This paper investigates the applicability of tube-based robust MPC to control the blood glucose concentration in type 1 diabetic subjects through simulation studies. An existing, widely used nonlinear time-invariant metabolic simulation model is adopted and extended with several physiologically motivated uncertainties to account for the underlying intra-patient variability observed in real patients. Results on a virtual subject show that robustness to uncertainties can be improved and the risk of hypoglycemia reduced, without increasing the online computational demands compared to the conventional model predictive control approach. © 2013 IEEE.