Do you feed a fever and starve a cold? Or is it the opposite? I can never remember. I never cared to. I thought it was just something people said when they didn’t know what else to say.

Ruslan Medzhitov cares. He’s a distinguished professor of immunobiology at Yale. And in the journal Cell today, his team showed some dramatic benefits to starving a bacterial illness—but feeding a viral illness.

This is contrary to the standard advice of doctors, which is reflected on the bastion of solace, WebMD:

Do you starve a cold and feed a fever when you're feeling under the weather? … Good news—starving is never the correct answer. When you eat a nutritional, well-balanced diet, many other factors fall in place that keep your body functioning optimally.

It’s true that the fever distinction is useless. Both bacteria and viruses can give us fevers. And the holistic benefits of a well-balanced diet are tough to overstate. But in cases of infectious disease, the rule that Medzhitov has discovered seems to have merit. As he first put it, “Starve a bacterial infection and stuff a viral infection.”

To be more precise, we do not feed or starve the bacteria or viruses themselves, but we may be able to modulate the different types of inflammation that these infections cause. “I want to be cautious here not to oversimplify and generalize,” Medzhitov warned. The most accurate and compelling way to put it, then, really: “Fasting has opposite consequences in different types of inflammation.”

And by opposite, he means opposite, life and death. In that way, his new findings could change not just the way we eat when we come down with a common cold, but how doctors treat the end stages of infections—when they spread throughout the blood and becomes known as sepsis, a condition that kills thousands of people every year.

* * *

It all starts with the idea that losing your appetite is a symptom of a lot of illnesses. Why? Wouldn’t it be best to fortify ourselves with all the nutrients we can?

A temporary loss of appetite is known to doctors as anorexia (not to be confused with anorexia nervosa). Medzhitov counts anorexia alongside other common but rather mysterious “sickness behaviors” including altered sleep patterns, depression, and social withdrawal. He has been fascinated with how these things might contribute to our survival. Should they be embraced—listen to what your body is telling you—or fought?

A popular idea behind temporary anorexia is that it seems to happen for a reason: it protects us against certain infections. We alter our metabolisms to deprive our invasive species of fuel. As nutrients and minerals become scarce, the infectious organism will starve before our bodies do. It’s a race to the bottom.

Consider the classic food-poisoning bacterium Listeria monocytogenes. Medzhitov’s team infected a bunch of mice with Listeria, and, predictably, the mice stopped eating. They eventually recovered. But when the researchers fed the mice the same food—force-feeding them, as they had no appetites—they died.

Why, exactly, would that happen? Was some specific element in the food keeping the infection alive?

To figure that out, the Yale team broke down the food by macronutrients (fats, lipids, and carbohydrates). And, indeed, it seemed that the mice could survive the illness when they were forcibly fed proteins or fats. What they couldn’t take was the sugar glucose.

(Glucose comes to us not just by way of what we traditionally think of as sugar, but from any starchy food like bagels or crackers where carbohydrate chains break down into glucose in our mouths and stomachs.)

To double-check that these negative effects of sugar were real, the researchers fed glucose to some mice and then administered a rescue drug (2-deoxy-D-glucose) that blocks the body’s ability to metabolize that glucose—and they survived. In the case of this infection, it could seem, the bacteria need to be starved of sugar.

This is the sort of misguided thinking that leads people to go on cleanses, to deprive themselves of food unnecessarily, and to risk making things even worse. The story is much bigger than avoiding sugar. In other diseases, glucose seems to be beneficial. Critical even.

When Medzhitov infected the mice with the influenza (flu) virus, the mice were more likely to survive if they were force fed. Denying them food—especially glucose, either by withholding it or administering the antagonist 2-deoxy-D-glucose—caused the mice to die. As the researchers write in the journal, in influenza infection, “inhibition of glucose utilization is lethal.” Whereas glucose was “required for survival in models of viral inflammation, it was lethal in models of bacterial inflammation.”

How could that be?

The mechanism doesn’t seem to have anything to do with starving the infectious agent. Rather, it has to do with modulating our own responses to the infections. Here we are dealing with two very different types of inflammation. In one case, glucose exacerbates inflammation. In the other, it is critical to survival.

* * *

When you stop eating, the body starts using fat reserves for calories. Keep fasting, and you start to convert some of that fat into ketones. That switch from burning glucose to burning fat and generating ketones (ketogenesis) is commonly referred to as moving to a “fasting metabolism.”

This switch seems to be important—necessary, even—in the body’s response to bacterial sepsis. Ketogenesis limits the body's formation of substances known as reactive oxygen species, which can damage cells. When you introduce glucose (as in, if you eat sugar, or any carbohydrate that breaks down into sugar), that switch to a fasting metabolism is undone. The sugar triggers the release of the hormone insulin, which tells the body that we don't need to use our fat reserves, bringing ketogenesis to a grinding halt. So when the mice were given glucose, the inflammatory process caused damage to neurons in the brain, causing the mice to convulse and die.

Meanwhile the virus triggered a different type of sepsis, in which removing glucose was uniformly lethal. In that case, glucose seemed to be necessary for adapting to the stress of viral inflammation, by preventing stress-mediated apoptosis (cell death). Without that, an area in the brainstem was destroyed by inflammation, and the mice would stop breathing.

Together, these mechanisms reframe the typical approach to treating infectious disease. Medzhitov’s manipulations weren't focused on killing the pathogen, per se, or starving it, but on using nutrition to modulate the host’s responses.

But would the same thing work in people? Could we eat to acutely modify our immune responses? Could we take 2-deoxy-D-glucose to temporarily block metabolism of glucose?

Medzhitov thinks the idea “very exciting,” especially because sepsis is an intractable condition. The only known treatment is antibiotics and mechanistic measures to keeping a person’s heart beating until the process abates. It’s unpredictable. So the Yale team plan to move expeditiously to testing this in human clinical trials.

Reducing caloric intake has been tested in people with sepsis before, but the results have been mixed. As a consequence, the recommendation in critical-care medicine is to keep people on a balanced diet. (As one Stanford School of Medicine guide directs physicians, “Nutrition, whether enteral or parenteral [via the gut or into the veins], should not be neglected given the high metabolic demands of the septic patient.”)

But Medzhitov thinks the reason the clinical results of fasting during sepsis have been mixed is that patients weren’t separated based on whether their inflammation was the result of a bacterium or virus: “Hopefully if we divide patients based on the cause of sepsis,” he said, “that could provide a way to manage this terrible condition.”

In principle, one day a doctor could give a diagnosis along with a specific dietary recommendation. That could speed recovery and limit the global crisis of antibiotic overuse. It might even be—I hesitate to say in the middle of a diabetes epidemic—an excuse to eat sugar.

This emphasizes the critical point that carbohydrates, like the other macronutrients, are not simply good or bad. Despite whatever diet fad comes about next week, or the week after. As the Yale researchers conclude, their work “implicates a differential need for metabolic fuels as a function of infection.” That is, as more research accumulates in this area, it adds to the understanding that we do well to count food as medicine.

In the immediate term, Medzhitov’s takeaway is to follow our cravings when we’re sick, because they may reflect evolutionary mechanisms that evolved to be protective: “So, for example, when you have the flu, you kind of feel like having some tea and honey. That may be the body’s way of telling us that we need some glucose. I suspect we have these mechanisms that tell us what we prefer to eat (or not to eat) when we’re sick. Those are the mechanisms we should probably listen to.”