Single Blinding: Participants, Investigators, or Assessors 184
Implementation Issues: Participants 185
Sham Procedures 186
Implementation Issues: Assessors 187
Double Blinding: Participants, Investigators, and Assessors 187
Implementation Issues: If No Effective Treatment Exists 188
Implementation Issues: If an Accepted Effective Treatment Exists 188
The Importance of Placebos 190
This chapter focuses on implementation of blinding in randomised trials. Investigators use single blinding frequently with nonpharmaceutical treatments where total blinding is not possible. In those circumstances, often participants are those blinded and the treatments involve surgery or a medical device. The concept is to standardise the experiences of the participants among all the treatment groups throughout the trial. Some trials just use simple visual screening such as draping. Other trials might use complicated approaches such as sham operations in which skin incisions are made and then closed. If healthcare personnel or participants cannot be blinded, investigators should always consider blinding the outcome assessor. That form of single blinding can frequently be done.
Most double blinding involves drugs, which include vaccines and other medicinal interventions. Placebos are generally used in double-blind trials. Hospital pharmacies and pharmaceutical companies may provide valuable assistance in these blinding activities. When examining the effects of a proposed new treatment for a condition for which no effective treatment already exists, the drug and placebo must be prepared identically, such as in capsules or pills, and packaged identically, such as in bottles or blister packs. They should appear, taste, and smell the same while also being administered identically. The experiences of the control group should mimic the experiences of the treatment group.
If an accepted effective treatment exists, sometimes ethics review boards or investigators determine that it should be administered to everyone enrolled in a trial. In that circumstance, the trial compares a new treatment plus the accepted effective treatment versus the placebo plus the accepted effective treatment. Frequently, however, if an effective standard treatment exists, it is used in the control group for comparison against a new treatment. Thus trialists compare two active treatments. In that circumstance, they usually have three options to double blind their trial. The research pharmacy could obtain the drugs in raw form and prepackage the two drugs equally, encapsulate the two standard formulation drugs in a larger capsule, or utilise a double-dummy, double-blinding approach. Frequently, investigators will encounter fewer implementation difficulties with the simpler double-dummy approach.
Many potential benefits derive from participants, investigators, and outcome assessors not knowing the group to which the participants have been assigned. This is commonly referred to as blinding ( Chapter 16 ). It can reduce differential assessment of outcomes (ascertainment bias) stemming from knowing the treatment group assignment of the participants being observed. In addition, and not well appreciated, blinding may improve compliance and retention of trial participants while reducing biased supplemental care or treatment (sometimes called co-intervention). Furthermore, empirical evidence from methodological studies indicates that indeed blinding prevents bias.
Notably, without successful blinding, subjective outcomes pose greater opportunities for bias. Pain-rating scales and wound-healing scores are examples. Generally, blinding becomes less important in reducing ascertainment bias as the outcomes become less subjective. Hard, objective outcomes, such as death, allow slight opportunity for ascertainment bias.
In Chapter 16 we avoided discussions of implementation. Frequently, medical researchers have a reasonable idea of how to implement blinding. At least they understand it better than how to randomise and adequately follow up participants. However, since we wrote the first edition of our book, we have noticed gaps in blinding implementation knowledge for investigators who conduct randomised trials. This chapter has been added to complement our earlier chapter by focusing on implementation of blinding in trials.
Descriptions of strategies for treatment blinding invariably refer to ‘single blinding’ and ‘double blinding’. These terms are best avoided due to confusion over definitions, as described in the literature and in Chapter 16 . However, avoiding their use is nearly impossible because the terms are used ubiquitously. Moreover, they permeate the methodological literature. We believe that investigators can reasonably use those terms given that they explicitly define them in the paper in which they are used. We use the terms ‘single blinding’ and ‘double blinding’ in this chapter and clearly define them.
We seek to clarify issues around implementation of blinding. We will first address single blinding of participants, investigators, or assessors. Some approaches are described that often neglect to achieve at least some level of blinding, particularly in surgical and medical device trials. Thereafter, we will address the more complex issues surrounding double blinding. The double-dummy, double-blinded approach persists as perhaps the most confusing aspect of blinding, while actually often simplifying the process of double-blinding. The role of placebos remains misunderstood as well. Our goal is to clarify those concepts and present examples. We hope these presentations will stimulate solutions to blinding implementation issues.
Single Blinding: Participants, Investigators, or Assessors
We define single blinding as keeping trial participants, investigators (usually healthcare providers), or assessors (those determining outcome data) unaware of the assigned treatment so that they will not be influenced by that knowledge. Our definition means that one of those three categories would be blinded.
Typically, single blinding connotes a trial in which participants remain ignorant to the treatment received. Participants usually are easier to blind than investigators. The key is to prevent the participant from knowledge of the treatment received.
As discussed in Chapter 16 , several potential benefits emanate from blinding participants. If blinded, participants are less likely to have biased psychological or physical responses to intervention, less likely to seek additional adjunct interventions, and less likely to leave the trial without providing outcome data, which leads to losses to follow-up. Also, they are more likely to comply with trial regimens. For example, a participant’s discovery that he or she is receiving placebo and not the desired new drug can lead to disappointment and dropping out of the study.
Usually, single-blind trials do not refer to the investigator being blinded. Indeed, if the investigator can be blinded, then likely the participants and assessors can also be blinded. We address that situation in the later section on ‘total blinding’ (what others frequently call ‘double blinding’).
The other form of single blinding that occurs relatively frequently in the medical literature is the blinding of assessors. Often, trial investigators and participants would not be blinded in this case. If the assessors are blinded, they would be less likely to have biases affect their outcome assessment, especially with subjective outcomes of interest ( Chapter 16 ). This can be a major benefit in a trial in which participants and investigators cannot be blinded.
Implementation issues: participants
Blinding of anyone tends to be easiest in trials of pharmaceutical agents, but these are instances where total blinding (usually referred to as double blinding) is frequently possible. We address trials involving pharmaceutical agents under the double-blinding section. In this section, we focus on nonpharmaceutical treatments (NPTs) where total blinding is generally not possible. Often, such treatments involve surgery, a medical device, or a treatment involving a collaboration among participants and investigators/caregivers such as rehabilitation, education, or psychotherapy.
In many instances, investigators have difficulty blinding participants. Yet they often can achieve some level of success with thorough and often ingenious planning. The concept is to standardise the experiences of all participants among all the treatment groups in the trial. The actual treatment and its administration must be perceived as the same for those participants. Usually, investigators manipulate the experiences of the control group to mimic the experiences of the treatment group. Given the innumerable potential treatments involved, the trial blinding implementation approaches need to be customised. Although we provide some examples in this chapter, readers should consult the paper by Boutron et al. for an excellent cataloguing of methods used for blinding in NPT trials.
As an example, consider a randomised trial of two different intrauterine devices (IUDs) for contraception. Blinding the clinician who inserts the IUDs would be impossible. The providers will know which IUD is inserted in each woman. However, procedures could be used that would blind the participant. A visual screening, such as a paper drape over the woman’s legs, might suffice.
In another example, a randomised controlled trial examined acupressure wristbands for treating nausea and vomiting in pregnancy. Those randomised to the active treatment group had the pressure beads applied to the P6 meridian point on the anterior forearms. Those assigned to the placebo group had the bead applied to the dorsal forearm, a site thought to be ineffective. Women in both groups had the same bands worn for the same duration, but with pressure applied to different spots.
In another, a trial of paracervical anaesthesia for first-trimester abortion used a sham treatment arm. All participants received the same oral premedications. All then had cervical injection of buffered lidocaine 2 mL for tenaculum placement (a holding instrument on the cervix). Participants assigned to the active treatment had an additional 18 mL of buffered lidocaine injected at four sites around the cervix. Those assigned to the sham arm had a covered needle pressed against the vaginal sidewall in two sites for a similar amount of time. That sham procedure was a reasonable choice, because paracervical injection of saline as a placebo may provide anaesthesia by tissue distention.
Blinding in surgical trials is difficult, and, of course, blinding the surgeon is invariably impossible. However, to blind the participant and assessor, an operation could be done on one group, and a sham operation, such as a skin incision only, could be done on the comparison group.
Indeed, in NPTs, methods for blinding of participants mainly entail ‘sham procedures such as simulation of surgical procedures, attention-control interventions, or a placebo with a different mode of administration for rehabilitation or psychotherapy. Trials assessing devices reported various placebo interventions such as use of a sham prosthesis, identical apparatus (e.g., identical but inactivated machine or use of activated machine with a barrier to block the treatment), or simulation of using a device’.
Although we are focusing on examples of sham surgery in the next section, much of what we discuss extends to medical device trials as well. In a survey of 123 NPT trials that reported blinding, 58% utilised a sham procedure. Of note, successful blinding of participants usually enables successful blinding of assessors as well.
Sometimes investigators go to great lengths to ensure blinding of participants. For example, in a randomised trial, investigators allocated patients with osteoarthritis of the knee to arthroscopic debridement, arthroscopic lavage, or placebo surgery. Those allocated to the placebo surgical group received skin incisions and underwent a simulated debridement without insertion of the arthroscope. Patients and assessors apparently were unaware of the treatment assignments. Interestingly, this trial revealed that the outcomes from arthroscopic lavage or arthroscopic debridement were not superior to the placebo procedure. Sham knee incisions were invasive but were an innovative approach that probably eliminated bias.
Another randomised trial involved sham burr holes in the skull. Investigators allocated patients who had severe Parkinson disease to receive a transplant of nerve cells into the putamen bilaterally or undergo sham surgery in which burr holes were drilled in the skull with the dura not being penetrated. The authors concluded that their intervention yielded some clinical benefit in younger but not older patients. In this challenging blinding design, surgeons did not know which procedure would be done until they were in the operating room. The blinded participants assessed the primary outcome.
Blinding efforts can be arduous. The two examples of sham operations presented previously were useful to reduce bias, but they were difficult to implement. Particularly in surgery trials, investigators must go to extraordinary lengths to properly blind participants. Too frequently, blinding challenges investigators; they become frustrated and shun proper blinding because of difficulties. In addition, they may excuse their lack of attention to proper blinding by claiming ethical problems. However, those claims ring hollow if they never made a diligent attempt.
As with all trials, surgery trials pass through clearance with ethical review boards (ERBs), sometimes called institutional review boards (IRBs). Certainly, trials involving sham operations face higher hurdles than pharmaceutical trials. ERBs balance the need for unnecessary comparative surgeries to reduce bias against the risks to participants. Criteria for use of sham surgery controls have been proposed:
Satisfy all the general standards for ethical conduct of clinical research.
Assure that no reasonable alternative research design exists.
Involve a procedure for minimising the risk–benefit ratio.
Recruit a minimum sample size to answer the trial question.
Establish an exceptionally vigilant, independent safety monitoring board.
When an investigator or ERB decide that sham surgery is not appropriate, then a less invasive sham procedure may be warranted. For example, in a trial of laparoscopic compared with open appendectomy, an analogous approach to a sham treatment was used, although it was probably not as effective: ‘At the end of each procedure, 3 wound dressings and an abdominal binder were applied to every patient to blind the patient, the nursing and the medical staff, and the independent data collector as to the nature of the procedure’. The investigators used placebo wound dressings that appeared identical and were equivalently applied. In this instance, however, greater risk for bias prevails in that the incisions were not similar. Hence examining under the wound dressing could reveal the operative method. Of note the authors term this a double-blind randomised trial. However, because the operators could not be blinded, this does not fit our definition of a double-blind trial.
Implementation issues: assessors
Consideration should be given to blinding assessors in trials. The process involves an assessor (such as a research nurse) blinded to the intervention group of the participants, who determines the outcomes of those participants. In some trials, blinding of participants makes that easier. In other trials, such as the knee-surgery trial discussed previously, the potential blinding of assessors necessitates the blinding of participants. Yet often the assessors can be blinded even if the participants are not blinded. In a study of NPT trials, the ‘methods reported for blinding outcome assessors relied mainly on centralized assessment of paraclinical examinations, clinical examinations (i.e., use of video, audiotape, photography), or adjudications of clinical events’. Investigators should consider blinding the outcome assessor more frequently than currently happens.
As an example, consider an HIV prevention, open-label randomised trial of two highly effective, reversible methods of contraception. Injectable depot medroxyprogesterone acetate (DMPA) would be compared with a copper IUD to evaluate whether DMPA increases the risk of acquiring HIV infection. Blinding the investigators and participants would be impossible. However, with the outcome being HIV seropositivity based on a blood sample, the assessor could be blinded to the intervention group of the participants and make the outcome determination based on a blinded reading of laboratory results. Even with this hard, laboratory outcome, blinding the assessor guards against any hint of bias, which includes fraudulent assessment. ‘Investigators should . . . attempt to blind the outcome assessment whenever possible’.
Blinding the assessor becomes much more important with subjective outcomes. If the outcome in the previous example had been some clinical assessment of disease status, then having the outcome assessor be blinded to the treatment group of the participants would be critical to eliminating as much bias as possible. This has been done, for example, by relying on assessment of outcomes ‘by a blinded adjudication committee or a blinded assessment of the extract of the case report form’.