Medication adherence framework: A population‐based pharmacokinetic approach and its application in antimalarial treatment assessments

Abstract We reported here on the development of a pharmacometric framework to assess patient adherence, by using two population‐based approaches – the percentile and the Bayesian method. Three different dosing strategies were investigated in patients prescribed a total of three doses; (1) non‐observed therapy, (2) directly observed administration of the first dose, and (3) directly observed administration of the first two doses. The percentile approach used population‐based simulations to derive optimal concentration percentile cutoff values from the distribution of simulated drug concentrations at a specific time. This was done for each adherence scenario and compared to full adherence. The Bayesian approach calculated the posterior probability of each adherence scenario at a given drug concentration. The predictive performance (i.e., Youden index, receiver operating characteristic ROC curve) of both approaches were highly influenced by sample collection time (early was better) and interindividual variability (smaller was better). The complexity of the structural model and the half‐life had a minimal impact on the predictive performance of these methods. The impact of the assay limitation (LOQ) on the predictive performance was relatively small if the fraction of LOQ data was less than 20%. Overall, the percentile method performed similar or better for adherence predictions compared to the Bayesian approach, with the latter showing slightly better results when investigating the adherence to the last dose only. The percentile approach showed acceptable adherence predictions (area under ROC curve > 0.74) when sampling the antimalarial drugs piperaquine at day 7 postdose and lumefantrine at day 3 postdose (i.e., 12 h after the last dose). This could be a highly useful approach when evaluating programmatic implementations of preventive and curative antimalarial treatment programs in endemic areas.