Sleep and the Evidence Base of Digital Health

A lot of my research had been on sleep, but some of it had been on epilepsy. I was interested in both, but sleep was really the main thing that caught my attention, because it seems so central to people's mental health and wellbeing. I often think of sleep as what nature provides for our mental health — from the outset, what's there to help us. Sleep seems to be a fundamental thing for emotional wellbeing and for cognitive functioning.

So I was always fascinated by the fact that sleep disturbance was very common when people became unwell, and often seemed to usher in periods of ill health. The challenge was always: how do you treat a condition as common as insomnia? What I did initially was train a lot of people in evidence-based cognitive behavioural therapies. We did big trials, published and acclaimed, in primary care and in cancer care, training nurses to deliver CBT, often to small groups. But the reason I founded Big Health was because, in the end, you can't scale up in-person care the way you can scale up molecules.

So when I met Peter Hames, who co-founded the company with me, I thought there was an opportunity here to develop something built on algorithms that were as scalable as molecules — a way to take CBT out at scale and offer an alternative to prescribing. Given that CBT is the number one treatment of choice according to clinical guidelines, we wanted to make it available somehow.

Well, I'll start with the family — they thought it was hilarious, because Dad had no idea about technology. Why would I get involved in a digital business? "You can hardly use your phone." It wasn't quite as bad as that, but technology is not my forte, it's not my strength.

But throughout my academic career and clinical practice, I've believed in a counsel of excellence: if you want the best person to do a certain thing, you find them, you collaborate with them, you do the best studies, you do the best work. So the fact that I didn't know anything about technology didn't put me off — I'd collaborate with people who were experts. What we're really trying to do is help people change the way they think and the way they act, and so improve their sleep. So it seemed like a natural thing to do.

But it was a very difficult decision professionally, because it felt like I was crossing a Rubicon. I'd always been a traditional UK clinical academic — you work independently, you're not conflicted, you publish the science, and it's someone else's responsibility to do something with it. You don't get your hands dirty actually making inroads; you pass on the information and it's up to other people to commercialise it. And I should say, I still bear that conflict of interest very honestly, because I think it is a conflict of interest.

But I began to realise that what you do is report conflicts — you say you're conflicted, you recognise you've got an interest in the business. At the same time, the reason for founding the business was to solve a problem. One of the biggest things that helped me overcome this was recognising that the NHS doesn't make anything at all. It doesn't manufacture anything. All the drugs, all the wheelchairs, all the syringes, all the imaging equipment — anything at all — is bought in or outsourced. So it was okay to make something of quality that the NHS would procure against its own standards. That helped a lot: "Oh, that's true — the NHS doesn't make anything." I never really thought the GPs and clinicians were grinding their own powder, don't get me wrong. But it was like an aha. Somebody's got to do this. And if we're going to do it, it should be done properly.

I realised that within the broad reach of medicine, partnership with industry is actually very common — in the biomedical sciences, so much would never have happened were it not for partnerships with industry, and many leading academics were involved. To some degree, I think you have a duty of care not to stand off on some high moral ground, but to get involved and influence things. But where I was coming from wasn't an equipment- or technology-enabled area. I come from the world of talking therapies — you, the person, and maybe some paper and pencil. So it was much more unusual for me as a clinical psychologist back then, about 12 years ago, than it might have been for medical colleagues who'd been working with the medical device or pharmaceutical industries.

For the university sector it wasn't at all unusual. But for my peer group within my own profession, it was. One of the things people were concerned about was that by developing a fully automated intervention, wouldn't this take our work away? Wouldn't it remove the need for in-person therapy? I still find that an extraordinary idea. But people's fears and perceptions are really important.

The reason I find it extraordinary: in the UK you've got about a one-in-10,000 chance of seeing someone for your sleep problem, compared with getting a pill for it. Now, I'm not at all against medication or any evidence-based treatment — I'm for evidence-based treatment. And that's the problem, because CBT just isn't available to people. So the idea that it's going to take someone's job away, when there's a potential 10,000-to-one increase in throughput, and CBT is the treatment of choice for insomnia — it would surely grow the availability of the stepped-care system to look after people's sleep.

And it would lead to more people getting help in person too, because this isn't a panacea for everyone — a digital pill, as it were. So that was one of the perceptions people had: that this was somehow destroying the opportunity for therapy. I saw it as advancing the opportunity for therapy, and giving patients and doctors choice.

Yeah, there is that tension, and it's something I constrain. This is why I'd differentiate between treatments and toys — that's an extreme comparison, but there's a lot of activity in the sleep space, and in the mental health and wellbeing space generally, that's driven by marketing and market claims and pays scant attention to evidence. They just want people to use their products, because they want you to believe you need them. I think it's really important, in terms of clinical governance, that things making claims in the market actually deliver against those claims in clinically meaningful ways. So I'm interested in people who have a disorder of their sleep primarily — how we can find an effective treatment for a disorder, whether that's insomnia...

...or other aspects of mental health, be it anxiety or depression. Big Health has another product called Daylight, which is for generalised anxiety disorder, as well as Sleepio for insomnia. And we'll look at developing other products, for depression and so on. I'm interested in providing a choice of evidence-based alternatives to the millions of people who have a diagnosed illness and are currently being treated, but don't have a choice among evidence-based alternatives — because the only alternatives that are scalable are pharmaceuticals. As I said, I have no beef with pharmaceuticals; I think they've transformed psychiatric care, and that's great. But they're not the only treatment that's effective or in the clinical guidelines, and not the only treatment patients want or doctors wish to prescribe.

So I hold our company to the highest possible standard: we will only make available products that have already been demonstrated, in a clinical trials pipeline, to be effective in a clinically meaningful way...

...and in a sustained way, against known outcomes used in the clinical domain. Other people don't believe that necessarily, and other companies don't have that perspective. But the thing is, Sleepio and Daylight aren't apps. They're treatments. They're interventions for a problem. A good way of thinking about it: we know there's pills and there's pills. There's a treatment for a condition — something you'd be given by your doctor because of a problem she or he thinks you have, and they don't want you to have it, so let's treat it effectively. That's a proper medical intervention, often in pill format. But we can also go and buy pills for all kinds of things — in shops, in homeopathic or health food stores. I'm not saying people shouldn't do that. But those aren't treatments; they're based on marketing claims. Now, the pharmaceutical industry, of course, backs up its drugs with huge marketing efforts as well.

But in this nascent digital therapeutics — or digital medicine — industry, we need to scrupulously observe the highest possible standards, because otherwise we'll do the public a big disservice and deceive them. People are easily misled, particularly if they're already vulnerable and have a mental health problem. So there's a duty of care for that governance to be inside the company as well as outside it.

That's a good question. Many of the trials are almost zero expense to us. We've never commissioned a trial. We've conducted several studies in-house, using available programs like Qualtrics to gather information, run with the equivalent of a research assistant, and we've paid trials units to analyse the data independently. So the average cost of a study is probably in the low tens of thousands — 10, 20,000, thereabouts. Much less — an order of magnitude less — than an HTA trial or anything like that. Most of the other studies don't cost us anything, because they're funded by external bodies. These are inbound requests to use the program.

Let me give you a couple of examples. There's a study conducted at UCSF, published early this year in JAMA Psychiatry, on sleep problems in pregnant women, led by someone called Jen Felder. She approached us a few years ago, interested to see if we could improve sleep during pregnancy using CBT. We had a set of principles we published 10-plus years ago: anyone could use the product as long as they pre-registered trials with primary outcomes, it was a properly conducted study, and they published the results independent of us. My approach really is about developing our knowledge base around something, not about making it commercially more attractive. We'd never have done anything like this number of studies if we just wanted to show that it works — there wouldn't be a strong argument for that. But there's a big pipeline of studies, many of them ongoing.

So Jen published this study in the end. We didn't know she'd finished it. We'd never been involved, we didn't have any of the data, we weren't on the authorship list. And I didn't really expect it to work, I'll be honest, because sleep problems in the last trimester of pregnancy are often quite situational — hormonal, or to do with iron levels. But equally, they're not the kinds of things you want to treat with drugs, because of the pregnancy. So I was very happy — the results were very positive, showing improvements in women's sleep. And she's gone on to apply to NIMH for a further study, because they found these women treated with Sleepio were less likely to develop depression postnatally. So many of the studies are inbound, where people come to us. That's really why we have the volume we have.

I've never thought about it as a portion of the marketing spend — it's never occurred to me. Although people have said, "That was very smart, what you did." And I said, "What do you mean?" They said, "You did a placebo-controlled trial. That was very clever." And I said, "That's what you're meant to do." Internally, it caused chaos in the company.

Right at the beginning, I said, "You're going to build a placebo version of Sleepio now, because we're going to do a placebo-controlled trial." And people said, "You don't believe us, you don't trust us, you don't think we're doing anything of value." And I said, "Well, I don't know until I've seen that it works better than a placebo." Because if you do something fancy that people think is going to help them, it often does. And we need to know it isn't just a placebo effect. We did that not just for Sleepio, but because we were at the beginning of a new industry, and it had to be demonstrated that interacting with your computer or smartphone, doing things with avatars or characters, is not just fanciful. A company should do that with all its products, because there's a risk we lead people astray just because we can prove that doing something is better than doing nothing. We need to prove these are clinically meaningful benefits, attributable to the content of the program.

Well, the thing is, you won't get a trial published in a reputable journal if it's not pre-registered. You won't get a randomised controlled trial published, full stop. You'll get it published somewhere — someone will publish it.

And that's where the stealth comes in, in terms of your motivation for doing research on something. Your motivation really should be to protect the public and to do good. People will say, "Well, maybe randomised controlled trials aren't the way to test digital therapeutics." I'm open to methodological arguments with anybody, and I get that point. But sometimes it's suspicious that what people actually want is to find a way around it. And I'll just say that isn't right. Full stop. At my age, I don't need to justify that to anybody. It's deception, that's what it is. If you're promising to help someone's depression...

...unless we don't think mental health matters as much as physical health conditions — and I've spent my whole career advocating that mental health should have parity in all respects — we're dealing with vulnerable people who want to believe somebody is going to help them. They can easily get misled by an app store or digital offerings that claim this, that and the other. They can't differentiate between them. And neither can the clinicians, because someone goes in and says, "I don't know if this data from this device is true. What do you think?" And they have no idea. Well, I'd say they have every idea: if we just adopt the same standards as medicine, it is just plain medicine. That's it. No different. People say, "But how can you be agile? How can you move quickly?" What they mean is, "How can I get to market as quickly if I have to do that?" And I say, "Well, that's your problem."

We need to be able to test that things work before we offer them at scale to the population. Does that not make sense?

Yeah. But there's a lot of work that goes into that pivotal study even before then — a big investment of time, energy and commitment. Look, healthcare is a serious business. Pharmaceutical companies have many, many drugs they have hopes about that never make it, because they're associated with too many problems, or the risk-benefit profile isn't great, or the results aren't clinically important.

I suspect many of the things out there right now that are meant to help your mood have trivial benefit. That's okay if you're not unwell, but if that's becoming your treatment for your condition, that's really not okay. The industry just has to do the hard yards. It's not complicated — you just have to do the hard yards. What people sometimes want to do is build apps in their spare time because of a great idea and then make them popular — and good luck to you. But that's not what I want to do.

And here we've talked about conflict of interest around profit and marketing — but there's also conflict of interest around opinion. In the academic world, people are conflicted anyway: you want to get your trial published, you want the results to be positive. There have to be constraints against that, even with no commercial motive whatsoever — reputational issues, personal advancement. You need the checks and balances. That's why you register a trial: you say, this is the primary outcome, you measure and report the primary outcome, you follow the statistical analysis plan, conducted independently, you follow CONSORT guidelines. The methodology is there — it just has to be adopted. I don't think digital medicine is anything different from any other kind of medicine. I don't see where it gets its exemption clause. What gives people the right to say, "But this is different"? And if they say, "It's because it's mental health," then it's a red rag to a bull to me.

That's part of the reason I got involved in this to start with — that we're not managing to get things out at scale that make a difference to population health — and digital is a way of achieving that. But it has to be done with the same mentality and governance we'd apply to anything else designed to have a meaningful effect on a condition.

To my mind, the ends don't justify the wrong means. There are serious discussions to be had about how we use big data, real-world evidence, how we construct studies to use the potential power of the digital space and the data you can't avoid gathering — especially with software.

All you've got is algorithms and data and interactions, data transferring back and forth. So there's an opportunity to transform our approach to interventional studies going forward — I've always said that. But you can't have those discussions with people who don't believe in evidence, because you have to start from the belief that the evidence really matters. The current standard for what your taxpayer money pays for is that things meet authentic standards.

For many years, warts and all, the trials pipeline — proof of concept, efficacy studies, clinical effectiveness studies, long-term monitoring — has been the standard of evaluation, and I think that should remain the case. If people want to spend £2.99 of their own money, they can do whatever they want with it — it's none of my business, it's a free world. But I don't think public money should be spent on products that aren't proven to work. Half the problem is already solved: there's already a methodology, there's a playbook. Follow the playbook. So there's a difference between the wellbeing market — a buyer-beware environment — and healthcare. I'm not saying all of that should be regulated through the NHS. But I'm concerned about healthcare. That's where my focus is.

So that's a bit similar to the pill-and-the-pill argument from earlier. I fully respect it. I just say: that's fine, but don't cross into the world of healthcare with it. Because the idea that "it won't do any harm" is not the approach we take when we sit down in a consulting room with a patient.

Yeah, I think that's right. But if somebody is depressed — and some of these things can look similar, which is part of the difficulty. Take talking therapies that are evidence-based, like cognitive behavioural therapy. Suppose somebody was actually depressed. We might know that seeing a clinical psychologist would help that person's depressive illness...

...and there's an evidence base that CBT for depression is effective — perhaps as effective as the major SSRI drugs. So the clinician has a choice of these two things. What they don't have the choice of doing, I believe, is saying to someone, "Just try to think more positively — see you next week," or using something that sounds a bit like therapy and saying, "Well, we do CBT." This is the stealth piece: there are hard-won, evidence-based therapeutics which involve certain things being true, and then translating those — which isn't trivial — into something that doesn't involve a person, and demonstrating they actually work in that situation. It's a bit like a drug: this works intravenously, but does it work sublingually? You don't just get to say, "Well, it worked that way, so it must work this way too, so we'll just swallow it now." You've still got to demonstrate it. These look-alike things that say, "Our programme includes a little bit of that" — it's not the same until it's shown to be the same.

Now, it's unlikely the doctor's primary treatment recommendation would be "take a hot bath." If they were saying that, they probably don't think the person has a medical condition that needs treatment. I'm differentiating those things. Wellbeing is really important — I've spent a lot of time this past year during the pandemic doing talks on sleep and wellbeing. Some people may simply be trying to optimise something that's already okay, or have an occasional bad night, or wonder what to do when the kids won't settle. All that is hugely important. But what I'm saying is, I'm not treating those people as patients with a condition. That's the binary differentiation I'm making. I know there's a continuum in every illness — hypertension, whatever, everything's a continuum. But it's only if you don't really want to understand that differentiation that you struggle with it.

If you want to pick holes in it, yeah, it's challenging. But you've got to wonder what your motivation is for doing that.

Years and years ago — maybe 25 years ago — my wife gave me a sheet of paper she'd printed off. I put it in an acetate and stuck it on the wall of my office at the University of Glasgow, when I was head of department there. It was a saying from a writer called Henri Nouwen. It said, roughly: "I used to get annoyed and frustrated at all the interruptions to my work, until I realised they were my work." It's a slightly philosophical point, but we can be so absorbed in what we want to do, and how we're going to plan our day, that we sometimes fail to realise that when people need something — a fellow, a PhD student, somebody inquiring about something — you need to structure and manage those things efficiently...

...but dealing with frustration in the workplace is really important, and recognising that interruptions are just part of the job — often important parts of the job. I mean, I'm doing this interview with you, right? This wasn't planned — well, it was planned, but I wasn't looking for it, and I've got other things I could do. But it's important.

And one of the best ways to deal with this, in my experience, is when there's that knock on the door, that interruption, that email pinging — "you haven't responded to the email I sent five minutes ago, or the Slack message" — it's to say to people, "Okay, I can give you five minutes just now. Or, if this can wait till Friday or next Monday, I can give you half an hour. What do you want to do?" And what often happens is people solve the problem themselves. Or they wait. And sometimes waiting for a more thoughtful discussion, rather than one in the moment when you're anxious about something, is better. There's a good message there around what we call worry time in psychological care: it's better sometimes to set aside a period when you're going to work through things constructively, rather than being so reactive. So deal with things when they arise, but also give people the time they need to discuss things.