Background: When conducting population pharmacokinetic (PK) analyses, total body weight (WT) is often chosen as a covariate to explain between subject variability in clearance (CL). Given that 99% of the body’s clearance takes place within lean tissues, it has been suggested that WT is unlikely to be suitable for obese subjects and lean body weight (LBW) should be considered as a biologically more plausible covariate .
Aim: The aim of this study was to determine the probability of choosing LBW over WT as a covariate on CL for trial designs that stratified subject inclusion based upon varying body mass index (BMI) or WT.
Methods: Six clinical trials were constructed. Each trial had a different range of inclusion criteria for BMI or WT, from normal weighted to overweight, to normal weighted to morbidly obese. 60 subjects matching these criteria were randomly selected from a database of 152 subjects pooled from 2 prior population PK studies [2,3]. For each of the 6 trial designs, 1000 bootstrapped datasets were created and the 2 competing covariate models (LBW or WT on CL) fitted to each bootstrapped dataset. The structural model (a 2 compartment linear model) was fixed at the previously identified best model . The differences in the objective function values (ΔOBJ = OBJLBW – OBJWT) between the 2 models were computed for those datasets where both the LBW and WT models converged successfully. Negative values of ΔOBJ indicated that the LBW model was preferred to the WT model. The probability of LBW being preferred was computed as the ratio of the number of negative ΔOBJ values to the number of successful runs.
Results: Of the 6 trial designs that stratified on BMI, the trial with the smallest sized subjects had a BMI from 14.9 to 25.4 kg/m2 while the trial with the largest sized subjects had a BMI that ranged from 15.7 to 44.1 kg/m2. The probability of LBW being preferred over WT increased from 0.830 to 0.968 for the datasets that included BMI up to 25.4 kg/m2 and 44.1 kg/m2 respectively.
Of the 6 trial designs that stratified on WT, the trial with the smallest sized subjects had a WT from 41 to 80 kgwhile the trial with the largest sized subjects had a WT that ranged from 43 to 131 kg. The probability of LBW being preferred over WT increased from 0.672 to 0.940 for the datasets that included WT up to 80 kg and 131 kg respectively.
Conclusion: The probability of choosing a model with LBW on CL compared to a model with WT on CL increases as larger sized subjects were allowed to be recruited in the trial. Typical Phase III studies that do not stratify according to weight or exclude obese subjects are less likely to determine that LBW is the preferred covariate for CL.
References:  Green B, Duffull SB. Br J Clin Pharmacol. 2004;58:119-33.  Green B, Duffull SB. Br J Clin Pharmacol. 2003;56:96-103.  Barras M et al. J Am Coll Cardiol. Under Review