In-Sun Nam, Leon Aarons, Feiko ter Kuile, Nick White
School of Pharmacy, University of Manchester, U.K. Faculty of Tropical Medicine, Mahidol University, Bangkok, Thailand
The purpose of the current study was to investigate a Bayesian approach to propagate pharmacokinetic information from a normal volunteer study to a patient study. The specific application was the analysis of halofantrine in the treatment of falciparum malaria. Normal volunteer data was cumulated from 6 healthy male subjects aged between 21 and 34, weighing 50 – 85 kg who received 250, 500 and 1000 mg of halofantrine in a cross-over design. Patient data came from two studies conducted in displaced persons of the Karen ethnic minority in camps situated along the Thai-Burmese border [1]. In one study (HF24M), patients received halofantrine 24 mg/kg as 8 mg/kg every six hours, and in the other (HF72M) they received 72 mg/kg as 8 mg/kg every 6 to 10 hours. For HF24M, weights varied from 8 to 58 kg and ages from 1 to 54, while in HF72M weight ranged from 8 to 62 kg and age from 1 to 58 year. A two compartment model was used to analyse the normal volunteer data with the NLME function in SPLUS[2]. Estimates of clearance (CL), volume of compartment one (V1), distributional clearance (CLD) and volume of compartment two (V2) with their inter-individual variability (cv) were 429.28 L/day (30%), 728.59 L (20%), 177.8 L/day and 1350.9 L. PKBUGS[3] was then utilised to analyse the patient data using the results of the normal data analysis as an informative prior. The Bayesian approach requires the specification of prior probability distributions for the population parameters as well as the corresponding variance-covariance matrix. The prior distribution of parameter estimates was a multinormal distribution and its inverse variance-covariance matrix was Wishart and the corresponding values were obtained from the normal volunteer data analysis. Estimates of CL, V1, CLD and V2 were 354.3 L/day, 727.8 L, 162.4 L/day and 1939.1 L while those of the most equivalent analysis using NLME were; 421.3 L/day, 733.1 L, 180.5 L/day, 1902.3L. Moreover, while only the random effects for CL and V1 could be estimated with NLME, all the parameters were allowed to be random in the Bayesian analysis. In conclusion, both fits were similar and the Bayesian approach showed more stable performance in covariate selection.
References
[1] ter Kuile, F. O., Dolan, G., Nosten, F., Edstein, M. D., Luxemburger, C., Phaipun, L., Chongsuphajaisiddhi, T., Webster, H. K. and White, N. J. Halofantrine versus mefloquine in treatment of multidrug-resistant falciparum malaria, Lancet, 341:1044-1049, 1993.
[2] Lindstrom, M. J., and Bates, D. M. Nonlinear mixed-effects models for repeated measures data, Biometrics, 46:673-687, 1990.
[3] Lunn, D. J., Wakefield, J., Thomas, A., Best, N. and Spiegelhalter, D. Pkbugs User Guide, Epidemiology & Public health at Imperial College School of Medicine, London, 1999
Reference: PAGE 10 (2001) Abstr 174 [www.page-meeting.org/?abstract=174]
Poster: poster