We have recently reported a validation study of a prototype low-field strength quantitative magnetic resonance (QMR) instrument for measurement of human body composition (EchoMRI-AH). QMR was very precise, but underreported fat mass (FM) by 2-4 kg when compared to a 4-compartment (4C) model in this cross-sectional study. Here, we report the performance of an updated instrument in two longitudinal studies where FM was decreasing. Healthy obese volunteers were given a modest energy deficit diet for 8 weeks (study A) and obese patients with heart failure and/or at high cardiovascular risk were prescribed a low energy liquid diet for 6 weeks (study B). FM was measured at the start and end of these periods by QMR, dual-energy X-ray absorptiometry (DXA) and 4C. A higher proportion of the weight lost came from fat in study A compared with study B, where loss of total body water (TBW) played a greater part. The intraclass correlation between QMR and 4C estimates of FM loss (DeltaFat) was 0.95, but 20 of 22 estimates of DeltaFat by QMR were lower than the corresponding estimate by the 4C model. Bland-Altman analysis demonstrated that estimates of FM loss by QMR were ~1.0 and 0.7 kg lower than those obtained with 4C (P = 0.0008) and DXA (P = 0.049), respectively. Measurement precision remained high. QMR measurement should prove valuable for quantifying modest changes of FM in small trials.