Is an InBody scan accurate? Not accurate enough to trust a single body fat percentage as your true one, and genuinely useful for tracking change in the same person under the same conditions. Against DEXA, the reference method, an InBody reading sits a few percentage points low on body fat and a couple of kilogrammes high on fat-free mass. That bias is consistent in direction, which is the point: most of it cancels when you subtract one scan from the next. We sell these scans, so you should know what you are buying and what it cannot tell you. The device at our studio is an InBody 270, and the rest of this piece is about that machine specifically, because the model matters more than most providers admit.

For how measurement fits into the bigger job, start with our guide to losing fat without losing muscle.

Expertise reviewed by Mariyan Voykov, Manager and Senior Coach at Body Transformation London. MSc Sports and Exercise Science, BSc Biomedical Science, Level 3 Personal Trainer.

Is an InBody scan accurate compared with DEXA?

One study has compared the InBody 270 directly with DXA. In 88 adults tested under a controlled protocol (three-hour fast, no prior exercise), percent body fat read 24.6 ± 9.0% on the 270 against 27.7 ± 7.5% by DXA, fat mass 17.8 ± 8.2 kg against 19.7 ± 7.3 kg, and fat-free mass 53.7 ± 13.0 kg against 51.1 ± 12.5 kg, all at p<0.0001 (Garcia 2020). In round numbers: body fat under-read by about 3.1 percentage points, fat mass under-read by about 1.9 kg, fat-free mass over-read by about 2.6 kg. The authors put it plainly: the InBody 270 “under-predicts fat mass and percent body fat; conversely, it over-predicts fat-free mass.”

Here is the part nobody else will tell you. That paper reports paired t-tests only. It publishes no correlation coefficient, no intraclass correlation and no limits of agreement for the 270 against DXA, and no other study has published them either. So we can tell you the average direction and rough size of the error. We cannot tell you how wide the error runs from person to person on this model, because that number does not exist in the literature. Anyone quoting you a margin of error for this model is quoting something else.

What that means in practice: treat the roughly three-point gap as a typical offset, not a correction factor you can subtract to get your “real” number. On other InBody models where limits of agreement have been published, they are wide. In 1,000 US Marines tested on an InBody 770, the mean body fat bias was −4.2 ± 3.0 percentage points in men, but the limits of agreement ran from −10.01 to +1.57 (Potter 2025). That is a different, higher-specification device and those figures are not our machine’s. They do tell you what the spread around a tidy-looking average tends to look like in this method.

The direction is the reliable part, and it is the same on every model tested: bioimpedance tends to underestimate fat and overestimate fat-free mass. If your printout flatters you slightly, that is expected behaviour, not a compliment.

Two further caveats belong here. A second, smaller study of 47 exercise-trained adults found the same direction of error but did not reach statistical significance at that sample size (Czartoryski 2020). And no peer-reviewed validation of this model in a defined athletic population exists at all, so if you are lean and heavily trained, you are outside the tested range.

Another study puts it bluntly. In 93 participants, only 35% of InBody body fat measurements fell within 10% error of DEXA, and the authors concluded these devices “should not be used in lieu of criterion measures such as DEXA” (Merrigan 2022). Which is why we never call an InBody number your body fat percentage. It is the scanner’s estimate of it.

The ESPEN technical review sets the method’s boundaries: within-day reproducibility of 2.7 to 4.0%, prediction errors of 3.5 to 6% for fat-free mass, and results valid up to 34 kg/m², above which most predictive equations are not reproducible for individuals over time. Its summary judgement is that bioimpedance is “more suitable for population groups than for individuals” (Kyle 2004).

Why does a correlation coefficient not answer the question?

Accuracy for these devices is often summarised as a correlation with DEXA. Correlation is a real statistic. It answers a different question from yours.

Correlation measures association: whether two methods rank people in the same order. Agreement measures whether they produce the same number. Two devices can correlate almost perfectly while one reads three points low on everybody, because the order holds and every value is wrong by the same amount. That is the pattern across bioimpedance as a method. So a scanner can correlate strongly with DEXA and still tell you nothing about whether your 21% is really 25%. Ask for the bias and the limits of agreement instead.

For our model you will not get either from us, because neither has been published. What we can give you is the mean difference from the one study that exists, and the honest caveat that it comes with.

What does NICE actually say about bioimpedance?

NICE guideline NG246, recommendation 1.9.9: “Do not use bioimpedance as a substitute for BMI as a measure of general adiposity in adults.” (NICE NG246)

What that means: nobody should present a bioimpedance reading as a superior replacement for BMI. If a provider says their scanner makes BMI obsolete, UK guidance does not support that.

What it does not mean: it is not a ban on using a scanner, and it does not say the readings are worthless. It is about substitution in adiposity assessment, not tracking progress in a gym.

So the scanner sits alongside the simple measures rather than replacing them. NICE recommends waist-to-height ratio as well as BMI in adults with a BMI below 35 kg/m², phrased as keeping your waist to less than half your height (NG246, 1.9.8 and 1.9.15).

What is an InBody scan genuinely good at?

Repeating itself. Under controlled conditions, bioimpedance test-retest reliability is close to perfect: ICC ≥0.998 for body fat percentage and 1.000 for fat-free mass, skeletal muscle mass and total body water, with day-to-day biological variability of 0.6 percentage points for body fat and 0.5 kg for fat mass (Looney 2024). Because the bias against DEXA is largely systematic, most of it subtracts out when you compare scan two with scan one. That is the honest basis for tracking the trend, not the number.

No adult test-retest study of the InBody 270 itself has been published, so that reliability figure comes from other devices and we are not going to pretend otherwise. The closest thing to a fair proxy tested the InBody 230, 720 and 770 across two test days in 67 adults: test-retest ICC ≥0.98 for body fat percentage, fat mass and fat-free mass, and a minimum detectable change of 2.12 to 2.73 percentage points of body fat, 1.49 to 2.39 kg of fat mass and 1.60 to 2.32 kg of fat-free mass (McLester 2020). The InBody 230 runs the same measurement engine as the 270, two frequencies at 20 and 100 kHz through eight-point tetrapolar electrodes, so we treat those figures as a reasoned inference about our device rather than a finding on it.

That same study is quietly reassuring about the hardware: “the research grade models, InBody720 and InBody770, added minimal benefit over the portable InBody230.” The expensive machines are better at outputs the cheaper ones do not print, not obviously better at the numbers both of them do.

It holds up in practice, with a caveat. Across a four-week diet in 41 trained adults, the change in body fat was −1.3 ± 0.9% by DXA and −1.4 ± 1.8% by InBody, no significant difference in the group means (Antonio 2019). The means matched. The spread did not: the scanner’s variability on the change was double DXA’s.

The counter-evidence cuts against us and deserves stating. Over a 12-week intervention, InBody’s agreement with DEXA for change was poor, concordance 0.33 (95% CI 0.03 to 0.58) (Merrigan 2022). Two studies, two different answers, and we are not going to pretend that is resolved. The fair reading: group-level change tracking is well supported, individual change tracking is noisier than the marketing suggests, and one person’s twelve-week delta deserves moderate rather than high confidence. That argues for scanning consistently and reading it cautiously, not for scanning less.

Some outputs beat others. Muscle agrees with DXA far better than fat does: appendicular skeletal muscle index reached ICC 0.965 in 84 older adults, while regional fat estimates failed badly, arm fat mass managing only ICC 0.615 with a mean difference of +2.32 kg (Buch 2022). That second finding is academic for us, because this model does not print segmental fat at all. What it does print is Segmental Lean Analysis: five segments, each in kilogrammes and as a percentage of what would be expected, referenced to your age and sex. That is the useful half of the segmental idea, and it is what we read.

It also means the sheet cannot tell you where you carry fat. There is no visceral fat area, no visceral fat level and no limb-by-limb fat breakdown on this model, so if anyone quotes you a visceral fat number off this sheet, it did not come from the machine.

What moves an InBody reading, and what should you use it for?

Fluid, mostly. About 832 mL of a glucose solution raised measured body fat by roughly 1.2 percentage points at 20 minutes, and the control arm drifted up 0.7 points over two hours with no intervention (Schierbauer 2023). Across a day fluid redistributes from the legs upwards: hand-to-hand impedance rose 27.2 Ω and foot-to-foot fell 31.8 Ω between morning and evening, while whole-body hand-to-foot impedance was unchanged (Oshima 2006). So the segmental lean columns move across a day even when your whole-body figure does not, which is why we compare them morning to morning or not at all.

There is a study on the other side. Researchers deliberately broke the rules (a litre of water, a large meal, 15 minutes on a treadmill, a full bladder) and found body fat differences of only −1.9 to +0.4 percentage points, none significant, 86% of tests within 2%. It used consumer devices rather than a professional scanner, was underpowered (182 participants were needed on a post-hoc calculation), and tested only single violations (Randhawa 2021).

Fair synthesis: breaking protocol probably moves a reading by one to two percentage points of body fat. Small next to the gap versus DEXA, and the same size as a month of genuine progress. Standardisation barely matters for a one-off scan. It matters enormously if you intend to compare two.

And it matters more on this device than on a dearer one. Our scanner measures at two frequencies, 20 and 100 kHz. The lower of those is still too high to isolate the fluid sitting outside your cells, so the machine cannot report an extracellular-to-total-body-water ratio. On a higher-specification InBody that ratio is the standard check when someone’s lean mass appears to jump two kilogrammes in a week: a rising ratio says fluid, not tissue. Here that check simply is not available, so a hydration-driven swing looks exactly like a real change in lean mass with nothing on the sheet to contradict it. The protocol below is what replaces it.

The protocol we ask clients to follow:

• Same time of day, ideally morning
• Fasted, or a consistent three to four hours after food
• No fluid immediately before; empty your bladder first
• No training, sauna or hot bath for 8 to 12 hours
• No alcohol for 24 hours
• Same device, same clothing, bare feet and hands
• A consistent point in the menstrual cycle where practical
• Stand still and settle first

That protocol sets the use case. Take the scan as evidence of direction and rate of change over months, alongside a tape measure, your bodyweight trend and your training numbers. Do not take it as a definitive body fat percentage, an assessment of adiposity, or anything that resolves a single month. ESPEN puts changes under 1.5 to 2 kg inside the noise of bioimpedance, and on this class of device a shift in body fat percentage smaller than roughly 2 to 3 points cannot be confidently separated from measurement noise either (McLester 2020, on the 230, 720 and 770). Plainly: nobody can tell you from a scanner that you gained 0.4 kg of muscle this month. That is noise, and a coach who reads it back as a win is selling you a rounding error.

If you want one accurate snapshot, DEXA is the better tool, and we compare the two in DEXA vs InBody scan. If you want to know whether what you are doing is working, a standardised scan every three to four weeks does that job well, at a fraction of the hassle, provided you read the line rather than the last point. A single three-week gap sits inside the noise. Ten of them, which is exactly what the printout’s Body Composition History holds, is roughly thirty weeks of trend and answers the question properly.

If you want it measured properly

Our scanner sits at the Finchley Road studio, every test on it runs to the protocol above, and each result is logged so the next one is comparable rather than merely new. A coach goes through your sheet with you, marks the figures that carry weight, strikes out the ones that do not, and tells you what would have to shift before either of us calls it progress. Nobody here builds a training block on a 0.4 kg swing. To have your body composition measured on those terms and your programme built around what it shows, book a call.

Free InBody scan until 30 September

We are running complimentary body composition scans at our Finchley Road studio, Monday to Friday between 10am and 2pm. Thirty minutes: the scan itself, plus a coach taking you through what your numbers mean and which ones to ignore. No charge, no obligation, one per person.

Book your free scan or see what’s included.

This article is general information, not medical advice. If you have a medical condition, are pregnant, are taking any prescribed medication, or have a history of disordered eating, speak to your GP or a registered dietitian before changing how you train or eat.

Frequently asked questions

Is an InBody scan accurate enough to replace a DEXA scan?
No. For a single absolute number, DEXA is the reference method. In 93 participants, only 35% of InBody body fat readings fell within 10% error of DEXA (Merrigan 2022). For tracking change under a standardised protocol, it is reasonable and far more practical.

Does an InBody scan overestimate or underestimate body fat?
It underestimates body fat and overestimates fat-free mass. In the one study of our model against DXA, percent body fat read 24.6% versus 27.7%, fat mass 17.8 kg versus 19.7 kg and fat-free mass 53.7 kg versus 51.1 kg, all p<0.0001 (Garcia 2020).

How much does a reading have to change before it is real?
More than you would like. ESPEN puts changes under 1.5 to 2 kg within the noise of bioimpedance, and the minimum detectable change for body fat percentage on this class of InBody is 2.12 to 2.73 percentage points (McLester 2020, tested on the 230, 720 and 770). Judge progress over months.

How often should I rescan?
Every three to four weeks is a good interval for accountability, and it is what we work to, because it builds the trend line and keeps the habit going. Judge progress on the shape of that line across two to three months rather than on any single gap between scans, because one three-week gap sits inside the measurement error. The sheet holds your last 10 tests, so ten three-weekly scans is about thirty weeks of trend on one page.

Does the InBody 270 measure visceral fat?
No. There is no visceral fat area or level on this model’s result sheet. For central adiposity, use waist-to-height ratio alongside BMI, which is what NICE recommends in adults with a BMI under 35.

Why does NICE say not to use bioimpedance instead of BMI?
NG246 recommendation 1.9.9 says not to use it as a substitute for BMI as a measure of general adiposity in adults. That is about substitution, not a ban on using a scanner to track your own progress alongside BMI and waist-to-height ratio.

Do I really need to fast before an InBody scan?
For a one-off scan it makes little difference. For scans you intend to compare, yes. Fluid and food can shift a reading by one to two percentage points of body fat, the same size as a month of real progress, so an unstandardised scan can hide or invent a result.

Sources

• Garcia JR, Manimaleth R, Czartoryski P et al. Journal of Exercise and Nutrition, 2020;3(3):10. https://journalofexerciseandnutrition.com/index.php/JEN/article/view/65
• McLester CN, Nickerson BS, Kliszczewicz BM, McLester JR. Journal of Clinical Densitometry, 2020;23(3):443–450.
• Czartoryski P et al. Journal of Exercise and Nutrition, 2020;3(1).
• Potter AW et al. European Journal of Clinical Nutrition, 2025;79:1235–1244. https://doi.org/10.1038/s41430-025-01664-4
• Merrigan JJ et al. Journal of Exercise and Nutrition, 2022;5(4). https://journalofexerciseandnutrition.com/index.php/JEN/article/view/133
• Kyle UG et al. Clinical Nutrition, 2004. https://www.espen.org/documents/BIA1.pdf and https://www.espen.org/documents/BIA2.pdf
• Looney DP et al. Frontiers in Nutrition, 2024;11:1491931. https://doi.org/10.3389/fnut.2024.1491931
• Antonio J et al. Journal of Functional Morphology and Kinesiology, 2019;4(2):23. https://doi.org/10.3390/jfmk4020023
• Buch A et al. Nutrition & Diabetes, 2022;12:45. https://doi.org/10.1038/s41387-022-00223-1
• Schierbauer J et al. Metabolites, 2023;13(4):473. https://doi.org/10.3390/metabo13040473
• Oshima Y, Shiga T. European Journal of Clinical Nutrition, 2006;60(8):938–941. https://doi.org/10.1038/sj.ejcn.1602402
• Randhawa AK et al. BMC Public Health, 2021;21:495. https://doi.org/10.1186/s12889-021-10552-y
• NICE NG246, Overweight and obesity management. Published 14 January 2025, last updated 8 January 2026. https://www.nice.org.uk/guidance/ng246/resources/overweight-and-obesity-management-pdf-66143959958725

{“@context”: “https://schema.org”, “@type”: “FAQPage”, “mainEntity”: [{“@type”: “Question”, “name”: “Is an InBody scan accurate enough to replace a DEXA scan?”, “acceptedAnswer”: {“@type”: “Answer”, “text”: “No. For a single absolute number, DEXA is the reference method. In 93 participants, only 35% of InBody body fat readings fell within 10% error of DEXA (Merrigan 2022). For tracking change under a standardised protocol, it is reasonable and far more practical.”}}, {“@type”: “Question”, “name”: “Does an InBody scan overestimate or underestimate body fat?”, “acceptedAnswer”: {“@type”: “Answer”, “text”: “It underestimates body fat and overestimates fat-free mass. In the one study of our model against DXA, percent body fat read 24.6% versus 27.7%, fat mass 17.8 kg versus 19.7 kg and fat-free mass 53.7 kg versus 51.1 kg, all p<0.0001 (Garcia 2020)."}}, {"@type": "Question", "name": "How much does a reading have to change before it is real?", "acceptedAnswer": {"@type": "Answer", "text": "More than you would like. ESPEN puts changes under 1.5 to 2 kg within the noise of bioimpedance, and the minimum detectable change for body fat percentage on this class of InBody is 2.12 to 2.73 percentage points (McLester 2020, tested on the 230, 720 and 770). Judge progress over months."}}, {"@type": "Question", "name": "How often should I rescan?", "acceptedAnswer": {"@type": "Answer", "text": "Every three to four weeks is a good interval for accountability, and it is what we work to, because it builds the trend line and keeps the habit going. Judge progress on the shape of that line across two to three months rather than on any single gap between scans, because one three-week gap sits inside the measurement error. The sheet holds your last 10 tests, so ten three-weekly scans is about thirty weeks of trend on one page."}}, {"@type": "Question", "name": "Does the InBody 270 measure visceral fat?", "acceptedAnswer": {"@type": "Answer", "text": "No. There is no visceral fat area or level on this model's result sheet. For central adiposity, use waist-to-height ratio alongside BMI, which is what NICE recommends in adults with a BMI under 35."}}, {"@type": "Question", "name": "Why does NICE say not to use bioimpedance instead of BMI?", "acceptedAnswer": {"@type": "Answer", "text": "NG246 recommendation 1.9.9 says not to use it as a substitute for BMI as a measure of general adiposity in adults. That is about substitution, not a ban on using a scanner to track your own progress alongside BMI and waist-to-height ratio."}}, {"@type": "Question", "name": "Do I really need to fast before an InBody scan?", "acceptedAnswer": {"@type": "Answer", "text": "For a one-off scan it makes little difference. For scans you intend to compare, yes. Fluid and food can shift a reading by one to two percentage points of body fat, the same size as a month of real progress, so an unstandardised scan can hide or invent a result."}}]}