Seventy Years of Data

In 1950 a physician named Stewart Wolf conducted a (morally questionable) experiment. A woman had come into his office because she couldn’t stop vomiting. She was pregnant, had lost a lot of weight, couldn’t keep any of her food down, and was getting really dehydrated. The condition she was presenting with is called hyperemesis gravidarum, and Wolf told her he had just the thing for her; he had a powerful new drug that would make all of her symptoms go away.

Then he handed her ipecac.

If you don’t know what ipecac is, it’s not an anti-vomiting medication. In fact, it’s the exact opposite; ipecac is what you give to someone if they have ingested poison and you need them to immediately and violently clear the contents of their stomach. It was so touted and so trusted at this time that parents were encouraged to always have a bottle on hand in their medicine cabinet in case their children ate something they shouldn’t. And this is what Dr. Wolf handed to this woman who couldn’t stop vomiting, this drug that should have amplified her symptoms even more. He gave ipecac to his vomiting patient, told her it would cure her, and then just waited and watched and documented what happened.

And what did actually happen? Well, her nausea disappeared, and her vomiting…stopped. Completely. And Wolf documented it all. He measured her body’s response, like her gastric contractions and the other physical mechanisms of nausea and vomiting, and he saw all of it reverse. The waves of contractions that were forcing her stomach contents up through her esophagus were replaced by the normal, downward peristaltic waves of healthy digestion. And this was clearly not just perception, her body’s physical measurable responses completely reversed, and Wolf meticulously documented it all. He had given her a drug that was widely used, and that predictably caused violent vomiting, and its functional mechanism had been not just overridden but reversed by her expectation.

Wolf published his findings in the Journal of Clinical Investigation. He had demonstrated that belief could override pharmacology; that the body's expectation could reverse a very well proven chemical mechanism. And this wasn't the first time Wolf had documented remarkable observations from a single patient, and it wasn't the first time his colleagues dismissed him for it. When he'd presented earlier findings on stress and gastric function to the Gastroenterological Society? They…laughed. One case, they said. How could anyone draw conclusions from a single patient?

They had a point, I mean, one case truly isn’t proof of anything. But proof is not the same as intriguing evidence. It was just one case, but it was a case with measurable, documented, physiological changes. This should have at least sparked curiosity. It should have prompted the question: has anyone else seen this? Does this happen with other patients, other conditions? But instead, it was dismissed as an anomaly. An interesting outlier at best. Nothing worth investigating further.

How science should handle unexpected findings

To understand how unusual this response is in the name of science, we need to look at what was going on in physics at around the same time. In 1956 physicists noticed something weird about particle decay; the math didn’t seem to be working the way they expected it to. Most physicists at the time assumed it was a measurement error, as anomalies typically are, but they investigated it further anyway, just to be sure. And through this investigation two physicists, Lee and Yang, proposed something radical: maybe one of physics’ fundamental laws was wrong…maybe parity wasn’t actually conserved in weak interactions. Within months, experiments within the field confirmed their hypothesis, and the following year they won the Nobel Prize for their discovery.

Now let’s look at what happened in 2015 when physicists at CERN noticed a bump in their data at 750 GeV, which could have meant the discovery of a new particle. They were ecstatic, and excited, but also cautious. Not because they'd proven anything yet, but because they'd found something that didn't fit. They were so excited at the possibility of something new that over 250 papers were published analyzing this single anomalous signal at 750 GeV. Physics chases anomalies because occasionally one rewrites the textbooks. And physicists want to rewrite the textbooks. The scientific spirit is captured in statistician George Box's observation: 'All models are wrong, but some are useful.' Physics embodies this - the curiosity, the willingness to be wrong, the excitement when something doesn't fit.

Don’t get me wrong, medicine is different, because human lives are at stake. You can’t just rush new findings into treatment, and testing things on humans and animals absolutely has different ethical considerations than some of the other sciences need to contend with. Caution is not only justified but necessary. But caution about treatment is not the same as incuriosity about novel findings. In the case of Wolf’s finding for example, he wasn’t proposing a new therapy be implemented immediately, he had just documented something he had measured that had already happened: expectation mediated physiological changes that reversed the expected pharmacological results. Understanding how and when belief could override chemistry could be tested ethically and safely. If anything the stakes at hand should have made investigating the claims more urgent not less, because human lives are involved, and this if true could have far reaching implications.

Five years later, in 1955, Henry Beecher, an anesthesiologist at Massachusetts General Hospital, looked at 15 clinical trials spanning different conditions and different treatments to assess the placebo response Wolf was demonstrating at scale. Wolf had only one patient’s worth of data, but Henry Beecher was looking across 1,082 patients, and he found that 35% of these patients experienced “satisfactory relief” from placebo alone. This was no longer an isolated case, or an anomaly, it was a pattern across hundreds of patients and multiple conditions, and Beecher published his findings in a paper titled “The Powerful Placebo” in The Journal of the American Medical Association otherwise known as JAMA, which is one of the most prestigious journals in medicine. So now Wolf had shown it could happen - belief reversing pharmacology in one dramatic case, and Beecher had quantified it - 35% of patients showed satisfactory relief across different diseases. The phenomenon was clearly real, measurable, and reproducible. So what did medicine do with this discovery?

Beecher himself framed it as a problem. And Ted Kaptchuk, the director of Harvard's Program in Placebo Studies, has described Beecher's framing as treating placebo like 'the devil' in clinical trials—something to be controlled for and eliminated. His paper argued for placebo-controlled trials as a way to subtract the placebo response away and isolate 'real' drug effects. His focus wasn't on understanding how belief could produce physiological changes - it was on eliminating it from the data so that the ‘real’ science could show through.

When placebo trials became law

Then in 1962 Thalidomide, a drug that was being prescribed to pregnant women for morning sickness, caused over 10,000 birth defects worldwide. In the US alone, roughly 20,000 patients had received the drug in unregulated clinical trials, and the drug had seemed safe because no one had tested it rigorously enough to catch the effect it had on developing fetuses. There was enough public outrage and attention during this time that congress responded with new requirements for drug testing before they would be allowed to enter the market. Pharmaceutical companies would now have to prove both safety and efficacy before approval - not through observation or doctor testimonials, but through standardized controlled trials. This became the beginning of what we now know of as randomized controlled trials, and through this, placebo controls became law.

By the 1970s, beating placebo became the gold standard of proving efficacy. Every drug seeking approval had to demonstrate that it worked better than placebo alone. Which meant that every study was actively measuring placebo effects, and rigorously recording their outcomes. The irony was perfect. Medicine needed placebo controls to prove drugs worked, and in doing so, it accidentally documented - with expert precision, over seventy years, in hundreds of thousands of trials - the very mechanism it was trying to eliminate. For seventy years, medicine measured placebo in nearly every clinical trial ever conducted, with millions of patients all showing the same thing: placebo produces measurable physiological changes. And for seventy years, medicine asked only one question: how do we make it disappear?

In 2001, the Cochrane Collaboration appeared to have the answer. Hróbjartsson and Gøtzsche published a systematic review analyzing 130 placebo-controlled trials across 40 different clinical conditions, and their conclusion was definitive: placebo effects were weak, clinically insignificant, and possibly even nonexistent. The paper appeared in the New England Journal of Medicine - another of medicine's most prestigious journals - and medicine ate it up. Finally, rigorous analysis had confirmed what many suspected: placebo was mostly myth. The emperor had no clothes. The paper then became the most cited work on placebo ever published, it functioned as the definitive reference that shut down conversations about whether placebo effects were real. Researchers could now focus on real mechanisms instead of this troublesome variable that had complicated seventy years of drug trials.

Now to anyone examining the analysis with the same rigor physics applies to anomalous findings, it was pretty clear that Hróbjartsson and Gøtzsche's methodology wasn't scientifically sound. They had 130 trials showing placebo effects across different conditions, and some conditions showed consistently strong effects. But Hróbjartsson and Gøtzsche didn't report it that way - they averaged everything together. Parkinson’s with smoking cessation trials. Nausea with schizophrenia. Asthma with infertility. All 40 conditions pooled and averaged into a single analysis, measured against an arbitrary marker of “significance” decided on by the writers themselves, and through this pooling and averaging, the real effects disappeared into the noise.

This would be like averaging the effectiveness of antibiotics across bacterial infections, viral infections, broken bones, and depression, then concluding antibiotics don't work because the average effect was small. No drug would survive that analysis. No physicist would average the behavior of electrons across completely different experimental conditions and call it rigorous.

When they wrote up their results, Hróbjartsson and Gøtzsche emphasized outcomes where placebo showed no effect while minimizing the outcomes where it did show effects. They did acknowledge that placebo might have "possible small benefits" for pain and subjective outcomes, but then spent most of the paper explaining why even those effects probably weren't real - that they were probably just reporting bias, or possibly regression to the mean. Their conclusion wasn't that placebo works for some conditions but not others; it was that placebo effects don't exist in any clinically meaningful way at all.

And then there was the circular exclusion criterion buried in their methodology. They brazenly excluded trials where "the alleged placebo had a clinical effect not associated with placebo" - which meant they excluded cases where placebo worked too well, reasoning these placebos must have contained some active ingredient or in some other way represented something that wasn’t truly placebo. Any particularly strong placebo effect was, by their definition, not a placebo effect, and so they just…quietly removed it from their data set. When other researchers reanalyzed the same data, they reached opposite conclusions. Wampold and colleagues found the effect sizes were essentially identical to what Hróbjartsson and Gøtzsche had calculated - the difference wasn't in the numbers but in the interpretation. An effect size of 0.28 could be called "small and clinically insignificant" or "robust and meaningful" depending on your framing. For context, that effect size - the one dismissed as meaningless - exceeds many accepted medical interventions.

But the reanalysis came too late. For two decades, the Cochrane paper shut down investigation. Placebo was myth. The data had spoken.

Placebo research?

Then in 2005, Harald Walach and his colleagues decided to look at placebo data differently. They analyzed over a hundred clinical trials and found something medicine had been documenting but never noticed: when patients improved significantly on a drug, they also improved significantly on placebo. When a drug showed weak effects, placebo showed weak effects too. The two moved together, trial after trial, with a correlation of 0.78. When patients improved on the drug, patients on placebo improved by similar amounts. When the drug didn't work well, placebo didn't work well either. The correlation coefficient was 0.78 - meaning roughly 60% of the treatment outcomes could be explained by what was happening in both groups. Think about what that means: in a trial where 40% of patients improve on the drug and 35% improve on placebo, medicine celebrates the 5-point difference. The drug works! But 87.5% of what patients actually experienced - the relief they felt, the symptoms that resolved - was happening in both groups. Something other than the drug's pharmacology was producing most of the improvement. This pattern held across over a hundred trials. The thing that determined whether a trial would show strong effects or weak effects wasn't primarily the drug - it was something happening in both arms.

Decades of data showing that most improvement in clinical trials occurs in both groups, pointing to something massive and obvious, and the response was…pretty much nonexistent. Nothing much was said, and nothing much changed.

There were a small number of researchers looking at placebo effects. The world's first and only research center dedicated to understanding placebo mechanisms was established at Harvard in 2011. And in 2014, some researchers formed the Society for Interdisciplinary Placebo Studies, an international association that holds a tiny conference every two years. These teams made and are making real discoveries - they’ve documented endogenous opioid release in response to placebo, showed dopamine production in Parkinson's patients given fake treatments, and found that conditioning and expectation could produce measurable physiological changes. They proved Wolf and Beecher had been documenting something real. Yet despite seventy years of consistent findings and growing evidence of physiological mechanisms, medicine couldn't answer basic questions: When do placebo effects occur? Why do some conditions show massive placebo responses while others show none? What determines whether a patient will respond? The research was fragmentary, studying individual pieces without a unifying framework.

To understand why progress around placebo response remained so limited, let’s look at the numbers. As of the time of publication the NIH - which funds the majority of publicly funded medical research in the United States - is spending roughly $47 billion annually across all research. Cancer research gets $7.3 billion. Alzheimer's gets $3.8 billion. Diabetes gets $1.1 billion. The NIH tracks spending across more than 300 research categories, from rare genetic disorders to common chronic diseases, publishing detailed breakdowns of where every dollar goes. So where does placebo research rank on that list?

It doesn't.

Placebo isn't a category. Rare genetic disorders affecting a few thousand people have dedicated funding streams. Diseases with clear treatments and established protocols get billions. But the mechanism documented in hundreds of thousands of trials, affecting millions of patients, explaining 60% of treatment outcomes? Medicine doesn't even track what it spends trying to understand it, indicating it receives essentially no dedicated funding.

Because of this lack of funding, and the lack of investigation into findings that don’t match what we think we know, no one has thought to ask certain questions - to step back and look at the whole of the data and think: what if there IS something here. Not just does or doesn't this mean something, not just why might this be happening, but even more importantly - is it possible placebo is telling us something about our bodies we didn't even realize we were measuring?

What placebo actually is

The data is there if you step back and look at it - clues are everywhere, documented but virtually unexplained - and it is telling us something vital. Take blood pressure for example. When doctors measure blood pressure in the office there is a very large documented placebo effect. But when they strap a blood pressure monitor on you and send you home for the day, there is virtually none. Why, when they are measuring blood pressure in the same patient in both cases, does placebo present so differently? Cardiology literature has little to say about this mystery. Most of what you will see if you look for cardiology's explanation about this is just that home monitoring is more accurate than in the office, so that is the gold standard when precise measurement is needed. When they do try to explain it they invoke "white coat syndrome" most often, which is the effect of feeling nervous in the office which causes your blood pressure to rise. But this doesn’t explain it at all, because the placebo effects they're showing in the office cause your blood pressure to LOWER, not rise. Despite decades of research and billions in funding, their best explanation for why these two vital measurements vary so wildly doesn't actually account for what they're seeing. And to show how wild this is, Cardiology receives $3 billion annually from the NIH alone - which is only about a third of all medical research and development funding, meaning there is probably around $9 billion being spent after you factor in money coming from pharmaceutical companies. And they can't explain this basic discrepancy.

And blood pressure isn’t the only paradox. When measuring air flow out of the lungs, doctors conduct two main tests regularly, FEV1 and Peak expiratory flow (PEF). They test these two things in the same patients at the same time, and the placebo effects are opposite each other. One shows placebo, the other actually goes down. And what’s interesting is that respiratory medicine hasn’t even tried to explain this variation, they just do what cardiology does and recommend one as more reliable than the other. However unlike in cardiology where they’re recommending the more reliable and accurate test, respiratory is doing the opposite unknowingly.

Respiratory medicine chose FEV1 as the "gold standard" because it seemed more sensitive, more responsive to treatment. They interpreted its greater responsiveness as superiority. But that greater responsiveness includes the placebo effect. Medicine preferred the measurement that changed more - without realizing that greater sensitivity was partly placebo responsiveness. Peak flow, which was dismissed as "too variable" and "effort-dependent," was actually capturing something different: voluntary muscle force, which doesn't respond to placebo. The measurement chosen as superior was unknowingly chosen partly because it had higher placebo responsiveness.

But blood pressure wasn't the only paradox, and respiratory function wasn't the only contradiction. Medicine has been documenting these discrepancies for decades - measurements that should show the same thing but don’t, conditions that respond differently to placebo for no clear reason, and patterns that are noted but never explained.

So now is a good time for us to take a step back and look at what we have in front of us; really spread it all out and see what the data actually says. And there is a lot of it. Seventy years of placebo-controlled trials. Hundreds of thousands of studies documenting placebo effects across every area of medicine. Mountains of data showing which conditions respond to placebo and which don't, which measurements show effects and which show none. And when you actually look at all this data together - not trial by trial, but the whole picture - a pattern emerges that seems, in retrospect, almost obvious.

The prevailing assumption in medicine is that placebo only works on subjective measures - pain, nausea, things patients report feeling. Despite seventy years of data showing otherwise, that's what medicine, and subsequently researchers and physicians and the general public believe. Placebo is psychological. It's about perception, not physiology. So let's test that assumption, pull some levers and see what we find. Let's look at measurements that can't be explained by belief or perception alone - objective physiological changes that happen whether a patient is aware of them or not.

Coming back to blood pressure, let’s look at what the data is actually showing us. When patients are given sugar pills and told it will decrease their blood pressure, in study after study the data shows that their blood pressure drops. And not just a little bit - an average of 5-7 mmHg systolic, sometimes more. Real, measurable changes recorded by automatic cuffs, measurements that can’t be influenced by perception in the way we have been taught to think of placebo. These are the same measurements doctors use to do things like diagnose hypertension and prescribe medications. And though this is pretty astonishing, blood pressure can be kind of hard to conceptualize, you can’t see it and what does 5-7 mmHg really mean? Could it be a measurement error?

So let’s look at something a little different, something a little more visible. When you think of Parkinson’s you probably think of the shaking that defines it in our perception of the disease. It’s measurable and even visible to the naked eye. If you’re not familiar, Parkinson’s rest tremor is called a “rest” tremor because it happens when you’re at rest, and stops happening when you voluntarily engage your muscles. And the way doctors have found to treat this tremor is by increasing the dopamine in your brain using a drug called Levodopa, which gets converted to dopamine once it’s in your body. This drug reduces the shaking in rest tremor and is the first line of defense against this tremor in patients.

So when you look at the clinical trials for drugs that reduce shaking by increasing dopamine in the brain, and then look at what happens in the placebo arm of the trial, when these patients are told they’re taking a drug to reduce their shaking, you will see that their shaking measurably decreases. Not a little, but a LOT, by 25-45%. With some patients showing over 70% improvements. And this is pretty substantial evidence on its own, but what will strike you more is seeing that what is happening in the brain to produce these changes is the exact same thing happening in the active drug arm of the trial. The patients' brains are creating more dopamine. The studies show measurable increases in dopamine production in these patients receiving the placebo, creating improvements that, for some patients, rival any drug currently on the market.

We could go through more examples but I think you get the picture. Placebo is not just perceptual, it is acting on real physiological and biochemical activity in the body, and since placebo is required for most every clinical trial, there is a lot of data that shows just this.

Now to be crystal clear on what we’re talking about here, we need to also look at that same dataset and see if there are things that don’t show placebo effects. Because surely if it acted on every single thing the prevailing assumption wouldn’t be that it was merely psychological. And this is what the data shows as well. Not everything shows a placebo effect, in fact if you look at the two most studied functions from each of the major body systems and assess the amount of placebo effects on each one, only 10 of 22 of those functions show high placebo effects. Things like bone mineral density and fracture healing, ovulation and wound healing, all show low to no placebo effects reliably across the medical testing landscape. So what do the things with high placebo effects have in common? What do the things with low placebo effect have in common? What could this data be showing us that medicine has failed to see across the decades?

Let’s think about what the high-responders have in common. Heart rate. Blood pressure in the clinic. Breathing patterns. Gastric motility. Pain perception. Tremor at rest. These aren't things you consciously control, they happen automatically. So what about the non-responders: bone density. Wound healing. Blood pressure over 24 hours. Red blood cell production. Ovulation. These are also automatic, so it's not about voluntary versus involuntary. The difference is something else.

Let's look at heart rate again and see if it has anything to show us. Your heart beat can change pretty quickly. If you were to stand up right now your heart rate would increase within seconds. And then if you take a few deep breaths, it will slow down momentarily. Compare that to bone density: changes in your bones are happening over weeks or months, there's no moment-to-moment adjustment. So speed seems like it could be part of it, but maybe not quite the whole picture.

Let’s look at pain perception; that responds to placebo. But so does nausea. Yet pain signals travel through nerves at hundreds of miles per hour, while nausea involves chemical signals in the gut that move much slower. Different speeds, but both show placebo effects. So it's not just speed…what else? Maybe if we look at how these systems work: heart rate changes through direct nerve signals but bone density changes through hormones. Different types of control; direct neural signaling versus hormonal messengers. And look - pain and nausea both work through nerves. Heart rate, breathing, digestion - all controlled by nerve fibers. The things that respond to placebo all share this: direct neural control, real-time regulation, moment-to-moment adjustments based on your body's needs and your brain's perception.

But wait - muscle strength is also neural. When you flex your bicep, nerve signals travel from your brain to your muscle. That's direct neural control too. Yet muscle strength shows essentially no placebo response. So it's not just about having neural control - maybe it's about what KIND of neural control?

Voluntary muscle movement, the kind you control consciously when you decide to lift your arm, is mediated by your somatic nervous system. Heart rate, breathing rhythm, digestion, blood pressure regulation, these are all mediated by your autonomic nervous system (ANS). Somatic function you control consciously, and autonomic is automatic, meaning it keeps things running in the background and can adjust based on your body’s needs and your brain’s perception of threat or safety.

So let's look back to those apparent paradoxes we were looking at earlier and see if this helps explain what we were seeing. Clinical blood pressure, like when your doctor puts that cuff on you in the office, shows large placebo effects. And ambulatory, which is the kind you wear home for a day, shows virtually none. So are these being mediated by different systems? It turns out yes, they are. Clinical, namely short-term, blood pressure is controlled by baroreceptor reflexes - moment-to-moment adjustments your body makes in response to context and perception. ANS regulated. Whereas ambulatory blood pressure is controlled by renal mechanisms that operate over hours and days, and these aren't under ANS control.

What about the lung function example? FEV1 measures how much air you can expel in one second - think of it like a balloon that's really filled up versus one that's only half filled up. If you unplug the end and let air come out, which one will expel more air? The fuller one, right? Well it turns out that the smooth muscles involved in this involuntary process of pulling more or less air into your lungs are ANS mediated, which tracks with our theory. Whereas peak flow measures the maximum speed of airflow - how hard you can push air out using your voluntary muscles. That's not ANS mediated, it's somatic. So yeah, this explains them.

So what we are seeing is that the functions that respond to placebo all seem to be controlled by the ANS. And when you look at the data, it bears this out with remarkable accuracy. Those 22 functions we talked about earlier? They span every major body system and all match perfectly to this framework. This pattern holds up with remarkable precision across all 22 functions the likelihood of which, if this were to happen by chance, is less than one in four million.

For seventy years, medicine had been measuring placebo in hundreds of thousands of trials, documenting which conditions responded and which didn't. The pattern was there the entire time, medicine just never asked the right question. In fact medicine barely asked any questions at all. They just corrected for this nuisance of a noise, while systematically documenting something fundamental about how our bodies work, how disease actually works, and more importantly, what we can do about it. The ANS mediation we just discovered is one piece of the elephant. One, admittedly large, piece of the elephant, all sitting in their data, largely unexamined. And on its own it admittedly almost raises more questions than it answers. So next let's look at some of the diseases they have been spending time and money on, billions of dollars and decades of research, and see what that part of the elephant will uncover for us. Let’s start with cancer; arguably one of the most investigated diseases in history. Let's see what it can teach us about disease that we have been missing.