Have you ever heard of a Perkins Tractor? If you lived in London in the 18th Century, you could buy a pair of Perkins Tractors (shown above) for the steep price of five guineas, and wave them over an aching part of your body for about twenty minutes as a way of relieving your pain. They worked, sort of. People who bought and used Perkins Tractors
I wonder whether the placebo effect might have gotten larger in studies, because we got better at making good convincing placebos?
I also read somewhere that placebos work better when they are more expensive. Medicine has famously gotten more expensive in the US (and more than in other countries). So perhaps that causes a bigger placebo effect?
Just speculation. I'd be very interested in more placebo studies.
I’ll read the paper, but my first thought is that his has more to do with better trial design (e.g. the push to preregister and avoid p-hacking) and stricter scrutiny of drugs by US agencies like the FDA.
Not necessarily. P-hacking and biased trials don't really care how the effect size is generated, and artificially reducing the control group would work. In fact, if a trial is biased by the human designers, I wouldn't be surprised if the control group was lower, because that is what the trial designers would expect. (It's very likely that there is another explanation to the effect reported in this paper. But I do think that trial designs have improved in recent years, especially with the advent of preregistration.)
And no, the FDA is not rubbish. US regulators are generally stricter about approving drugs and devices. For example, the EU's CE standard is much lower bar than FDA approval.
I see your point re control group being reduced - that's possible. But as it's the US changing and everyone else staying as they were, and the effect is big, it would need the US to be doing their stats pretty drastically differently from the rest.
RE the FDA 'being rubbish'... I simply mean that, overall, US regulators/government have a reputation for not showing up well versus other western countries and so was questioning assumption that the FDA would be doing a lot more than others in terms of better statistical techniques.
To quote the Cochrane review of "Placebo interventions for all clinical conditions":
"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."
Hence the measured "placebo effect" is mostly response bias. It is a change in symptom reporting/questionnaire answering behaviour, rather than a change in the underlying disease.
The only "real" placebo effect is a conditioning of the endorphin system which is an acute response that allows for a temporary reduction in pain, so that mammals can escape danger, or continue to fend for themselves despite that pain (and a similar pattern for acute nausea).
Notably, "placebo effects" tend to disappear over the long term - suggesting those response biases revert back to the mean over time.
This means that placebos are useless for anything serious.
Is anyone arguing evidence supports the opposite viewpoint: that placebos are effective over long periods for a variety of ailments? Wondering if this is a consensus view or if there is (serious) debate about this.
As I understood it, the placebo effect is only a real thing for pain. Your Mum really can 'kiss it better'. This makes a certain amount of evolutionary sense. If part of what pain is telling you is to 'get help' then if the help arrives, there's less need to be motivated to get more help.
Note that this is consistent with your warm/competent graph. If you're getting help from someone warm and competent then you've already done well. But if the warm and competent person is trying to hurt you, that doesn't matter, you still need help.
In all other cases, it's either reversion to the mean, or it's an effect that's seen when patients fill out self-report forms, and don't want to disappoint.
Note that that's consistent with the 'placebos getting stronger' thing. The trials are done better, less p-hacking, pre-registration, so the statistical artifacts go away and the drugs look weaker.
The childhood epilepsy thing I can't explain, but I wonder if the study involved some sort of self-report questionnaire instead of more solid methods of measuring the effect?
The same is at work with anti-depressants. Some studies show that placeboes are very nearly or even as effective as the actual medication, especially when the placebo produces mild side-effects and those side effects are listed among the side-effects patients might expect from medication. The side effects convince the patient that they received the real drug.
I wonder whether the placebo effect might have gotten larger in studies, because we got better at making good convincing placebos?
I also read somewhere that placebos work better when they are more expensive. Medicine has famously gotten more expensive in the US (and more than in other countries). So perhaps that causes a bigger placebo effect?
Just speculation. I'd be very interested in more placebo studies.
Another speculation: "weak sickness" is increasing, whilst placebo-resistant sickness remains the same.
That might be something we can test with existing data.
I’ll read the paper, but my first thought is that his has more to do with better trial design (e.g. the push to preregister and avoid p-hacking) and stricter scrutiny of drugs by US agencies like the FDA.
That might be the case if the effect of the drugs was higher in other countries but the paper suggests it's the placebo that varies
Also, is it not a common place that US regulators are a bit rubbish compared to other western countries?
Not necessarily. P-hacking and biased trials don't really care how the effect size is generated, and artificially reducing the control group would work. In fact, if a trial is biased by the human designers, I wouldn't be surprised if the control group was lower, because that is what the trial designers would expect. (It's very likely that there is another explanation to the effect reported in this paper. But I do think that trial designs have improved in recent years, especially with the advent of preregistration.)
And no, the FDA is not rubbish. US regulators are generally stricter about approving drugs and devices. For example, the EU's CE standard is much lower bar than FDA approval.
I see your point re control group being reduced - that's possible. But as it's the US changing and everyone else staying as they were, and the effect is big, it would need the US to be doing their stats pretty drastically differently from the rest.
RE the FDA 'being rubbish'... I simply mean that, overall, US regulators/government have a reputation for not showing up well versus other western countries and so was questioning assumption that the FDA would be doing a lot more than others in terms of better statistical techniques.
To quote the Cochrane review of "Placebo interventions for all clinical conditions":
"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."
Hence the measured "placebo effect" is mostly response bias. It is a change in symptom reporting/questionnaire answering behaviour, rather than a change in the underlying disease.
The only "real" placebo effect is a conditioning of the endorphin system which is an acute response that allows for a temporary reduction in pain, so that mammals can escape danger, or continue to fend for themselves despite that pain (and a similar pattern for acute nausea).
Notably, "placebo effects" tend to disappear over the long term - suggesting those response biases revert back to the mean over time.
This means that placebos are useless for anything serious.
Is anyone arguing evidence supports the opposite viewpoint: that placebos are effective over long periods for a variety of ailments? Wondering if this is a consensus view or if there is (serious) debate about this.
As I understood it, the placebo effect is only a real thing for pain. Your Mum really can 'kiss it better'. This makes a certain amount of evolutionary sense. If part of what pain is telling you is to 'get help' then if the help arrives, there's less need to be motivated to get more help.
Note that this is consistent with your warm/competent graph. If you're getting help from someone warm and competent then you've already done well. But if the warm and competent person is trying to hurt you, that doesn't matter, you still need help.
In all other cases, it's either reversion to the mean, or it's an effect that's seen when patients fill out self-report forms, and don't want to disappoint.
Note that that's consistent with the 'placebos getting stronger' thing. The trials are done better, less p-hacking, pre-registration, so the statistical artifacts go away and the drugs look weaker.
The childhood epilepsy thing I can't explain, but I wonder if the study involved some sort of self-report questionnaire instead of more solid methods of measuring the effect?
You might find this interesting, particularly the part at the end on Super Placebos https://rogersbacon.substack.com/p/the-tale-of-the-shaman-science-magic?s=w
The same is at work with anti-depressants. Some studies show that placeboes are very nearly or even as effective as the actual medication, especially when the placebo produces mild side-effects and those side effects are listed among the side-effects patients might expect from medication. The side effects convince the patient that they received the real drug.
I suspect that this reflects the yanks being bit more mental than everyone else, but the rest of the rest of the world will start to catch up soon.