Analysis of digoxin data using Non-Parametric Maximum Likelihood (NPML)
Comets, Emmanuelle(1) and Céline Verstuyft (2)
(1) Inserm U738, Paris 7 University, Paris, F-75018 France; (2) Centre d'investigation Biologique (CIB), Pharmacology Department, Saint Antoine University ofedicine, University Pierre et Marie Curie, Paris Sud, France.
Objectives: Genetic factors contribute to the processes of absorption, transport and metabolism of drugs. Parametric methods for parameter estimation usually assume the normality of pharmacokinetic parameters, but the influence of genetic factors could introduce multimodality in the distribution. Non- parametric methods can be used as a tool to detect multimodality. The objectives of this work were (i) to reprogram the non-parametric method NPML [1] in the C language, and (ii) to apply this non-parametric method to digoxin data, a well-known probe for the activity of P-glycoprotein (PgP).
Methods: We pooled the data from three drug interaction studies in 32 healthy volunteers with extensive pharmacokinetic sampling. The data has been previously analysed using non-compartmental approaches [2]. All patients were genotyped for the two main mutations in the MDR-1 gene which controls the expression of PgP (C3435T in exon 26 and G2677T/A in exon21).
We have programmed the Non-Parametric Maximum Likelihood method developed by Alain Mallet [1] in C language. The program was developed under GNU C on a personal computer running the Linux operating system (RedHat and Ubuntu distributions). The programming makes extensive use of the papers and the original Fortran code from Alain Mallet.Results: In a previous work, we have shown an increase in bioavailability in homozygote carriers of the T allele for the exon 26 polymorphism in MDR-1. Here, we used NPML to study the same dataset and we compare with the results of the previous analysis.
In the current version of the program, the user must write his or her own models and provide the routines to read the data. The program can accomodate models written as differential equations using the Gnu Scientific Library (GSL).Conclusion: The NPML algorithm has been reprogrammed in C language and we have applied it to digoxin pharmacokinetic data. We intend to package the C version of the software as a library for the statistical software R to improve its user-friendliness.
References:
[1] A Mallet. A makimum likelihood estimation method for random coefficient regression models. Biometrika, 73:64556, 1986
[2] C Verstuyft, M Schwab, E Schaeffeler, R Kerb, U Brinkmann, P Jaillon, C Funck-Brentano, and L Becquemont. Digoxin pharmacokinetics and MDR1 genetic polymorphisms. Eur J Clin Pharmacol, 58(12):80912, 2003.