Body area networks (BAN) offer amazing perspectives to instrument and support humans in many aspects of their life. Among all possible applications, this paper focuses on body monitoring applications having a body equipped with a set of sensors transmitting in real time their measures to a common sink. In this context, at the application level, the network fits with a star topology, which is quite usual in the broad scope of wireless networks. Unfortunately, the structure of the network at the physical layer is totally different. Indeed, due to the specificity of BAN radio channel features, all radio links are not stationary and all sensors suffer from link losses during independent time frames. In wireless networks, link losses are often coped with multi-hop transmission schemes to ensure a good connectivity. However, since the radio links are not stationary, the multi-hop routes should adapt quickly to BAN changes. We instead propose in this work a different approach based on opportunistic relaying. The concept relies on electing some sensors to support the transmission of other ones having a worst connection. Instead of changing the relay time to time, we rather select a relay node from a statistical perspective. We evaluate this approach from a theoretical point of view and on realistic simulations using the packet error rate outage probability as a performance criterion.