The continued growth and popularity of IoTbased wearables coupled with wireless body area sensor network (BASN) communication architectures continues to gain widespread adoption for e-Health applications. Challenges facing the adoption of this vision require e- Health communication architectures for wearable IoT/BASNs to make effective use of bandwidth, energy stores for extended continuous transmission of physiological data, mitigation of computational burden when analyzing the data in the cloud at a global scale and, most importantly, ensuring the features of clinical significance in the biomedical signal are not compromised. In this paper, we present a novel physiologically-aware communication architecture to address these challenges for wearable IoT/BASNs for e-Health applications. The architecture works by extracting patient health state using local in-node pre-diagnosis (or pre-screening) to help guide the operation of the communication architecture in deciding when data should be transmitted, its type and format, and given quality. This latter property on preserving signal 'quality' is central to the architecture, which employs a feature-based diagnostic distortion measure to ensure retention of features of clinical significance during source coding in order to guarantee their reconstruction; a notion which carries greater emphasis for biomedical signals as opposed to ordinary multimedia signals. Simulation work of the architecture is presented in MATLAB using Electrocardiograph (ECG) signals from PhysioNet's ECG database, where it is demonstrated energy savings are realized by a factor of 30 with a reduction in biomedical data volume as high as 86% for tested cases while preserving clinical features.