This paper evaluates the performance of an analog least mean square (ALMS) loop employed to cancel self-interference in in-band full-duplex (IBFD) orthogonal frequency division multiplexing (OFDM) systems. Cyclostationary analysis is applied to investigate the behavior of the ALMS filter. It is revealed that the performance of the ALMS filter for OFDM systems primarily depends on windowing function rather than pulse shaping as in single carrier systems. It is also noticed that the ALMS loop in OFDM systems provides a much higher level of sel-interference (SI) suppression because OFDM signals lead to reduced the error of the interference channel modelling with the adaptive filter. Simulations are then conducted to verify the theoretical findings.