The future of depression care
30:47 Share

Hi, everyone. It's Russ Altman here from the Future of Everything. We're starting our new Q&A segment on the podcast. At the end of an episode, I'll be answering a few questions that come in from viewers and listeners like you.

If you have a question, send it our way either in writing or as a voice memo, and it may be featured in an upcoming episode. Please introduce yourself, tell us where you're from, and give us your question. You can send the questions to thefutureofeverythingatstanford.edu. The future of everything, all one word, no spaces, no caps, no nothing, at stanford.edu.

S-T-A-N-F-O-R-D dot E-D-U. Thanks very much.

We've seen this huge transformation in psychiatry thanks to there's a mapping of the human brain that has occurred, a little like the human genome. There's an explosion of data that we have access to and data sciences approaches. So that gives us a way to make sense of what we've learned from neuroscience about understanding how the brain functions. Music

This is Stanford Engineering's The Future of Everything, and I'm your host, Russ Altman. If you're enjoying the podcast, please rate and review it, because that helps spread the word about the podcast and get everybody on board with the future of everything. Today, Leanne Williams from Stanford University will tell us about new ways to precisely diagnose depression. It turns out there are multiple types, and they may need to respond to different types of treatments. It's the future of depression.

Before we get started, a reminder to rate and review the podcast. It'll spread the word about the podcast and it'll help us make it better. Depression is a terrible disease. Many of us have experienced it ourselves or through our friends, family and colleagues. One of the frustrating things about depression is it can be very difficult to treat.

Sometimes you take a medication, try another medication. It'll take weeks or even months to find the right regimen. And during that time, you're at risk of all the terrible side effects of being depressed.

Well, Leanne Williams is a professor of psychiatry and behavioral science at Stanford University and an expert on depression. She has been developing new ways to diagnose depression that are more precise. In fact, she finds six types of depression. And guess what? Each of those types may have a slightly or very different need for the therapies that will work. She'll tell us about that, and she'll also tell us about how

depression is becoming less stigmatized with our ability to be more precise about defining it. Leanne, you're an expert at depression and depression has both colloquial and clinical definitions. Colloquially, we all feel sad sometimes, but that's not the same as depression. So I thought we could start out just making sure that everybody understands what we're talking about when we talk about clinical depression. Absolutely. And thanks for covering this topic, Russ.

When we're talking about clinical depression, we are specifically talking about symptoms that are so extreme that they are actually disrupting your ability to function. And that would be be able to function at school or at work. Maybe it means even not being able to get out of bed. So at that level of an extreme,

An analogy might be to think of high blood pressure where maybe sometimes it's part of normal variation, but at higher levels it'll become and it'll have an impact on your function.

And one of the things you've both spoken and written about is that there needs to be an understanding of just among the lay public that this is a real disease with real mechanisms, that this is not a character flaw. Did you want to just address that for a moment? I think this is a really key topic, that depression is a real illness. It's a real disease. We can situate it in terms of how the brain is functioning, right?

And in absence of thinking of it as a real disease, it's more straightforward to make an assumption that, oh, this is someone not trying hard enough or something like weak about them or a character flaw, which we wouldn't do for other illnesses. We wouldn't do if someone had cancer or heart disease or diabetes. Yeah.

So, okay. Now, thank you. So now we've established two important pillars. And I want to start with the fact that treating depression is kind of famously difficult. It involves multiple medications, trial and error. There are multiple kind of

factors that help determine whether a patient will respond. And so as we start to talk about your work, your exciting work, trying to make progress in this area, give us a feeling for what the landscape for treatment of depression looks like now or before some of your new innovations kind of take hold fully. Right. The current landscape, if we think of what would we do as a clinical standard, I

is the same way that we've been doing things for decades. And that would be where we observe someone, we would get a sense of what they're experiencing, we would ask them to report to us, what are you experiencing? We wouldn't be using any tests. We'd ask about family history and prior life experiences and

From that conversation and interview, we would arrive at a kind of a one size fits all diagnosis like major depressive disorder. And we would apply that to someone assuming they fit the average of what we know about depression. That diagnosis would not tell us about which specific treatment might be effective for that individual person.

So what we would do is follow this clinical heuristic, which means we would try several treatments. First, most typically SSRIs, the typical antidepressants we know about. We try one. Usually we'd wait around eight weeks, see if it works. And by works, I mean does the person, does the individual tell us that we're feeling better, they're feeling better?

And if yes, we would continue with it. If not, we would try another. And that's literally the trial and error. On average for depression, the latest statistics tell us on average it takes yourself, our loved ones, seven years to find the right treatment. So it's very extended and

If you were thinking of this for cancer, where I know you specialized, we'd be kind of waiting to stage for cancer. And of course, implicit in what you're saying is during that entire seven years where treatment is not adequate, the patient themselves is at risk for terrible consequences of the disease. Right. Right. It's not neutral. Yes.

Okay, so great. So we've now established, now we can get into the excitement, which is there is an explosion of discovery happening both at the molecular, cellular, and clinical level of psychiatry. And can you take us through what are some of these exciting developments that are giving you and others new approaches towards depression treatments? Absolutely. It's a really exciting time, and this has been particularly in the last decade that we've

We've seen this huge transformation in psychiatry thanks to there's a mapping of the human brain that has occurred, a little like the human genome. There's an explosion of data that we have access to and data sciences approaches.

So that gives us a way to make sense of what we've learned from neuroscience about understanding how the brain functions. And as you said, that's across molecular through to what I would refer to as circuits, large-scale circuits of the brain where I focus. And that is in the human brain thanks to imaging techniques like functional magnetic resonance imaging techniques

we can actually see your brain in action. So we can take a kind of a movie of the brain and look at the circuits that are regions of the brain communicating with each other. And they're governing the very experiences that not only make us human, so the emotions we feel, the thoughts we have, the behaviors we take, but that when disrupted result in the disease of depression.

So because of these imaging technologies, can you see...

forgive my phraseology, short circuits. Can you see when in disease the things are not working the way they would typically in a non-disease state? Exactly. So we can see those different degrees of functioning and then dysfunctioning. So there's a circuit or network in the brain that is often referred to in the field. It's called the default mode circuit.

It's when we're not focusing on something in particular and we literally see when it gets stuck because it can get stuck in one of two ways. You can either get stuck because it's over-connected and it can't break out of that state or it becomes under-connected, fragmented.

And so that means you can't flexibly switch into other states or you can't easily switch to another circuit function.

Now, okay, so that's fascinating. And it leads to some, my head exploding is what it leads to in a good way. So does that mean we're looking at a time where we might have diagnostic criteria that are not purely the assessment of a psychiatrist, of the feelings and the emotions and the functionality levels, as you described earlier, but there might be imaging evidence to help us also make the diagnosis? Right. That is exactly where we're moving. Yeah.

With the technology I've developed, we are starting to use it clinically. It's early days, but it means we have tests that they don't replace the psychiatrist or that detailed observation. We would need that like we do in other areas of medicine, but they give us a way to have an objective measure. And you could think of particular aspects of depression where that's especially important, right?

For example, one criterion symptom for depression is actually cognitive problems or problems concentrating, which are very difficult for people to report about and say, oh, I can tell that I have a problem with my executive function and I can source it in how my brain is functioning.

So, when you look, so this is very exciting and it's amazing that it's actually being, starting to be used clinically. Is it the case that one size fits all for the diagnosis? In other words, is there a standard set of these circuit malfunctions that lead to depression or are you seeing heterogeneity and are there different types of depression when you start looking at them with a combination of these new modalities? Right.

There are definitely different subtypes. I've been referring to them as biotypes, as have colleagues at other institutions. And biotypes is a term to specifically highlight that it's a biological type, but they do map on to specific types of symptoms and, importantly, treatment outcomes.

And what we know, if we just think of the symptoms used to get that one size fits all diagnosis, there's an enormous amount of heterogeneity. People don't have the same exact type of depression. There are nine key symptoms that we consider for defining depression. And the diagnosis is based on having at least five of them.

Two are considered cardinal. You have to have one of them. But still, just mathematically, that's a lot of different. That's a lot of, yes. So what we're finding when we image the brain is where there's at least six from what we've discovered, six biotypes. We reported on those recently in Nature Medicine.

Prior to our finding, other groups have found four when you look at only the brain at rest. We look at the brain when it's engaged in different tasks, and that's where we can see some more.

I would be surprised if that's, that it's not more than that. And certainly as we go on and we use more sophisticated machine learning AI, I'm sure we'll refine these further. Okay, so this is very exciting. So let's say for now we have six of these biotypes.

I'm guessing that they must correspond a little bit to what psychiatrists have known for a long time or else they would be very irritated and they wouldn't accept your findings, among other things. So I guess my question is, are they mapping indeed onto some of the well-known kind of syndromic depressions? And are there any surprises? Right.

It's a great question. The answer is a bit of both. So some of these biotypes do map on to what we already know and the symptoms. So for example, we can see if the circuit of the brain that helps us feel good, like respond to reward, if it's flattened, kind of blunted,

It's associated with a symptom called anhedonia, like lack of anhedonia, and that's a key criterion symptom for depression. In other cases, we've seen surprises. So one of them is the importance of these cognitive impairments. And I guess we think of depression as sadness and kind of low moods.

really because of the development of SSRIs, which targeted sadness. So the biotypes help us remember that there are these other features and symptoms, and we found that

cognitive biotype is important for understanding the heterogeneity and who doesn't respond to SSRIs but may respond to other treatments. Yes, and that last comment you made about responding to SSRIs, I'm guessing that a big hope is that once you figure out which of these six biotypes a patient falls into, it might make treatment decisions less

And is that true? And how is that looking in terms of mapping treatment patterns to biotype patterns? It's looking very promising. And that I say based on the actual evidence from my lab and many others, I undertook one of the very first of these studies to use imaging to predict treatment response across three different most commonly used medications.

medications. That was years ago. We completed it 2013. We're still analyzing the data. And since then, I've led a series of studies in other labs, a series of them. So what we know right now is if you look at the traditional approach, typically only one third of people will get better on those first few tries. And then it's kind of diminishing returns.

And if we use the imaging, we can boost that predictive accuracy quite substantially and we can get to a point where we're doubling the chances of getting better.

by having that information. So that is very exciting. And may I ask, and I know this is new, so forgive me for being greedy, but is it the medications that change or the doses? Can you give us a sense of how the medication regimens change across the biotypes or may change in the future? It's a key question. So there's two related approaches. One is there's

There's quite an array of what you might consider the standard antidepressant treatments, different types of SSRIs, what we call SNRIs that tackle multiple brain chemistries.

behavior therapy, cognitive behavior therapy. So what the biotypes do is help determine which of these is likely to be most effective for each person and which not likely to be. So we can rule some out and that's the process by which we can get someone to the right treatment much sooner than cycling through.

For some people, it might be the standard SSRI, but that's going to be a relatively small number, maybe maximum one-third of people. And even then, it's like which SSRI? They don't all act the same way. So we can help select between them. So that's a kind of prospective decision. And then one thing that is really exciting about depression, we have a lot of new treatments.

We have transcranial magnetic stimulation and neuromodulation technique. We have all sorts of exploratory therapeutics and now looking at repurposing some medications. All of those are being developed for treatment-resistant depression, which means waiting until you've failed the other. Uh-huh.

And typically people, maybe they're failing seven or nine. That's like disheartening. And as we're seeing before, not neutral. So the second exciting aspect of having biotypes is you have a way to potentially fast track people to a new treatment and an argument to say, we don't need to wait for them to fail five or seven or nine.

This is the Future of Everything with Russ Altman. More with Leanne Williams, next.

Welcome back to the future of everything. I'm Russ Altman and I'm speaking with Leanne Williams from Stanford University. In the last segment, we set the basics of depression and the exciting new ways of finding biotypes, six of them, which are different subtypes of depression which may map to different treatment requirements. In this segment, Leanne will tell us about how AI is helping revolutionize her analysis of her data,

She'll also tell us that psychedelic drugs may be part of the treatment regimens for some of these biotypes. And finally, she'll tell us that stigmatization of depression is actually being reduced with our more precise ability to diagnose it.

Leanne, I wanted to ask you about the role of AI in your work, because I know it's involved and it's been striking to me across all fields of science, how AI kind of raises its head and its utility in all kinds of different ways. So has AI made a difference in your research program? Absolutely. And in an increasing way. So we're explicitly planning an AI program for expanding and refining the biotypes and

In the biotypes I've discovered so far, we've used more kind of classical machine learning to identify those clustering algorithms and so on. And now we're in a position to really make use of the rich data sets we've already developed or acquired. And they are very deeply clinically phenotyped. We have a lot of rich data there.

on those datasets and we've also been acquiring them in standard ways. So we have the same kind of imaging formats across all the datasets. What that means is we can now go to much bigger datasets like UK Biobank or others and take more of the foundation models with the kind of large language models but of course in this case

We are using them with brain data and we're going to the raw time series of the brain activity. What's interesting is then we are applying neural nets. So we're actually applying neural nets to the actual neural networks. And that gives us a way to really get deep into the language of the brain and the kind of dynamics that we may not have accounted for

in the biotypes that we have now, how the biotypes might interact, how there may be sub-biotypes or like just really getting into the detail. And once we go across those larger data sets, so Biopank and those acquired by other labs,

we can come back to our very richly phenotype data and fine-tune what we find using those large language models applied to brain data. So that's something we're very excited by, and it will give us a way to use AI to get more at the kind of details of the essential structure of the biotypes

and really refine also the prediction of treatments. So it sounds like the AI, you're confident and probably already have some evidence that it can see patterns in the data that even skilled researchers might miss. I'm confident about that. And

It comes about because to do the more traditional machine learning, we are abstracting slightly from the raw data using kind of calculated metrics. So when we look a bit deeper just in smaller sets, we can see there are more dynamics that we may not be picking up.

Really fascinating. And a couple of other things I wanted to ask you about. One is, in your work, you've mentioned the use of these psychedelic drugs, things like ketamine and MDMA and others. And that's, of course, that gets people's attention. And I'm just wondering, are we imagining that these will be part of the armamentarium for when we have these biotypes that some of them or some or all of them may benefit from these medications? Yeah.

Right. I imagine some will, and certainly the evidence so far from our studies indicates that there are specific biotypes that respond to those, what we could call rapid acting or exploratory therapeutics. Personally, I'm agnostic to what treatment is.

standard meds, behavior, kind of whatever works, what works. And certainly I think that's the same for individuals that whatever works and we do it in an informed way and safely and

For MDMA, we have work that's being reviewed right now where we see that one biotype that is the threat circuit of the brain being switched on. It kind of stays on in this kind of alarm mode, and that's when people feel very stressed. They may feel almost like overwhelming sense of anxiety or arousal, maybe trauma feeling stressed.

That we see change with MDMA. And that's one example. Very exciting. So the other thing I wanted to ask you is, we initially started talking about the biotypes in the context of these kind of novel and exciting new imaging technologies.

But I know from reading your papers that it's not just a single modality. You already talked about the importance of the clinical perspective and the elicitation of the patient history. But I know that you're thinking about other things and that we really should probably think of these biotypes as multimodal integrations of multiple sources of evidence. And so what are the other sources of evidence that you imagine either now or in the future folding in so that the biotypes are very rich?

Right. I imagine at least two modalities and there'll be more. I just want to take a step sideways for a moment. When we're doing the imaging, we already have a couple of modalities. And here I draw the analogy to cardiology. So we image the brain when it's at rest, kind of freely thinking, and then we measure it doing specific tasks, which are a little like cardiology with

Heart at rest and heart when it's being stressed. What do they call them? Treadmill tests. Right, right. And that gives us already important complementary information.

The other modalities we've found are really important is asking in life history about experiences of early trauma, also current trauma. But early trauma can shape already how our circuits are functioning or functioning in a relative way to each other. So that's important. It also helps us predict treatment when we combine early life trauma experiences with the biotypes.

The other modality we're looking at is pharmacogenomics. And there'll be other aspects of omics, but pharmacogenomics, very practical. And I know this is something that you've been pioneering, where we can pin down, if it's a medication, the dose or how to personalize the dose. And

For example, that's been important with standard SSRIs, but now we're seeing also with the psychedelics. So the genetics is infiltrating the AI. It sounds like multiple ways of using the imaging. Are there other major new sources of data that might inform the biotypes? I believe so. So ones I'd be fascinated to look at would be inflammatory measures,

There's evidence that with the chronic stress that can lead to kind of having someone tip from everyday depression to a disease may also have a kind of ongoing inflammatory effect. So there's some indication that relates to some of the biotypes, for example. So looking at that.

We're also looking in a current project funded by NIMH at getting what you might call surrogates, the behavioral and digital surrogates that are kind of readouts of the biotypes, but more objective than asking someone to report on their symptoms. Yeah, you mentioned cognitive where we might not be the best at reporting our own cognitive status, but...

not to diminish it, but an app might be helpful for, for evaluating cognitive status. Yeah. To,

And it could be in a more passive way. We can get a little sense of the speed of someone's processing or actively asking them to do some simple tests. Great. So in the last minute, I wanted to ask you about stigma. I know you've thought about this. You've written about it. And people might say, well, what does all of this have to do with stigma? But as you know better than I, depression is a stigma problem.

stigmatized disease. So how could these, the work that you're doing actually impact the societal reception of depression as an entity? One of the most pleasantly surprising aspects of our work is just what an impact it has on the stigma.

We hadn't foreseen how much of an impact. So what we've observed is when someone actually sees their brain, they literally see their own brain, and then they can see if they're experiencing depression where some of the circuits are stuck or where they're overactive. It's such a transformative moment for them.

you can literally see them shift from, I thought this was my fault, like that sense of blame they're carrying to being able to see that it's tangible, it's something outside of them.

And that just in that moment changes the experience of stigma. That is remarkable because I thought you were going to talk about other people's views of the patient, but it sounds like it starts with the patient themselves having a different understanding of what's going on.

Yes, and then absolutely other people as well. So I had one parent recently who said how much it changed how she could talk about anxiety with her daughter because she could say, now we can talk about, you know, when your brain's in that alarm mode, like that amygdala in your brain that gets fired, how are we going to deal with that?

And it's not about, oh, you're anxious and that sense of something that's less tangible. Well, there you have it. We have better diagnoses. It's going to hopefully lead to better treatments and there'll be less stigma. So that is a very positive outlook towards a very tragic disease.

Thanks to Leanne Williams. That was The Future of Depression. You've been listening to The Future of Everything. I'm Russ Altman. You know, we have more than 250 back episodes in our archives, and you can listen to these conversations about the future of anything.

Please remember to hit follow in whatever app you're listening to so you'll always be alerted to new episodes. It's like clockwork and we want you to be part of our clock. You can follow me on many social media including Threads, Blue Sky, Mastodon, and LinkedIn. And you can do the same for Stanford Engineering at StanfordENG.

If you'd like to ask a question about this episode or a previous episode, please email us a written question or a voice memo question. We might feature it in a future episode. You can send it to thefutureofeverythingatstanford.edu, all one word, thefutureofeverything, no spaces, no underscores, no dashes, thefutureofeverythingatstanford.edu. Thanks again for tuning in. We hope you're enjoying the podcast.