Indirect Calorimetry

For physicians

What is measured

Oxygen consumption and carbon dioxide production at rest, from which resting metabolic rate, resting energy expenditure and the respiratory quotient are derived. Acquisition uses a self-calibrating single-use sensor, and results are available within minutes.

Why prediction equations underperform

Predictive equations are population regressions and inherit the variance of the population they were built on. In any individual whose body composition, thyroid status, weight history or age places them away from that population’s center, the estimate diverges from the measurement — and the divergence is not detectable without measuring. This is the practical case for calorimetry over calculation in exactly the patients a metabolic clinic sees.

The respiratory quotient

The ratio of carbon dioxide produced to oxygen consumed indicates the predominant substrate being oxidized: a lower value indicates predominantly fat, a higher value predominantly carbohydrate. Read it as a physiological descriptor at the moment of measurement, under the recorded conditions — it is not a fixed trait and it moves with recent intake, which is why the fasting requirement is absolute.

Conditions that invalidate the measurement

• Non-fasted state, or caffeine or nicotine within the preceding hours.
• Insufficient rest before acquisition — the walk from the parking lot is enough.
• Talking, sleeping or restlessness during the recording.
• Mask or mouthpiece leak; check the seal rather than trusting the trace.
• Acute illness, fever or recent significant exertion.

Where it fits

Read against body composition, which is acquired first for exactly that reason, and against insulin and glycemic markers from the laboratory panel.

Evidence and limitations

Written for a reader who wants the literature rather than the summary. Where the evidence is thin we say so and give both sides, and where a claim is mechanistic rather than demonstrated we label it as such.

Mechanism

Substrate oxidation consumes oxygen and releases carbon dioxide in fixed stoichiometric proportions that differ between fat and carbohydrate. Measuring both gas exchanges at rest therefore yields energy expenditure and, from their ratio, the predominant fuel being oxidized. The measurement is direct in the sense that matters: it observes the patient’s own metabolism rather than inferring it from a regression on other people.

What the evidence shows

Predictive equations are population regressions and inherit that population’s variance. A 2023 study enrolling 722 Caucasian adults — normal weight, overweight and with obesity, including non-hospitalized subjects with medically controlled diabetes, cardiovascular disease and thyroid disease — measured resting metabolic rate by indirect calorimetry and compared it with the Harris-Benedict, Mifflin-St Jeor, FAO/WHO/UNU and Owen equations. The authors developed revised equations that outperformed the existing ones on the same anthropometric inputs, reporting R-squared of 0.95 for men and 0.86 for women, and noted that the older equations were derived under different conditions from contemporary ones.

The clinically important reading of that paper is not which equation wins. It is that a family of widely used equations, applied to the same 722 people whose metabolic rate was actually measured, disagreed with the measurement enough for a new equation to be worth publishing.

Limitations of that evidence

• Single-population derivation. The 2023 cohort was Caucasian and adult, excluding active infection and pregnant or lactating women. Equations derived there carry the same transferability caveat as the ones they replace.
• Group accuracy is not individual accuracy. A high R-squared at group level coexists with clinically meaningful error in individuals, which is the whole reason to measure someone rather than predict them.
• Measurement conditions are unforgiving. Non-fasted state, caffeine, nicotine, insufficient rest, talking, dozing or a mask leak each invalidate the acquisition.
• Respiratory quotient reflects the moment. It moves with recent intake and is a physiological descriptor under recorded conditions, not a stable trait.

What remains uncertain

What has not been demonstrated is that building a nutrition plan on a measured rather than estimated resting energy expenditure produces better weight or metabolic outcomes in a randomized comparison. It is a reasonable expectation and it is not yet an evidenced claim, so this site does not make it. What the measurement does establish is what a given person’s resting expenditure actually is, which converts an argument about willpower into a number that can be checked again in four months.