Prediction of Zamorano cheese quality by near-infrared spectroscopy assessing false non-compliance and false compliance at minimum permitted limits stated by designation of origin regulations

Near-infrared transmittance (NIT) spectroscopy was used to predict the percentage in weight of the fat, dry matter, protein and fat/dry matter contents in Zamorano cheeses, protected with a Designation of Origin by the European Union. A total of 42 cheeses submitted to official control were analysed by reference methods. Samples were scanned (850–1050 nm) and predictive equations were developed using Partial Least Squares regression with a cross-validation step. Eight pretreatments independent from the remaining calibration samples were first considered. The most adequate one was that performing the second derivative (using a Savitzky–Golay method with a nine-point window and a second-order polynomial) followed by the standard normal variate transformation. Percentages of the root mean square error in cross-validation, the coefficient of determination and the mean of the absolute value of relative errors found were, respectively, for fat (0.62; 96.16; 1.05), dry matter (0.76; 96.03; 0.83), protein (0.41; 97.82; 0.81) and the fat/dry matter ratio (0.61; 92.51; 0.66). At a 99% confidence level, the trueness of the NIT+PLS methods for fat, dry matter and protein was verified. The official regulation for Zamorano cheese demands minimum permitted limits on the percentages in weight for protein (25%), dry matter (55%) and the ratio of fat to dry matter (45%). The adaptation of both the decision limit and the detection capability to the case of a minimum permitted limit (CDα and CDβ, respectively) when a Partial Least Squares calibration is used has been applied for the first time for a food product protected by a Designation of Origin. The values of CDα with a probability of false non-compliance equal to 0.05 and of CDβ when, in addition, the probability of false compliance was equal to or less than 0.05, both provided by the corresponding NIT+PLS-based method, were, respectively, for protein (24.78%; 24.57%), dry matter (54.14%; 53.28%) and the fat/dry matter ratio (44.39%; 43.78%).