Evaluating prediction methods for estimating tacrolimus, mycophenolate and prednisolone exposure following adult kidney transplantation

Abstract

BackgroundTacrolimus, mycophenolic acid (MPA) and prednisolone are cornerstone immunosuppressant agents prescribed after kidney transplant, however their use is complicated by significant intra- and inter-individual pharmacokinetic variability and narrow therapeutic windows. Therapeutic drug monitoring (TDM) and individualising immunosuppression improves drug efficacy and safety, and transplant outcomes. The most accurate measurement of drug exposure is the area under the concentration-time curve (AUC), but this is not practical for routine use. Currently, single time measurements (commonly pre-dose or trough (C0)) are used routinely for dose adjustment of some immunosuppressant drugs, however these poorly reflect total drug exposure. Limited sampling strategies (LSSs), including multiple linear regression and Bayesian forecasting techniques can estimate total drug exposure from a limited number of drug-concentration measurements. TDM using blood or plasma concentrations is invasive and cumbersome. Saliva may offer a more practical alternative, and preliminary studies have shown a correlation between saliva and plasma concentrations for free MPA and prednisolone. Further investigations are required to validate LSSs and to determine how well these predict future immunosuppressant exposure and to clarify the relationship between saliva and plasma MPA and prednisolone concentrations.AimsThis thesis aims to review Bayesian forecasting of tacrolimus following whole organ transplantation, evaluate existing LSSs for estimation of MPA AUC from 0 to 12 hours post-dose (AUC0-12) following administration of enteric-coated mycophenolate (EC-MS), evaluate the performance of Bayesian forecasting programs/services for tacrolimus, and investigate any correlation of free MPA and prednisolone concentrations between saliva and plasma.MethodsA literature search was performed to identify published manuscripts and data on TDM, LSSs and saliva measurements of MPA, tacrolimus and prednisolone in solid organ transplantation (Chapter 2 and 3). A prospective, clinical study was conducted involving twenty adult kidney transplant recipients, approximately four to eight weeks post-transplant. All patients were receiving oral EC-MS (Myfortic™, twice daily), tacrolimus (Prograf™, twice daily) and prednisolone (Panafcortelone™, daily), and had blood samples collected on two occasions, approximately one week apart, for measurement of tacrolimus, total and free MPA and total and free prednisolone concentrations.  Twelve blood samples were collected at specified times between pre-dose and 8 hours post-dose on the initial occasion, and thirteen blood samples were collected at specified times between pre-dose and 12 hours post-dose on the second occasion, from each participant. Saliva samples were collected simultaneously with blood samples, using a synthetic saliva collection device (Salivette™ with nylon wad).  Predicted MPA and tacrolimus AUC0-12 were estimated on the first and second occasions and compared to an estimated AUC0-12 calculated using the trapezoidal rule (Chapter 5 and 6, respectively). Plasma and salivary MPA and prednisolone concentrations and AUC were assessed for any association (Chapter 7 and 8, respectively).ResultsIn Chapter 3, studies that assessed the population pharmacokinetics and/or the predictive performance of MAP Bayesian estimation of tacrolimus AUC from 0 until the end of the dosing interval were identified. Evaluation using data from the clinical study showed bias and imprecision associated with prediction of tacrolimus AUC ranged from -15 to 9.95% and from 0.81 to 40%, respectively. r2 values ranged from 0.27 to 0.99%. The combination of sampling times at 0, 1 and 3 hours post-dose consistently showed bias and imprecision of <15%.In Chapter 4, LSSs for estimation of MPA AUC0-12 following EC-MS were evaluated. Bias and imprecision for prediction of full MPA AUC0-12 were <15% for four LSSs, using the data from the same (second) occasion. One equation achieved bias and imprecision <15% for prediction of a future MPA AUC0-12, where the AUC0-12 predicted from the first occasion was compared to the full AUC0-12 estimated from the second. All LSSs with acceptable predictive performance included concentrations taken at least 8 hours after the dose.In Chapter 5, three Bayesian forecasting programs/services for tacrolimus were evaluated using different LSS (C0; C0, C1, C3; C0, C1, C2, C4; and all available concentrations). With the exclusion of LSS using only C0, bias and imprecision for the prediction of tacrolimus full estimated AUC0-12 were <15% for all Bayesian forecasting programs/services, using data from the same (second) occasion. Two Bayesian forecasting programs/services, as well as the third when patients with tacrolimus dose changes between sampling occasions were excluded, showed bias and imprecision <15% for the prediction of future tacrolimus AUC0-12, where the occasion one predicted AUC using C0, C1 and C3 was compared with occasion two estimated AUC.In Chapter 6, the relationship between plasma and saliva MPA was investigated. Correlation between total plasma and saliva MPA and free plasma and saliva concentrations was r2 = 0.51 and r2 = 0.41, respectively. The correlation between total plasma MPA AUC or free plasma AUC and saliva MPA AUC was r2 = 0.25 and r2 = 0.13, respectively.In Chapter 7, the relationship between plasma and saliva prednisolone concentrations was evaluated. Correlation was poor between saliva and free (r2 = 0.003) and total (r2 = 0.01) prednisolone plasma concentrations but improved when concentrations prior to the maximum free prednisolone plasma value were excluded (free plasma versus saliva r2 = 0.57). There was a poor correlation between free and total plasma prednisolone AUC0-12 and saliva AUC0-12 (r2 = 0.07; r2 = 0.17). ConclusionsOnly one MLR LSS had acceptable bias and precision for future estimation of MPA AUC. Accurate dosage prediction using a MLR LSS was not possible without concentrations up to at least 8 hours after the dose. Bayesian forecasting programs/services using tacrolimus concentrations at C0, C1 and C3 could be used for accurate prediction of future tacrolimus exposure in adult kidney transplant recipients. Measurement of MPA and prednisolone concentrations in saliva cannot currently replace plasma measurements for therapeutic drug monitoring of MPA following EC-MS administration, or prednisolone. Findings from this thesis provide useful insights and advancements into our understanding of measuring immunosuppressant exposure in adult kidney transplant recipients.