Placebo effect: a psychosomatic component, or only an aggregate of other biases?

We have previously discussed interventions that manipulate how patients report symptoms as a separate form of bias, rather than as part of the placebo effect.

Critics of the placebo effect have pointed out that according to the available evidence, "placebo effects are weak: regression to the mean is the main reason ineffective treatments appear to work". This seems to be based on the Cochrane review of placebo interventions by Hróbjartsson and Gøtzsche (2010), which concludes (bolding and underlining mine):

We did not find that placebo interventions have important clinical effects in general. However, in certain settings placebo interventions can influence patient‐reported outcomes, especially pain and nausea, though it is difficult to distinguish patient‐reported effects of placebo from biased reporting. The effect on pain varied, even among trials with low risk of bias, from negligible to clinically important. Variations in the effect of placebo were partly explained by variations in how trials were conducted and how patients were informed.
​

Recently, two reviews of the placebo effect were published in NEJM (Colloca and Barsky, 2020) and JAMA (Jones et al, 2021, in treatment-resistant depression, S4ME thread). These found, respectively, that the placebo (and nocebo) effect were "powerful, pervasive, and common in clinical practice" and "large and consistent across treatment modalities".

However, these reviews conducted within-arm comparisons, not between-arm like the Cochrane review, which subjects them to confounding by regression to the mean. In their rebuttal of Colloca and Barsky, Dahly and Zad's conclusion falls in line with that of the Cochrane review (bolding mine):

Instead, the evidence presented seems to support the plausibility of therapeutically useful placebo effects based on indirect evidence (an argument placebo researchers have been making for decades now), and some limited empirical evidence of them (systematic effects beyond regression to the mean and measurement error) in a very limited number of areas.

(...) to suggest that placebos have potent, widely-applicable therapeutic effects is clearly a substantial overreach, and in our opinion the suggestion that we should explore subgroups of patients who are most likely to benefit from such therapeutic effects will result in much more noise than signal.
​

Jones and Maher's rebuttal of Jones et al highlight the same problems.

Overall, this begs the question: is there a standalone psychosomatic component to the placebo and nocebo effects, or can they be wholly explained by other sources of bias -- first and foremost regression to the mean, and response bias on subjective outcomes --?

 

Regarding the nocebo effect, Dahly and Zad also point out that adverse events are often mistakenly attributed to a treatment, in which case they may be incorrectly understood as psychogenic (bolding mine):

Importantly, two of the three reviews cited to support this argument report on adverse events, which is a technical term that specifically refers to any and all adverse events recorded during the course of a trial without consideration of any causal link to a treatment, placebo or otherwise. For example, a patient enrolled in a drug trial who fell down some stairs resulting in injury would likely be recorded as having experienced an adverse event. This is in contrast to the term side-effects, which would surely suggest to many or most readers that the effect was the result of a treatment. Making this leap from adverse events to side-effects (or more formally, adverse effects or adverse reactions) requires expert adjudication, so it is concerning to see the authors shortcut this important step.​

 

My rather firm opinion after years seeing it validated is that it's an imprecision error that comes from not measuring something real. So it simply doesn't exist at all, it's an illusion physicians want to see because it grants them magical healing powers from simply being there.

If you try the same with something that can be measured it disappears entirely. When you can account for multiple factors interacting it also falls within any imprecisions. You can get a -cebo effect if you ask people to rate temperatures by a subjective low-mid-high arbitrary scales, completely disappears if you actually measure using a standard unit.

Placebo and nocebo are both pure illusions that exist in all pre-science fields of expertise, right up to the day you can measure what's really there and confirm that there is no magical demon doing something magical, it's just something you can't see or account for. It's basically just bias, especially of the confirming kind.

 

In a tweet, as a response to whether PACE accounted for a placebo effect. Can't search for it but it's probably still there. IIRC he said "one of the most powerful interventions we have". Which may be technically true, in that these people truly have nothing more powerful than literally the control for nothing and have simply been taking credit for natural courses on which they had no effect.

 

There is no "healing" effect associated with placebo.

There is however a conditioned effect that allows temporary reduction in nausea or pain (mediated by endogenous opioids) which is an evolutionarily derived system that allows animals to escape danger despite experiencing pain or nausea.

This effect on nausea and pain is what the Cochrane review showed.

Studying questionnaire answers within single arms in clinical trials is a joke as there is no means at all for controlling for response bias.

 

I also wonder if there is an evolutionary bias towards responding positively to being helped.

Let’s face it if you are unwell or hurt, and require help to survive, you are much more likely to continue to receive said help if you actively encourage the helper with appreciation.

So even if the help doesn’t make a blind bit of difference, your evolutionary instinct is going to encourage you to “say” it helps.

In the long run, if the helper stays around and continues to try and help you, then you are more likely to survive - even if the real thing “helping” you is not the “magic remedy” they gave you but their protection or something entirely different.

We are basically wired to respond kindly to being helped.

 

You might also get more help if you hide just how bad you are.

So another evolved response to being helped might be to conceal how bad things are, because a helper is more likely to continue helping you if they judge you have a reasonable chance of survival.

What you indicate to others, is therefore not always a true reflection of reality. The need to seem optimistic when receiving any sort of help, is surely an evolutionary bias?

 

Friends with MS were part of a series of research studies. even if it did not help them they were reluctant to dismiss something that could help someone else so they tended to answer "a little bit".

It can be difficult to decide if pain has gone down a little bit or if you feel a bit brighter with a depression treatment and I think this is where a lot of the placebo effect comes from.

 

How frustrating to see this is still going on.

I think it goes back to the very first randomized clinical trials after the second world war. Some people were surprised by how much the control group improved in those studies as they didn't receive active treatment, so they said it was due to a placebo effect. Henry Beecher popularized this idea with his paper "The powerful placebo." Only later it was pointed out that other factors such as regression to the mean, Hawthorne effect, and other biases explain reported improvements in the control group. See: How much of the placebo 'effect' is really statistical regression? - PubMed (nih.gov)

In order to measure a placebo effect, one would need to compare a group that receives a placebo with a control group that didn't. The review by Hrbojartsson and Gotzsche looked for such comparisons. It showed that only a small placebo effect on subjective outcomes such as pain remained. This could be due to response bias. After all: the placebo group received something while the control group did not, so the former is more likely to say they're doing a bit better even if they're not.

I think that as long as we don't have any better evidence that is able to differentiate a "placebo effect" from response bias, there is little to suggest that there is an effect of placebo's that has clinical relevance.

Also, in my view, skeptics do not need to provide evidence that the thing some people believe in is fully due to methodological weaknesses. It's the believers in a true placebo effect that need to provide reliable evidence, not the other way around.

 

Just had a look at these reviews and it is really astonishing that these were published in prestigious journals.

The one in NEJM seems like an opinion piece. It doesn't provide data or systematic review. It cites the literature selectively and sometimes incorrectly. I note that the senior author of this paper is Arthur Barsky which some may know from his horrible but influential paper on functional somatic syndromes. Functional somatic syndromes - PubMed (nih.gov)

The one in JAMA simply looked at improvement over time in the control group of trials for depression and called this the placebo response. With this method, you can demonstrate that everything is effective. Sad to see how many healthcare professionals on Twitter do not understand this basic point. It's clinical trials 101 that you need a control group to measure a clinical effect. This is what's on the first pages of the textbooks...

 

I do wonder how this keeps happening.

Do so many researchers not understand basic clinical trial methodology, like the need for a control group and the risk of response bias? I find this hard to believe because this is as basic as it gets.

So it must be that the idea of a "powerful placebo" is very attractive to clinical researchers.

 

This effect seems to be distinct from the adrenaline rush in a fight-or-flight situation, which tends to block pain altogether. The first example that comes to mind is someone getting stabbed, or shot at by a firearm, only to notice it once they are out of danger.

While this effect is only meant to be a short-lived distraction in an acute situation, the aim of CBT, "brain retraining" and similar forms of psychotherapy for chronic pain appears to be -- at least in part -- to potentiate the longevity and intensity of this effect. The evidence, however, shows that they fail at attaining this goal.

It is interesting to think about our response to help in an unfavorable situation from the lens of evolution. Are you aware of any research on this topic, @Keela Too?

 

No, not aware of anything. Just my interpretation of how certain behaviours might give a selective advantage over evolutionary time.

I imagine these sorts of behaviours could be either genetically predisposed, and/or passed on as cultural memes.

So just a concept I’ve been playing with to try and help explain our tendency to tell others that “cures” they give us “help”. Maybe there is also logic there, in that even if we know this thing didn’t help, our helper might try something else when suitably encouraged, and perhaps that next thing might be the key?

 

Maybe it's a way to avoid having to admit the treatments you can offer are close to ineffective. If your treatment is barely better than placebo, the idea that placebos are some mysterious and powerful healing effect would sound appealing. Then your treatment is a bit better than the powerful placebo effect. If placebos are powerless, then your treatment is just a bit better than nothing and even that might be due to publication bias and incomplete blinding...

 

At this point I'd say sunk cost and avoiding embarrassment are the main reasons to keep this charade, probably accounts for 90% of it. The magical powers of the mind have been a truism in medicine for well over a century, it's too much to accept that it's been nothing but a bunch of nonsense for rubes all along. Especially after millions of lives have been sacrificed to it, with a massive record of its victims objecting to it, the responsibility of medicine here is total, this was 100% the product of conscious choices, nothing was forced or inevitable.

Like a war that has gone on for too long, but those in charge can't bring themselves to pull the plug so they keep it going, years, many generations, if they have to. As long as the cost of perpetuating a lie is less than its acknowledgement, it almost never happens that people will simply do the right thing even at personal cost, this is the just world fallacy, people rarely act against their interest, or go against the tribe. Medicine isn't special enough to be exempt from this, it's human nature, probably honed in from thousands of generations living in small tribes as hunter-gatherers where reputation and status was all there was.

 

So the placebo effect to control for is any placebo effect attributable to the treatment. But if the treatment itself includes strong elements of placebo, such as happened in PACE (telling participants how well proven their treatments were, lots of morale-boosting material, etc), then how do you control for that? How do you null out the placebo elements of a treatment when that is mostly what the treatment is anyway!? How can you have a sham treatment in that case, having only the placebo elements of the treatment, when the distinction between placebo and active is so woolly? (Note: I realise the notion of a "sham treatment" has a double relevance here).

It's relatively easy to control for any placebo effects attributable to being part of a trial itself (changed routine, change of scenery, more/different human interactions, etc), as that is an effect applicable to all groups.

Feels like trying to run a trial where active and control groups all take the same looking tablet, as per usual, but the active group also get a full advertising pack with theirs, complete with high pressure sales person, promoting how great participants will feel after taking the tablets. But the control group is not able to have the same placebo-inducing components.

 

I've been speculating about this evolutionary perspective, too, with a slightly different slant but ultimately similar.

No science to hand but I've definitely seen some papers discussing the matter in animals, though to me it looks like it applies just as much to humans.

Basically, showing signs of weakness is dangerous in various ways. It may lead to being targeted by predators or to being abandoned or even attacked and killed by one's own group, which from a species survival point of view is 'rational' behaviour; weak individuals may be infectious or using up scarce resources. Chances at procreation also drop drastically.

So it's a major advantage to look and act strong and healthy. Once illness or weakness is obvious, many animals retreat and hide as part of their sickness behaviour response. Again, on a species level this reduces the risk an individual may pose to the group (infection, bad genes) while at the individual level it may increase chances of survival if it allows the individual to escape attack.

In some social species, including humans, altruistic behaviour modifies the situation somewhat; on a species level helping individuals from one's group survive is an advantage - up to a point. Sure, some individuals risk their lives for others, but perceived 'lost causes' are just as likely abandoned or killed. Which way it goes depends on circumstances at the time which in the case of humans go beyond food availability and include what value a specific culture places on helping others.

So I think the drive to pretend we're well and strong - or at least have the potential to become well and strong again - is a deeply conserved biological drive, more nature than nurture. It's not necessarily a conscious thing but at some level we 'know' that the willingness of others to help us is fickle and contingent on us 'rewarding' their efforts by getting better.

This subconscious fear could help explain two response biases that often get mashed up with the placebo effect; the need to please the therapist as well as the need to believe that you're getting better could well spring from this self-protection drive.

 

Not just in medicine but right through society, and that's a big part of the problem. It suits far too many people far too well to believe this and that's why nobody wants to listen to those spoilsports who say hang on a minute, something doesn't quite add up here...