or may have favorable expectations. In addition, subjects are more likely to seek adjunct intervention in an open trial, and more likely to dropout. If the treating clinician is aware of the upcoming allocation, (s)he could exclude patients (s)he considers unsuitable for that treatment or include patients considered specially suitable. This can introduce bias. During the course of the trial, a clinician, who knows the group assignment, is more likely to administer co-intervention and more likely to adjust dose. All these potential sources of bias can be avoided if subjects and the observers are blinded.
Single blinding refers to subjects not told about the treatment allocation. This eliminates the possibility of participants psychologically changing their response when they know that they are in a particular group. Blinding works as an insulation against such biases as the subjects are generally more committed when told in advance that they can get any of the regimens under trial and give consent to participate. Single blinding occurs automatically if the subjects are unconscious or anesthetized when a treatment is given or a procedure undertaken such as using or implanting a device, or changing the policy of intensive care. The term single blinding can also be used when the assessors are blinded but not the subjects. When this is done, this is specified upfront else the usual single blinding is assumed that the subjects are blinded and not the assessors. Although any single blinding is far from ideal, it may be unavoidable if some features or effects of treatments are apparent to trial personnel and observers but not to the subjects, or vice-versa.
If the assessor knows what treatment a particular subject is receiving, this may affect the way the questions are asked, investigations done, or interpretations made. Thus it is desirable that the assessor also is kept blind. This removes possible bias of the physicians or nurses involved in patient assessment— at least mitigates any subconscious influence on the outcome assessment. Unblinded assessments can also be unbiased when done by credible people but blinding improves the confidence in the results without much additional cost. When observers, assessors and other study-related personal are blinded, in addition to the subjects this is called double blinding. Such precaution is an important criterion for validity of the results of a trial. Double-blind randomized controlled trial is considered a gold standard for assessing the efficacy and safety of a new regimen or for any other intervention.
Sometimes even the data analyst has preconceptions, expectations and prejudices. (S)He might be interested in particular findings and can gear the analysis and interpretation accordingly. Most such biases at the data management stage occur in coding of adverse events when manually done, or in interpretation of handwriting when the case sheets are hand written. If the data analyst is aware of the treatment allocation, (s)he could be selective in whether to seek clarification when something is not clear and may impose his or her own interpretation. To avoid this, the analyst can also be kept blind about the codes. The codes are broken only after the data analysis is complete. This makes the trial triple blind. However make sure blinding the data analyst is not jeopardizing the analysis and presentation. Ii is better to specify in advance what analysis would be done and what tables would be prepared.
In a blinded trial, other than during the phase of analysis, treatment allocations should not be available to anyone, even in the form of treatment-A and treatment-B. It is also important that nobody involved in the trial knows which patients are receiving the same treatment. Blinding can be implemented by involving a third party who keeps the record that a subject has received treatment or control. Rigid coding systems, such as code X for the treatment and code Y for the placebo, should be avoided because breaking the code for one patient breaks it for the rest of the trial. This can easily happen in a medical care setup where laboratory investigations or side-effects can identify that the subjects has received active regimen or placebo.
Details of how blinding is actually intended to be implemented should be fully specified in the protocol and how was actually implemented should be stated when reporting a trial. Merely stating that blinding was done is not enough. In fact, such details should be given in the protocol itself.
Difficulties in Blinding
Blinding can be difficult and sometimes not achievable. One problem is masking as described later in this section and the second problem is its feasibility. For assessing the outcomes such as quality of life, readmissions, and falls after hip surgery, blinding is just not possible if one maneuver is keeping the patients in hospital for a specified number of days, and the other is early discharge and home rehabilitation. In most surgical interventions, the control has to be another kind of surgery, and not a placebo. A sham surgery may be unethical many times because it exposes a patient to surgical risks. In either case, it is extremely difficult to enforce blinding in a surgical trial. The patient can be kept blind after proper consent but the surgeon definitely knows. However, a mechanism can possibly be developed wherein all assessments subsequent to the operations are done by another surgeon who does not know and cannot decipher whether the patient belongs to the test surgery or the control surgery.
If the effects of one of the treatments are such as to be apparent (e.g., marked facial flushing after taking a drug or developing different symptoms after different types of surgery), it is almost impossible to maintain the blindness. A similar problem arises if, for example, two different durations of hospitalizations are being compared.
Blinding is rarely feasible in studies requiring long follow-up. At some stage something happens that can break the code. Field trials are very difficult to be blinded except in some typical situations. Special efforts may be needed. For example, Lwegaba [1] reports a single-blinded field trial on educational material for tobacco prevention. The trial was conducted on school students and the schools were distantly located so that contact between students did not occur and the blinding could be maintained.
Some trials keep a provision of interim appraisals. This has the potential to unblind an otherwise blinded trial. Unblinding is necessary to assess whether the treatment arm is giving sufficient evidence of efficacy or of futility relative to the control or for sample reestimation. Efforts are made that this unblinding remain confined to an unconcerned group such as Data Safety and Monitoring Board and does not extend to the participants and the assessors and the blinding for them remains in effect till such time that the trial is completed.
Morality issues are attached to blinding because the information is withheld from the participants on the one hand who would be keen to know what they are getting in case an unusual side-effect appears or unusual recovery occurs, and on the other hand, i If the doctor does not know, he may not be able to take remedial measures if anything happens to the detriment of the participant. Nonetheless, this is considered acceptable so long as the subjects are fully informed about blindness and they provide consent. Also, breaking the code must be axiomatic if the care providers consider it necessary in the interest of the patients. ... ...
Masking
The term blinding applies to the human beings involved in the project and masking to the regimens and the procedures. An obvious prerequisite for maintaining blinding is that the treatments must not be distinguishable. Regimens and procedures in a trial must be disguised to look similar to the subjects and to the assessors as much as possible. This can be easily illustrated for drugs where the active drug under trial and the placebo should exactly look alike in appearance (color, size, weight, packaging, smell, etc.) and possibly in taste. Note that the treating physician would know the treatment because (s)he is giving the treatment unless the masking of the regimen is done and the regimen such as the tablets are in the envelopes for administration. For masking, both the regimens have to be administered equally often. If one is once a day and the other twice a day, the former has to be given a placebo second time each day to look like similar treatments. This looks easy in this example but consider one regimen requiring six tablets a day and the other just two. You can have four placebos but the subjects may find six too cumbersome to swallow and may decide to take three. All these may be placebos. This may be difficult to track. Thus masking is not always easy. Consider beforehand what is feasible. A classic example is two anesthetic agents, one of which is to be given 30 ml and the other 60 ml. Dilution in this case is not advisable. For masking in this situation, a placebo (say, saline) 60 ml to the first group and of 30 ml to the other groups of the same appearance is given. The order is randomized. This might look like strange as both the groups are getting placebo. This becomes necessary in this case to mask the regimen. This is called double dummy technique. ... ...
Concealment of Allocation
Allocation concealment is that the person allocating the treatments in a clinical trial does not know what treatment the next person is going to get. The allocation of the subjects to the treatments under trial is generally done with the help of opaque sealed envelopes that contain the allocation. They are opened only after the subject’s name and other identification is written on the envelope so that the treatment cannot be changed. Concealment means that the envelopes are in random sequence and the serial on the envelope is not able to reveal what treatment it contains. A third party keeps all the record of the random sequence. This prevents bias of the allocating person in choosing which subject will get what treatment. Using pharmacy as the third party is common for concealment of allocation in case of drug trials.
Do not confuse concealment of allocation with blinding. ... ...
For final and full version, see
Concise Encyclopedia of Biostatistics for Medical Professionals