Doubly Labeled Water, Explained

Two heavy waters, and the difference between how fast they leave#

Doubly labeled water measures the calories you burn by watching two tracers leave your body at different speeds. You drink a small dose of water in which some hydrogen atoms are the heavier isotope deuterium and some oxygen atoms are the heavier oxygen-18. Over the next one to three weeks you give occasional urine samples, and a mass spectrometer tracks how quickly each tracer disappears. The hydrogen leaves only as water. The oxygen leaves as water and as carbon dioxide, because oxygen in body water swaps freely with the bicarbonate that becomes exhaled CO2. So the oxygen label falls faster, and the difference between the two decay rates, scaled by the size of your body-water pool, lets researchers calculate your carbon dioxide output³.

Carbon dioxide is the exhaust of burning fuel, so knowing how much you produced tells you how much energy you spent. That is the whole trick: no mask, no chamber, no activity diary, no wearable. You live your normal life, and the isotopes keep the books. It is why this one method, more than any other, turned "people eat more than they write down" from a clinician's suspicion into a measured quantity — a story told in full in what the underreporting studies found. This article is about the instrument itself: what it measures, how good it is, and the one thing it cannot see. For where it sits among all the errors in counting calories, start at how accurate calorie counting is.

From four volunteers in 1982 to a routine reference method#

The idea came from animal physiology, and the question in the early 1980s was whether it would work in people at doses anyone could afford. Dale Schoeller and Edward van Santen gave four healthy adults about 10 grams of oxygen-18 and 0.5 grams of deuterium as water, then followed the isotopes' disappearance in urine for 13 days while the volunteers ate a measured diet. The isotope estimate of energy expenditure differed from measured intake plus change in body stores by an average of 2 percent, with a coefficient of variation of 6 percent. The authors' last line is the one that mattered: the method "is noninvasive, and the subjects could maintain their daily activities without restriction"¹.

Six years later Schoeller summarized the state of the field in a review that set the numbers people still quote. Doubly labeled water, he wrote, "is a form of indirect calorimetry" that "measures integral CO2 production for up to 3 wk," and validations against near-continuous respiratory gas exchange showed it "is accurate and has a precision of 2-8% depending on the isotope dose and the length of the elimination period"². A later review of three decades of use reports the same picture against respiration chambers: accuracy within 2 percent, with individual precision in the single digits³.

A method that is off by about 2 percent on average, over two weeks of a normal life, is the closest thing nutrition science has to a referee.

How it compares with everything else that estimates your burn#

The value of the method is easiest to see next to its alternatives. Every other way of estimating daily expenditure either confines you to a room or guesses from a proxy.

Method	What it actually senses	Free-living?	Typical time window	Main weakness
Respiration chamber	Oxygen in, CO2 out, directly	No — a sealed room	Hours to a few days	Your life in a chamber is not your life
Doubly labeled water	CO2 production from isotope decay	Yes	1–3 weeks	Gives one average; no day-by-day detail
Prediction equation	Age, sex, height, weight	Yes	None — a formula	Individual errors are large
Wearable tracker	Motion and heart rate	Yes	Minute by minute	Converts a proxy to calories

The trade is visible in the table. Doubly labeled water gives up resolution — it returns one average number for the whole collection period and cannot tell you what Tuesday cost — in exchange for being both free-living and accurate. That makes it poor for coaching and excellent for checking. It is a reference method for testing TDEE calculators and wrist trackers, and the data behind the finding that total expenditure resists climbing as much as exercise arithmetic predicts (constrained energy expenditure).

Converting CO2 production into energy expenditure is the last step. Turning CO2 into energy needs an assumption about the fuel mix being burned; for an ordinary Western diet, the error from estimating that mix from the diet's composition is small — Westerterp puts it at no more than about ±2 percent³.

The method measures burning. Intake is an inference.#

This is the point most summaries skip, and it changes how every famous finding should be read. Doubly labeled water never looks at food. It measures expenditure. The step from "you burned 2,600 calories a day" to "you must have eaten about 2,600 calories a day" is an assumption: that your body's energy stores did not change over the measurement window.

The authors of the largest analysis of the method state the condition plainly: converting energy expenditure into an estimate of food intake "assumes that individuals are in energy balance over the time course of measurement"⁴. Average intake equals average expenditure plus the daily change in stored body energy, and that change is negative while you are losing tissue. Careful studies therefore measure body composition at both ends and convert the change into energy. The 1982 validation ran the logic the other way, checking the isotopes against measured intake adjusted for changes in body stores.

The classic demonstration involved ten self-described diet-resistant patients whose diaries said they were eating very little. Over 14 days, the group's average total expenditure and resting metabolism were within 5 percent of predicted values, so low expenditure did not explain their reported diet resistance, while their reported intake ran an average of 47 percent below actual intake, and reported activity was overstated by 51 percent⁶. The finding only works because expenditure was measured by something that did not depend on the patients' own reports. That independence is the method's real contribution: it gives self-report something to be wrong against.

What 6,497 measurements added — and the correction that halved the headline#

Individual studies can check individual diaries. The harder problem is population surveys, which collect food records from thousands of people; earlier screening methods had limitations, and a 2025 analysis built a prediction equation from a much larger isotope dataset. Pooling 6,497 measurements from people aged 4 to 96, a large consortium fitted an equation that predicts total expenditure from body weight, height, age, sex, ethnicity and elevation, explaining 69.8 percent of the variation in the log of expenditure, and defined 95 percent limits around each prediction. A reported intake outside those limits for someone of a given size falls outside the model's expected range and is flagged for scrutiny⁴.

Applied to two national surveys — the UK's National Diet and Nutrition Survey and the US NHANES — the screen flagged 27.4 percent of dietary reports as falling outside the expected range. The flags were not neutral either: larger reporting discrepancies went with higher reported protein shares and lower fat shares, and screening changed the associations between diet and body weight.

That 27.4 percent figure has a history worth knowing. The paper as first published said misreporting exceeded 50 percent. Months later the authors issued a correction: the equation's output was in megajoules, the surveys' intakes were in kilojoules, and the wrong unit had been inserted into the calculation, overstating underreporting. The corrected level is 27.4 percent, and tables, figures and supplementary material were revised⁵.

Two things follow. First, the accurate reading is that roughly one in four survey reports is flagged — common, but not the majority, and a flag is a reason for scrutiny rather than proof of error: by construction, about one valid observation in twenty falls outside 95 percent limits, and ordinary short-term swings in intake contribute too. Second, the window is wide: it only catches reports far off the prediction for someone of that size. A diary that is 15 percent low can sit comfortably inside the limits. The screen finds the large errors; it cannot certify the rest as correct.

What this means if you are the one keeping a log#

The method requires a specialist testing service, and an ordinary tracker does not need one. What the method offers an ordinary tracker is a principle rather than a measurement. It shows that self-reported intake and actual intake are different quantities, that the gap is measurable, and that it is a property of the tool — pen, app, memory — more than a character flaw. It also shows what the reference standard for your own intake looks like: not a better-kept diary, but an independent measure of the other side of the ledger.

Your weight trend over several weeks offers context in the same spirit, though it is not an expenditure measurement: water shifts and body-composition changes blur it. If your log promises a deficit and the trend stays flat for a month, a persistent mismatch warrants checking both your intake log and your expenditure estimate; the trend alone does not say which one is off. The practical troubleshooting lives in why you might not be losing weight.

FAQ#

How does a doubly labeled water test actually work for the person taking it?#

You drink a measured dose of isotope-labeled water, give a baseline urine sample beforehand and a few samples over the following one to three weeks, and otherwise live normally. There is no mask, chamber or activity tracker. The laboratory then calculates your average daily carbon dioxide production, and from it your average daily energy expenditure.

How precise is a single doubly labeled water measurement?#

Against respiration chambers, the method's average error is within about 2 percent, and the precision of an individual measurement runs roughly 2 to 8 percent depending on the isotope dose and how long the collection period lasts. For an estimated average expenditure of 2,500 calories a day over the collection period, that is roughly 50 to 200 calories (our conversion). It describes the average across the whole period, not any single day.

Can doubly labeled water tell me how many calories I ate?#

Not directly. It measures expenditure. Intake can be inferred from it only if your energy stores stayed constant during the measurement window, or if the change in body composition is measured and added back. That is why studies pair it with body composition measurements at the start and end.

Why don't fitness trackers use this method?#

Because it returns a single average over one to three weeks, not a minute-by-minute reading, and each measurement requires laboratory isotope analysis. It is used to check trackers and prediction equations rather than to replace them.

Sources#

Source: BurnWeek — "How doubly labeled water measures the calories you burn", https://burnweek.fit/blog/doubly-labeled-water-explained/. Licensed CC BY 4.0: free to quote or reuse with a link to this page.

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 →