Stabilizing blood glucose level for type 1 diabetic patients is a continuous challenge, the patient tries always to avoid hyperglycemia and hypoglycemia. In fact, the analysis of blood glucose measurements is one of the most important tasks in order to assess the glucose metabolic control. In this paper, we propose to use a numerical method to show that the blood glucose evolution of type 1 diabetes patients is a chaotic phenomenon. Indeed, we used Mutual Information (I) and the False Nearest Neighbors (FNN) to obtain the embedding delay (τ) and the embedding dimension (m) of our system and then to reconstruct the attractor. From the long time series of the blood glucose evolution of a type 1 diabetes patient including 17000 data, we have computed the Largest Lyapunov Exponent (LLE) that we found positive and which is the signature of a chaotic behavior. Thus, we have found that the characteristic time of this phenomenon should be about 11 minutes. As a consequence, it is possible to define a range of confidence in which we may predict the blood glucose evolution level of type 1 diabetic patients. Out of this range the process is chaotic.