Vehicular safety applications require vehicles to maintain a high awareness level of the local neighborhood through broadcasting safety beacons on the control channel. However, the existing 10-MHz control channel in the IEEE 802.11p based Dedicated Short Range Communication (DSRC) standard can be easily congested by frequent beaconing in a dense environment which therefore degrades the performance of network and safety level of vehicles. Existing congestion mitigation approaches aim to fairly distribute the channel resources based on channel load measurements, but fail to incorporate the road safety requirements of vehicles. In this paper, we model the congestion control problem of Vehicular Ad-hoc Networks (VANETs) as a utility maximization problem leveraging i) the contribution of every vehicle to channel load with respect to its location, and ii) a car-following risk factor which is defined as the rear-end crash risk perceived by each vehicle. A distributed game-theoretic rate adaptation mechanism is then proposed to address the problem. Numerical results demonstrate that the proposed scheme dominates IEEE 802.11p CSMA/CA based beaconing mechanism in terms of packet loss, packet delivery rate and aggregate throughput.