16 Quality Management
geographical origin of an olive oil. This is possible because the fingerprints
of e.g. Spanish, Italian or Greek olive oils are slightly different from each
other. These differences are often hard to see visually in a single NMR-fingerprint
or even invisible to the human eye, but statistics are used to make
these differences visible and to assign a country of origin to a sample.
A great way to visualize olive oil fingerprints are so-called quantile
plots. An example of such a quantile plot is given in figure 2. The black line
represents the measured sample, an authentic olive oil from Morocco in
this case, and the colored background reflects data from the database of
authentic samples. The NMR fingerprint of authentic olive oil is expected
to be within the colored envelope. The red band represents the most frequently,
the blue band less commonly detected signal intensities in authentic
samples. If a sample shows signal intensities outside the range of the
authentic samples in the database, this is an indication that something is
wrong and the sample is judged as not authentic.
Adulteration of olive oil with other oil can easily be recognized in
this way. Figure 3 depicts the NMR-fingerprints of adulterated olive oil.
Canola, sunflower and soya oil were used for adulteration. Significant
deviation from of the NMR fingerprint from database data is indicating
a discrepancy from the profile of authentic samples. In addition to the
comparison of each sample’s profile to the database, five parameters
of each sample are quantified and put into relation to the average values
all authentic olive oil samples in the database.
Evaluation of the quality of olive oil is also possible by application of
linear discriminant analysis (LDS) to the complex fingerprint of all olive
oils in the database. The result of the LDA of database samples for the
grades extra virgin (nativ extra), virgin (nativ) and lampant is depicted
in figure 4. The three groups of samples are very well separated. An
unknown sample could now be tested against the database and its
quality could be derived based on this data.
In the same way, the country of origin can be made differences visible.
Again, an LDA is applied to the complex fingerprint of all olive oils in the
database makes it possible to separate them in different groups according
to their geographical origin. An unknown sample can be analyzed the
same way and depending on where it shows up in the graph, the analyst
can derive if the geographical origin stated on the label is true or false.
Complex matrices need complex analytical methods. For a matrix like
olive oil with its diverse qualities and geographical origins and the extremely
high risk of adulteration, NMR-fingerprinting has shown to be a highly
valuable screening tool to uncover fraudulent products. However, as for
many other matrices and methods, there is no one-size-fits-all solution.
This is also true for NMR-fingerprinting. Though very powerful and a perfect
screening method, in certain cases additional (targeted) methods
might be necessary to complement the NMR-fingerprint.
Screening with NMR-fingerprinting yields a comprehensive view
on the olive oil sample in question on which decisions for further analysis
can easily be made.
The Tentamus Center for Food Fraud (TCF²) would like to thank Quality
Services International (QSI), Bremen, Germany, Laboratorio Tello, Jaén,
Spain and Analytical Food Laboratories (AFL), Grand Prairie, Texas, for
sharing data and figures on this topic.
Figure 2: Quantile plot of an authentic olive oil sample from
Morocco (numbers at x-axis are the chemical shifts in ppm).
Spectra were obtained using CDCl3:DMSO-d6 (9:1) as solvent.
The NMR fingerprint contains the fatty acid composition of
the sample in the region from 0.6-6.0 ppm. Compounds that
are associated with positive health effects, like α-tocopherol
as well as oxidation products, like hydroperoxides and aldehyds,
can be found in the region between 6.0-11.0 ppm.
Figure 3: Section of the NMR fingerprint of an olive oil
sample adulterated with 10% canola oil (A), 20% sunflower
oil (B) and 10% soya oil (C). Deviation of samples from
data in the database of authentic samples is emphasized
by using red color for deviating signals in the fingerprint.
Figure 4: Separation of olive oil samples by quality using
statistical methods, e.g. linear discriminant analysis (LDA).