Abstract
Dose scaling is a critical component of dose individualisation that helps ensure that the required amount of drug exposure is achieved in each and every patient. This is typically achieved by scaling the pharmacokinetic (PK) parameters by relevant covariates, on the assumption that variability in drug response is linked to PK. Body size is the most important covariate when ๐ถ๐ฟ is scaled for differently sized populations, e.g. obese or paediatric patients, in order to calculate the maintenance dose. Conventionally, total body weight (๐๐) has been used as the size scaler, however, fat-free mass (๐น๐น๐) is a suitable alternative as it accounts for body composition as well.
Accurate scaling of ๐ถ๐ฟ by body size needs an accurate value of the scaling exponent (i.e. the allometric exponent). Since there is no consensus on the true value of the exponent, it is often empirically chosen from the a priori recommended values (2/3, 3/4, 1) or estimated. In either situation, the accuracy of the empirical value would depend on the study design, which is assessed in this thesis by modelling and simulation. The finding indicates that sub-optimal study designs risk bias to the allometric exponent, which in turn risks bias to ๐ถ๐ฟ prediction.
For sub-optimal study designs, it is recommended to use a biologically plausible a priori value for the allometric exponent. While a theoretical value exists for ๐๐, it is currently lacking for ๐น๐น๐. Therefore, the value for ๐น๐น๐ was investigated by modelling the relationship between liver size and ๐น๐น๐. Further attempt was made to empirically estimate the exponent from literature data. A literature based meta-analysis revealed a disconnect between the theoretical (expected) and the empirical values; this needs further research.
Given the issue with the choice of the allometric exponent, an alternative โbottom-upโ approach could be used to scale ๐ถ๐ฟ by in vitro-in vivo extrapolation (IVIVE). Unlike the classical โtop-downโ approach above, the accuracy of scaling ๐ถ๐ฟ by IVIVE would depend on choosing an accurate descriptor of โfunctionalโ liver size, instead of study design. In this thesis, lean liver volume (๐ฟ๐ฟ๐) was identified as a more accurate descriptor of โfunctionalโ liver size by comparing its predictive performance with the conventional descriptor, total liver volume (๐ฟ๐).
Another issue is that ๐น๐น๐ is not a readily measurable covariate and hence, accuracy of ๐น๐น๐ prediction is of utmost importance in dose scaling. Janmahasatianโs ๐น๐น๐ model, although extensively used, is known to over-estimate ๐น๐น๐ in Indian population. Therefore, an extended version of Janmahasatianโs ๐น๐น๐ model was developed for Indians. The extended ๐น๐น๐ model structure includes (estimable) ethnicity specific body composition parameters, which can be further estimated for other relevant populations as well.