Spectrum leasing has been widely regarded as one of the most effective ways to improve the utilization of limited spectrum resources. In this paper, we propose a novel traffic-adaptive spectrum leasing (TASL) scheme by allowing secondary users (SUs) to lease part of licensed spectrum channels from primary users (PUs) temporarily for transmitting the dynamically generated secondary packets. As the time length of each leasing period is variable according to the dynamic generation of secondary packets, the proposed TASL can effectively satisfy the quality-of-service requirement of SUs and also benefit PUs with the financial payoff provided by SUs. By establishing a three-dimensional continuous Markov chain for the proposed TASL, we formulate the average utilities of PUs and SUs in terms of the expected buffering time of primary and secondary packets, as well as the expected transmission throughput of PUs and SUs. Moreover, to coordinate the interests of PUs and SUs in a noncooperative manner, we also propose a Stackelberg game model for PUs and SUs to negotiate various spectrum leasing parameters and further apply two specific rules to guarantee the existence of a unique equilibrium solution. Numerical simulation shows that, compared with those existing spectrum leasing schemes that preset a fixed-time length for leasing periods, the proposed TASL can effectively improve the utilization of the leased channels, increase the average utilities of both PUs and SUs, and be more suitable for newly emerging applications that are sensitive to packet transmission delay and buffering overhead.