In this work, we evaluated different sets of features derived from wearable physiological signals to assess their feasibility in predicting interstitial glucose (IG) values for a patient with Type 1 Diabetes Mellitus (T1DM) over a 30-min prediction horizon. These features were tested both independently and in combination with data from a Continuous Glucose Monitoring (CGM) sensor. In addition, we analyzed predictions based solely on CGM data. For each prediction horizon, a machine learning regression model was trained using the corresponding set of characteristics with a 10-fold cross validation scheme. Model performance was assessed using analytical metrics and the Clarke Error Grid was used to evaluate clinical relevance. The best results were obtained for the five-minute prediction horizon when combining wearable sensor features with CGM data, achieving a MAE of 5.18 mg/dL, a RMSE of 7.62 mg/dL, and an R \(^2\) of 0.95, all predicted values falling within clinically acceptable zones. These findings highlight the potential of incorporating wearable sensor data to improve glucose prediction accuracy.