Determination of the Sugar Content in Commercial Plant Milks by Near Infrared Spectroscopy and Luff-Schoorl Total Glucose Titration

Thirty-nine samples of plant milks (rice, soy, oat, barley, spelt, quinoa, almond, and a variety of wheat called kamut) were analyzed for their reducing sugars content by NIR spectroscopy, using the Luff-Schoorl official method as reference to build the calibration models. The amount of reducing sugars, expressed as grams of glucose/100 mL of beverage, ranged from 0.5 g/100 mL (soy) to 7.6 g/100 mL (rice). Both partial least squares (PLS) and interval-partial least squares regression (iPLS) were used to build multivariate calibration models, testing different spectra preprocessing methods. The performance in prediction of the best calibration model was evaluated on an external test set of nine randomly selected samples (RMSEP = 0.98 g/100 mL, R2PRED = 0.84), and its statistical significance was assessed using a randomization t test based on Monte Carlo simulations. The results obtained suggest that NIR spectroscopy can be a valid alternative to the laborious reference titrimetric method for the determination of total glucose content in plant milks.