When a Placebo Is Not a Placebo: Problems and Solutions to the Gold Standard in Psychotherapy Research (2018) Blease et al.

[Cheshire]

Randomized placebo-controlled trials are recognized as the gold-standard of evidence-based medicine but when it comes to psychotherapy research all that glitters is not gold. Translation of this standard from medicine to clinical psychology is fraught with difficulties. While a wealth of robust evidence shows that psychotherapy is effective for a range of mental health conditions the use of placebo controls to assess the effectiveness of specific psychological interventions faces serious conceptual and methodological challenges (Gaab et al., 2018).

In this Opinion article we identify two under-appreciated placebo-related problems which substantially risk the validity of clinical trials in psychotherapy. The first is a common misconception about the nature of placebos; the second is the problem of double-blinding. We review current solutions and future prospects for the gold-standard in psychotherapy research.

https://www.frontiersin.org/articles/10.3389/fpsyg.2018.02317/full

 

[Barry]

Not had a chance to do more than skim this yet, but initial impression encouraging. Will look more later - hopefully still encouraging.

 

[wdb]

First, we advocate the need for “placebo-literacy” among investigators.... Second, we propose that psychotherapy trials be obliged to report comprehensive descriptions of all placebo control components including the rationale provided to patients.... Third, we argue that published articles should include a declaration of researchers' allegiance.

I'd add that to ensure unbiased review peer review should only be undertaken by peers who do not share the same allegiance as the authors.

 

[rvallee]

If only that was the only problem. It's in addition to allowing non-controlled subjective studies to be arbitrarily considered as somehow having the same significance as objective RCTs, but only when it fits prior biases. Even Cochrane and the Lancet clearly have no problem with non-controlled studies with subjective outcomes being elevated to the highest rated clinical evidence. The issue is that science is being ignored to fit a political agenda and the gatekeepers of science are colluding with this nonsense.

Very little of the psychosocial research is controlled, because control participants would either be horrified at how unprofessional it is or they would have to get a different protocol to avoid making it obvious that it is gaslighting therapy.

Above placebo control, PACE and other studies did everything they possibly could to maximize bias, rather than control for it. It is impossible to be more biased than promising patients they will get cured if they follow the treatment, along with glowing testimonies of its success during a trial meant to test the efficacy of the treatment. PACE is the highest maximum level of bias. The problem is way worse than just placebo illiteracy.

 

[Barry]

If only that was the only problem. It's in addition to allowing non-controlled subjective studies to be arbitrarily considered as somehow having the same significance as objective RCTs, but only when it fits prior biases. Even Cochrane and the Lancet clearly have no problem with non-controlled studies with subjective outcomes being elevated to the highest rated clinical evidence. The issue is that science is being ignored to fit a political agenda and the gatekeepers of science are colluding with this nonsense.

Very little of the psychosocial research is controlled, because control participants would either be horrified at how unprofessional it is or they would have to get a different protocol to avoid making it obvious that it is gaslighting therapy.

Above placebo control, PACE and other studies did everything they possibly could to maximize bias, rather than control for it. It is impossible to be more biased than promising patients they will get cured if they follow the treatment, along with glowing testimonies of its success during a trial meant to test the efficacy of the treatment. PACE is the highest maximum level of bias. The problem is way worse than just placebo illiteracy.

Valid points. But the article is focusing on one crucial aspect that is so misunderstood and abused. By focusing on one aspect, the message has a much better chance of getting through I think.

 

[Snow Leopard]

It's good, but doesn't mention the alternatives:

Objective outcomes
Comparing psychotherapies to a drug arm and a pharma placebo arm.

Yes, the set of biases are different between pharma and non-pharma, yet it elicits some placebo and reporting bias effects in the pharma placebo arm.

 

[Barry]

A placebo effect is essentially a psychological effect - the recipient's response, at some level, to knowing they have been administered a placebo. I suspect that in any given instance, although the psychological placebo effect might seem quite straightforward, it likely implicates quite a number of different psychological mechanisms, and will therefore amount to a multi-variable effect. The 'control' aspect of an RCT is intended to cope with controlling for such multi-variable effects, including but not confined to placebo effects. The crucial part of an RCT I would think, is for the intervention to be the opposite: a single variable effect ideally, probably minimal (number of) variable effect in practice. Otherwise confounding effects will be huge, and unravelling intervention effects from placebo effects challenging to say the least.

But surely a psychotherapy intervention can never be a minimal (number of) variable intervention? Certainly not something like CBT or GET anyway. The whole point of such an intervention is to be a broad spectrum intervention, intervening across a range of psychological mechanisms, and must in its way implicate multiple variables. Given that the placebo effect is itself a form of psychological 'intervention', it is hardly surprising that some - maybe many - of the intervention variables possibly overlap with the placebo variables. Intervention effects and placebo effects one and the same thing in some aspects. e.g. Participants wanting to show how hard they are trying to push past their symptoms. So the confounding effects will be massive, and unravelling intervention effects from placebo effects likely impossible.

So can the notion of Randomised Controlled Trials ever be legitimately applied to psychotherapy trials, given they cannot be controlled for placebo effects, and such placebo effects can so overlap with intervention effects? RT yes, but RCT no? So does this mean that psychotherapy trials need to simply accept this, and not presume otherwise? Does it mean that if the condition under consideration is genuinely psychological in nature, then might it be that RTs are sometimes acceptable even if not RCTs? That the unavoidable unblinding of psychotherapy interventions is sometimes OK if trialling for a genuinely psychological condition? I suspect that the answer is sometimes yes.

But for physiological conditions, only RCTs make sense I would think.

ETA: Added "(number of)" in a couple of places, because realised ambiguous.

 

[Lucibee]

@Barry I would agree with most of that, but psychological treatments should not be given a free pass on this one. RCTs could be better designed with such trials in mind.

I don't think any condition is "genuinely psychological in nature". Psychology itself is a label we have put on a set of outcomes, in this case, "behaviour" - but who's to say that some (most) behaviours are physiologically determined? Any condition that is mediated by stress has the potential to be "helped" by some kind of psychological intervention. In terms of trials, it's how you unpick what factor is doing what that is important in terms of correct blinding and placebo usage. Psychology clearly has some way still to go to adequately allow for this.

 

[Barry]

@Barry I would agree with most of that, but psychological treatments should not be given a free pass on this one. RCTs could be better designed with such trials in mind.

So what genuinely interests me, is how do you control for placebo effects in an unavoidably open label trial, if the psychological condition in question can only be measured on subjective outcomes? And especially if the intervention itself produces effects that can so overlap with placebo effects? Given intervention and placebo are both psychological, I don't see how you get round that confusion. Genuine question.

 

[Lucibee]

I guess it's all in the comparison. Psych trialists want the comparison to be as big as possible to maximise the effect they see. But from a RCT perpective, you want the comparison to be as small as possible to isolate the true interventive effect. So maybe there needs to be a negotiation process whereby the objective trial scientists try to come up with a fake treatment that is as close to the intervention as possible but that doesn't deliver the hypothesised intervention that is agreeable to the psychologists as a fair test. Even that process might be quite interesting.

That's why wait list is not a fair trial. But I guess the problem is that an active placebo intervention might be hard to control for all the other subtle effects it might introduce.

Ultimately, if you can't blind, you can't control for those placebo effects that are manifest by knowing that you are receiving the treatment. But they might be less important if knowing you are receiving the right treatment is part of the treatment.

 

[Barry]

I guess it's all in the comparison. Psych trialists want the comparison to be as big as possible to maximise the effect they see. But from a RCT perpective, you want the comparison to be as small as possible to isolate the true interventive effect. So maybe there needs to be a negotiation process whereby the objective trial scientists try to come up with a fake treatment that is as close to the intervention as possible but that doesn't deliver the hypothesised intervention that is agreeable to the psychologists as a fair test. Even that process might be quite interesting.

That's why wait list is not a fair trial. But I guess the problem is that an active placebo intervention might be hard to control for all the other subtle effects it might introduce.

Ultimately, if you can't blind, you can't control for those placebo effects that are manifest by knowing that you are receiving the treatment. But they might be less important if knowing you are receiving the right treatment is part of the treatment.

Can't help thinking that the psychs have gone off half cocked for years presuming they are doing RCTs and making big claims to that effect, when in fact that have not. Feels like they, and maybe trialists from other disciplines, need to get round the table and work from the ground up to work out what kind of trial methodologies would make more sense for psychotherapy interventions. Might come up with quite something different to what people have thought of so far.

 

[Snow Leopard]

I'd add that to ensure unbiased review peer review should only be undertaken by peers who do not share the same allegiance as the authors.

That is quite a bold ask, "peer review" generally means people with the same interests...

 

[Snowdrop]

That is quite a bold ask, "peer review" generally means people with the same interests...

You can be interested in the same illness or subject matter but not have the same perspective regarding it. I think that might be what is meant. So you would still be considered a peer.

 

[obeat]

So what genuinely interests me, is how do you control for placebo effects in an unavoidably open label trial, if the psychological condition in question can only be measured on subjective outcomes? And especially if the intervention itself produces effects that can so overlap with placebo effects? Given intervention and placebo are both psychological, I don't see how you get round that confusion. Genuine question.

One way round this might be to have the LTFU and initial results published together. A small feasibility study could be published prior to the randomised trial.

 

[wdb]

That is quite a bold ask, "peer review" generally means people with the same interests...

Yes it's obviously unlikely to happen any time soon but while any quack can get their research reviewed by like-minded quacks nothing much is going to improve.

Perhaps as a first step they could just exclude any reviewers who have previously worked in the same department or co-authored a publication with any of the authors.

 

[Flying Dutchman]

The article explains:
"In every randomized controlled trial (RCT) the placebo should ideally be “bespoke” – tailored to mimic an intervention under investigation without consisting of any it's hypothesized characteristic constituents. While fulfilling this goal is enormously challenging, in drug trials a placebo should preferably mimic the particular taste, appearance, and method of administration of the specific intervention without comprising the treatment's characteristic pharmacological ingredient(s). Placebos should also ideally be administered double-blind: practitioners/researchers and patients should not be able to distinguish the placebo from the treatment."

One problem in psychological RCTs is that if you create a "placebo" that successfully mimics the "treatment", then how can you be sure that the placebo is inert?

You can be confident a sugar pills won't work, but not so with a credible sham psychological treatment.

Eg If your hypothesis is designed to test the efficacy of CBT directed at false illness beliefs, a suitable mimic-placebo might be CBT targeted at alternative false health-related beliefs... But who's to say your "placebo" isn't as likely to be effective as your treatment?

If both achieve 20% improvement in your primary outcome measure, then it's possible:
- Both are efficatious
Or
- Neither are efficacious (ie the placebo generates a placebo response, and the treatment does likewise, just as frequently)

 

[Snow Leopard]

One problem in psychological RCTs is that if you create a "placebo" that successfully mimics the "treatment", then how can you be sure that the placebo is inert?

You can't. That's why non-pharmacological studies cannot be considered Randomised Control Trials, because there is no true control group. They should instead be refered to Randomised Comparison Group Trials.