Background: Infliximab is an anti-tumour necrosis factor alpha monoclonal antibody used to treat inflammatory bowel diseases. Unfortunately, not all patients demonstrate an initial response during induction, whilst others respond initially but relapse within one year of treatment. Adapting doses based on clinical outcomes and trough concentrations can improve response and reduce the proportion that develop antibodies to infliximab, but interpretation in the presence of time-varying patient factors is complicated.
Methods: Several adaptive dosing strategies (label recommendations versus therapeutic drug monitoring with an established stepwise algorithm or proportional dose adjustments or population pharmacokinetic model-based dosing through Bayesian estimation) were simulated on a virtual patient population (constructed with time-varying covariates and random effects on individual pharmacokinetic parameters) using R to assess their relative performance. Strategies were evaluated on their ability to maintain trough infliximab concentrations above an established target, 3 mg/L, during the maintenance phase.
Results: Model-based dosing was superior in maintaining target trough concentrations, showing individuals achieving concentrations above the target faster in the maintenance phase and a lower proportion of individuals who developed antibodies to infliximab. Model-based dosing results were consistent across a range of baseline covariate groups.
Conclusion: This in silico assessment of adaptive dosing strategies demonstrated that, when challenged with dynamic covariate and random effect changes occurring in individual pharmacokinetic parameters, model-based approaches were superior compared to other strategies. The potential clinical benefits of model-based dosing strategies for infliximab is worthy of investigation, and the in silico methods described here could be used to optimise the design of these strategies.