Accuracy and precision of the the doubly labeled water (DLW) measurements of energy expenditure are commonly estimated with models that do not account for 2H and 18O natural abundance variation. A new and simple treatment that completes the ratio-product model is derived in the present study. It uses both linear regression of multipoint data with time and information on within-subject 2H and 18O natural abundance variation. Use of this treatment is demonstrated in a group of seven subjects in whom 2H and 18O natural abundance variation was assessed by collecting predose samples for 6-8 days before dosing. In this set of 16 measurements, the precision (coefficient of variation) of individual DLW estimates of CO2 production was 4.90 +/- 2.14%, of which 3.23 +/- 1.20% arose from natural abundance variation.