Thermic Effect of Food: What Digestion Costs

Digestion's bill is real, and smaller than the number implies#

Eating costs energy. Chewing, secreting enzymes, absorbing nutrients across the gut wall, converting what arrives into something storable — all of it runs on ATP, and the cost shows up as a measurable rise in metabolic rate for several hours after a meal. That rise is the thermic effect of food, and across a mixed diet it comes to roughly a tenth of what you ate. Protein is the expensive macronutrient to process and fat is the cheap one; the macro accounting behind that ordering sits with the pillar.

The interesting part is what happens when you ask where the tenth came from. In the largest standardized dataset published on the question — 136 older and 141 younger adults studied at the Mayo Clinic across a decade — the measured thermic effect was 7.3 ± 0.3% of meal energy in young adults and 6.4 ± 0.2% in adults aged 60 to 88, or 38.2 versus 32.0 kcal over four hours¹. Thirty-eight calories. Not per day — per meal, measured over four hours, in the group that scored highest.

Those two figures, the tidy tenth and the measured thirty-eight, are not in conflict, and reconciling them is most of what there is to understand about this number. The rest of this article is about where the measurement comes from, why it is one of the least reproducible quantities in nutrition science, and the one property of it that nobody mentions: it shrinks the moment you start dieting.

Four hours captures three-quarters of a six-hour event#

Start with the protocol, because the protocol explains the gap. A thermic-effect test measures resting metabolic rate before a meal, feeds the meal, then tracks metabolic rate under a ventilated hood or inside a respiratory chamber and integrates the area above the pre-meal baseline. The question that has never been settled is how long to keep measuring.

Analyzing 131 such tests across subjects eating meals of varying size and composition, each run for a full six hours, Reed and Hill found that of the total six-hour thermic effect, 60% had been measured after three hours, 78% after four, and 91% after five. Their conclusion was blunt — "an inadequate measurement duration of the TEF could lead to errors" — and they recommended measuring for at least five hours².

Now put the two studies together, because the arithmetic is worth doing explicitly rather than leaving implied. The Mayo figure of 7.3% was integrated over four hours. Reed and Hill's truncation curve says a four-hour window captures about 78% of the six-hour total. Dividing 7.3 by 0.78 gives roughly 9.4% — which lands, unsurprisingly, right on the textbook tenth. That division is mine, not either paper's: the two studies used different meals and different subjects, so it is an illustration of how truncation produces the gap, not a measurement of anything. But it does mean the textbook number and the measured number are the same number seen through different clocks, and that a great many published thermic-effect values are three- and four-hour readings quietly compared with six-hour ones.

The least reproducible number in the room#

Truncation is the tidy problem. The untidy one is that the thermic effect of food is a small difference between two large, wobbly quantities — postprandial metabolic rate minus a fasted baseline — and small differences between large numbers inherit all of both numbers' noise.

A 2026 validation study makes the point about as cleanly as it can be made. Researchers put a 7,209-litre whole-room calorimeter through technical validation with propane combustion and then through biological replicate measurements in ten healthy adults. Technically the instrument is superb: coefficients of variation from 0.23% for respiratory exchange ratio to 1.74% for energy expenditure. Biologically, postprandial energy expenditure repeated with an intraclass correlation of 96% — excellent. The thermic effect derived from those same measurements repeated with an intraclass correlation of 24%, a coefficient of variation of 27.5%, and a minimal detectable change of 43.8%³. The authors also note that reported reproducibility across the existing literature runs from a coefficient of variation of 21% to 68%.

Read the two intraclass correlations side by side, because that contrast is the whole finding. The machine measured the same person's postprandial energy expenditure almost identically twice. The thermic effect calculated from those measurements barely repeated at all. Nothing was wrong with the instrument. The subtraction is what is fragile — and a minimal detectable change of 43.8% means a single study cannot honestly distinguish a thermic effect of 7% from one of 10% in the same person.

This is why the macronutrient percentages you see quoted arrive as bands rather than values, and why they have barely narrowed in three decades. It is not vagueness. It is the measurement.

Why protein is expensive: two different bills#

The reason protein costs more to process than carbohydrate or fat is not one mechanism but two, and separating them explains why the effect is reliable in direction and unreliable in size.

The first is obligatory thermogenesis: the unavoidable ATP cost of doing the work. Reviewing the physiology, Tappy described this component as the "stimulation of adenosine triphosphate (ATP) hydrolysis during intestinal absorption, initial metabolic steps and nutrient storage"⁴. Amino acids are simply more expensive customers here than glucose or fatty acids. Their nitrogen has to be stripped and packaged into urea, they are frequently converted to glucose before use, and building them into body protein costs peptide bonds. Fat, at the other extreme, arrives as a molecule the body already wants to store and can shelve almost as-is.

The second is facultative thermogenesis — the part that is regulated rather than obligatory. Tappy defined it as "the part of glucose induced thermogenesis which is eliminated by beta-adrenergic antagonists," occurring at least partly in skeletal muscle and driven by sympathetic nervous system activation. That is the component that varies with who you are and what state you are in, and Tappy notes it is "reduced in obese, insulin-resistant patients." A number with a regulated component will not behave like a physical constant across people, which is a second reason the published range is wide.

What a high-protein diet actually buys#

The review most often cited for protein's thermic advantage examined 15 randomized controlled trials of thermogenesis in 187 people. Its result is consistent in direction and vague in magnitude: "all 6 studies that assessed the thermic effect of food as a percentage of ingested energy reported a greater energy expenditure for the higher protein versus the lower protein diet," and all three reporting the effect in kilojoules found the higher-protein meals significantly more expensive⁵.

Six out of six and three out of three is about as clean a directional signal as nutrition produces. Two caveats belong beside it. The review's own conclusion was hedged — higher-protein diets "might increase weight loss in the short term, but further longer term research is required." And the independent quality assessment of the review, by the UK's Centre for Reviews and Dissemination, is unusually direct: "the limited search, lack of a validity assessment and poor reporting of review methods mean it is difficult to assess the reliability of these conclusions"⁶. The direction is solid. The size, from this review, is not established.

Here is the size done as arithmetic rather than as a finding. Take a 2,000-calorie day and move 100 g of carbohydrate into 100 g of protein — a large shift, from roughly 20% to 40% of energy as protein. Using the midpoints of the commonly reported bands, 25% for protein against 7.5% for carbohydrate, the 400 calories that changed hands now cost about 100 calories to process instead of 30. That is a gain of roughly 70 calories a day for a substantial dietary overhaul. It is real, it is worth having for free, and it is smaller than the error bar on the measurement that produced it. Protein's genuine contribution to a deficit runs through appetite instead, which is why the case for it is made in protein for weight loss rather than here.

	What is established	What is not
Direction	Protein costs more to process than carbohydrate; fat costs least	—
Size	Roughly a tenth of intake across a mixed diet	Any per-person value; the minimal detectable change is ~44%
Measurement	Needs ≥5 hours; 4 h captures ~78%	A standard protocol — reported CVs span 21–68%
Mechanism	Obligatory ATP cost of absorption, metabolism, storage	How much of the facultative component you personally have

The property nobody mentions: it shrinks when you diet#

Every other component of your energy budget is a property of your body. Your resting rate depends on your organs and lean mass; your activity burn depends on what you do. The thermic effect is the only line that is a fixed fraction of your intake — which means it is the only component that falls automatically the moment you eat less.

The arithmetic is unavoidable, and it is arithmetic rather than a research finding. Cut 500 calories a day from a 2,500-calorie diet and you have also removed roughly 10% of those 500 — about 50 calories — from the expenditure side. Your deficit is not 500. It is about 450 before any other adaptation has occurred, and this one is not adaptation at all; it is definitional. It arrives on day one, it is permanent for as long as you eat less, and it is a real part of why measured expenditure falls further than a calculator predicts during weight loss and why a total-expenditure estimate is a moving target rather than a constant.

Which is the correct way to hold this whole topic. The thermic effect is not a lever — you cannot raise it without eating more, and eating more raises intake faster than it raises the tax on intake, an accounting trap that also sinks every "metabolism-boosting food" claim. It is a bookkeeping line: a tenth off the top, worth understanding so that you know why the calories you eat and the calories you keep were never the same number, and why the gap between them is smaller than the label's own error bars.

FAQ#

How many calories does digesting food actually burn?#

About a tenth of what you eat across a mixed diet. In a Mayo Clinic sample of 227 adults, a standardized meal produced 38.2 ± 1.8 kcal of extra expenditure over four hours in young adults and 32.0 ± 1.3 kcal in adults aged 60 to 88¹. Four hours captures only about 78% of the full six-hour response, so the true per-meal figures are somewhat higher².

Why do published thermic-effect numbers vary so much?#

Because the quantity is a small difference between two large measurements, and it inherits both of their errors. In a whole-room calorimeter that repeated postprandial energy expenditure with 96% intraclass correlation, the thermic effect derived from those same readings repeated at only 24%, with a minimal detectable change of 43.8%³. Measurement duration adds more spread: three-hour tests capture 60% of the response, five-hour tests 91%².

Does a bigger meal have a bigger thermic effect than several small ones?#

Per meal, yes — the response scales with meal size. Per day it makes no difference, because the total is a fraction of total intake rather than of meal count; meal frequency and metabolism works through the controlled evidence. What does change with meal size is the measurement window: larger meals take longer to resolve, so a short test underestimates them more.

Sources#

This article was researched and drafted with AI assistance and reviewed for accuracy by the BurnWeek team. It is general information, not medical advice. How we research and correct our articles →