An adaptive iterative decision multi-feedback detection algorithm with constellation constraints is proposed for multiuser multi-antenna systems. An enhanced detection and interference cancellation is performed by introducing multiple constellation points as decision candidates. A complexity reduction strategy is developed to avoid redundant processing with reliable decisions along with an adaptive recursive least squares algorithm for time-varying channels. An iterative detection and decoding scheme is also considered with the proposed detection algorithm. Simulations show that the proposed technique has a complexity as low as the conventional decision feedback detector while it obtains a performance close to the maximum likelihood detector.