Operational problems arising in the planning of integrated supply chains have been increasingly studied in the past decade. Among these, the production routing problem (PRP) is a difficult problem which aims to jointly optimize production, inventory, distribution and routing decisions in order to satisfy the dynamic demand of customers and minimize the overall system cost. This paper introduces an optimization-based adaptive large neighborhood search heuristic for the PRP. In this heuristic, binary variables representing setup and routing decisions are handled by upper-level search operators, while continuous variables associated with production, inventory and shipment quantities are set by solving a network flow subproblem. Extensive computational experiments have been performed on benchmark instances from the literature. The results show that our algorithm generally outperforms existing heuristics for the PRP and can produce high quality solutions in short computing times.