Relative Body Weight and Standardised Brightness-Mode Ultrasound Measurement of Subcutaneous Fat in Athletes: An International Multicentre Reliability Study, Under the Auspices of the IOC Medical Commission

Introduction Fat is a metabolic fuel, but excess body fat is ballast mass, and therefore, many elite athletes reduce body fat to dangerously low levels. Uncompressed subcutaneous adipose tissue (SAT) thickness measured by brightness-mode ultrasound (US) provides an estimate of body fat content. Methods The accuracy for determining tissue borders is about 0.1–0.2 mm and reliability (experienced measurers) was within ± 1.4 mm (95% limit of agreement, LOA). We present here inter- and intra-measurer scores of three experienced US measurers from each of the centres C1 and C2, and of three novice measurers from each of the centres C3–C5. Each of the five centres measured 16 competitive adult athletes of national or international level, except for one centre where the number was 12. The following sports were included: artistic gymnastics, judo, pentathlon, power lifting, rowing, kayak, soccer, tennis, rugby, basketball, field hockey, water polo, volleyball, American football, triathlon, swimming, cycling, long-distance running, mid-distance running, hurdles, cross-country skiing, snowboarding, and ice hockey. SAT contour was detected semi-automatically: typically, 100 thicknesses of SAT at a given site (i.e., in a given image), with and without fibrous structures, were measured. Results At SAT thickness sums DI (of eight standardised sites) between 6.0 and 70.0 mm, the LOA of experienced measurers was 1.2 mm, and the intra-class correlation coefficient ICC was 0.998; novice measurers: 3.1 mm and 0.988. Intra-measurer differences were similar. The median DI value of all 39 female participants was 51 mm (11% fibrous structures) compared to 17 mm (18%) in the 37 male participants. Discussion DI measurement accuracy and precision enables detection of fat mass changes of approximately 0.2 kg. Such reliability has not been reached with any other method. Although females’ median body mass index and mass index were lower than those of males, females’ median DI was three times higher, and their percentage of fibrous structures was lower. The standardised US method provides a highly accurate and reliable tool for measuring SAT and thus changes in body fat, but training of measurers is important. Electronic supplementary material The online version of this article (10.1007/s40279-019-01192-9) contains supplementary material, which is available to authorized users.


Ad: Obtainable accuracy of US thickness measurements
The accuracy obtainable with brightness-mode (B-mode) US depends on the probe frequency (high frequency increases resolution, but reduces image depth due to increased attenuation), on the appropriate setting of the US system (primarily: gain, time-gain compensation, image depth, number and position of foci, image analysis parameters), and on the skills of the investigator (small movements or tilts of the probe can change image quality substantially). Linear probes should be used for quantitative measurements. Tissue compression can be avoided by including a thick layer of gel between the probe and the skin [3,6]. The resolution of US imaging is determined by the US wavelength (λ): at 18 MHz probe frequency (f), which is the best choice for the thin subcutaneous adipose tissue layers found in most athletes, a resolution approximately equal to the wavelength (0.1 mm) can be obtained because diffraction and spatial length of the US pulse are the limiting factors for transverse and longitudinal resolution: = / = 1450 m s −1 /1.8 • 10 7 s −1 ≈ 0.1 mm. A sound speed deviation from the real speed in the given tissue of 30 ms -1 would result in 2% thickness measurement error. For example, a 10 mm thick layer would erroneously be measured as 10.2 mm. Sound velocity data can be found in [18][19][20][21] and detailed discussions of the US thickness measurement accuracy can be found in [3,6,22].
Note: It is not the technically given high accuracy that limits the usefulness of the US method, but rather biological reasons (furrowed borders and visco-elastic deformations of adipose tissue) mostly affect the reliability. The technically given accuracy of about 0.1 mm cannot be outperformed by any other method (including mechanical measurements of tissue thickness layers using a micrometre screw [2]).

Choice of US sites
Deviations of repeated measurements depend primarily on the micro-anatomical structure at a given site (provided that the measurers are trained sufficiently). In most cases, these deviations are larger than the technically given accuracy limitations. To obtain the highest possible reliability, it is of paramount importance to choose US sites where SAT thickness is mostly constant around the centre of the site, to choose sites that can be easily and reliably marked, and to capture the US images in standardised positions (because fat is visco-elastic) [3,4,6,7]. The sites selected for the standardised US measurement of SAT obey these criteria and they represent trunk (3 sites), arms (2 sites), and legs (3 sites).

Linear transformation of SAT thickness measurement error into fat mass error
The small errors resulting from the accuracy and reliability limitations of US thickness measurements of SAT layers transform linearly into the error of subcutaneous adipose tissue mass mSAT because the fat volume is proportional to the ( For this approximate calculation, a factor of 0.7 was used for calibrating the mean SAT thickness derived from the eight sites [22] (this considers that the standardised sites over-represent the mean SAT thickness because these sites were chosen to represent typical fat patterning, but not mean SAT thickness). From this approximate assessment of fat mass, we learn that a measurement error of 1.4 mm (95% LOA; see Tables 2   and 3) transforms into a SAT mass error of less than 0.2 kg.

Body fat measurements in sport:
In all computed tomography methods, the accuracy of volumetry (and thus the determination of fat mass) depends strongly on the choice of the image segmentation parameters and the measurement parameter settings. Additionally, MRI pixel size used for total body scans (typically 1.3-2.0 mm) is not small enough for measuring the thin fat layers in lean athletes with sufficient accuracy. X-ray computer-tomography (CT) cannot be used for routine measurements because of the high radiation exposure.  [2,8,27], which is based on a highly reductionist measurement concept that aims to assess body fat mass by simply measuring the alternating current resistance (e.g., at 50 kHz) of the human body [2,27].
BMI and the derivation of the MI formula for relative body weight: The body mass index (Quetelet's index) BMI = ℎ 2 ⁄ (m body mass, h stature) is not useful for assessing body fat of individuals, particularly in athletes (compare to Fig. 2a in the main text). When using the BMI for assessing 'relative body mass', there is a further important limitation: the BMI ignores individual body properties [14]. Therefore, W. Müller has developed an improved measure for relative body weight, which considers the sitting height s (and thus, implicitly, also the leg length l): the mass index MI [15,16,17]. The MI is a modified BMI. The general formula for this modified BMI is: The exponent k weights the modification term. The Cormic index C = /ℎ. At a given h, a large s is associated with a low l (according to h=s+gl; g is a geometry factor of the individual). Therefore, both s or l can be used for designing a measure that considers the individual's geometric properties. C = 0.53 represents mean sitting height (this reference value is chosen in the middle of the Cormic index continuum).
The choice of = 1 (thus 'MI1') takes both into consideration, the dependency of relative body weight on stature h and on sitting height s: MI1 is chosen symmetrically between MI 2 = 0.53 2 · / 2 ( = 2 would ignore the impact of stature h on relative body mass as h would cancel out in this case), and = 0 ignores the impact of sitting height s (and thus of leg length, too): MI 0 = /ℎ 2 ≡ BMI. For a person with long legs, the MI is higher than the BMI, and vice versa for a person with short legs. With the same stature h, a person with shorter legs (large sitting height) can be expected to have higher body mass m because his volume is higher due to the relatively large dimensions of his upper body. Particularly in low weight sports, assessing 'relative body mass' of athletes is a crucial health parameter.
A BMI of 17.5 kgm -2 is one of the four criteria for diagnosing anorexia nervosa: when using the MI instead of the BMI, different diagnoses would result whether the MI is equal to the BMI (due to mean leg length), or differs by one or even more units. In the group of elite athletes described in this publication, differences MI-BMI ranged from -1.7 kgm -2 to +1.3 kgm -2 . At a BMI of 17.5 kgm -2 , a difference of -1.7 kgm -2 would result in 15.8 kgm -2 , which is far below the weight criterion for anorexia nervosa, whereas +1.3 kgm -2 would result in 18.8 kgm -2 , which is far above this criterion, and even above the WHO criterion for underweight (which is 18.5 kgm -2 ). In addition to using the MI, the accurately measured SAT amount should be included in diagnostics and therapeutics of low weight problems in athletes and in anorexia nervosa patients [9].
The MI is defined such that the WHO cut-off points for underweight, overweight, and obesity (18.5, 25, and 30 kgm -2 ) can remain the same when replacing the BMI by the MI: their means are equal for groups with a mean Cormic Index of C = 0.53. The most important advantage of the MI over the BMI is the appropriate assessment of relative weight of the individual, although group means may be similar or the same. In the group of mainly Caucasians (and a small number of Hispanics) studied here, median BMI was 22.6 kgm -2 , and median MI 22.2 kgm -2 . Using the MI1 instead of the BMI will also contribute to the discussion about BMI cut-off points for ethnic groups with shorter or longer legs [26]. These populations show large differences between MI and BMI. The same abbreviations as in Table A2 are used here. Additionally: w (weightsensitive), and nw (non-weight-sensitive). Ordered according to increasing body mass index (BMI).

Table A4: Thickness value differences ABS(δE) at the individual sites
Absolute values of differences of the three measurers from their mean values for thickness measurements with fibrous structures excluded (index: "E"). For number of comparisons and abbreviations see Table A3. Data correspond to Figs. 6c and 6d.