Placebo Effect Stronger Than We Thought?
Double-blind trials have long been considered the gold standard to determine drugs’ effectiveness. Do we need to rethink that assumption, given the power of the placebo effect?
In July 2001, the Amgen Corporation announced the failure, in a second-stage clinical trial, of an experimental drug to treat Parkinson’s disease, a degenerative illness that affects nerve cells in the brain. Such a failure was hardly unusual; only a minority of the drugs that undergo trials make it to the marketplace.
But for Perry Cohen, who had been diagnosed with Parkinson’s several years before, at age 50, the announcement brought both surprise and disappointment. Cohen, an MIT-trained PhD who had spent decades advising health-care organizations on how to evaluate medical care, had hopes for the Amgen drug, called NIL-A. Studies on animals had suggested that the drug could help regrow damaged nerve cells. Cohen had figured that if it worked, and he could get that kind of therapy early in his disease, he might have a chance of slowing or reversing the process that would otherwise rob him of the ability to control his body. He knew that the announcement very likely meant the end of Amgen’s investment in NIL-A.
Cohen suspected, and still believes, that something was wrong with the conclusions being drawn about some of these trials. Over the intervening years, a number of Parkinson’s treatments that appeared even more promising also failed in Phase II trials, and in just the same way, after producing significant and often long-lasting improvement in patients who took them in Phase I. In the case of a treatment called GDNF, patients who took part in Phase I trials believed so strongly that the treatment was effective that, when the drug was declared a failure in Phase II, they sued to force the manufacturer to continue providing it to them. The courts ultimately rejected their claim. These treatments failed, according to the scientists evaluating them, because, usually at six months, results were no better than those from a placebo. But the enduring improvement that some patients did experience, Cohen says, should not have been discounted. Some results showed intriguing findings. In one patient who died of unrelated causes after taking one of the experimental drugs, an autopsy revealed new growth of the very nerve fibers that Parkinson’s affects.
The Jan-Feb 2012Miller-McCune
This article appears in our Jan-Feb 2012 issue under the title "Head Games." To see a schedule of when more articles from this issue will appear on Miller-McCune.com, please visit theJan-Feb 2012 magazine page.
For Cohen and fellow members of the Parkinson’s Pipeline Project, an advocacy network that he founded and now directs, there is evidence to suggest that treatments such as GDNF worked; it was the method used to judge them that didn’t. The problem with the rejected experimental treatments stems, the group claims, from overreliance on the long-revered, double-blind, placebo-controlled clinical trial — the method that science has considered for decades the most reliable in determining medications’ effectiveness. That so-called “gold standard” has, Cohen says, become a “golden calf.”
Because of research in recent decades, new understandings of how the brain works are undermining the assumptions that support one of the main elements of that traditional trial design: the placebo control. Finding the truth about experimental drugs — and the new medicines for diseases like Parkinson’s that patients need — now requires, he says, a better understanding of placebos.
• • • • • • • • • • • • • • •
In the decade since the NIL-A announcement, Cohen has given talks at scientific meetings, helped maintain a website that tracks and critiques research, served on governmental and industry advisory boards, and written papers published in peer-reviewed journals. Building on his expertise, and his decades-long health-care advisory career, he has become an internationally known advocate for reform in trial design. He has presented his views in person and in writing at some of the highest councils of medical decision-making in the United States and Europe.
Yet while his activism has advanced, so have the effects of his disease. A former skier and bicyclist, he has used a number of available treatments, but his strength, energy, and ability to control his muscles have inexorably declined. He now stands with the forward-bending stoop, often walks with the small steps, and speaks with the hushed voice typical of advanced Parkinson’s patients. Retaining his strength and mobility requires a complicated daily regimen of medications. “Time is not neutral for someone with a serious chronic illness,” Cohen says.
The process of bringing a new drug to market generally takes about 10 years. The company sponsoring the treatment must first show the Food and Drug Administration extensive data from animal research in order to get permission to try it on humans. Once permission is granted, the first phase of clinical testing on people is intended to produce initial information on safety, side effects, and dosage, but not on whether the drug actually works as hoped. This phase uses a small group of subjects who know which drug they are receiving; for this reason it is called an “open label” trial, in contrast to so-called “blinded” testing, when subjects or researchers (or both) don’t immediately know who is getting the active drug. Sometimes, as with “failed” Parkinson’s drugs, volunteers in Phase I trials experience what appear to be beneficial effects. Yet even though the trial seems to provide temporary medical treatment, it is not meant to serve as a remedy for patients. The trial’s primary purpose is the gathering of scientific evidence.
Should the drug prove sufficiently safe, it moves to Phase II, the first formal test of its effectiveness. This phase pits the experimental treatment against something administered to a control group, which can be either a treatment known to be active against the disease or a placebo that is presumed to be inactive. Patients are randomly assigned to an experimental group or to a control group. When an existing — or “standard” — treatment is available, as with many cancers, the control group may receive that therapy. The trial then tests whether the new treatment exceeds the results of the standard treatment. When no established treatment exists, such as in Parkinson’s trials, the control group can receive a placebo. In a double-blind trial, neither the researchers, study evaluators, nor the subjects know who has received what until conclusions are drawn and the results are revealed, or “unblinded.”
In some trials the placebo may be a prototypic “sugar pill,” but in certain Parkinson’s trials it can be something as elaborate as an operation implanting catheters in the brain, which then administer only a saline solution. Or it can be a “sham operation” in which doctors drill into patients’ skulls but do not implant a device for delivering either a drug or a placebo. These supposedly “inert” procedures require control group members to undergo the anxiety, stress, and anesthesia associated with brain surgery — for purely experimental purposes.
To critics, such methods are rooted in a faulty premise. They are rigorously designed to protect patients at nearly all costs from what statisticians call a Type I error, or a false positive — concluding that a treatment is effective when in fact it is not. But Cohen maintains that for Parkinson’s, this design frequently permits Type II errors, false negatives — concluding that a treatment is not effective when in fact it is. That second kind of error, he and others say, deprives patients of potentially valuable ways of stopping or reversing chronic degenerative diseases. In general, Cohen notes, the better a trial is designed to protect against Type I errors, the likelier it is to produce a Type II error.
• • • • • • • • • • • • • • •
Cohen and his collaborators are not the first patient activists to challenge the methods of testing drugs. At the dawn of the AIDS epidemic in the 1980s, patient groups demanded changes in the rules of drug testing to permit dying patients access to potentially life-saving experimental medications before they gained approval. Back then, when very little was known about how to treat HIV and AIDS, the FDA stringently restricted unapproved treatments. The agency was acting on a strong commitment to protect the public from potentially unsafe medications, which grew out of a high-profile battle in the early 1960s over the drug thalidomide. (In Europe, many pregnant women took thalidomide as a sleeping pill and to mitigate morning sickness. And many of those women bore children with severely deformed or absent limbs. Despite strong pressure from manufacturers, the FDA refused to approve thalidomide for use in the United States. The decision came to be considered one of the agency’s regulatory triumphs.)
The thalidomide controversy also inspired Congress to give greater authority to the FDA to require both strong evidence that drugs are safe and effective, and promulgate strict rules concerning who could take them before approval. In the 1980s, AIDS activists argued that, for people dying of a disease with no treatments, the chance that an unproven drug could prolong their lives far outweighed the risk that it could do them harm. Ultimately, the FDA agreed to alter the rules to permit terminally ill patients to use experimental, unapproved drugs outside of clinical trials. This balance, between regulatory vigilance and experimental freedom, is once again being called into question by activists, especially those concerned with Parkinson’s disease — for which placebo-based trials present a special kind of challenge.
• • • • • • • • • • • • • • •
The cluster of Parkinson’s symptoms that afflict an estimated 1 million people in the United States — including some 60,000 newly diagnosed each year — first came to medical attention when a small monograph was published by a London doctor named James Parkinson in 1817. Still considered a masterpiece of clinical observation, An Essay on the Shaking Palsy describes six individuals, three of them examined by Parkinson and the rest strangers he observed in public places. The unidentified malady, Parkinson correctly noted, first shows itself subtly, then takes years to reach its ultimately crippling effects, while often sparing the “senses and intellects.” Parkinson also presciently surmised a source in the brain but guessed the wrong part. He knew of no effective treatment, nor did the next major researcher to tackle the disease, the French physician Jean-Martin Charcot. Charcot, however, added significantly to Dr. Parkinson’s description and suggested naming the disorder after the man who had first identified it.
There is still no way to cure the disease or even to slow its course. There are some treatments that can alleviate — or, in the words of the advocacy group Parkinson’s Action Network, “mask” — some of the symptoms, at least for a time. Eventually, and often as soon as four to eight years, side effects of those treatments increase, limiting the doses that people can tolerate.
The immediate cause of the disease is loss of the brain’s natural ability to produce dopamine, though why this happens is unknown. Dopamine belongs to the class of molecules called neurotransmitters, which carry signals among brain cells. Its many roles involve mood and voluntary motion. Standard treatment includes supplying drugs that brain cells convert into dopamine or treatments that, according to the National Institutes of Health, “mimic the role of dopamine in the brain.” Other medications can help control symptoms. A technique called deep brain stimulation, which surgically implants electrodes in the brain, has produced marked improvement in a number of people.
• • • • • • • • • • • • • • •
There’s one more treatment that seems to have, in some cases, eased symptoms of Parkinson’s disease: the administration of placebos. How should such results be understood?
The prevalent assumption of researchers has long been that placebo treatments are inert, and that “placebo and treatment are independent, and that there’s no interaction effect between” them, Cohen says. From this idea flows the conclusion that any sign of improvement among patients receiving a placebo must indicate the presence of some kind of experimental bias, by which scientists mean some factor that systematically distorts experimental results in favor of a particular conclusion. This logic served medical science well for many decades and has accounted for huge advances in both knowledge and cures. In cases of treatments intended to cure infections, kill microbes, or serve as vaccines, for example, inactive controls — placebos — can reveal whether or not a treatment really works.
But suppose the assumption that the control is inert — that it has no effect and is independent of any results from the treatment being tested — is not always correct. The human brain doesn’t act like infectious microbes or research animals that cannot, so far as is known, believe that a treatment will make them well. Nor, as healers have understood for centuries, are positive changes from inactive treatments necessarily bogus. Rather, a placebo effect, even if caused by a well-intentioned sugar pill, can bring real improvement in a human patient’s condition.
This has been well accepted since 1955, when Harvard anesthesiologist Henry Beecher published an influential article in The Journal of the American Medical Association called “The Powerful Placebo.” An incident of benevolent deception that he had witnessed during World War II inspired Beecher to undertake postwar research on the placebo effect. He had seen a nurse tell a wounded soldier he was getting a shot of morphine when all she had to give him was salt water. The man’s severe pain abated nonetheless. After the war, Beecher studied existing research and became convinced that about 35 percent of patients showed improvement that could be attributed to placebos. The news riveted the medical world.
In a 2010 article in The Lancet, Damien G. Finniss of the University of Sydney Pain Management and Research Institute, and co-authors, wrote, “Placebo effects are genuine psychobiological phenomenon” that occur because of “the overall therapeutic context” and can happen both in the laboratory and the clinic. Research well supports the fact that, unlike experimental animals, people given placebos hope or expect that treatment will make them better, but recent findings also show that there is a connection between Parkinson’s and placebo use that is even deeper than hope or expectations.
“In Parkinson’s disease … the placebo effect is associated with release of endogenous dopamine” — the very neurotransmitter in short supply because of the illness — in two areas of the brain, write Sarah C. Lidstone and colleagues at the Pacific Parkinson’s Research Centre, in the Archives of General Psychiatry. It’s possible that the placebo effect may act on the very brain pathways involved in the disease.
Such findings about placebos and dopamine in Parkinson’s patients indicate that old assumptions need revision, argue Finniss and other researchers. The late J. Stephen Fink, a Parkinson’s researcher who chaired the department of neurology at Boston University School of Medicine, for example, noted in an article on the American Parkinson’s Disease Association website that the gains that placebo recipients experience in Parkinson’s trials may even result in part from “actual physiological changes in the damaged brain dopamine nerve cells.”
These developments require researchers to “reconsider placebos and placebo effects,” looking not at what they can’t accomplish, but at what they are “actually doing to the patient,” write Finniss and co-authors. A paradox lies at the center of the traditional concept, they note, because an inert agent, by definition, can’t have an effect. Therefore, the role that a patient’s expectations might have in the results of a trial needs to be understood as an important factor. Could placebo responses possibly trigger or even enhance the effects of treatments? The reconsideration that Finniss and co-authors are advocating could give placebo reactions a specific role in treatment techniques and even “encourage the use of treatments that stimulate placebo effects,” they write.
This view, however, hasn’t won over the medical community, at least in the United States and Canada. A 2005 survey of North American Parkinson’s researchers found that 97 percent of those responding preferred the most drastic kind of placebo control — sham brain surgery — to unblinded studies of surgical treatments; half considered a study lacking a blinded control “unethical because it may lead to a false positive,” according to Scott Y. H. Kim of the University of Michigan and colleagues in Archives of Neurology. Fewer European experts share these beliefs, according to a 2004 survey in which only about half of the European researchers responding thought sham surgeries justified. From presentations and comments made at a National Institutes of Health symposium on sham surgery held in Washington in 2010, it is clear that opinions have not shifted.
As for Cohen, his argument is that placebos need to be used — but with a greater understanding of what they do and don’t reveal about Parkinson’s. He favors a shift toward research that takes seriously the possibility of relationships between placebos and treatments. If placebo-controlled trials are used, he wants criteria of success or failure that allow sufficient time for placebo effects — known to be especially powerful and long-lasting in Parkinson’s — to wear off before judging whether real differences exist between the people who took experimental medication and those who did not. Even better, Cohen maintains, would be single-blind trials testing experimental treatments against best standard treatment. The patients would know which treatment they got, but the experts evaluating the responses would not. This, he believes, could equalize the placebo effect without subjecting patients to the considerable emotional stress some suffer from uncertainty about which treatment they are receiving, or from learning they had been given a placebo. FDA laws, he points out, don’t require placebo controls but only “well-controlled” trials. Experiments in deep brain stimulation, for example, received approval without a placebo trial. All patients in this trial had electrodes surgically implanted in their brains. When electric current was turned on, they experienced improvements in their symptoms. The differences observed when the current was off served as the trial’s control.
Roger Barker, a Cambridge University neuroscientist who is leading an international study called Transeuro, believes the failure of some earlier studies resulted not only from placebo effects, as scientists said, but from enormous variability in the patients and methods used within a number of the studies done for this complex disease. These include variation in the nature and stage of the patients’ disease, and in the techniques and materials used in studies that have surgically infused treatments into patients’ brains.
To confront these issues, Transeuro will evaluate, over a number of years, a consistent surgical treatment. A group of patients that is uniform in a number of factors relevant to the trial’s outcome, such as age and the stage of disease, will be selected for the evaluation. A subset of this group will receive an experimental surgical treatment. After two years, the condition of the patients who received the treatment will be compared with that of the remaining patients, who will have received standard therapy. Only if the experimental treatment uniformly produces benefits will a subsequent phase of the study take place to compare the experimental treatment with a placebo, which could involve sham surgery. Should the first phase not produce a significant effect on the patients who receive the experimental treatment, the researchers will not proceed to the second, placebo-controlled phase. Instead, they will work on refining the first-stage procedure until a consistent effect emerges.
In such experiments, researchers are working their way toward modifications to the experimental model that scientists originally devised for studying simpler problems, in hopes of making it more appropriate for investigating diseases of the human brain. Cohen echoes Finniss and co-authors in arguing that the human brain’s normal response to treatment is not experimental bias or distortion, but rather an element of natural human healing processes that are inherent in all medical interactions, both the purely curative and the frankly experimental. Studies designed to recognize the possibility of real interactions among all factors — including placebo effects — would harness the results of hope and expectation for patients’ benefit, rather than dismissing them as detrimental to science.
Whether they be the conceptual breakthroughs of Galileo’s heliocentric planetary system, Einstein’s theory of relativity, or of innumerable smaller advances, reconsideration of assumptions — what the philosopher of science Thomas Kuhn termed paradigm shifts — have been fundamental to scientific progress. Perhaps it’s time for another.