Population PK/PD modelling of AZD9773 in patients with severe sepsis.
James W.T Yates1, Shampa Das1, Guy Mainwaring1 and John Kemp1
1AstraZeneca, Alderley Edge, SK10 4TG
Objectives: AZD9773 is an intravenously infused ovine-derived polyclonal anti-TNF-α Fab fragment currently being developed as a treatment for severe sepsis. The aim of this analysis was to develop a population PK/PD model in patients with severe sepsis. This model was then used to simulate proposed dosing regimens for a phase IIb trial (NCT01145560).
Methods: A phase IIa trial in patients with severe sepsis was conducted (NCT00615017). A total of 71 patients were recruited, 47 of whom received study treatment. Serum AZD9773 and TNF-α levels were measured.
A population pharmacokinetic analysis was performed in NONMEM and serum pharmacokinetics of total AZD9773 were described by a two-compartment model. The variation in TNF-α levels between individuals over time and the drug effect were captured using an indirect response model with a variable production rate of TNF-α. Two routes of elimination of TNF-α were accounted for in the model: natural removal of TNF-α, TNFout, and a drug dependent process, kbind.
The apparent appearance of TNF-α in the serum was described by a quadratic function parameterised by: slope describing the initial slope; peak describing the peak (or trough) rate of appearance; ttp (time to peak) describing the time at which this peak (or trough) was achieved.
Results: The population PK model had one covariate: creatinine clearance on drug clearance.
A mixed effects model with random effects on peak, slope, ttp and TNFout was used to describe the TNF-α data. The parameter kbind was kept as a fixed effect only because a random effect on this parameter might have been confounded by TNFout inter-individual variability. Parameter estimation in NONMEM terminated normally.
A number of different regimens were simulated for 100 patients in each arm. 100 trials were simulated and the effects on TNF-α levels were summarised across these.
Conclusions: TNF-α serum levels and the pharmacodynamic effect of AZD9773 on TNF- α were well characterised by the model across the cohorts in the Phase IIa study. The model was used to simulate different dosing options to support design of a now ongoing phase IIb trial.