Do Subsamples Provide Reproducibility Estimates Comparable to the Full Sample in Trial-Based Ultrasound Muscle Measurements?

Reproducibility estimates are commonly used to determine whether ultrasound-derived muscle changes exceed measurement error, yet they are often obtained from subsamples for practical reasons. This study compared reproducibility estimates of ultrasound-derived muscle morphological and architectural measurements calculated using the full sample versus subsamples of different sizes. One hundred ultrasound image sequences of the vastus lateralis were analyzed from the right and left limbs of 50 healthy adults participating in a randomized controlled trial registered at the Brazilian Registry of Clinical Trials (RBR-57v9mrb). Muscle cross-sectional area, muscle thickness, fascicle length, and pennation angle were assessed in a test-retest design separated by 72 hours. Reproducibility was quantified using the typical error expressed as a coefficient of variation. Monte Carlo simulations with 1,000 iterations assessed estimate stability. Mean subsample estimates were generally close to full-sample values, but dispersion increased as sample size decreased, indicating lower stability and precision. Under the operational equivalence criterion, only the 90% subsample condition was equivalent to the full-sample reference across all four outcomes.