The Accuracy Myth - What Body Composition Testing Actually Measures.

Few words get thrown around more carelessly in health and fitness than “accurate”. Walk into any gym, scroll through any wellness brand’s website, and you’ll find devices, trainers and influencers making bold claims about accuracy, usually to sell you something, and almost never backed by a clear definition of what that word actually means in the context of body composition testing.

I want to change that conversation. Because the science is clear, the literature is robust, and most of what people believe about body composition testing, especially BIA, is wrong.

First, let’s establish what “the gold standard” actually measures.

DXA (Dual-Energy X-ray Absorptiometry) is frequently cited as the gold standard for body composition testing. But here’s what’s rarely said out loud: DXA is the gold standard for bone mineral density, not body composition as a whole. When researchers use DXA as the reference comparator for body composition studies, they’re making a pragmatic choice based on availability, safety and convenience, not because DXA provides a perfect direct measure of fat mass or lean mass.

In fact, outside of MRI, which is expensive, inaccessible and impractical at scale, no technology provides a true direct measure of whole-body composition in living humans. The only truly direct measurement of body composition in its entirety remains post-mortem cadaveric analysis.  

Everything else, DXA, BIA, hydrostatic weighing, air displacement plethysmography, is an estimate, derived from models and assumptions (Wang et al., 1992; Heymsfield et al., 2005; Wells & Fewtrell, 2006). This isn’t a criticism of any technology. It’s simply the physiological reality that every honest conversation about body composition testing must start with.

DXA has the same measurement conditions as BIA. All of them do.

Here is something the DXA advocates rarely mention - DXA results are just as sensitive to measurement conditions as BIA. Hydration status, time of day, recent food intake, supplement and medication timing, training history in the preceding 24–48 hours, all of these introduce variance into DXA-derived body composition estimates, just as they do with BIA.

This is not speculation. It is well-documented in the peer-reviewed literature. Lohman et al. (1992) established foundational standardisation protocols for body composition testing across all methods. A systematic review by Toomey et al. (2017) confirmed that DXA measurements are meaningfully affected by hydration status, food intake and exercise timing, with fat mass estimates varying by up to 1.5 kg and lean mass estimates by up to 2 kg depending on pre-scan conditions.

The same applies to BIA. Deurenberg et al. (1989) demonstrated that even mild dehydration significantly alters BIA-derived impedance values, and Kushner et al. (1996) confirmed that food and fluid intake before a scan introduces clinically meaningful variance. The practical conclusion is identical regardless of the device you use, standardise your conditions, scan at the same time of day, in a consistent pre-scan state, and treat the change between scans as the meaningful data, not any single absolute number.

The five inputs that make BIA reliable. And why removing any of them should be a red flag.

BIA works by passing a small electrical current through the body and measuring the resistance (impedance) to that current. Fat mass resists electrical current; lean mass, which contains water and electrolytes, conducts it. From that impedance value, body composition is estimated using a validated prediction equation.

The quality of that equation determines everything. The peer-reviewed literature is unambiguous: BIA equations that incorporate five variables, height, weight, age, gender and impedance, consistently produce the highest agreement with reference methods (Kyle et al., 2004, Clinical Nutrition, Parts I and II). These five variables are the established covariates required to account for the physiological diversity of human bodies across age, sex and body size.

If a manufacturer tells you their device doesn’t use age or gender in its equation, treat that as a significant red flag. The physiological relationship between impedance and body composition changes substantially with age (Roubenoff et al., 1997) and differs fundamentally between males and females. Omitting these variables doesn’t make the device simpler or more objective, it makes it less physiologically valid for precisely the populations who need reliable data most.

The physiology is obvious. The gender difference in body composition is not a detail — it’s a fundamental biological reality.

Men and women differ profoundly in body composition. On average, women carry a higher percentage of essential fat (approximately 10–13% vs 2–5% in men), have different fat distribution patterns, different hormonal profiles affecting fat storage and mobilisation, and different reference ranges for what constitutes healthy body fat percentage (Lohman, 1992; American College of Sports Medicine, 2018).

These differences are not minor adjustments, they are physiologically fundamental. A BIA algorithm that treats male and female bodies identically is not neutral or inclusive. It is scientifically inaccurate. The literature consistently demonstrates that sex-specific equations produce significantly better agreement with reference methods than generalised equations (Kyle et al., 2001; Sun et al., 2003). This is settled science. Any device that markets the omission of gender as a feature deserves scrutiny, not applause.

The number on the screen is not an indictment. It’s a baseline.

This might be the most important point of all, and the least clinical.

People are emotionally attached to their perceived body fat percentage. When a medically graded body composition device returns a number higher than expected, the instinctive response is often to distrust the device rather than reconsider the assumption. This is entirely human, and entirely counterproductive.

BIA devices registered as medical devices across multiple regulatory jurisdictions, the FDA, TGA, CE and Health Canada, are non-invasive, validated instruments for body composition assessment. Their clinical value lies not in any single absolute number, but in the trajectory of change over time under standardised conditions. A baseline reading tells you where you are. Subsequent readings, taken consistently, tell you whether your intervention is working.

The narrative that needs to change is this: body fat percentage is not a judgement. It is a data point. And a single data point, taken once, under unknown conditions, tells you relatively little. What tells you a great deal is a consistent series of measurements showing whether lean mass is being preserved, whether visceral fat is trending down, whether hydration is stable.

The mirror lies. Your body composition data doesn’t have to.

Being visually lean does not mean having low body fat. And it certainly does not mean having low visceral fat.

Visceral adipose tissue, the metabolically active fat stored around the abdominal organs, is not visible in the mirror. It does not correlate reliably with appearance. Individuals with a lean aesthetic presentation can carry clinically significant visceral fat accumulation, a phenomenon sometimes described as TOFI (Thin Outside, Fat Inside), which is associated with elevated cardiometabolic risk despite normal or low BMI (Thomas et al., 2012; Stefan et al., 2008).

This is exactly the category of insight that body composition assessment exists to provide, and that the mirror, BMI, and even bodyweight on a scale cannot. A comprehensive body composition scan, conducted consistently over time, surfaces data that no external observation can.

The Evidence Supports BIA unequivocally

Let me be direct - the claim that BIA is an unreliable or inferior measure of body composition is not supported by the current scientific literature, when appropriate, validated equipment and standardized protocols are used. What follows is not a marketing claim. It is a summary of peer-reviewed evidence spanning decades, populations and clinical contexts.

Kyle et al. (2004), one of the most cited reviews of BIA methodology ever published, appearing across two landmark papers in Clinical Nutrition concluded that BIA incorporating the five-variable algorithm provides body composition estimates with high agreement to reference methods across diverse populations. This wasn't a single study. It was a comprehensive synthesis of the field, and it remains the foundational scientific reference for how BIA should be designed, validated and applied.

Raeder et al. (2018) took this further by validating BIA against DXA in colorectal cancer patients — a physiologically complex population characterized by significant changes in hydration, muscle mass and body composition during treatment. Even in this challenging clinical context, R² values of 0.94 to 0.98 were reported. If BIA holds up in oncology, it holds up in a gym.

Sergi et al. (2015) demonstrated R² = 0.92 against DXA in free-living Caucasian adults aged 60–85 years, the demographic most susceptible to sarcopenia and the one with the most to gain from consistent body composition monitoring.  

Lukaski et al. (1986), in one of the earliest and most influential validation studies, established the foundational agreement between tetrapolar BIA and established reference methods, laying the groundwork for everything that followed.  

Sun et al. (2003) developed population-validated BIA prediction equations using a multicomponent model across a large and diverse sample, further cementing the scientific credibility of the method.

Desport et al. (2003) validated BIA-derived fat-free mass in patients with amyotrophic lateral sclerosis, a neurological condition associated with asymmetric, complex changes in muscle mass and hydration, and found no significant difference from DXA-derived reference values (p = 0.43). That BIA performs at this level in ALS patients is a remarkable demonstration of the robustness of the measurement principle.

Across healthy adults, clinical populations, older adults, athletes and patients with complex disease, the literature tells a consistent story. BIA, when properly designed and correctly applied, produces body composition estimates that agree strongly with reference methods. The criticism that BIA is unreliable is almost always a criticism of poorly designed devices, poorly standardized protocols, or both. It is not a criticism that applies to validated, medically registered BIA technology used correctly.

These are not marginal findings. They are compelling, replicated, peer-reviewed — and they span four decades of independent research.

What you should actually be asking.

Instead of “is this device accurate?”, ask –

–  Does this device use a validated five-variable algorithm including height, weight, age, gender and impedance?

–  Is it registered as a medical device in regulated markets (FDA, TGA, CE, Health Canada)?

–  Are measurement conditions standardized across scans?

–  Am I tracking change over time rather than fixating on a single absolute number?

If the answer to all four is yes, you have a clinically defensible, scientifically supported body composition assessment tool. The conversation about “accuracy” is a distraction from the conversation that actually matters - consistency, context, and what you do with the data.

‍Guest Author, Kylie Bruno, Product Manager, Evolt Health

‍

‍References

Cho W. EV360 Clinical Study Results. Chungnam National University, Republic of Korea. April 2026. Pre-publication.

Deurenberg P, Weststrate JA, Hautvast JGAJ. Changes in fat-free mass during weight loss measured by bioelectrical impedance and by densitometry. Am J Clin Nutr. 1989;49(1):33–36.

Heymsfield SB, Lohman TG, Wang Z, Going SB (eds). Human Body Composition. 2nd ed. Human Kinetics; 2005.

Kushner RF, Gudivaka R, Schoeller DA. Clinical characteristics influencing bioelectrical impedance analysis measurements. Am J Clin Nutr. 1996;64(3 Suppl):423S–427S.

Kyle UG, Bosaeus I, De Lorenzo AD, et al. Bioelectrical impedance analysis — Part I: review of principles and methods. Clin Nutr. 2004;23(5):1226–1243.

Kyle UG, Bosaeus I, De Lorenzo AD, et al. Bioelectrical impedance analysis — Part II: utilization in clinical practice. Clin Nutr. 2004;23(6):1430–1453.

Kyle UG, Genton L, Karsegard L, Slosman DO, Pichard C. Single prediction equation for bioelectrical impedance analysis in adults aged 20–94 years. Nutrition. 2001;17(3):248–253.

Lohman TG. Advances in Body Composition Assessment. Human Kinetics; 1992.

Raeder H, et al. Validity of bioelectrical impedance analysis in estimation of fat-free mass in colorectal cancer patients. Clin Nutr. 2018;37(1):292–300.

Roubenoff R, Baumgartner RN, Harris TB, et al. Application of bioelectrical impedance analysis to elderly populations. J Gerontol A Biol Sci Med Sci. 1997;52(3):M129–M136.

Sergi G, De Rui M, Veronese N, et al. Assessing appendicular skeletal muscle mass with BIA in free-living Caucasian older adults. Clin Nutr. 2015;34(4):667–673.

Stefan N, Kantartzis K, Machann J, et al. Identification and characterisation of metabolically benign obesity in humans. Arch Intern Med. 2008;168(15):1609–1616.

Sun SS, Chumlea WC, Heymsfield SB, et al. Development of bioelectrical impedance analysis prediction equations for body composition. Am J Clin Nutr. 2003;77(2):331–340.

Thomas EL, Frost G, Taylor-Robinson SD, Bell JD. Excess body fat in obese and normal-weight subjects. Nutr Res Rev. 2012;25(1):150–161.

Toomey CM, McCormack WG, Jakeman P. The effect of hydration status on the measurement of lean tissue mass by dual-energy X-ray absorptiometry. Eur J Appl Physiol. 2017;117(3):567–574.

University of South Queensland. Summary of the Reliability of the Evolt 360. Internal research summary, physically active adults 20–40 years (ICC = 0.99).

Wang ZM, Pierson RN Jr, Heymsfield SB. The five-level model: a new approach to organizing body-composition research. Am J Clin Nutr. 1992;56(1):19–28.

Wells JC, Fewtrell MS. Measuring body composition. Arch Dis Child. 2006;91(7):612–617.