Accepted answer
Believe the DEXA for tracking, treat the BodPod as a reasonable independent fat estimate, and treat the BIA as a hydration meter with a body-composition skin on it. Here is the reasoning device by device.
What each one physically measures
| Device | Physical measurement | Key assumption | Assumption most violated by | Typical precision on repeat |
|---|
| DEXA | Differential attenuation of two X-ray energies, pixel by pixel | Soft tissue over bone can be estimated from adjacent non-bone pixels; hydration of lean tissue is ~73% | Very large body habitus, oedema, arms outside the scan field | Fat mass 1-2%, lean mass 1.0-1.5% coefficient of variation |
| Air-displacement plethysmography (BodPod) | Whole-body volume, hence density | Fixed densities for fat (0.9007 g/mL) and fat-free mass (1.100 g/mL) | Anyone whose fat-free-mass density is off: high hydration, low bone density, very muscular, very old | Fat percentage 1.5-2.5 percentage points absolute |
| Single- or multi-frequency BIA | Impedance to a small alternating current | The body is a fixed-geometry conductor whose resistance maps to total body water, and TBW maps to fat-free mass at a fixed ratio | Hydration swings, glycogen state, recent exercise, food and fluid in the gut, skin temperature, menstrual cycle phase | Fat percentage 2-4 points; lean mass can swing 1-2 kg within a single day |
| Four-compartment model (research) | Combines densitometry, total body water by dilution, and bone mineral by DEXA | Measures hydration rather than assuming it | Availability and cost | Best available; the reference the others are validated against |
The pattern: every one of these is a two-compartment or three-compartment estimate built on an assumed constant. DEXA assumes lean tissue hydration. BodPod assumes fat-free-mass density. BIA assumes both, plus geometry, and is the only one whose input signal is directly a water measurement. That is why BIA is the worst choice for someone whose water balance is moving, which is exactly your situation in month five.
Why your three numbers disagree in the direction they do
Your BodPod reads highest for fat. During sustained weight loss, fat-free mass tends to be relatively over-hydrated, and water is less dense than the 1.100 g/mL the model assumes for fat-free mass. Over-hydrated lean tissue lowers measured whole-body density, and the model can only explain low density by assigning more fat. This is a known, directional artefact, not a malfunction.
Your BIA reads lowest for fat, and its 36.2 kg is not comparable to the DEXA 57.8 kg at all: it is reporting an estimate of appendicular skeletal muscle, whereas DEXA lean soft tissue includes organs, blood, connective tissue and water. Both can be internally consistent and share no scale.
Least significant change
Precision error tells you the smallest change you can call real. The usual formula is LSC = 2.77 x precision error for a 95% confidence interval on a two-scan comparison.
Suppose your DEXA centre's lean-mass precision error is 1.2% of 57.8 kg:
precision error = 0.012 x 57.8 = 0.69 kg
LSC = 2.77 x 0.69 = 1.9 kg
So a 1.2 kg drop in lean mass between two DEXA scans on that machine is not a detectable change. A 2.5 kg drop is. Ask the centre for its in-vivo precision figures; a good one has them, a bad one has never measured them. If they cannot tell you, assume roughly 1.5-2 kg for lean and 1-1.5 kg for fat as your threshold, and stop reacting to anything smaller.
Corollary that matters more than the device choice: never mix devices in a trend line, and never switch machines mid-programme even between two units of the same model. Cross-device differences are systematic and larger than the changes you are trying to detect.
2Asking a scan centre for its in-vivo precision error is a great filter. Most cannot answer. – Dr_Signe_Baldursdottir 9 months ago Same-machine, same-technician, same-positioning. Otherwise the LSE calculation is fiction. – juan_esquivel 7 months ago add a comment