USING BREATH TO DIVINE the inner workings of the human body is a practice as ancient as Hippocrates, who described the fetor hepaticus of liver failure that manifested itself in the fishy odor of patients’ exhalations. Other revelations from breath, just as distinctive, are also well known to clinicians: the fruity smell of acetone in uncontrolled diabetes, the freshly baked–bread vapors of typhoid fever, the ammonia reek of kidney disease. And although breath as a clue to disease was long ago replaced by more precise diagnostic tests, scientists have recently returned to breath’s rich repository of chemicals—about 3,000 compounds have been measured so far—to look for markers that could indicate the presence of a particular disease or gauge how well a patient is responding to treatment.

Breath has several potential advantages as an analytic tool. Unlike blood and urine, it can be sampled frequently and noninvasively, it’s never in short supply, it generates little or no infectious wastes, and results could be produced in real time using newer technology that can collect breath and parse its compounds instantly. Breath tests might be used, for example, for fast, inexpensive screening of large populations for exposure to a flu strain. And breath may hold a surprising trove of information in addition to the expected gases from the environment and metabolism. “We can collect microdroplets of water and particles to give us access to substances we could find only in blood and urine, such as fragments of DNA, messenger proteins, inflammatory cytokines, bacteria and cellular material,” says Joachim Pleil, research physical scientist at the Environmental Protection Agency, who studies molecules in breath to distinguish between those emanating from human physiology and inhaled compounds in the environment.

At June’s International Conference on Breath Research in Germany, scientists presented research on breath tests for Parkinson’s disease, bacterial infections in cystic fibrosis, cancer, diabetes and other conditions. During the past two decades, thousands of articles have been published on breath’s potential biomarkers. But in this young field, researchers have often had to depend on jerry-rigged instruments and ad hoc methods to collect and analyze breath, and the quality of much of the work has been suspect. Because people wrongly believe it’s easy to collect breath samples, there have been a lot of bad studies, and that has led many scientists to dismiss breath research as fringe science, according to Terence Risby, professor emeritus of environmental health sciences at Johns Hopkins University.

Now, however, that attitude may be changing, thanks in part to more accurate, commercially available instruments that collect breath as well as to advanced statistical algorithms that enable researchers to find associations between what they are measuring and clinical outcomes. “The instruments we use to analyze breath can detect compounds in our breath in concentrations measured in parts per billion,” says Raed Dweik, professor of medicine and director of the pulmonary vascular program at the Cleveland Clinic. “That’s the equivalent of finding one red Ping-Pong ball in a baseball stadium full of white Ping-Pong balls.” Such sensitivity is crucial, because abnormal physiology may announce itself through extremely subtle fluctuations in the normal range of a compound in breath.

Despite recent progress, only a handful of breath tests have made their way into common use. But researchers have been energized by the widespread acceptance of a breath test that measures concentrations of nitric oxide to reveal airway inflammation and show whether a patient needs corticosteroid treatment and is likely to respond to the therapy. Dweik notes that testing blood only gradually became standard practice, “and there’s still much we don’t understand about blood,” he says. He believes breath analysis could eventually ascend to prominence as a new frontier in medical testing.

A GUINEA PIG WAS THE SUBJECT of the first breath test, conducted in the early 1780s by French scientists Antoine-Laurent Lavoisier and Pierre Simon Laplace. The experiment revealed that animals exhale CO2—the result, the men surmised, of oxygen combustion in the lungs. A century later, the British physician Francis Anstie first isolated ethanol from breath and came up with Anstie’s limit, which held that it was safe to consume up to 1.5 ounces of pure ethanol daily. That led to the invention of the Drunkometer in 1931, an instrument that tested breath for alcohol, which was followed by the Breathalyzer in 1953.

Today’s most widely used clinical breath test also got its start decades ago. Since the 1970s, anesthesiologists have used capnographs, medical devices that measure the concentration of carbon dioxide in the breath of patients undergoing surgery or who have breathing tubes inserted in intensive care. The correct amount of CO2 in breath indicates that the endotracheal tube is correctly positioned to ventilate the lung (not the stomach) and shows that vital cellular metabolism is taking place.

The advent of gas and liquid chromatography and mass spectrometry—instruments that can separate gases and liquids into their many trace components—got breath analysis going in earnest. In 1971, Nobel laureate Linus Pauling used gas chromatography to analyze breath and found, beyond the usual nitrogen, oxygen, carbon dioxide and water vapor, an additional 250 compounds. That seminal study “was proof of concept that exhaled compounds may include biomarkers that indicate disease,” says Cristina Davis, director of the Bioinstrumentation and BioMEMS Laboratory at the University of California, Davis. Three years later, scientists discovered that a high concentration of ethane in breath indicated elevated oxidative stress, which can create a dangerous imbalance in cell metabolism. Administering antioxidants, in turn, reduced the oxidation of fatty acids in the body—and the amount of ethane in breath. “That was the first study to show that you could do an intervention based on breath analysis,” says Risby.

As cells consume energy and protect themselves against stress, they release metabolites—byproducts of those biochemical processes—into the bloodstream in the form of volatile organic compounds (VOCs). Blood carries the VOCs into the lungs, where they diffuse from capillary walls and across a thin pulmonary membrane to mix with the air that is exhausted from the body as breath. When disease produces abnormal cell metabolism, that may show up in breath in the form of altered metabolites. VOCs are also created and emitted in breath when the body metabolizes drugs of any kind, therapeutic or street.

To find biomarkers in breath, researchers can use either of two main analytic methods. The most accurate—and most costly, requiring instruments that may run $500,000 or more—is gas chromatography combined with mass spectrometry. It’s highly sensitive and can detect trace amounts of any chemical present in breath. But that can be too much information if you don’t know what you’re looking for. The machines are also slow, taking 20 to 40 minutes to analyze a breath sample.

A sensor array known as an electronic nose is much simpler, though it can provide information only about the distribution of compounds in breath; it doesn’t identify them. Still, electronic noses are much closer to the point-of-care breath tests that clinicians may ultimately use, says Peter Mazzone, director of the Lung Cancer Program and Pulmonary Rehabilitation Program at the Cleveland Clinic. In his research on the breath biomarkers of lung cancer, Mazzone uses a sensor that changes color as it reacts to different chemicals. “The pattern is the biomarker,” says Mazzone. “If the sensor is accurate in identifying the pattern, it’s not necessary to know which compounds you’re measuring.”


ANOTHER CRUCIAL ISSUE in refining breath research is the potential “contamination” of breath samples by molecules from outside and inside the body. “Breath contains compounds from the air you inhaled as long as a week ago, plus what you’ve eaten, the drugs you take and the microbes that have entered your bloodstream from your skin,” says Risby. He recalls how excited he was to discover a unique molecule in the breath of a patient, only to find out that it was Freon she’d inhaled in her car, produced by a leaking air conditioner. To try to sidestep that difficulty, some researchers measure the compounds in room air and subtract those from the breath sample of the subject in that room, or they take samples by having people blow through filters. Yet those approaches ignore important information about where someone has been and that person’s habits and diet, Risby says.

Moreover, no matter how well you control testing variables, it can be hard to tease out the origin of a particular molecule. For instance, there could be several reasons for ammonia levels in breath to be elevated. It could be because of abnormal protein metabolism or a sign of possible renal disease, or there could be a problem with how the liver is converting ammonia to urea. But normal bacteria in the mouth could also lead to an excess of ammonia, as could recent intense exercise.

It also can be tough to trace a line between abnormal physiology and compounds found in breath. It would be convenient if the advent of a disease created distinct, measurable substances. But cancer, for example, doesn’t produce new biochemistry, says Risby. Rather, it modifies the existing concentration of molecules. And it’s often not clear whether the concentration of a particular compound in breath is normal.

“There’s a very broad normal distribution of certain compounds such as acetone in the breath of healthy people, and that range overlaps with the range in those who have diabetes,” says Matthias Frank, leader of the Advanced Instrumentation and Diagnostics Group at Lawrence Livermore National Laboratory. “So where do you draw the line for diagnostic purposes?” In many cases, there’s no clear answer yet.

Bacteria, fungi and viruses, on the other hand, may produce some distinctive metabolites not normally found in the body. In one study, Pleil at the EPA is working with researchers at the University of North Carolina School of Medicine to find the “breath print” of bacterial infections that can be life threatening for people with cystic fibrosis. “Opportunistic infections kill people with cystic fibrosis because they can’t clear their lungs,” says Pleil. By growing specific strains of bacteria in the lab and measuring their “headspace”—the gas molecules floating above petri dishes—Pleil hopes to discover the molecules they emit once the bacteria invade the body. The next step would be to develop an at-home breath test to detect trace amounts of bacteria that could be treated before there’s a full-blown infection.

FOR THOSE WHO BELIEVE breath tests could one day become as useful and ubiquitous as blood tests, any progress toward practical applications is welcome. One success, Dweik says, is a test to diagnose whether someone is infected with Helicobacter pylori, the bacterium that causes stomach and intestinal ulcers. A patient swallows a solution containing urea labeled with carbon 13, a non-radioactive, stable and relatively harmless isotope. If H. pylori is in the gut, an enzyme it produces, urease, will break down the urea in the stomach, producing CO2 marked with the carbon 13, which is absorbed by the blood and exhaled 15 minutes later. Yet while Dweik considers this tool to be quick and definitive, most gastroenterologists prefer traditional endoscopy and biopsy to diagnose H. pylori.

Similarly, the Heartsbreath test, approved by the Food and Drug Administration, has had difficulty gaining traction in the clinic—in this case, because insurance companies may not pay for it. When the immune system begins to reject a transplanted heart, the heart cells create increased amounts of oxidative stress, which degrade fatty acids and produce an abundance of hydrocarbons called alkanes that are emitted as VOCs. If the breath test doesn’t show an elevated level of alkanes, patients can avoid the 10 to 15 biopsies of heart muscle obtained from a catheter inserted in a neck vein that are typically done in the first year after transplant. “Patients hate those biopsies, but physicians feel compelled to do them to find the very infrequent rejection that can be life threatening,” says Michael Phillips, founder and chief executive of Menssana Research in Newark, N.J. “The result of a Heartsbreath test can guide a physician’s decision on whether to do a biopsy.” But the Centers for Medicare & Medicaid Services has asked for more studies before deciding whether it will cover the cost of the test. Phillips’s company is wrapping up a three-year trial that it hopes will sway the agency.

A pioneer of breath research, Phillips believes tests such as Heartsbreath—designed not to confirm a disease or condition but to rule it out—will be the norm for analyzing breath. “A breath test is not diagnostic by itself; rather, it’s the bottom rung of a ladder leading to diagnosis,” he says. For example, Phillips is also searching for the breath print of lung cancer, and if he succeeds, he expects that the resulting breath test would precede rather than replace the CT scan ordinarily used to screen for lung cancer. “If the breath test was negative, perhaps you could say with 99% certainty that you didn’t have lung cancer,” says Phillips, “whereas those with a positive breath test might get a chest CT scan followed by additional testing.”

But a definitive breath signature for lung cancer has yet to be found. To diagnose cancer, a breath test would have to be at least 95% accurate, and to date, results in the lab have also lagged what has been the gold standard for sniffing out lung cancer in breath: work by specially trained dogs that, in small studies, has achieved up to 99% accuracy. The canines improved with practice, and Mazzone believes that human diagnosticians will also get better as they examine the breath signatures of more lung cancer patients.

Beyond its potential for screening and diagnosis, breath may emerge as a tool for planning or monitoring treatment. One current research effort involves a breath test that could determine whether a particular cancer patient can metabolize the chemotherapeutic agent 5-fluorouracil. The patient swallows uracil labeled with carbon 13. Because C13-labeled uracil is metabolized via the same pathway as 5-fluorouracil, a high concentration of CO2 marked with carbon 13 in a subsequent breath sample indicates that the patient should be able to tolerate the drug, says Anton Amann, director of the Breath Research Institute of the Austrian Academy of Sciences and editor in chief of the Journal of Breath Research.

Such work heartens the scientists involved in this nascent field. And while they have yet to identify as many useful biomarkers as they may have hoped, they have observed sufficiently significant differences in the breath of healthy and sick people to be convinced that many more breath signatures exist, says Dweik. “At our breath meetings, the idea that most physicians reject the validity of breath tests comes up all the time,” he says. “To get over that hump, we need to prove that breath tests are not only faster and easier but also better than existing diagnostic tests. Then physicians will use them.”