We describe an optimization technique for multiuser resource allocation assuming adaptive modulation and coding (AMC) in OFDMA radio downlink communications. The resource allocation process is based on the dual solution of a weighted sum-rate maximization problem. In particular, to realistically address IP-based video and data communications, the set of considered constraints include the number of bits actually present in the different transmission buffers. Furthermore, we propose an innovative AMC solution, based on M-QAM modulation and LDPC codes, jointly working with the scheduler and capable to maximize the throughput while guaranteeing a given set of performance requirements. A simple weight adaptation algorithm is introduced to properly tune the trade off between opportunistic transmissions and fairness among users. Several simulation results are presented to validate the proposed approach and a comparison with more traditional schemes is provided. In particular, we show that a good trade off between user fairness and opportunistic exploitation of the radio channel can be reached with the proposed strategies, improving the communication quality for both real-time video and non real-time data applications.