Does a DEXA repeated every three weeks give me a usable lean-mass trend?

I would like to know what to measure as much as what to do.

Both of these get recommended confidently by different people, which suggests neither is obviously right.

My constraints are cost, measurement resolution and how much handling I am prepared to do — in roughly that order.

What is the actual trade-off, and does it matter at the scale I am working at?

3 weeks is 21 days, and the only question is whether 3 weeks accumulates more signal than the instrument's own noise. Scan-to-scan precision on lean mass is about one to one and a half per cent under standardised conditions, so on a 60 kg lean mass the least significant change is roughly 0.9 kg. At half a kilogram of total loss a week with a quarter of it lean — 0.125 kg of lean per week — 3 weeks accumulates 0.38 kg. 0.38 kg is under the 0.9 kg threshold, so a repeat every 3 weeks mostly reports positioning and hydration. On those figures you need about 7 weeks between scans before the trend outruns the error bars. Standardised means the same machine, the same time of day, the same hydration state, the same clothing. Change any of those and the precision figure this arithmetic rests on is no longer yours. Three scans on a line beat two scans and a difference at any interval: the third is what tells you whether the second was a measurement or an outlier.

Start with standardising the conditions, because hydration, food, recent exercise and time of day all move the result by more than the change you are trying to detect.

Different manufacturers use different edge-detection and tissue-decomposition algorithms, so cross-machine comparisons can differ by several percentage points of body fat on the same person on the same day.

Protein target arithmetic

At roughly 4 kcal per gram, 141 g of protein is about 564 kcal — a substantial fraction of a 900 kcal budget, which is the real constraint.

Put another way, scans are usually reported with a total and with regional segments. Regional lean mass in the limbs is the appendicular skeletal muscle measure used in the sarcopenia definitions, and it is more useful than the total for training questions.

Bioimpedance is not a substitute and is far more sensitive to hydration than DXA is.

Ask the facility for its precision error. If it does not have one, that is information.

Worth being precise here: regional analysis — trunk versus limbs — is more informative than the total for anyone tracking training effects.

Hydration is the dominant confounder. A litre of fluid is a kilogram, and it is allocated mostly to the lean compartment, so an under-hydrated scan flatters the body-fat percentage and a post-carbohydrate-loading scan does the opposite.

On the detail: bone mineral density falls modestly during sustained weight loss, and having the baseline on the same series is the only cheap way to see it.

Take the bone density seriously; it is the part of the scan nobody reads.

The short version: same machine, same time of day, fasted, rested, and then the series means something.

Typical precision error on a good DXA is around one per cent for fat mass and one to two per cent for lean mass, which means the least significant change is roughly two and a half to three per cent — about one to one and a half kilograms of lean mass for most people.

Standardise: same scanner, morning, twelve hours fasted, no exercise in the preceding twenty-four hours, no unusual carbohydrate intake in the preceding two days, and the same clothing.

Between-manufacturer differences in body-composition output on the same subject are documented in cross-calibration studies and are not small.

The caveat is that DXA is a two-compartment estimate in a three-compartment world, and its accuracy against reference methods is decent rather than excellent.

Precision is good, accuracy is conditional. Track change, do not compare absolutes.

edited 23 Oct 2025 by Dr_Yusuf_Adeyemi — fixed an arithmetic slip in the third paragraph

The honest answer is that a single scan tells you very little and a series of three tells you a great deal.

Radiation exposure is very low — of the order of a few microsieverts, comparable to a day or two of natural background — so scan frequency is limited by cost and by precision, not by dose.

Same machine, same conditions, twelve weeks apart. Anything else is noise collection.

Answering this needs the precision error of the specific scanner, which good facilities publish and most do not.

Scanning more often than every twelve to sixteen weeks generally produces changes inside the precision error, which reads as noise and gets interpreted as signal.

Read the regional segments, not just the total.