36 THE SCIENTIST | the-scientist.com
© ISTOCK.COM, MATSABEwith Mann to investigate the suicide-
inflammation link. These pathways
haven’t “been interrogated enough,”
Melhem says. “We’ ll be looking more
into that in future work.”Predicting suicide
One of the defining moments in psychia-
trist David Brent’s career happened dur-
ing his medical residency some 40 years
ago. Brent had been assigned to work
with young people admitted for inten-
tional drug overdoses at the Univer-
sity of Pittsburgh Medical Center Chil-
dren’s Hospital. He had to determine
who should be referred to a psychiat-
ric ward and who could safely go home.
“I found that I really didn’t have a very
good way of making that determination,”
says Brent, now a professor at Pitt. As he
learned more about how other clinicians
made such decisions, “I realized I was in
good company—that nobody really knew
what they were doing.”
It’s still a dilemma facing anyone
attempting to provide care for people at
risk of suicide. Today’s clinicians often
rely on patients to report their inten-
tions in order to decide on appropriate
interventions. But the approach has
limitations. One 2019 meta-analysis of
studies on suicidal ideation found that
around 60 percent of people who ended
their lives had denied having suicidal
thoughts when asked by a clinician or
doctor in the weeks or months before
their death.^15
This problem has led some research-
ers to look for ways to translate findings
from neurobiology, however prelim-
inary, into the identification of bio-
markers to predict the onset of suicidal
behaviors. Given its strong association
with suicide, the H PA axis has long
been a focus of this work, and there’s
some evidence that abnormal cortisol
levels—higher or lower than normal—
in blood or saliva could hold prom-
ise as a biomarker. A few months ago,
for example, Melhem, Brent, and col-
leagues published findings from a long-
term study of teenagers that suggested
a person’s baseline cortisol levels mightbe used to predict future suicidal think-
ing, with higher cortisol associated with
increased ideation within the next cou-
ple years.^16
Cortisol tests may help provide pre-
dictive power to other measures of sui-
cidality, such as questionnaires about
social and academic stress. One recent
analysis showed that while survey data
were good predictors of who among 220
teenage girls with mental health con-
cerns would be thinking about suicide
within the next few months, they were
poor predictors of who would attempt
suicide during that period. But when the
researchers focused only on girls who
had shown blunted cortisol responses
in lab tests, the questionnaire data pre-
dicted suicide attempts much better.^17Looking beyond stress responses,
other groups have attempted to identify
biomarkers related to neurotransmis-
sion. A few years ago, Mann’s group used
positron emission tomographic (PET)
imaging to assess levels of 5-HT1A sero-
tonin receptors in the midbrains of 100
patients with major depressive disorder.
The scientists found that higher 5-HT1A
levels predicted greater suicidal ide-
ation and more-lethal suicidal behavior
within the next two years.^18 Last sum-
mer, a team led by Yale University neuro-
psychologist Irina Esterlis reported that
levels of glutamate receptor mGluR5,as measured by PET, was linked to cur-
rent suicidal ideation in patients with
post-traumatic stress disorder—though
the results didn’t hold for patients with
major depressive disorder.^19
Opinions differ among researchers
about the potential of such biochemical
signatures to assess suicide risk. Greg
Ordway, a pharmacologist studying
depression at East Tennessee State Uni-
versity, says that while biology might
identify people predisposed to suicidal
behavior, it’s unlikely to produce one or
a handful of biomarkers that reliably
reveal whether a person is about to end
their life. “Suicide is extremely difficult
to predict,” he says. “People are always
trying to do it—people like me are look-
ing for markers. But in reality, I don’t
think we’ll probably ever find that.”
Some of the most promising tools for
assessing immediate risk might instead
come from other areas of neuroscience
that measure more-complex, emotional
signals in the brain as opposed to bio-
chemical signatures. In 2017, Brent,
along with Carnegie Mellon University
neuroscientist Marcel Just and other col-
leagues, used functional MRI to image
the brains of 34 people as they contem-
plated words such as “death,” “trouble,”
and “carefree.” Using machine learn-
ing algorithms to process the data, the
team could distinguish between people
who were thinking about suicide, as self-
reported during the study, and those who
weren’t with 91 percent accuracy. Among
those who were, the team identified peo-
ple who’d already attempted suicide with
94 percent accuracy.^20
The researchers recently received
$3.8 million from the National Institute
of Mental Health to scale up the project
and are planning long-term monitor-
ing of people with and without various
types of mood disorders. As part of the
study, the researchers hope to extend
their tool to identify people who might
attempt suicide in the future, not just
those who are thinking about it at the
time of the scan or who have attempted
it in the past. Just tells The Scientist
that the team also plans to adapt theNEARLY 800,000
PEOPLE WORLDWIDE
DIE BY SUICIDE
EVERY YEAR.
Source: WHO, 2016 dataDIE BY SUICIDE
EVERY YEAR.
Source: WHO, 2016 dataNEARLY
PEOPLE WORLDWIDE800,000
WORLDWIDE800,000