Ultrasound-derived equations incorporating both muscle thickness and cross-sectional area estimated whole-body muscle mass in healthy Caucasian adults with high accuracy, with the top model achieving a standard error of estimate of 1.7 kg compared with magnetic resonance imaging, according to a recent study.
Researchers developed and validated ultrasound-derived equations to estimate whole-body muscle mass using magnetic resonance imaging (MRI) as the reference standard. The study included 211 healthy Caucasian adults (age 42.0 years [29.0–58.0], 52% female) who underwent MRI and ultrasound examinations on the same day. Eight muscle thickness and seven cross-sectional area (CSA) measurements were assessed across the arm, trunk, and leg. The participants were randomly assigned to development (two-thirds) and cross-validation (one-third) groups.
Using stepwise multiple regression, the researchers generated models based on muscle thickness, CSA, and a combination of both. The models combining muscle thickness and CSA showed the best performance based on adjusted R² and standard error of estimate (SEE). In the development group, the combined model (Model 5) achieved an adjusted R² of 0.948 with a SEE of 1.6 kg. In cross-validation, this model showed a bias of −0.03 kg, an intraclass correlation coefficient of 0.965, and limits of agreement ranging from −3.99 to 3.92 kg.
When applied to the full sample, the most accurate equation included 10 variables: sex, weight, body mass index, and seven ultrasound muscle measurements. These comprised muscle thickness of the forearm extensor, rectus abdominis, rectus femoris, biceps femoris, and tibialis anterior muscles; along with CSA measurements of the triceps brachii and tibialis anterior muscles.
The most practical equation included six variables—sex, height, and four ultrasound measurements—but showed lower accuracy, with an adjusted R² of 0.927 and a SEE of 2.0 kg, reflecting a trade-off between accuracy and measurement burden.
Models combining muscle thickness and CSA measurements demonstrated improved prediction accuracy, with higher adjusted R² values and lower SEE compared with models using muscle thickness alone.
Ultrasound measurements were performed using reference lines and extended-field-of-view imaging based on established protocols. Intra-rater reliability ranged from 0.75 to 0.98. Muscle mass on MRI was derived from semimanual segmentation of muscle volume and conversion using a density factor.
The researchers also evaluated the use of doubled right-sided measurements to estimate whole-body muscle mass. Although a statistically significant difference was observed, the standard error of measurement was 0.2 kg, corresponding to less than 1% of total muscle mass and considered not clinically relevant.
The study was conducted in a healthy Caucasian population, and the researchers noted that the developed equations can only be applied to this group. Further research is needed to determine their applicability in nonhealthy populations. Measurements of muscles with complex morphology, including the triceps brachii and tibialis anterior, may be prone to inter-rater variation, particularly when anatomical landmarks are less clearly defined. In addition, the limitations of MRI required the use of unilateral measurements in some of the participants.
Previous ultrasound-derived equations have been based solely on muscle thickness measurements, and MRI-based equations have been developed in Asian samples.
“The addition of CSA measurements—specifically of the triceps brachii and tibialis anterior—to models based on muscle thickness resulted in improved accuracy, as reflected by higher R2 values and lower SEE,” said lead study author Jona Van den Broeck, of the Experimental Anatomy Research Group in the Department of Physiotherapy and Human Anatomy at the Vrije Universiteit Brussel in Belgium, and colleagues.
The authors reported no conflicts of interest.
