Estimation of Precision in DLW Studies Using the Two-Point Methodology

Previous attempts to estimate precision of doubly labeled water (DLW) estimates of CO₂ production, using propagation of error analyses, have necessarily made simplifying assumptions which may compromise the resultant error estimate. Using an empirical iteration approach, error distributions for the DLW calculation were generated which overcome these problems. The error distribution for CO₂ estimates generated by DLW is symmetrical but not normal. The distribution is significantly truncated such that the 99% inclusion limits are 2.034 SD and not 2.58 SD. The precision error (99% CI for mean as percent of the mean) in DLW experiments, when using duplicate analyses, varies between approximately 3% and 47% depending on the ratio of the elimination constants of the two labels (k_o/k_d), experimental duration and initial isotope dose. The error could be improved by approximately 10 fold by increasing the number of replicates at all six isotope determinations from 2 to 5. Estimating precision in actual experiments can be made using the same empirical approach. The resultant estimates can be of extreme value in evaluating the role of precision as a factor influencing deviations during validation studies, and also for weighting mean estimates in applications of the technique. The deviations of DLW estimates from those made simultaneously by indirect calorimetry in a small mammal validation study were generally greater than the precision of the DLW estimates of CO₂ production. This may indicate there are more problems with the technique than precision alone.