Characterization of Omega Dietary Supplements by combining FT-IR and

Pattern Recognition Tools Michael J. Wenstrup and Luis Rodriguez-Saona

Department of Food Science and Technology, The Ohio State University, Columbus, OH 43210

Methodology

Introduction

Abstract

The objective of this research was to create a model

for rapid determination of fatty acid composition and

classification of fish oil supplements.

25 Samples of Fish,

Flaxseed and CLO oils

Derivatization and

methylation

Cary 630 portable benchtop FT

mid-infrared spectrometer

Agilent 6890 gas

chromatography system

CH

CH2

CH3

C-H

stretching

C=O

stretching

Fingerprint

region

IR Spectral Data

+ Internal Standard

+ External Standards

GC Fatty Acid Composition Data

CHEMOMETRICS

Quantification Model

Partial Least Square Regression

(PLSR)

Soft Independent Modeling of Class

Analogy (SIMCA)

Classification Model

+

IR

GC-FAME

Interest in omega-3 dietary supplements has increased due to their health and nutritional

benefits. Scientific studies on these products have uncovered anti-inflammatory effects,

decreased cardiovascular disease mortality, and the potential reduction in blood

pressure. The compounds primarily responsible for these effects, polyunsaturated fatty

acids (PUFA) such as eicosapentaenoic acid (EPA), 20:5, and docosahexaenoic acid

(DHA), 22:6, are almost exclusively found in seafood products and are essential fatty

acids for humans (de Leiris 2009). The concentration of EPA and DHA in fish depends

on the type of fish, with salmon, herring, mackerel, tuna, and sardines having the highest

concentrations. As each PUFA has a different role in health promotion and disease

prevention, it is important to understand the variability in composition within fish oil

products, and whether they are supplied as triglycerides, esters, or fatty acids. Alkyl

esters of omega-3 fatty acids are more prone to oxidation than their triglyceride

counterparts, and some evidence suggests that the bioavailability of alkyl esters is lower

than that of intact triglycerides. Modern, rapid techniques to determine composition,

authenticity, and quality of fish oils are not yet readily available, but development of

Fourier transform infrared (FT-IR) and attenuated total reflectance (ATR) spectrometry

equipment has opened the door to novel methods. FT-IR systems provide a more robust

alternative to older dispersive systems, and ATR-IR systems can accommodate a wide

Figure 1. ATR-IR systems offer

higher signal intensity and are

compatible with a wider array

of sample types than other

sampling accessories.

Objective

GC

Sa

mp

le

GC Determination Label

EPA (%) DHA (%) EPA

(%) DHA

(%)

Fis

h A

K1 41.8 3.5 25.2 3.3 30 20

K2 58.6 3.1 35.2 2.8

MF1 43.0 0.7 34.7 1.7 40 30

MF2 48.7 2.1 37.9 3.0

S 47.6 0.7 34.7 0.0 36 21.6

N 43.5 1.1 28.3 0.4 27 18

CE 42.1 4.8 30.9 3.7 32 24

CS 41.1 3.4 26.8 2.7 30 20

NF 23.9 1.2 13.5 1.5 18 12

Fis

h B

W1 24.3 3.6 7.4 1.9 30

W2 26.2 1.4 7.5 0.5

ON1 22.0 1.8 10.4 1.5 30

ON2 21.2 0.4 11.3 0.7

NM1 22.9 1.6 14.5 4.1 15 10

NM2 21.7 0.0 10.3 0.2

CN 21.3 0.2 10.6 0.2 18 12

U 21.7 1.5 11.2 1.2 16.5 11

Raw F 6.7 0.1 3.8 0.2 6 5

CLO

T 14.5 1.0 10.1 1.0 12.3 8.2

VC 13.1 0.3 11.7 0.8 7.7 11.4

CLO 13.5 0.2 14.6 0.4 -

Factors SECV r Val SEC r Cal RPD

EPA 4 2.73 0.98 2.65 0.98 5.1

DHA 5 1.76 0.99 1.66 0.99 8.0

Figure 4. PLS Regression models of EPA and DHA content. Reference EPA/DHA data from GC data in Table 1.

Table 2. PLSR fit and error parameters. SECV

is Standard Error of Cross-Validation, SEC is

Standard Error of Calibration, RPD is Relative

Percent Difference to reference data.

Table 1. GC-FAME fatty acid analysis with label fatty acid

estimates. Left side groups names are SIMCA classes.

Figure 2 (above). SIMCA classification plot from 900-

1300 cm-1 region, showing 5 classes and EPA content

trend (red arrows).

Figure 3. Spectral comparison of the 1700-

1800 cm-1 regions of Fish groups A & B,

showing the shift in the ester peak.

SIMCA

PLSR

Discussion SIMCA classification reveals relationships between samples and can be used to

classify unknown samples into groups. Fish oils clustered into two groups, A and B,

Cod Liver Oil (CLO), Virgin Fish Oils (Raw) and Flaxseed oils clustered into

distinct groups with significant interclass distances (Figure 2). The amount of

EPA/DHA increased down the PC3 axis, and the Fish A group was separated from

By correlating the MIR spectral data with

fatty acid composition information, a PLSR

model was built.. Table 2 shows the

parameters for evaluating the model, as

well as comparison to the reference data.

A separate SIMCA model that included only Fish A &Fish B groups was constructed,

with a similar separation to that seen in Figure 2. Inspection of the spectral overlays

revealed a downward shift of the triglyceride ester band, from 1745 cm-1 to

1737 cm-1 (Figure 3). This shift is associated with fatty acid methyl and ethyl esters.

the other groups down PC1.

The discriminating power plot

revealed that the band at 1038 cm-1

was primarily responsible for

classification in this model.

References

Figure 5. Sample

application during

spectral collection.

The ATR is skirted

by a temperature-

controlled collar,

ensuring a

homogenous liquid

phase. Required

sample volume is

between 60 and 100

μL.

• Analysis of fatty acid composition by gas chromatography revealed

a wide range of EPA (6-60%) and DHA (3-40%) concentrations.

These values are comparable to the label claims, with most

samples having more EPA/DHA than claimed.

• With correlation coefficients of cross-validation of ≥ 0.98 and

SECVs of 2.73% and 1.76%, for EPA and DHA respectively, PLSR

fatty acid determination is comparable to modern chromatography

techniques (Table 2). Ratio of performance deviation (RPD) of the

models versus the reference data shows that this method is

capable of quantifying EPA & DHA across a broad

compositional range.

• SIMCA classification revealed clustering by oil source, as well as

3D spacing with regards to EPA/DHA content. This region

responsible for this significant separation, 900-1300 cm-1, is

associated with C=C and C-O stretching. The band primarily

responsible for discrimination, 1038 cm-1, has been found to

correlate with the C-O stretching in esters.

• The Fish A group has a distinct carbonyl band shift (from 1745 cm-1

to 1737 cm-1) that is associated with alkyl esters. This may reflect

the delivery of these supplements as primarily fatty acid alkyl

esters instead of triglycerides.

Conclusions

We have shown a fast, simple, and reliable method for determining

the characterization of the fish oils in these supplements and for

quantification of the important omega-3 fatty acids present, to

support and validate labeling claims. The portable ATR-MIR method

not only provides precise fatty acid composition information, but can

be used to determine the state of these fatty acids; whether they be

methyl/ethyl esters, free fatty acids, or triglycerides.

Omega-3 dietary supplements have been linked with health benefits including

decreased cardiovascular disease mortality, anti-inflammatory effects, reduced

blood pressure, and antiatherogenic activity. The sources of these effects,

polyunsaturated fatty acids (PUFA) such as eicosapentaenoic acid (EPA)

docosahexaenoic acid (DHA), are almost exclusively found in seafood products and

are essential fatty acids for humans. The concentration of EPA and DHA in fish is

highly variable; depending on species, catch location, environment, season and

processing parameters in determining the ratio of fatty acids found in the oil. Our

objective was to characterize the composition of commercial Omega-3 dietary

supplements using mid-infrared spectroscopy, gas chromatography, and

chemometrics. Supplements were purchased from retailers that included a large

amount of source, processing, and compositional variability. Fatty acid composition

of oils was determined by fatty acid methyl ester gas chromatography. The

supplements encompassed a wide range of fatty acid profiles and delivery methods.

Mid-infrared spectral data was collected on portable infrared systems and analyzed

by pattern recognition (SIMCA and PLSR). SIMCA analysis using the 900-1300 cm-1

region allowed for tight clustering of samples into distinct classes, with significant

(>3) interclass distances. Discriminating power showed a strong influence of the

1038 cm-1 band, which is typically associated with C=C and C-O stretching

vibrations. PLSR generated multivariate models with high correlation (R ≥ 0.98)

between the infrared spectrum and fatty acid composition, and SECV of 2.73 and

1.76 for EPA and DHA, respectively. Our results indicate that ATR-IR

spectroscopy combined with pattern recognition analysis provides for robust

screening and determination of fatty acid composition of fish oil supplements.

variety of sample types, from gases, to liquids, to powders.

Although the food and supplement industries have utilized

near-infrared (NIR) spectroscopy for decades, the

applications of mid-infrared (MIR) spectroscopy have only

recently gained ground. ATR FT-MIR is a non-destructive and

rapid technique that allows for analysis of all the chemical

bonds in a system. When combined with multivariate

statistical analysis and modern computational power

(collectively known as chemometrics), the large amount of

information provided by MIR can be used to replace

traditional techniques (Bosque-Sendra et al., 2012). In oil

analysis, these traditional techniques are both time

consuming and expensive. Gas chromatography (GC) and

high-performance liquid chromatography (HPLC) are used in

the analysis of oils, but the extensive sample preparation,

long run-times, and destructive nature of the analysis has

piqued interest in spectroscopic alternatives.

Bosque-Sendra, JM; Cuadros-Rodriguez, L; Ruiz-Samblas, C; de la Mata, AP. Combining chromatography and

chemometrics for the characterization and authentication of fats and oils from triacylglycerol compositional

data-A review. ANALYTICA CHIMICA ACTA, v. 724, 2012, p. 1-11.

Cabo, N. Infrared spectroscopy in the study of edible oils and fats. JOURNAL OF THE SCIENCE OF FOOD

AND AGRICULTURE, v. 75 issue 1, 1997, p. 1-11.

De Leiris, J; de Lorgeril, M; Boucher, F. Fish Oil and Heart Health. JOURNAL OF CARDIOVASCULAR

PHARMACOLOGY, v. 54 issue 5, 2009, p. 378-384.