Studies on the influence of single nucleotide polymorphisms (SNPs) on drug pharmacokinetics (PK) have usually been limited to the analysis of observed drug concentration or area under the concentration versus time curve. Nonlinear mixed effects models enable analysis of the entire curve, even for sparse data, but until recently, there has been no systematic method to examine the effects of multiple SNPs on the model parameters. The aim of this study was to assess different penalized regression methods for including SNPs in PK analyses. A total of 200 data sets were simulated under both the null and an alternative hypothesis. In each data set for each of the 300 participants, a PK profile at six sampling times was simulated and 1227 genotypes were generated through haplotypes. After modelling the PK profiles using an expectation maximization algorithm, genetic association with individual parameters was investigated using the following approaches: (i) a classical stepwise approach, (ii) ridge regression modified to include a test, (iii) Lasso and (iv) a generalization of Lasso, the HyperLasso. Penalized regression approaches are often much faster than the stepwise approach. There are significantly fewer true positives for ridge regression than for the stepwise procedure and HyperLasso. The higher number of true positives in the stepwise procedure was accompanied by a higher count of false positives (not significant). We find that all approaches except ridge regression show similar power, but penalized regression can be much less computationally demanding. We conclude that penalized regression should be preferred over stepwise procedures for PK analyses with a large panel of genetic covariates.