Source: Wired, Adam Rogers
The symptoms of post-traumatic stress disorder include uncontrolled memories of a traumatic event, anxiety and panic—“hyperarousal” is the technical term—depression, avoiding anything that’s a reminder of the event, self-destructive behavior, and more. It’s the only psychiatric disorder where people are pretty sure of the cause: emotionally traumatic events, from the death of a loved one to an experience of fear or violence. By some estimates 5 to 10 percent of all US adults have PTSD, more women than men. Wars in Afghanistan, Iraq, and other places US troops are deployed put those numbers even higher among people in the military and veterans.
But the biology of PTSD—neurological changes, elevated or depressed levels of something a blood test could pick up, genetic vulnerabilities—is … multifactorial. Difficult to correlate. Under investigation.
Which is to say, nobody’s really sure. So this week, a long-in-the-works research study called Aurora is starting to recruit human subjects. Working with 19 hospitals around the country, Aurora will ask 5,000 people to become part of the study. The goal is to try to figure out what biomarkers connect a traumatic event to the development and eventual diagnosis of PTSD.
The researchers at the University of North Carolina and Harvard who are running the study have turned to Verily, the health-focused Google spinoff, for help with data collection and management. That means Aurora will test more than how and why people get PTSD. It’ll also test how the methods used to ask scientific questions invariably constrain the answers.
Aurora’s design is, without doubt, slick. When eligible people come to an emergency room at one of Aurora’s 19 partner institutions, caregivers will try to sign them up. People who agree will, within hours or days of the inciting event, get a baseline biologic and psychiatric assessment, and they’ll return for more tests every month. They’ll also get a Verily Study Watch, a wearable that captures data like heart rate, skin electrical conductivity, and movement.
If all goes as planned, participants will also get an experimental app on their smartphones from the health startup Mindstrong. By monitoring things like keystroke speed and pressure, speed of word choice while texting, and even simple time on-screen—invisibly, in the background—the Mindstrong algorithms are supposed to be able to pick up early signs and symptoms of psychiatric disorders. It’s the “digital phenotype” of an illness.
Right now, no one knows which people who experience trauma are going to develop PTSD, much less how to prevent it or what the underlying biology is. “So we’ll get people right after the traumatic event, and then see if six months later they develop one of these outcomes, and then we’ll see what’s changed,” says Sam McLean, director of the Institute for Trauma Recovery at UNC-Chapel Hill and Aurora’s principal investigator. “Developing better treatments depends on getting a much deeper, more accurate understanding of what’s going on.”
A “biomarker” used to be something that only a blood test or imaging like an MRI would find—a physical manifestation of an illness or injury. Aurora expands that definition. “Historically in studies you’d just get snippets of data. Come in for an assessment, fill out forms, take some physiological data. This is getting data in real time,” McLean says. That means they can look for patterns. Does a person panic every time they leave the house? In that case, “the biomarker is some combination of GPS and, from the watch, seeing your heart rate spike.”
With 5,000 people spewing 24/7 data, you can see why partnering with Verily makes sense. Aurora got $21 million from the National Institutes of Health to spin up; that’s not a lot for a big, multicenter, prospective study. So they can take advantage of Verily’s Google-spawned data-handling heritage, “keeping track of the data and ensuring it’s intact, secure, and well-organized,” says Menachem Fromer, Verily’s mental health science lead, and using Verily’s skill with analytics and machine learning. Aurora also gets to take advantage of Mindstrong’s clever data collection and interpretation algorithms.
Meanwhile Verily and Mindstrong both get to play with the Aurora database to validate and improve their own work. (Verily also has a separate effort, Project Baseline, that will use its watch to monitor 10,000 otherwise healthy people.) “Validation, while it seems straightforward, is actually a really hairy problem,” says Tom Insel, a neuroscientist, ex-director of the National Institute of Mental Health, and currently a partner at Mindstrong. “We’ve got a novel, objective measure of mood and cognition. So how do we prove that?”
In fact, that’s exactly what Insel tried to codify when he was head of NIMH, where he championed finding those more objective measures. As Insel himself acknowledged, it didn’t quite work—and the attempt polarized the mental health community. Insel left NIMH in 2015 for what was then called Google Life Sciences … now known as Verily. Then in May he left Verily for Mindstrong.
Both companies hope to help people, and also sell products—not that there’s anything wrong with that. But explicitly looking for PTSD biomarkers that might provide therapeutic hooks could also pose a forest/trees problem, because mental health in general doesn’t have clearly-defined objective measures of its disorders. “That has made the hunt for biomarkers particularly difficult,” says epidemiologist Sandro Galea, dean of the Boston University School of Public Health and chair of a committee that produced an Institute of Medicine report on PTSD in 2014, “You can have PTSD in many different ways,” Galea says, “because the diagnostic criteria are essentially a mix of different symptoms and categories.”
Big data approaches like what Verily and Mindstrong bring to Aurora have a chance to unify the belts, as it were, collecting huge amounts of information and then sorting through it to try to quantify the qualitative standards. It could be a powerful approach, says Galea, and may well add to knowledge about and the potential to treat PTSD, “but it also introduces a distorting force. Not the data itself—data is value-neutral. But the data applications have what the companies hope are clear commercial approaches.”
In other words, if you go looking for biomarkers, you are actually looking for therapeutic targets, and while those targets might be hittable by a drug, they might also oversimplify the actual disorder. It’s a step away from holistic theories that try to encompass socioeconomic or public health-style approaches—because, fundamentally, it’s easier to sell a drug.
Verily’s people understand this potential pitfall, of course. “The best information, to a person living in a social environment that is not that healthy, may not be that useful,” former FDA commissioner Robert Califf, now at Verily, said at a recent biotech conference—not talking about PTSD specifically, but health in general. “Social determinants far outweigh the biological determinants.”
Aurora’s investigators know that working with big-data companies means they have to strike a balance, too. And it’s worth it. “They have been and are being terrific,” McLean says. “Personally, as PI, if some other entity maybe wouldn’t do direct patient assessment but had another way to analyze data or to bring discoveries faster, then hey, I’m totally game.” Right now, Aurora, Verily, and Mindstrong have aligned but not equivalent reasons for wanting to bridge the cause-to-symptom gap for PTSD. For people with the disorder, maybe those reasons don’t matter, as long as they get results.