Journal of Chromatography A, 1299 (2013) 94–102

Contents lists available at SciVerse ScienceDirect

Journal of Chromatography A
journal homepage: www.elsevier.com/locate/chroma

A database of chromatographic properties and mass spectra of fatty

acid methyl esters from omega-3 products
Ziar Wasta a , Svein A. Mjøs a,b,∗
a
Department of Chemistry, University of Bergen, P.O. Box 7803, N-5020 Bergen, Norway
b
Nofima BioLab, Kjerreidviken 16, NO-5141 Fyllingsdalen, Norway

a r t i c l e i n f o a b s t r a c t

Article history: Fatty acids in products claimed to contain oils with the omega-3 fatty acids eicosapentaenoic acid (EPA)
Received 2 March 2013 and docosahexaenoic acid (DHA) were analyzed as fatty acid methyl esters by gas chromatography–mass
Received in revised form 9 May 2013 spectrometry using electron impact ionization. To cover the variation in products on the market, the 20
Accepted 22 May 2013
products that were studied in detail were selected from a larger sample set by statistical methodology.
Available online 29 May 2013
The samples were analyzed on two different stationary phases (polyethylene glycol and cyanopropyl)
and the fatty acid methyl esters were identified by methodology that combines the mass spectra and
Keywords:
retention indices into a single score value. More that 100 fatty acids had a chromatographic area above
Fatty acid methyl esters
Gas chromatography–mass spectrometry
0.1% of the total, in at least one product. Retention indices are reported as equivalent chain lengths, and
Retention indices overlap patterns on the two columns are discussed. Both columns were found suitable for analysis of
Equivalent chain lengths major and nutritionally important fatty acids, but the large number of minor compounds that may act as
Omega-3 products interferents will be problematic if low limits of quantification are required in analyses of similar sample
Marine fatty acids types. A database of mass spectral libraries and equivalent chain lengths of the detected compounds has
been compiled and is available online.
© 2013 Elsevier B.V. All rights reserved.

1. Introduction
The long chain polyunsaturated omega-3 fatty acids eicosapen-
taenoic acid (EPA) and docosahexaenoic acid (DHA) are abundant in
most lipids of marine origin. They are claimed to have several health
effects, such as reduced inflammation [1–3] and protection against
cardiovascular disease [1,4,5]. They have critical roles in brain func-
tion and neurodevelopment [1,5–7] and enhanced intake may have
positive effects on psychological and cognitive function [1,5,6].
The human body has a limited ability to synthesize EPA and DHA
from alpha linolenic acid (ALA), and in most western diets there is a
gap between the recommendations and the average intake [8–10].
As a result, a large number of pharmaceutical products and food
supplements containing omega-3 fatty acids are available. The raw
material and production methods vary, and in a recent study it was
shown that products on the market had extreme variations in the
content of EPA and DHA and in the general fatty acid profile [11].
Other fatty acids than EPA and DHA may be of significance. This may
be other omega-3 fatty acids that have similar effects or that can be
converted to EPA and DHA, such as stearidonic acid and docosapen-
taenoic acid (DPA). Since several of the positive effects of omega-3
∗ Corresponding author at: Department of Chemistry, University of Bergen, P.O.
Box 7803, N-5020 Bergen, Norway. Tel.: +47 5558 3553; fax: +47 5558 9490.
E-mail address: svein.mjos@kj.uib.no (S.A. Mjøs).
fatty acids are linked to the balance between omega-6 and omega-3
fatty acids [2,3] correct assessment of the omega-6 content of the
products is of relevance. In addition to the omega-3 and omega 6
series of polyunsaturated fatty acids (PUFA) the products also con-
tain a large number of other fatty acids, for which there is limited
knowledge about the health effects. This include other PUFA series,
such as omega-1 and omega-4, long chain monoenoic fatty acids
(MUFA), diunsaturated fatty acids (DUFA), branched fatty acids, and
furan fatty acids. The biological roles of many of these compounds
are not fully understood, but it has for instance been speculated
that the furan fatty acids may be partly responsible for some of the
positive effects that are assigned to omega-3 fatty acids [12].
As a consequence it is important to have more detailed knowl-
edge about the fatty acid composition of the omega-3 products, not
only to assess the nutritional value of the products on the market,
but also because omega-3 supplements are used in studies of lipid
metabolism and nutrition. Most commercial fish oils have been pre-
viously analyzed in detail and there is general consensus on the
identity of major fatty acids of nutritional importance. However,
the large variety of alternative sources and production methods,
such as urea fractionation, molecular distillation and preparative
chromatography, lead to enrichment of many minor fatty acids in
omega-3 products. The purpose of this work has been to identify
fatty acids that are present in the products down to 0.1%, and to
provide information that may facilitate identification of fatty acids
in products of marine origin.

0021-9673/$ – see front matter © 2013 Elsevier B.V. All rights reserved.
http://dx.doi.org/10.1016/j.chroma.2013.05.056
 Z. Wasta, S.A. Mjøs / J. Chromatogr. A 1299 (2013) 94–102
Marine fatty acids are typically analyzed by gas chromatography
(GC) as fatty acid methyl esters (FAME), using polar columns with
polyethylene glycol (PEG) or cyanopropyl (CP) stationary phases. In
this work, selected products were analyzed on both types of phases
using electron impact mass spectrometry (MS) for detection.
The compounds were identified using novel methodology that
combines the mass spectral similarity and retention index match
into a single score. Databases of mass spectra and retention indices
on both columns were thereafter built. Overlap patterns were eval-
uated from the difference in retention indices and a threshold level
based on the separation efficiency.
2. Theory
2.1. Nomenclature
The applied fatty acid nomenclature for common saturated and
unsaturated fatty acids follows the shorthand notation a:b n−c,
where a designates the number of carbons in the fatty acid chain, b
designates the number of double bonds, and c designates the posi-
tion of the first double bond relative to the methyl end of the carbon
chain. When this system is used it is assumed that all double bonds
have cis geometry and that double bonds are separated by a single
methylene group if there is more than one.
The major omega-3 fatty acids 20:5 n−3 and 22:6 n−3 may be
referred to in the text by their common abbreviations EPA (eicosa-
pentaenoic acid) and DHA (docosahexaenoic acid). Trans isomers
of EPA and DHA are designated by the system used in [11,13].
Furan fatty acids are designated by the system used in [14]. Methyl
branched saturated ‘iso-’ and ‘anteiso’ fatty acids are designated by
‘i-’ and ‘ai-’ followed by the number of carbons.
In addition to these short hand notations, a system with a unique
code of three letters and a running number, and a long name that
usually specifies the position and geometry of double bonds or
a full systematic name, are applied by the software used in this
study. The tri-unsaturated fatty with common name ␣-linolenic
acid will be specified in this system by short name 18:3 n−3, long
name c9,c12,c15-18:3 and the code POU-032. ‘POU’ designates a
polyunsaturated fatty acid. Other common three letter codes are,
SAN for saturated unbranched, SAB for saturated branched, MOU
for monounsaturated, DIU for diunsaturated and FUR for furan fatty
acids. Codes and long names are used in tables in addition to the
short names. A comprehensive list of fatty acids and related com-
pounds in this system can be found at www.chrombox.org/data.
2.2. Retention indices of FAME
Retention indices based on homologous series of reference com-
pounds are widely applied for identification of analytes in gas
chromatography. Kováts indices based on n-alkanes as references
is the dominating general-purpose system, but a large number of
alternative systems with other calibration compounds have been
developed for special purposes [15]. Equivalent chain lengths (ECL)
[16,17] are applied for analysis of fatty acid derivatives. The ECL
system uses the saturated straight chain FAMEs as reference com-
pounds and the ECL value of the references are by definition equal
to the number of carbons in the saturated fatty acid chain. In
temperature-programmed GC, the ECL of a compound i can be cal-
culated by a modification of the van den Dool and Kratz equation
originally developed for Kováts indices [18]:
tR(i) − tR(z)
ECL(i) = u +z (1)
tR(z+u) − tR(z)
where tR(i) is the retention time of compound i, tR(z) is the reten-
tion time of a saturated unbranched FAME eluting before i and z
is the number of carbons in the fatty acid chain of this molecule
95
(excluding the carbon in the methyl group in the FAME). tR(z+u) is
the retention time of a saturated unbranched FAME eluting after i
and u is the difference in the number of carbons between the two
saturated FAMEs. Ideally, u should be 1.
The fractional chain length (FCL) of a FAME molecule is defined
as the difference between the ECL value and the number of carbons
in the fatty acid chain. FCL is therefore calculated by the following
equation:
FCL(i) = ECL(i) − nC(i) (2)
where FCL(i) is the fractional chain length of compound (i) and nC(i)
is the number of carbons in (i), excluding the carbon in the methyl
group.
2.3. GC–MS of FAME
Although there are alternative derivatives that often give more
informative spectra [19,20], fatty acids are often analyzed by elec-
tron impact MS of FAME. Important reasons for preferring FAME
are the ease of preparation and good chromatographic properties.
FAMEs are also the preferred derivatives in GC analyses of fatty
acids by flame ionization detection (FID) and the use of FAMEs with
GC–MS instead of other derivatives may facilitate optimization and
method transfer to GC–FID.
Overviews of FAME fragmentation and important diagnostic
ions are given in [21–25]. Mass spectra of FAMEs can also be found
in commercial MS libraries, but it should be emphasized that many
FAMEs have highly similar spectra, and spectral differences caused
by different instruments may be larger than the differences caused
by molecular structure [26]. It may therefore be necessary to com-
bine mass spectral information with retention indices for reliable
identifications. This can for instance be done as explained in Section
3.5.
Most electron impact spectra of FAMEs have abundant frag-
ments in the region below m/z 110, which may be referred to as
the fingerprint region [23]. Except that spectra can be assigned to
the main classes saturated, monounsaturated, diunsaturated and
polyunsaturated fatty acids, the fingerprint region contains little
information suitable for manual interpretation. However, the use
of mathematical methods on spectra from this region can provide
additional information [23,27,28]. The fingerprint may also be suit-
able for searching in libraries of already identified compounds.
Since the range of masses are smaller than in full scan spectra,
the spectra are less affected by the sensitivity for large ions that
often vary between instruments [26]. Scanning only the fingerprint
region when data are acquired can be regarded as a compromise
between acquiring full scan spectra and using selected ion moni-
toring (SIM). Higher signal strength and spectra of better quality
can be achieved because time is not wasted by scanning for weak
or absent ions.
3. Materials and methods
3.1. Samples
The samples in this study consisted of 20 different fish oils and
omega-3 products. The majority of the samples are a subset of
samples from a previous study of fatty acids in 77 omega-3 prod-
ucts purchased on the European market (Italy, Spain, UK, Germany,
Norway and France) [11]. Eight samples that are representative for
large groups of products, i.e. cod liver oil, seal oil, typical South
American oils with approximately 18% EPA and 12% DHA (‘18/12-
oils’) and products having approximately 33% EPA and 23% DHA
(‘33/23 concentrates’) were selected from a principal component
analysis (PCA) score plot of the fatty acid profile from the previous
study [11]. PCA will explain the main trends in a data set but may
96 Z. Wasta, S.A. Mjøs / J. Chromatogr. A 1299 (2013) 94–102

Table 1
Product overview.

No Product type Product/vendor Country of origin Weight % EPA Weight % DHA

5 Algal + vegetable oila DHA Neuromins/Solaray ES 0.01 22.68

6 High DHAa Oligen DHA/Ifigen ES 4.35 79.92
8 Cod liver oila Higabac/Soria Natural ES 7.50 8.61
9 High EPAa AquiOmega/Uriach Aquilea ES 53.21 23.32
10 33/23 concentrate ArkOmega3/Arkopharma ES 35.03 23.94
13 Cod liver oil Aceite de Higado de Bacalao/Santiveri ES 8.67 11.76
17 33/23 concentrate Feniko, Seefishöl konzentrat/Urgo DE 30.84 23.77
18 18/12 oil Eicosan 750/Truw DE 17.44 12.23
23 18/12 oil Omega-3 Lachsöl-kaspseln/Feingold DE 17.71 11.50
25 18/12 oil Ameu 500 mg/Casella med DE 15.00 11.26
27 Seal oila Sel-Olje/Biopharma, Nordsveen NO 6.61 9.42
61 High EPAa Pulse Cardiomax/Seven Seas UK 57.32 15.05
64 Fish + vegetable oila Eye Q/Equazen UK 17.91 5.68
65 High DHAa Efalex Active 50+/Efamol UK 8.65 53.01
73 33/23 concentrate Omega 3/Vitarmoryl IT 32.41 23.88
76 High EPAa Esapent/Pfizer IT 45.07 41.92
77 Cod liver oil Møllers tran/Møller Collet NO 9.12 13.03
78 Salmon oil (farmed salmon)a Super Omega-3/Shift NO 2.97c 4.74c
79 Krill oil (Superba)a Omege Red/Medica Nord NO 15.89c 8.40c
80 Calamaris oila Vitomaris/Vitomega NO 17.50c 40.78c
KP Herring oila , b Control sample/Nofima NO 4.63c 5.30c
a
Regarded as product with unique FA composition, other samples are representative for a group (CLO, 18:12 oil or 33/23 concentrate).
b
Control sample at Nofima BioLab, not a commercial product.
c
Estimated by GC–MS (BP-20 short scan), other values are analyzed by GC–FID as described in [11].

not adequately explain single objects with unique features. A math-
ematical method for selecting extreme objects [29] was therefore
used to pick eight additional samples having the most extreme fatty
acid compositions. Four sample types that were not represented in
the original study [11], squid oil, salmon oil from farmed salmon,
pure krill oil and herring oil were also added to the sample set. The
products included in the study are listed in Table 1 and the initial
sample set is further described in [11].
3.2. Sample preparation
Fatty acid methyl esters were prepared from two drops of the
products (25–35 mg, depending on viscosity) by heating at 100 ◦ C
for 5 min in 1.5 mL 0.5 N methanolic NaOH followed by heating
at 100 ◦ C for 10 min after addition of 2.0 mL 12% methanolic BF3
(Sigma–Aldrich). Samples where thereafter vortex mixed for 30 s
after addition of 1 mL isooctane, and 5 mL saturated aqueous NaCl
solution was added. The samples were vortexed again and the
isooctane layer was collected. The extraction with isooctane was
thereafter repeated and the isooctane layer was combined with the
first aliquot. The isooctane extract was thereafter diluted further to
two different concentrations of 1:100 and 3:100 in isooctane and
1 ␮l of each dilution was injected on the GC–MS.
3.3. Reference compounds
A mixture of all saturated FAMEs from 12:0 to 28:0 (except
23:0) was used for calibration of the relationship between retention
times and ECL values. The GLC-793 reference mixture that contains
28 common fatty acids (Nu-Chek Prep. Elysian, MN, USA) was also
analyzed on a routine basis. In addition, 8 more reference mixtures
containing less common fatty acids were analyzed once in each
sequence. A full list of the content in these mixtures is given in the
supplementary material, Table S1.
3.4. Analytical conditions and sample sequence
All samples and standards were analyzed by two different
GC columns, one polyethylene glycol column, BP-20 (L = 50 m,
i.d. = 0.22 mm, df = 0.25 ␮m) and one cyanopropyl column, BPX-70
(L = 60 m, i.d. = 0.25 mm, df = 0.25 ␮m). Both columns were from SGE
(Ringwood, Australia).
The BP-20 column was applied on an Agilent (Santa Clara, CA,
USA) 6890/5975 GC–MS system. The samples were injected at
an oven temperature of 60 ◦ C. After 4 min the temperature was
increased by 30 ◦ C/min to 160 ◦ C followed by 1.4 ◦ C/min to 260 ◦ C.
Estimated average carrier gas velocity was 30 cm/s (constant flow
mode). MS interface, ion source and mass filter temperatures were
300 ◦ C, 230 ◦ C and 150 ◦ C, respectively.
The BPX-70 column was applied on an Agilent 5890/5972
GC–MS system. The samples were injected at an oven temperature
of 60 ◦ C. After 4 min the temperature was increased by 30 ◦ C/min
to 160 ◦ C followed by 2.0 ◦ C/min to 260 ◦ C. Estimated average car-
rier gas velocity was 26 cm/s (constant flow mode). MS interface
temperature was 270 ◦ C.
On both systems the 1 ␮L sample was injected by autosamplers.
The injection was done splitless (split valve opened after 4 min)
with an injector temperature of 250 ◦ C. Helium (>99.996%) was
used as carrier gas. Both the 1:100 and 3:100 concentrations of
the samples were analyzed using both full MS scan from m/z 50
to 430 and short MS scan from m/z 50 to 110. Scan frequencies
for the full scan were 0.97 s−1 and 1.18 s−1 for the 6890/5975 and
the 5890/5972 systems, respectively. Corresponding values for the
short scan were 2.79 s−1 and 2.01 s−1 .
The analytical sequences were set up so that the saturated FAME
mixture and the GLC-793 reference mixture were analyzed for
every 10th sample and at the beginning and at the end of the
sequences. Other reference mixtures and the two concentrations
of each sample were analyzed once in each sequence. In total each
compound were analyzed eight times (two different columns × two
scan modes × two concentrations)
3.5. Chromatographic data handling
The handling of GC–MS data was performed in Chrombox Q
(12-01) running under Matlab 7.9. Chrombox Q is designed for
identification of compounds by combining information from mass
spectra and retention indices. In addition it has the possibility
for deconvolution of overlapping peaks using the alternating least
squares principle [30,31]. Initial estimates for pure spectra or
pure chromatographic profiles are estimated according to [29]. A
 Z. Wasta, S.A. Mjøs / J. Chromatogr. A 1299 (2013) 94–102
Fig. 1. Principle for calculating retention score from difference in retention index
(RI) and window size (w).
description of the applied methodology for identification is given
below.
Spectrum scores (SS ) are calculated from the regression coeffi-
cient (r) between library spectra and the spectrum that is tested
according to Eq. (3).
SS = 50r + 50
Two identical spectra will get a score of 100 because r is equal to
1. Uncorrelated spectra will get a score of 50 because r is 0, and
negative correlations will give scores from 0 to 50. In Chrombox Q
the information about the scanned mass range is stored with the
spectra. It is therefore possible to match spectra from short scans
or selected ion monitoring against full scan spectra and vice versa.
The calibrations between retention times and ECL values were
calculated by second order local regressions as explained in [32].
Retention scores (SR ) based on the difference in retention index
between library indices and the tested compound is calculated
according to Eq. (4).
2
SR = e−32(RI/w)
RI is the difference in retention index units between the two
compared indices, in this case ECL values. w defines a window of
retention indices around the compound that is tested. Any com-
pound with RI outside ±0.5w is excluded from the test. e is the
base of the natural logarithm. This is in principle equivalent to posi-
tion a normal distribution curve with maximum 1 and standard
deviation of 8/w within the window and reading the retention score
from the curve. The principle is illustrated in Fig. 1, where two com-
pounds with ECL of 0.11 will be given a SR of 0.62 if the window
size is 0.9.
A total score (ST ) is calculated by combining SS and SR according
to Eq. (5).
ST = SS · SR
Adjusting the window size in Eq. (4) is a practical way of adjusting
the weight between SS and SR . By increasing the window size the
penalty for differences in retention indices will be lower and more
weight will be put on the spectral similarity. By experience we have
found that a window size of 0.9 works well with FAME and ECL, and
this value was therefore applied.
Identification was to start with based on existing libraries
(www.chrombox.org/data) that have been built from previous
projects [13,23,26,32–37]. A few compounds (basically branched
fatty acids) was identified by search in the NIST 05 MS library
(National Institute of Standards and Technology, Gaithersburg, MD,
USA) by linking Chrombox Q to the MS Search 2.0 software (NIST).
New libraries for each of the two columns were gradually built
using the following criteria. Any compound that had a chromato-
graphic area above 0.1% of the total area in a sample should be
97
assigned a code and added to the library. Fronting peaks with asym-
metry factor [38] below 0.75 should not be included in the libraries.
Retention indices for all samples and standards were always cali-
brated against the nearest saturated FAME mixture in the analytical
sequence.
4. Results and discussion
4.1. Overview
In total 103 different compounds were detected above the
threshold of 0.1%, and 101 of these were identified as FAME. Of
these there were 14 FAME from normal SFA, 28 from normal MUFA,
9 from DUFA, 32 from PUFA, 12 from branched SFA or MUFA, and 6
from furan fatty acids. The compounds are listed in Table 2 together
with maximum abundance and ECL values on the two columns.
Standard deviations of the ECL values of all compounds that were
saved to the libraries were below 0.01 and 0.02 ECL units on BP-20
and BPX-70, respectively.
All normal SFA from 8:0 to 24:0, except, 9:0, 11:0 and 13:0, were
detected above 0.1%. SFA with odd number of carbon atoms, partic-
(3) ularly 19:0, 21:0 and 23:0 are commonly used as internal standard
in quantitative analysis of marine fatty acids. The maximal levels of
these were so low (0.19%, 0.58% and 0.11%) that their presence in
the products should have limited influence on quantitative studies
as long as a sufficient amount of internal standard is added.
The 28 MUFA had carbon numbers varying from C14 to C26 with
double bonds from n−4 to n−13. In addition, a few late eluting
MUFAs with double bonds in unknown positions were detected.
These are probably n−3, n−2 or n−1. MUFAs were basically iden-
tified from their retention indices in existing databases from [36].
The different MUFA series are basically produced by 9-desaturase
giving n−5, n−7, n−9 and n−11 from 14:0, 16:0, 18:0 and 20:0,
respectively [39]. The n−11 series that is abundant in oil from North
(4) Atlantic species, such as cod liver oil and seal oil, can also arise from
metabolism of the 22:1 n−11 fatty alcohol from Calanus [40].
The majority of abundant DUFA belonged to the n−6 series,
which is dominating in most living organisms. Minor amounts of
DUFA belonging to the n−7 and n−4 series were also found. The
n−7 and n−4 series are common in marine micro algae [41,42].
Two MI DUFA with unknown double bond positions (18:2 n−x and
20:2 n−x) and an NMI 20:2 were also detected in cod liver oils and
seal oil. The double bond positions in the 20:2 NMI are unknown,
but 5,11-20:2; 5,13-20:2; 7,13-20:2 and 7,15-20:2 are frequently
reported in shellfish and may be found in lipids of species feeding
on bottom organisms.
The majority of detected PUFA belonged to the common n−3
and n−6 series but the n−4 and n−1 series were also abundant.
These two series originate basically from 16:2 n−7, which is desat-
(5)
urated to 16:3 n−4 and 16:4 n−1 [39]. These compounds can then
be metabolized further to other members of the same series. Two
PUFA with odd number of carbon atoms were found. 21:5 n−3 is
frequently reported in marine lipids [43,44]. 19:5 n−3 was also
detected in some samples with high amounts of 21:5 n−3. Small
amounts of 24:5 n−3 and 24:6 n−3, which are important interme-
diates in the formation of 22:6 n−3 [39], were also detected. Trans
isomers of 22:5 n−3 and 22:6 n−3, probably occurring as a result of
high temperature oil refining [11], could be detected in most prod-
ucts. All previously reported mono-trans isomers [13,39], except
DHA isomer D1c, were found above the threshold. The D1c isomer
is also included in Table 2 to have a complete list of the mono-trans
DHA isomers. No other peaks that could be assigned to common
trans fatty acids were detected in the study.
The majority of the branched fatty acids belonged to
the common saturated iso and anteiso series. The saturated
98 Z. Wasta, S.A. Mjøs / J. Chromatogr. A 1299 (2013) 94–102

Table 2
Codes, short names and long names, maximal area percent and ECL on BP-20 and BPX-70.

Code Short name Long name Max Apct ECL BP-20* ECL BPX-70*

SAN-001 8:0 8:0 2.60 8.000 8.000

SAN-002 10:0 10:0 1.88 10.000 10.000
SAN-003 12:0 12:0 2.83 11.999 12.000
SAN-005 14:0 14:0 13.43 14.000 14.000a
SAB-324 4,8,12-Me 13:0 4,8,12-Trimethyltridecanoic acid 0.28 14.083 14.011a
MOU-020 14:1 n−5 c9-14:1 0.72 14.386 14.598
SAB-078 i-15:0 13-Methyltetradecanoic acid 0.27 14.533 14.508
SAB-077 ai-15:0 12-Methyltetradecanoic acid 0.12 14.678 14.710
SAN-006 15:0 15:0 0.55 15.001 15.001
SAB-072 i-16:0 14-Methylpentadecanoic acid 0.14 15.527 15.504b
SAB-070 Pristanic acid 2,6,10,14-Tetramethylpentadecanoic acid 0.32 15.792 15.534b
SAN-007 16:0 16:0 26.13 16.000 16.001
MOU-274 16:1 n−11 c5-16:1 0.56 16.126 16.225
MOU-275 16:1 n−9 c7-16:1 0.38 16.198 16.378
MOU-021 16:1 n−7 c9-16:1 17.89 16.261 16.472c
MOU-255 16:1 n−5 c11-16:1 0.51 16.383 16.607
MOU-297 16:1 n−x x-16:1 0.27 16.516a SAB-071d
SAB-074 i-17:0 15-Methylhexadecanoic acid 0.97 16.525a 16.496c
MOB-289 16:1 n−10, 7Me (a) 7-Methyl-6-hexadecenoic acid (a) 0.17 16.544a 16.548
DIU-091 16:2 n−7 c6,c9-16:2 0.19 16.598 16.967f
SAB-073 ai-17:0 14-Methylhexadecanoic acid 0.24 16.677 16.710
DIU-201 16:2 n−4 c9,c12-16:2 1.26 16.831 17.253
MOB-286 16:1 n−10, 7Me (b) 7-Methyl-6-hexadecenoic acid (b) 0.31 16.889 16.903
SAB-071 Phytanic acid 3,7,11,15-Tetramethylhexadecanoic acid 2.03 16.955b 16.763de
SAB-599 Phytanic acid isom. Tetramethylhexadecanoic acid isomer 0.13 16.978bc 16.805e
SAN-008 17:0 17:0 0.52 17.001c 17.002f
POU-046 16:3 n−4 c6,c9,c12-16:3 1.82 17.155 17.720g
MOU-436 17:1 n−8 c9-17:1 0.41 17.221 17.429
MOU-480 17:1 n−4 c13-17:1 0.17 17.467d 17.683g
SAB-076 i-18:0 16-Methylheptadecanoic acid 0.15 17.495d 17.479
POU-051 16:4 n−3 c4,c7,c10,c13-16:4 0.13 17.550 18.073
POU-052 16:4 n−1 c6,c9,c12,c15-16:4 3.40 17.708 18.392h
SAN-009 18:0 18:0 6.76 18.001 18.002
MOU-098 18:1 n−13 c5-18:1 0.13 MOU-277e 18.232
MOU-277 18:1 n−11 c7-18:1 4.65 18.170ef 18.326
MOU-023 18:1 n−9 c9-18:1 65.52 18.187f 18.389h
MOU-079 18:1 n−7 c11-18:1 7.25 18.261 18.476
MOU-278 18:1 n−6 c12-18:1 0.14 18.325 18.537
MOU-258 18:1 n−5 c13-18:1 0.58 18.393 18.612
MOU-279 18:1 n−4 c14-18:1 0.13 18.475g 18.686
DIU-102 18:2 n−x x,x-18:2 0.20 18.513g 18.859
DIU-027 18:2 n−6 c9,c12-18:2 16.58 18.637 19.045i
DIU-158 18:2 n−4 c11,c14-18:2 0.51 18.842 19.263
POU-030 18:3 n−6 c6,c9,c12-18:3 2.31 18.934 19.479j
SAN-010 19:0 19:0 0.19 18.998 18.999i
NFA-288 Unknown Unknown (not FAME) 0.20 19.129h 19.711k
POU-050 18:3 n−4 c8,c11,14-18:3 0.28 19.131hi 19.732k
MOU-144 19:1 n−10 c9-19:1 0.11 19.158i 19.357
MOU-580 19:1 n−8 c11-19:1 0.12 19.218 19.429j
POU-032 18:3 n−3 c9,c12,c15-18:3 4.79 19.272 19.840
POU-059 18:4 n−4 c5,c8,c11,c14-18:4 0.17 19.352 20.036l
FUR-185 DiMeF(9,3) 11,12-Dimethyl-10,13-epoxyhexadeca-10,12-dienoic acid 0.17 19.555j 19.906
POU-053 18:4 n−3 c6,c9,c12,c15-18:4 6.84 19.573j 20.286m
POU-056 18:4 n−1 c8,c11,c14,c17-18:4 0.41 19.689 20.408n
POU-068 18:5 n−1 c5,c8,c11,c14,c17-18:5 0.17 19.915 20.729
SAN-011 20:0 20:0 0.74 20.001 20.004l
MOU-097 20:1 n−11 c9-20:1 5.42 20.143k 20.331m
MOU-024 20:1 n−9 c11-20:1 14.50 20.183k 20.387n
MOU-087 20:1 n−7 c13-20:1 0.62 20.271 20.485
DIU-094 20:2 NMI x,x-20:2 0.11 20.333 20.642o
MOU-180 20:1 n−5 c15-20:1 0.17 20.404 20.626o
DIU-589 20:2 n−x x,x-20:2 0.33 20.468 20.851p
FUR-186 MeF(9,5) 11-Methyl-10,13-epoxyoctadeca-10,12-dienoic acid 0.31 20.538 20.890p
DIU-028 20:2 n−6 c11,c14-20:2 1.14 20.648 21.057
POU-314 19:5 n−3 c4,c7,c10,c13,c16-19:3 0.21 20.826 21.525r
POU-033 20:3 n−6 c8,c11,c14-20:3 0.27 20.908 21.488qr
SAN-012 21:0 21:0 0.58 20.998 21.000
POU-035 20:4 n−6 c5,c8,c11,c14-20:4 3.34 21.124l 21.782
FUR-188 DiMeF(9,5) 11,12-Dimethyl-10,13-epoxyoctadeca-10,12-dienoic acid 0.12 21.154l 21.476q
POU-034 20:3 n−3 c11,c14,c17-20:3 0.53 21.292 21.858
POU-054 20:4 n−3 c8,c11,c14,c17-20:4 2.63 21.553m 22.307st
FUR-189 DiMeF(11,3) 13,14-Dimethyl-12,15-epoxyoctadeca-12,14-dienoic acid 0.69 21.584m 21.934
POU-117 E1a Mono-trans 5,8,11,14,17-20:5 0.28 21.706 22.289s
POU-036 20:5 n−3 c5,c8,c11,c14,c17-20:5 64.96 21.776 22.616
POU-118 E1b Mono-trans 5,8,11,14,17-20:5 0.30 21.960n 22.527u
 Z. Wasta, S.A. Mjøs / J. Chromatogr. A 1299 (2013) 94–102 99

Table 2 (Continued)

Code Short name Long name Max Apct ECL BP-20* ECL BPX-70*
no
SAN-013 22:0 22:0 0.60 22.001 22.005
POU-558 E1c Mono-trans 5,8,11,14,17-20:5 0.30 22.049o POU-119v
POU-119 E1d Mono-trans 5,8,11,14,17-20:5 1.38 22.128p 22.716v
MOU-262 22:1 n−11 c11-22:1 22.72 22.134p 22.326t
MOU-025 22:1 n−9 c13-22:1 1.38 22.194 22.400
MOU-271 22:1 n−7 c15-22:1 0.47 22.287 22.506u
FUR-190 MeF(11,5) 13-Methyl-12,15-epoxyeicosa-12,14-dienoic acid 0.14 22.551 22.898
POU-069 21:5 n−3 c6,c9,c12,c15,c18-21:5 2.00 22.854 23.752
SAN-014 23:0 23:0 0.11 23.005 23.011
POU-037 22:4 n−6 c7,c10,c13,c16-22:4 0.65 23.170q 23.901
FUR-191 DiMeF(11,5) 13,14-Dimethyl-12,15-epoxyeicosa-12,14-dienoic acid 0.79 23.174q 23.495
POU-066 22:5 n−6 c4,c7,c10,c13,c16-22:5 6.20 23.450 24.142
POU-055 22:4 n−3 c10,c13,c16,c19-22:4 0.31 23.575 24.322w
POU-038 22:5 n−3 c7,c10,c13,c16,c19-22:5 6.72 23.832 24.740
SAN-015 24:0 24:0 0.20 24.002r 24.007
POU-138 D1a Mono-trans 4,7,10,13,16,19-22:6 0.26 24.036r 24.655
MOU-570 24:1 n−11 c13-24:1 0.14 POU-039s 24.324w
POU-039 22:6 n−3 c4,c7,c10,c13,c16,c19-22:6 85.80 24.120s 24.997x
MOU-026 24:1 n−9 c15-24:1 3.71 24.211 24.407
MOU-453 24:1 n−7 c17-24:1 0.10 POU-139t 24.525
POU-139 D1b Mono-trans 4,7,10,13,16,19-22:6 1.01 24.297t 24.895
POU-142 D1c Mono-trans 4,7,10,13,16,19-22:6 0.03 POU-140u 25.070y
POU-140 D1d Mono-trans 4,7,10,13,16,19-22:6 1.03 24.478u 25.101y
POU-141 D1e Mono-trans 4,7,10,13,16,19-22:6 1.28 24.556 25.202
DIU-526 24:2 n−6 c15,c18-24:2 0.25 24.692 POU-039x
POU-067 24:5 n−3 c9,c12,c15,c18,c21-24:5 0.27 25.849 26.776
POU-318 24:6 n−3 c6,c9,c12,c15,c18,c21-24:6 0.34 26.147 27.239
MOU-455 26:1 n−9 c17-26:1 0.29 26.224 26.404
OTH-178 Cholestadiene Cholesta-x,x-diene 0.23 26.497 27.326
a–x
Letters in the ECL columns denote peaks separated by less than 0.05 ECL units. Codes in the columns denote that the peak is hidden under the peak with corresponding
code.

isoprenoid fatty acids 4,8,12-trimethyltridecanoic acid, 4.2. Column differences

2,6,10,14-tetramethylpentadecanoic (pristanic) acid and 3,7,11,15-
tetramethylhexadecanoic (phytanic) acid were also found. All
these are frequently found in marine lipids [45]. Two compounds
matching the characteristic spectra of 7-methyl-6-hexadecenoic
acid were also detected. These have previously been reported in
bottom-living marine organisms [46–48]. The highest levels were
found in cod liver oil and cod is a typical bottom feeder. The Z
and E isomers are usually found together [47,48], which fits well
with the observation of two peaks with similar spectra. The two
compounds are reported without geometry in Table 2 since the
elution order of the E and Z isomer on polar columns is currently
unknown.
Six furan fatty acids were found above the threshold. All these
have been reported in fish already in 1974 [49]. Furan fatty acids
typically occur in small amounts in marine lipids and are easily
identified by their characteristic mass spectra [49,50]. The largest
levels of the two most abundant furan fatty acids (0.79% of FUR-191
and 0.69% of FUR-189) were found in omega-3 concentrates with
more than 50% EPA and DHA. Furan fatty acids are bulky molecules
and the most common have relatively high mass (typically 308–378
as FAME). Application of techniques such as urea inclusion and dis-
tillation to concentrate EPA and DHA may therefore also lead to
higher levels of furan fatty acids than in unprocessed fish oils.
Although fatty acids in omega-3 concentrates are basically of
marine origin it should be emphasized that some of the fatty acids
may come from other sources. The saturated fatty acids 8:0 and
10:0, which are usually of low abundance in marine oils, were for
instance found in significant amounts (up to 2.6% 8:0 and 1.9% 10:0),
but only in the capsuled products. Many of the products with high
levels of 8:0 and 10:0 were distilled “33/23 concentrates” where
short fatty acids should have been removed during distillation, so
it is obvious that these fatty acids must have been added later
in the manufacturing process. Some of the products in the study
were also labelled as blends of marine and terrestrial vegetable
oils.
It is common to assume that FCL values (Eq. (2)) are similar for
members of a homologous series (e.g. a:3 n−3, where a is a varying
number of carbons) and it has been a common basis for identifi-
cation of FAME from gas chromatographic retention indices [51].
While the assumption is fairly accurate for PEG phases, there are
larger deviations on CP phases. The FCL values of different homol-
ogous series on the two columns are plotted in Fig. 2. In addition
to the compounds listed in Table 2, Fig. 2 also contains FCL values
of relevant compounds from the standard mixtures, or compounds
that were present in the samples below 0.1%.
For the BPX-70 it can be seen that there are large variations
within some of the groups, and that the first members have lower
FCL values than other compounds in the same group. This is most
visible for 22:6 n−3, 19:5 n−3 and 16:4 n−3, which all have the
first double bond in 4 position relative to the carboxyl group, and
it can also be observed for 20:5 n−3 and 16:1 n−11, which have
the first double bond in 5 position. This effect has been observed
previously, and it has been shown that it is necessary to include
the double bond positions relative to the carbonyl group to achieve
models that can accurately predict ECL values of PUFA on the BPX-
70 phase [33]. It is emphasized that also 22:5 n−6, which is the only
member of the a:5 n−6, has a 4 double bond. Any higher member
of the same series should therefore be expected to have a higher
FCL value.
Also among the FAMEs that have no 4 or 5 double bond
there is a tendency that the FCL values increase with increasing
chain length. This can be explained by the temperature depend-
ent polarity of cyanopropyl phases, where the apparent polarity
increases with the temperature [13,33,36,52]. The higher mem-
bers of a homologous series will therefore experience a more
polar column since they elute at higher temperature. As a conse-
quence of the temperature dependence, the retention pattern on
BPX-70 is to a large degree tunable. Higher temperatures, higher
temperature gradients and lower carrier gas velocities increase
100 Z. Wasta, S.A. Mjøs / J. Chromatogr. A 1299 (2013) 94–102
Fig. 2. Fractional chain lengths of homologous series of FAME on BP-20 and BPX-70. Numeric values next to the boxes refer to mean values for each group.
ECL values, and the effects increase with the number of double
bonds in the molecules. This is an advantage when methods are
optimized to avoid chromatographic overlaps, but ECL values can
usually not be easily transferred with high accuracy between sys-
tems. Before using the retention indices reported in Table 2 for
identification purposes it is therefore advised to check how well
the reported indices match with reference compounds and adjust
the conditions if necessary. The polarity of PEG phases has been
shown to be virtually independent of the temperature [52] and the
retention indices are therefore less dependent on chromatographic
conditions. When ECL values for PUFA reported in Table 2 are com-
pared with results reported for a CP-Wax 52 column (L = 25 m,
i.d. = 0.25 mm, df = 0.25 ␮m, Agilent) [www.chrombox.org/data] the
mean absolute difference for PUFA is 0.012 and maximal difference
(22:6 n−3) is 0.075. The maximal deviation is slightly more than
a peak with at baseline (4 ), which varies between 0.04 and 0.05
ECL units. For less unsaturated compounds the deviations between
these two data sets are much lower.
The differences between the two columns in average FCL val-
ues for the separate classes pictured in Fig. 2 reflect the difference
in polarity between the two phases. For classes with FCL values
above 2 there is a risk of chromatographic overlap with abundant
saturated and monounsaturated fatty acids with two more carbon
atoms. For fatty acids with odd number of carbons in the chain this
problem arises when FCL is above 1. For the BP-20 phase FCL values
above 2 are only observed for the a:6 n−3 group, but FCL is close to
2 also for a:5 n−1. For BPX-70 FCL values above 2 are observed for
a:4 n−4 and the more polar classes. FCL values above 2 may also be
observed for a:3 n−3 and a:4 n−6 when the BPX-70 is used under
conditions that give a higher apparent polarity.
In addition to increased risk of chromatographic overlap, FCL
values above 2 makes identification more challenging because the
number of possible structures for a peak increase. Overlap patterns
on the two columns are discussed in more detail below.
4.3. Overlap patterns
The separation number, SN, is usually applied as a measure of
column efficiency in temperature programmed GC and is calculated
according to Eq. (6) [38],
tR(z+1) − tR(z) 1
SN = −1≈ −1 (6)
wh(z+1) − wh(z) 2wh,ECL
where tR denotes the retention times of two members of a
homologous series with z and z + 1 carbon atoms and wh is the
corresponding peak widths at half height. Since the difference in
retention between the two homologues is given by definition (1
for ECL, 100 for Kováts indices) on a retention index scale there is
 Z. Wasta, S.A. Mjøs / J. Chromatogr. A 1299 (2013) 94–102
a direct relationship between SN and peak widths given in reten-
tion index units. SN was therefore estimated from the peak widths
of the GLC-793 reference mixture as explained in [53]. In the ECL
region from 16 to 24, where the majority of the peaks elute, the
estimated SN was 18.3 and 18.1 for BP-20 and BPX-70, respectively,
with corresponding ranges 15.7–20.0 and 16.8–18.9.
The resolution (Rs ) between two peaks, a and b, as a function
difference in retention and peak width at half height is given by Eq.
(7).
1.177 · (tR(a) − tR(b) ) 1.177 · ECL
Rs = ≈ (7)
wh(a) + wh(b) 2 · w̄h,ECL
This equation is valid also if retention and peak width are
measured in retention index units. Eqs. (6) and (7) can there-
fore be combined into Eq. (8), which explains the relationship
between chromatographic resolution and the difference in ECL val-
ues between two compounds.
Rs ≈ 1.177 · ECL · (SN + 1)
A chromatographic resolution slightly above 1 is usually required
for accurate quantification of two closely eluting peaks. It can be
estimated from Eq. (8) that ECL should be approximately 0.05
or better for sufficient resolution when SN is approximately 18.
However, it is emphasized that the required resolution is highly
dependent on the chromatographic efficiency, and it is common to
analyze FAME on shorter and less efficient columns than used in
this work.
Peaks that were separated by less than 0.05 ECL units from their
nearest neighbour are regarded as overlapping and denoted by
matching letters behind the ECL values in Table 2. In cases where
accurate ECL values could not be calculated because peaks were
completely hidden under larger peaks, the code of the overlapping
peak is given instead of the ECL value. The total number of over-
lapping peaks were 39 on BP-20 and 47 on BPX-70, which means
that more than a third of the peaks had interferents. This illustrates
how challenging it will be to do accurate quantification of minor
peaks in a broad range of omega-3 products by GC FID. ECL values
on BPX-70 and other cyanopropyl columns in the upper range of the
polarity scale are highly dependent on the chromatographic condi-
tions, and ECL values of the most unsaturated FAMEs can typically
vary with more than 0.3 ECL units on the same column [54]. The
number of potentially overlapping peaks on these columns is there-
fore much higher than the 47 overlaps that were detected in this
case. However, the temperature dependent polarity also makes it
much easier to resolve overlaps of critical compounds if conditions
are optimized properly.
Most overlaps involved minor peaks (e.g. less than 1%), and
none of the compounds that are usually considered as nutritionally
important had overlaps with other major peaks. On BP-20 there
is poor separation between n−9 and n−11 isomers, but monoenes
are often reported as a sum value for each chain length. Several n−6
and n−3 PUFA (18:4 n−3, 20:4 n−6 and 20:4 n−3) were overlap-
ping with furan fatty acids on BP-20. The furan fatty acids are minor
compounds and these overlaps are not expected to give significant
contribution to the estimation of total n−3 and n−6 content of the
products, but a PEG phase may bee a poor choice for quantifying
furan fatty acids in marine lipids.
On BPX-70 the most problematic overlap may be between 18:4
n−3 and 20:1 n−11 because both these peaks may constitute more
than 5% of the total area. A similar overlap occurs between 20:4
n−3 and 22:1 n−11. However, these overlaps can be resolved by
tuning the chromatographic conditions to give slightly lower polar-
ity. Except that 22:6 n−3 overlaps with 24:2 n−6, which is a minor
compound found in only one sample, the other major n−3 PUFA
are free of interferents.
101
Fig. 3. Highest levels for compounds (maximal area percent in any sample) plotted
against the number of compounds that were detected above the level.
(8)
Particular attention has recently been paid to the trans iso-
mers of 20:5 n−3 (EPA) and 22:6 n−3 (DHA) [11,55,56]. These
are minor compounds in omega-3 products that form during high-
temperature refining of the oils [57,58]. Several of the mono-trans
isomers have potential overlaps, and careful optimization of chro-
matographic conditions is therefore recommended when these
compounds are quantified.
4.4. Discussion of threshold level
Although samples used in this study were selected by statistical
methods to be representative for commercial products it is empha-
sized that the sample set is still a subset of the products that are on
the market. Production processes and raw materials are also chang-
ing over time and new products may contain fatty acids above the
0.1% threshold in addition to those listed in Table 2.
Since GC–MS was used in this study, response factors are also
different than those achieved with GC–FID, and they vary depend-
ing on instrument and m/z scan range. The detector response
relative to 18:0 for the most unsaturated FAME in the reference
mixture, 22:6 n−3, varied from 0.70 to 0.78. This will affect whether
a compound with mass percent near 0.1% will be regarded as above
the threshold or not. On the one hand, a low response reduces
the chance that highly unsaturated compounds are detected above
the threshold, relative to more saturated compounds. On the other
hand it reduces the total chromatographic area in products domi-
nated by 20:5 n−3 and 22:6 n−3, since both these have relatively
low response. This gives an increased chance for all minor com-
pounds of being detected above the threshold.
With modern GC–FID and GC–MS instruments it is possible to
detect and quantify pure peaks far below the 0.1% threshold used in
this study. However, a traditional detection or quantification limit
may not be meaningful when complex samples are analyzed. For
a chromatographic method it is also important to consider how
many peaks that one can expect to resolve. In Fig. 3 the maximal
percents from Table 2 are plotted against the number of peaks that
were found above the same level. It illustrates clearly how much the
number of peaks increase with a decreasing threshold level. Going
lower than 0.1% may give an enormous increase in the number of
compounds, and at 0.1% approximately 1/3 of the peaks already had
less than 0.05 ECL units to their nearest neighbour.
If the level is set to 1% of total chromatographic area there are
only 35 peaks above the threshold and three overlaps on each col-
umn. But the 68 compounds that were found in the range 0.1–1%
are still present in the samples and they may give significant
102 Z. Wasta, S.A. Mjøs / J. Chromatogr. A 1299 (2013) 94–102
interference in quantitative studies, particularly if they interfere
with compounds close to the threshold.
In spite of the many interferences it is emphasized that both
columns could resolve the major fatty acids that are considered to
be of nutritional importance without overlaps from other abun-
dant fatty acids. It is therefore concluded that both columns can be
suitable for evaluation of nutritional quality of the products. But
if minor compounds are of interest, techniques that give higher
selectivity than GC–FID, for instance GC–MS or comprehensive
two-dimensional gas chromatography (GC × GC) should be consid-
ered for quantification.
4.5. Supplementary information
Retention index maps (ECL values) of the FAMEs listed in Table 2,
which can be of help for identification of compounds in omega-3
products and other marine lipids, are given as supplementary mate-
rial. The databases of mass spectra and retention indices can be
downloaded from www.chrombox.org/data. The database also con-
tains spectra and ECL values from compounds that were detected
below the 0.1% threshold. Retention index maps of all compounds
in the databases can be downloaded from the same source.
5. Conclusions
More than 100 fatty acids were found with a chromatographic
area above 0.1% of the total in at least one product. The identi-
fied compounds included FAMEs of normal and branched saturated
fatty acids, normal and branched monounsaturated fatty acids,
methylene-interrupted and non-methylene interrupted diunsat-
urated fatty acids, methylene-interrupted poly unsaturated fatty
acids belonging to the n−1, n−3, n−4 and n−6 series, and furan
fatty acids.
Peaks were regarded as not fully resolved if they were sepa-
rated by less than 0.05 ECL units from their nearest neighbour and
39 and 47 peaks were found to have chromatographic overlaps
on BP-20 and BPX-70, respectively. However, both columns were
found suitable for analysis of omega-3 products because major and
nutritionally important fatty acids are resolved from other major
peaks, but the large number of minor compounds that may act as
interferents will be problematic if low limits of quantification are
required.
Appendix A. Supplementary data
Supplementary data associated with this article can be found,
in the online version, at http://dx.doi.org/10.1016/j.chroma.2013.
05.056.
References
[1] Anon., EFSA J. 9 (2011) 2078.
[2] W.E.M. Lands, FASEB J. 6 (1992) 2530.
[3] R. Wall, R.P. Ross, G.F. Fitzgerald, C. Stanton, Nutr. Rev. 68 (2010) 280.
[4] C. von Schacky, Curr. Atheroscler. Rep. 5 (2003) 139.
[5] D. Mozaffarian, E.B. Rimm, J. Am. Med. Assoc. 296 (2006) 1885.
[6] G. Fontani, F. Corradeschi, A. Felici, F. Alfatti, S. Migliorini, L. Lodi, Eur. J. Clin.
Invest. 35 (2005) 691.
[7] M.C. Morris, D.A. Evans, J.L. Bienias, C.C. Tangney, D.A. Bennett, R.S. Wilson, N.
Aggarwal, J. Schneider, Arch. Neur. 60 (2003) 940.
[8] A.C. Patterson, K.D. Stark, J. Am. Coll. Nutr. 27 (2008) 538.
[9] A.P. Simopoulos, A. Leaf, N. Salem, Prostag. Leukotr. Ess. 63 (2000) 119.
[10] P.M. Kris-Etherton, D.S. Taylor, S. Yu-Poth, P. Huth, K.M. Moriarty, V. Fishell, R.L.
Hargrove, G. Zhao, T.D. Etherton, Am. J. Clin. Nutr. 71 (2000) 179S.
[11] C. Sciotto, S.A. Mjøs, Lipids 47 (2012) 659.
[12] G. Spiteller, Lipids 40 (2005) 755.
[13] S.A. Mjøs, J. Chromatogr. A 1100 (2005) 185.
[14] C.H. Rahn, D.M. Sand, T.P. Krick, R.L. Glass, H. Schlenk, Lipids 16 (1981) 360.
[15] G. Castello, J. Chromatogr. A 842 (1999) 51.
[16] F.R. Gonzalez, A.M. Nardillo, J. Chromatogr. A 757 (1997) 109.
[17] F.P. Woodford, C.M. van Gent, J. Lipid. Res. 1 (1960) 188.
[18] H. van den Dool, P.D. Kratz, J. Chromatogr. 11 (1963) 463.
[19] D.J. Harvey, in: W.W. Christie (Ed.), Advances in Lipid Methodology—One, The
Oily Press, Ayr (Scotland), 1992, p. 19.
[20] G. Dobson, W.W. Christrie, Eur. J. Lipid Sci. Technol. 104 (2002) 36.
[21] R.T. Holman, J.J. Rahm, Progr. Chem. Fats Lipids 9 (1966) 13.
[22] W.W. Christie, Gas Chromatography and Lipids, The Oily Press, Ayr (Scotland),
1989.
[23] S.A. Mjøs, Eur. J. Lipid. Sci. Technol. 106 (2004) 550.
[24] C. Härtig, J. Chromatogr. A. 1177 (2008) 159.
[25] The AOCS Lipid Library: http://lipidlibrary.aocs.org/ms/masspec.html; 2013
(accessed 11.02.13).
[26] L. Zhang, S.A. Mjøs, S. Meier, O.M. Kvalheim, Y. Liang, J. Chromatogr. A 1217
(2010) 5986.
[27] F. Brakstad, Chem. Int. Lab. Syst. 19 (1993) 87.
[28] S.A. Mjøs, J. Pettersen, Eur. J. Lipid Sci. Technol. 105 (2003) 156.
[29] B.-V. Grande, R. Manne, Chemom. Int. Lab. Syst. 50 (2000) 19.
[30] E. Karjalainen, Chemom. Int. Lab. Syst. 7 (1989) 31.
[31] A. de Juan, R. Tauler, Anal. Chim. Acta 500 (2003) 195.
[32] S.A. Mjøs, J. Chromatogr. A 1061 (2004) 201.
[33] S.A. Mjøs, O. Grahl-Nielsen, J. Chromatogr A. 1110 (2006) 171.
[34] S.A. Mjøs, Eur. J. Lipid Sci. Technol. 110 (2008) 547.
[35] L. Xiao, S.A. Mjøs, B.O. Haugsgjerd, J. Food Compos. Anal. 25 (2012) 198.
[36] S.A. Mjøs, B.O. Haugsgjerd, J. Agric. Food Chem. 59 (2011) 3520.
[37] K.J. Redmond, T. Magnesen, P.K. Hansen, Ø. Strand, S. Meier, Aquaculture 298
(2010) 202.
[38] L.S. Ettre, Pure Appl. Chem. 65 (1993) 819.
[39] J.P. Bergé, G. Barnathan, Adv. Biochem. Engin./Biotechnol. 96 (2005) 49.
[40] C.L. Scott, S. Kwasniewski, S. Falk-Petersen, J.R. Sargent, Mar. Ecol. Progr. Ser.
235 (2002) 127.
[41] N.V. Zhukova, N.A. Aizdaicher, Phytochemistry 39 (1995) 351.
[42] V.M. Dembitsky, H. Rezankova, T. Rezanka, L.O. Hanus, Biochem. Syst. Ecol. 31
(2003) 1125.
[43] L.N. Larsen, K. Høvik, J. Bremer, K.H. Holm, F. Myhren, B. Børretzen, Lipids 32
(1997) 707.
[44] P. Mayzaud, R.G. Ackman, Lipids 13 (1978) 24.
[45] W.M.N. Ratnayake, B. Olsson, R.G. Ackman, Lipids 24 (1989) 630.
[46] N.M. Carballeira, C. Cruz, S. A. J. Nat. Prod. 59 (1996) 1076.
[47] N.M. Carballeira, A. Sostre, A.D. Rodriguez, Comp. Biochem. Physiol. B 118
(1997) 257.
[48] N.M. Carballeira, C. Miranda, A.D. Rodriguez, Comp. Biochem. Physiol. B 131
(2002) 83.
[49] R.L. Glass, T.P. Krick, A.E. Eckhardt, Lipids 9 (1974) 1004.
[50] W. Vetter, S. Laure, C. Wendlinger, A. Mattes, A.W.T. Smith, D.W. Knight, J. Am.
Oil Chem. Soc. 89 (2012) 1501.
[51] R.G. Ackman, Anal. Chim. Acta 465 (2002) 175.
[52] G. Castello, S. Vezzani, G. D‘Amato, J. Chromatogr. A 779 (1997) 275.
[53] L.K. Skartland, S.A. Mjøs, B. Grung, J. Chromatogr. A 1218 (2011) 6823.
[54] S.A. Mjøs, J. Chromatogr. A 1015 (2003) 151.
[55] C. Ferreri, S.A. Grabovskiy, M. Aoun, M. Melchiorre, N. Kabalnova, C. Feillet-
Coudray, G. Fouret, C. Coudray, C. Chatgilialoglu, Chem. Res. Toxicol. 25 (2012)
687.
[56] V. Fournier, F. Destaillats, B. Hug, P.A. Golay, F. Joffre, P. Juanéda, E. Sémon, F.
Dionisi, P. Lambelet, J.L. Sébédio, O. Berdeaux, J. Chromatogr. A 1154 (2007)
353.
[57] S.A. Mjøs, M. Solvang, Eur. J. Lipid Sci. Technol. 108 (2006) 589.
[58] V. Fournier, P. Juanéda, F. Destaillats, F. Dionisi, P. Lambelet, J.-L. Sébédio, O.
Berdeux, J. Chromatogr. A 1129 (2006) 21.