The formalism for simulating scanning electrochemical microscopy (SECM) experiments by boundary element methods in three space coordinates has been extended to allow consideration of nonlinear boundary conditions. This is achieved by iteratively refining the boundary conditions that are encoded in a boundary condition matrix. As an example, the simulations are compared to experimental approach curves in the SECM feedback mode toward samples modified with glucose oxidase (GOx). The GOx layer was prepared by the layer-by-layer assembly of polyelectrolytes using glucose oxidase as one of the polyelectrolytes. The comparison of the simulated and experimental curves showed that under a wide range of experimentally accessible conditions approximations of the kinetics at the sample by first order models yield misleading results. The approach curves differ also qualitatively from curves calculated with first order models. As a consequence, this may lead to severe deviations when such curves are fitted to first order kinetic models. The use of linear approximations to describe the enzymatic reaction in SECM feedback experiments is justified only if the ratio of the mediator and Michaelis-Menten constant is equal to or smaller than 0.1 (deviation less than 10%).