Population pharmacokinetic modelling of roflumilast and roflumilast N-oxide by total phosphodiesterase-4 inhibitory activity and development of a population pharmacodynamic-adverse event model.

Background: Roflumilast is an oral, selective phosphodiesterase (PDE)-4 inhibitor in development for the treatment of chronic obstructive pulmonary disease (COPD). Both roflumilast and its metabolite roflumilast N-oxide have anti-inflammatory properties that contribute to overall pharmacological activity.

Objectives: To model the pharmacokinetics of roflumilast and roflumilast N-oxide, evaluate the influence of potential covariates, use the total PDE4 inhibitory activity (tPDE4i) concept to estimate the combined inhibition of PDE4 by roflumilast and roflumilast N-oxide, and use individual estimates of tPDE4i to predict the occurrence of adverse events (AEs) in patients with moderate-to-severe COPD.

Methods: We modelled exposure to roflumilast and roflumilast N-oxide (21 studies provided the index dataset and five separate studies provided the validation dataset), extended the models to COPD (using data from two studies) and assessed the robustness of the parameter estimates. A parametric bootstrap estimation was used to quantify tPDE4i in subpopulations. We established logistic regression models for each AE occurring in >2% of patients in a placebo-controlled trial that achieved a p-value of <0.2 in a permutation test. The exposure variables were the area under the plasma concentration-time curve (AUC) of roflumilast, the AUC of roflumilast N-oxide and tPDE4i. Individual AUC values were estimated from population models.

Results: Roflumilast pharmacokinetics were modelled with a two-compartment model with first-order absorption including a lag time. A one-compartment model with zero-order absorption was used for roflumilast N-oxide. The final models displayed good descriptive and predictive performance with no appreciable systematic trends versus time, dose or study. Posterior predictive checks and robustness analysis showed that the models adequately described the pharmacokinetic parameters and the covariate effects on disposition. For roflumilast, the covariates of sex, smoking and race influenced clearance; and food influenced the absorption rate constant and lag time. For roflumilast N-oxide, age, sex and smoking influenced clearance; age, sex and race influenced the fraction metabolized; bodyweight influenced the apparent volume of distribution; and food influenced the apparent duration of formation. The COPD covariate increased the central volume of distribution of roflumilast by 184% and reduced its clearance by 39%; it also reduced the estimated volume of distribution of roflumilast N-oxide by 21% and reduced its clearance by 7.9%. Compared with the reference population (male, non-smoking, White, healthy, 40-year-old subjects), the relative geometric mean [95% CI] tPDE4i was higher in patients with COPD (12.6% [-6.6, 35.6]), women (19.3% [8.2, 31.6]), Black subjects (42.1% [16.4, 73.4]), Hispanic subjects (28.2% [4.1, 57.9]) and older subjects (e.g. 8.3% [-11.2, 32.2] in 60-year-olds), and was lower in smokers (-19.1% [-34.0, -0.7]). Among all possible subgroups in this analysis, the subgroup with maximal tPDE4i comprised elderly, Black, female, non-smoking, COPD patients (tPDE4i 217% [95% CI 107, 437] compared with the value in the reference population). The probability of a patient with tPDE4i at the population geometric mean [95% CI] was 13.0% [7.5, 18.5] for developing diarrhoea, 6.0% [2.6, 9.4] for nausea and 5.1% [1.9, 8.6] for headache.

Conclusions: Covariate effects have a limited impact on tPDE4i. There was a general association between tPDE4i and the occurrence of common AEs in patients with COPD.