Heart Tracings Reveal Sleep Patterns for Health and Disease
Stable and Unstable Sleep Found Distinct from Traditional Non-REM Stages
During the most restful sleep, the resting heart speeds up and slows down
slightly with each breath in and out. But when the heart rhythm drops out
of sync with breath-to-breath respiration, slumber becomes more fitful and
Photo by Graham Ramsay
An electrocardiogram could provide more information on sleep stability and
quality than a dozen monitoring devices, say (clockwise from top left) Chung-Kang
Peng, Robert Thomas, Joseph Mietus, and Ary Goldberger.
Normally, a person must be wired and monitored from head to toe to
assess sleep quality. An overnight sleep evaluation continuously tracks
eye movements, snoring, leg twitches, teeth grinding, and much more. Now,
just one of the many tracings, a simple continuous electrocardiogram (ECG),
may be able to do the job of a suite of independent instruments, researchers
at Beth Israel Deaconess Medical Center report in the September Sleep.
Efforts Wake Science to Sleep Classification
Almost 40 years ago, a dozen sleep scientists grew concerned about
the reliability of the way they scored stages of sleep in their
research. One weekend in California, they met to refine a set
of standards for classifying the cycle of sleep as wakefulness, rapid
A major sticking point came in discussions about how to further
classify the obvious variations in non-REM sleep. At one point
during the extended
debate, Allan Rechtschaffen from the University of Chicago, who
co-authored the resulting manual, locked the door and said he would
not let them leave
until they could come to a consensus.
It was an empty threat, but
it worked. They divided non-REM sleep into four stages known
as arousal thresholds, which roughly correlate
with the amount of shouting and shaking required to wake someone
The manual was published in 1968, providing an important tool for
modern sleep research. The rapidly expanding field of sleep medicine
also relied on the manual, even though the supporting data was
mostly based on
observations of healthy people.
“It gave people a language and a tool with which they could compare
observations,” said Wolfgang Schmidt-Novarro at the Sleep Medicine Institute
at Presbyterian Hospital of Dallas, “but it doesn’t do a particularly
good job of providing insights or parameters for diagnosing disease.”
For the first time, the manual is being revised to include more
clinically relevant standards, such as the addition of respiratory
and cardiac criteria.
Extensive evidence from people with sleep disorders will underlie
some of the changes. Where sufficient evidence does not exist,
the reviewers will
revert to consensus.
The staging of non-REM sleep has remained problematic. The reworked
manual is not due to be finished until summer 2006, but it appears
that scoring of non-REM sleep may be compressed into three stages,
said Conrad Iber of
the University of Minnesota, who is heading the process for the
American Academy of Sleep Medicine.
Digital analyses, such as the new study by Robert Thomas and his
colleagues, will be considered, but the studies are probably too
preliminary to be included in the new manual, Iber said.
“It’s missed that boat,” said Thomas. “Maybe
in the next reclassification. ”
The new standards will undergo review
every five years, said Iber. In the meantime, Thomas has integrated
the new knowledge into his
clinical practice by more precisely tuning treatment to enhance
stable sleep. And he
and his colleagues have launched new studies to evaluate clinical
correlates of stable and unstable sleep.
“This is a distal but clean biomarker that tells us if the system
is oscillating in synchrony with each breath or over multiple breaths,” said
first author Robert Thomas, head of the BID sleep laboratory and HMS instructor
in medicine. “This reflects stable and unstable sleep behavior. Disease
expands the unstable behavior of the system. The goal of treatment is to
enhance the stable behavior.”
If validated by further studies, the
ECG as a measure of sleep stability may be an easier and less expensive
way of diagnosing and guiding the therapy
of sleep disorders.
The work also lends credence to a nonconventional way
of thinking about the stages that make up most of a good night’s sleep,
known as non–rapid-eye-movement
(non-REM) sleep. The researchers’ ECG analysis revealed two states
of non-REM sleep, stable and unstable. In contrast, the traditional staging
system divides non-REM sleep into four grades ranging from light to deep
sleep, which correlate with the effort needed to wake someone up. The sleep
staging standards are now being reevaluated (see sidebar).
Heart of Sleep
In overnight evaluations, multiple lines of data spike and plunge across
a large computer screen in real time as the seconds and minutes go by.
When one measurement changes—reflecting a gasp, a snore, or a shift in body
position—most of the other markers change in synchrony. “A sleep
study is like an orchestral score of the music of the sleeping body,” Thomas
“The breathing and heart rate control turn out to have a profound
connection to what’s going on in the brain during sleep,” said senior author
Ary Goldberger, director of the Margret and H.A. Rey Institute for Nonlinear
Dynamics in Medicine at BID and an HMS professor of medicine. “It ratifies
the growing consensus of the importance of cross talk and systems biology.
You have one set of conversations going on between the heart and lungs and
nervous system in health. In pathology, the frequency and tone of that conversation
literally changes and you see a new conversation emerge.”
to wonder about non-REM sleep about five years ago. He was becoming both
intrigued and frustrated by spontaneous flips back and forth between
stable and unstable patterns of sleep in patient after patient. Bad sleep
could suddenly change to good and vice versa without any intervention and
while remaining in the same grade of non-REM sleep.
Another Look at Non-REM
Brain waves are the gold standard in sleep medicine. Thomas noticed that
the electroencephalogram patterns of these Jekyll-and-Hyde sleep patterns
had been described by a group of Italian researchers as cyclic alternating
patterns (CAP) and non-cyclic alternating patterns (non-CAP). In several
published studies, Thomas correlated the CAP/non-CAP to measurements of
sleep quality and sleep disorders. Seeking an independent validation,
to Goldberger’s group.
Meanwhile, co-author Joseph Mietus, a BID bioengineer,
had shelved an initially disappointing algorithm he had devised to deconstruct
a single jagged ECG
plot to show the link between the heart rate and breathing dynamics. He
had hoped it would help diagnose sleep apnea, but he could not correlate
results with the classic non-REM sleep staging standards. Colleague and
co-author Chung-Kang “C.K.” Peng, a statistical physicist and co-director
of the Rey lab, thought the approach had promise and urged him to pursue
When Thomas first asked them for a mathematical way to distinguish
between CAP and non-CAP sleep in his sleep study datasets, Goldberger,
Peng were daunted by what they envisioned as an insolvable problem.
“You have one set of conversations going on between the heart
and lungs and nervous system in health. In pathology, the frequency
and tone of that conversation literally changes and you see a new
But because Goldberger directs PhysioNet, a research resource funded by
the National Institutes of Health, his group felt a responsibility to assist
Thomas, an NIH-funded researcher. So Mietus ran a sample dataset through
his algorithm, expecting it would probably be a dead end.
“It turned out to be enormously exciting,” Goldberger said. They refined
the technique with 70 sleep studies on patients from BID and other accredited
sleep centers. The stable and unstable sleep patterns overlapped, but were
not identical to, the CAP/non-CAP brain wave patterns. When they retested
the trained algorithm on data from 15 healthy people who were part of a different
study, they discovered that stable and non-stable sleep were a feature of
normal sleep in healthy individuals.
The landscape of sleep. The spectrogram above,
created by Joseph Mietus, demonstrates the difference between stable
and unstable sleep. This nocturnal
electrocardiogram of a healthy man, 24, shows normal fluctuations between
stable (top range) and unstable (bottom range) sleep throughout the night.
The top mountain range nearly disappears in people with uncorrected sleep
disorders. Stable and unstable sleep patterns appear to be independent
of classic sleep staging for the same person, shown in the upper bar:
wake, rapid eye movement (REM), and non-REM stages 1 through 4.
(Courtesy Joseph Mietus)
Mietus devised a quick way to visualize the results using a “sleep
spectrogram” with two distinct mountain-range bands. Healthy people
show more stable sleep; people with untreated sleep disorders show more unstable
sleep. The stable and unstable sleep patterns do not correlate with conventional
non-REM sleep staging, suggesting a complementary new view of sleep regulation
“We are not proposing a new sleep classification system,” Thomas said. “We’re
saying this is how non-REM sleep works. The field can decide what to
do with the new information and how to use this new tool.”