In this paper, novel event-triggered strategies for the design of model predictive (MPC) controllers are presented. The MPC framework consists in finding the solution to a constraint optimal-control problem at every time-step. The case of triggering the optimization of the MPC only when is needed, is investigated. The centralized case is treated first and the results are then extended to a decentralized formulation. We consider a system composed by a number of interconnected subsystems, each one of them controlled by a robust MPC algorithm. Using the Input-to-State (ISS) property of the decentralized MPC controller we reach to an event-triggering rule, for each of the subsystems.