The power of “know-cebo”: 5 Questions for data scientist Balázs Szigeti

Szigeti discusses expectancy effects, unblinding, and how researchers might design better studies on psychedelics.

Balázs Szigeti’s path to studying psychedelics has been a meandering one. As an undergraduate at Imperial College London, he studied physics, then moved to the University of Edinburgh for a PhD in computational neuroscience. Just as he was finishing his doctorate, Szigeti started thinking about designing his first psychedelics study, which he initially envisioned as a randomized controlled trial. But such studies are costly and he didn’t have funding so Szigeti moved to New York for a postdoctoral scientist position where he built software to simulate bacterial metabolisms.

In his free time, Szigeti continued planning his psychedelic study, and settled on an innovative design, in which he recruited citizen scientists and trained them to use a “self-blinding” procedure to investigate whether microdosing was more effective than placebo in changing participants’ mood, creativity, anxiety, and energy levels. He’d been corresponding with Imperial College London researcher David Erritzoe while designing the study, and eventually, he decided to move back to London, living off his savings to work on his research. He eventually landed a job as a postdoctoral researcher at Imperial College London working on psychedelics studies.

In 2023, Szigeti moved to the U.S. to work as a data scientist at UCSF’s Translational Psychedelic Research Program, where he’s been a staunch proponent of open science, and an advocate for innovative study designs and research methods. Some of the concepts that Szigeti explores in his work include expectancy effects (how participants’ expectations can affect clinical trial outcomes) and unblinding effects (how participants’ knowledge of what treatment they receive can introduce bias) — and how each of those might contribute to a “nocebo” effect in psychedelic studies, where participants may experience negative outcomes if they believe they’ve been given a placebo instead of a psychedelic drug. The Microdose spoke with him about how researchers might design better studies on psychedelics.

What is “nocebo,” and where have you seen it in psychedelic studies?

Nocebo effects traditionally refer to placebo effects that affect people in a negative way. The mechanism behind a placebo and nocebo are exactly the same — it’s just that our value judgment of it is different.

I got interested in this recently because there was a metaanalysis that compared results from the placebo arms of psychedelic studies with studies of traditional antidepressants. What they found was that the changes between baseline and end point measures using the Hamilton Depression Scale — one of the most frequently used scales — was smaller for psychedelics than for traditional antidepressants. What this means is that you’re actually seeing a smaller placebo effect for psychedelics than with antidepressants, which is interesting because typically people are worried that the placebo effects for psychedelics are too high.

I saw other examples of this in two recent announcements from psychedelics companies, both with depression trials — one used 5-MeO-DMT, and the other psilocybin. In both of those trials, patients in the placebo group actually get worse. That led me to look at another metaanalysis using traditional antidepressants, which included 304 placebo conditions, which they compared to results from antidepressants. In all of those cases, participants improved in the placebo arm. Patients getting worse in the placebo arm? You only see that in the psychedelic trials, which tells me that there’s something going on here.

So what is happening in psychedelic trials that would lead to patients in the placebo condition showing worse outcomes when in the antidepressant studies the people getting the placebo felt better?

I think this all has to do with the issue of unblinding. Patients in psychedelic studies often very easily can figure out whether they have been randomly assigned to the active psychedelic or control condition. There’s a term gaining momentum in psychedelic circles: the “know-cebo effect.” If patients know they are in the control condition, that could lead to some disappointment. I haven’t met anybody who would sign up for one of these studies and want to be in the control condition; everyone wants the active psychedelic. That’s not surprising; that’s true for any drug trial, but what’s different for psychedelics is that it’s very easy to figure out which group you’ve been assigned to. For SSRIs, the unblinding rate — the percentage of participants who can guess which condition they were assigned to — is much lower.

Last year, I published a paper analyzing expectancy effects in the well-known clinical trial comparing the antidepressant escitalopram with psilocybin. In that study, at baseline, patients had much lower expectations regarding escitalopram than psilocybin. That was not surprising. But here’s the thing: if you analyze the data from the trial as it is, the improvement in the psilocybin arm is about 11 points on the Hamilton scale, and 5 points in the escitalopram arm — so a 6 point difference. But if you adjust the trial results for this expectancy imbalance at baseline, that 6-point difference shrinks to just 2 points: the psilocybin arm improvement remains 11 points, but it increases to 9 points of improvement in the escitalopram arm. What seems to be happening is that expectancy effects are affecting control arms more than the active arms of studies.

In the psychedelic space, everyone seems to be worried about overenthusiastic hippies being in these trials, and their positive expectations biasing the results – but I think that’s a red herring. The real issue seems to be the negative effects happening in the control arm.

What methods might help reduce these effects?

I think, given the data, that people’s expectations tend to be stronger in the control arm than in the active arm, and so I’m starting with the assumption that this is the case. If we accept that, there’s one study design, called the Zellen design, that I think could be a useful solution. It’s a study design from the 1970s that no one is really using any more, but it’s really simple and elegant: it switches the order of consent and randomization of participants. In a traditional trial, a researcher sets out to study how psilocybin can treat depression, and when people reach out to learn more, they’re told the details about what study arms there are, and then they consent. After consent, they’re randomized into one of those study arms. At that point they already know there are different arms.

In the Zellen design, when you are recruiting patients, you’re just telling them you’re running, say, a clinical trial about depression. That’s all you’re saying. As soon as the patient is in touch with the trial team, they’re being randomized into a condition, like receiving psilocybin or escitalopram. The study team would tell them it’s an open label study with psilocybin or with escitalopram; they would not say there is another study arm. Then the patient would consent or not. Those in the escitalopram arm — or whatever other control researchers want to use — would not be aware that they could have received psychedelics at all. It’s a clever way to eliminate the know-cebo effect because they don’t even know there was another condition to be jealous of.

You mentioned that this Zellen design has been around since the 1970s. Why isn’t anyone using it?

It solves a very specific problem: namely, eliminating disappointment in the control condition. And that problem is not really present with most drugs, but it is highly relevant to psychedelics, and I think it’s time to revisit it. The challenge is finding the funding for this kind of study.

It also has some practical difficulties: let’s say I randomize you into a group where you’re going to receive a microdose of a psychedelic, and you’re not interested in receiving a microdose. What do you do with that patient? This kind of design would be good for studying effectiveness, but not efficacy; effectiveness captures how an intervention might work in the real world, whereas efficacy looks at how it works under idealized circumstances, with the perfect randomization and blinding measures in place. But it would be really interesting to replicate previous studies — in particular, that escitalopram and psilocybin study — with the Zellen method just to see if the results would be different. In that study, there was no statistically significant difference between the depression scores of the escitalopram group and the psilocybin group; from that, you might come away with the impression that this is a negative result, but in many other studies, psilocybin easily beats escitalopram. A design that eliminates some of the biases associated with blinding would help sort out what’s really going on here.

We can’t just keep trying to do placebo controlled trials; we need to move on and find an alternative.

Why do we keep trying these placebo controlled trials?

We have a generation of scientists and medical doctors who were trained that placebo controlled studies are the gold standard, that this is what makes medicine more objective than what so-called social scientists do. But if you don’t measure the quality of blinding in these trials, that doesn’t help, and we’re seeing how psychedelics are forcing us to consider this under researched area; the magnitude of participants guessing which condition they’re in is high enough that we can’t ignore it.

The other case against placebo controlled trials is that it’s not a very patient-centric method. In those trials, researchers are essentially asking: is this treatment effective over and above knowing that you’re getting this drug? But most patients don’t care about that — what they want to know is: Will I get better?

This interview has been edited and condensed for clarity and length.