Chester Clinic

 

Pii: s0140-6736(02)07816-6

EPIDEMIOLOGY SERIES Blinding in randomised trials: hiding who got what Kenneth F Schulz, David A Grimes Blinding embodies a rich history spanning over two centuries. Most researchers worldwide understand blindingterminology, but confusion lurks beyond a general comprehension. Terms such as single blind, double blind, and tripleblind mean different things to different people. Moreover, many medical researchers confuse blinding with allocationconcealment. Such confusion indicates misunderstandings of both. The term blinding refers to keeping trialparticipants, investigators (usually health-care providers), or assessors (those collecting outcome data) unaware ofthe assigned intervention, so that they will not be influenced by that knowledge. Blinding usually reduces differentialassessment of outcomes (information bias), but can also improve compliance and retention of trial participants whilereducing biased supplemental care or treatment (sometimes called co-intervention). Many investigators and readersnaively consider a randomised trial as high quality simply because it is double blind, as if double-blinding is the sinequa non of a randomised controlled trial. Although double blinding (blinding investigators, participants, and outcomeassessors) indicates a strong design, trials that are not double blinded should not automatically be deemed inferior.
Rather than solely relying on terminology like double blinding, researchers should explicitly state who was blinded,and how. We recommend placing greater credence in results when investigators at least blind outcome assessments,except with objective outcomes, such as death, which leave little room for bias. If investigators properly report theirblinding efforts, readers can judge them. Unfortunately, many articles do not contain proper reporting. If an article claims blinding without any accompanying clarification, readers should remain sceptical about its effect on bias reduction.
The rich history of blinding in clinical trials spans a couple received.3 Participants who know that they have been of centuries.1 Most researchers worldwide appreciate its assigned to a group who will receive a new treatment meaning. Unfortunately, beyond that general appreciation might harbour favourable expectations or increased lurks confusion. Terms such as single-blind, double-blind, apprehension. Those assigned to standard treatment, and triple-blind mean different things to different people.2 however, might feel deprived or relieved. Despite evidence Moreover, many medical researchers confuse the term to suggest that new treatments are as likely to be worse as blinding with allocation concealment. The fact that such they are to be better than standard treatments,6 confusion arises suggests that both terms are participants probably assume that new treatments will be misunderstood. Clear theoretical and practical differences better than standard treatments—new means improved.
separate the two. Blinding prevents ascertainment bias In any case, knowledge of the intervention received, and and protects the sequence after allocation.3,4 By contrast, perceptions of that treatment, can affect the psychological researchers use methods of allocation concealment or physical responses of the participants. Knowledge of primarily to prevent selection bias and to protect an treatment allocation can also affect compliance and assignment sequence before and until allocation.
retention of trial participants (panel 1).
Furthermore, in some trials, blinding cannot be Blinding investigators—those who contribute to a successfully implemented, whereas allocation conceal- broadly defined trial team including, but not limited to, ment can always be successfully implemented.4,5 trial designers, participant enrollers, randomisation Blinding represents an important, distinct aspect of implementors, health-care providers, intervention randomised controlled trials.3 The term blinding refers to counsellors, and routine data collectors—is also keeping trial participants, investigators (usually health- important.3 Investigators especially pertinent to blinding care providers), or assessors (those collecting outcome include health-care providers (such as an attending data) unaware of an assigned intervention, so that they are physician or nurse) and intervention counsellors—eg, not influenced by that knowledge. Blinding prevents bias someone who delivers a behavioural prevention message— at several stages of a trial, although its relevance varies who might interact with the participants throughout the according to circumstance. Although initial forays into trial. If investigators are not blinded, their attitudes for or blinding might have used a blindfold,1 the processes have against an intervention can be directly transferred to now become much more elaborate. In this article, we participants.7 Their inclinations could also be manifested focus on the attributes and benefits of blinding. in differential use of ancillary interventions ofsupplemental care or treatment (co-interventions), Potential effects of blinding differential decisions to withdraw participants from a trial, If participants are not blinded, knowledge of group or differential adjustments to the medication dose assignment can affect responses to the intervention (panel 1). Investigators might also encourage ordiscourage continuation in a trial on the basis of Lancet 2002; 359: 696–700 knowledge of the intervention group assignment. Perhaps most importantly, blinding helps to reduce Family Health International, PO Box 13950, Research Triangle differential assessment of outcomes (often called Park, NC 27709, USA (K F Schulz PhD, D A Grimes MD) information or ascertainment bias) (panel 1). For Correspondence to: Dr Kenneth F Schulz example, if outcome assessors who know of the treatment allocation believe a new intervention is better than an old THE LANCET • Vol 359 • February 23, 2002 • www.thelancet.com For personal use. Only reproduce with permission from The Lancet Publishing Group.

EPIDEMIOLOGY SERIES Panel 1: Potential benefits accruing dependent on those individuals successfully blinded Individuals blinded Potential benefits Less likely to have biased psychological or physical responses to interventionMore likely to comply with trial regimensLess likely to seek additional adjunct interventionsLess likely to leave trial without providing outcome data, leading to lost to follow-up Less likely to transfer their inclinations or attitudes to participants Less likely to differentially administer co-interventionsLess likely to differentially adjust doseLess likely to differentially withdraw participantsLess likely to differentially encourage or discourage participants to continue trial Less likely to have biases affect their outcome assessments, especially with subjective outcomes of interest one they could register more generous responses to that not define double-blind trials consistently, in particular, intervention. Indeed, in a placebo-controlled trial in but they make matters worse by frequently failing to patients with multiple sclerosis8 the unblinded, but not the report their definitions clearly in their articles. Building on blinded, neurologists' assessments showed an apparent the original blindfolding efforts,1 and the once common benefit of the intervention. double blindfold terminology,12 we further obfuscate by Subjective outcomes—eg, pain scores—present great offering additional definitions of single and double opportunities for bias.3 Furthermore, some outcomes blinding (figure 1). More seriously, when we use double- judged objective can be fraught with subjectivity, for blind or its derivatives in this article, we mean that steps example, salpingitis. In general, though, blinding becomes have been taken to blind participants, investigators, and less important to reduce observer bias as the outcomes assessors to group assignments. In reporting randomised become less subjective, since objective (hard) outcomes controlled trials, we urge researchers to explicitly state leave little opportunity for bias. Knowledge of the what steps they took to keep whom blinded. intervention would not greatly affect measurement of a Sparse reporting on blinding, however, is common.
hard outcome, such as death.
Many investigators neglect to report whether or not theirtrial was blinded. For example, reports of 51% of 506 Lexicon of blinding trials in cystic fibrosis,13 33% of 196 trials in rheumatoid Non-blinded (open or open label) denotes trials in which arthritis,14 and 38% of 68 trials in dermatology15 did not everyone involved knows who has received which state whether blinding was used. When researchers have interventions throughout the trial. Blinding (masking) reported their study as double-blind, they frequently have indicates that knowledge of the intervention assignments not provided much further clarification.14,16—18 For example, is hidden from participants, trial investigators, or of 31 double-blind trials in obstetrics and gynecology, only 14 (45%) reports indicated the similarity of the The terminology single blind usually means that one of treatment and control regimens (for example, appearance, the three categories of individuals (normally participant taste, administration) and only 5 (16%) provided rather than investigator) remains unaware of intervention statements to indicate that blinding was successful.18 assignments throughout the trial.9 A single-blind trialmight also, confusingly, mean that the participant and Masking or blinding investigator both know the intervention, but that the Some people prefer the term masking to blinding to assessor remains unaware of it. describe the same procedure. Masking might be more In a double-blind trial, participants, investigators, and appropriate in trials that involve participants who have assessors usually all remain unaware of the intervention impaired vision, and could be less confusing in trials in assignments throughout the trial.3 In view of the fact that which blindness is an outcome.3 Blinding, however, three groups are kept ignorant, the terminology double conveys a strong bias prevention message. Apparently, blind is sometimes misleading. In medical research, blinding terminology emerged when Benjamin Franklin however, an investigator frequently also assesses, so in this and colleagues19 actually blindfolded participants to shield instance the terminology accurately refers to two them from knowledge in their assessments of the Triple blind usually means a double-blind trial that also maintains a blind data analysis.10 Some investigators,however, denote trials as triple-blind if investigators andassessors are distinct people and both, as well asparticipants, remain unaware of assignments.
Investigators rarely use quadruple blind, but those that douse the term to denote blinding of participants,investigators, assessors, and data analysts.11 quintuple blind must mean that the allocation schedulehas been lost and nobody knows anything. Contrary toMae West's claim that "too much of a good thing can bewonderful", such is not always the case in blinding.
Confused terminology of single, double, and triple blinding permeates the literature,3 textbooks, and journal articles all offering differentinterpretations and definitions.2 Not only do investigators Figure 1: The authors: double blinded versus single blinded THE LANCET • Vol 359 • February 23, 2002 • www.thelancet.com For personal use. Only reproduce with permission from The Lancet Publishing Group.

EPIDEMIOLOGY SERIES Does blinding prevent bias?Some investigators, readers, and editors overstate theimportance of blinding in prevention of bias. Indeed,some consider a randomised trial as high quality if it isdouble blind—ie, as if double blinding is the sine qua nonof a randomised controlled trial.3 Unfortunately, scientificlife is not that simple. A randomised trial can bemethodologically sound and not be double blind or,conversely, double blind and not methodologically sound.
Lasagna12 captured that notion long ago: "Let us examinethe placebo somewhat more critically, however, since itand ‘double blind' have reached the status of fetishes inour thinking and literature. The Automatic Aura ofRespectability, Infallibility, and Scientific Savoir-fairewhich they possess for many can be easily shown to be Figure 2: The authors blinded and masked undeserved in certain circumstances."12 Although doubleblinding suggests a strong design, it is not the primary therapeutic claims made for Mesmerism. The imagery of indicator of overall trial quality. Moreover, many trials blindfolding, a total covering of the eyes, conveys stronger cannot be double blinded. Such trials must, therefore, be bias prevention than masking, where eye holes could judged on overall merit rather than an inapplicable permit viewing (figure 2). Blinding also suggests a more standard based on double blinding.
secure procedure to some. The International Conference We do not, however, suggest that blinding is on Harmonization (ICH) guidance,20 unimportant.3 Intuitively, blinding should reduce bias, primarily uses blinding terminology. (The ICH is an and available evidence supports that impression.
intensive tripartite collaboration between regulatory Methodological investigations tend to show that double authorities in Europe, Japan, and the USA to develop blinding prevents bias but is less important, on average, in common guidelines for the design, implementation, and prevention of bias than is adequate allocation reporting of clinical trials). We prefer blinding because it has a long history, maintains worldwide recognition,creates strong imagery, and permeates the ICH What to look for in descriptions of blinding In general, if researchers describe a trial as double-blind,readers can assume that they have avoided bias. Empirical Placebos and blinding evidence lends support to this recommendation. As Interventions (treatments) sometimes have no effect on suggested in the CONSORT guidelines,24,25 however, the outcomes being studied.3 When an ineffective investigators should not use only the single-blind, double- intervention is administered to participants in the context blind, or triple-blind terminology, but should also of a well-designed randomised controlled trial, however, explicitly state who was blinded, and how. Moreover, if beneficial effects on participants' attitudes sometimes the researchers contend that the trial investigators, occur, which in turn affect outcomes.10 Researchers refer participants, and assessors were blinded—ie, double to this phenomena as the placebo effect.
blind—then they should provide information about the A placebo refers to a pharmacologically inactive agent mechanism (capsules, tablets, film, &c), similarity of that investigators administer to participants in the control treatment characteristics (appearance, taste, group of a trial.3 The use of a placebo control group administration), and allocation schedule control—eg, balances the placebo effect in the treatment group, location of the schedule during the trial, when the code allowing for independent assessment of the treatment was broken for the analysis, and circumstances under effect. Although placebos can have a psychological effect, which the code could be broken for individual instances.
they are administered to participants in a trial because Such additional information can lend support to or they are otherwise inactive. An active placebo is a placebo undermine claims of double-blinding (panel 2).26–29 with properties that mimic the symptoms or side-effects— If researchers properly report their blinding efforts, eg, dry mouth, sweating—that might otherwise reveal the readers can judge those efforts. Unfortunately, many identity of the (pharmacologically) active test articles will not contain proper reporting. If a researcher treatment. Most researchers agree that placebos should be claims to have done a blinded study, but does not provide administered, whenever possible, to controls when accompanying clarification, readers should remain assessing the effects of a proposed new treatment for a sceptical about its effect on bias reduction. For example, condition for which no effective treatment already one trial30 of prophylactic antibiotics claimed to be exists.9,10 Indeed, blinding frequently necessitates the use blinded, but the methods section of the report revealed that little or no blinding occurred.
However, a proven effective standard treatment, if such Ideally, researchers should also relate if blinding was exists, is usually given to the control group for comparison successful. Investigators can theoretically assess the against a new treatment.3 Thus, investigators might success of blinding by directly asking participants, health- compare two active treatment groups without a placebo care providers, or outcome assessors which intervention group. Even then, however, investigators frequently they think was administered (panel 3). In principle, if attempt to achieve blinding by use of the double-dummy blinding was successful, these individuals should not be method, in essence two placebos.11,21 For example, for able to do better than chance when guessing the comparison of two agents, one in a blue capsule and the intervention, for example. In practice, however, blinding other in a red capsule, the investigators would prepare might be totally successful, but participants, health-care blue placebo capsules and red placebo capsules. Then providers, and outcome assessors might nevertheless both treatment groups would receive a blue and a red guess the intervention because of ancillary information.
capsule, one active and one inactive.
Disproportionate levels of adverse side-effects might THE LANCET • Vol 359 • February 23, 2002 • www.thelancet.com For personal use. Only reproduce with permission from The Lancet Publishing Group.
EPIDEMIOLOGY SERIES the queries—in other words, if they have deciphered Panel 2: Descriptions of blinding group assignments, they might provide responses contraryto their deciphering findings to disguise their actions.
"No patient, research nurse, investigator, or any other medical That difficulty, along with interpretation difficulties or nursing staff in the ICU was aware of the treatment stemming from adverse side-effects and successful clinical assignments for the duration of the study. All statistical outcomes, leads us to question the usefulness of tests of analysis was also done with masking maintained.
blinding in some circumstances. Investigators should Randomisation authorities were instructed to report any carefully consider the usefulness of assessing the success suspected breach of the masking procedures. No report was of their blinding efforts, but if they proceed, should filed . . The drug or placebo (vehicle without active drug) was provide the results of any assessments. At the very least, prepared for syringe pump infusion or for volumetric pump they should report any failure of the blinding procedure, infusion in indistinguishable syringes or bags."26 such as non-identical placebo or active preparations.
". . in a double-blind, placebo-controlled manner . . Neither Published reports rarely contain assessments of blinding, the patients nor doctors could distinguish the placebo from but, if provided, readers should sceptically assess the sibutramine capsules. The taste of the capsules was identical provided they were swallowed whole as instructed. . Results Double blinding proves difficult or impossible in many of biochemical analyses were completed before the trials. For instance, in general, surgical trials cannot be randomisation code was broken at the end of the completed double blinded. Specifically, a trial that compares degrees of pain associated with sampling blood from the ear orthumb cannot be double-blinded.34 If researchers do not "The study was double-blinded—that is, neither the women nor describe their trial as double-blind or the equivalent, it the study staff, including the biostatisticians at Family Health could still be scientifically strong. Apart from assessment International, knew which group was using the nonoxynol 9 of the other methodological aspects of the trial, readers film. The nonoxynol 9 film contained . . The placebo film would have to assess how much bias might have ensued contained . . The films were identical in appearance, due to absence of blinding. Readers should identify if packaging, and labeling."28 anybody was blinded in the trial and what benefits might "The doxycycline and placebo were in capsule form and have accrued (panel 1). Indeed, blinding of outcome identical in appearance . . The randomization code was kept assessors is often possible and advisable, even in open in the USA." (Note: the trial was conducted in Kenya) "Thus, trials.11 For example, lesions can be photographed before all administration and assessments were done blinded to and after treatment and assessed by someone not involved treatment assignment, and the investigators and patients were in the study.11 We recommend placing greater credence in also blinded to the ongoing results of the study. The code was results when someone unaware of treatment assignments broken only after data collection had been completed."29 judges outcome measures. Even that recommendation, however, is not absolute.
ICU=intensive care unit.
As noted earlier, some hard outcomes, such as death,leave little room for ascertainment bias. In other words, provide strong hints as to the intervention. Irrespective of blinding the assessor to hard outcomes might have little painstaking efforts to do double-blinded trials, some interventions have side-effects that are so recognisablethat their occurrence will unavoidably reveal the intervention received to both the participants and the Blinding embodies a rich history spanning over two health-care providers.11,24 Even more fundamental than centuries. Most researchers worldwide understand hints from adverse effects are the hints from clinical blinding terminology, but confusion lurks beyond a outcomes. Researchers usually welcome large clinical general comprehension. Investigators should clearly effects (except perhaps in equivalence trials). If they arise, explicate those blinded and not blinded in their trial, health-care providers and participants would likely rather than only labeling their trial as single-blind, double- deduce—not always accurately of course—that a blind, or triple-blind. Readers should expect such clarity participant with a positive outcome received the active when reading and judging a trial report.
(new) intervention rather than control (standard). If We thank Willard Cates and David L Sackett for their helpful comments indeed the active (new) intervention materialises as on an earlier version of this report. Much of this material stems from our helpful (highly desirable) then their deductions would be 15 years of teaching the Berlex Foundation Faculty Development Course.
correct more often than chance guesses.24,31 Irrespective oftheir suspicions, end-of-trial tests of blindness might actually be tests of hunches for adverse effects or Kaptchuk TJ. Intentional ignorance: a history of blind assessment and placebo controls in medicine. Bull Hist Med 1998; 72:
Furthermore, individuals might be reluctant to expose any unblinding efforts by providing accurate responses to Devereaux PJ, Manns BJ, Ghali WA, et al. Physician interpretations
and textbook definitions of blinding terminology in randomized
controlled trials. JAMA 2001; 285: 2000–03.
Schulz KF, Chalmers I, Altman DG. The landscape and lexicon of Panel 3: Assessment of the success of blinding blinding in randomized trials. Ann Intern Med 2002; 136: 254–59.
Schulz KF, Chalmers I, Hayes RJ, Altman DG. Empirical evidence "We asked 126 staff members their opinions of which film of bias: dimensions of methodological quality associated with was the placebo. Eighteen percent thought film A (the placebo) estimates of treatment effects in controlled trials. JAMA 1995; 273:
was the placebo, 13 percent thought film B (nonoxynol 9) was the placebo, and 69 percent had no opinion about which film Schulz KF, Chalmers I, Grimes DA, Altman DG. Assessing was the placebo. Of the 68 peer educators (the staff members the quality of randomization from reports of controlled trials
published in obstetrics and gynecology journals. JAMA 1994; 272:
most likely to reflect the opinion of the participants), 16 percent thought film A was the placebo, 13 percent thought Chalmers I. What is the prior probability of a proposed new film B was the placebo, and 71 percent had no opinion."28 treatment being superior to established treatments? BMJ 1997; 314:
74–75.
THE LANCET • Vol 359 • February 23, 2002 • www.thelancet.com For personal use. Only reproduce with permission from The Lancet Publishing Group.
EPIDEMIOLOGY SERIES Wolf S. Effects of suggestion and conditioning on action of chemical 22 Khan KS, Daya S, Collins JA, Walter SD. Empirical evidence of bias agents in human subjects: pharmacology of placebos. J Clin Invest in infertility research: overestimation of treatment effect in crossover 1950; 29: 100–09.
trials using pregnancy as the outcome measure. Fertil Steril 1996; 65:
Noseworthy JH, Ebers GC, Vandervoort MK, Farquhar RE, Yetisir E, Roberts R. The impact of blinding on the results of a randomized, 23 Moher D, Pham B, Jones A, et al. Does quality of reports of placebo-controlled multiple sclerosis clinical trial. Neurology 1994; 44:
randomised trials affect estimates of intervention efficacy reported in meta-analyses? Lancet 1998; 352: 609–13.
Meinert CL. Clinical trials: design, conduct, and analysis. New York: 24 Altman DG, Schulz KF, Moher D, et al. The revised CONSORT Oxford University Press, 1986.
statement for reporting randomized trials: explanation and elaboration.
10 Pocock SJ. Clinical trials: a practical approach. Chichester: Wiley, Ann Intern Med 2001; 134: 663–94.
25 Moher D, Schulz KF, Altman D. The CONSORT statement: revised 11 Day SJ, Altman DG. Statistics notes: blinding in clinical trials and recommendations for improving the quality of reports or parallel-group other studies. BMJ 2000; 321: 504.
trials. Lancet 2001; 357: 1191–94.
12 Lasagna L. The controlled trial: theory and practice. J Chronic Dis 26 Bellomo R, Chapman M, Finfer S, Hickling K, Myburgh J. Low-dose 1955; 1: 353–67.
dopamine in patients with early renal dysfunction: a placebo-controlled 13 Cheng K, Smyth RL, Motley J, O'Hea U, Ashby D. Randomized randomised trial. Australian and New Zealand Intensive Care Society controlled trials in cystic fibrosis (1966–1997) categorized by time, (ANZICS) Clinical Trials Group. Lancet 2000; 356: 2139–43.
design, and intervention. Pediatr Pulmonol 2000; 29: 1–7.
27 James WP, Astrup A, Finer N, et al. Effect of sibutramine on weight 14 Gøtzsche PC. Methodology and overt and hidden bias in reports of maintenance after weight loss: a randomised trial. STORM Study 196 double-blind trials of nonsteroidal antiinflammatory drugs in Group. Lancet 2000; 356: 2119–25.
rheumatoid arthritis. Control Clin Trials 1989; 10: 31–56.
28 Roddy RE, Zekeng L, Ryan KA, Tamoufe U, Weir SS, Wong EL. 15 Adetugbo K, Williams H. How well are randomized controlled trials A controlled trial of nonoxynol 9 film to reduce male-to-female reported in the dermatology literature? Arch Dermatol 2000; 136:
transmission of sexually transmitted diseases. N Engl J Med 1998; 339:
16 Mosteller F, Gilbert JP, McPeek B. Reporting standards and research 29 Sinei SK, Schulz KF, Lamptey PR, et al. Preventing IUCD-related strategies for controlled trials: agenda for the editor. Controlled Clin pelvic infection: the efficacy of prophylactic doxycycline at insertion. Trials 1980; 1: 37–58.
Br J Obstet Gynaecol 1990; 97: 412–19.
17 DerSimonian R, Charette LJ, McPeek B, Mosteller F. Reporting 30 Baker KR, Drutz HP, Barnes MD. Effectiveness of antibiotic on methods in clinical trials. N Engl J Med 1982; 306: 1332–37.
prophylaxis in preventing bacteriuria after multichannel urodynamic 18 Schulz KF, Grimes DA, Altman DG, Hayes RJ. Blinding and investigations: a blind, randomized study in 124 female patients. exclusions after allocation in randomised controlled trials: survey of Am J Obstet Gynecol 1991; 165: 679–81.
published parallel group trials in obstetrics and gynaecology. BMJ 31 Quitkin FM, Rabkin JG, Gerald J, Davis JM, Klein DF. Validity of 1996; 312: 742–44.
clinical trials of antidepressants. Am J Psychiatry 2000; 157: 327–37.
19 Franklin B, Bailly JS, Lavoisier A. Rapport des commissaires chargés 32 Sackett DL, Gent M, Taylor DW. Tests for the blindness of par le roi, de l'examen du magnetisme animal. Nice: Chez Gabriel randomized trials may not. Clin Res 1986; 34: 711A.
Floteron, 1785.
33 The Canadian Cooperative Study Group. A randomized trial of aspirin 20 Department of Health and Human Services, FDA. International and sulfinpyrazone in threatened stroke. N Engl J Med 1978; 299:
conference on harmonisation: guidance on statistical principles for clinical trials. Fed Regist 1998; 63: 49583–98.
34 Carley SD, Libetta C, Flavin B, Butler J, Tong N, Sammy I. An open 21 Altman DG. Practical statistics for medical research. London: prospective randomised trial to reduce the pain of blood glucose Chapman and Hall, 1991.
testing: ear versus thumb. BMJ 2000; 321: 20.
THE LANCET • Vol 359 • February 23, 2002 • www.thelancet.com For personal use. Only reproduce with permission from The Lancet Publishing Group.

Source: http://www.anaesthesia.ie/archive/Courses/MSc%20Units/Appendix%202%20Unit%203.pdf