Recently, increasing attention is paid to improving the spectrum utilization in cognitive OFDM systems. To enable coexistence between secondary users and primary users, a scheme, called as extended active interference cancellation (EAIC), was proposed for the effective sidelobe suppression of OFDM signals. However, with the EAIC method, the cancellation signals cause self-interferences to OFDM data subcarriers, which seriously degrades the symbol-error-rate performance, especially, when a high order modulation is used. In this paper, we analyze the self-interferences and formulate the minimization of both the total sidelobe power and the self-interferences as a optimization problem, which could be solved by least square method with acceptable complexity for OFDM communication systems nowadays. Simulation results show that the proposed EAIC-IC scheme provides a good sidelobe suppression performance of about 38.0 dB with signal-to-noise ratio loss less than 0.10 dB at symbol-error-rate of 10 − 3 when 64 QAM modulation is employed. Moreover, the proposed EAIC-IC scheme can provide an arbitrary tradeoff between the sidelobe suppression and the symbol-error-rate performance via adjusting the constraint parameter.