This paper presents a novel iterative hybrid algorithm for subcarrier and power allocation in a cognitive orthogonal frequency division multiple access (OFDMA) downlink. In the considered setup a primary base station forwards information to K distant receivers by using a single OFDM waveform, whereas a secondary base station – subject to stringent per-user rate constraints – interferes with the former by sending information from users to the same set of destinations. Power and user allocation at both base stations is jointly performed by the proposed algorithm to maximize the overall throughput of the setup while satisfying, at the same time, the imposed rate constraints. Our proposal, which stems from an hybridization of the harmony search (HS) and differential evolution (DE) algorithms along with a greedy local repair method, is shown – through computer simulations over the extended vehicular A ITU channel model – to be an effective and practical resource allocation procedure for cognitive OFDMA downlinks.