Get access to over 1,300 articles and reader Q&As
Recently, before a long run, I consumed a shot of broccoli sprout juice. I did this because my friend Shana told me that all the Norwegian cross-country skiers use it and that when she tried it on a run she felt “weightless.” I did not feel weightless; I felt like garbage. But when I complained to Shana about this later, she told me that maybe the problem was that I just didn’t believe in the broccoli juice enough.
She’s probably right.
The placebo effect — where a fake treatment changes your health outcomes — is a real thing. It’s one of our most reliable effects. It’s also a huge problem when it comes to designing and interpreting research.
I spend a lot of time pushing back on various wellness trends. Invariably, when I do this, I get reactions from people who insist the trend works for them. And, at least sometimes, their claims are supported by some studies or data. But the fact that people feel something works is not actually evidence that it does. And unless studies are well-designed, they can easily mistake placebo effects for real impacts. This is a problem, at least sometimes (it’s also an opportunity).
What is the placebo effect? How do we know it’s real, and how big is it? And why does this matter so much when we are trying to learn from data?
What is the placebo effect?
Put simply, the placebo effect is when people respond to a treatment when they think it’s real, even if it’s fake. If you give someone a sugar pill and tell them it’s a painkiller, they will report less pain than someone who has been given nothing.
This effect happens in clinical trials, but also in some version in our lives all the time. Our beliefs play an incredibly important role in how we feel. If someone tells you you slept great, you’re going to feel better during the day than if they tell you you slept poorly, even if the sleep is exactly the same.
The placebo effect has been recognized, presumably, for centuries. But the best-known reference is a 1955 paper by Henry Beecher, which showed that patients in clinical trials improved with a placebo (sometimes a lot). Interestingly, data shows that placebo treatments can have impacts even if people know they are getting a placebo.
This isn’t just reporting bias or some kind of desire to please researchers. Brain imaging studies of placebo pain relief show evidence that the brain regions responsible for pain actually visibly react to a placebo. In a 2021 meta-analysis covering about 600 people, researchers identified some particular (pain-related) regions that were responsive when placebo pain relief was used. In other words, the placebo effect is a real thing in the brain.
One of my favorite analyses of placebo effects is this paper from 2006. The author finds that the placebo effect — the impact on the untreated group in the trial — is larger if the treatment probability is higher. In other words, among people who do not get the treatment, the response is greater if people are more likely to think they got it. This is both very strong confirmation of the existence of a placebo effect in the first place and evidence that the more you believe in this, the more it matters.
Why does this matter?
I want to answer this question by talking about knee surgery for arthritis. Arthroscopic debridement is a type of minimally invasive knee surgery in which a surgeon makes some small incisions and “cleans out” the knee joint. The idea is that this would improve pain and function for people with arthritis and related conditions.
This surgery was very common in the 1980s and 1990s and still is to some extent. This was largely because evidence suggested that patients’ conditions seemed to improve. An example paper: among 160 patients followed for five years after surgery, more than half improved — often by a meaningful amount.
It turns out, though, that the improvement was actually just a placebo effect. A seminal 2002 paper in the New England Journal of Medicine looked at 180 patients with knee arthritis. They divided them into three groups — the first two got a type of surgery commonly used to treat this issue. The third group got a sham surgery. In other words, they were put into surgery, and the surgeon made the incision in the same way it was done in the true surgery but didn’t do anything else. The surgery was therefore a true “sham placebo” — designed to make it as hidden as possible to the subjects which group they were in.
The study found no differences across groups on pain or function. That is, there was no evidence the surgery was better than a placebo. Everyone improved — as you’d expect based on the existing evidence — but the actual surgery was no better than just cutting people’s knees and sewing them up.
This study is both very practically useful (don’t do this kind of knee surgery!) and also a clear illustration of why we need placebo controls. This study also illustrates the importance of a really thoughtfully designed placebo. The authors could have compared the people with knee surgery to a group who had nothing done, but that would not be a good placebo because the people without surgery would have known they were the untreated group. It was only because they did a true sham placebo here that the authors were able to draw the conclusions they did.
When we interpret studies, including those with control groups, the details of the placebo are really important. For example, meta-analysis evidence suggests that sham acupuncture (done at the “wrong” points) is just as efficacious as true acupuncture. As a result, if you’re studying the impact of some particular acupuncture approach, you cannot just compare people who get the acupuncture to people who do nothing — you need everyone to at least think they are being treated.
A question you may ask is why don’t all trials have placebos? Is there any reason not to have a full, no-treatment placebo in a medical setting? The answer is yes, often, especially when we are testing a new treatment against something that is already known to work. In oncology, for example, new chemotherapy regimens are tested against the current standard of care, not against doing nothing, since that would be unethical in addition to not telling us whether the new treatment is better. Similarly, when we test new vaccine formulations, they are tested against the existing vaccines, not against nothing, since that would violate ethics.
But if the placebo effect works, what’s the problem?
A lot of things are sold to us as effective that are probably just a placebo effect. Red light masks come to mind. And we all know people who swear by some particular combination of supplements.
One take on this is — who cares? Yes, you might waste some money on a red light mask, but if you think it works and you’re happier with your face, then what does it matter why? It has achieved the goal that you intended. Even though you may have paid for it, it’s not really “different” than paying for a true treatment. Placebos can also be really helpful — there is evidence that even sham acupuncture can reduce nausea and vomiting in pregnancy.
Where I think we want to be really cautious is when studies that do not correctly adjust for the placebo effect are driving us to do things that are costly or have significant downsides. Knee surgery is an example. There are also things like injectable peptides, which may have serious risks, and in many cases, the effects we see reported could easily be explained as a placebo. In the fertility space, there are expensive and invasive treatments like “endometrial scratching” that may look effective, but, in fact, the results are entirely placebo effect.
The reality is that there are a lot of cases where we really do want to know if a treatment actually works, and for that, we need to address the elephant (placebo) in the room.
P.S.
The placebo effect can also be very useful, including with your children. People really do feel better when given fake treatment. And although I wouldn’t advocate lying to your children, it might not be so terrible to have a bottle of sugar pills that “cure” stomach aches. I’m just saying.
There actually is a bit of very good randomized, double-blind, placebo-controlled evidence on the broccoli juice. The placebo control is alfalfa sprouts, which is a good control because it probably also tastes awful. So now I’m back to being a believer, and I’m going to try again. If nothing else, my belief should strengthen the placebo.
Community Guidelines
Log in
The question this article raises for me is what to do when the placebo has a measurable and significant effect. For example, the knee surgery: what should people do then to create this effect, since it does actually improve pain and function? Totally agreed that people shouldn’t be having unnecessary knee surgeries but it probably also isn’t feasible to do sham surgeries. So is there a way to produce the effect with some other method?