Combining 1H NMR spectroscopy and multivariate regression techniques to quantitatively determine falsification of porcine heparin with bovine species.

(1)H NMR spectroscopy was used to distinguish pure porcine heparin and porcine heparin blended with bovine species and to quantify the degree of such adulteration. For multivariate modelling several statistical methods such as partial least squares regression (PLS), ridge regression (RR), stepwise regression with variable selection (SR), stepwise principal component regression (SPCR) were utilized for modeling NMR data of in-house prepared blends (n=80). The models were exhaustively validated using independent test and prediction sets. PLS and RR showed the best performance for estimating heparin falsification regarding its animal origin with the limit of detection (LOD) and root mean square error of validation (RMSEV) below 2% w/w and 1% w/w, respectively. Reproducibility expressed in coefficients of variation was estimated to be below 10% starting from approximately 5% w/w of bovine adulteration. Acceptable calibration model was obtained by SPCR, by its application range was limited, whereas SR is least recommended for heparin matrix. The developed method was found to be applicable also to heparinoid matrix (not purified heparin). In this case root mean square of prediction (RMSEP) and LOD were approximately 7% w/w and 8% w/w, respectively. The simple and cheap NMR method is recommended for screening of heparin animal origin in parallel with official NMR test of heparin authenticity and purity.