Resting energy expenditure, calorie and protein consumption in critically ill patientsa retrospective cohort study

Picture an intensive care unit bed. A patient on a ventilator is sedated and unable to swallow. Someone has to decide how many calories to pump through a feeding tube. At most hospitals, that decision is made by plugging height, weight, and age into a formula. A formula that has never met this particular patient and cannot know how fast their metabolism is running today, tomorrow, or next week. Oren Zusman and colleagues followed one thousand one hundred seventy-one patients over twelve years and found that this standard approach gets it wrong in ways that cost lives. Their study, published in Critical Care, is the largest cohort to date that measured actual metabolic needs rather than predicted them. What it found should change how every intensivist thinks about the feeding tube. The core problem is that critically ill patients are metabolically unpredictable. Their resting energy expenditure — the calories their body burns just to stay alive — shifts constantly as inflammation rises and falls, as organs fail or recover, and as sedation levels change. Predictive equations, the Harris-Benedict formula and its descendants, give a static estimate from demographic data. Zusman and colleagues describe them plainly as "known to be less accurate and provide only an approximate snapshot of metabolic needs." The alternative is indirect calorimetry, or IC, which is a bedside measurement that tracks oxygen consumption and carbon dioxide production, and then uses those values to calculate true resting energy expenditure. It’s more accurate, more responsive to change, and far less commonly used than it should be. The team at Rabin Medical Center built their study around IC measurements taken on patients admitted between 2003 and 2015 to a single sixteen-bed tertiary intensive care unit. Of nearly seven thousand admissions, they identified one thousand one hundred seventy-one patients who had IC measurements, met inclusion criteria, and stayed long enough to generate meaningful data. The calorimeter was calibrated monthly with ethanol and before each use with test gases. Measurements were only accepted if the patient had been stable for at least thirty minutes and the recording ran for at least twenty minutes. The mean measured resting energy expenditure across the cohort was just under one thousand nine hundred eighty kilocalories per day, but the range ran from eight hundred to over four thousand five hundred, which tells you exactly why a one-size formula fails. The key variable in their analysis was what they call percent administered calories over resting energy expenditure — administered calories divided by measured resting energy expenditure, expressed as a percentage. This is the fraction of actual metabolic need that was actually delivered. They modeled this continuously using a statistical technique called a restricted cubic spline, which allows the relationship between that variable and mortality to curve and bend rather than forcing it to be a straight line. That methodological choice turned out to be crucial. The result was a U-shaped curve. Not a line or a plateau, but a U. Increasing caloric delivery from zero up to seventy percent of measured resting energy expenditure was associated with steadily falling mortality: a hazard ratio of zero point ninety-eight per percentage point, with a confidence interval of zero point ninety-seven to zero point ninety-nine. A hazard ratio below one indicates a lower risk of death. Each additional percentage point of delivered calories, up to that seventy percent mark, was associated with roughly a two percent relative reduction in the instantaneous risk of dying. That's the left side of the U — underfeeding kills. Then the curve turns. Above seventy percent, more calories meant more deaths. The hazard ratio flipped to one point zero one per percentage point, with a confidence interval of one point zero one to one point zero two. Once delivery exceeded one hundred percent of measured resting energy expenditure, the hazard ratio climbed above one and kept climbing. The non-linearity was statistically significant with a p-value below zero point zero zero seven eight, and the association held after adjustment for other variables with a p-value below zero point zero zero six. The model's area under the curve was zero point seventy-five, corrected to zero point seventy-four after internal validation — a solid performance for an observational dataset this complex. That's the central finding: both underfeeding and overfeeding appear harmful, and the optimal target sits at about seventy percent of measured energy expenditure. Not one hundred percent. Not as much as the patient can tolerate. Seventy percent. Protein told a different story. Zusman and colleagues tracked protein delivery separately because its effects may be independent of total caloric load. Higher daily protein was independently associated with lower sixty-day mortality — a hazard ratio of zero point ninety-nine per gram per day, with a p-value of zero point zero one eight. The paper frames this as roughly a one percent relative reduction in mortality for each additional gram of daily protein. More protein, within the ranges studied, appeared protective regardless of where the patient sat on the calorie curve. The secondary outcomes reinforce the overfeeding warning from a different angle. The team divided their cohort into three groups by administered calories over resting energy expenditure: below seventy percent, between seventy and one hundred percent, and above one hundred percent. Among survivors, the median intensive care unit length of stay was twelve days in the lowest group, fifteen days in the middle group, and sixteen point five days in the highest group. Median days on mechanical ventilation were ten, thirteen, and fourteen days for those same groups. Both comparisons were significant at a p-value below zero point zero zero one. More calories wasn't just failing to help — it stretched out the very interventions patients most want to be free of. The authors offer a plausible mechanism: higher caloric delivery may increase metabolic and ventilatory load, making it harder to wean patients off the ventilator. That's speculative, but it fits the pattern. These secondary findings also help resolve an apparent contradiction with prior literature. David Heyland and colleagues had reported that delivering more than two-thirds of prescribed calories was beneficial and suggested above eighty-five percent as optimal. But in that analysis, indirect calorimetry was used in only zero point five percent of cases. The rest relied on predictive equations. Zusman and colleagues argue that when your target is based on an inaccurate estimate of resting energy expenditure, hitting a high percentage of that target tells you very little about whether you're actually meeting metabolic needs. You could be massively overfeeding a patient whose true resting energy expenditure was far lower than the formula predicted, while thinking you're providing excellent nutrition. That's the methodological heart of this paper. The finding that seventy percent of resting energy expenditure is optimal means nothing unless resting energy expenditure is actually measured. A predictive equation might tell you a patient needs two thousand kilocalories when they actually need one thousand four hundred or two thousand six hundred. Giving seventy percent of the wrong number lands you in the wrong place entirely. The precision of the target depends entirely on the precision of the measurement. The study's limitations are real, and the authors are direct about them. It's retrospective, single-center, and observational. Causality cannot be established. What gives the findings unusual credibility is the scale — twelve years, one thousand one hundred seventy-one patients, over five thousand individual indirect calorimetry measurements, repeated for many patients across their intensive care unit stay — and the consistency of the results across multiple sensitivity analyses. For clinicians, the practical implication is clear. Indirect calorimetry is technically demanding and more expensive than running a formula, and most hospitals haven't adopted it as standard practice. This study provides those pushing for its broader use with a concrete, data-backed argument: not that indirect calorimetry is theoretically preferable, but that patients fed to targets based on measured resting energy expenditure appear to survive at higher rates and leave the intensive care unit sooner. Zusman and colleagues call for randomized controlled trials using indirect calorimetry-derived caloric targets as the logical next step. Given that this is the largest indirect calorimetry-based cohort yet assembled, it's a call with real weight behind it. The protein signal is worth carrying forward too. It showed up independently of the caloric curve, suggesting that even as clinicians work to hit precise energy targets, protein delivery deserves its own attention. The two levers are not the same and do not move in lockstep. Feeding a critically ill patient is a medical intervention. Like any intervention, it has a therapeutic window. Zusman and colleagues have used twelve years of measured data to sketch the shape of that window — and it turns out to be narrower and more asymmetric than most clinical practice currently assumes. This lecture was created by ennepō. Go to https://ennepo.ai to Discover, Create and Follow the latest research in your field. Read when you can. Listen when you want to.