Clinical pharmacokinetic/pharmacodynamic modelling of 5a-reductase inhibitors for the treatment of benign prostatic hyperplasia

Sammanfattning: Dihydrotestosterone (DHT), a potent androgen necessary for the development of benign prostatic hyperplasia (BPH) is formed from testosterone by the enzyme 5α-reductase, of which there are two types. The irreversible 5α-reductase inhibitors dutasteride, which inhibits both isozymes, and finasteride, which inhibits only one, are intended for the treatment of BPH. Models, based on clinical data, were developed for dutasteride pharmacokinetics and for the effects of 5α-reductase inhibitors on enzyme activity and thereby on circulating DHT. Further was a relationship characterised between DHT suppression and changes in prostate specific antigen (PSA), a potential marker for disease progression. Model development progressed along three axes, from healthy volunteers to patients, from single dose to long-term treatment and from pharmacokinetics via biomarker to possible surrogate endpoint. Model predictiveness was validated and methods for propagation of information between studies were examined. Dutasteride pharmacokinetics was best described by a two-compartment model with parallel linear and nonlinear elimination, with a long terminal half-life (5 weeks) at clinical doses. A physiologically based model of DHT and 5α-reductase turnover and of irreversible 5α-reductase inhibition described the data well. This model could explain differences in rates of onset and offset of effect and offers a way to determine the potency of these irreversible inhibitors. The DHT model, derived from single dose data in healthy men, was found to be valid and predictive when compared to data from a 28-day repeat dose study in BPH patients. Using data from a six-month study, PSA was related to DHT by a model describing disease progression and an effect of DHT suppression on elimination of PSA producing cells. When propagating information, the simultaneous modelling of data proved a powerful approach to compare datasets, while the use of priors could stabilise the fitting of complex models to sparse data.