Survival Analysis Applied to the Sensory

Shelf-Life Dating of High Hydrostatic Pressure

Processed Avocado and Mango Pulps
D.A. Jacobo-Velázquez, P.A. Ramos-Parra, and C. Hernández-Brenes

Abstract: High hydrostatic pressure (HHP) pasteurized and refrigerated avocado and mango pulps contain lower microbial
counts and thus are safer and acceptable for human consumption for a longer period of time, when compared to fresh
unprocessed pulps. However, during their commercial shelf life, changes in their sensory characteristics take place and
eventually produce the rejection of these products by consumers. Therefore, in the present study, the use of sensory
evaluation was proposed for the shelf-life determinations of HHP-processed avocado and mango pulps. The study
focused on evaluating the feasibility of applying survival analysis methodology to the data generated by consumers in
order to determine the sensory shelf lives of both HHP-treated pulps of avocado and mango. Survival analysis proved to
be an effective methodology for the estimation of the sensory shelf life of avocado and mango pulps processed with HHP,
with potential application for other pressurized products.
Keywords: high hydrostatic pressure processing, sensory shelf life, survival analysis

Practical Application: At present, HHP processing is one of the most effective alternatives for the commercial nonthermal
pasteurization of fresh tropical fruits. HHP processing improves the microbial stability of the fruit pulps significantly;
however, the products continue to deteriorate during their refrigerated storage mainly due to the action of residual
detrimental enzymes. This article proposes the application of survival analysis methodology for the determination of
the sensory shelf life of HHP-treated avocado and mango pulps. Results demonstrated that the procedure appears to be
simple and practical for the sensory shelf-life determination of HHP-treated foods when their main mode of failure is
not caused by increases in microbiological counts that can affect human health.
S: Sensory & Food
Quality

Introduction Jacobo-Velázquez and Hernández-Brenes 2009). Although these

High hydrostatic pressure (HHP) processing is an effective non-
thermal technology for the pasteurization and shelf-life enhance-
ment of tropical fruit pulps and other food products. In many
cases, properly HHP-pasteurized fruit pulps are microbiologically
stable during storage. However, the quality of these products is
affected by the residual activity of enzymes that are not com-
pletely inactivated by the commercial HHP treatments. Therefore,
low- and high-acid fruit pulps processed by HHP are commonly
stored under refrigeration to reduce quality losses. Avocado and
mango pulps processed by HHP are products commercially avail-
able, and have been reported to be stable to the action of spoilage
microorganisms during refrigerated storage (López-Malo and oth-
ers 1998; Guerrero-Beltrán and others 2006; Ramos-Parra 2006;
MS 20090893 Submitted 9/10/2009, Accepted 3/24/2010. Authors Jacobo-
Velázquez, Ramos-Parra, and Hernández-Brenes are with Dept. of Biotechnology
and Food Engineering, School of Biotechnology and Health, Tecnológico de Monterrey-
Campus Monterrey, E. Garza Sada 2501 Sur, C.P. 64849, Monterrey, N.L.,
México. Author Jacobo-Velázquez is also with Dept. of Horticultural Sciences, Texas
A&M Univ., HFSB Building, Room 202, MS 2133, College Station, TX 77843-
2133, U.S.A. Direct inquiries to author Hernández-Brenes (E-mail: chbrenes@
itesm.mx).
products are safe for human consumption for long-storage peri-
ods, consumers reject them after a certain period of time due to
physicochemical and sensory changes.
Previous shelf-life estimation reports for HHP-treated products
generally base their determinations on physicochemical measure-
ments, which can result in the overestimation of the shelf life
of these products. However, when food products have enhanced
microbiological stability, their shelf life is limited by changes in
their sensory properties (Hough and others 2003). Commercial
HHP-treated foods generally contain low levels of microorgan-
isms and their acceptability is mainly affected by storage-related
sensory changes, therefore, survival analysis methodology is be-
ing proposed herein as a suitable option to determine their shelf
life. Survival analysis is a statistical data analysis procedure widely
used in clinical, biological, and epidemiological studies, and it has
been proposed as a potential tool to estimate the sensory shelf
life of microbiologically stable food products. The key concept
of this shelf-life determination methodology is to focus on the
risk of rejection by consumers rather than on product deteriora-
tion (Hough and others 2003; Gámbaro and others 2004a, 2004b,
2006; Salvador and others 2006).
The objective of this study was to determine the sensory shelf
life of HHP-treated avocado and mango pulps. To the best of
our knowledge, there are no previous reports focused on the

C 2010 Institute of Food Technologists

 R

S286 Journal of Food Science r Vol. 75, Nr. 6, 2010 doi: 10.1111/j.1750-3841.2010.01656.x
Further reproduction without permission is prohibited
Survival analysis and HHP-treated fruits . . .

determination of the sensory shelf life of these commercially avail- establishes a random variable T (storage time at which consumers
able HHP-treated fruit pulps. reject the product); and then a probability of a consumer rejecting
the product before or at a storage time t, F(t) = P(T ≤ t), which
Materials and Methods is estimated based on the interval obtained from each consumer.
Survival times generally do not follow a normal distribution.
Materials and high hydrostatic processing Therefore, models such as the Weibull, log-normal, and log-
Avocado (Persea americana) and mango (Mangifera indica L.) logistic are used (Klein and Moeschberger 1997). Survival analysis
fruits were obtained from the regions of Michoacan and Oaxaca, methodology requires the estimation of the likelihood function
México, respectively, and transported to Avomex Inc. (Sabinas, (L) based on the acceptability data obtained from consumers. The
Coahuila, Mex., U.S.A.) for high-pressure processing. Briefly, av- likelihood function is a mathematical expression that describes
ocados and mangos were washed, sanitized, peeled, and macerated the probability of obtaining the data observed as a function of
without food additives. Individual avocado and mango pulp sam- estimated parameters of the selected mathematical models (Klein
ples (200 g) were vacuum-packaged into plastic bags impermeable and Moeschberger 1997). The product of each interval obtained
to oxygen (WinpaK Ltd., Winnipeg, Manitoba, Canada). Avo- from each consumer [τ j , τ k ] defines the likelihood function. The
cado and mango pulps packages were commercially processed at contribution of each interval to the likelihood function is given
600 MPa for 3 min at 23 ◦ C, using a 215L ULTRA HHP- by P(τ j < T < τ k ) = S(τ j ) − S(τ k ). Therefore, the complete
processing unit (Avure Technologies, Kent, Wash., U.S.A.). Pro- likelihood is given by Eq. (1).
cessing conditions were closely monitored and recorded to assure
experimental reproducibility. After processing, the samples were   
◦ L= S(Ri ) S(l i ) − S(r i ) (1)
placed in a cold-water bath (1 to 3 C) and transported under
i ∈R i ∈I
refrigerated conditions (4 ◦ C) to the Biotechnology Center main
laboratory at Tecnológico de Monterrey (Monterrey, Nuevo León, where R is the set of right-censored observations (consumers that
México). accepted HHP-treated avocado and mango pulps for 45 and 40 d,
respectively); I is the set of observations censored in an interval.
Storage study In Eq. (1), S(t) is substituted by the expression that corresponds
HHP-processed avocado and mango pulps samples (30 plastic to the survival function for the selected parametric model. The
bags, 200 g each) were stored in the dark at 4 ◦ C for 45 and data obtained from the sensory study were analyzed using the
40 d, respectively. Refrigerated samples were collected (every 5 d program S-PLUS 7.0 (Insightful Corp., Seattle, Wash., U.S.A.).
for avocados and every 4 d for mangoes) and placed in a freezer Regressions were carried out for all parametric models (Weibull,
at −80 ◦ C until they were analyzed by consumers. Since freezing log-normal, log-logistic, extreme, normal, and logistic) and the
at −80 ◦ C occurred quickly, the formation of water crystals was values of μ (intercept) and σ (dispersion) were obtained for the
prevented and therefore nonsignificant organoleptic differences 2 models that showed best fit for the collected data of each fruit
were detected in the products. pulp. Once the models were selected, the survival function was
defined. For instance, when using the log-logistic model for T,

S: Sensory & Food

Consumer sensory analyses the survival function for the log-logistic model is the expression
Fruit pulps from each storage time were assessed by 60 con- given in Eq. (2).

Quality
sumers (for each fruit) recruited among students and employees
from Tecnológico de Monterrey with ages ranging from 18 to ez t − μ
55 y. In the selection process, the consumers were questioned on S(t ) = σ t (1 + e z)2 , where z = σ and t  = ln(t ) (2)
the frequency of consumption of each fruit and those who con-
sumed avocados or mangoes at least once a week were selected
for the sensory evaluations. Consumers generated sensory accept- Finally, the shelf life with their 95% confidence interval was
ability data for samples with different storage times, by answering predicted for each HHP-processed fruit pulp using values for
the question “would you normally consume this product?” with F(t) = 0.25 and 0.50 (representing 25% and 50% of product re-
a “yes” or “no” answer. Prior microbiological analyses confirmed jection probability from the consumers, respectively).
that both fruit pulps were fit for human consumption during the
evaluated storage period. Results and Discussion
As it has been previously described, various mathematical mod-
Statistical analyses els (commonly used in survival analysis) were evaluated for ad-
Because of the nature of the study, each consumer tasted sam- justment to the acceptability data generated by consumers for
ples with different storage times (τ 1 < τ 2 < · · · < τ m ). The HHP-treated avocado and mango pulps (Figure 1 and 2, respec-
information from each consumer was then reduced to an interval tively). Suitability of each model was assessed visually since, to
[τ j , τ k ] that represented the period during which the consumer date, there are no goodness-of-fit tests to handle interval-censored
rejected the product. In other words, a consumer with the above- data (Hough and others 2003; Salvador and others 2006).
mentioned interval accepted the product before τ j and rejected it Parametric models that presented the best fit for avocado pulp
before or at the time τ k . Therefore, the exact time at which the data were the log-logistic and Weibull (Figure 1A and 1E). On
consumer rejects the product is located within τ j and τ k . the other hand, the log-normal and log-logistic models showed
Survival analysis methodology was then used as previously de- the best fit for mango pulp data (Figure 2C and 2E). These math-
scribed (Hough and others 2003; Salvador and others 2006) to ematical models were selected and used in subsequent steps of the
estimate the shelf life of avocado and mango pulps processed by sensory shelf-life determination. The parameter estimates (μ and
HHP using the acceptability data obtained from consumers. In σ ) of the selected models were calculated and are shown in Table 1.
order to estimate the shelf life of a product (P), the methodology Estimated values for μ and σ were then used to obtain prediction

Vol. 75, Nr. 6, 2010 r Journal of Food Science S287

 Survival analysis and HHP-treated fruits . . .

curves for the time-dependent consumer rejection probabilities
of avocado (Figure 3) and mango (Figure 4) pulps treated with
HHP. For both fruit pulps, the mathematical models used resulted
in similar rejection probability curves. As observed in Figure 3,
the probability of rejection of the HHP-processed avocado pulp
after 45 d of storage was almost 90%. On the other hand, HHP-
processed mango pulp almost reached 50% of consumer rejection
at the end of the evaluated storage period of 40 d (Figure 4). The
obtained rejection probability curves for both pulps suitably de-
scribed undesirable changes in sensory perception that took place
during the storage period and decreased consumer acceptability of
the products. From the 60 consumers that evaluated the avocado
pulp, the acceptability data generated by 54 consumers were in-
terval censored. Whereas, 32 of the 60 consumers that evaluated
mango pulp accepted the product with 40 d of storage resulting
in right-censored data.
Solution of the mathematical models (Figure 3 and 4), for av-
ocado and mango, respectively, resulted in the estimation of the
shelf life of both products. In order to predict the shelf life of
each HHP-processed fruit pulp F(t) values of 0.25 and 0.50 were
used (25% and 50% of product rejection probability from the con-
sumers). As previously indicated, the estimated values for μ and
σ from the selected models for avocado and mango are shown
in Table 1. In the case of avocado pulp by solving F(t; μ =
2.84, σ = 0.43) and F(t; μ = 3.21, σ = 0.69), values of t = 17
and 19 d were obtained for the log-logistic and Weibull models,

.999 A weibull B extreme

.9
.5
.2
P ro b a b ility

.1
.05
.02
.01
.003

.0005
.0002
10 20 30 40 10 20 30 40
Failure Time Failure Time
.9999 lognormal normal
C D
.999
.995
.98
S: Sensory & Food

.9
P ro b a b ility

.7
Quality

.5
.3
.1
.02
.005
.0005
.00005
10 20 30 40 10 20 30 40
Failure Time Failure Time
.9999 loglogistic logistic
E F
.9995
.998
.99
.95
P ro b a b ility

.8
.5
.2
.05
.01
.002
.0005
.00005
10 20 30 40 10 20 30 40
Failure Time Failure Time

Figure 1–Mathematical prediction lines for consumer acceptance/rejection data for avocado pulps processed by HHP (600 MPa, 3 min) and stored
refrigerated at 4 ± 1 ◦ C for 45 d. Models evaluated included (A) Weibull, (B) extreme, (C) log-normal, (D) normal, (E) log-logistic, and (F) logistic.

S288 Journal of Food Science r Vol. 75, Nr. 6, 2010

Survival analysis and HHP-treated fruits . . .

respectively. Therefore, results indicated that the estimated shelf Table 1–Values for μ (intercept) and σ (scale) for selected mathe-
life for avocado pulp processed by HHP (600 MPa, 3 min, and matical models used to predict consumer-rejection probabilities
during storage of avocado and mango pulps processed by high
23 ◦ C) for a 50% rejection probability was approximately 17 to hydrostatic pressure (HHP, 600 MPa, 3 min).
19 d (Table 2). The estimated shelf life of avocado pulp using 25%
Parameter estimates
rejection probability was 10 d on both of the parametric models
evaluated. By the same procedure, it was determined that mango HHP-processed Mathematical μ ± standard σ ± standard
product model error error
pulp shelf life was 41 d and 31 d using 50% and 25% consumer
rejection probabilities, respectively (Table 2). Avocado pulpa Log-logistic 2.84 ± 0.14 0.43 ± 0.17
As stated previously, 60 consumers (N) were used to determine Weibull 3.21 ± 0.13 0.69 ± 0.17
the sensory shelf life of HHP-treated avocado and mango pulps Mango pulpb Log-normal 3.74 ± 0.10 0.45 ± 0.15
Log-logistic 3.76 ± 0.10 0.27 ± 0.15
by the mathematical approaches of the survival analysis method-
ology used herein. The acceptability data collected showed a data
a
Storedfor 45 d at 4 ◦ C.
b
Stored for 40 d at 4 ◦ C.
trend that indicated the possibility to mathematically describe the

.999 A weibull B extreme

.9
.5
.2
P ro b a b ility

.1
.05
.02
.01
.003

.0005
.0002
8 9 10 20 30 40 10 15 20 25 30 35 40
Failure Time Failure Time
.9999 C lognormal D normal
.999
.995
.98

S: Sensory & Food

.9
P ro b a b ility

.7

Quality
.5
.3
.1
.02
.005
.0005
.00005
8 9 10 20 30 40 10 15 20 25 30 35 40
Failure Time Failure Time
.9999 E loglogistic F logistic
.9995
.998
.99
.95
P ro b a b ility

.8
.5
.2
.05
.01
.002
.0005
.00005
8 9 10 20 30 40 10 15 20 25 30 35 40
Failure Time Failure Time

Figure 2–Mathematical prediction lines for consumer acceptance/rejection data for mango pulps processed by HHP (600 MPa, 3 min) and stored
refrigerated at 4 ± 1 ◦ C for 40 d. Models evaluated included (A) Weibull, (B) extreme, (C) log-normal, (D) normal, (E) log-logistic, and (F) logistic.

Vol. 75, Nr. 6, 2010 r Journal of Food Science S289

 Survival analysis and HHP-treated fruits . . .
rejection of both food products through time with this approach.
However, a recent study suggested a larger number of consumers
(N = 120) to determine the sensory shelf life of food products such
as minced beef, yogurt, and lettuce when using the survival analy-
sis methodology (Hough and others 2007). Therefore, it would be
relevant for future studies that use survival analysis methodology to
explore the impact of the N value on the shelf-life determination
of foods. Particularly, it would be interesting to validate its effects
on the shelf-life estimation of HHP products, which is the focus
of the present study.
It is well known that avocados are usually consumed as a fresh
fruit product and that thermal processing of avocado pastes can
generate bitter off-flavor (Ben-et and others 1973). The estimated
sensory shelf life of HHP-processed avocado pulp (17 d consider-
ing a 50% rejection probability) demonstrates the effectiveness of
this nonthermal technology on the conservation of this fruit. The
S: Sensory & Food
Quality
S290 Journal of Food Science r Vol. 75, Nr. 6, 2010
use of additives such as resveratrol has been reported to extend the
quality attributes and shelf life of avocado products ( Jiménez and
others 2005). Nevertheless, the use of additives is not always well
accepted by consumers since it can modify avocado flavor.
Other strategies oriented to extend sensory shelf life of mini-
mally processed mangoes are also reported in the scientific liter-
ature. Robles-Sánchez and others (2009) reported that fresh-cut
mangoes dipped on a solution containing ascorbic acid, citric acid,
and calcium chloride (CaCl2 ) allowed the retention of the nutri-
tional quality on the product. However, the authors observed that
after 15 d of storage at 5 ◦ C close to 50% of the panelists in their
study rejected the product. The authors attributed this finding to
negative sensory changes produced during storage of the fresh-cut
mangoes as a result of the dipping treatment. Martı́nez-Ferrer and
others (2002) studied the shelf life minimally processed mangoes
stored under modified atmospheres using a mixture of 10% CO2 ,
Figure 3–Consumer rejection probability curve
of avocado pulp processed by HHP (600 MPa,
3 min) as a function of storage time. The
mathematical models selected to obtain the
curves were the log-logistic (- - -) and Weibull
(—). Rejection probability curves were obtained
using the software S-PLUS 7.0 (Insightful Corp.,
Seattle, Wash., U.S.A.).
Figure 4–Consumer rejection probability curve
of mango pulp processed by HHP (600 MPa,
3 min) as a function of storage time. The
mathematical models selected to obtain the
curves were the log-normal (- - -) and log-logistic
(—). Rejection probability curves were obtained
using the software S-PLUS 7.0 (Insightful Corp.,
Seattle, Wash., U.S.A.).
Survival analysis and HHP-treated fruits . . .
Table 2–Estimated sensory shelf life (d) of avocado and mango pulps processed by HHP (600 MPa, 3 min).
HHP-processed product Mathematical model
Avocado pulpa Log-logistic
Weibull
Mango pulpb Log-normal
Log-logistic
a
Storedfor 45 d at 4 ◦ C.
b
Stored for 40 d at 4 ◦ C.
4% O2 , and 86% N2 . The authors reported that the minimally
processed mango product maintained good color, texture, odor,
and taste after 25 d of storage at 5 ◦ C. The shelf-life estimated
herein for HHP-treated mango pulps (30 d considering 25% re-
jection) is significantly higher than previously reported studies,
demonstrating the effectiveness of HHP for the conservation of
the sensory quality of minimally processed mango pulp.
Conclusions
Survival analysis allowed the determination of the sensory shelf
life of HHP-treated avocado and mango pulps. The estimated sen-
sory shelf life of HHP-processed avocado pulp (17 d considering
a 50% rejection probability) demonstrated the effectiveness of this
nonthermal technology on the conservation of pulp acceptabil-
ity during its refrigerated storage. Likewise, the sensory shelf life
estimated for HHP-treated mangoes was extended in compari-
son with previous reports for nonpressure treated precut mangoes
(>45 d compared with 15 d, respectively), demonstrating the ef-
fectiveness of HHP for the enhancement of their refrigerated shelf
life. Survival analysis methodology has potential application for
other pressurized products. A significant aspect to remark is the
simplicity of the sensory approach used in the study in comparison
with other sensory shelf-life estimation procedures. The “yes” or
“no” answers, obtained from consumers, to the question about
their willingness to consume the stored samples generated enough
data to model the probabilities of consumer rejection for each
product. Shelf-life estimations were made from the mathematical
models showing good adjustment to the consumer acceptability
data. The methodology can also be applied to study the effects of
changes in processing parameters or product formulations on the
sensory shelf life of microbiologically stable foods.
Acknowledgments
This research was possible thanks to the research funds from the
Tecnologico de Monterrey-Research Chair Initiative (CAT-05)
Sensory shelf life (d) ± confidence interval (95%)
25% rejection probability 50% rejection probability
10 ± 6 17 ± 5
10 ± 6 19 ± 6
31 ± 5 41 ± 5
32 ± 7 42 ± 7
and from the joint Research Program of The Univ. of Califor-
nia Inst. for Mexico and the United States (UC MEXUS) and
the Mexican Natl. Council for Science and Technology (CONA-
CYT). We also would like to thank Fresherized Foods, A Div.
of Avomex Inc. for providing avocados, mangoes, and the use of
HHP equipment and for their continuous commitment to the
advancement of nonthermal processing technologies.
References
Ben-et G, Dolev A, Tatarsky D. 1973. Compounds contributing to heat-induced bitter off-flavor
in avocado. J Food Sci 38:546–7.
Gámbaro A, Fiszman S, Giménez A, Varela P, Salvador A. 2004a. Consumer acceptability com-
pared with sensory and instrumental measures of with pan bread: sensory shelf-life estimation
by survival analysis. J Food Sci 69(9):S401–5.
Gámbaro A, Giménez A, Varela P, Garitta L, Hough G. 2004b. Sensory shelf-life estimation of
a Alfajor by survival analysis. J Sens Stud 19(6):500–9.
Gámbaro A, Gastón A, Giménez A. 2006. Shelf-life estimation of apple-baby food. J Sens Stud
21(1):101–11.
Guerrero-Beltrán JA, Barbosa-Cánovas GV, Moraga-Ballesteros G, Moraga-Ballesteros MJ,
Swanson BG. 2006. Effect of pH and ascorbic acid on high hydrostatic pressure-processed
mango puree. J Food Process Pres 30(5):582–96.
Hough G, Langohr K, Gómez G, Curia A. 2003. Survival analysis applied to sensory shelf life
of foods. J Food Sci 68(1):359–62.
Hough G, Calle ML, Serrat C, Curia A. 2007. Number of consumers necessary for shelf life
estimations based on survival analysis statistics. Food Qual Prefer 18(5):771–5.
Jacobo-Velázquez DA, Hernández-Brenes C. 2010. Biochemical changes during the storage
of high hydrostatic pressure processed avocado paste. J Food Sci. DOI:10.1111/j.1750-
3841.2010.01654.x.
Jiménez JB, Orea JM, Montero C, González-Ureña A, Navas E, Slowing K, Gómez-Serranillos
MP, Carretero E, De Martinis D. 2005. Resveratrol treatment controls microbial flora, pro-
S: Sensory & Food
longs shelf life, and preserves nutritional quality of fruit. J Agric Food Chem 53(5):1526–
30.
Klein JP, Moeschberger ML. 1997. Survival analysis, techniques for censored and truncated data.
Quality
New York: Springer-Verlag. 502 p.
López-Malo A, Palou E, Barbosa-Cánovas GV, Welti-Chanes J, Swanson BG. 1998. Polyphe-
noloxidase activity and color changes during storage of high hydrostatic pressure treated
avocado puree. Food Res Int 31(8):549–56.
Martı́nez-Ferrer M, Harper C, Pérez-Muñoz F, Chaparro M. 2002. Modified atmosphere
packaging of minimally processed mango and pineapple fruits. J Food Sci 67(9):3365–71.
Ramos-Parra PA. 2006. Factores que afectan la estabilidad de mango (Mangifera indica L.) proce-
sado por alta presión hidrostática [Master of Science Thesis]. Tecnológico de Monterrey.
Robles-Sánchez RM, Islas-Osuna MA, Astiazarán-Garcı́a H, Vázquez-Ortiz FA, Martı́n-Belloso
O, Gorinstein S, González-Aguilar GA. 2009. Quality index, consumer acceptability, bioactive
compounds, and antioxidant activity of fresh-cut “Ataulfo” mangoes (Mangifera Indica L.) as
affected by low-temperature storage. J Food Sci 74(3):S126–34.
Salvador A, Varela P, Fiszman SM, Gómez G. 2006. Estimating the shelf life of brown pan bread,
suitability of survival analysis methodology. J Food Sci 71(4):S321–5.
Vol. 75, Nr. 6, 2010 r Journal of Food Science S291