In cognitive radio (CR), a cognitive user must cease transmission or reduce the transmit power at the band where a primary system (PS) is detected. To this end, a popular method for orthogonal frequency division multiplexing (OFDM) based CR is the active interference cancellation (AIC) which tunes the modulation of some subcarriers to minimize the interference to PS. This paper focuses on complexity reduction and performance enhancement of AIC. The former relies on the findings that (i) many matrices in AIC depend only on the PS bandwidth and are irrelevant to where PS resides, and (ii) statistical relation among the data carried by subcarriers for interference suppression is available. For the latter, performance metric of AIC that minimizes the total residual power at the PS band is adjusted to minimizing the largest residual power, avoiding the undesirable situation that, although the total power is the least, part of the band is severely interfered. Spectrum overshoot problem occurred in an ultra-wideband (UWB)-CR system is also addressed. Simulation results demonstrate that complexity reduction of AIC induces negligible performance loss, and the new performance metric significantly reduces the number of subcarriers required to suppress the interference below a certain level.