E138 What Does Science Reveal About Therapy? (w/ Prof. Pim Cuijpers)

Pim Cuijpers. Ph.D. is professor emeritus of Clinical Psychology at the Department of Clinical, Neuro and Developmental Psychology, Amsterdam Public Health research institute, Vrije Universiteit Amsterdam, The Netherlands. He is also director of the WHO Collaborating Centre for Research and Dissemination of Psychological Interventions in Amsterdam.

Professor Cuijpers is specialised in conducting randomised controlled trials and meta-analyses on prevention and psychological treatments of common mental disorders.

Today, we focus on what is known about the psychological treatment of depression. We start off a little technical - so that you hopefully can come away with an initial understanding of what a meta-analysis is and how they are useful. Prof Cuijpers then tells us about what is known about how effective psychological therapies are in depression, where we’re at with personalising and what his vision for the future is.

Get in touch or read more from with Prof Cuijpers here: https://www.pimcuijpers.com/blog/

Interviewed by Dr. Anya Borissova - Give feedback here - thinkingmindpodcast@gmail.com Follow us here: Twitter @thinkingmindpod Instagram @thinkingmindpodcast Tiktok - @thinking.mind.podcast 

Check out The Thinking Mind Blog on Substack: https://substack.com/home/post/p-170117699

Give feedback here - thinkingmindpodcast@gmail.com Follow us here: Twitter @thinkingmindpod Instagram @thinkingmindpodcast

Give feedback here - thinkingmindpodcast@gmail.com Follow us here: Twitter @thinkingmindpod Instagram @thinkingmindpodcast

Transcript

 80% of the people with mental health problems live in low and middle income countries, and that's, there are 1 billion people in the world with a mental health program. There are more people with depression in China than the whole population of Spain. There are more adolescents with depression in India than the whole population of the Netherlands.

Uh, but you cannot imagine that you train enough therapists to, to deliver therapy to all these people. That's, that's impossible. So digital solutions are really, uh, one of the possibilities that if you have an intervention that is scalable. Has a very small effect, but we can reach, let's say, 10% of the 50 million people of with depression in China than the impact of that, of that intervention is still [00:01:00] much, much larger than anything you do in the UK for the treatment of people.

Small effects can still have a huge impact when you, uh, can disseminate it on a large scale.

Hi everyone. Welcome back to the podcast this week. My name is Anya and I'm a psychiatry registrar working in South London. I'm very excited to bring you a conversation with Professor Pim Kiper today. Professor Kiper is Professor Emeritus of Clinical Psychology at the Department of Clinical Neuro and Developmental Psychology at the Amsterdam Public Health Research Institute.

He's also the director of the WHO Collaborating Center for Research and Dissemination of Psychological Interventions in Amsterdam. Now, professor Kiper is an absolute authority on the conduct of research into psychological therapies. With today our conversation focusing on his work [00:02:00] in depression. A lot of his recent work has involved conducting meta-analyses, which are amazing bits of research that pull together all of the other existing studies that there are in a field in order to draw some quite firm and strong conclusions about what the state of the field is.

So our conversation today starts off a little bit technical to explain what meta-analyses are, what they involve, how his team conduct them, and from there we go into thinking together about what meta-analysis can tell us about psychological therapies in depression. What is known, what isn't known. From his position of expertise, what should be the focus of future research and perhaps where should energy actually not be focused?

I really hope that you enjoy this conversation, and as always, if you found it useful, I'd love it if you could share it with someone else who you [00:03:00] think would find it interesting. If you could give us a rating or a comment as this really helps other people to find us and now sit back, relax, enjoy. You have found The Thinking Mind Podcast here.

You can access conversations all about psychiatry, psychology, and self-development, which should be accessible to all. Hello, professor Kiers. Thank you so much for joining us on the podcast today. You. Have done amazing research when it comes to psychological therapies, understanding what works, some of how it works, what we can do to improve things.

Um, and I hope that today we can have a conversation about what the treatment of depression looks like and how we can improve it as we think ahead to the future. But to begin with, to start to delve into your ideas, uh, your, your research focus is really on using sort of big data sets, pulling data together that, um, from trials and studies that have been done in the past to really get an overview of, of what's happening in the field.[00:04:00] 

Why, why is this direction that you chose to work? Why, why is this the way that you decided to address the question of, of psychological therapies and how they work? Well, it's not the only, uh, research I've done. Uh, I'm retired now, and now I only work in getting all these data together. But before I retired, I have, I was involved in.

I think about 80 randomized trials myself, and I've done a lot of research in global mental health, in, uh, mental health and college students in epidemiological, uh, research. So it's not only that, that I've done this, I've done a lot of more research, but I find this, I mean, we do, we do what we call meta-analysis, and that's, uh, uh, the statistical integration of the results of individual studies examining the effects of therapies.

The problem with these [00:05:00] studies is that they, they don't have all the same outcomes. If they would have all the same outcomes, then we wouldn't have to integrate them. But they, they vary. And so if you have a lot of studies in a specific fields, you cannot just look at these studies and say, okay, this works and these are the effects, et cetera.

You have to integrate them statistically. And I've done that a lot. And, um, uh, there are now so many trials on, I mean, for psychological treatments of depression, there are more than a thousand randomized trials. And so it's impossible to look at these trials and say, okay, uh, it works. And, uh, these are the effects.

And it works for these people and not for these people. And so you have to do that meta analytic work. And now it's so, uh, there are, so there are really very randomized trials, so more than a [00:06:00] thousand. But there are now also all kinds of meta-analysis. So the number of studies integrating all these studies is also becoming so big that we need to step up.

So to say we have to, uh, what we did from trials to meta-analysis, we now have to do that with meta-analysis to the next step. And that's, that's what I, uh, what I, what I've been doing. And I like that because, um, uh, it gives an overview of the whole field. So what we do is we systematically collect all these data.

Every four months we extract the data so that we have a complete overview of what can be known from all these trials. And so I think it's the best, uh, resource of knowledge on, uh, on [00:07:00] psychological treatments. What, what in your view, makes a good me meta-analysis? Or how, how do we do them better than what, than some of the things that exist?

Well, basically what you do is you, you, you pull the effects you find for a study and you, you calculate a mean, that's basically what you do in a meta-analysis. But it's a bit more complicated because, uh, some studies are very large and some studies are very small. And so the one, the big ones should have a higher weight when you calculate the, the pool results, so to say.

Another problem is that they use different outcome measures. So if you, there are all kinds of, I think we have more than 30. Outcome measures used in psychological treatments for depression. And so how can you pull the, the, the, how can you [00:08:00] calculate the mean if the outcome measures are not the same? And so what we do is we, we calculate, uh, and then it becomes a bit technical.

We calculate the standardized mean difference, and that indicates the difference between the treatment and the control group in terms of standard deviations. And these are the two basic principles of what you do in a meta-analysis that you, you standardize the outcomes and you pull the outcomes according to the size of the study.

And that we, I mean, that's the basics, but uh, over time the methods have become so. Sophisticated that, uh, that are all kinds of things you have to analyze in a study. For example, what do you do with negative studies that are not published and how do you, in, how do you [00:09:00] estimate that and what do you do with all these studies that are so different that you, we call that heterogeneity.

How do you handle that? All these differences in a meta-analysis. And so that kind of thing has been, has been, um, yeah, there are so many methodological and statistical developments that's too complicated to explain, but the basics are still the same. So pooling the outcomes based on the size of the study and using a standardized outcome.

So in, in its most simple sense, you, you have all these different studies that have maybe done things in different ways. And you make it so that it doesn't matter, that one has maybe measured something about how someone's sadness changes. One has maybe focused more on questions about energy and anxiety.

Uh, you know, another study has maybe included something completely different. [00:10:00] But because each, in each study you can compare the difference within the study, that's what helps you make, no, I mean, depression is measured with all kinds of different measures and the exact questions are different, but it's not about, uh, anxiety or quality of life or social support.

These are really different measures. So all the main outcome measures, they measure depression, but they use different questions to, to measure it. And, and even though those questions are different. By doing the standardization, it means that actually you can get an idea of what's happening with each treatment at each time.

Okay, so let's talk numbers. When we look at psychological therapies and depression, what is it that we know about the how well they work, how well they get people better? Well, they are, they are definitely better than doing nothing. That's, uh, that's, that's very clear. [00:11:00] And so when we examine the effects of psychological treatments, we have to compare them with a control condition where people get nothing.

Uh, that is a bit problematic, uh, because what is doing nothing, um, you can, one way is to put people on a waiting list. Then, then, then they will get a treatment within two months. Or you can compare it with the usual care that people get. Or you can, for example, give them a pill placebo like you do in drug trials.

And all these different control conditions have an impact on the effects that you find. And so one problem with, with talking about the effects of psychological treatments is that they're always reported in terms of the standardized mean difference. And I just already said that a little. That's a [00:12:00] bit of a com complicated, uh, uh, concept.

Uh, it is, it is the best way to examine, uh, the, the outcomes because it's, it's, it show, it, it assumes that depression is not something you have or don't have. It's a continuous thing. You can have it a little or more, or really a lot. Most people don't have it, but you can have it a little or severe or very severe.

And so it's a continuum. It's not something you have or don't have. And so if you wanna measure depression, you measure it on that continuum and that continuum you, what you do in trials is you compare the continuum of the treatment group to the continuum of the control group. Then you look at the difference between these two continuums [00:13:00] and that's, you, you, you report that in terms of the standard deviation.

So when you have, uh, an effect size or a standardized mean difference of 0.5, that means that the treatments and the control group differ 0.5 standard deviation after treatment. But of course, I mean, that's so complicated to to explain what it means. So what we did in, in, and we do that more and more, uh, is that we also look at just simply at absolute outcomes.

So how many people get better and how many people are completely recovered after treatment? How, how many people get worse because that also happens. And so if you look at these numbers. Statistically, you have to be very careful with that. So that's why a lot of pe, a lot of researchers don't do it because it's statistically too complicated to do that.[00:14:00] 

But it's so important to have these absolute outcomes. So well, basically, when you get therapy, the most used absolute outcome is a 50% symptom reduction. So when you, you take your base baseline depression score, and when somebody responds that baseline depression score is 50% after treatments, then it was before treatment.

So that's treatment response. And when you look at treatment response, that's about 42% of people who respond after psychotherapy for depression and in the control groups. So in care as usual or wait list control groups, that's about 16 or 17%. And so the nets benefits of getting treatments is roughly 25%.[00:15:00] 

If you look at placebo control groups, so in drug trials where people get psychotherapy or a drug or a pill, placebo, the response rate, so the 50% symptom reduction is about, I think 31%, or it can also be 32%, but something like that. Yeah. So these are the outcomes. And then you have remission when people are completely recovered.

So there you cannot differentiate them from the general population. That's about 25% after treatment, and I think it's about 12%. I don't, don't have the exact percentages. So, uh, it could. Be different for one or 2%, but in, if in the control groups, that's about 12% and then you have the people who get worse, that's about 5% in the, uh, in the, in the treatment groups.

[00:16:00] But fortunately, that proportion of people getting worse is much lower in the psychotherapy group than in the control groups where it is about 10 or 11%. So there are people who get worse during therapy, but these, that number is smaller than the number of people who don't get treatment. There are more people who get worse.

Good. That's, that's good news. So about a quarter of people or, well, so 40% of people can expect to feel better if they've had a psychological therapy treatment, but a proportion of them would have gotten better anyway because of. Just, that's how, that's how things go. And so leaving us with, with about a, a quarter, 25% of people overall, overall getting better from a treatment.

I mean, what, what, what's, what's your view on, on those numbers? Well, I think it's, um, [00:17:00] it's modest, but then if, if you compare it with treatments in, uh, in others, in other sectors of biomedicine, then it's not, not so, not, and it's normal. So it's not, it's, it's, it's a good outcome. And you, and you have to remember that these treatments are usually pretty brief.

So most treatments are, let's say 10 to 12 sessions, and that can be a little less, can be six, or it can be 18. Some are very long. Uh, but overall, these, these, these, these treatments take, take. One to three months. And so it's very brief and it means that if you, that's the first treatment you get. And so if you get that, that doesn't mean that after that it's done.

I mean, if you, if you have a good therapist and you, you get the first treatment and it doesn't work the first time, which is [00:18:00] the case for many people, then you can get another treatment or you can continue with the treatment you have, or you can add antidepressants to it. And so in the end, most people get better.

Uh, but, uh, we, it depends. You, you can be lucky and you get better after the first treatment or you can have that luck and you need five treatments. And unfortunately, we cannot predict that yet. We do not know. How you will be on that scale of number of treatments you need to get better. And that's one thing I know from clinical practice is that clinicians are often not very clear about this.

They, they, they think themselves that, okay, well we have an evidence-based treatment, so, uh, everybody should get better because it is evidence-based. But that's not the case. Many people, most people do according to our [00:19:00] st estimates, do not get better after the first treatment. Uh, so, uh, it's much better to know as a patient from the beginning that you can be lucky and you get better with the first treatments, relatively lucky, and you get better with the second treatments, or you can have bad luck and you need, you need multiple treatments to get better.

My understanding is that we. What we know so far is that there doesn't seem to be a huge difference between what kind of psychological therapy you get when we sort of look at these big data sets. But do we know anything about, on an individual level, you know, is is there a possibility that an individual person might respond better to something rather than, than, than other, or that certain symptoms might respond better than others to a certain kind of therapy?

Do we know anything about that, that level of personalization? Well, briefly, no, [00:20:00] we don't. And, uh, the, there is a lot of research going on, uh, to, to try to disentangle that. Uh, but unfortunately we do not know that. And, uh, it is true that most psychotherapies, uh, are affected for depression, but we have ac we can examine that only.

In the therapies that have been examined sufficiently. And so that means that we, we, we know about, uh, many therapies, uh, that they work, but that doesn't mean that anything works because we just don't know that. So you have to be bit careful with saying that everything works because we don't know. And, uh, we know, for example, that cognitive behavioral therapy works and that behavioral activation and psychodynamic therapies, but for example, for non-directive counseling, it [00:21:00] is not that clear.

We find in our comparative outcome studies that non-directive therapies are a little less effective than other therapies. But that's also because they're often used as a control condition. So instead of wait list, care as usual, they, these studies used. Counseling as a control condition, and in the studies in which counseling is used as an active treatments, people are trained to, when they deliver it, et cetera, then it's just effective as other therapies.

But if you look at all the studies on supportive therapies, we don't know whether it works. So we should also be a little careful with saying that all therapies work. That's only true for the therapies that we have examined sufficiently. And regarding the, the issue that we don't know who benefits from which treatments.

I mean, [00:22:00] we know from clinical practice that some people benefit, uh, from one treatment and not from the other. You can see that we know that also from the drug field, that people get one drug and nothing happens. People get another drug and nothing happens. And then the fourth drug they get suddenly they are better.

And that can be a coincidence. We don't know. Because if you look at the, if you have 40%, 42% of people getting better at the first treatment, and then you have if of the remaining 50, 58%, the 42% of that remaining, 58% respond to the second treatment, and then they respond to the third treatment. And so it could be, be chance that people respond to the fourth treatments, but.

In clinical practice, it's so obvious that there are some people who can handle [00:23:00] their problems very well with behavioral activation. But when you talk about cognitive restructuring, like in CBT, they try it and they try it again, but it doesn't work. Unfortunately, it's very difficult to examine who benefits from which treatment.

And so all the randomized, most randomized trials of the thousands of the thousand that have examined the effects of the psychological treatment of depression, they have enough patients included to examine if the therapy works. But if, if they wanna examine if it works for a specific subgroup better than for another subgroup.

The sample size of that study needs to be much, much larger, and that can easily be 10, be be 10 times as big. And these [00:24:00] trials are extremely expensive. And, uh, we do psychological treatments, don't have pharmaceutical companies behind them who fund their trials. So we have a thousand trials, but very few are large enough to examine who benefits from which treatment.

There are all kinds of new methods to, to, uh, to examine that machine learning. Uh, uh, we, what we do is we also get the primary data from these studies for specific questions so that we do have a large enough data sets so that we can examine who benefits from what treatments. And there are all kinds of, uh, uh, studies.

Being done now to explore who benefits from what treatments. Unfortunately, that has not been examined very well yet, or it has [00:25:00] been examined very well, but are very few indicators. One indicator probably is baseline severity. Uh, but that's, that just says that, uh, with higher baseline severity, the effects of therapies are getting bigger.

It doesn't say which treatment you need. Another exception is that if you are severe, more moderate to severely depressed, and if you add. Drugs. So antidepressants in a second step that a lot of people get better. So the combination of antidepressants and uh, psychotherapy is significantly better than the effects of psychotherapy alone.

And that's why, uh, usually people are recommended if, if the first treatments, psychological treatment doesn't work, that it, it would be good to try to see if [00:26:00] the addition of antidepressants has a, an, uh, will do the trick, so to say. And what about, I guess the, the other side of that, that question, so not so much about what we know about, on a personal level, what might help people get better, but what are the kind of factors in, in the therapy delivery, the therapy itself, that, that seem to link with better improvements?

Well, there are a few things that we know or beginning to understand. To know. It's, I think it's too early to say that we know a lot. We just don't know very much. There are also studies showing that therapists differ in their, in their effects in specific problem areas. So, uh, that has been examined in very large data sets where you look at, uh, [00:27:00] uh, uh, therapists how good they are in solving problems in specific areas, and then you see that, that that therapist.

Clearly differ in terms of how good they are in handling specific problem areas. And there is also one randomized trial in which patients were randomized to a therapist who was very good in handling their type of problem, uh, compared to just random, giving them a therapist. And that clearly showed better outcomes.

But that's one trial and that's, um, uh, based on analysis of big data sets of trials and therapists. But that has not been yet translated to, yeah, to the use in, in regular routing care. But it should be because that's, uh, that's beginning to [00:28:00] emerge. As a pretty stable, strong predictive factor. Another thing that we find is that, um, the frequency of sessions is related to the outcome.

So it's not so much the number of sessions. Strangely enough, it doesn't matter how long, how many sessions you get. The, the therapy works regardless of the number of sessions, the, the total contact time between you as patient and the therapist. But what does matter is the frequency. So if you have two sessions per week, that's better than one session per week.

We clearly find that in meta-analysis, but we also did a new randomized trial in which we randomized patients to one session a week or two sessions a week, and we confirmed that two sessions per week is better. [00:29:00] Another thing that can improve the outcomes is getting feedback from patients regularly. And if you do that systematically and you discuss it with patients, uh, then we see that the effects are not very much, but they are significantly better than, uh, than if you don't get systematic feedback from patients now.

And then to go into those a little bit more, I mean, what are your, I guess, what are your hypotheses for, for why more than one session a week may make such a difference? Uh, well, we don't know. I mean, that's true for all randomized trials. You find that this is the outcome. So, um, we find that two sessions per week are better than one session, but we don't know why.

Uh, I can hypothesis, I, I have ideas about that, but [00:30:00] there's no research supporting that. So that's, uh, uh, and randomized trials are the best experimental design we have. So if you have randomized trial trials, that evidence is very strong. But if you wanna examine why that is the case, that's much more complicated to examine and that evidence is much less strong.

But if you ask for the frequency, I mean, I think what happens if you go to a therapist, uh, as a patient and that therapist talks, says to you, okay. Your problems are very serious, and we're gonna work on it two days a week. Two, two times a week. That's, that gives a, a different kind of urgency than when you get once a week, you talk about how you feel, you get ideas about and homework for the next week, and then you come back next week [00:31:00] and talk about it again.

But if you go that twice a week, there's no way you will be able to forget it or to, uh, let things go on as they usually do, because you are, you have to go there. That urgency is there, and I think that that's good for patients. It also gives hope because patients are forced to work on it very hard and go back again later this week.

And yeah, so something about the learning and almost the speed of the behavior change that might occur. Yeah, and the, the pressure you have to have to do something, but also the hope that something will change and the expectancies that something will happen soon because you go that twice a week. And when it's a case of the therapist being better able to handle the person's problems.

Could you give an example? I'm, I'm trying to picture what [00:32:00] exactly, what that looks like and what it might mean for a therapist that who can't solve the problems or who can't solve the problems. What, what are the kinds of things? Yeah. So, um, I don't have, they, the, the, the problems for which people go to therapy.

I mean, they go to therapy for depression because of depression. That's obvious. Uh, but. In addition to, to that they can have all kinds of different problems like, uh, interpersonal problems, relationship problems, problems at work, uh, uh, they have personality problems. They, so the kind of problems associated with the depression can vary, uh, co uh, considerably in the studies examining that.

They have used different types of way to categorize that. I don't have the, the exact categories in front of me, but they usually do that empirically. So they look into their own data sets, [00:33:00] look into the problems that people report, and then try to classify them. And that's, uh, um, yeah, that's what I said.

That can be done in all kinds of different ways. Like you also have comorbid problems, like you also have anxiety or you don't sleep well, or things like that. I mean, something that I was quite interested that you've, you've spoken about before is the, uh, the, the benefit of digital interventions and the how, how useful delivering therapy online, uh, still is, how good it still is to do things like self-help, which in just in my professional experience, uh, patients are often quite dismissive of, or disinterested in, they feel like they're getting a raw deal if what you offer them isn't a in-person face-to-face therapeutic encounter.

But actually the, the research tells us something quite different. Is that right? Yeah, and I, I think it's, [00:34:00] it depends on the patients you talk. You, you speak to, there are also a lot of people who don't want go into, wanna go into therapy at all because it's the stigma because they, they don't want to talk with a therapy therapist.

Um, uh, my colleague from who were, who I worked with a long time ago, Isaac Marks, who was one of the pioneers in his area, he's from, uh, he worked a long time in London, in the uk. He said a computer has no eyebrows. And I, I think that, that, thats summarizes very why Well, that some people wanna talk to a person and, and other people Exactly.

Not, they wanna do it on their own when nobody sees it, when they can do it in their own time. So it's, it's not. Um, either or you, it's, it's not, not [00:35:00] the same thing. Um, if you, and if you, if you look into clinical practice, it's also different. It's not that, that patients don't, don't want it. If, if you, as a therapist say, says this is the best treatment, patients will not be that dismissive.

Um, my experience is that it's also therapists who are not inclined to do that because they feel that this is not the optimal care. Because they feel that they send our patients home. Okay, here's a website and I'll talk to you online. That's, uh, but it's, patients are not always, I mean, there are, of course, there are patients who just wanna talk with people and don't want anything, uh, to be done with, uh, uh, uh, with the internet.

But, uh, many [00:36:00] patients are not, not completely negative about it. What we see now more and more is also blended care so that you, you talk with a therapist, you go there and you talk with therapists just like always. Uh, but you do part of the treatments online at home. And why not talk with a therapist about your, uh, problems, uh, with depression or with your at work or at home, and then do a social scale straining online.

I mean, why waste your therapist's time with that? And I, so I, I don't think it's the, either this or that, it's always, uh, uh, matching the right care with a, with a problem of the patient in the right way. I do think that, uh, routine care should be a bit more open to digital [00:37:00] solutions. Uh, but that's a process and that will not go automatically.

But I, I think that will have to change because there are so many people needing help. That's, it's, it'll not be possible to, to deliver that in a very traditional, old fashioned 16 session individual therapy. And then, then we have the problem of low and middle income countries. I mean, I. 80% of the people with mental health problems live in low and middle income countries, and that's, there are 1 billion people in the world with a mental health program, 800 million people with mental health programs in low and middle income countries.

There are more people with depression in China than the whole population of Spain. There are more adolescents with depression in India than the whole population [00:38:00] of the Netherlands. So when you, when you at one solution for that, there are more, more solutions, uh, but you cannot imagine that you train enough therapists to.

To deliver therapy to all these people. That's, that's impossible. So digital solutions are really, uh, one of the possibilities. One thing which is important is that in our research, we have found that digital interventions are as effective as face-to-face therapies. There is no difference as long as, as human support.

So the most digital interventions are self-help interventions. You go to websites where you learn on the website, where depression comes from, what you can do about it, you get homework, et cetera, et cetera. And then you get [00:39:00] feedback from a professional every week if you don't get feedback. And that's, uh, there, there's also a lot of research on that.

Then we usually see that the effects are smaller. They're not zero, but the effects are smaller. And so unguided digital interventions are less effective than, than guided digital interventions. And you must remember that, uh, the guidance delivered to digital interventions, it's much less than face-to-face therapy, and it doesn't, it doesn't have to be delivered by fully trained clinicians.

We work with clinical psychology students who are trained for a couple of days, and they deliver the support during the digital interventions, uh, very well. So guided interventions are [00:40:00] more effective than unguided interventions. But we, when we look at low and middle income countries, we do not see the difference between guided and unguided interventions, so that, that makes the possibilities of using digital interventions in low and middle income countries.

Yeah. That, that's, that's, that's really one of the big opportunities to get help to people who may benefit from it. And that also, that goes hand in hand by the increasing number of people having smartphones. I don't know the exact number, but I think it's above 80% of the world population now who has a smartphone.

Um, and so we did, for example, one trial in Lebanon. Where we delivered, uh, guided self-help interventions through the [00:41:00] smartphone, uh, to people living in Lebanon. And we did a separate trial in Syrian refugees living in Lebanon. And it just worked. And, uh, so we're now implementing it in, in a na, on a national scale because you can implement these things without having an office.

You can just do that digitally. Yeah. So I think that's a lot of possibilities. I also must say that, um, when you think of, uh, getting therapies to people, uh, there are two things. First, also in high income countries, there are many people who don't get treatments. There are 200 million people with mental health problems in high income countries, but less than 50% of those get therapy.

And, uh, that's not only true [00:42:00] for the, these 200 million people include all mental health programs, but if you only look at depression, for example, in most high income countries, not more than 50% of people get treatment whatsoever. And that's often treatment from not highly trained clinicians. And then if you look at specific target groups, they often, the, the, the uptake is much lower.

So, for example, in older adults or in student populations, or in adolescence or in minority groups, or in people with lower socioeconomic. Uh, groups, the uptake of therapies is much, much lower than that 50%. So actually, if you look at it, psychological treatments are only for a few high income groups. In high income countries, which are not marginalized in any [00:43:00] way.

Um, uh, that's the people who get therapy. Uh, and then the people going into therapy can say, okay, we only want face-to-face therapy, but we also should think of that more than 90% of the people who don't get treatment, but who could benefit from it? And then you have to think in a completely different way.

So then you have these enormous numbers of people with depression in China and India, that if you have an intervention that is scalable. Has a very small effect. So, um, let's say that, uh, that the 25 benefits of therapy, we, we find in our studies, suppose we can reduce that, that is reduced with a new intervention to 5%, but we can, uh, we can reach, let's say [00:44:00] 10% of the 50 million people of with depression in China.

Then the impact of that, of that intervention, uh, with a very small effect size is still much, much larger than anything you do in the UK for the treatment of people. So you, small effects can still have a huge impact when you, uh, can disseminate it on a large scale. So these are all considerations you have to take into account when you, when you think about digital interventions.

It's not just for the people who go into therapy and who don't want, uh, uh, digital therapies. It's the, the use of it and the benefits can have a much broader benefit, so to say. Yes. And it's for all the people that can't be reached to be that here in, in sort of my home [00:45:00] patch in London, um, but also across the world who, for, for whatever reason, but perhaps don't, you know, I can think of, don't have time, don't have the, you know, kind of energy to travel.

Don't, as you say, don't want to deal with the raised eyebrows, um, aren't ready yet to, to actually speak to a person. But the idea that they can get a huge amount of benefit from, from doing something online with some remote personal support is quite powerful. What are your thoughts on the developments of the language models?

Things like chat, GPT. Because I think the, the level at which they are now, the prompts that you can put in and, you know, saying like, help me solve this problem in the voice of Ramdas. Um, and it does a, doesn't do a terrible job from on, on a kind of face to face level. But what, do we know anything about how it is?

Is it, is it, is it coming to us? What do you think? [00:46:00] Yeah, there is a, I mean, it's, it's a very interesting development. Of course there is an increasing number of trials examining that. There are very few trial. There are, I find only one trial yet in people with mental health problems. Um, so the evidence we, we don't know.

Um, I, there was this one study in, in, in the New England Journal of Medicine. Um, recently we're using large language models to, to help people with, uh, depre, I think it was depression. And they were very happy that they use, that. They found significant effects. But yeah, my, I'm, as a scientist, I'm also always a bit skeptical because I, uh, that's my job as a scientist.

And so, uh, the effects were larger, were pretty large, but that's because they used [00:47:00] waiting list control groups. So that's, that We know that that overestimates the true effects in, in trials. And if you look at it from that perspective, then it's, it's what, it's actually nothing new. I mean, unguided digital interventions.

Not a large language models, but if you go to a website and you, you learn on the website how you can do cognitive restructuring or behavioral activation or whatever, they also find effects. And this study was in that sense, nothing new because it's an unguided intervention and it had effects which are comparable, which we find for other unguided interventions.

So yeah, it's, uh, it's, it's, it's, it's interesting, but it's not, not, not yet so clear that it will definitely improve the outcomes. [00:48:00] And so one problem we have with psychological treatments is that we find comparable effects for all therapies. There are no therapies which are better. We find no superior effects for the treatment format.

So you can do it individually in, in groups, by telephone, digitally, uh, guided, that doesn't matter. The effects are there and it's the 42% of people getting better. And I doubt whether when you use large language models, whether you can improve that. I don't think so. I don't think that it works that way.

I'm, I, I would be very surprised if a computer can do that better than, uh, uh, a person. I, I, I doubt it. Another thing here is that, um, what researchers think [00:49:00] is that the alliance between a therapist and a patient is very important. And, um, uh, that's the motor or the drive of the therapy. If you know that it's not a human that you, that you are talking to, I don't know what happens with the alliance.

Maybe you get that alliance and you, uh, you, that that would mean that, uh, that the unguided digital tools would be as effective as the guided digital tools. But going beyond that, I don't think that's, that's, that's a reasonable, uh, expectation of the large language models. So they can help in, uh, getting therapies to people who need it, and they can.

Uh, they are, they are definitely, I mean, if you do it well, then [00:50:00] they have definite, they, you can go on forever with 'em and that kind of thing. But I'm not sure whether, I don't believe yet. That they will be able to improve the outcomes. And that's what we need. We need to have better outcomes. We know need to go beyond that 42%, that we get the right treatment to the right people, that we get, uh, add-ons that, uh, that we know more about the sequence of therapies, that we know what combination works best.

We have to go beyond that 42% to get, uh, 80% of people better with the first treatments. That should be the goal. And I'm not sure whether, uh, large language models will really speed up that process of getting these therapies better. But with all these things, if it [00:51:00] happens, I'm completely happy. Uh, but I first have to see it.

We have seen a lot of. Promises and a lot of, uh, people saying, yes, this is so new and this is so different and this will, this is a revolution. And, uh, but I, I have not seen one yet in the last 50 years. The effect sizes over time are completely flat. There's, there are no improvements over time of the therapies, and I certainly hope that we will be able to improve them, but because of all the promises and hypes and the revolutions that have been promised in the past five decades, I am prefer to wait until I see the evidence.

Yeah. So I, I mean, it sounds like you're saying the. Future and what's, what's ahead on the horizon. [00:52:00] Shouldn't be trying to develop a new psychological therapy or a new, not even necessarily a new delivery method, but something about improving what we already have and, and improving how we use the, how we use the tools that we already have.

I mean, what, what do you think is on the horizon for improving therapies for depression? Yeah. I, I, I do not think that it's a good idea to be developed more therapies for depression. We already have a lot of them, uh, and although many of them promise to be much better than the ones we had, uh, none of them has shown that they're better.

So they're, they're, i, I think it's a waste of energy. What happens is that there is somebody coming with a new therapy. And they, they write a book about it and they say, based on my clinical experience or based on my research, I have developed this new fantastic way [00:53:00] of treatments of depression much better than the things we had.

It's much more sophisticated, much more advanced, and clinicians being confronted with the limited effects of their therapies think, oh, this, this will solve it. Now we go, can go beyond the 42% that we will get the 80%. Uh, and they, they, and it sounds so good and it matches with my own clinical experience.

And they get, they go all to go to conferences, hear about from the developers. They read books about it, they get training, they get accreditation for it, and they start implementing it. And then at the end, no patient got better because of all the energy spent. And I think that's the wrong innovation cycle and we should stop with that because it takes away money and [00:54:00] resources from the real innovations that are needed.

And so I think we should acknowledge that we have limited effects and that, that we should start. Innovating care, psychological treatments, but also pharmacological treatments, uh, in a much more rational way. So when we say, okay, in the first treatment you get 42%, and then in the second treatments you get, uh, 42% of the remaining 58%.

And then, and that means that we can get, um, 80% of people better with three treatments. Roughly what do we do with the 20%? And what treatment do we start with? Do we start with a, with a treatment that the patients prefer? Do we start with a, with the treatments for which we have the expertise as clinician, or start with, do we start with an easy therapy like behavioral activation?[00:55:00] 

And so these are things nobody has ever thought of. Well, I think that's the rational way to do that. And, um, we should think of what I talked about earlier, that we should more classify problems of patients, uh, according to the, the types of problems. And look how good the therapist in our institutions are, and then assign them to the right patients that will improve the effects.

And, um, so I, I think we should be much more rational about these things. Our treatments work. They have modest effects. What do we do with that? And then you, you, you get a completely different way of thinking than, ah, we have this new revolutionary treatments while we start introducing psychedelic therapies.

That [00:56:00] will be the, the thing that will change is everything. No, it will not. If we wanna improve the care for people with depression, we will have to do that with small incremental steps by looking at the, uh, at, uh, the, uh, the order of therapies, looking at the combinations of therapies, looking at matching of therapies.

With, of therapists, with patients, uh, doing more research, uh, with machine learning to explore who benefits from which, uh, therapy. And these will all be small incremental steps moving us away from these 42% to the 80% that we want. When you talk about combination therapies, what, what would that look like?

What, what are the kinds of ideas that are being tested or that you're thinking about? Uh, well, usually [00:57:00] combination therapy is the, is the combination of psychotherapy with pharmacotherapy. We know a lot about that. Uh, and we know that, uh, uh, at the short term, psychotherapy and pharmacotherapy have comparable effects.

So you can either take any of them, um, uh, and the effects are comparable overall, but the combined treatment is better than either of them. If you look at the longer term, then it's very clear that psychological treatments are better. So if you get therapy once, so you get one, uh, not one session, but one, one psychotherapy for eight or 12 or 16 sessions, uh, that's better than taking antidepressants for a whole year.

So after one year, that one time psychotherapy has [00:58:00] better outcomes than taking antidepressants the whole year. But if you look at combination therapy, that's the best. That's better than pharmacotherapy, but it's all also better than psychotherapy. And so, um, this, it's a very large study we did a couple of years ago in which we found this.

That was also the reason why the WHO in their updated guidelines now say that you should not give antidepressants without psychotherapy, if that's possible. Of course, many people recommend to start with psychotherapy, also for moderate and severe depression. And then if that's not sufficient enough, then you can, um, add antidepressants, uh, after the psychotherapy has not enough effects [00:59:00] so that you, in the long run, you get, uh, uh, and that's, that's what usually meant with, uh, with, uh, a combination therapy.

But you can combine therapies in all kinds of way. I mean, there, there are all kinds of add-ons being developed. So, um, you can, you can, uh, get individual therapy, but as an add-on, you get cognitive bias modification, which is, is a very computerized way of learning. Uh, how to change the biases. Uh, the biases and thinking you have, uh, the effects are not traumatic and they're probably not, uh, should be much examined, much better.

Or you have cognitive remediation where you learn how to rationally think about things, how to, how you can improve your memory, that kind of thing. And [01:00:00] the, these probably. The, these adults, uh, personally I think they will not have a strong additional benefits. I, I'm not sure whether they will get a 42% to 48%, so to say.

Uh, but I mean, there is a lot of research on that kind of thing. Yeah. I'm not aware of studies or, uh, uh, attempts to combine therapies. I mean, most, uh, cognitive behavioral therapies, you learn already a lot of different tools that you can use to improve your moods, like behavioral activation, like cognitive restructuring, like social skills training, like problem solving.

So that's already part of many therapies. But, so one of the things that we don't know is how we combine those existing therapies in a sort of smarter, savvier way with the medications that we have [01:01:00] in a more. Personalized way is, is that where some of the benefits might come? Yeah, that's what I think. And that's, that will also, I mean, it won't, uh, get us to the 80%, but maybe it'll get us from the 42 to 52% because that's, that's what we have to think of, these small steps to make things better.

And all the, all the, all the, the, there are, there are, I'm, I'm, I'm writing a paper now about all the innovations on in psychological, uh, therapies. There are dozens of them. There are more than 30 that are, uh, that have the potential to improve the outcomes. And um, uh, some of them will work and many of them will not work, but none of them will be silver bullets that will change anything.

And in [01:02:00] the, kind of the, the last point that I'm interested in is in the idea of personalizing treatments and where, where the, where the developments are. I think I've heard you talk about a tool that I think in your, in your group, um, you've developed where people that's combine a lot of the individual patient data and that actually allows clinicians to start to sort of put their individual patient's characteristics in a bit.

Could I, um, I may have got that wrong, but would you share a bit about that? So, um, we have this 1000 randomized trials and we do, uh, we have done, uh, more than a hundred meta-analysis looking at these studies. But in what we also do is we look at, uh, what we call individual patient data meta-analysis that we try to collect.

The primary data, uh, from, uh, from these studies and make one big data set and then examine them. And [01:03:00] that's, that does allow to, uh, look at personalized outcomes, so to say. But we have done that for specific comparisons. So for example, we have them for digital interventions for depression. I, I think there are more than 100 trials and we have the data from them.

And so then we can look at all these outcomes and we have done that. Um, and we have, for example, also data on CVT cognitive behavioral therapy trials comparing cognitive behavioral therapy with pharmacotherapy and, uh, combined, uh, pharmacotherapy and cognitive behavioral therapy with pharmacotherapy alone.

And so we do all these. IPD meta-analysis to examine predictors and the, uh, that, that's the outcomes we find cannot [01:04:00] be generalized to psychological treatments in general, because we look at specific subquestions, so to say. But for example, for digital interventions, we have a done a, my colleague Iki has done a very large IPD network meta-analysis in which we've looked at guided digital interventions, unguided digital interventions.

And then studies also comparing them with each other and with control conditions. And then you can see which individual benefits from guided and which from unguided interventions. And so we have made models and you go, if you, if you read the papers, there are models on the web where you, where you can go and you can fill in the characteristics of a patient and you can see what that patient will benefit from.

And overall, we [01:05:00] find that in mild depression, unguided interventions are as effective as guided interventions. And only when depression gets severe, then uh, uh, guided interventions are more effective than unguided. But these are all what we call. Indirect findings. If you wanna really wanna make, wanna transfer that to routine care, you have to do a new randomized trial showing that the personalized, uh, uh, treatment is better than a standardized treatment and that has not been done.

Uh, so, and what I said about these IPD meta-analysis, that's a ton of work. There are huge projects and lots of work, and that's the problem with all these machine learning and artificial intelligence trials, [01:06:00] trying to examine who benefits you and what, there is a lot of research, but to show to, to have these robust outcomes that can be transferred to routine care.

That's, that's costs a ton of work, a ton of money and lots of research and we're not there yet. But, but there is definitely progress and, uh, I mean that kind of thing that we do, these IPD network meta-analysis and looking at all who inwitz from what, that has not been done before. And so there, there is a lot going on and a lot.

But as I said, uh, before we can get, we are certain, okay, by personalizing it through this tool, we can go from 42 to 48%. We haven't, we're not there yet. So if anyone is listening with a ton of money, uh, investing in some trials that are testing [01:07:00] the personalization, testing how to sequence therapies, testing how to combine therapies, uh, what, what would, would that summarize your prescription or would there be other things that you think the future.

Of psychological therapy needs? No, I think, uh, there are now very large trials, so, uh, that there, it's not that enough. That's, there is a lot of research going on also with a lot of money and a lot of, we need a lot of research in all these different types of innovations and in all these more than 30 innovations, there are people working on it.

So money alone won't solve it. It's, it's uh, it's also, uh, a rational research agenda realizing we, there is no silver bullets. We have to work on each of these small areas, which takes a lot of research, money and [01:08:00] resources. To make these small incremental steps. Uh, so I think that's the most important thing.

And then make, then, then if you look at it from that perspective, it's actually very easy because then you can make a rational research agenda for all these six, eight, or 10% that we can improve the outcomes. And you can say if, if you would have limitless amount of money, you can say, okay, we're in 10 years time, we're gonna examine 25 of these innovations and see how much percent they can get up to the 42%.

And then we're there in 10 years time. Professor Keifers, thank you so much for your time. If people want to hear more about your work, follow you, hear more your thinking. Where, where would you point people to? Well, I, I stopped using, uh, Twitter XI [01:09:00] should say, because I don't think it's a good, uh, I also don't use Facebook and Instagram anymore.

I, I am on Blue Sky and I am on, uh, LinkedIn and, uh, I get a lot of, and I, I wanna start a block on psychological treatments in general. Uh, and I wanna do that on Substack, but I haven't started with it yet. But I will do that in the coming, uh, six months. And, um, uh, I think if you wanna know something, uh, just send me an email and I'll be happy.

I got a lot of emails from a lot of people, patients, clinicians, researchers, and I usually try to respond to them all. Thank you very much. Thank you so much for speaking to us today. It's a pleasure to meet you. You're welcome.