A model-driven design and validation of closed-loop medical device systems is presented. Currently, few if any medical systems on the market support closed-loop control of interconnected medical devices, and mechanisms for regulatory approval of such systems are lacking. We present a system implementing a clinical scenario where closed-loop control may reduce the possibility of human error and improve safety of the patient. The safety of the system is studied with a simple controller proposed in the literature. We demonstrate that, under certain failure conditions, safety of the patient is not guaranteed. Finally, a more complex controller is described and ensures safety even when failures are possible. This investigation is an early attempt to introduce automatic control in clinical scenarios and to delineate a methodology to validate such patient-in-the-loop systems for safe and correct operation.