Purpose: The implications of undertaking DXA scans using best practice protocols (subjects fasted and rested) or a less precise but more practical protocol in assessing chronic changes in body composition following training and a specialized recovery technique were investigated. Methods: Twenty-one male cyclists completed an overload training program, in which they were randomized to four sessions per week of either cold water immersion therapy or control groups. Whole-body DXA scans were undertaken with Best Practice (BEST) or Random Activity (RANDOM) protocols at baseline, after 3 weeks of overload training and after a 2 week taper. Magnitudes of changes in total, lean and fat mass from baseline-overload, overload-taper and baseline-taper were assessed by standardization (?mean/SD). Results: The standard deviations of change scores for total and fat-free soft tissue mass (FFST) from RANDOM scans (2-3%) were approximately double those observed in the BEST protocol (1-2%), owing to extra random errors associated with RANDOM scans at baseline. There was little difference in change scores for fat mass. The effect of cold water immersion therapy on baseline-taper changes in FFST was possibly harmful (-0.7%; 90% confidence limits ᱮ2%) with BEST scans but unclear with RANDOM scans (0.9%; Ხ0%). Both protocols gave similar possibly harmful effects of cold water immersion therapy on changes in fat mass (6.9%; ᱳ.5% and 5.5%; ᱴ.3%, respectively). Conclusions: An interesting effect of cold water immersion therapy on training-induced changes in body composition might have been missed with a less precise scanning protocol. DXA scans should be undertaken with the Best Practice Protocol.