The Pitfalls of Body Fat “Measurement”, Part 6: Dual-Energy X-Ray Absorptiometry (DEXA)

In the first five parts of this series, I discussed various 2-compartment models of estimating body composition.  These models divide the body into fat and fat-free mass.  Now let's talk about a common 3-compartment model:  dual-energy x-ray absorptiometry (DEXA).


DEXA, once used for determining bone density, evolved into a technique for also estimating body composition.  DEXA represents a 3-compartment model for estimating body composition, because it can divide the body into 3 compartments:  fat, bone mineral, and all other fat-free mass that does not include bone.  Thus, unlike 2-compartment models, DEXA is not subject to errors caused by variations in bone density among different ethnicities.  However, there are other sources of error which I will discuss in this article.

DEXA offers a number of advantages.  Improved technology has dramatically decreased scan times (years ago, a DEXA scan would take 20-25 minutes; now scans can take 5-10 minutes), so the method is convenient and non-invasive.  DEXA can provide bone density estimates, and can provide regional estimates of body composition (meaning it can estimate the body composition of individual parts of your body).  DEXA works by measuring your body's absorbance of x-rays at two different energies.  Fat, bone mineral, and fat-free soft tissue have different absorption properties.  Thus, we can get estimates of your body composition by scanning your entire body.

DEXA Sources of Error

Like all other body fat estimation techniques, DEXA has numerous sources of error.  There can be inconsistent results between different machines from different manufacturers, and even different results between machines from the same manufacturer.  Software upgrades can change the algorithms that the device uses to calculate body composition.  Different hardware and software configurations can result in different interpolations of soft tissue over bone, and different treatment of pixels of which a small portion is bone.  The type of X-ray beam (fan beam or pencil beam) can also be a source of error; DEXA machines with fan beams can suffer from beam magnification (also known as parallax error).  A final source of error is the same error that all 2-compartment models also have:  the hydration of fat-free mass.  In fact, a 5% variation in fat-free mass hydration can change your DEXA-determined body fat percentage by nearly 3%.  This can be a problem when comparing different ethnicities or body types where fat-free mass hydration can vary.  It can also be a problem when trying to measure change over time.

DEXA vs. 4-Compartment Models

A number of studies have compared DEXA to 4-compartment models.  When looking at group averages, DEXA does pretty well, with errors of 1-2%.  However, like with all other body fat testing techniques, individual error rates can be much higher.  The error rates vary by what study you look at and which DEXA machine was used; error rates range from 4% in one study to up to 8-10% in another study. Also, the accuracy of DEXA can be affected by sex, size, fatness, and disease state of subjects.

There are also issues when trying to use DEXA for tracking change over time.  In one study, researchers wrapped lard around the legs of subjects to simulate weight gain.  However, this changed the apparent bone mineral content, questioning the ability of DEXA to accurately assess body composition during weight change.   In another study on bodybuilders, DEXA was reasonably accurate when looking at the group change in body fat percentage, but the individual error rate was as high as 4%.  Thus, you could decrease your body fat percentage by 4% yet DEXA might show no change, or DEXA might show a 4% change when you didn't really have a change.  Another study showed DEXA to overestimate decreases in percent body fat, while it underestimated increases in percent body fat.  In fact, the overestimations got as high as 5%.  Finally, here's a chart from a study that compared changes in body fat over time using DEXA to using a 4-compartment model:

Change in % body fat, comparing DEXA to a 4-compartment model

The X-axis of this chart shows the change in % body fat for DEXA, while the Y-axis shows the change for the 4-compartment model.  You can see there was very little agreement between the methods.  For example, one person showed an increase in body fat percentage by 5% with the 4-compartment model, but a decrease by 5% according to DEXA.  Another person had nearly a 10% loss of body fat according to the 4-compartment model, but only a 3% decrease according to DEXA.  These results question the reliability of DEXA for measuring change over time in individuals.

DEXA:  The Verdict

Despite the fact that DEXA represents a 3-compartment model, its error rates are no better than hydrostatic weighing, and in some cases is worse.  Like other techniques, DEXA does well when looking at group averages, but not so well when looking at individuals.  Individual error rates tend to hover around 5%, although some studies have shown error rates as high as 10%.  When looking at change over time in individuals, error rates have hovered around 5% in some research, although other research has indicated DEXA to perform much more poorly.  For these reasons, I do not recommend DEXA for tracking change over time in individuals.  If you do use DEXA for tracking change over time, I recommend very long time periods between measurements (a minimum of 3-6 months), as you will need a minimum of a 5% change in body fat to reliably detect a true change in body fat in most people.

That sums it up for the most widely available techniques for determining body composition.  In the final part next week, I will discuss the practical application of everything you have learned in this series.

104 Responses to “The Pitfalls of Body Fat “Measurement”, Part 6: Dual-Energy X-Ray Absorptiometry (DEXA)

  • Where is the article that ties all this together. If all these methods are so unpredictable, should I even bother with body fat monitoring. It is nice to know the error ranges for these – especially BIA because most health clubs use it. I don’t see any value in it now.

    So, what should I do? Just measure my waist circumference and weight myself? Look in the mirror?

    • Patrick
      2 years ago

      waist circumference is the most relevant measure for BF. For men, BF accumulates at the belly. Reducing your belly size or waist measure, regardless of general fitness or muscle development, is the most reliable measure of your actual BF%

      For a typical male 5’10”, a 32″ waist equates to a 18-20% BF
      32.5″ might be 21-22%
      34″ would be closer to 28-30%
      36″ would be 35-37%

      regardless of muscle development, as fat pools mainly in the gut.
      muscle development is in the limbs and trunk.

      • Not true . You can be densely muscled and have overly developed obliques and have a thicker midsection . Doesn’t mean your bodyfat is that high lol that’s stupid . Calipers are most reliable method in predicting subcontaneous boydfat.

  • Douglas Ravski Pereira
    2 years ago

    +1 to above comment. Would love to see the last article here that ties it up with a recommendation

  • Walking Walnut
    2 years ago

    Hello? Can we still consider th DXA scan so wildly inaccurate since the tests sited are from 1999,2003,2006,2007? Technology does improve over time.

    • Yes…even more recent comparisons haven’t made things any better. It’s because no matter how much you improve the technology, DEXA is still based on a number of assumptions that don’t necessarily hold true (like the hydration of fat-free mass).

  • Archie Williams
    1 year ago

    Having performed nearly 3000 DEXA scans, the error rates stated are a load of rubbish. I have never done one scan where the results didn’t make sense (i.e no change in body composition but 4% down on scan results). This just doesn’t happen-not even anywhere close to it. I would put money on doing a back to back scan and the results being within grams difference (the bf% wouldn’t change more than 0.1%)

    You can only test how ‘out’ a method is by knowing the exact true value in the first place so how is this found to know if there’s an error.

    Maybe whoever wrote this article should get a DEXA and then return to the same machine and get another one. There’s variations between machines for sure as they are using different calibrations but with the same machine, reproducibility is very very close.

    • Archie, how many DEXA scans you’ve done is irrelevant. Since you don’t have a research-based gold standard to compare to (like a 4-compartment model), you cannot state that DEXA reliably estimates change in body composition within an individual over time. You need a reference standard to make that claim.

      Doing back-to-back scans is irrelevant. That only determines test-retest reliability…i.e., how close the estimates are when repeated measurements are made. It says nothing on whether those measurements accurately determine body fat percentage. Test-retest reliability is NOT the same thing as accuracy.

      Please reread the article. To determine the accuracy of change over time of DEXA, you compare it to the reference method, which is a 4-compartment model. Numerous scientists have done this, and I have referenced some of their studies in this article. DEXA isn’t as good as people think on an individual level, and there are reasons why (which I also mention in the article, like changes in FFM density).

  • As part of a weight/fat loss challenge, I had my first Dexa done 5 weeks ago. Last week I decided to repeat the test to track my progress. The results showed that I gained 1 pound of muscle and lost 1.5% fat in 4 weeks, almost proportionally in the entire body. However, I also gained 2% of fat in the right arm. I was on 34.5% and am 33% right now.

    I can tell by touching myself that my arms are super strong. In fact, right before the second test I asked my coach to change the exercise routine for arms because they are getting bulky. The point is, how is it possible to gain and lose fat at the same time while being in calorie deficit? When I asked the guy from Dexa the reason for that he answered that the Dexa machine does not lie.

    Any thoughts?

    • You can still have elevated muscle protein synthesis in a deficit, which means you can gain muscle. The amount you gain won’t be optimal, though, and the leaner you get, the less likely you’ll simultaneously put on muscle while losing fat.

  • I had a full body DEXA scan back in March and it came out with a BF% of 31.9 which is much higher than an electrical resistance test (~22%) had showed the week prior, but I took the DEXA to be the ‘gold standard’ in BF measurement. A few months down the line, after increasing training intensity, I went back in for a repeat of the DEXA and it came out at 25.2%. However, I was told that the algorithm that had been used on this particular machine was now known to be incorrect so not to take either measurement at face value (I should mention this was free!).

    I’ll have to get a PT to do the caliper measurement and electrical resistance test and see how far out the DEXA was, but my thoughts are that it was overestimating my BF by a few percentage points at least. Vascularity is back and I’ve lost a notable amount of waist size so I would like to think that the recent 25.2% was an overestimate. To summarise, the DEXA might be a very advanced and capable machine, but if it’s not operated correctly the results could still be off.

  • 5% inaccuracy is good enough for most of us. Who cares if DXA shows 10% body fat and in reality it is 10.5% or 9.5%?!

    • No, it’s 5 percentage points. So if DXA showed 10%, it could be 5% or 15%.

      • It is stated as +/- 3% to 5% of fat free mass. Hence measure the ffm Error maximum is 5% of the measured value. Stefan is correct. I assume you do not understand error calculations.

        • Barry, you are incorrect. Please read the papers I referenced in the article. They are percentage points. I even show a Bland-Altman plot in the article.

  • Well there are a few other body fat measurement techiques not mentioned here namely CT and MRI.

    CT is the most accurate and shown to be 1.3% within cadaveric studies eg they did a CT and measured body fat and then cut all the fat off a dead guy and compared the two. I’d believe the CT was even more accurate as they probably couldn’t cut all the fat off the cadaver without dropping some on their clothes/floor.

    This makes sense since CT gives a 3d volumetric view of your body with densities of those sections which is easily correlated to fat/muscle/bone/water/etc.

    DXA is the gold normal standard and compares extremely well with CT. Sure there are errors which are probably related to people that don’t fit inside the DXA machine but it’s way better than the other methods.

    • CT is not practical, cost effective, and there’s the issue with radiation exposure.

      DXA is not the normal “gold standard.” Only a 4-C model is the gold standard. The errors go well beyond people not fitting in the machine. I described much of the sources of error in this article (like assumptions about the density of FFM).

  • I had a DEXA it said I was up a pound in fat and a pound in muscle since my last one. I feel this is incorrect since I gained 1/2 inches on my arms and lost 2 1/2 inches off my waist.

Trackbacks & Pings

Leave a Reply to Kelly Cancel reply

Your email address will not be published. Required fields are marked *