Amonafide is extensively metabolized, including conversion by N-acetylation to an active metabolite. Our previous studies have shown that fast acetylators of amonafide have increased toxicity, and we have recommended doses of 250 and 375 mgm-2day-1for 5 days, for fast and slow acetylators, respectively. Despite phenotype-specific dosing, significant variability in leukopenia persisted. The goal of this study was to construct and validate a pharmacodynamic model-based dosing strategy for amonafide, to try to further decrease inter-patient variability in leukopenia. The model was based on a training data set of 41 patients previously treated with amonafide. The first cycle nadir WBC was modelled as a function of dose, acetylator phenotype and baseline patient factors. This model was validated prospectively on patients similar to those in our previous studies. Based on the training data set, the optimal model was defined by three factors: acetylator phenotype, gender, and pretreatment WBC. Using this model and a target WBC nadir of 1700μ 1-1, six dosing strata were prospectively evaluated. A total of 24 fast acetylators received either 238 or 276 mgm-2day-1and 20 slow acetylators received between 345 and 485 mgm-2day-1The mean (±SE) error (deviation from target nadir) was 430 (±240) cells μI-1. Submaximal treatment (yielding grade 0-1 leukopenia) was limited to 20% of patients, while 55% experienced grade 2-3 toxicity. A complex dosing strategy for amonafide is feasible, employing prospective acetylator phenotyping, model-guided dosing, and adaptive control