The Placebo Effect

"We know that four sugar pills a day beats two sugar pills a day . . . and that's an outrageous finding.”
-- Ben Goldacre, Epidemiologist

“[Alternative Medicine Advocates] tend to argue that what they do is useful and good because it’s 'harnessing the placebo effect' for therapeutic purpose . . . [but] a good placebo requires on some level at least some deception of the patient by either saying or implying that he is receiving an active treatment or medicine of some kind.”
-- David Gorski, M.D., Ph.D., Surgical Oncologist at the Barbara Ann Karmanos Cancer Center.

Slides for this lecture can be found here: PDF Version of Lecture Slides

Introduction

Medical claims abound. Check the headlines in any newspaper, on any news website, or on social media on any given day, and you will see amazing claimed medical benefits from a new treatment. Or, perhaps you will see claims about new medical benefits from an existing treatment, unrelated to the original purpose of the existing medical treatment. This is often labeled as "emerging science" or "new medical breakthrough." But, before we can trust new medical claims, or even old medical claims left over from before the Germ Theory of Disease, we need to ask a few critical questions:

What is the evidence for the claim?
Is the evidence derived from gold-standard scientific testing: double-blinded, randomized, controlled trials?
Even if the claim is based on evidence from such an experiment, is this a single experiment or have the findings been replicated at least once, and ideally more than once?
What is the "control" used when assessing the claim?

It's this last question that we will be focused on in this lecture. Specifically, it is not appropriate to test a new claim against literally doing nothing at all. Rather, it's important to test the new claim against placebo. What is placebo, and why is it so important? Let's find out.

The Definition of "Placebo"

“From the Latin 'I shall please', placebo entered ecclesiastical English in the 13th century, but did not appear in medical parlance until the late 18th century. Hooper's Medical Dictionary of 1811 defined the term as 'any medicine adapted more to please than benefit the patient'.”
-- Martin Edward. “Placebo.” The Lancet. Volume 365, Issue 9464, 19–25 March 2005, Pages 1023.

Simply put, a “placebo” is . . .

a medically inert substance given to a patient as if it were an effective medicine
- e.g. water, saline, sugar, a ritual.
a diagnostic tool, not a medical intervention
- Doctors can give a placebo to a patient and instruct them to call back in 24 hours – this gives the body a chance to simply do what it usually does: heal and/or fight infection.
- If the patient calls back and is the same or worse, the doctor is dealing with a potentially real disease requiring real medical intervention.

At this point, we can make some hypotheses based on the idea that a placebo is a non-medical intervention delivered to a patient as if it were medicine. We can then assess these hypotheses by looking at the medical literature to see how they stand up to reality.

People will not respond to a non-medical intervention as if a medicine had been applied
If there is a response, that placebo response can be predicted by factors measurable in the individual
Placebo response is inherent and cannot be manipulated (that is, an individual's response is pre-determined by their specific biological factors, e.g. genes)

The Placebo Effect

It turns out there there is a response in some fraction of patients who are unknowingly given a non-medical intervention; they can respond genuinely as if they are being given a real medicine. This is called "The Placebo Effect".

This is when a patient responds to placebo as if they had been given real medicine

they feel better, or the disease actually clears up
this is a real effect, but its power comes from the BELIEF and EXPECTATION that the placebo is real medicine. Degree of belief determines outcome.
effects vary – typically 20-30% of people respond positively to placebo, but it can be higher or lower.
- for instance, about 34% of patients in a placebo control group for treatment of irritable bowel syndrome reported positive effects.
  A. C. Ford and P. Moayyedi. “Meta-analysis: factors affecting placebo response rate in the irritable bowel syndrome.” Alimentary Pharmacology and Therapeutics. Vol 32. 2:144-158. 2010.

What factors influence belief?

Cultural associations
- For example, in the United States and other countries a white laboratory coat is associated with a medical professional. You might believe that a person wearing a lab coat is a better doctor than one wearing a ripped tee shirt and jeans, even if the latter person has twenty more years of medical experience than the former. This is an example of how authority can be concocted to promote a cultural idea about the medical abilities of an individual, conferring upon you the belief and expectation of healing.
Elaborateness and/or ritual
- For example, if a claimed healer goes through a complicated series of bodily motions (arms, legs, a dance of some kind, chanting, etc.) you might believe the healer is going to extraordinary lengths, so you should expect extraordinary outcomes.
color, context, history, etc
- As we will see, cultures that assign meaning to colors, or to events based on historical association, or any other context-based meaning, can be influenced at the level of a placebo response based on those expectations.

In short, a more complex ritual implies in the patient's mind a more desirable outcome. This is generall of the class of psychological phenomena known as "conditioning." See, for instance, the work of researchers like Ivan Pavlov, who famously conditioned dogs to salivate when a bell was rung by first ringing the bell before giving them food; at first, it is the food that causes them to salivate, but later, the bell could be rung and the food not given, yet salivation in anticipation of food being given would occur. This is known as "classical conditioning," and suggests that physical outcomes can be effected by first priming the subject.

In the case of placebo effect, we can ask: does it matter how the priming occurs? Is physical priming important? Is verbal priming important? Are both important? Is one more important than the other? Let's look at some examples.

Physical Placebo Priming: An Example Using Pain

Researchers now recognize that placebo control is especially important when assessing treatment of pain. Pain is very subjective. There is no objective measuring device that can tell a medical doctor the precise level of calibrated pain that you are feeling. Instead, subjective pain scales have been developed (this is a whole research area in and of itself!) which rely on the patient to self-report the degree to which they feel pain. What is excruciating for one person may me only mildly discomforting for another.

Studies have shown it's possible to condition a group of people to expect a positive or negative response from a treatment. Researchers like Pavlov (1927) and Herrnstein (1962) observed animals could be conditioned to expect medical results from non-medical interventions. What about Humans?

The peer-reviewed and published medical literature, including replication with different methodologies by different researchers, suggests that humans can be easily conditioned by physical means to expect relief or even MORE pain when given a placebo treatment for real, physical pain. Here are a few resources:

Voudouris, N. et al. “Conditioned Placebo Responses.” Journal of Personality and Social Psychology, Vol. 48, No. 1, 47-53. 1985.
Voudouris, N. et al. “Conditioned response models of placebo phenomena: further support.” Pain 38 (1989) 109-116.
Voudouris, N. et al. “The role of conditioning and verbal expectancy in the placebo response.” Pain 43 (1990) 121-128.
Montgomery, G and Kirsch, I. “Classical Conditioning and the placebo effect.” Pain 72 (1997) 107-113.

The Voudouris et al. Pain Trials

Let's look at just one example that is illustrative: the trials of placebo pain cream by Voudouris et al. Here is the idea:

Begin by giving a pain rating scale to participants in the study. Expose them to physically induced pain levels using electric current. For each person, calibrate the machine to their subjective pain rating using a 0-10 scale (0 is no pain, while 10 is the worst pain they've ever felt). Since each person experiences pain differently, it doesn't make sense to give each person the same amount of electric current since one person might rate that a 3 and another an 8. Instead, find out what a "5" means to each person, or a "7", or a "3", and keep those settings handy for the next phase of the study.
You can read the papers by Voudouris et al. to see what happens in each of their study groups. We'll discuss just two of them here. These researchers took their subjects, divided them randomly into 4 groups, and then experimented slightly differently on each group to see if placebo response could be conditioned differently, but using the same operating hypothesis, in each group.
For instance, one group was given a level 5 of pain in phase 1. They received a level 5 first without the application of placebo pain cream. Then they were given an application of placebo pain cream. The cream smelled "mediciny" and the patients were told it was a powerful new analgesic cream that was being tested. They were then given the same level of pain as before and asked to assess it. The researchers found a small placebo response during this phase - patients tended to rate the same level of induced pain lower after the application of the cream.
Phase 2 was conditioning. Unbeknownst to the patients, they were again fiven a level 5 induced pain without the cream applied. Then the cream was applied, and the patients were given a level 2 of induced pain; this is physical conditioning - they thought they were getting level 5, but really got level 2. They, of course, rated the cream as effective again.
Phase 3, the final phase, was designed to observe whether or not the physical conditioning in Phase 2 would hold. The patients were given a level "5" of pain with and without the cream, and asked to rate the pain. Compared to Phase 1, they reported a STRONGER placebo effect - a superior reduction of reported pain - compared to Phase 1. The placebo effect can be physically conditioned.

When thinking about what this means, it's important to keep in mind that at no time were the patients actually given pain medicine. They were always given a smelly cream advertised to them verbally as pain medicine. In phase 1, with only verbal conditioning, they already showed a placebo response. After phase 2, when they are tricked into believing the pain medicine is really working (by secretly being given less pain), they show an EVEN STRONGER placebo response. No medicine is ever given.

What is powerful about the Voudouris et al. studies is what happened in the other three groups. One of them was given a level "8" of pain - much worse than the group discussed above. That was secretly reduced in phase 2 to level "5". They reported an EVEN STRONGER placebo response conditioned in phase 3. The other two groups were given levels of pain that were INCREASED, not decreased, in phase 2 after the application of the fake cream. This physically conditioned them to believe the cream DIDN'T work. In fact, the researchers found in phase 3 that these groups showed a DECREASED placebo response. So placebo response can be physically conditioned in either direction - you can amplify the placebo response in patients, or you can degrade it.

Montgomery and Kirch studied whether or not verbal priming could undo physical conditioning - the opposite of what Voudouris et al. studied. Their findings suggested that verbal priming is very important, too, since it can UNDO the positive effects of physical conditioning. Again, at no time was actual medicine given during these trials.

Other Factors in the Placebo Effect: Ritual, Cultural Association, and Degree of Belief

Other studies have tested one placebo against another to see what role "degree of belief" plays in outcome. For instance, you might ask: does it matter if I get 2 sugar pills or 4 sugar pills?

The relative effects of different placebo interventions have been assessed (Meissner, K. and Linde, K. “Placebo and Pain”. “Chapter 23: What are the best placebo interventions for the treatment of pain?”. Academic Press. 2013.). For instance, Meissner and Linde reviewed trial literature on 79 placebo-controlled migraine pain prophylaxis studies. They found that sham acupuncture and sham surgery was more effective than oral placebo at relieving migraine or other pains, though neither is a real medical intervention.

But pain is subjective. What about less subjective illness? Duodenal ulcer is a great place to test placebo. Most ulcers are caused by bacteria, and modern treatment involves antibiotics and antacids (to control discomfort). However, unlike pain, you can stick a scope down a patient's throat and quantitatively measure the size of a duodenal ulcer. This allows you to objectively measure the rate of healing in a patient, and whether or not the cause of their discomfort has gone or not.

The role of placebo in affecting the outcomes of patients afflicted with duodenal ulcers has been studied. Craen et al. (“Placebo effect in the treatment of duodenal ulcer”, Br. J. Clin. Pharmacol. 1999, 48(6), 853-860) found that patients who got 4 placebo interventions each day over 4 weeks recovered at a rate of 44.2%, while those getting 2 a day recovered at a rate of 36.2%. (a 22% increase in recovery rate!). In other words, if you give a patient with duodenal ulcer more placebo treatments per day, the perception that this is a more dramatic intervention appears to induce more rapid healing - a strong placebo effect - than in patients with the same condition getting 2 placebo interventions per day.

What about cultural meaning and placebo? Blackwell et al (“Demonstration to Medical Students of Placebo Responses and Non-Drug Factors”, The Lancet 1:1279-1282, 1972) found that inert pills, colored red and blue, will be classified by US participants (who are told one is a stimulant and one is a depressant) more often as "red=stimulant" and "blue=depressant". A similar study in Denmark in 1996 repeated the cultural effect (de Craen, AJ et al. “Effect of color of drugs: systematic review of perceived effect of drugs and of their effectiveness.” BMJ (Clinical Research edition). 1996. 21-28;313(7072):1624-6.)

The de Craen et al. Drug Color Review found that, as regards the perceived actions of drugs: red, yellow, orange pills are associated with stimulant effect (but no stimulant is given) while blue and green pills were associated with tranquilizing effects (but no tranquilizer is given). Color and its effects on drug effectiveness color could degrade the effect of real drugs, and the drug industry knows this and uses it to color pills in a culturally effective way.

Taking Placebo Into Account: Medical Experimental Design

There is, as yet, no conclusive theoretical framework that can predict, for a single patient, their degree of placebo response when faced with a new medical treatment. Placebo is the result of the patient's extensive history of cultural, physical, and verbal conditioning. This means that there is no means by which you can reliably test a medical claim without also testing the same claim but substituting a fake version of the treatment. You must use MANY patients in the study - single patients and anecdotes won't do. They are confounded by the placebo effect to a degree that cannot be predicted.

Given the complexity of the placebo effect, we can see that one needs great care in designing a trial of a medical claim:

At least one placebo control group must be in place
Placebo and medicine must match: smell, color, size, texture, taste, ritual, verbal description, etc.
No cues can be given as to which is the placebo.
The trial must be double-blind, ideally: the researchers should be fire-walled from knowing what treatment (none, placebo, medicine) each patient gets and the patient should not know which they are receiving.
Randomization: patients should be randomly assigned to groups, and the only care here is to make sure the randomization doesn't skew a group in some way (age, medical history, etc.)

Any medical claim based on evidence gathered by any means other than the above should NEVER be trusted.

Thinking about Placebo and Medical Claim Testing

Let's think about how to design placebo for different claims

a physical intervention (e.g. acupuncture, chiropractic, surgery)
a pill
a ritual, like a faith-healing event

How would YOU design a placebo control for claims about a medical benefit from any of the above?

We might also ask, "does science-based medicine have anything to learn from alternative medicine?" Since alternative medicine is, by its own definition, alternative to science- and evidence-based medicine (requiring no gold-standard testing to make claims and sell its benefits), can real science- and evidence-based medicine learn anything from a field entirely devoted to packaging the placebo effect?

We'll investigate the efficacy of the actual treatments in a few cases (see Alternative Medicine) and test them against placebo. But, the only clear lesson so far from Alternative Medicine is that patient-doctor relationship and attentiveness of the doctor to the patient can have an effect on outcome, independent of the effectiveness of the treatment (ritual, personal interaction, etc.). That is the only thing that alternative medicine peddlers have to teach the medical profession; the rest of their art is smoke and mirrors.