Determination of pharmaceuticals in river water by column switching of large sample volumes and liquid chromatography-diode array detection, assisted by chemometrics. An integrated approach to green analytical methodologies.

MARÍA MARTÍNEZ GALERA; MARÍA DOLORES GIL GARCÍA; MARÍA JULIA CULZONI; HÉCTOR CASIMIRO GOICOECHEA

JOURNAL OF CHROMATOGRAPHY - A

ELSEVIER SCIENCE BV

Año: 2010 p. 2042 - 2049

0021-9673

An analytical method for the simultaneous determination of nine -blockers (sotalol atenolol, nadolol, pindolol, metoprolol, timolol, bisoprolol, propanolol and betaxolol) and two analgesics (paracetamol and phenazone) in river water by liquid chromatography and diode array detection is reported. The method involves a modified precolumn switching methodology replacing the small precolumn with a short C18 liquid chromatography column (50mm×4.6mm,5mparticle size), thus allowing the preconcentration of large water sample volumes whereas interferences eluting at the first of the chromatogram were discarded to waste. This approach allowed to preconcentrate 30mL river water samples, modified with 0.4% MeOH, achieving univariate method detection and determination limits ranged between 0.03 and 0.16gL−1 and between 0.2 and 0.5gL−1, respectively, with precision values lower than 9.4% for spiking levels at the quantitation limits of each analyte and lower than 4.0%, except bisoprolol (8.3%), for higher spiking levels (1.0gL−1 of all analytes). Matrix background was reduced in three way data by a baseline correction following the Eilers methodology, whereas multivariate curve resolution and alternating least squares in combination with the standard addition calibration method, applied to these data, coped with overlapping peak, systematic (additive) and proportional (matrix effect) errors. The method was successfully applied for the determination of the target pharmaceuticals in river water from three places in a river stream with acceptable recoveries and precision values, taking into account the complexity of the analytical problem. The joint statistical test for the slope and the intercept of the linear regression between the nominal concentration values versus those predicted, showed that the region computed contained the theoretically expected values (0) for the intercept and (1) for the slope (at a confidence level of 95%), which indicates the absence of both constant and proportional errors in the predicted concentrations.