Four improved finite-difference numerical deconvolution methods and one nonlinear regression numerical deconvolution method are proposed and implemented using IMSL/IDLTM. These five numerical deconvolution methods are evaluated using simulated data generated with and without added noise under six different dosing cases. Comparisons between these methods are made in terms of the superimposability of the calculated cumulative amount of drug released or absorbed-time profiles with the theoretical data. The results indicate that the proposed fixed step number equal step length numerical deconvolution method is simple and accurate and therefore is appropriate for pharmacokinetic and biopharmaceutic studies. When an analytic function is legitimate to represent the drug input rate, the nonlinear regression numerical deconvolution method will yield enhanced numerical accuracy and stability.