A lightweight distributed MAC protocol is proposed in this paper to regulate the coexistence of high-priority (primary) and low-priority (secondary) wireless devices in cognitive wireless sensor networks. The protocol leverages the available spectrum resources while guaranteeing stringent quality of service requirements. By sensing the congestion level of the channel with local measurements and without any message exchange, a novel adaptive congestion control protocol is developed by which every device independently decides whether it should continue operating on a channel, or vacate it in case of saturation. The proposed protocol dynamically changes the congestion level based on variations of the non-stationary network. The protocol also determines the optimal number of active secondary devices needed to maximize the channel utilization without sacrificing latency requirements of the primary devices. This protocol has almost no signaling and computational overheads and can be directly implemented on top of existing wireless protocols without any hardware/equipment modification. Experimental results showsubstantial performance enhancement compared to the existing protocols and provide useful insights on low-complexity distributed adaptive MAC mechanism in cognitive wireless sensor networks.