The study compares standard addition (SA), stable isotope dilution assay (SIDA) and multiple headspace extraction (MHE) as methods to quantify furan and 2-methyl-furan in roasted coffee with HS-SPME-GC-MS, using CAR-PDMS as fibre coating, d(4)-furan as internal standard and in-fibre internal standardization with n-undecane to check the fibre reliability. The results on about 150 samples calculated with the three quantitation approaches were all very satisfactory, with coefficient of variation (CV) versus the U.S. Food and Drug Administration (FDA) method, taken as reference, almost always below the arbitrarily-fixed limit of 15%. Furan was detected in the 1-5 ppm range, 2-methyl-furan in the 4-20 ppm range. Moreover, experimental exponential slopes (Q) and linearity (r) of both furan and 2-methyl-furan MHE regression equation on 50 samples were very similar thus making possible to use the same average Q value for all samples of the investigated set and their quantitation with a single determination. This makes this approach very rapid and competitive in-time with SA and SIDA. A non-separative method (HS-SPME-MS) was also developed in view of possible application on-line monitoring of furan and 2-methyl-furan in a pilot-plant with the aim of optimizing the roasting process to reduce these compounds to a minimum. Sampling times of 20 and 5 min were tested, the latter enabling total analysis time to be reduced to about 9 min. The results on 105 samples with both SIDA and MHE approaches were again highly satisfactory most of the samples giving a CV% versus the conventional methods below 20%. In this case too average Q values for both furan and 2-methyl-furan were used for MHE. The separative method presented very good repeatability (RSD% always below 10%) and intermediate precision over three months (RSD% always below 15%); performance were similar for the non-separative method, with repeatability (RSD%) always below 12% and intermediate precision over three months (RSD%) always below 15%. The sensitivity of both separative and non-separative methods was also very good, LOD and LOQ being in the ppb range for both furan and 2-methyl-furan, i.e. well below the amounts present in the roasted coffee samples. (C) 2010 Elsevier B.V. All rights reserved.
