Shelf Life of Extra Virgin Olive Oil and Its Prediction Models

Abstract

Extra virgin olive oil (EVOO), with high unsaturation degree (oleic acid, linoleic acid, and linolenic acid), is prone to oxidation during production and storage even with the presence of abundant antioxidants (e.g., phenolic compounds, alpha-tocopherol, and chlorophyll). The level of oxidation degradation is greatly affected by the EVOO chemical composition (free fatty acids, saturated and unsaturated fat ratio, total phenol content, etc.) and storage conditions (packaging material, oxygen, temperature, and light). With the increasing demand on qualitative acceptability and food safety of an EVOO product, consumers rely heavily on “shelf life” as a good indicator. Hence, it is critical for olive oil producers to provide accurate and practical information on shelf-life prediction. This review analyzes ten shelf-life prediction models that used various parameters and approaches for model establishment. Due to the complexity of chemical interactions between oil phase and environment under real-time storage and rapid accelerated testing conditions, further investigation is needed to scrutinize and minimize the discrepancies between real-time shelf life and predicted shelf life of EVOO products.

Full text

Review Article
Shelf Life of Extra Virgin Olive Oil and Its Prediction Models
Xueqi Li 1and Selina C. Wang 1,2
1University of California Davis Olive Center, Davis, CA 95616, USA
2Department of Food Science and Technology, University of California, Davis, Davis, CA 95616, USA
Correspondence should be addressed to Selina C. Wang; scwang@ucdavis.edu
Received 26 October 2017; Accepted 2 January 2018; Published 31 January 2018
Academic Editor: Amani Taamalli
Copyright ©  Xueqi Li and Selina C. Wang. is is an open access article distributed under the Creative Commons Attribution
License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly
cited.
Extra virgin olive oil (EVOO), with high unsaturation degree (oleic acid, linoleic acid, and linolenic acid), is prone to oxidation
during production and storage even with the presence of abundant antioxidants (e.g., phenolic compounds, alpha-tocopherol, and
chlorophyll). e level of oxidation degradation is greatly aected by the EVOO chemical composition (free fatty acids, saturated
and unsaturated fat ratio, total phenol content, etc.) and storage conditions (packaging material, oxygen, temperature, and light).
With the increasing demand on qualitative acceptability and food safety of an EVOO product, consumers rely heavily on “shelf life”
as a good indicator. Hence, it is critical for olive oil producers to provide accurate and practical information on shelf-life prediction.
is review analyzes ten shelf-life prediction models that used various parameters and approaches for model establishment. Due to
the complexity of chemical interactions between oil phase and environment under real-time storage and rapid accelerated testing
conditions, further investigation is needed to scrutinize and minimize the discrepancies between real-time shelf life and predicted
shelf life of EVOO products.
1. Introduction
Known as a key component in the Mediterranean diet for
centuries, extra virgin olive oil (EVOO) has become globally
recognized and appreciated by consumers due to its unique
sensory characteristics and high nutritional values. In recent
years, there has been considerable interests in correlating
monounsaturated fatty acids (mainly oleic acid) and minor
components (phenolic compounds, alpha-tocopherol and
carotenoids, squalene, simple triterpenes, and volatile com-
pounds)inEVOOwithhealthbenets{e.g., antihypertensive
activity [], chemopreventive activity [], tumor-inhibitory
activity [], and anti-inammatory activity []}and positive
sensory attributes [–]. However, due to high levels of unsat-
urated fatty acids and the presence of endogenous enzymes
such as lipase, polyphenol oxidase, and peroxidase, EVOO
is also prone to lipid oxidation and enzymatic hydrolysis
which favors autoxidation [–]. EVOO oxidation is highly
dependent on factors including the storage of olive fruit
prior to processing [], the techniques of oil extraction
[], the exposure degree of oxygen, light, and temperature
[], and the packaging [] and storage conditions of the
nal product [], which could greatly change the chemical
composition of the oil, leading to unpleasant o-avors and
eventually degrading the quality.
For consumers, one of the most important characteristics
in EVOO is freshness, as freshness is typically associated with
high quality and ensures food safety []. e term “shelf life”
is commonly referred to when determining the freshness and
consumer acceptability of EVOO []. Specically, EVOO
shelf life could be dened as the length of time under normal
storage conditions within which no o-avors or defects are
developed and quality parameters are within accepted limits
for this commercial category []. Consumers rely on shelf-
life determinations to dierentiate between products that are
acceptable for consumption from those that are no longer
acceptable. us, it is obligatory for the olive oil industry
to monitor oil quality throughout the production line []
and to be able to provide realistic information on shelf-life
prediction considering the temperature changes and light
exposure during transport and commercial activities [].
EVOO shelf-life testing is oen conducted under real-
time conditions or accelerated conditions []. Real-time
shelf-life testing allows data collection under normal storage
Hindawi
Journal of Food Quality
Volume 2018, Article ID 1639260, 15 pages
https://doi.org/10.1155/2018/1639260

Journal of Food Quality
conditions and reects actual changes in EVOO matrix over
time []. On the other hand, this process requires consistent
storage conditions and can be extremely time consuming
when the quality depletion of EVOO proceeds fairly slowly
under normal storage conditions []. Hence, accelerated
shelf-life testing (ASLT) methods such as Rancimat, Active
Oxygen Method (AOM), and Oil Stability Index (OSI) are
also employed to determine EVOO shelf life under conditions
which are dierent from normal storage conditions within
a short period of time []. Noticeably, as convenient and
rapid as the accelerated methods can be, Kaya et al. []
reported that extrapolation from the Rancimat values led to
either underprediction or overprediction of the actual shelf
life of sunower and olive oil due to drastic ASLT conditions.
Nonetheless, analytical data generated from either or both
conditions can be applied to the development of EVOO shelf-
life prediction models.
In general, two types of shelf-life prediction models are
widely used to simulate EVOO degradation: kinetic models
and empirical models. Kinetic models are developed based
on how reaction rates in critical chemical parameters (Table )
are inuenced by experimental conditions related to variables
such as storage time, temperature, and light []. Data
describing the changes of these parameters under conditions
simulating actual storage are submitted to modeling based
on the known rate of a particular reaction. e limitation
ofkineticmodelingisthatclassicalkineticequationscannot
easily accommodate the complexity of oxidation reactions
and oil deteriorations. Empirical models are developed based
on the correlations between individual chemical parameters
and experimental condition variables. Typically, advanced
statistical analyses are performed on analytical data to
develop regression models which enable the prediction of
maximum shelf life as a function of chemical parameters
[, ]. e limitation of empirical modeling is the diculty
to extend beyond the measured setup (e.g., storage condition)
and simplication and approximation can fail when the setup
is changed.
Previous studies have been done intensively on how
dierent ratios of chemical composition, packaging systems,
and storage conditions would aect the quality of EVOO
[,–].However,theoliveoilindustryisstillingreatneed
of practical and eective shelf-life prediction models that can
beeasilyusedoradoptedaermoderatemodicationsin
order to reasonably predict EVOO shelf life and to ensure the
EVOO products complying with the current regulations for
its category [].
In this review, ten shelf-life prediction models using var-
ious parameters and approaches are discussed. In addition,
future directions of shelf-life prediction models are proposed
aiming at minimizing the discrepancies between real-time
shelf life and predicted shelf life of EVOO products.
2. Prediction Models for EVOO
Shelf-Life Determination
e development process of EVOO shelf-life prediction
models is streamlined in Figure . While ASLT provides a
more rapid and less-expensive method of predicting shelf life
than real-time storage condition monitoring, some acceler-
ated conditions may lead to erroneous shelf-life predictions
because of complicated chemical reaction mechanisms from
real-time conditions []. us, shelf-life prediction models
arebestdevelopedbasedonresultsfrombothreal-time
and accelerated storage conditions, followed by extensive
evaluation and adjustment.
Table  provides a summary of olive oil sample size,
chemical/sensory analysis, and statistical approach of the ten
shelf-life prediction models discussed in this review.
2.1. Pagliarini et al. (2000) [20]. is Tuscan EVOO shelf-life
prediction model used induction time, hydroxytyrosol, and
tyrosol to predict the time (in days) to reach an acceptable
limitof.forUV232.
e research team analyzed a total of  samples from ve
dierent lots which are categorized in Table . e samples
were subjected to dierent bottling, transport, and storage
conditions in supermarkets, although the authors found that
the stability of the oil was not signicantly aected. is
couldbeduetoreasonsthat()theoilwasstoredproperly
in the tanks at processing facility in Italy (OL.MA.) before
getting bottled; () the oil did not experience extreme travel
stress during transportation to either Italian supermarket
or Australia supermarket; () while the oil was stored in
supermarkets, the uncontrolled light and temperature were
still in favor of maintaining the quality of olive oil.
e research team tracked the changes in oil during stor-
agewithphysiochemicalparametersandsensoryanalysis
and via multivariate analysis procedure, it was concluded that
the most signicant parameters were 232, induction time,
chlorophyll, carotenoid, alpha-tocopherol, hydroxytyrosol,
and tyrosol. Since the only parameter that had established
limit in the standards was 232, three empirical models were
set up to predict the time to reach a given value for 232 and
. was chosen as a reference value:
(a)=1130.84Ln (induction time)−2388.13
(b)=329.02−38.11hydroxytyrosol
(c)=580.34−68.11tyrosol.
()
In these equations isthetime(indays)toreachanacceptable
limit of . for 232. According to the authors, this model
underestimates the experimental storage time by  days for
Rancimat induction time,  days for hydroxytyrosol content,
and  days for tyrosol content.
e above models could be useful for selecting new
olive/oil suppliers and comparing dierent suppliers, olive
harvest years, and storage conditions. While these three equa-
tions consist of simple calculations, the output of estimated
time is when 232 reaches . instead of . which is the
upper limit of 232 for EVOO category in the International
Olive Council (IOC) trade standard []. Hence, the results
may not be reective and reliable in their current form.
2.2. Guti´
errez and Fern´
andez (2002) [35]. e quality indices
(specied in the European Union standards EC Reg-
ulation /) of EVOO samples produced from two

Journal of Food Quality 
T : Critical chemical parameters used in olive oil shelf-life prediction model development.
Parameter Determination Indicator Methodology
Free fatty acids (FFA)
Free fatty acids are formed by the
hydrolysis of triglycerides during
ripening, processing and storage
An elevated level of free fatty acid
indicates hydrolyzed fruits
and/or poor quality oil made
from unsound fruit, improperly
processedorstoredoil
Analytical titration
Peroxide value (PV)
Peroxides are primary oxidation
products that are formed when
oils are exposed to oxygen,
producing undesirable avors
and aroma
An elevated level of peroxides
indicates oxidized and/or poor
quality oil
Analytical titration
Ultraviolet
absorbance (UV)
Conjugated double bonds are
formed from natural
nonconjugated unsaturation in
oils upon oxidation. e 232
measures primary oxidation
products and 270 measures
secondary oxidation products
An elevated level of UV
absorbance indicates oxidized
and/or poor quality oil
UV spectrophotometry
, -Diacylglycerols
(DAGs)
Fresh EVOO contains a high
proportion of ,-diacylglycerols
to ,- and ,-diacylglycerols,
while olive oil from poor quality
fruits and rened olive oils have
higher level of ,-DAGs than
fresh EVOO
A low ratio of ,-diacylglycerols
to ,- and ,-diacylglycerols is
an indicator for oil that is
hydrolyzed, oxidized, and/or of
poor quality
Gas chromatography (GC)
Pyropheophytins
(PPP)
Chlorophyll pigments break
down to pheophytins and then
pyropheophytins upon thermal
degradation and aging of olive oil
An elevated ratio of pheophytin a
to pyropheophytins is an
indicator for oil that is oxidized
and/or adulterated with rened
oil
High performance liquid
chromatography (HPLC)
Sensory evaluation Sensory refers to avor and
aroma attributes
Sensory evaluation can help
identify oils that are of poor
quality, oxidized, and/or
adulterated with other oils
A recognized panel of –
people evaluates oils for sensory
characteristics
Induction time
e oxidation process is
accelerated by means of heating
up the reaction vessel while
passing air continuously through
the sample
Oxidative stability (in hours)
denotes the resistance of oils to
oxidation. e longer the
induction time, the more stable
the oil
Rancimat
Total phenols
esumofuptoindividual
phenols which have antioxidative
ability
A low level of total phenols can
indicate a shorter shelf life while
ahighleveloftotalphenolscan
indicate a longer shelf life
UV spectrophotometry/High
performance liquid
chromatography (HPLC)
Volati l e s ( e . g.
hexanal/nonanal,
E--hexenal/hexanal),
Volatile compositions change
during oxidation. For example, as
the oil oxidizes, the
concentration of hexanal
decreases while concentration of
nonanal increases
e ratios of hexanal/nonanal
and E--hexenal/hexanal can
indicate oxidized oil
Headspace-gas chromatography
(GC)
Fatty acid prole
(FAP)
Saturated and unsaturated fatty
acids consist of the principal
components of fats. Fatty acid
proles are distinguishable
markers between olive oils and
some seed/nut oils (FAPs vary
slightly depending on the
varieties and growing region of
olives)
Analysis of the fatty acid prole
provides information on the
authenticity of the olive oil and
canbeusedasanindicatorfor
adulteration
Gas chromatography (GC)

Journal of Food Quality
Select
Critical
chemical
parameters and
their
acceptable
limits related to
shelf life
Evaluate
Changes in
products under
real-time and
accelerated
storage
conditions
Develop
Shelf-life
models based
on the results
from real-time
and accelerated
storage
conditions
Tes t
Prediction
models and
make
adjustment
based on the
known data
F : Shelf-life prediction model development process.
cultivars, Picual and Hojiblanca, were monitored throughout
two dierent storage conditions, together with the evolution
of the oxidative stability and sterols, polyphenols, alpha-
tocopherol, chlorophyllic and carotenoid pigments, and FAP.
In this study, a total of  L of EVOO was extracted and
packed for each cultivar in Spain. irty-four  L transparent
glass bottles of each cultivar were stored inside a thermostated
chamber at C with illumination ( lx;  h/day), which
was similar to commercial storage conditions. Other twelve
bottles of each cultivar were stored at Cindarkness.In
addition,  L EVOO of each cultivar was purchased in a
local market as commercial reference samples and stored
in the thermostated room. Bottles were sampled weekly
during the rst  days and subsequently every  days for
 months of storage. It is worth mentioning that the EVOO
samples produced from each cultivar had similar initial
values on most of the chemical parameters other than acidity
(Picual: .%; Hojiblanca: .%), stability (Picual: . h;
Hojiblanca: .h), and o-diphenols (Picual: . mg/kg;
Hojiblanca: . mg/kg).
Overall, samples stored at C in darkness remained min-
imal to unaltered throughout the entire storage period. us,
theregressionanalysiswasperformedonselectedchemical
parameters from samples stored at Cwithilluminationfor
each cultivar (Table ). Similar changes in PV of two olive
cultivars were observed in a -fold increase during the rst 
days and followed by a linear decrease until the end of storage.
e evolutions of alpha-tocopherol, chlorophyllic pigments
(CP), total polyphenols (TP), and o-diphenols were well tted
to rst-order kinetics. Most importantly, the coecient 270,
which measures the accumulation of secondary oxidation
productsthatcauseo-avorsinoliveoil,showedasharp
increase along the storage period in all the samples stored at
C with illumination in spite of cultivar and sample source.
As a result, an excellent correlation between initial stability
and time to reach the limit of 270 =.wasestablishedfor
EVOO samples bottled in glass containers regardless of olive
cultivar (Table ).
e correlation between initial stability ()andstorage
time to achieve 270 of . has demonstrated when an
EVOO no longer retains its extra virgin quality. Being a
critical indicator of oxidation level, 270 is required by the
IOC standard [] and can be easily obtained by producers.
Nonetheless, the validation of this model is in need for other
cultivars with a larger sample size in an extended storage
period. Storage containers other than glass type should also
betakenintoconsiderationwhenapplyingthismodel.
2.3. Psomiadou et al. (2003) [19]. To establish this empiri-
cal model, y-two Greek virgin olive oil (VOO) samples
(Koroneiki cv) from three consecutive crops (–)
were obtained as the training set for quality parameters
measurement. e measured parameters included FFA, PV,
UV, FAP, and the ratio of unsaturated and saturated fatty
acids, alpha-tocopherol, total phenols, total chlorophylls,
and OSI. Collinearity diagnostics, variable selection, and
regression analysis were performed on the obtained analytical
data to determine the contribution of each parameter to
maintaining VOO quality.
rough statistical analyses, the research team located
alpha-tocopherol, PV, total chlorophylls, and total phenols to
be the most important factors that aected OSI values and
yielded below model:
OSI =5.081+0.0102 alpha-tocopherol
−0.364 (PV)+0.0477 total chlorophylls
+0.0259 total phenols.
()
As shown in the above model, all antioxidants contributed in
a similar way to the OSI factor while PV posed clear negative
impact on the oxidative stability of the oil. e predictability
of this model was further examined and conrmed by a test
set of  VOO samples of the same cultivar from -
crop, which showed a negligible prediction bias and a low
square root of the mean square error of ., indicating an
eective prediction of OSI was achieved in this model for
VOO of Koroneiki cv.
In this study, the eect of many oxidative parameters
on oils from dierent crop years was examined with com-
prehensive statistical analyses, yielding a simple predictive
equation, and followed by validation on another  samples
fromthesamecultivar.However,whilethismodelgives
useful information regarding the oil stability which impacts
shelf life directly, it would require producers to incur the
expense for three tests (alpha-tocopherol, total phenols, and
total chlorophylls) that are not currently required in the
standards []. Besides, producers can request OSI analysis
(by Rancimat) for less of the cost than these three tests
although the correlation between OSI and actual shelf life
was not elaborated. Regardless, this model still has practical
inuence on the routine control of Koroneiki cv VOO in the
industry and future development of prediction models for
VOO made from other olive cultivars can be derived from
this validated model with minor modications.

Journal of Food Quality 
T : Summary of shelf-life prediction models discussed in the review.
Model Sample (bottles) Storage condition Chemical/physic al analysis Sensory analysis Statistical analysis Shelf-life prediction indicator Ref.
TuscanEVOOfromLots
(i) Dark glass
(ii) In the dark
(∘C)/uncontrolled light and
temperature
(iii) /// months; sampled
every  months
FFA, PV, UV, polyphenol,
alpha-tocopherol, tyrosol and
hydroxytyrosol, secoiridoid
aglycons, chlorophyll
absorbance, carotenoid
absorbance, induction time
(∘C,  L/h)
(i)  trained panelists
(ii) Bitterness and astringency
(i) Sample classication: principal component
analysis (PCA) and partial l east-squares
analysis (PLS) on Unscrambler . sow are
package (Camo As, Trondheim, Norway)
(ii) Comparison of re gression lines:
Statgraphi cs Plus . soware pack age
(Manugest KS Inc., Rockville, MD)
Time (in days) t o reach 232 of.  ():
(i) =1130.84 Ln (induction time)−2388.13
(ii) =329.02−38.11 (hydroxy tyro sol)
(iii) =580.34−68.11 (tyrosol)
Pagliar ini et al.
() []

 experimentalSpanish
Picual and Hojiblanca
EVOO and  commercial
Picual and Hojiblanca
EVOO
(i) Clear glass
(ii) ermostated chamber
(∘C)/in the dark (∘C)
(iii)  months; sampled weekly
during the rst  days and ever y
 days aerward
FFA, PV, UV, total phenols,
o-diphenols, tocopherols,
chlorophyllic and carotenoids,
FAP and sterols, induction time
(∘C,  L/h)
(i)  trained panelists
(ii) Highest quality (score ) to
lowest quality (score )
(i) Comparison of means: Duncan’stest
(ii) Analysis of vari ance and correlation:
CoStat. soware (CoHort Soware,
Berkeley, CA)
Time (in days) t o reach 270 of. () according toinitial
stability ():
=1.01−12.84
Guti´
errez and
Fern´
andez ()
[]

 Greek Koroneiki virgin
olive oil (VOO) from  to
 for model developme nt
and  Koroneiki VOO from
- for model
validation
Stored at −∘C till analysis
FFA, PV, UV, total phenols,
alpha-tocopherol, total
chlorophylls, FAP and the ratio
of unsaturatedand saturated fatty
acids (U/S), induction time
(∘C,  L/h)
NP∗
(i) Eect of the production year: linear
regression, -test and variable selection
(ii) Possible collinearity among independent
variables: variance ination factor(VIF), PCA
and singular value decomposition(SVD)
(iii) Selection of analytical parameter
aecting OSI the most: re gressions, forward
and backward selection, stepwis e selection,
Mallow’sindex and Akaike’s information
criterion (AIC)
(iv) Statistical packages: SAS version  (SAS
Institute Inc., Cary, NC), SPSS vers ion 
(IBM Corporation, Chicago, IL), and JMP IN
version .. (SAS Institute Inc., Cary,NC)
OSI =5.081+0.0102(alpha-tocopherol)−0.364(PV)+
0.0477(total chlorophylls)+0.0259 (total phenols)Ps omiadou et al.
() []

Large EVOO sample size
from Mediterranean areas
from  to  formodel
development and  EVOO
from  formodel
validation
NP
 chemical parameters
including FFA,PV, UV, minor
component content, lipid
oxidation status and antioxidant
activity
(i)  trained panelists
(ii)  sensory parame ters
including green olive
aroma/avor (e.g. bitterness),
astringency, and pungency
(i) Fractional factorial design (FFD): Modde
. soware package (Umetri, Umea,Sweden)
(ii) PCA and PLS: Unscrambler . soware
package (Camo AS, Oslo, Norway)
EVOO degradation paramete r value () as a function of
FFA (1),oleic acid content (2) and bitterness score
(3):
=+11+22+33+412+513+623
Zanoniet al. ()
[]
 Portuguese organic
EVOO
(i) PET/PVC/glass bottle
(ii) Covered with aluminum foil
inside berboard boxes/under
uorescentlightat//
∘C
with relative humidity of%
(iii)  months; sampled monthly
Hexanal N P
(i) Determination of dierences between
treatments for the hexanal evolution rate: SAS
.
(ii) Separation of t he means of GC areas:
Tukey test and Duncan test
(iii) Model development: nonuniform
nite-dierence scheme with an upwinding
numerical algor ithm (Newton method for
nonlinear systems)
Probability for the olive oil t o reach the end of its shelf life
during a certain period oftime (safe):
safe =1−∫𝑡2
𝑡1hexanal()
∫𝑡2
0hexanal()
Coutelieris and
Kanavouras ()
[, ]

Journal of Food Quality
T : C o n t i n u e d .
Model Sample (bottles) Storage condition Chemical/physic al analysis Sensory analysis Statistical analysis Shelf-life prediction indicator Ref.
Seven Spanish Cornicabra
VOO
(i) Opened amber glass
(ii) In the dark at ///∘C
(iii) sampled aer ///
weeks
PV, UV, FAP, phenolic
compounds, tocopherols,
induction time (∘C,  L/h)
NP
(i) Dierences between treatments: Duncan
test on SPSS version 
(ii) Linear and nonlinear reg ression:
GraphPad Prism . (GraphPad Soware,
Inc., San Die go, CA)
(iii) Oxidation and antioxidantdegradation
rates: calculated from the slopes of respective
concentration versus time curve
(iv) Eect of temperature on ratesof
oxidation: Arrhenius equation
Time (in weeks ) to reach 232 of.atamildtemperature
(T≤∘C):
TRUL =𝑏
Mancebo-Campos et
al. () []
Sixsingle-cultivarVOO
(i) Amber glass
(ii) In the dark at room
temperature/∘C
(iii)  months; sampled monthly
Chlorophyll pigments
(pheophytin a and
pyropheophytin a)
NP
(i) Dierences between means: one-way
ANOVA
(ii) Post hoc comparison: Brown and Forsythe
test []
(iii) PCA, PLS, and nonlinear regression:
Statisti ca . and Statgraphi cs Centurion XV
for Windows (StatSoInc., Round Rock, TX)
e percentage of pyroph eophytin a (%PPP) over time
(): % PPP ()=
(𝑎1−𝛽1/𝑇) [PP]0/(𝛼2−𝛽2/𝑇) −(𝛼𝑡𝑎−𝛽𝑡𝑎/𝑇)
⋅
−(𝑒(𝛼𝑡𝑎−𝛽𝑡𝑎/𝑇) )𝑡 −−(𝑒(𝛼2−𝛽2/𝑇))𝑡
×[PP]0−(𝑒(𝛼𝑡𝑎−𝛽𝑡𝑎/𝑇))𝑡
+(𝑎1−𝛽1/𝑇) [PP]0/(𝛼2−𝛽2/𝑇) −(𝛼𝑡𝑎 −𝛽𝑡𝑎/𝑇) 
⋅
−(𝑒(𝛼𝑡𝑎−𝛽𝑡𝑎/𝑇) )𝑡 −−(𝑒(𝛼2−𝛽2/𝑇))𝑡−1
Aparicio-Ruiz et al.
() []
Nineoliveoilsamples
Stored at ∘Cuntilanalysis
FFA, PV, total tocopherols and
total phenols, total polar
compounds, conjugateddiene
value, ratio of mono- and
poly-uns aturated fatty acids
(M/P ratio), and induction time
(///∘C,  L/h)
NP
ANOVA and regression analyses: MStatC
(Michigan State University, East Lansing, MI)
and SlideWrite(Sigma Aldrich, St. Louis,
MO)
Calculate shelf life at ∘C(SL
50) as a fu nction of
induction time (OSI):
SL50 =0.9985100−0
100 (OSI −)
+ 1((−OSI)/100)
1+OSI −−2/32+
Farhoosh an d
Hoseini-Yazdi ()
[]

A wide range of commercial
olive oil samples for model
development and  olive oil
and  EVOO for model
evaluation
(i) Dark glass
(ii) In the dark at  ±∘C
(iii)  months; sampled when
estimated best before date was
reached
FFA, PV, UV, PPP, DAGs, and
induction time (∘C,  L/h)
(i)  trained panelists
(ii) e median values of defect,
fruitiness,bitterness, and
pungency
NP
Best before date (i n months) from the lowest value
obtained from induction time, DAGs, FFA Factor (derived
from FFA) and PPP:
(a)Hours of induction time at 110∘C
(b)DAGs −35%
FFA Factor
FFA factor = .%(if FFA <.%); .%(if FFA >.%
and <.%); or .%(if FFA >.%)
(c)17%−PPPs
0.6%
Guillaume and
Ravetti () []
  EVOO
(i) Amber glass
(ii) In the dark/exposed to
naturalandarticiallightat
room temperature ( –∘C)
(iii)  months; sampled aer
//// months
FFA, PV, UV, induction time;
electronic tongue s ignal prole
(i)  trained panelists
(ii) Olfactory sensations,
gustatory-retronasal sensations,
and nal olfactory-gustatory
sensations
See Table   best linear discriminant analysis (LDA) and simulated
annealing (SA) prediction models
Rodrigues et al.
() []
∗NP:notprovided.

Journal of Food Quality 
T : Sample Lot information in the Pagliarini et al. () model.
Lot  Lot 
(reference lot) Lot 1Lot 2Lot 1Lot 2
Time taken from
freshly made batch
Immediately aer
processing
Aer  days of
storage in tanks
Aer  days of
storage in tanks
Aer  days of
storage in tanks
Aer  days of
storage in tanks
Bottling
 mL dark glass,
closed with screw
caps
 mL dark glass,
closed with screw
caps
 mL dark glass,
closed with screw
caps
 mL dark glass,
closed with screw
caps
 mL dark glass,
closed with screw
caps
Shipping
destination aer
bottling
Processing facility
in Italy (OL.MA.)
A supermarket in
Aust ra li a
A supermarket in
Aust ra li a
A supermarket in
Italy (close to
OL.MA.)
A supermarket in
Italy (close to
OL.MA.)
Storage condition
at destination In the dark at CUncontrolled light
and temperature
Uncontrol led light
and temperature
Uncontrol led light
and temperature
Uncontrol led light
and temperature
Storage period at
destination  months  months  months  months  months
T : Correlations between storage time (in days) and selected
parameters by Guti´
errez and Fern´
andez ().
Cultivar Picual Hojiblanca
PV (mequiv/kg) PV =−0.04+7.2;
=0.9532 PV =
−0.03+6.6;
=0.9600
TP (mg/kg, caeic acid)
Ln (%TP)=
−4.45×10−3+2.03;
=0.9879
Ln (%TP)=
−2.55×10−3+
1.97;=0.9965
CP (mg/kg)
Ln (CP)=
−0.11+12.34;
=0.9848
Ln (CP)=
−0.26t+18.92;
=0.9810
Initial stability ()forthe
achievement of 270 =0.25
(h) =1.01−12.84; =0.9823
2.4. Zanoni et al. (2005) [10]. A phenomenological model
was introduced for the rst time to predict the stability of
EVOO based on combined stability/instability composition
indices. e experimental design comprised two steps: ()
stability/instability indices screening and () signicant rela-
tionships between screened indices and EVOO degradation
investigation and conrmation. e screening of composi-
tion indices was carried out by multivariate analysis on data
derived from  chemical and  sensory parameters obtained
from oils purchased from four dierent Mediterranean area
duringthe–crops.Basedonthestatisticalanalysis,
the research group proposed that acidity value was indirectly
relatedtooilstabilitywhileoleicacidcontentandbitter
taste were directly related to oil stability. e predictability
ofthesemostrelevantindicestooilstabilitywasthen
checked by measuring six major degradation parameters
on eleven oil samples diering in screened indices planned
by a fractional factorial design (FFD) and processed with
principal component analysis (PCA) and partial least squares
(PLS) regression aerward. Parameters of PV, UV, minor
polar component content (oleuropein and ligstroside deriva-
tives), oxidative status of fatty acids, antioxidant activity, and
sensoryevaluationweremeasuredinthisstep.
Combining the results from PCA mapping and PLS
modeling has proved the hypothesis that in EVOO samples
() the more acidity the more degradation; () the more
oleic acid content the less degradation; and () the more
bitter the taste the less the degradation. Furthermore, PV, UV
232, and lipid oxidation status (oxidized fatty acid content
at  nm and dienoic and trienoic conjugated fatty acids
content) were found to be the most critical parameters when
measuring EVOO degradation. A mathematical model was
established to predict EVOO degradation as a function of the
combination of the three most relevant indices (acidity, oleic
acid content, and bitter taste):
=+11+22+33+412+513
+623,()
where is the selected degradation parameter to be pre-
dicted, 1is the acidity, 2is the oleic acid content, and 3
isbittertastescore.Constantvaluesa
are listed in Table .
Unlike what had been found in previous studies [, ,
], the antioxidant component content which consisted of
antioxidant activity and minor polar component content in
thisstudyshowedinsignicantimpactonEVOOdegradation
andwasexcludedfromtheproposedmodelasaresult.
Predictive models of oil degradation degree can be
obtained based o of the above proposed mathematical
model, which may be useful to predict the rate of oil
degradation if the oil degradation history was known. In this
regard, a major limitation of this model is that it was based
on constant indices without taking the composition changes
under storage conditions into full consideration. at is,
any changes of oil composition that occurred during lipid
oxidation would require a new set of quality index values to
maintain the model validation. Although this drawback may
be overcome by replicating the experimental design several
times, the application of this model may be limited to EVOO
samples being stored under optimal conditions that have
minimal eects on the change of stability/instability indices
and/or samples that yield similar rate of degradation kinetics
for the same stability/instability indices combination.

Journal of Food Quality
T : Constant values of selected degradation parameters for empirical polynomial models by Zanoni et al. () [].
Degradation parameter 
123456
PV −77.921 115.143 1.148 28.910 −1.411 −3.679 −0.367
232 −6.364 10.911 0.106 2.981 −0.133 −0.328 −0.037
Lipid oxidation status
Dienoic trans-trans −0.051 −0.103 0.002 0.136 0.002 −0.040 −0.001
Dienoic cis-trans, trans-cis 0.990 −0.510 0.006 0.054 0.023 −0.262 −0.002
Trienoic conjugated fatty acid content −1.562 2.960 0.020 0.563 −0.034 −0.183 −0.006
Oxidized fatty acid content at  nm −0.280 4.805 0.032 1.164 −0.054 −0.099 −0.020
T : Sample packaging and storage conditions for a -month study in the Coutelieris and Kanavouras () model.
Sample Portuguese organic EVOO
Packaging material . L PET bottle . L PVC bottle . L glass b ottle
Oxygen transmission rate at . atm driving force (cc/m2/day)  . N/A
Storage location Half covered with aluminum foil inside berboard boxes; half
exposed to uorescent light
Storage temperature (C) , , and 
Relative humidity (%) 
2.5. Coutelieris and Kanavouras (2006) [34, 36]. As listed in
Table , volatile compounds are good indicators of olive oil
quality as they are mainly produced through lipoxygenase
pathway and chemical oxidation during processing and
storage and contribute greatly to the olive oil avor []. In
this study, the evolution of hydroperoxide in the packaged
Portuguese organic EVOO samples was monitored and the
progression of hexanal, which was assumed as the most
prominent volatile compound posing higher impact on the
sensory attributes of olive oil, was quantied over a -month
storage period. Table  shows the packaging materials and
storage conditions of analyzed EVOO samples.
A mathematical model for the mass transfer taking place
in the oil-package material interacting system was deduced
based on four assumptions in the oil phase and two assump-
tions in the oil-package system. In the oil phase, the assump-
tions were () the oil quiescent; () all the hydroperoxides
eventually transformed to hexanal during lipid oxidation; ()
at time () = 0, there was a measurable amount of oxygen,
fattyacid,andhexanalintheoilphase;and()thepackaging
materials adsorbed hexanal according to Langmuir isotherm.
In the oil-package system, the assumptions included the
following: () oxygen and hexanal had constant concentration
outside the bottles at spatial coordinate () = 0; and ()
at =0, oxygen and hexanal concentrations inside the
packaging material were zero. e mass transfer phenomena
were elaborated explicitly by using diusion equations for dif-
fusion of oxygen and hexanal and Langmuir-type adsorption
for hexanal adsorption in the oil-package (cylinder bottle)
system. A numerical algorithm, along with a nonuniform
nite-dierence scheme, was then applied with modications
to solve the issue of nonlinearity of the studied system for
various combinations of the storage conditions mentioned in
Table .
A
B
⟨＃Ｂ？Ｒ；Ｈ；Ｆ⟩ol
t1t2
246810120
Time (months)
F : e graphical representation of the denition of safe.
e threshold of the hexanal concentration is represented by the
long dash dot line {adopted from Coutelieris and Kanavouras ()
[]}.
e study showed that samples kept under light had
yielded much higher concentration of hexanal when com-
pared to the samples stored in dark. In addition, the highest
hexanal concentration was found in samples stored in PET
bottles at Cwithlightexposure,followedbythosestored
in glass, and samples stored in PVC bottles had a lower
hexanal concentration. As shown in Figure , safe ,the
probability for the olive oil to reach the end of its shelf life
during a certain time period is comparable to the ratio of the
areasbelow(areaA)andabove(areaB)aroughlydened
threshold of the hexanal concentration (long dash dot line).
e estimation of saf e was proposed in the following model
duringthesametimeperiod[1,2]:
saf e =1−∫2
1hexanal()
∫2
0hexanal(),()

Journal of Food Quality 
where hexanal is the concentration of hexanal, 1is the time
when concentration reaches the upper limit for the oil’s
quality acceptance, 2is set to  months in this study, and the
brackets indicate spatial averaging of hexanal concentration
being used.
e sensitivity of this model was tested on samples kept
under dierent storage conditions and safe values were
compared for four dierent thresholds (%, %, %,
and % over the initial hexanal concentration). One of
the key ndings suggested that the predictions diverged
from experimental results under specic storage conditions
due to the low concentrations of hexanal in oil stored in
dark at any temperature. Moreover, the determination of
hexanal concentration threshold was ambiguous without
knowing data generated from additional chemical analyses
and sensory evaluation. Most importantly, the amount of
hexanal does not always allow oxidized olive oils to be distin-
guished from virgin ones, as this compound can come from
both lipoxygenase and oxidative pathways []. Nonetheless,
theproposedmodelhadundertakenacomprehensiveand
extensive investigation on the EVOO degradation in the oil-
package system by factoring in the chemical reactions and
diusion of compounds both in the oil phase and through
packaging materials, granting a promising parameter for
better monitoring the shelf life of packaged olive oil stored
under various conditions. e validation of safe model can
be further strengthened by adding sensory analysis.
2.6. Mancebo-Campos et al. (2008) [23]. During the storage
of seven Cornicabra cv VOO samples (varied in total phenol
concentrations) in dark and at mild temperatures (, ,
, and C), the autoxidation kinetic behavior of the main
oxidation indices (PV, 232,and270) and the oxidizing
substrate [unsaturated fatty acids (UFA)] were reported for
the rst time. In addition, the extrapolated time (in weeks)
required to reach the upper limit (TRUL) of each main
oxidation index in the EU regulation for the VOO category
was also calculated based on the experimental results from
this study and a previous study conducted by the same
research group [].
According to the evolution of measurements in this study,
PV did not reach its upper limit (meq/kg) in any samples
stored at Cbytheendofa-weekstorage,nordiditdo
that in some of the samples stored at higher temperatures.
Stabilization of PV was reached below or slightly above the
limit in all cases in spite of more harsh storage temperatures
and intensive air exposure in opened bottles. is similar
observation was also conrmed by other research groups as
a reduction in PV would occur due to the breakdown of
peroxides into secondary products [, ], indicating the
unreliability of PV being used as a quality marker for olive oil
shelf life. On the contrary, the upper limit of 232 (. K1%1cm)
wasreachedinsamplesstoredatanyconditionsalthough
232 and PV tended to stabilize at a similar value in each
sample stored at higher temperatures, following pseudo zero-
order kinetics before reaching the plateau. On the other hand,
the upper limit of 270 (. K1%1cm)wasreachedinall
sampleswithonlytwoexceptionsat
C, yielding pseudo
3.2 3.4 3.6 3.8 4.0 4.23.0
Ln Temperature
0.0
1.0
2.0
3.0
4.0
5.0
Ln TRUL
F : Correlation between TRUL and temperature for 232
(TRUL =
). Seven samples denoted in seven shades {adopted
from Mancebo-Campos et al. () []}.
rst-order kinetics. Furthermore, the polyunsaturated fatty
acids (PUFA) linoleic and linolenic acids showed a linear
decrease at a rate increasing with storage temperature, and
the best correlation was drawn between loss of PUFA and
increase of 232 at all temperatures as described by the linear
Arrhenius equation.
As a good indicator of primary oxidation level and an
easy parameter to determine, 232 showed high linearity
in the early stages of oxidation and presented excellent
correlation with loss of UFA. us, 232 was selected as
the best normalized oxidation index for potential shelf-life
estimation of VOO, dened as TRUL, at a mild temperature
(≤C):
TRUL =.()
BasedontheTRULresultsof232 generated at dierent
temperatures, the above model can be further explained by
Figure . As a result, the predicted TRUL at Cwasvery
close to the experimental TRUL at the same temperature
when applying the proposed model to accelerated storage
temperatures (,  and C).
Unlike drastic ASLT conditions where olive oil samples
aretestedonoxidativestabilityatover
C[,],
thismodelconductsanacceleratedstabilitytestatmild
temperatures below C and allows a time-saving shelf-life
prediction to reasonably estimate the actual shelf life of VOO
samples stored under normal storage conditions (C). It is
worthnotingthatVOOsampleswerestoredinopenbottles
throughout the study with intensive oxygen exposure, which
did not reect the actual storage conditions from a commer-
cialstandpoint.Besides,VOOsamplesusedinthisstudywere
from the same cultivar with similar initial concentrations on
the majority of the measured parameters. A follow-up study
focusing on the contribution of antioxidants content and fatty
acids unsaturation degrees to oxidation rates is also necessary
to test the applicability of the proposed model.

 Journal of Food Quality
2.7. Aparicio-Ruiz et al. (2012) [16]. Chlorophyll pigments
are sensitive to small amounts of degradation, which would
eventuallytakeplaceinanEVOOevenunderoptimalstorage
conditions. During storage, pheophytin a (PP) degrades to
PPP (Table ). e ratio of these two compounds therefore is
a useful parameter to track olive oil degradation over time.
is kinetic prediction model is established based on PPP
because PPP changes predictably with time under specic
temperatures [].
In developing this model, the research team stored six
single-cultivar VOO samples (Blanqueta cv,Arbequinacv,
Cornicabra cv, and Picual cv)inmLamberglassjarswith
% (v/v) headspace, in the dark at room temperature. e
monthly temperatures range from .Cto.
Cthrough-
out the year, with an average annual temperature of 19.3 ±
1.9C. Chlorophyll pigments were quantied every month up
to one year. e degradation of PP was found tting rst-
order kinetics aer applying multivariate statistical analysis
to the experimental data. e statistical results also showed
that time, temperature, and initial PP concentration were
the main variables that aected PPP prediction for shelf life.
Percent PPP (% PPP) over time was dened as the quotient of
the concentration of PPP ([PPP]) and the sum of [PPP] and
[PP]. A mathematical model to predict % PPP as a function
of time and temperature was then developed as shown below:
% PPP ()=(1−1/)[PP]0/(2−2/) −(𝑡𝑎−𝑡𝑎 /)−((𝛼𝑡𝑎−𝛽𝑡𝑎 /𝑇)) −−((𝛼2−𝛽2/𝑇))
[PP]0−((𝛼𝑡𝑎−𝛽𝑡𝑎/𝑇)) +(1−1/)[PP]0/(2−2/) −(𝑡𝑎−𝑡𝑎/)−((𝛼𝑡𝑎−𝛽𝑡𝑎/𝑇) ) −−((𝛼2−𝛽2/𝑇)).()
In this equation, [PP]0is the initial concentration of PP, 
is temperature in Kelvin, is the storage time in hour, and
values 1,1,2,and2are related to kinetic constants and
are protected by industrial license according to the authors.
Accordingtotheproposedmodel,%PPPatanytimepoint
canbecalculatediftheinitialPPandPPPconcentrationand
storagetemperatureareknown.
is study also compared the change of % PPP under
a well-controlled storage temperature of Candroom
temperature for six single-cultivar VOO samples. Overall,
% PPP increased under both temperatures, indicating the
degradation of olive oil quality occurred over time in spite of
cultivars. However, it is clear that the same samples stored at
room temperature had a signicant increase in % PPP from
 to above %, especially during summer time (– storage
months) when room temperature was typically higher. e
development of this parameter tended to be linear with a
smaller slope (from  to %) throughout the entire storage
period at C. is nding conrms the temperature impact
on PPP generation over time which should be taken into
consideration when developing the kinetic model.
Aer being validated on and compared with another
set of empirical data calculated from chlorophyll pigment
experimental data obtained by Gallardo-Guerrero, et al. [],
the model was adopted to develop a % PPP prediction graph
between Cand
C as shown in Figure . e authors
suggested that the % PPP acceptable limit could be set at %,
which would allow VOO to have one year of shelf life if stored
under C. However, this value seems arbitrary as it did take
into account any other chemical parameters and/or sensory
results.
In a follow-up study published in  [], the same
research team applied this % PPP prediction model to single-
cultivar olive oils (Arbequina cv) with various levels of initial
% PPP at bottling. e samples were stored at dierent
average annual temperatures, ranging from Cto
C.
e authors concluded that the initial value of % PPP is of
great importance to be included for a better monitoring of
the storage conditions of VOO. Table  shows shelf life (in
months) for VOO samples stored at Cand
Cbefore
reaching the Australian/California upper limit for PPP of
%. For instance, if % PPP is .% at bottling, the oil will
have more than  months and  months before it reaches
thelimitof%ifstoredat
Cand
C, respectively. ese
temperatures are likely to be cooler than the actual storage
temperature; thus a follow-up study with oil stored at a typical
store shelf temperature is recommended.
is model only consists of two chlorophyll pigments (PP
and PPP) and can be used to monitor the changes of storage
temperature and to detect undesired storage conditions based
on the rate of pyropheophytinization. Knowing the value of
% PPP at any moment during a storage period would also
allow a timely adjustment on proper storage temperature
and later on provide a better shelf-life estimation. However,
without knowing the values of other quality parameters
of VOO samples, the % PPP alone may not reect the
storage condition correctly as light exposure can cause the
complete breakdown of chlorophylls and all of its derivatives
therefore yield a zero value of % PPP []. Hence, inclusion of
other quality parameters of samples would benet the model
optimization.
2.8. Farhoosh and Hoseini-Yazdi (2013) [25]. e empirical
model was developed based on the relationship between
oxidative stability measurements (OSI) taken at high tem-
peratures (–C) and the chemical composition data
obtained at a low temperature (C).
To study the contribution of each compositional parame-
tertotheoxidativestabilityinoliveoil,nineoliveoilsamples
in  L glass b ottles were purchased from local shops and stored
at C until analysis on accelerated stability at –C.
And the ratios between mono- and polyunsaturated fatty
acids (M/P ratio), PV and FFA, total tocopherols (TT) and
total phenols (TP), total polar compounds (TPC), conjugated
diene value (CDV), and induction period (IP) were tested for
thestoragestabilityonthesamesamplesincubatedat
C.

Journal of Food Quality 
T : Shelf life (in months) for VOO stored at Cand
C
before reaching Australian/California limit of % summarized
from Aparicio-Ruiz et al. ().
% PPP at bottling Shelf life if stored at
C
Shelflifeifstoredat
C
. > 
. > 
. > 
.      
.  
0
14
28
42
56
70
84
% PPP
246810120
(Month)
35∘C
30∘C
25∘C
20∘C
15∘C
F : Predicted % PPP during one year of storage at temper-
atures between Cand
C{adopted from Aparicio-Ruiz et al.
() []}.
During the storage stability test conducted at C, the
evolution of hydroperoxides and conjugated dienes showed
two pseudo zero-order kinetic curves: a gradual slope of
linear stage which was considered as the initiation phase
oflipidoxidationandthenasteepslopeofanotherlinear
stage known as the propagation phase. e storage time (in
days) at intersection points of the PV and CDV curves was
identied as the induction period IPPV and IPCDV for the olive
oil sample. e level of hydroperoxides increased gradually
duringIPandthenelevatedrapidlyinthepropagation
phase, where decomposition of hydroperoxides to aldehydes,
ketones, and other secondary oxidative products occurred
and o-avors were accumulated []. us, the IP-based
oxidative stability values IPPV and IPCDV were selected as
better parameters to measure the oxidative stability and
determine the shelf life of olive oil at C.
Positive correlations were found between oxidative sta-
bility (IPCDV at CandOSIat–
C) and M/P ratio,
tocopherols, and phenolics. To further elucidate, the higher
theM/Pratiois,thelesspronetoranciditytheoliveoilis;the
higher the content of tocopherols and/or phenolics of the oil
is, the better the antioxidative ability the oil has. It is worth
mentioning that the order of the IP-based oxidative stability
of olive oil samples at Cwassample7>2>6≈3>
9>4≈1>8>5, whereas the order of that determined
by the OSI from the accelerated stability test at –C
followed sample 7>9>3>8>6≈2≈1>5>4.e
dierence may be indicative of the fact that the extrapolation
from the OSI obtained at accelerated temperature to ambient
conditions could lead to over- or underprediction of the
actual shelf life due to complicated kinetics involved at higher
temperature [, ].
Regression models developed under low- (model (a))
and high-temperature (model (b)) were also provided based
on the analytical data generated from either condition. By
incorporating the chemical composition data collected at
CintotheOSImeasurementat–
C, an empirical
model (c) of shelf-life prediction (SL50) was derived from
model (a) and model (b):
(a)SL50 =24.06393

=1∗+104.7369, ()
where and are regression coecients and standardized
compositional variables.
(b)SL50 =10(50+−1.2272),()
where and values are slopes and intercept of the linear
equation to the log OSI versus accelerated temperature.
(c)SL50 =0.9985100−0
100 (OSI −)
+ 1((−OSI)/100)
1+OSI −−2/32+, ()
where ,,,andarethevaluesofthelinearregression
models developed at high and low temperatures, respectively.
e values of 0,1,2,and3were calculated and shown
in Table . According to the conclusion of 10 factor from
Mancebo-Campos et al. [], which is that a decrease of C
in the storage temperature increases the shelf life of olive
oil more than two folds, a value of . of 10 was used to
estimatetheoilshelflifeat
C(normalstoragetemperature)
by using the regression model (a) and promising estimation
was obtained (.–. months) which was considered to
be representative of the typical shelf life claimed by olive oil
producers (– months aer production).
Model (c) permitted the estimation of olive oil shelf life to
be achieved within acceptable errors less than ±% by using
only one measurement, OSI, at accelerated temperatures. e
interrelated mathematical equation of the low- and high-
temperature regression models also allows real-time shelf-
life prediction from the accelerated testing results to be
done rapidly. A limitation of this model is that only two
EVOO samples were analyzed; without performing further
validation on the empirical model on a larger size of EVOO
samples,thecalculatedvaluesprovidedinTableandthe
correction coecient of . may considerably deviate and
notreecttheactualsituationofEVOOcategory.
2.9. Guillaume and Ravetti (2016) [18]. is empirical model
uses four quality parameters, induction time, DAGs, FFA
Factor (derived from FFA), and PPP, to identify a best before

 Journal of Food Quality
T : e values of 0,1,2,and3in the SL50 prediction model calculated at Cand
C by Farhoosh and Hoseini-Yazdi ().
Tempe r a t u r e ( C) 0123
 −174.4761 −201.4675 32898.6507 19757.9412
 0.0108 0.0230 −0.0106 0.0190
date (BBD, in months) using the lowest value obtained from
the following three equations:
(a)Hours of induction time at 110C
(b)DAGs −35%
FFA Factor
()
FFA factor = .% (if FFA <.%); .% (if FFA >.%
and <.%); or .% (if FFA >.%):
(c)17%−PPPs
0.6%()
is model recognizes that induction time generally cor-
relates with olive oil FAPs and antioxidant content. DAGs
andPPPhavebeenshowntobepredictableandchange
linearly with time whereas FFA provides a value for the initial
oil quality and does not change signicantly under proper
storage conditions. ese four quality parameters represent
factors that can aect olive oil shelf life over time.
To evaluate this empirical model, the research team
analyzed  samples for FFA, PV, UV, PPP, DAGs, and
sensory evaluation during a -month storage period. e
sampleswerestoredinadarkenvironmentat
C±C,
and tested immediately aer reaching their estimated best
before date. Of the  samples, only one sample (.% of
total samples) exceeded the Australian limit of .% for FFA;
no sample failed the Australian limit for PV ( meq O2/kg)
or 232 (. K1%1cm); two samples (.%) failed 270 limit
(. K1%1cm); twelve samples (.%) failed the Australian
limit of % for PPP; six samples (.%) failed the Australian
limit of % for DAGs; and ten samples (.%) failed sensory
evaluation. In addition to testing  samples at the end of
shelf life under controlled storage condition,  samples with
predicted shelf life were randomly collected from dierent
retailers every three months during a -month storage
period to validate the model from retailers’ standpoint (
samples in total). Only one sample (.% out of  samples)
exceeded the limit for 270 and two samples (%) exceeded
the limit for DAGs at their predicted BBD. By recalculating
and comparing the actual and predicted BBD, the data
suggested that producers may want to deduct - months
from the BBD given from the model to compensate for the
potential exposure to heat and light during transportation,
handling, storage, and display on the retail shelves.
is model was validated on a total of  samples,
including  commercial samples from real-time storage
conditions on the market, with simple and straightforward
calculations and yielded clear output. Modications to the
predicted BBD are necessary when storage condition is not
ideal; however, this would be true for any models that are
designed for the ideal packaging and storage conditions for
oliveoilshelflife(Table).
T : Recommended packaging and storage conditions for olive
oil shelf life.
Packaging Temperature Light
Dark glass, aluminum cans with
food-grade enamel coating, coated
paperboard, and bag-in-box provide
protection from light and oxygen.
Bag-in-box also has the advantage of
maintaining minimum oxygen in
headspace []
Stored at a
reduced
temperature
of C[]
Stored in
the dark to
minimize
light
exposure
2.10. Rodrigues et al. (2017) [38]. Coupled with powerful sta-
tistical linear discriminant analysis (LDA) [, ] and sim-
ulated annealing (SA) variable selection algorithm approach
[–], a most recent study on the evaluation of EVOO shelf
life was conducted by applying a potentiometric electronic
tongue (E-tongue) with nonspecic cross-sensitivity lipid
membranes to assess the commercial storage conditions
including light exposure and storage time.
e research group had analyzed  amber glass-bottled
EVOO samples on sensory attributors (conducted by four
trained panelists to classify samples based on olfactory sen-
sations, gustatory-retronasal sensations, and nal olfactory-
gustatory sensations), physicochemical parameters (FFA, UV,
and PV) and oxidative stability (OSI), and electrochemical
signal proles (E-tongue device with two print-screen poten-
tiometric arrays containing  sensors on each one). To fur-
ther elucidate the sample storage and testing conditions, four
fresh samples were analyzed immediately aer processing
at T ( month) while  samples were kept under room
temperature (–C) for one year in the lab, with  samples
being stored in dark and  samples being exposed to natural
light and articial light ( h/day from eight uorescent
lamps) to create a  ×× experimental factorial design.
During the one-year storage period, four samples were taken
out and analyzed every three months at time points of T, T,
T, and T. As a result, the quality parameters and oxidative
stability of the tested EVOO samples were indeed aected by
both the storage time and light conditions. It was inferred
by the authors that, being stored in amber bottles, light
conditions played less signicant role on the olive oil quality
deterioration during the storage period. In addition, not all
the positive attributors of EVOO samples were aected by
the storage conditions aer one-year storage period, although
samples exposed to light showed the strongest correlations
among the respective sensory attributors (-Pearson ≥.).
To evaluate the possibility of correctly categorizing olive
oil samples based on storage time and/or light conditions
(dark/light), the Kennard-Stone selection algorithm (a uni-
form mapping algorithm that generates a at distribution
of data suitable for regression model development) []

Journal of Food Quality 
T : Summary of statistical analysis performed by Rodrigues et al. ().
Objective Statistical analysis∗Statistical
analysis ref.
Compare the impact of dark/light storage conditions on olive oils for each
storage time Student’s -test []
Assess the eect of the storage time on olive oils stored in dark/light One-way ANOVA, Tukey’s post hoc
multicomparison test []
Evaluate the existence of bivariate correlations within the olive oil’s
physicochemical parameters
Linear Pearson correlation coecient
(-Pearson) []
Test the capability of the E-tongue to correctly classify the EVOO based on
storage time or storage conditions as a supervised pattern recognition
method
Linear discriminant analysis (LDA) [, ]
Evaluate the qualitative classication capability of physicochemical and
sensory data LDA [, ]
Select the best subsets of independent predictors among  E-tongue
potentiometric signals
Metaheuristic simulated annealing (SA)
variable selection algorithm [–]
Compare the current and the new subsets of (⊆K) variables Tau quality criterion []
Evaluate the LDA classication models Leave-one-out cross-validation (LOO-CV) [, ]
Minimize the risk of overtting from LOO-CV when sample size is large
and generate a at distribution of data for regression model development
 olive oil samples used as “training set” for
LOO-CV [, ]
oliveoilsamplesusedas“testingset”using
Kennard-Stone algorithm []
∗All statistical analyses were performed using Subselect [, ] and MASS [] packages of the statistical program R (version ..) at a signicant level of %.
was adopted by splitting  bottled olive oil samples into
two subsets ( for internal validation and  for exter-
nal validation). By applying the metaheuristic SA variable
selection algorithm, the best subset to be included in each
LDA model was selected from physicochemical parameters,
sensory attributors, and E-tongue signal proles for the deter-
mination of the eect of dierent storage conditions on the
quality of EVOO samples. e internal validation statistical
data showed that E-tongue signal proles yielded an over-
all better predictive discrimination performance, enabling
the establishment of three best LDA-SA prediction models
(from  to  sensor/sensor-replicas as independent variables)
without redundant variables. e external validation further
justied the predictive capability of E-tongue by giving a
representative ngerprint of the polar compounds in olive oil
samples.
As a promising chemometric approach, combining E-
tongue measurement and comprehensive statistical analysis
(Table ) could successfully determine the freshness of
EVOO samples during normal commercial storage condi-
tions (stored in dark or exposed to light for one year)
and provide accurate shelf-life prediction. Nonetheless, it is
important and necessary that at least eight trained panelists
were presented to provide sensory data as the lack of sen-
sory data could signicantly inuence the statistical analysis
results.
3. Conclusion
e global production and consumption of olive oil has
escalated signicantly in the past decade []. According to
the IOC Market Newsletter released in September, , the
producerpricesofEVOOhaveincreasedbymorethan%
(ineuros)inSpain,Italy,Tunisia,andGreececomparedtothe
same period in the previous year []. us, to maintain the
high quality of EVOO products during commercial activities
hasbecomeanurgentmattertooliveoilproducersandbeing
able to accurately predict the shelf life of EVOO products
would greatly benet both producers and consumers.
EVOO quality can be safeguarded by using proper pack-
aging, ideal storage conditions (cool and dark), and having
an accurate best before date. Currently in literature, common
parameters that are being used to track the changes in olive
oil include FFA, PV, UV, DAGs, PPP, sensory evaluation,
induction time, FAP, total tocopherols and total phenols, and
volatiles. A mathematical model for tracking deterioration
using sensitive and accurate quality parameters can be a
powerful and aordable tool for accurately predicting olive
oil shelf life.
In this review, ten practical mathematical models that
have potential to be adopted and utilized by olive oil pro-
ducers are summarized. Nonetheless, each of the models can
benet from further study with a large set of samples under
real-life transport and storage conditions, monitoring both
compositional and environmental variables. To establish a
robust and systematic model for shelf life assessment, the
most urgent tasks are () to remove unnecessary parameters
and to conrm the acceptable limits without losing the
predictability and accuracy and () to continue developing
and ne-tuning accelerated methods to minimize their ten-
dency for overprediction or underprediction of actual shelf
life. By reducing inessential parameters used in a model,
the processing time and cost of shelf life assessment are
also reduced. Since sensory evaluation remains to be one

 Journal of Food Quality
of the most sensitive methods for olive oil quality and
freshness, a working model should be calibrated with sensory
evaluation and complement sensory evaluation for olive oil
freshness assessment in the future. Temperature, airow rate,
and oil sample size have signicant impacts on shelf life
prediction when using accelerated methods. It is critical to
adjust and optimize the operational settings to minimize the
discrepancy between the real-time shelf life and accelerated
prediction of an EVOO product.
Conflicts of Interest
e authors declare that there are no conicts of interest
regarding the publication of this paper.
Acknowledgments
isreviewwasmadepossiblewiththenancialsupport
from the Olive Oil Commission of California. e authors
wouldalsoliketothankLeandroRavettiandDanFlynnfor
helpful discussions.
References
[] S. Teres, G. Barcel´
o-Coblijn, M. Benet et al., “Oleic acid content
is responsible for the reduction in blood pressure induced by
olive oil,” Proceedings of the National Acadamy of Sciences of the
United States of America,vol.,no.,pp.–,.
[] R.W.Owen,A.Giacosa,W.E.Hull,R.Haubner,B.Spiegel-
halder, and H. Bartsch, “ e antioxidant/anticancer potential of
phenolic compounds isolated from olive oil,” European Journal
of Cancer, vol. , no. , pp. –, .
[] H.L.Newmark,“Squalene,oliveoil,andcancerrisk:areview
and hypothesis,” Cancer Epidemiology, Biomarkers & Preven-
tion,vol.,no.,pp.–,.
[] R. Mateos, M. M. Dom´
ınguez,J.L.Espartero,andA.Cert,
“Antioxidant Eect of Phenolic Compounds, -Tocopherol , and
Other Minor Components in Virgin Olive Oil,” Journal of
Agricultural and Food Chemistry,vol.,no.,pp.–,
.
[] F.Angerosa,R.Mostallino,C.Basti,andR.Vito,“Virginoliveoil
odour notes: eir relationships with volatile compounds from
the lipoxygenase pathway and secoiridoid compounds,” Food
Chemistry,vol.,no.,pp.–,.
[] A. M. Inarejos-Garc´
ıa, S. G´
omez-Alonso, G. Fregapane, and
M. D. Salvador, “Evaluation of minor components, sensory
characteristics and quality of virgin olive oil by near infrared
(NIR) spectroscopy,” Food Research International,vol.,no.,
pp.–,.
[] F. Angerosa, M. Servili, R. Selvaggini, A. Taticchi, S. Esposto,
and G. Montedoro, “Volatile compounds in virgin olive oil:
Occurrence and their relationship with the quality,” Journal of
Chromatography A, vol. , no. -, pp. –, .
[] S. Fadiloglu and Z. Soylemez, “Kinetics of lipase-catalyzed
hydrolysis of olive oil,” Food Research International,vol.,no.
-, pp. –, .
[] E. Frankel, Lipid Oxidation, Elsevier, .
[] B.Zanoni,M.Bertuccioli,P.Rovellini,F.Marotta,andA.Mattei,
“A preliminary approach to predictive modelling of extra virgin
olive oil stability,” Journal of the Science of Food and Agriculture,
vol. , no. , pp. –, .
[] A. Kiritsakis, G. D. Nanos, Z. Polymenopoulos, T. omai, and
E. M. Sfakiotakis, “Eect of fruit storage conditions on olive oil
quality,” Journal of the American Oil Chemists’ Society,vol.,
no. , pp. –, .
[] L. Di Giovacchino, M. Solinas, and M. Miccoli, “Eect of
extraction systems on the quality of virgin olive oil,” Journal of
the American Oil Chemists’ Society, vol. , no. , pp. –,
.
[]G.Pristouri,A.Badeka,andM.G.Kontominas,“Eectof
packaging material headspace, oxygen and light transmission,
temperature and storage time on quality characteristics of extra
virgin olive oil,” Food Control, vol. , no. , pp. –, .
[] S. Wang, X. Li, R. Rodrigues, and D. Flynn, Packaging Inuences
onOliveOilQuality:AReviewofeLiterature,UCDavisOlive
Center, .
[] X.Li,H.Zhu,C.F.Shoemaker,andS.C.Wang,“eeectof
dierent cold storage conditions on the compositions of extra
virgin olive oil,” JournaloftheAmericanOilChemists’Society,
vol.,no.,pp.–,.
[] R. Aparicio-Ruiz, M. Roca, and B. Gandul-Rojas, “Mathemat-
ical model to predict the formation of pyropheophytin a in
virgin olive oil during storage,” Journal of Agricultural and Food
Chemistry,vol.,no.,pp.–,.
[] N. Sinelli, M. S. Cosio, C. Gigliotti, and E. Casiraghi, “Prelimi-
nary study on application of mid infrared spectroscopy for the
evaluation of the virgin olive oil “freshness”,” Analytica Chimica
Acta, vol. , no. , pp. –, .
[] C. Guillaume and L. Ravetti, “Shelf-Life Prediction of Extra
VirginOliveOilsUsinganEmpiricalModelBasedonStandard
Quality Tests,” Journal of Chemistry,vol.,ArticleID
, .
[] E. Psomiadou, K. X. Karakostas, G. Blekas, M. Z. Tsimidou,
and D. Boskou, “Proposed parameters for monitoring quality of
virgin olive oil ( Koroneiki cv),” European Journal of Lipid Science
and Technology,vol.,no.,pp.–,.
[] E. Pagliarini, B. Zanoni, and G. Giovanelli, “Predictive study on
tuscan extra virgin olive oil stability under several commercial
conditions,” JournalofAgriculturalandFoodChemistry,vol.,
no.,pp.–,.
[] M. Nicoli, e Shelf Life Assessment Process,CRCPress,Boca
Raton, Fla, USA, .
[] A. Kanavouras, P. Hernandez-Munoz, and F. A. Coutelieris,
“Packaging of olive oil: quality issues and shelf life predictions,”
Food Reviews International,vol.,no.,pp.–,.
[] V. Mancebo-Campos, G. Fregapane, and M. D. Salvador,
“Kinetic study for the development of an accelerated oxidative
stability test to estimate virgin olive oil potential shelf life,”
European Journal of Lipid Science and Technology,vol.,no.
, pp. –, .
[] A.Kaya,A.R.Tekin,andM.D. ¨
Oner, “Oxidative stability of
sunower and olive oils: comparison between a modied active
oxygen method and long term storage,” LWT- Food Science and
Technolog y ,vol.,no.,pp.–,.
[] R. Farhoosh and S.-Z. Hoseini-Yazdi, “Shelf-life prediction of
oliveoilsusingempiricalmodelsdevelopedatlowandhigh
temperatures,” Food Chemistry,vol.,no.,pp.–,.
[] A. Kanavouras, P. Hernandez-M¨
unoz,F.Coutelieris,andS.
Selke, “Oxidation-derived avor compounds as quality indica-
tors for packaged olive oil,” JournaloftheAmericanOilChemists’
Society,vol.,no.,pp.–,.

Journal of Food Quality 
[] S.Dabbou,I.Gharbi,S.Dabbou,F.Brahmi,A.Nakbi,andM.
Hammami, “Impact of packaging material and storage time on
olive oil quality,” African Journal of Biotechnology,vol.,no.,
pp. –, .
[] D. A. Tsimis and N. G. Karakasides, “How the choice of con-
tainer aects olive oil quality - A review,” Packaging Technology
and Science,vol.,no.,pp.–,.
[] D. Krichene, M. D. Salvador, and G. Fregapane, “Stability of
virgin olive oil phenolic compounds during long-term storage
( months) at temperatures of -C,” Journal of Agricultural
and Food Chemistry, vol. , no. , pp. –, .
[] A. I. M´
endez and E. Falqu´
e, “Eect of storage time and
container type on the quality of extra-virgin olive oil,” Food
Control,vol.,no.,pp.–,.
[] L. Rastrelli, S. Passi, F. Ippolito, G. Vacca, and F. de Simone,
“Rate of degradation of -tocopherol, squalene, phenolics, and
polyunsaturated fatty acids in olive oil during dierent storage
conditions,” JournalofAgriculturalandFoodChemistry,vol.,
no.,pp.–,.
[] IOC, Trade Standard Applying to Olive Oils and Olive Pomace
Oils, COI/T./NC No /Rev. , .
[] E. N. Frankel, “In search of better methods to evaluate natural
antioxidants and oxidative stability in food lipids,” Tre n ds in
Food Science & Technology,vol.,no.,pp.–,.
[] F. A. Coutelieris and A. Kanavouras, “Experimental and the-
oretical investigation of packaged olive oil: development of a
quality indicator based on mathematical predictions,” Journal
of Food Engineering,vol.,no.,pp.–,.
[] F. Guti´
errez and J. L. Fern´
andez, “Determinant parameters and
components in the storage of virgin olive oil. Prediction of
storage time beyond which the oil is no longer of ‘extra’ quality,”
Journal of Agricultural and Food Chemistry,vol.,no.,pp.
–, .
[] A. Kanavouras and F. A. Coutelieris, “Shelf-life predictions for
packaged olive oil based on simulations,” Food Chemistry,vol.
,no.,pp.–,.
[] M. B. Brown and A. B. Forsythe, “Robust tests for the equality
of variances,” JournaloftheAmericanStatisticalAssociation,vol.
, no. , pp. –, .
[] N. Rodrigues, L. G. Dias, A. C. A. Veloso, J. A. Pereira, and A.
M. Peres, “Evaluation of extra-virgin olive oils shelf life using
an electronic tongue—chemometric approach,” European Food
Research and Technology,vol.,no.,pp.–,.
[] C. M. Kalua, M. S. Allen, D. R. Bedgood Jr., A. G. Bishop, P. D.
Prenzler, and K. Robards, “Olive oil volatile compounds, avour
development and quality: a critical review,” Food Chemistry,vol.
, no. , pp. –, .
[] M. T. Morales,J. J. Rios, and R. Aparicio, “Changes in the volatile
composition of virgin olive oil during oxidation: Flavors and
O-avors,” Journal of Agricultural and Food Chemistry,vol.,
no.,pp.–,.
[] S. G´
omez-Alonso, V. Mancebo-Campos, M. D. Salvador, and
G. Fregapane, “Evolution of major and minor components and
oxidation indices of virgin olive oil during  months storage at
room temperature,” Food Chemistry,vol.,no.,pp.–,
.
[] S. A. Vekiari, P. Papadopoulou, and A. Koutsaakis, “Compar-
ison of dierent olive oil extraction systems and the eect of
storage conditions on the quality of the virgin olive oil,” Grasas
yAceites,vol.,no.,pp.–,.
[] R. Farhoosh, “e eect of operational parameters of the Ranci-
mat method on the determination of the oxidative stability
measures and shelf-life prediction of soybean oil,” Journal of the
American Oil Chemists’ Society, vol. , no. , pp. –, .
[] M. Mart´
ın-Polvillo, T. Albi, and A. Guinda, “Determination of
trace elements in edible vegetable oils by atomic absorption
spectrophotometry,” Journal of the American Oil Chemists’
Society,vol.,no.,pp.–,.
[] S. J. Schwartz and J. H. Elbe, “Kinetics of Chlorophyll Degrada-
tion to Pyropheophytin in Vegetables,” Journal of Food Science,
vol.,no.,pp.–,.
[] L. Gallardo-Guerrero, B. Gandul-Rojas, M. Roca, and M. I.
M´
ınguez-Mosquera, “Eect of storage on the original pigment
prole of Spanish virgin olive oil,” Journal of the American Oil
Chemists’ Society,vol.,no.,pp.–,.
[] R. Aparicio-Ruiz, R. Aparicio, and D. L. Garc´
ıa-Gonz´
alez,
“Does “best before” date embody extra-virgin olive oil fresh-
ness?” Journal of Agricultural and Food Chemistry,vol.,no.
,pp.–,.
[] M. O’Mahony, Sensory Evaluation Of Food: Statistical Methods
and Procedures, CRC Press, Boca Raton, Fla, USA, .
[] A. J. Izenman, Modern Multivariate Statistical Techniques,
Springer,NewYork,NY,USA,.
[] J. N. Miller and J. C. Miller, Statistics and Chemometrics for
Analytical Chemistry, Pearson Education Limited, England,
.
[] J. Cadima, J. O. Cerdeira, and M. Minhoto, “Computational
aspects of algorithms for variable selection in the context
of principal components,” Computational Statistics & Data
Analysis,vol.,no.,pp.–,.
[] S. Kirkpatrick, J. Gelatt, andM. P. Vecchi, “Optimization by sim-
ulated annealing,” American Association for the Advancement of
Science: Science,vol.,no.,pp.–,.
[] D. Bertsimas and J. Tsitsiklis, “Simulated annealing,” Statistical
Science,vol.,no.,pp.–,.
[] L.G.Dias,A.Fernandes,A.C.A.Veloso,A.S.C.Machado,
J. A. Pereira, and A. M. Peres, “Single-cultivar extra virgin olive
oil classication using a potentiometric electronic tongue,” Food
Chemistry,vol.,pp.–,.
[] J. M. Guti´
errez, Z. Haddi, A. Amari et al., “Hybrid electronic
tongue based on multisensor data fusion for discrimination of
beers,” Sensors and Actuators B: Chemical, vol. , pp. –,
.
[] R. W. Kennard and L. A. Stone, “Computer aided design of
experiments,” Te c h n o m e t r ic s , vol. , no. , pp. –, .
[] M. Kuhn and K. Johnson, Applied Predictive Modeling,Springer,
New York, NY, USA, .
[] W. N. Venables and B. D. Ripley, Modern Applied Statistics with
S-PLUS, Springer Science & Business Media, New York, NY,
USA, .
[] J. Ayton, R. Mailer, and K. Graham, e Eect of Storage Condi-
tions on Extra Virgin Olive Oil Quality, Australian Government
Rural Industries Research and Development Corporation, .
[] B. Lynch and A. Rozema, “Olive oil: conditions of competition
between us and major foreign supplier industries,” United States
International Trade Commission, Wash, USA, .
[] IOC, September market newsletter No , .