DEXA vs InBody vs 3D Scans: How to Read Your Results

The short answer. No body composition method available outside a research lab measures fat directly. DEXA, bioelectrical impedance and 3D optical scanning all estimate it, from different physical signals and different built-in assumptions, which is why the same body can return three different percentages on the same afternoon. That is not a malfunction. Once you know each method's error bars, the useful move is to pick one instrument, hold your test conditions constant, and read the trend rather than the number — which is exactly how we use 3D body composition scanning.

Most scan reports are presented as if they were measurements. Knowing which part of your report is a measurement and which part is a modeled estimate is most of the skill of reading it.

What each method is actually doing

DEXA passes two x-ray energies through the body and infers fat, lean soft tissue and bone from how much each is attenuated. It is the most established of the three, and it is validated against genuinely direct methods, including chemical analysis of cadavers. Its agreement with whole-body CT for fat mass is very high, with correlations around 0.99 — though the same review notes DEXA underestimating whole-body fat mass by as much as 5 kg on average against CT (Shepherd and colleagues, Bone, 2017). Even the reference standard carries a systematic offset.

Bioelectrical impedance devices pass a small current through the body, estimate total body water from the resistance they meet, and derive fat-free mass from that water estimate. The chain of inference is longer, and it runs through hydration.

3D optical scanning does something different in kind. It builds a surface model of your body, derives circumferences and segmental volumes from that geometry, then estimates composition from shape. The first half is close to a direct measurement. The second half is a prediction model.

Styku body scan: repeatability and shape

The most useful published numbers come from a validation study of 188 adults comparing a 3D optical scanner against both DEXA and tape-measure anthropometry.

On circumference, agreement was excellent: concordance correlation coefficients of 0.97 for waist and 0.98 for hip (Bennett and colleagues, Clinical Nutrition, 2022). Test-retest precision was tight — 0.60% for percent fat, under 0.60 cm for both waist and hip circumference, and 1.1 L for whole-body volume (Bennett and colleagues, PubMed record). Fat-free mass tracked DEXA closely, with a mean difference of 1.2 ± 3.4 kg and a concordance correlation of 0.97.

Precision, not accuracy, is the property that matters for tracking. A method can carry a consistent offset and still be the right tool for detecting change, provided the offset stays consistent. That is the real case for optical scanning, and it is narrower than most marketing makes it.

It also lands on a measure clinicians increasingly care about. An international consensus statement recommends treating waist circumference as a vital sign, notes that it provides information both independent of and additive to BMI for predicting morbidity and death, and identifies decreases in it as a critically important treatment target (Ross and colleagues, Nature Reviews Endocrinology, 2020).

Where 3D scanning is weak: the body fat percentage itself

Here is the number that should temper expectations. In that same study, DEXA returned a mean body fat of 28.2% and the scanner returned 26.3% — a mean difference of only 1.9 percentage points, which sounds excellent. But the 95% limits of agreement spanned roughly 7 points below to 11 points above, and the concordance correlation for body fat percentage was 0.86, well below the 0.97 achieved for fat-free mass (Bennett and colleagues, Clinical Nutrition, 2022).

A small average gap with wide limits of agreement means the method is well calibrated across a population and can still be meaningfully off for you individually.

An independent comparison is blunter still. Four commercial 3D scanners were tested in 171 participants against a four-compartment model, the strongest reference available. Reliability was good across the board — root-mean-square coefficients of variation of 2.3–4.3% for body fat percentage, intraclass correlations of 0.975–0.996. Validity was patchier: concordance correlations of 0.74–0.90 for body fat percentage, root-mean-square errors of 3.7–6.1%, proportional bias in every scanner tested, and three of the four — but not the Styku unit — meeting the study's equivalence criterion against the four-compartment model (Tinsley and colleagues, Clinical Nutrition, 2020).

We would rather publish that finding than leave you to discover it. It is one search away.

Where impedance readings differ

Impedance shows the same pattern in a different place: highly repeatable, systematically offset. Across 67 adults tested against DEXA on two separate days, three InBody analyzers all produced intraclass correlations of at least 0.98 for body fat percentage, with standard errors of measurement between 0.77% and 0.99%. All three also showed systematic bias against DEXA — underestimating body fat percentage and fat mass, overestimating fat-free mass — with the smallest detectable difference for body fat percentage falling between 2.12 and 2.73 percentage points (McLester and colleagues, Journal of Clinical Densitometry, 2020).

That last figure is the practical one. A 1.5-point shift on an impedance reading is not necessarily a change in your body.

Fluid is the specific vulnerability. A randomized crossover trial in 18 healthy adults gave participants different oral solutions and re-measured impedance over two hours. The authors concluded that all the fluids tested may affect conductivity and therefore the calculation of body compartments, and recommended measuring without recent fluid intake — while noting the mean differences they saw were under 1% and may not be clinically meaningful (randomized crossover trial of fluid intake and impedance measurement, Metabolites, 2023).

Body composition scan results: reading your report

Four rules cover most of it.

Compare like with like. A percentage from one method is not interchangeable with a percentage from another. Switching devices mid-program destroys your trend line.

Know your method's smallest detectable change. If the published figure is 2.12–2.73 points for a given impedance device, a two-point move is inside the noise.

Standardize the conditions, not just the device. The most commonly skipped step, and it matters even for the reference method: in resistance-trained athletes, precision error on consecutive days was nearly twice as large for fat mass (1261 g versus 660 g) and more than three times as large for lean mass (2083 g versus 617 g) as two scans taken the same day (Zemski and colleagues, Journal of Clinical Densitometry, 2019).

Watch the measurements before the estimates. Circumference and segmental volume are closer to observation. Body fat percentage is a model output. Weight it accordingly.

The limits of all of this

Three caveats belong on the record.

The validation studies above are cross-sectional comparisons in mostly healthy adult samples, several of them small — 18 participants in the fluid trial, 21 in the consecutive-day precision study. They describe how instruments behave, not how a particular person's program should change.

Body composition is a process measure, not an outcome. It indicates the direction your training and nutrition are pushing you. It diagnoses nothing, and reading a scan is not a substitute for a clinician.

And every commercial device carries proprietary calibration the published literature can only partly see through. Some optical systems ship more than one composition model, calibrated against different reference methods — so two people scanned on nominally the same equipment can get numbers that are not comparable.

What this means at Aeon

We run 3D scanning for what it is genuinely good at: repeatable circumference and shape data, captured quickly, with no radiation, in a form you can see rather than a single number to argue with. We do not present it as a clinical-grade body fat measurement, because the published limits of agreement do not support that claim.

What it buys you is a stable baseline and a trend — same device, same protocol, same landmarks, across a training block, read alongside how much weight you are moving and how you are recovering. If a decision genuinely depends on a precise body fat percentage or on bone density, that is a referral, and we will say so rather than stretch the tool we have.

What the scan captures and how it fits into an assessment is on our body composition scan page.