Conference of Amine Kassouf in "Innovations in Food packaging, shelflife and food safety, 15-17 september 2015, Munich, Germany

1. Safety evaluation of plastic food contact materials using analytical
fingerprinting methodologies coupled with chemometric tools
Amine Kassouf ¹ ² ³
Douglas N. Rutledge², Jacqueline Maalouly³, Hanna Chebib³, Violette Ducruet¹
1-Research group I2MC, UMR 1145 Ingénierie Procédés Aliments, Massy, France
2-Research group IAQA, UMR 1145 Ingénierie Procédés Aliments, Paris, France
3-Research group “Lebanese Food Packaging”, Faculty of Sciences II, Lebanese University, Fanar, Lebanon
Lebanese University

2. Joint Research Unit 1145
FOOD PROCESSING AND ENGINEERING
Partnership between AgroParisTech and INRA
Staff: 90 including 57 Scientists, ~45 PhDs and Post-docs
6 research groups
 Modeling (transversal)
 Food structuring
 Reactions in food
 Interactions between materials and media in contact
 Consumer – food – process
 Analytical chemistry
Tools: industrial plant (food, biofuels and materials…), instrumented laboratories
(analytical chemistry, physico-chemistry, sensory analysis, ESR, microscopy,
chemometrics…), cluster of computers… 2

3. Group: Interactions between Materials and Media in Contact
(I2MC)
Research issue : Transport phenomena and reactivity in dense polymeric materials
 Design of materials (petro or bio based) with optimized properties (efficiency, quality, safety, sustainability) for
given applications in a reverse engineering approach.
 Decision tools for the design and control of separation processes involving dense membranes.
Contact medium
Contactmaterial
3

4. 4
Outline
• Safety evaluation of p-FCMs according to EU regulations and
alternative approaches
• Proposed methodology
o What are analytical fingerprints
o Data handling: exportation and preprocessing
o Chemometric tools: ICA and CCSWA
• Application example: Identification of NIAS in PET
by HS-SPME/GC-MS coupled with PCA, ICA and CCSWA
• Conclusion
, 16th September 2015Munich

5. 5
EU regulation and alternative approaches
Migration testing in food or food simulants is complex, costly and time-consuming:
• Consume large amount of solvents
• Low concentrations of migrants in food/food simulants
• Incompatibility between simulants and tested materials
• Problem of non intentionally added substances (NIAS)
• Laborious analytical methodologies (need to lower detection limits, forests of peaks etc.)
Alternative approaches based on mathematical
calculations (modelling, worst case calculation…)
, 16th September 2015Munich
Introduction
Need of potential migrants identities/quantities:
information transfer, certificates, predicted IAS/NIAS…
Sole use of substances from the
« Union list of authorized
substances »
Compliance with overall
(OML) and specific migration
(SML) limits
Issuance of a declaration
of compliance

6. 6
Lack of information transfer across the production chain
Producer of polymers,
starting substances etc. Food industries
Importers, trade
Retailers
Consumers
Producer of FCMs
Need for composition analysis (initial
composition and evolution across the
production chain)
, 16th September 2015Munich
Introduction
Once again: laborious analytical methodologies,
complex sample preparations, forests of peaks, need
for data reduction/selection.

7. 7
Objectives
Development of new, direct and fast, analytical approaches, complementary or even
alternative to common methodologies applied for composition analysis of FCMs, in
quality control systems, for food simulants/foodstuffs analysis (interaction tests)…
Chemometric tools
For a pertinent data treatment
Analytical fingerprinting
• Chromatographic and/or spectroscopic techniques
• Fingerprints of FCMs/simulants/foodstuffs
, 16th September 2015Munich
• Direct analytical methodologies
With no or little sample preparation
• Mainly non-targeted approaches
Methodology

8. 8
Proposed methodology: analytical fingerprinting
Analytical data matrices X (n samples × m variables) used in
“chemometrics - assisted analytical chemistry” come
from two sources:
, 16th September 2015Munich
Raw signals coming directly
from the instrument
Fingerprinting
Continuous signal data
Data from derived information
such as measured intensity,
composition/concentration/%
results /peak areas etc.
Profiling
Discrete data set
• Keeps the complete information
• Difficult to analyze
• Need of powerful chemometric tools
Valuable information could be lost
Methodology

9. 9
Proposed methodology: data handling of 2D data
, 16th September 2015Munich
Transpose
Concatenation
Minutes
0 10 20 30 40 50 60 70 80
mAU
0
200
400
600
800
1000
1200
1400
1600
mAU
0
200
400
600
800
1000
1200
1400
1600
DAD-CH1 200 nm
PET sample dopé_ACN_17122012
TIC: GC-MS/LC-MS
Intensity=f(tr)
One vector « variables »
per sample
1 2 3 …. n: samples
1
2
3
4
5
.
.
.
.
m
variables
2D data matrix X
(n samples × m variables)
Mass spectra
Intensity=f(m/z)
MIR spectra
Absorbance=f(wavenumber)
1
2
3
4
5
.
.
.
.
n
samples
1 2 3 4 …..……. m variables
2D chemometric tools
(PCA, ICA)
Methodology

10. 10
Proposed methodology: data handling of 3D data
, 16th September 2015Munich
1
2
3
4
5
.
.
.
l
variables
1 2 3 4 .…….…. m variables
GC-MS/LC-MS
Intensity =f(tr)=f(m/z)
3D front face fluorescence spectroscopy
Intensity =f(λ excitation)=f(λ emission)
One data matrix « variables »
per sample
3D data cubic array
(n samples × l variables × m variables)
123….nsamples
1 2 3 4 …………… l variables
Cube unfolding: ICA
CCSWA
Sensory data and/or
physicochemical data from various
techniques and/or toxicological data
Methodology

11. 11
Proposed methodology: chemometric tools - ICA
, 16th September 2015Munich
• ICA: a method of blind source separation(BSS).
• ICA aims to extract nF "source” signals, considered independent, knowing only n "observed” signals (n ≥ nF), considered as a
weighed sums of theses « source » signals and in unknown proportions.
Rutledge et al. Trends Anal Chem. 2013
General model
X = A.S
X: matrix of observed signals (mixtures)
A: matrix of proportions
S: matrix of “source” signals (ICs)
ICA
(JADE algorithm)
ICA attempts to recover the “source” signals by
estimating a linear transformation, using a
criterion that measures statistical
independence among the sources, by
maximizing their non-gaussianity
Choice of optimal number of ICs
• No natural order for the extraction of the signals.
• Too few ICs: non-pure signal
• Too many ICs: may decompose the signal.
• ICA_by_blocks (J-R Bouveresse et al., Chemo lab 2012)
• New methods: Random_ICA, ICA_corr_y,
multi_ICA_corr, ICA_DA etc.
Methodology

12. 12
Proposed methodology: chemometric tools - CCSWA
, 16th September 2015Munich
123……..nsamples
Same number of
samples (n)
Different variables describing the
same samples
(different characterization :
analytical, sensory, toxicological etc.)
• Objective: simultaneously study several matrices with different variables describing the same
samples, searching for common spaces for all m data tables (Common Components: CCs).
• Saliences: contribution of each matrix (data table) to the definition of each dimension of this
common space.
ComDim: implementation of the CCSWA method - toolbox SAISIR (Bertrand & Cordella, 2011)
Qannari et al., J Food Qual Pref, 2000
J-R Bouveresse et al., Chemo lab, 2011
1 2 3 4 ………….. l variables
Methodology

13. 13
Identification of NIAS in PET
by HS-SPME/GC-MS coupled with PCA, ICA and CCSWA
 The production process of a PET bottle could induce the degradation of the polymer and its additives as well as
introduce impurities and contaminants. Therefore, emergence of new potential migrants.
Aim/Mean:
 Development of a qualitative tool to monitor the apparition of NIAS during the production process of PET bottles.
 Compare volatile fingerprints of PET pellets, preforms and bottles, combining 2D and 3D GC-MS data with
chemometric tools
NIAS in PET
15Munich, 16th September 2015
NIAS in PET

14. 14
NIAS in PET Materials and methods
16
NIAS in PET
Kassouf et al., Talanta 2013
Munich, 16th September 2015
Pellets Preforms Bottles
Generation of NIAS during the process
TIC* EIC*PCA
ICA
CCSWA (ComDim)
Compensation of potential information
loss due to the use of TICs
10277 variables
18samples
18samples
10277 variables
Follow-up by HS-SPME/GC-MS
(using optimized extraction parameters)
* TIC: total ion current chromatogram
* EIC: extracted ion chromatogram
• Two PET grades: R and J
• Pellets (G), preforms (P),
bottles (B)
• 3 repetitions
• Total: 18 samples
HS-SPME/GC-MS fingerprint
Chemometric calculations: Matlab version
R2007b (The MathWorks, Natick, USA)

15. 15
NIAS in PET Results and discussion
Example of TIC chromatograms obtained by HS–SPME/GC–MS
of ground PET pellets, preforms and bottles
Relevant information may be difficult to extract
17
NIAS in PET
Kassouf et al., Talanta 2013
Munich, 16th September 2015

16. 16
NIAS in PET
• Two PET grades: R and J
• 3 repetitions
• Matrix: 18 × 10277
• number of PCs = 6
NIAS responsible for the discrimination on IC1
Compound
1 2-methyl-1,3-dioxolane
2 Ethylene glycol
3 Toluene
4,5,6 Ethylbenzene, Xylene isomers
7,8 Nonanal, Decanal
9 Diethylphthalate (DEP)
10 Di-isobutylphthalate (DIBP)
Results and discussion (PCA)
18
NIAS in PET
Kassouf et al., Talanta 2013
Munich, 16th September 2015
Volatile compounds
responsible for the
discrimination of pellets
and preforms
Semi-volatile compounds
responsible for the
discrimination of bottles

17. 17
NIAS in PET
•Two PET grades: R and J
• 3 repetitions: 1,2,3
• Pellets (G), preforms (P), bottles (B)
• Matrix: 18 × 10277
Results and discussion (ICA)
Discrimination of
Bottles J due to DEP
19
NIAS in PET
Kassouf et al., Talanta 2013
Munich, 16th September 2015
Optimal number of ICs = 6 obtained by ICA-
by-blocks
• nF = maximum number of ICs= 9
• B = number of blocks = 2
• Optimal number of ICs = nFopt = 5

18. 18
NIAS in PET Results and discussion (ICA)
Discrimination of
Bottles J and R due
to linear aldehydes
Discrimination of
pellets J due to EG
Not clear in PCA
20
NIAS in PET
Kassouf et al., Talanta 2013
Munich, 16th September 2015

19. 19
NIAS in PET Results and discussion (CCSWA)
CCSWA scores on CC1, CC2, CC4 and CC6 related to the batches types (R and J), the nature of
the samples (G:pellets; P:preforms and B: bottles) and the repetitions (n=3).
21
NIAS in PET
Kassouf et al., Talanta 2013
Munich, 16th September 2015
CCSWA applied on the segmented
cubic array
18 × 10277 × 255 with 6 CCs
10277 variables
18samples

20. 20
NIAS in PET
m/z = 191, 192, 206, 221, 277 and 292 (Saliences >
0.6)
m/z= 48, 91, 92, 106, 117, 118, 119, 120, 134 and 272
(Saliences > 0.1)
2,4-bis(1,1-dimethylethyl) phenol
(Degradation product of antioxidants:
Irganox 1010 and Irgafos 168)
2-methyl-1,3-dioxolane( m/z=48); toluene,
ethylbenzene and xylene isomers (m/z=91, 92, 106)
and dichlorobenzene (m/z= 117, 118)
Results and discussion (CCSWA)
22
NIAS in PET
Kassouf et al., Talanta 2013
Munich, 16th September 2015

21. 21
NIAS in PET Conclusion
22Munich, 16th September 2015
Analytical fingerprints
GC-MS, LC-MS, 3D front-face fluorescence and ATR-MIR
Chemometric tools
 The quality of results depends on the chosen analytical methodology: repeatability, clean blanks,
data handling, sensitivity etc.
 Qualitative and quantitative determinations may be performed.
 Versatile methodologies: composition analysis, quality control systems (different grades of raw
materials, different industrial processes ...), interaction tests...
 Approaches can be extrapolated to other analytical techniques (NMR, direct mass spectrometry ...).
 High potential for multi-block chemometric tools: simultaneous analysis of sensory data,
toxicological data, composition data...

22. 22
NIAS in PET
22Munich, 16th September 2015
• A. Kassouf, A. Ruellan, D. Jouan-Rimbaud Bouveresse, D.N. Rutledge, S. Domenek, J. Maalouly, H. Chebib, V. Ducruet
(2015). Attenuated total reflectance-mid infrared spectroscopy (ATR-MIR) coupled with independent components analysis
(ICA): a fast method to determine plasticizers in polylactide (PLA). Accepted in Talanta.
• A. Kassouf, J. Maalouly, D.N. Rutledge, H. Chebib, V. Ducruet. (2014). Rapid discrimination of plastic packaging materials
using MIR spectroscopy coupled with independent components analysis (ICA). Waste management, 34, 2131-2138
• A. Kassouf, M. El Rakwe, H. Chebib, V. Ducruet, D. N. Rutledge, J. Maalouly. (2014). Independent components analysis
coupled with 3D-front-face fluorescence spectroscopy to study the interaction between plastic food packaging and olive oil.
Analytica Chimica Acta, 839, 14-25
• A. Kassouf, J. Maalouly, H. Chebib, D. N. Rutledge, V. Ducruet. (2013). Chemometric tools to highlight non-intentionally
added substances (NIAS) in polyethylene terephthalate (PET). Talanta, 115, 928-937
• J. Maalouly, N. Hayeck, A. Kassouf, D. N. Rutledge, V. Ducruet. (2013). Chemometric tools to highlight possible migration
from packaging to sunflower oils. Journal of Agricultural and Food Chemistry, 2013, 61 (44), 10565-10573
Thank you
aminekassouf@hotmail.com violette.ducruet@agroparistech.fr