Ethical Debate on Placebo Use May Prompt New Trial Designs

For a thing that is "nothing," placebo has been much in the news lately. Whenever the media have mentioned placebo this year, it often has been in the context of clinical trials overseas for treatments to prevent perinatal HIV transmission. The studies were controversial because most of them employed placebo controls with the various treatments being tested, although trials in the United States and France already had indicated that the antiretroviral drug zidovudine (AZT) reduced the incidence of mother-to-child transmission of type 1 HIV (E.M. Conner et al., New England Journal of Medicine, 331:1173-80, 1994). A debate erupted over whether it was proper to give people placebo when an effective therapy was available, even though the new studies were attempts to develop treatments that would be affordable in these developing countries, where the full AZT course is too expensive to be widely obtained. BLIND OR BALANCED? Subjects who were aware of receiving caffeine (left) tested higher for physiological signs of tension than subjects (right) who received the drug under double-blind conditions. Also contributing to current interest in placebos is the popularity of alternative medicine, mind-body concepts, and behavioral treatments for various disorders, which may involve psychological responses similar to those induced by placebo use. However, Anne Harrington, a Harvard University professor of the history of science, notes, "Alternative medicine is just as reluctant as any other form of medicine to say their results belong to the placebo effect, because it has been seen as nothing." She edited one of several books published recently on the subject (A. Harrington, ed., The Placebo Effect: An Interdisciplinary Exploration, Cambridge, Mass., Harvard University Press, 1997). "For me, placebos were a wonderfully subversive phenomenon in the world," she says in explaining why she produced the book. "They seemed to function [in] narrative, symbols, human relationships, and other things thought about in the humanities and social sciences, but they also happened in the body, and looked a lot like biology." Defining "placebo" can be tricky, in part because of the stigma historically connected to the word. It derives from the Latin phrase "I shall please," the first word in the Catholic vespers for the dead, which in the 12th century were commonly referred to as placebos. Professional mourners paid to sing placebos were deprecated, which is the likely source of the connotation the word took on, suggesting a flatterer or sycophant. In modern medicine, placebo became known as a pharmacologically inert treatment provided to please patients rather than benefit them (W.A. Brown, Scientific American, Jan. 1998, page 93). Nowadays, it is considered to be primarily a tool of clinical trials. But in both experimental research and medicine, it can take many guises. Placebos can be sugar pills, saline injections, sham incisions or other physical procedures, behavioral treatments, even the aura of professionalism created by credentials on the wall, and white-coated figures offering words of reassurance. The placebo effect varies greatly in degree and size, frequently involving many more people than the often-cited one-third of a given study group (J.A. Turner et al., JAMA-Journal of the American Medical Association, 271:1609-14, 1994). It also varies in strength and scope, depending on the disorder, but because very few studies on the effect itself have been done, little is known about which disorders are most susceptible to it. Much work remains to determine whether biochemical and neural pathways are activated by the psychological mechanisms involved, to produce physiological changes that affect the disease itself. What is known is that placebos seem capable, in some instances, of relieving the pain and suffering that can attend disease. "There are uncomfortable, ambiguous, raggedy-edged places in medicine, and placebos bring that fact back to us," Harrington observes. In the introduction to the book, she writes, "Placebos are the ghosts that haunt our house of biomedical objectivity, the creatures that rise up from the dark and expose the paradoxes and fissures in our own self-created definitions of the real and active factors in treatment." Irving Kirsch Irving Kirsch, a psychology professor at the University of Connecticut, has studied the role of a person's expectations in contributing to the placebo effect. He explains that classical conditioning, involving stimuli and responses, is supplemented in the case of placebo effect by specific beliefs concerning, say, relief of pain or depression, and by more global mechanisms of belief, such as hope, optimism, and possibly emotional states. The immune system, for example, could be affected by global expectancies. "Having that kind of distinction to these very different sorts of effects is very important to our designing of clinical trials," he says. According to FDA regulations, clinical trial designs may employ five different kinds of controls: placebos, no treatment, active agents, historical controls, and dose comparisons. However, for three decades the "gold standard" design has been the placebo-controlled, double-blind configuration. Theoretically, a placebo control group provides a baseline against which the effectiveness of an active agent can be measured, to show that the drug is better than nothing. But the placebo effect has repeatedly been shown to be something other than merely nothing. Yet, why and how it works remains largely unknown. Even the magnitude of the placebo effect in a study is hard to determine. In part, this is because normal, spontaneous remission of a disease cannot be distinguished from the placebo effect, unless there is a control group that has received no treatment at all. No- treatment controls are not common, although they have been used in alcohol studies, often employing the "balanced placebo design." In the four cells of such a design, participants: a) are told they will get a drug, and do get it; b) are told they will get a drug, but receive placebo; c) are told they will get placebo, and do get it, d) are told they will get placebo, but get a drug. TO LEARN MORE: One of several books published recently on placebos. Kirsch employed a variation of that design to assess effects of caffeine (I. Kirsch, M.J. Rosadino, Psychopharmacology, 110:437-42, 1993). Results showed that only subjects who knowingly received caffeine registered physiological and subjective increases in tension. In other words, only expectation of receiving the drug, coupled with actually receiving it, produced the tension effect. But what ethical guidelines pertain in such a trial, which deliberately misleads participants? Kirsch explains that ethics are served if several conditions are met: there is no better way to obtain the information sought; deception of subjects is minimized; prior consent to receive either an active agent or placebo is obtained; and subjects are made aware of the deception after the study. Another Kirsch paper that challenges the standard design is a meta-analysis published in an online journal of the American Psychological Association (I. Kirsch, G. Sapirstein, Prevention and Treatment, posted June 26, 1998). Examining 19 double-blind, placebo-controlled trials of various antidepressants involving 2,318 patients, he came to the surprising conclusion that only 25 percent of the response to any drug was actually the effect of the drug, and not the placebo effect. Kirsch stresses that a 25 percent response is large if the antidepressants are actually responsible for that effect. It is also possible, he notes, that the apparent response to the drug really may be an "active placebo" effect. An active placebo has no pharmacological properties to affect a given condition. However, it produces side effects that not only make it difficult for subjects to distinguish it from an active drug; it possibly will elicit a placebo response. Kirsch says a good way to make sure that the drug, and not merely its side effects, are eliciting a response, is to test it against an active placebo. This, he adds, is not often done. Another type of design is an equivalence or active control trial, matching a new drug against an existing treatment. But such trials can encourage sloppy methodologies, according to Food and Drug Administration guidelines on drug study designs, because researchers tasked with showing mere equivalence have little incentive to adopt meticulous methodology. "Carelessness of all kinds will tend to obscure differences between treatments," the guidelines note. Even so, equivalence trials can be appropriate when the standard therapy has been shown to be beneficial and the innovative treatment is easier to use, has fewer side effects, or is less costly (J.H. Ware, E.M. Antman, N. Engl. J. Med., 337:1159-61, 1997). Peter Lurie, research associate with the Washington, D.C.-based consumer rights group Public Citizen, emphasizes that equivalence trials can ensure that all subjects, including those in control groups, receive the best proven treatment, as required by the 1989 Declaration of Helsinki from the 41st World Medical Assembly in Hong Kong. New, less-encompassing language for that declaration is currently being drafted. "Write the rules, break the rules, get caught, change the rules. That's what's happening here," Lurie charges. "A Method of Evaluating a Therapeutic Agent," was the title Albert J. Stunkard gave to a paper about a clinical trial he devised almost a half-century ago to study the adrenergic blocking agent dibenamine (A.J. Stunkard, American Journal of Psychiatry, 107: 463-7, 1950). Then a psychiatrist at Johns Hopkins Hospital and now a professor of psychiatry at the University of Pennsylvania, Stunkard designed a double-blind, placebo-controlled trial. Double-blinding-in which neither researchers nor subjects know exactly who gets the active agent and who gets the placebo-has been around since a 1908 laboratory study on the influence of alcohol and other drugs on fatigue, but it did not achieve widespread usage until the 1960s (A.K. Shapiro, E. Shapiro, in The Placebo Effect, A. Harrington, ed., Cambridge, Mass., Harvard University Press, 1997, page 20). Although Stunkard helped to pioneer blinding, he is among many scientists who believe that its effectiveness is compromised by the numbers of subjects who can guess correctly that they've been given a pharmacologically active agent rather than an inactive placebo. Often, the side effects of the drug give it away. Albert J. Stunkard This creates a problem for researchers, because subjects who feel certain they're receiving a drug are more likely to be influenced by the placebo effect than if they have no idea whether they're getting an active agent. Trial results can be skewed, and the purpose of blinding, to eliminate bias, is undermined. One answer to the problem is to give a pharmacologically active placebo that will have side effects but will not directly treat the disorder in question. However, this creates an ethical difficulty, Stunkard points out: treatment is withheld from subjects who are then made sick. An obesity expert, Stunkard says that drugs being tested for binge eating disorders, which can be affected by behavior therapy, are matched against both the therapy and placebo, in order to demonstrate the new agent's effectiveness. "Now, some people are doing the worst possible behavior therapy they can, to prevent weight loss in the control group," he reveals. "I think trials will continue to have to use placebo, so it will be increasingly important--and increasingly problematic." He and Sidney M. Wolf, director of Public Citizen's health research group, were leading agitators for revision of 15 overseas HIV perinatal transmission trials. The majority of the trials were funded by the Centers for Disease Control and Prevention (CDC) and the National Institutes of Health, and most involved retroviral treatment alternatives to the expensive AZT regimen. CDC has now identified a less costly AZT course, and Lurie says the disputed trials either have been halted or placebo controls have been eliminated. However, Public Citizen is concerned about what it calls a "second generation" of such overseas trials that do not employ adequate AZT therapy for HIV-infected pregnant women. "There's not enough creative thinking in the field of clinical trials," Lurie believes. "People are not willing to go the extra yard, to think hard enough in order to avoid placebo-controlled trials." Alison Wichman, deputy director of the office of human subjects research at the National Institutes of Health in Bethesda, Md., observes, "It was very interesting how many people came out and said studies in Africa have to be done the exact same way as in the U.S. If that's not scientific imperialism, what is? The idea that AZT is the standard of care in those countries is ludicrous." Another type of FDA-approved design involves historical controls. But Wichman points out that such information gathered over time about a population of patients can be misleading, if the historical group has significantly different characteristics from that of the current group. For example, malnourishment compared to a healthy diet could affect the two groups differently, independent of drug interventions. Douglas E. Faries, research scientist for Eli Lilly and Co. of Indianapolis, has been involved in "adaptive designs" of clinical trials that incorporate information obtained during the study to alter the design. "We're not really eliminating placebo or active control, we're just limiting its use," he says. "The primary advantage is an ethical one, of reducing the allocation [of subjects] to ineffective treatments." Lilly has used such an adaptive design in Phase 1 cancer trials for years, but according to Faries, this is rare among pharmaceutical companies. Lilly also has experimented with drug dose escalation, in which dosages begin at pharmacologically ineffective levels to search for toxicity, until the mean tolerated dose is reached. Faries thinks eliminating an arm of such a trial, once another arm has been shown to have a more advantageous dosage, is an adaptive design with good potential. He is a strong supporter of randomization in trials, saying, "As a statistician, I think that's critical to avoid biases." But he has no comment on the use of double-blinding. Although innovative trial designs may emerge, it seems unlikely that placebo controls, despite their flaws, will be supplanted soon. Concerning Lilly's future reliance on placebo, Faries simply says, "That's the way we're headed."

Ethical Debate on Placebo Use May Prompt New Trial Designs

Interested in reading more?

Become a Member of