Impact of model misspecification at design (and/or) estimation step in population pharmacokinetic studies

Simulating clinical trials for model-informed precision dosing: using warfarin treatment as a use case

Treatment response variability across patients is a common phenomenon in clinical practice. For many drugs this inter-individual variability does not require much (if any) individualisation of dosing strategies. However, for some drugs, including chemotherapies and some monoclonal antibody treatments, individualisation of dosages are needed to avoid harmful adverse events. Model-informed precision dosing (MIPD) is an emerging approach to guide the individualisation of dosing regimens of otherwise difficult-to-administer drugs. Several MIPD approaches have been suggested to predict dosing strategies, including regression, reinforcement learning (RL) and pharmacokinetic and pharmacodynamic (PKPD) modelling. A unified framework to study the strengths and limitations of these approaches is missing. We develop a framework to simulate clinical MIPD trials, providing a cost and time efficient way to test different MIPD approaches. Central for our framework is a clinical trial model that emulates the complexities in clinical practice that challenge successful treatment individualisation. We demonstrate this framework using warfarin treatment as a use case and investigate three popular MIPD methods: 1. Neural network regression; 2. Deep RL; and 3. PKPD modelling. We find that the PKPD model individualises warfarin dosing regimens with the highest success rate and the highest efficiency: 75.1% of the individuals display INRs inside the therapeutic range at the end of the simulated trial; and the median time in the therapeutic range (TTR) is 74%. In comparison, the regression model and the deep RL model have success rates of 47.0% and 65.8%, and median TTRs of 45% and 68%. We also find that the MIPD models can attain different degrees of individualisation: the Regression model individualises dosing regimens up to variability explained by covariates; the Deep RL model and the PKPD model individualise dosing regimens accounting also for additional variation using monitoring data. However, the Deep RL model focusses on control of the treatment response, while the PKPD model uses the data also to further the individualisation of predictions.

External evaluation of published population pharmacokinetic models of tacrolimus in adult renal transplant recipients

AIM

Several tacrolimus population pharmacokinetic models in adult renal transplant recipients have been established to facilitate dose individualization. However, their applicability when extrapolated to other clinical centres is not clear. This study aimed to (1) evaluate model external predictability and (2) analyze potential influencing factors.

METHODS

Published models were screened from the literature and were evaluated using an external dataset with 52 patients (609 trough samples) collected by postoperative day 90 via methods that included (1) prediction-based prediction error (PE%), (2) simulation-based prediction- and variability-corrected visual predictive check (pvcVPC) and normalized prediction distribution error (NPDE) tests and (3) Bayesian forecasting to assess the influence of prior observations on model predictability. The factors influencing model predictability, particularly the impact of structural models, were evaluated.

RESULTS

Sixteen published models were evaluated. In prediction-based diagnostics, the PE% within ±30% was less than 50% in all models, indicating unsatisfactory predictability. In simulation-based diagnostics, both the pvcVPC and the NPDE indicated model misspecification. Bayesian forecasting improved model predictability significantly with prior 2-3 observations. The various factors influencing model extrapolation included bioassays, the covariates involved (CYP3A5*3 polymorphism, postoperative time and haematocrit) and whether non-linear kinetics were used.

CONCLUSIONS

The published models were unsatisfactory in prediction- and simulation-based diagnostics, thus inappropriate for direct extrapolation correspondingly. However Bayesian forecasting could improve the predictability considerably with priors. The incorporation of non-linear pharmacokinetics in modelling might be a promising approach to improving model predictability.