Blinding in Trials of Interventional Procedures Is Possible And
Total Page:16
File Type:pdf, Size:1020Kb
F1000Research 2017, 6:1663 Last updated: 20 SEP 2017 REVIEW Blinding in trials of interventional procedures is possible and worthwhile [version 1; referees: 1 approved] Karolina Wartolowska , David Beard, Andrew Carr Botnar Research Centre, Nuffield Department of Orthopaedics Rheumatology and Musculoskeletal Sciences (NDORMS), Oxford, UK v1 First published: 08 Sep 2017, 6:1663 (doi: 10.12688/f1000research.12528.1) Open Peer Review Latest published: 08 Sep 2017, 6:1663 (doi: 10.12688/f1000research.12528.1) Referee Status: Abstract In this paper, we have used evidence from our earlier review of surgical randomised controlled trials with a placebo arm to show that blinding in trials of Invited Referees interventional procedures is feasible, and that many creative methods can be 1 used to make the active and the placebo procedure as similar as possible. We give examples of ingenious strategies used to simulate the active procedure version 1 and make the placebo control indistinguishable from the active treatment. We published report discuss why it is important to blind of patients, assessors, and caregivers and 08 Sep 2017 the types of bias that may occur in interventional trials. Finally, we describe the benefits of blinding, from the obvious ones such as avoiding bias, as well as Ian A Harris, University of New South less evident benefits such as avoiding patient drop out in the control arm. 1 Wales, Australia Discuss this article Comments (0) Corresponding author: Karolina Wartolowska ([email protected]) Author roles: Wartolowska K: Writing – Original Draft Preparation, Writing – Review & Editing; Beard D: Writing – Review & Editing; Carr A: Funding Acquisition, Writing – Review & Editing Competing interests: No competing interests were disclosed. How to cite this article: Wartolowska K, Beard D and Carr A. Blinding in trials of interventional procedures is possible and worthwhile [version 1; referees: 1 approved] F1000Research 2017, 6:1663 (doi: 10.12688/f1000research.12528.1) Copyright: © 2017 Wartolowska K et al. This is an open access article distributed under the terms of the Creative Commons Attribution Licence, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited. Grant information: This work has received funding from the NIHR Oxford Musculoskeletal Biomedical Research Unit. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. First published: 08 Sep 2017, 6:1663 (doi: 10.12688/f1000research.12528.1) Page 1 of 8 F1000Research 2017, 6:1663 Last updated: 20 SEP 2017 Introduction Blinding of surgeons The aim of a trial is to produce unbiased evidence. As randomised Unlike drug trials, in which the physician gives a tablet prepared controlled trials (RCTs) with a placebo arm control for many somewhere else, the surgeon has to perform a specific procedure types of bias, they are regarded as the most reliable method of considered to be therapeutic; therefore, blinding of surgeons may demonstrating treatment efficacy and provide the highest level of not always be possible. evidence1. RCTs of interventional procedures are rare2–5, partly because they are challenging6; however, they are not impossible There have been attempts to blind surgeons, for example a to perform, even if they involve a placebo arm7. In this paper surgeon inserted a catheter under fluoroscopic guidance and handed we will discuss why trials should be blinded and summarise the over the procedure to a technician who delivered the radiofre- methods which have been used in the published placebo- quency energy (or not) according to the allocation16. In other trials, controlled trials of interventional procedures to achieve blinding. a palatal implant delivery system was prepared by the manufacturer to either contain the implant or not, which allowed for blinding Blinding in interventional trials is often necessary because of surgeons17,18. nowadays many procedures are performed to reduce pain and improve function and quality of life. Pain, function and quality Blinding of outcome assessors of life are sometimes regarded as preferable outcome measures In 81% of placebo-controlled surgical trials both patients because they reflect patients’ needs and point of view8. However, and assessors were blinded15. It has been demonstrated that, as these outcomes depend on patients’ subjective perception, non-blinded assessors of subjective outcomes cause less bias in they are prone to bias and may lead to an exaggerated treatment trials than non-blinded patients reporting their symptoms19. effect in open-label trials9,10. Using subjective outcomes in an Blinding of assessors prevents observer-related bias, detection open-label study undermines its internal validity because it is not bias, and the Pygmalion effect. The Pygmalion effect refers to a possible to determine how much of the reported effect is related situation when investigators looking for a particular response are to the investigated treatment and how much is related to bias. predisposed to interpret the result in a way that shows the response they expect even if it is objectively absent. For example, a study Blinding of patients, surgeons, outcome assessors by Hrobjartsson and colleagues demonstrated that non-blinded and caregivers assessors were over-optimistic and “over-rated” patients in the Blinding of patients treatment group rather than “under-rated” patients in the control Blinding means concealing the treatment allocation from group20. In some trials the assessment was done by people not patients and any other people involved in the trial who may bias involved in the surgery, for example blinded researchers, staff at the results of the trial by knowing which groups the patients were another hospital that they were operated on at21, or by pathologist randomised to. blinded to the treatment allocation22. Blinding of patients prevents reporting bias in patient- Blinding of caregivers reported measures. For example, it has been demonstrated that Apart from blinding patients and assessors, it is important non-blinded patients exaggerate the effects size by 0.56 standard that caregivers and clinical/research staff do not know patient deviations and that the effect is even larger in trials on interven- treatment allocation, because their behaviour and attitudes may tional procedures, such as acupuncture11,12. This bias may be influence patient responses23–25. Patient-clinician interaction plays caused by patients’ expectations of treatment effect and infor- an important role in treatment response, and patients in trials mation given to them before the treatment13. Patients may also do better as they get more attention and time from clinical staff report symptoms depending on their “hunches” about treat- than patients receiving standard care26,27. Therefore, the interac- ment being effective or they may give answers they believe are tions between patients and the trial team should be standardised “correct” or expected from them, for example, because it would so that the “treatment context” (similar attention from doctors, have been impolite not to report improvement11. Therefore, it expectations, and settings) are comparable between the groups. has been suggested that patients should be blinded whenever possible11. Strategies used to maintain blinding in interventional placebo-controlled trials Blinding of patients also reduces adherence bias, i.e. patients in A placebo control is necessary if we want to know whether the control group not following the protocol/treatment. It may also improvement is really caused by the investigated procedure. prevent so called “contamination of the control group”, i.e. seek- It compares the intervention of interest with a procedure that seems ing additional treatment/help outside the trial and receiving con- identical but does not involve the crucial element of believed to comitant treatment. Blinding improves patient retention in the be “the cure”. The aim of a placebo arm is to control for effects trial. Risk of attrition in blinded trials is about 4% whereas in of receiving treatment not specifically related to the investigated non-blinded trials it is 7%11. Specifically in placebo-controlled sur- intervention. gical trials, subject retention is often reported as “excellent”14, and in our analysis the withdrawal rate was low (4%) and comparable It is often difficult to determine what is a specific and what is between the treatment and the placebo arm15. a non-specific effect in a trial27,28, and to disentangle placebo Page 2 of 8 F1000Research 2017, 6:1663 Last updated: 20 SEP 2017 response from response bias or the effect of patient-doctor to blind as natural orifices were used to insert the scope, and interactions29. It is beyond the scope of this review to discuss the incision or actual procedure site was not visible to patients, definitions of placebo1,29. Whether something is or is not a placebo caregivers, and assessors. depends on the intervention and chosen outcome variables1, but in order for blinding to be successful, the control procedure has to be Simulation of an interventional procedure as similar as possible to the investigated procedure27. Interventional Typically, the preparation for the placebo procedure and the trials differ from drug trials as they require access to the anatomical active procedure was as similar as possible and immitated the vis- structure of interest; therefore, they involve