Background/ObjectivesSeveral studies have addressed the validity of ultrasound (US) for body composition assessment, but few have evaluated its reliability. This study aimed to determine the reliability of percent body fat (%BF) estimates using A-mode US in a heterogeneous sample.Subjects/MethodsA group of 144 healthy adults (81 men and 63 women), 30.4 (10.1) years (mean (SD)), BMI 24.6 (4.7) kg/m2, completed 6 consecutive measurements of the subcutaneous fat layer thickness at 8 anatomical sites. The measurements were done, alternatively, by two testers, using a BodyMetrix™ instrument. To compute %BF, 4 formulas from the BodyView™ software were applied: 7-sites Jackson and Pollock, 3-sites Jackson and Pollock, 3-sites Pollock, and 1-point biceps.ResultsThe formula with the most anatomic sites provided the best reliability quantified by the following measures: intraclass correlation coefficient (ICC) = 0.979 for Tester 1 (T1) and 0.985 for T2, technical error of measurement (TEM) = 1.07% BF for T1 and 0.89% BF for T2, and minimal detectable change (MDC) = 2.95% BF for T1, and 2.47% BF for T2. The intertester bias was −0.5% BF, whereas the intertester ICC was 0.972. The intertester MDC was 3.43% BF for the entire sample, 3.24% BF for men, and 3.65% BF for women.ConclusionsA-mode US is highly reliable for %BF assessments, but it is more precise for men than for women. Examiner performance is a source of variability that needs to be mitigated to further improve the precision of this technique.