Based on density matrix equations, we studied the interaction process among phase-modula-ted beams, unmodulated beam and a cascade three-level system. Calculations of the line shape of optical heterodyne modulation transfer spectroscopy via nearly-degenerate four-wave mixing were made. The results show that the Doppler background due to the near resonance of the in-termediate level is eliminated and the signal of two-photon transition can be distinguished from saturation absorption.