Tacrolimus and sirolimus are immunosuppressive drugs used to prevent graft rejection. Marked pharmacokinetic (PK) variability has been reported for both, which makes therapeutic drug monitoring potentially useflil. Although PK profiles of both drugs have been characterised, and provide useful information to help dose individualisation, a number of issues require further investigation. The global aim of this PhD thesis was to use modelling approaches to investigate how individualisation of drug therapy for tacrolimus and sirolimus could be used for maximum benefit.
For tacrolimus, concentration-time data, which arose as part of a previous study were analysed using a population PK approach and the resulting PK model for tacrolimus in liver transplant was used to (1) define appropriate sampling times for monitoring tacrolimus and (2) perform sequential population PK-pharmacodynamic (PK-PD) modelling. Only data from patients with stable conditions were used. A two-compartment model with proportional residual error variance provided the best fit to the data. Population PK parameters were: CL = 27 L/h, Vc = 388 L, Q = 55 L/h, Vp = 3580 L, and ka = 1.06 h-1. Between subject variability for CL, Vc and ka was 43%, 39%, and 70%, respectively. Between occasion variability for CL and Vc was 16% and 61%, respectively.
The most informative sampling times for monitoring tacrolimus was determined using two techniques:- a limited sampling strategy (LSS) and sensitivity analysis (SA - a model-based approach). For LSS, a linear mixed effects modelling approach (NONMEM) was used to estimate regression coefficients when Co, Ci, C2, C4, interpolated C5, and interpolated Ce were used as predictors of AUCo-6. Sensitivity analysis was performed using the PK structural model described above (MATLAB®). Results supported the use of a single blood sample at 5 hours post-dose as a predictor of both AUCo.6 (LSS) and AUCo-12 (SA). A Jackknife procedure was used to evaluate the predictive performance of the model, and this demonstrated that collecting a sample at 5 hours was the optimal sampling time for predicting AUCo-6 and also AUCo-i2-
The association between markers of tacrolimus exposure (predicted values of Co, Ci, C2, C4, Ce/g and AUC0-12) and the occurrence of hypertension, hyperkalaemia, hyperglycaemia, and nephrotoxicity were assessed using a sequential population PK-PD approach. Possible relationships between hypertension, hyperkalaemia, and hyperglycaemia and tacrolimus exposure was initially screened (NCSS). Linear regression using a mixed effects modelling approach was used to assess the relationship between exposure and renal function (NONMEM). No significant association between markers of tacrolimus exposure and adverse effects could be found in these patients.
Sirolimus concentration-time data arising as part of routine care from 25 kidney transplant patients were analysed under a fully conditional Bayesian framework (WinBUGS) with informative priors in order to estimate the parameter values and possible influence of covariates. One- and two-compartment models were fitted to the data. A two-compartment model with proportional residual error provided the best fit to the data. The population median value of CL was 12.94 L/h at a median age of 44 years. A negative correlation was found between age and CL, with the regression coefficient (slope) of In(CL) = -0.2987. The posterior probability was 0.734 that there was a clinically significant effect of age on CL.
Data available from the same patients were empirically screened for possible relationships between sirolimus dose or observed conentration and outcomes (WBC, PLT, and HCT). Pooled naive data analysis was performed (NCSS®). Results showed that WBC count and HCT were significantly lower when sirolimus dose was greater than 10 mg, or when concentration was greater than 12 µg/L.
Methodology developed for performing the sirolimus analysis included (1) a method called back analysis (BA), developed to convert non-compartmental variables to compartment model dependent PK parameters for both one- and two-compartment models using Excel®; (2) a prior elicitation method for quantitative data and (3) model discrimination tools for Bayesian population PK modelling.
Modelling was successfully performed for each immunosuppressive drug and the studies and techniques developed provide future directions to maximize benefits from tacrolimus and sirolimus in transplant recipients.