Efficient management of radio resources is crucial in maintaining peak performance of cellular wireless networks under resources constraints. As wireless networks evolve toward 3G/4G and beyond with broadband multimedia services, managing radio resources to satisfy diverse Quality-of-Service requirements is becoming even more critical. Battery energy conservation for portable terminals on the move is another important aspect in wireless networks. Due to the limit on battery life-time, each unit of battery energy saved directly translates to an increase in the value of communication for a subscriber. In this work, we investigate a uplink power control problem for packet-switched data services, with a focus on energy efficiency for mobile terminals. Packet-switched data differ fundamentally from circuit-switched data in the burstiness of traffic and the connectionless nature of communications. Based on a utility-maximization approach from microeconomics, we define a probabilistic utility model as a performance metric for a wireless data user, which takes into account both the traffic burstiness and average packet delay requirement. Game-theoretic approach is then utilized to study a distributed power control strategy to simultaneously maximize the utility for each individual user in the system. In general, the problem is mathematically intractable. Using several approximation methods, the problem is reduced into tractable format and is studied both analytically and by simulations. Results show that the proposed power control scheme converges to a unique Nash equilibrium which depends on mobile's location, average packet delay requirement, traffic burstiness, and the mean and variance of the interference and background noise at mobile's base station receiver. The scheme can be easily extended to multi-class user traffic environments. For performance evaluation, we establish two idealized slot-by-slot based power control strategies as performance benchmarks. It's shown that the performance achieved by the proposed scheme is close to those by benchmark schemes. Generally, this work provides a new approach based on approximation techniques for the investigation of uplink power control problem in packet-switched systems. The results and insights generated by this study provide guidance on the efficient management of transmit powers for energy-efficient packet-switched data services in current and future generations of wireless networks.