DEXA vs. InBody vs. Callipers: Real Accuracy, Real Limitations

The body-fat-measurement market sells you on accuracy numbers. DEXA is “the gold standard.” InBody is “clinical grade.” Callipers are “research validated.” All three are technically true. All three hide significant footnotes.

This is the realistic comparison: what each method actually measures, what the real-world accuracy looks like once you account for the hidden variables, and which one to use for which purpose. If you want to skip the comparison and run a free estimate from a tape measure, the Navy Body Fat Calculator is ±3–4% — competitive with callipers and better than most BIA scales.

---

DEXA: The Gold Standard With an Asterisk

Dual-energy X-ray absorptiometry uses two low-dose X-ray beams at different energies. Bone, lean tissue, and fat absorb the beams differently, and the scanner reconstructs a full-body composition map. The scan takes 6–10 minutes, you lie still on a table, and you get a region-by-region breakdown of fat mass, lean mass, and bone mineral density.

Claimed accuracy: ±1–2% body fat.

The asterisks:

• Hydration matters more than people admit. Lean tissue is ~73% water. If you’re 2% dehydrated, your “lean mass” reads lower and your “fat mass” appears proportionally higher. A 1.5L water difference between two scans can shift the reading 1–2 percentage points.
• Time of day matters. Morning scans (fasted, post-bathroom) read leaner than afternoon scans on the same person by ~1%. Glycogen and intramuscular water shift across the day.
• Recent exercise matters. A hard workout the day before increases lean-tissue water and can read your body fat ~1% lower than baseline.
• The machine matters. GE Lunar and Hologic scanners produce different absolute numbers for the same body. They agree on direction but not on absolute value. A reading of 18% on one machine might be 19.5% on another.
• Body size matters. Very tall (>6’3″) or very large (>300 lbs) bodies fall outside the scan area on some machines, and the scanner extrapolates.

The honest accuracy: ±2% for a given scan. ±2.5–3% when comparing scans on different machines. ±1% when comparing two scans on the same machine, same time of day, same hydration state.

Best for: anchor measurements, taken twice a year on the same machine, in the same prep state. The scan is also the only method that gives you regional fat distribution (visceral, abdominal, limb), which is clinically useful.

Worst for: week-over-week tracking. The within-day noise is larger than a week of fat loss.

---

InBody (Multi-Frequency BIA): The Gym Compromise

InBody (and its competitors Tanita, Seca, Tanita-MC) use eight-point multi-frequency bioelectrical impedance analysis. You stand on the platform, hold the handles, and the device measures impedance at 1, 5, 50, 250, 500, and 1000 kHz across hand-to-hand, hand-to-foot, and foot-to-foot pathways. The math reconstructs segmental body composition.

Claimed accuracy: ±3% body fat versus DEXA, in published studies.

The asterisks:

• Hydration sensitivity is enormous. A 2% hydration change shifts the body-fat reading by 4–6 percentage points. This is why InBody scans require the same pre-scan rules as DEXA: fasted, post-bathroom, no exercise within 12 hours.
• Recent food matters. A 300-calorie meal in your gut reads as ~0.5% body fat. A salty meal reads as more.
• The estimation equation matters. InBody uses proprietary algorithms that differ slightly by machine model. The 270 reads different absolute numbers than the 580 or the 770 for the same person — but the direction is consistent.
• Body type assumptions matter. The equations are trained on adult Asian and Caucasian body composition. Bodies far from the equation’s training distribution (very lean athletes, very obese, very elderly) read with 5–8% error rather than 3%.
• The bilateral electrodes can’t see what’s between them. Visceral fat estimation is an inference, not a measurement.

The honest accuracy: ±3% versus DEXA in controlled conditions. ±5% in the typical “walked into the gym, did the scan” case. The trend across multiple scans on the same machine, same time of day, same prep state is more reliable than any individual reading.

Best for: monthly check-ins where you can do the prep (fasted, morning, post-bathroom). Most gyms with an InBody include scans in membership or charge $15–$25.

Worst for: comparing your InBody to your friend’s DEXA. The two methods will disagree by 3–5%, and that disagreement is the method difference, not a measurement error.

---

Skinfold Callipers: The Technician Is the Variable

A calliper pinches a defined site (triceps, suprailiac, abdominal, etc.), measures the thickness of the skin-plus-fat fold in millimeters, and converts via a regression equation (Jackson-Pollock 3-site or 7-site is the dominant choice).

Claimed accuracy: ±3% body fat with a trained technician.

The asterisks:

• The technician is the variable. A trained technician can pinch the same site within ±0.5 mm session to session. An untrained person varies by ±1.5–2 mm — which propagates to ±5–8% body fat once the regression amplifies it.
• Site location matters. Anatomical landmarks (anterior superior iliac spine, midpoint between acromion and elbow) require practice to find. A 1-cm shift in site location can shift the reading 1–2%.
• Pinch pressure matters. Calliper springs are calibrated, but how long you wait after pinching (the fat compresses over 2–3 seconds), how firmly you isolate the fold, and tape resistance all affect the reading.
• The equation matters. Jackson-Pollock was developed on lean-to-average adult populations. It systematically overestimates body fat in lean athletes (1–3% too high) and underestimates in obese subjects (1–3% too low). Durnin-Womersley is friendlier to athletes; Brozek is friendlier to obese subjects. The equation you pick can shift the answer by 2%.
• Distribution matters. Skinfold methods only sample subcutaneous fat. Visceral fat is invisible to them. Two people with the same skinfold sums can have very different visceral profiles.

The honest accuracy: ±3% with a trained technician and a body that matches the equation’s training distribution. ±5–8% when you do it on yourself with a $15 plastic calliper.

Best for: people who train with a coach who measures them regularly, or people who do it themselves for trend tracking and have learned to ignore the absolute number.

Worst for: absolute body-fat estimation. The technician error swamps the equation error.

---

Side-by-Side: Realistic Accuracy

The numbers in “Typical Real-World” include the variables most people fail to control — hydration, time of day, prep state, technician inconsistency. The “Within-Month Repeatability” column is the more useful number for tracking purposes: how much would two readings differ on the same body, weeks apart, with nothing actually changed?

---

The Practical Recommendation

Pick one method and track over time. This is the single most useful piece of advice in the entire space. The 2% absolute difference between DEXA and InBody is irrelevant if you’re tracking change. What matters is that your method gives you the same answer this month when nothing about you has changed.

A defensible setup looks like this:

• Primary method (free or cheap, monthly): Navy tape, run through the Navy Body Fat Calculator. Repeatable, ±3–4% accuracy, no clinic visit.
• Anchor calibration (once or twice a year): DEXA scan at a sports-med clinic or walk-in body composition center. Use it to correct your at-home reading: if the DEXA says 18% and your tape said 16%, you know your tape reads 2 points low and you adjust.
• Optional weekly trend (low effort): a consumer BIA scale used at the same time every morning. The day-to-day number is noise. The 30-day rolling average tracks the direction.

What doesn’t work: switching methods mid-cut. If you started with the Navy tape, finish with the Navy tape. If you started with DEXA, finish with DEXA. The 2–4% absolute disagreement between methods looks like progress or regression and isn’t real.

---

When the Numbers Don’t Match

You weigh in. The DEXA says 19%. The InBody says 16%. The Navy tape says 17%. Your scale says 14%. The honest answer:

• The DEXA is probably closest to the truth.
• The Navy tape, at 17%, is doing exactly what it was designed to do — ±2% of DEXA in a typical body.
• The InBody, at 16%, is also within its expected error band.
• The consumer scale at 14% is reading low — almost all consumer BIA scales read 3–5% below DEXA in lean bodies because the equation training set is biased average-overweight.

The right move is to use the DEXA as your “real” number, the tape as your monthly tracker, and ignore the scale’s absolute number while still using its trend line.

---

The Verdict

DEXA is the most accurate widely-available method, with real ±1–2% accuracy under controlled prep — but the within-day noise is large enough that week-over-week tracking is meaningless. InBody is solid for monthly check-ins if you control hydration. Skinfolds are technician-limited and best left to people who train with a coach. The Navy tape method is competitive with skinfolds for self-measurement, costs $5, and is the right default for most people.

The right strategy isn’t to find the “most accurate” method. It’s to pick one, control the variables you can, track the trend, and use a DEXA scan once or twice a year as a reality check. The goal isn’t precision — it’s a consistent signal that tells you whether the work is working.