In order for the large-scale realization of vehicular networks to be feasible, the problem of congestion control must be addressed to ensure the reliability of safety applications. The latter rely on single-hop broadcasts of safety packets in the control channel to acquire up-to-date knowledge of the local neighbourhood. However, high transmission ranges of onboard radios and the highly dynamic mobility of vehicles may result in fast-forming pockets of high node density in the network. Subsequently, the excessive load caused by safety packets broadcasts may degrade the network performance and subsequently reduce the level safety provided by applications. Existing congestion control schemes in the literature aim to reach a fair rationing of available channel resources throughout the network. However, a particular vehicle, depending on its distance and relative velocity with respect to its neighbours may require less or more network resources than another vehicle to achieve the same level of safety benefit. We examine the problem of adapting the probability of transmission of each node under a slotted p-persistent vehicular broadcast medium access control (MAC) scheme. A network utility maximization (NUM) problem is formulated, in which utility incorporates both the expected delay and a notion of safety benefit. A distributed algorithm is proposed to solve this problem in a decentralized manner and its performance is studied through simulations.