The Effect of Study Design on Pharmacokinetics in Patients with Impaired Renal Function
HK. Kim(1), SB Duffull, (1,2), B. Green(1)
(1) School of Pharmacy, University of Queensland, Brisbane, Queensland, Australia (2) School of Pharmacy, University of Otago, Dunedin, New Zealand
Objectives: To ensure the effect of renal function on drug exposure is precisely quantified, FDA guidance recommends that studies recruit approximately equal subject numbers with normal renal function and mild, moderate and severe renal impairment. However, in population PK analyses it is common to pool data from various studies, resulting in an over-representation of subjects with normal renal function. The purpose of this simulation based experiment was to explore how varying the design, with respect to subject numbers in the different renal function groups, impacts upon the precision of PK parameter estimation.
Methods: A 1-compartment, first order input, first order elimination model was used to simulate concentrations following a single 1mg/kg dose of a drug. Clearance (CL) was defined as a composite of renal CL (0.7 L/hr) and non-renal CL (0.3 L/hr). During the simulation, subjects were stratified into the 4 renal function groups where normal renal function, mild, moderate and severe renal impairment were classified as creatinine clearances of ≥80, 50-79.9, 30-49.9 and <30 mL/min respectively. 1000 trials with 100 subjects per trial were simulated under five different designs. Design A included 25 subjects in each of the 4 renal function groups. Designs B, C, D and E included 33-33-33-0, 50-50-0-0, 100-0-0-0 and 76-8-8-8 subjects in each renal function group respectively. Both intensive and sparse samplings were evaluated for each design, as well as differing magnitudes of random effects. PK parameters were re-estimated and compared across the designs. The percentage of times that the PK estimate was within 10% of the true value was computed for each design.
Results: For dense sampling the median (SE) estimates of renal CL in designs A, B, C, D and E were 0.68(0.06), 0.67(0.08), 0.62(0.12), 0.53(0.17) and 0.68(0.06) respectively. For design A, the estimate of renal CL was within 10% of the true value on 71.6% of occasions. This percentage decreased to 57.8, 37.9 and 19.7% under designs B, C and D respectively. The percentage was 71.8% in design E.
Conclusions: Design, A which had equal numbers of subjects in the different renal function groups resulted in the lowest standard errors together with the design that pooled a renal impairment study with subjects that had predominantly normal renal function (Design E). Excluding subjects with severe renal impairment in pooled population PK analysis does not give a reliable estimate of the renal effect on CL.