Wireless Body Area Network applications demand reliable, low-latency task processing while operating under stringent energy and deadline constraints. Although mobile devices have achieved significant computational advances, they remain limited by processing capabilities and battery life. Mobile Edge Computing (MEC) presents a viable solution through computational task offloading; however, developing optimal offloading strategies poses considerable challenges due to the dynamic and distributed nature of WBAN environments. This paper introduces a Federated Multi-Agent Deep Deterministic Policy Gradient (FL-MADDPG) framework for intelligent task offloading and resource allocation in multi-WBAN MEC systems. Our approach simultaneously optimizes three critical objectives: minimizing mobile device energy consumption, ensuring task deadline compliance, and maximizing MEC server resource utilization efficiency. To address scalability concerns and maintain data privacy, federated learning is integrated to enable periodic parameter aggregation across distributed learning agents without exposing sensitive user data. Simulation results demonstrate the effectiveness of the proposed model in improving overall system performance.