Based on the theory of stochastic resonance, the signal to noise ratio (SNR) of HPLC/UV chromatographic signal of roxithromycin is enhanced by cooperation of signal, noise and nonlinear system. A simple new method for the determination of low concentration of roxithromycin in beagle dog plasma is presented. Using signal enhancement by stochastic resonance, this method extends the limit of quantitation from the reported 0.5 to 0.1 microg/ml. During validation of the new method, HPLC/MS was used as a comparison technique. The results indicate that the recovery and low concentrations of roxithromycin in beagle dog plasma were equivalent between the two methods (P>0.05). Stochastic resonance may be a promising tool for improving detection limits in trace analysis.