Solving matrix-effects exploiting the second order advantage in the resolution and determination of eight tetracycline antibiotics in effluent wastewater by modelling liquid chromatography data with multivariate curve resolution-alternating least squares

DE ZAN,; GIL GARCÍA, MARÍA; CULZONI MARÍA JULIA; GABRIEL SIANO; MARTINEZ GALERA, MARIA; GOICOECHEA, HÉCTOR C

JOURNAL OF CHROMATOGRAPHY - A

Elsevier

Lugar: Amaterdam; Año: 2008 vol. 1179 p. 108 - 114

0021-9673

The effect of Piecewise Direct Standardization (PDS) and Baseline Correction approaches was evaluated in the performance of Multivariate Curve Resolution (MCR-ALS) algorithm for the resolution of three-way data sets from liquid chromatography with diode-array detection (LC-DAD). Firstly, eight tetracyclines (tetracycline, oxytetracycline, chlorotetracycline, demeclocycline, methacycline, doxycycline, meclocycline and minocycline) were isolated from 250 mL effluent wastewater samples by Solid Phase Extraction (SPE) with Oasis MAX 500 mg / 6 mL cartridges and then separated  on an Aquasil C18 150 mm × 4.6 mm (5 mm particle size) column by LC and detected by DAD. Previous experiments, carried out with Milli-Q water samples, showed a considerable loss of the most polar analytes (minocycline, oxitetracycline and tetracycline) due to breakthrough. PDS was applied to overcome this important drawback. Conversion of chromatograms obtained from standards prepared in solvent was performed obtaining a high correlation with those corresponding to the real situation (r2 = 0.98). Although the enrichment and clean-up steps were carefully optimized, the sample matrix caused a large baseline drift, and also additive interferences were present at the retention times of the analytes. These problems were solved with the baseline correction method proposed by Eilers. MCR-ALS was applied to the corrected and uncorrected three-way data sets to obtain spectral and chromatographic profiles of each tetracycline, as well as those corresponding to the co-eluting interferences.   The complexity of the calibration model built from uncorrected data sets was higher, as expected, and the quality of the spectral and chromatographic profiles was worse.