This paper focuses on solving a variant of the vehicle routing problem (VRP) in which a time window is associated with each customer service and some services require simultaneous visits from different vehicles to be accomplished. The problem is therefore called the VRP with time windows and synchronization constraints (VRPTWSyn). We present a simulated annealing algorithm (SA) that incorporates several local search techniques to deal with this problem. Experiments on the instances from the literature show that our SA is fast and outperforms the existing approaches. To the best of our knowledge, this is the first time that dedicated local search methods have been proposed and evaluated on this variant of VRP.