The clean-up procedure involved in the trace analysis of polychlorinat
ed biphenyls (PCBs) was evaluated by studying GC-MS chromatograms of P
CB mixtures after the subsequent steps of the procedure. The complex c
hromatograms obtained were evaluated using two methods: the individual
PCB congener method and a chemometric approach based on a study of th
e autocovariance function. The results obtained are statistically comp
arable, proving that the statistical approach is able to determine the
total PCB content in the sample quantitatively. Moreover, since the a
utocovariance method is based on statistical evaluation of the whole c
omplex chromatogram, it can overcome the problems that usually arise b
ecause of peak overlapping in PCB mixture chromatograms. It also provi
des accurate results when the chromatogram shows interfering peaks and
low resolution. This is the case with Aroclor 1242, where some compou
nds released by the cartridges strongly interfere with the analysis, r
esulting in errors in the quantification of individual PCBs, Highly ch
lorinated PCBs (i.e., Aroclor 1260) can be quantitatively recovered (m
ean recovery, 100 +/- 1%), In contrast, lower proportions (less than 7
5%) of the less chlorinated compounds (i.e,, those containing two and
three chlorine atoms) are recovered because they are selectively retai
ned by silica and alumina columns.