Population Pharmacokinetics of Tigecycline in Phase 1 Subjects

Authors: Van Wart S, Cirincione B, Hirankarn S, Phillips L, Meagher AK, Troy S, Owen JS
Conference: ICAAC
Keywords: anti-infectives, antibiotic, full-profile, infectious disease, intravenous, IV, NONMEM, PK
Division: Clinical Pharmacology and Pharmacometrics (CPP)

Abstract

Background: TGC is the first glycylcycline antimicrobial agent in development and has potent in vitro activity against many multi-drug resistant organisms. Given observed differences in the PK profiles after single doses (SD) & multiple doses (MD) of TGC, separate PPK models were developed for SD & MD data pooled from 5 Phase 1 studies.

Methods: TGC (12.5 to 300 mg) was infused over 1 hour twice daily for 9 to 10 days. Serial blood samples were collected after a SD (2030 samples, 174 subjects) and on Day 9 or 10 (203 samples, 13 subjects). Both 2- and 3-compartment (CMT) models were fit to the serum
TGC data using NONMEM®. The models that best described the full-profile SD & MD data were evaluated on a PH 1 dataset reduced to the Phase 2/3 sparse sampling scheme and dose range (25 to 100 mg)

Results: 3-CMT models with 1st-order elimination best described the SD & MD data. Intersubject variability (IIV) of CL, distribution CL (Q1 & Q2), and volume (Vp1 & Vp2) for each peripheral CMT were described using an exponential error model. However, the IIV of Q2 & Vp2 could not be estimated for MD data and were removed from the MD model. A log error model best described residual variability (RV) for both the SD & MD models. The elimination half-life was longer following MD of TGC (115 hr) than a SD (51 hr). The MD Bayesian PK parameters were also used to predict SD data (reverse superpositioning), revealing that the SD AUC0-12 was underpredicted for most subjects. Thus, the full-profile SD & MD data were not pooled and fit with a single model. The reduced Phase 1 data collected out 12 hr for both SD & MD of TGC were adequately described using a 2-CMT model. The predicted SD & MD AUC0-12 values were unbiased relative to observed values; median prediction error (PE) and absolute PE were similar for both models and were ±1% and 3%, respectively.

Conclusions: A 3-CMT model best described the serial TGC data following a SD or MD, however, an empiric 2-CMT model provides unbiased estimates of AUC0-12 using the PK sampling strategy implemented in Phase 2/3.

Interscience Conference on Antimicrobial Agents and Chemotherapy (ICAAC), Washington, DC, October 2004

By Van Wart, S., Cirincione, B., Hirankarn, S., Luann Phillips, Meagher, A., Troy, S., Joel S. Owen