You are on page 1of 9

Journal of Food Engineering 289 (2021) 110182

Contents lists available at ScienceDirect

Journal of Food Engineering


journal homepage: http://www.elsevier.com/locate/jfoodeng

Development and optimization of beer containing malted and non-malted


substitutes using quality by design (QbD) approach
~o a, Joa
Simone Tessarini Esteva ~o Batista de Almeida e Silva b, Felipe Rebello Lourenço c, *
a
Departamento de Química Fundamental, Instituto de Química, Universidade de S~ ao Paulo, Av. Prof. Lineu Prestes, 748 – Bloco 1, CEP: 05508-000, S~ ao Paulo, SP,
Brazil
b
Departamento de Biotecnologia. Escola de Engenharia de Lorena, Universidade de S~ ao Paulo, Estrada Municipal do Campinho, s/n – Planta Piloto de Bebidas, CEP:
12602-810, Lorena, SP, Brazil
c
Departamento de Farm�acia, Faculdade de Ci^encias Farmac^euticas, Universidade de S~
ao Paulo, Av. Prof. Lineu Prestes, 580 – Bloco 15, CEP: 05508-000, S~
ao Paulo, SP,
Brazil

A R T I C L E I N F O A B S T R A C T

Keywords: Quality by Design (QbD) approach has been widely used in the pharmaceutical field, but its application in the
Beer development of food products is recent. The objective of this work was to apply QbD in the development of Ale-
Barley malt substitute type beer, with up to 45% of barley malt substitutes, with alcohol content of 5.50 � 0.55% v/v and a good
Mixture design
sensory evaluation. A simplex-centroid mixture design was used, in which five cereals were tested: maize, oat
Quality by design
malt, rice, rye malt, and sorghum. In the screening design, rice, maize, and oat malt were selected. The opti­
mization design allowed the definition of the design space to meet critical quality attributes. The standard (100%
barley malt) and optimized (55% barley malt, 42% oat malt, and 3% maize) formulations had similar physi­
cochemical, fermentative, and sensory parameters. The implementation of QbD helps improve quality and reduce
quality deviations.

1. Introduction (QTPP), which identifies the critical quality attributes (CQAs) of a


product; product design and understanding, including identification of
Quality by Design (QbD) approach is more than 25 years old and is critical material attributes (CMAs); process design and understanding,
relatively well established in industries whose products must comply including identifying critical process parameters (CPPs), critical mate­
with quality, safety, and performance or efficacy requirements. The rial attributes, and critical process parameters of critical quality attri­
notion of quality as a science was born in the 1980s and 1990s with the butes; a control strategy that includes specifications for product
work of Juran and Deming and has a relationship with the components and controls for each step of the manufacturing process;
manufacturing processes created by two major global companies: process capability; and continuous improvement. QbD tools and studies
Motorola’s Six Sigma and Toyota Lean Manufacturing (Menezes; Flo ^res; include prior knowledge, risk assessment, mechanistic models, data
Gouveia, 2014). analysis, process analytical technology (PAT), and design of experiments
In the pharmaceutical industry, the notions of QbD first appeared in (DoE).
the 21st century and in documents from the Food and Drug Adminis­ Traditionally, the development and optimization of food products
tration (FDA) from the International Council for The Harmonisation of were carried out by analyzing one factor at a time (univariate approach)
Technical Requirements for product registration Pharmacists for Human (Politis, 2017; Candioti et al., 2014). One of the factors is changed
Use (ICH), which provides recommendations and guidelines on the within an appropriate range or level, while the others are kept constant.
subject (Food and Drug Administration, 2009). In the food area, both in In addition to requiring a large number of experiments, the univariate
industries and research, these methodologies are not yet well approach cannot evaluate the existence of interaction between factors,
disseminated. which can lead to inadequate development and optimization (Politis,
Because it is a systematic scientific-based approach, according to Yu 2017). DoE is a tool in which experiments are planned according to a
(2014), QbD includes some elements: a quality target product profile pre-established matrix, interactions between factors are evaluated, and

* Corresponding author.
E-mail address: feliperl@usp.br (F.R. Lourenço).

https://doi.org/10.1016/j.jfoodeng.2020.110182
Received 7 January 2020; Received in revised form 25 April 2020; Accepted 1 June 2020
Available online 8 June 2020
0260-8774/© 2020 Elsevier Ltd. All rights reserved.
S.T. Estev~
ao et al. Journal of Food Engineering 289 (2021) 110182

the entire set of experiments can provide more information using a characteristics of beers. QbD approached allowed definition of an opti­
smaller number of experiments compared to the univariate approach mized formulation that complies with the parameters established in the
(Leng et al., 2017). QTPP.
DoE is designed from a structured and organized method to deter­
mine the relationships between input factors (xi - independent variables) 2. Material and methods
that affect one or more output responses (y - dependent variables), by
establishing mathematical models (y ¼ f (xi)). In the DoE approach, the To apply Quality by design (QbD) in the development of beer, the
input factors studied are systematically varied to determine their effects quality target product profile (QTPP) was established. The critical
on output responses, which allows determination of the most important quality attributes (CQAs), the critical materials attributes (CMAs), and
input factors, identification of the configuration of input factors leading the critical parameters of the process (CPPs) were identified. In addition,
to optimized output responses, and the elucidation of interactions be­ screening and optimization designs of experiments (DoE) were
tween input factors. To provide better understanding, experimental performed.
projects can be divided into two types: screening designs and optimi­ A simplex-centroid mixture design was adopted. The varied and
zation designs (Fukuda et al., 2018; Politis, 2017; Candioti et al., 2014). studied components were the proportions of the brewers: maize, oat
To design and develop a robust food product that has the desired malt, rice, rye malt, and sorghum. Mixture design occurred in two steps:
QTPP, development should consider the physical, chemical, biological, the screening design (SD) and optimization design (OD).
and sensory properties of food. The experimental design can be con­
structed through factorial designs or, depending on the type of food and 2.1. Screening design
the objectives established, through a mixture design.
When mixed variables are involved in an optimization, the result The control consisted of a formulation with 100% barley malt with
depends on the proportion of these components, and their levels cannot four replicas: F0I, F0II, F0III, and F0IV. In the other formulations, barley
be varied without taking into account the other components. Thus, malt was fixed at a concentration of 55% and the substitutes were varied
mixtures are systems whose properties depend on the relative pro­ in proportions between zero and 45%. The concentrations of hops and
portions of their components and not their concentrations. There are no yeasts were fixed in all formulations.
restrictions on physical state, so components can be solid, liquid, or The central point formulation (F1) was composed a mixture of barley
gaseous, provided that system properties are defined by their proportion malt (55%), sorghum (9%), rye malt (9%), maize (9%), rice (9%), or oat
in the mixture. Thus, the amount of each component of the system malt (9%). At the vertex points, 45% of the barley malt was substitute by
should be treated as a variable (mixing variable), which is not inde­ sorghum (F2), rye malt (F3), maize (F4), rice (F5), or oat malt (F6).
pendent of the others, because the sum of the proportions of its com­
ponents must be constant (Cornell, 1990). 2.2. Optimization design
Thus, mixture design consists of adjusting a polynomial mathemat­
ical model to a response surface obtained according to a specific The three barley malt substitutes selected in SD were submitted to
experimental design, known as mixture design. This name is used to the optimization design (OD) in which a new mixture design was per­
differentiate it from the factorial design employed in modeling process formed. The formulations of the vertex points of the selected substitutes
variables. The mixture design to be employed will depend on the had been performed in the SD (F4, F5, and F6). The double mixture
complexity of the mathematical model that is desired to adjust and the formulations (F7, F8, and F9) were manufactured with two substitutes
number of components of the mixture (Coscione et al. 2005). ranging in a ratio of 22.5%. The formulations of axial points (F10, F11,
Simplex-centroid design is a type of mixture design which employs and F12) were composed of one of the substitutes in the proportion of
binary and tertiary mixture of the components. The number of points of 30% and the others at 7.5%. A central point (F13) was also held in which
this project depends on the number of components of the mixture each substitute was used together in the proportion of 15%, which was
(Nunes, 1998). Although the simplex-centroid design is a simple appli­ manufactured in four replicas (F13I, F13II, F13III, and F13IV). Fig. 1 il­
cation, it cannot test for lack of fit of the model (Cornell, 1973). This lustrates the distribution of each formulation in the study space.
issue can be solved by including replicas of the central point (Cornell,
1990). 2.3. Manufacturing of formulations
According to Yu (2014), mixture designs fulfilling the role of DoE can
help identify ideal conditions, CMAs, CPPs, and finally design space. The manufacturing of each formulation was performed according to
Thus, the main objectives of applying the QbD approach in the food area the following steps: mashing (barley malt is added to water and starch is
are to develop a robust product that can deliver the target quality profile saccharified in smaller sugars by the enzymes present in malt); boiling
of the desired product during its lifetime, increase the capacity of the for 1 h (addition of hops in the brewer must for the development of
process, reduce the variability of the food manufacturing process, and bitterness and aromas); inoculation of the yeast Saccharomyces cerevisiae
incorporate quality into the product. In this work, QbD was applied to US-05 (Fermentis, France); fermentation at 20 � 1 � C for six days;
develop beer formulations produced with malted and non-malted maturation at 2 � 1 � C for thirty days; carbonation at 20 � 1 � C for seven
cereals. days; and filling steps. The formulations were not filtered.
Beer is the drink resulting from fermentation, by the brewing yeast, The barley malt substitutes generally entered production along with
of a malt barley or malt extract previously submitted to an additional barley malt. Rice and sorghum, which are difficult to saccharification,
cooking process of hops or hop extract. Part of malted barley or malt underwent a pre-preparation stage, which required cooking each cereal
extract may be replaced by brewer substitute (Brasil, 2019). In Brazil, soaked in mineral water (1:4) until the absorption of all water. In the
the use of substitutes may not exceed forty-five percent of the primitive case of sorghum, after cooking, the mixture was cooled to between 60
extract (Brasil, 2019; Brasil, 2009; Brasil, 2001). Barley malt substitutes and 70 � C, the enzyme thermostable alpha-amylase Brautec-α (Prozyn
are considered other cereals suitable for human consumption, malt or Biosolution, Brazil) was added to 0.3%, and the mixture kept at this
not, as well as starches and sugars. temperature range for about an hour. Maize syrup did not require any
With the premise of scientifically comparing beer produced with pre-preparation and was added to the formulation during boiling step.
barley malt and its substitutes, the objective of this work was to apply Beer formulations were manufactured at the Beverage Pilot Plant of
the QbD methodology to beer production. Mixture design of experi­ the Escola de Engenharia de Lorena at the Universidade de Sa ~o Paulo
ments were adopted aiming to observe the effects of the presence of non- (EEL-USP) and in the Faculdade de Ci^ encias Farmac^euticas at the Uni­
malted and malted cereal substitutes on physicochemical and sensory versidade de Sa ~o Paulo (FCF-USP) according to Good Manufacturing

2
S.T. Estev~
ao et al. Journal of Food Engineering 289 (2021) 110182

2.6. Fermentative and kinetic parameters and sensory analysis used to


characterize the optimized formulation and compare with control
formulation

Inferences about the proportions of the barley malt substitutes were


addressed by the regression model. Response surface analyses were
performed to obtain the optimized formulation (F14). Optimized (F14)
and control (F0) formulations were submitted to the same character­
ization analyses performed for the beers of the OD. In addition,
fermentative and kinetic parameters were calculated, such as extract
masses, the conversion factor of extract into alcohol, the real and
apparent degrees of attenuation, as well as the efficiency of the con­
version of extract into alcohol. The formulations were also submitted to
a 5-point hedonic scale purchase intention test, taking in their extremes
the terms “certainly buy” (5) and “certainly wouldn’t buy” (1). The
purchase intention test was performed by 109 tasters with a minimum
age of 18 years old.

2.7. Statistical analyses

The results of mixture design were subject to analysis of variance


(ANOVA) and regression analysis. The results of characterization ana­
lyses and sensory analysis were analyzed by the ANOVA tests and Tukey
tests, at 5% significance. All statistical analyses were performed using
Fig. 1. Mixture design of experiments that defines the proportions of maize, Minitab™ 18 software (Minitab Inc., USA). Principal component anal­
rice, and oat malt as barley malt substitutes in formulations F4 to F13. ysis (PCA) was also performed using Minitab™ 18. The charts and tables
were prepared in Microsoft Office Excel 2010), Microsoft Office Word
Practices (GMP) for food described in RDC 216 (Brasil, 2004) and in 2010, Minitab™ 18, and GraphPad Prisma 6.
random order to obtain statistical independence from errors.
3. Results and discussion
2.4. Physicochemical analyses
Implementation of QbD approach was initiated by defining the
The analyzed responses were alcohol content (% v/v), density (g/ quality target product profile (QTPP). This definition contributes to the
cm3), original extract (% w/w), real extract (% w/w), apparent extract development of the product design and concept. The QTPP defined was
(% w/w), calories (kcal/100 mL), real degree of fermentation (%, RDF an Ale beer, with up to 45% of barley malt substitutes, with an alcohol
¼ real extract – original extract/original extract), and apparent degree content of 5.50 � 0.55% v/v and well evaluated in sensory analysis.
of fermentation (%, ADF ¼ apparent extract – original extract/original Sensory analysis was measured by means of the hedonic scale, in which
extract), in triplicate, were obtained using a beer analyzer equipment a well-evaluated beer received, at least 6 out of 9 possible points (which
(Alex 500 model, Anton Paar, Brazil). Based on the values of original, means “liked slightly”) for each of the following attributes: color, aroma,
real, and apparent extracts, and according to Medeiros et al. (2018), real and flavor (totaling, at least, 18 points from a total of 27 possible points).
and apparent degrees of attenuation were calculated. Color analyses, The values of alcohol content and sensory analysis that comprise the
according to EBC (2005), and pH analyses were also performed, in QTPP were defined a priori, before manufacturing the beer formulations.
triplicate, for all formulations. Additionally, the efficiency of the The reactions that lead to the development of color also contribute to
mashing step process (%) was calculated for the formulations of the SD, the formation of aromas and flavors. The pH of the must is related to the
based on the masses of components used and the final amount of ob­ transformation of hops during boiling, affecting the extraction of alpha
tained for each formulation. acids (Olec, 2019) and consequently the production of iso-α-acids,
which is related to the bitterness of the final product, a parameter that
2.5. Sensory analysis can contribute to acceptance or rejection of the taste. For these reasons,
color and pH were defined as critical quality attributes (CQAs) directly
Sensory analysis was based on affective tests, using a hedonic scale of related to the sensory part of the QTPP. Thus, the defined CQAs were the
9 points, adapted according to the Likert scale (Likert, 1932), in which results of sensory analysis, alcohol content, color (measured according
the tasters (untrained, but consume beer products regularly) were able to EBC), and pH analyses.
to choose the option that best expressed their opinion about the samples The critical materials attributes (CMAs) were the proportions of
for the attributes: color, aroma, and flavor. The scale had at its extremes barley malt and its substitutes in the final products. The optimized
the terms “I really liked” (9) and “I really disliked” (1). The results of the proportions of barley and its substitutes was defined using mixture
sensory analysis were calculated in a total sensory score, the possible design of experiment. Moreover, time and temperature of mashing,
interval to be obtained in each formulation is 3–27 points. The sensory fermentation, maturation, and carbonation stages are potentially critical
analysis was performed by 109 tasters with a minimum age of 18 years, process parameters (CPPs). Because these parameters were kept con­
who provided a grade from 1 to 9 for color, aroma, and flavor for each stant for all formulations, we assumed that no CCPs were identified and
tasted formulation. The sensory responses of each formulation were studied in the work.
compared using a repeated ANOVA measures. Sensory evaluations were The efficiency of the mashing process in the screening design (SD)
approved by the Research Ethics Committee of the Faculdade de Ci^ en­ was not defined as a critical attribute but as a performance parameter,
cias Farmac^ euticas da Universidade de Sa ~o Paulo. which was evaluated because it was related to the proper control of
CPPs.

3
S.T. Estev~
ao et al. Journal of Food Engineering 289 (2021) 110182

3.1. Screening design

Product development according to QbD approach, adopting the


screening design (SD), provided the responses presented in Table 1.
The efficiency of the mashing process was determinant for the choice
of the first barley malt substitute selected in the screening step. The
calculation of the efficiency of the mashing process considered the pre-
preparation stage of rice and sorghum, which justifies that the formu­
lations containing these barley malt substitutes (F5 and F2, respectively)
were the least efficient. On the other hand, the formulations containing
maize as barley malt substitute (F4 and F1) presented the greatest effi­
ciency, since the addition of maize is in the form of syrup during the
boiling stage of the manufacturing process, dispensing with the need for
Fig. 2. Principal component analysis for the screening design (SD) for the
saccharification. Thus, maize was the first substitute selected for the physicochemical characteristics (color, pH, density, alcohol content, calories,
optimization design. AE – apparent extract, RE – real extract, OE – original extract, ADF – apparent
The results of the sensory analysis in Table 1, in the form of a total degree of fermentation, and RDF – real degree of fermentation) of formulations
sensory note, indicate that F6 was the formulation that presented the F1 to F6 (n ¼ 3 for each formulation). (For interpretation of the references to
best sensory evaluation, and for this reason, oat malt was the second color in this figure legend, the reader is referred to the Web version of
barley malt substitute selected in the SD. this article.)
The selection of the third substitute was through the principal
component analysis (PCA), in which the various variables are trans­ 3.2. Optimization design
formed into a few new components that are able to explain much of the
variability, according to Fig. 2. The PCA can explain, in total, 89.0% of The formulations of the optimization design (OD) were manufac­
variability with only two components (67.2% explained by PC1 and tured and the CQAs analyzed. According to Rosa and Afonso (2015), the
21.8% by PC2). pH in the brewing process acts directly in the regulation of enzymatic
The formulations at the right side of the chart (CP1 < 0) have higher activity. The enzymes involved in the saccharification of the must,
alcohol content, have greater real and apparent degrees of fermentation, especially amylases, can act at different levels of pH, but the optimal pH
are less dense, and have lower apparent extract in relation to the for­ range for the activity of these enzymes is between 5.2 and 5.5 (Gon­
mulations at the left side (CP1 > 0). Thus, PC1 was strongly related to çalves, 2006). The pH is also related, also according to Rosa and Afonso
the fermentation process. The formulations at the lower portion of the (2015), with the solubilization of astringent components, with color
chart (CP2 < 0) have lighter color (EBC) and higher values than the real variation and coagulation of the protein components of the must. The pH
and original extracts, besides being more caloric than the formulations analyses results are in Fig. 3.
at the upper portion (CP2 > 0). Thus, PC2 was strongly related to the Several statistically significant differences can be observed between
composition of the formulations. the final pH of the formulations. The definition of a pH range intended
The biplot chart obtained from PCA indicates the formulation with for the final product of each formulation was not one of the aspects of
45% rice (F5) is the one that appears closest to the formulation with 9% quality target product profile because, according to Briggs et al. (2004),
of each substitute tested in screening (F1). In other words, F5 presented the fermentation process tends to establish an adequate pH, as it usually
similar characteristics to F1, which demonstrates that rice was the falls with the secretion of carbonic acid.
determining substitute for the characteristics of the mix. Thus, it was Brazilian legislation does not define an ideal pH range for ready-
determined that rice would be the third cereal selected in the SD to made beer, but according to Lindstrom and Hansen (2014), beer usu­
compose as a substitute the OD formulations. ally has pH between 4.1 and 4.5. According to Rosa and Afonso (2015),
The results of all analyses of the SD formulations allowed the choice the pH of beer should be around 4.0, and the pH of 4.5 is considered very
of the three barley malt substitutes that best met the established QTPP, high, which increases the biological susceptibility of the product. Thus,
namely: maize, rice, and oat malt. These substitutes were adopted in the pH presented in the formulations of the optimization design (be­
optimization design, considering the high efficiency of the mashing tween 4.05 and 4.16) were adequate.
process, the sensory analysis results, and the similar characteristics of The results of the alcohol content allowed the elaboration of the
the formulation as the control formulation (100% barley malt formu­ graphs presented in Fig. 4.
lation F0). The formulations F8 and F10 presented a statistically significant
difference in alcohol content. Since F8 and F10 have identical

Table 1
Physicochemical responses, efficiency values of the mashing and sensory analysis results (n ¼ 3) for the formulations of the screening design (SD).
Variables F0I F1 F2 F3 F4 F5 F6

Alcohol (% v/v) 5.72 6.06 5.91 5.11 6.14 6.04 5.20


Density (g/cm3) 1.0061 1.0068 1.0047 1.0091 1.0055 1.0070 1.0083
Real Extract (% w/w) 4.1 4.4 3.8 4.6 4.1 4.5 4.5
Apparent Extract (% w/w) 2.0 2.3 1.7 2.8 1.9 2.3 2.6
Original Extract (% w/w) 12.7 13.5 12.8 12.4 13.4 13.5 12.3
Calories (Kcal) 46.0 49.0 46.0 44.6 48.3 49.0 44.5
RDF (%) 69.5 69.3 71.7 63.7 71.1 69.0 65.3
ADF (%) 84.2 83.9 86.9 77.1 86.1 83.4 79.0
pH 4.21 4.30 4.04 3.99 3.84 4.07 4.30
Color (EBC) 6.1 4.5 5.4 6.6 3.4 3.1 5.6
Efficiency (%) 74.54 80.09 63.98 75.54 89.36 72.04 78.53
Total Sensory 18.43 17.43 18.59 20.00 19.00 20.29 21.71

Legend: F0I: control; F1: mix; F2: 45% sorghum; F3: 45% rye malt; F4: 45% maize; F5: 45% rice; and F6: 45% oat malt. RDF ¼ real degree of fermentation and ADF ¼
apparent degree of fermentation.

4
S.T. Estev~
ao et al. Journal of Food Engineering 289 (2021) 110182

the EBC scale. This may be due to the use of only malted cereal, since the
malt color is mainly affected by toasting (Maillard reactions between
sugars and amino acids) and caramelization (sugar decomposition)
during malting process. The oat malt was the only one of the three
substitute that malted and it also contributed to an increase in color level
on the EBC scale in the formulations in which it was used in greater
proportion, such as in F8 and F12; the only formulation composed only
of non-malted substitutes, F7, was the one with the lightest color.
The F8 and F9 formulations that were manufactured with 22.5% oat
malt are statistically different, because the development of color does
not depend solely on the malting of ingredients. F8 was also composed of
22.5% maize and F9 by 22.5% rice. The difference in the development of
the color of both occurs by the interactions between the different sub­
stitutes, since the types of sugars formed during the saccharification of
the must is one of the factors that influence the Maillard reaction that
occurs during the boiling phase. Another factor that can influence the
Fig. 3. Results of pH values (n ¼ 3) for the formulations F0 and F7 to F13 Maillard reaction is the pH of the must. The higher pH during
manufactured as part of the optimization design (OD). manufacturing process, the greater the color of the formulation (Nas­
cimento et al., 2015). However, this interactive effect between the two
compositions, the difference in alcohol content could be due to the CQAs (pH and color) was not relevant when the pH values were within
random variability of the temperature during the fermentation process. the specifications for all tested formulations. Thus, the color of the inal
According to the mixture contour chart, formulations with higher pro­ product was mainly affected by the amount of barley malt and its
portions of maize in their composition contained higher alcohol content, substitutes.
while those with higher concentrations of oat malt were those with the Sensory analyses were also performed in the OD. These results are
alcohol content close to the intended target values (5.50 � 0.55% v/v). showed in Fig. 6.
The color analyses results (EBC) of the formulations are illustrated in There is no statistically significant difference between formulations
Fig. 5. (p-value > 0.05), but the presence of maize in the composition in the
The control formulation (F0) had with the highest level of color in formulations was preferred by the tasters. F10, with 30% maize,

Fig. 4. Results of alcohol content (n ¼ 3) and response surface counter plot for the formulations F0 and F7 to F13 manufactured as part of the optimization
design (OD).

Fig. 5. Results of color (EBC) (n ¼ 3) and response surface counter plot for the formulations F0 and F7 to F13 manufactured as part of the optimization design (OD).

5
S.T. Estev~
ao et al. Journal of Food Engineering 289 (2021) 110182

Fig. 6. Results of sensorial analysis (n ¼ 3) and response surface counter plot for the formulations F0 and F7 to F13 manufactured as part of the optimization
design (OD).

received the best evaluation, and F9, without maize in its composition,
had the worst. The good evaluation in the SD formulation with only oat
malt as barley malt substitute (F6) did not remain in the results of the
OD. The lack of statistical significance may be explained due to high
variability of sensory responses among the tasters, particularly for
screening and optimization designs, which required the tasters to test a
large number of formulations (up to 13 formulations). Despite the lack
of statistical significance, several formulations provided sensory re­
sponses above the minimal value defined in QTPP (minimal of 18 points
out of 27 possible points).
The mixture contour plot for the sensory response confirms that
maize formulations were better evaluated by tasters, as well as formu­
lations with higher proportions of rice, the two non-malted substitutes
used in OD.
Each mixture contour plot contributed suggestions. (a) The mixture
contour plot for alcohol content (Fig. 4) suggests that for the optimized
formulation to achieve the alcohol content close to the intended, which
is 5.50 � 0.55% v/v, it should be manufactured with a high proportion
of oat malt and with low proportions of the other substitutes. (b) The
mixture contour plot for color (Fig. 5) suggests the use of malt substitute
for a greater level of color on the EBC scale and use of non-malted to
lower level. This information is relevant to demonstrate that color Fig. 7. Overlapped counter plot obtained for the surface count plot of alcohol
attribute can be measured an related as part of the QTPP. (c) The content (blue lines), color (red lines), and sensory analysis (green lines) that
mixture contour plot for sensory analysis (Fig. 6) suggests that for an define the design space region (white region) and optimized formulation (F14).
(For interpretation of the references to color in this figure legend, the reader is
optimal sensory score (of 24 total points or more, including at least 6
referred to the Web version of this article.)
points in each category – color, aroma, and flavor), the optimized
formulation should contain higher concentrations of non-malted
substitutes. content (according to the contour plot of Fig. 4) resulted in a design
All the results obtained with the manufacture of the SD and OD space (white region) presented in Fig. 7. Formulations adopted within
allowed the construction of the overlapped contour plot (Fig. 7) for the the design space region ensure the requirements defined in QTPP.
development of the final optimized formulation (F14), in which the Alternatively, optimized formulation can be defined using a mathe­
results of the alcoholic content, color, and sensory analysis are simul­ matical method, based on desirability functions. The combined desir­
taneously optimized. The pH determination was not considered during ability value (D) was calculated by the square root of the product of the
the optimization step because all tested formulations provided pH values individual desirability values of alcohol content (dAlcohol) and sensory
within the specifications defined in QTPP. The optimization parameters analysis (dSensorial). D value ranges from 0 to 1, and the closer to 1, the
consider alcohol content values between 5 and 6% v/v (as defined in the better the results complies with the CQAs. If one of the parameters do
QTPP), color on the EBC scale of 6 (range between 5 and 7), and sensory not complies with CQAs requirements (d ¼ 0), then D ¼ 0, that is, this
analysis with a total evaluation greater than 18 (as defined in the combination is outside the design space region. The combined desir­
QQTP). Although the color was not considered in the QTPP, the color ability value (D) for the optimized formulation was found to be 0.55,
range around 6 was adopted from the mean results obtained with the OD which indicates that all CQAs were simultaneously achieved.
formulations. Although the design space region considers the use of rice as barley
Fig. 7 establishes the design space region, which defines the pro­ malt substitute, it was not included in the optimized formulation, since it
portions of barley malt substitutes that the optimized formulation would require the pre-preparation stage, without presenting significant
should have to comply with the CQAs. Although the formulations con­ improvement in compliance with the CQAs.
taining maize and rice are promising for sensory analysis (according to The design space region (Fig. 7) and the combined desirability
the contour plot of Fig. 6), the overlap of this parameter with the alcohol function helped to define the final optimized formulation (F14), which
contains 55% barley malt, 42% oat malt, and 3% maize.

6
S.T. Estev~
ao et al. Journal of Food Engineering 289 (2021) 110182

In addition to the composition of final formulation, the quality and and real extract were already constants.
compositional parameters of raw materials may affect the final charac­ In both formulations, the alcohol and real extract contents, which are
teristics of the products. All raw materials used in this work fulfilled the inversely proportional, were constants after about 48 h of fermentation.
quality and compositional parameters defined in literature. However, At the end of the process, the degree of attenuation could be measured.
these issues were not considered in this work because all formulations Attenuation, according to Costa (2017), is the measurement of how
were manufactured using the same raw materials (varying only the complete the fermentation of the must was, that is, how much of the
proportions of the raw materials in the final formulation). sugars present in the extract were consumed during fermentation. Re­
sidual sugars are present in beer as part of the complex set that generates
the characteristic flavor and aroma of this product. The real (ATR) and
3.3. Characterization of optimized formulation and comparison with apparent (ATA) degrees of attenuation were calculated based on the
control formulation results of the original, real, and apparent extracts. These results have the
same meaning as the real (RDF) and apparent (ADF) degrees of
Yeasts are the agents of alcoholic fermentation, and the substrate is fermentation. The difference lies in the fact that the RDF and ADF are
the extract obtained in the previous stages of the manufacturing process. responses from the beer analyzer of samples only from the end of
This extract is classified as original extract (EO). For growth and fermentation. Despite the greater accuracy that the results of ATR and
reproduction, yeasts preferably use sugars, to the detriment of other ATA offer by analyzing the beginning and end of the fermentation pro­
carbon sources present in the original extract. According to Bento and cess, throughout this experimental project, the RDF was used as the
Almeida (2006), the formation and concentration of products during fermentation evaluation parameter. These and other fermentative pa­
fermentation depend mainly on the metabolic routes that the yeast uses. rameters, which could be measured or calculated, are in Table 2.
The product formed after fermentation is also called extract. If the During fermentation, yeasts grow and reproduce forming a yeast
alcohol content is considered, this extract is classified as apparent cream that was separated from beer so that the beer can enter the
extract (As). On the other hand, if the measure does not suffer alcohol maturation process; thus, the volume available for maturation is lower
influence this extract is called real extract (ER). The fermentation than that of the beginning of fermentation.
response variables monitored over time were the alcohol content and the For a laboratory scale process, the efficiency of 93.4% for F0IV and
real extract. The fermentation kinetics of the formulations were 87.8% for F14 was considered good results. Comparing with values in
analyzed approximately every 12 h for 5 days, according to Fig. 8. The the literature, the real degrees of fermentation values of 64.1% (F0IV)
initial time showed in the real extract curve refers to the original extract. and 63.0% (F14) were higher than those obtained by Andrade (2007),
The fermentation stage was halted when the results of alcohol content which was 61.7%, and Curi (2006), which were 59.7% and 62.5%. The
real degrees of attenuation values of 65.6% (F0IV) and 65.7% (F14) were
higher than the 62.7% obtained by Cerri (2012). For the conversion
factor of extract into alcohol, the values of 0.452 g/g (F0IV) and 0.425
g/g (F14) were lower than 0.475 g/g of the laboratory scale formulation
obtained by Castro (2014). These authors also used barley malt sub­
stitutes in their beer formulations. The factors for conversion of extract
into alcohol obtained were also lower than the theoretical factor of
0.4839 g/g (Balling factor). The lower conversion factor of extract into
alcohol in relation to the Balling factor indicates less efficiency in the
fermentation process. Greater control of this parameter could have been

Table 2
Fermentative and kinetic parameters for control (F0IV) and optimized (F14)
formulations (n ¼ 3).
Parameters F0IV F14

di (g/cm3) 1.0500 1.0497


df (g/cm3) 1.0093 1.0094
Vi (mL) 9.000 9.500
Vf (mL) 6.500 6.500
OE(% w/w) 13.50 13.43
RE(% w/w) 4.65 4.61
AE (% w/w) 2.80 2.86
A(% w/w) 4.00 3.75
mEi (g) 1.275.7 1.339.3
mEf (g) 305.1 302.5
YA=E (g/g) 0.452 0.425
RDF (%) 64.1 63.0
ADF (%) 77.5 76.2
ATR (%) 65.6 65.7
ATA (%) 79.3 78.7
EfA (%) 93.4 87.8

Legend: di : initial density, df : final density, Vi : initial volume, Vf : final volume,


OE: original extract, RE: real extract, AE: apparent extract, A: alcohol content,
mEi : initial mass extract, mEf : final mass extract, YA=E : conversion factor of extract
in alcohol, RDF: real degree of fermentation, ADF: apparent degree of fermen­
Fig. 8. Fermentation kinetics of the control (F0IV) and optimized (F14) for­ tation, ATR : real degree of attenuation, ATA : apparent degree of attenuation, and
mulations (n ¼ 3). EfA : conversion efficiency of extract in alcohol.

7
S.T. Estev~
ao et al. Journal of Food Engineering 289 (2021) 110182

considered one of the CQAs of this experimental project, which would Table 3
allow greater success in obtaining the alcohol content intended in the Physicochemical, sensory analysis and purchase intention test results (n ¼ 3) for
formulations. control (F0IV) and optimized (F14) formulations.
However, according to Fermentis (2017), an 81% apparent attenu­ Variables F0IV F14 p-value
ation (ATA) value is expected after the proper use of US-05 yeasts, which Alcohol (% v/v) 5.93 (0.12) 5.73 (0.12) 0.1108
were used in this study. The values of 79.3% and 78.7% for control and Density (g/cm3) 1.0072 (0.0070) 1.0077 (0.0070) 0.9345
optimized formulations, respectively, were close to that indicated by the pH 4.35 (0.02) 4.33 (0.02) 0.1778
yeast strain manufacturer, which demonstrates that the fermentative Color (EBC) 6.7 (0.2) 4.9 (0.2) 0.0006
Calories (Kcal) 48.6 (1.5) 47.6 (1.5) 0.4741
process was well controlled, and that the final product had an adequate
Sensory
sweetness (residual sugars). Color 6.5 (1.6) 6.2 (1.7) 0.1870
The finalized F0IV and F14 formulations, the results of alcohol con­ Aroma 6.7 (1.7) 6.2 (1.7) 0.0055
tent determinations, density, pH, color, and calories, as well as the re­ Flavor 6.1 (2.2) 5.9 (2.2) 0.3802
sults of sensory analysis and purchase intention test for finished Total 19.4 (4.2) 18.3 (4.4) 0.0604
Purchase intention test 3.3 (1.2) 3.1 (1.3) 0.2491
products are presented in Table 3.
Among the physicochemical determinations, the formulations only Results are expressed as mean values (standard deviation).
presented a statistically significant difference in color analysis. This
difference was because the optimized formulation has 3% of non-malted selection of target customers, and understanding their needs. However,
maize syrup and oat malt are lighter than barley malt. The color vari­ the methodology can also be applied to improve existing products, to
ation of oat malt ranges from 2 to 4 EBC, while the barley malt used in add new manufacturing lines, to reduce cost, and even to reposition the
this experimental design ranges from 4 to 7 EBC. company in the market.
The pH values of both formulations were suitable and according to
expected, which, according to Rosa and Afonso (2015), is around pH 4.0. 4. Conclusion
The optimized formulation exhibited an alcohol content of 5.73 �
0.12% v/v, reaching the desired content in the QTPP. The lack of sta­ The Quality by Design approach was applied, from beginning to end,
tistically significant difference in alcohol between F14 and F0IV (5.93 � for the development of beer in this experimental project. The results
0.12% v/v) was because the optimized formulation contains 97% of indicate that industry and scientific research can promote innovation in
malted cereals (55% barley malt and 42% oat malt), thus the sugars the creation of new food products in a well-structured way using this
formed from the saccharification of starch were similar and fermenta­ methodology, contributing to the improvement of product quality and to
tion also occurred in a similar way. This fact was corroborated by the the reduction of quality deviations.
statistical similarity between the real degrees of fermentation (RDF) of
both formulations (Table 2). Declaration of competing interest
For sensory attributes color and flavor, the control and optimized
formulations did not present averages with statistically significant dif­ The authors declare that they have no known competing financial
ferences (p-value > 0.05). Only the aroma attribute of the control interests or personal relationships that could have appeared to influence
formulation was considered superior to the optimized formulation (p- the work reported in this paper.
value ¼ 0.0055). The value of 18 total sensory grade points for the
optimized formulation was the minimum score defined in the QTPP, CRediT authorship contribution statement
thus the optimized formulation obtained sensorially scores within the
pre-established requirements. There was no statistically significant dif­ Simone Tessarini Esteva ~o: Formal analysis, Investigation, Meth­
ference between control and optimized formulations for the total sen­ odology, Validation, Writing - original draft. Joa
~o Batista de Almeida e
sory values (p-value > 0.05). Silva: Investigation, Methodology, Resources, Supervision, Writing -
For the purchase intention test, the formulations received, on review & editing. Felipe Rebello Lourenço: Conceptualization, Fund­
average, the grade 3, which expressed the opinion “perhaps bought, ing acquisition, Project administration, Resources, Supervision, Writing
maybe did not buy”. These values did not present statistically significant - review & editing.
difference between them (p-value > 0.05).
As the optimized formulation was constructed throughout the
Acknowledgments
experimental project, it was expected that in addition to meeting the
attributes stipulated in the QTPP that the sensory evaluation of the F14
We thank the Programa de Po �s-Graduaça ~o Interunidades em Bio­
would be higher than the final result. Although Brazil is the third largest
tecnologia, Faculdade de Ci^encias Farmac^euticas and the Escola de
producer of beer in the world, the participation of craft beers in the
Engenharia de Lorena da Universidade de Sa ~o Paulo for offering the
Brazilian market is still negligible, with less than 1% of the total (Ayer,
necessary structure for the development of this work. Anton Paar
2017). The formulations produced in this experimental project fall into
gratefully loaned his Alex 500 beer analyzer.
craft beer and even with the increase in the availability of this product
on the market in recent years, the flavors and aromas differentiated from
Appendix A. Supplementary data
this type of beer are only beginning to be appreciated. The Brazilian
brewing market currently employs maize as a widely used substitute
Supplementary data to this article can be found online at https://doi.
(Mardegan et al., 2013) and the tasters demonstrated a tendency to
org/10.1016/j.jfoodeng.2020.110182.
appreciate more what they already know, since the formulations with
predominantly maize in the composition of the OD presented better
results in sensory analysis than the formulations produced with pre­ References
dominance of other substitutes.
Andrade, C.M., 2007. Obtenç~ao de chope utilizando arroz preto (Oryza sativa) como
The application of QbD approach in the development of a food adjunto de malte. 2007. 78f. Dissertaç~ao (Mestrado em Biotecnologia Industrial) –
product is still recent, but quite promising since the effective develop­ Escola de Engenharia de Lorena. Universidade de S~ ao Paulo, Lorena.
ment of a product is a form of innovation. Ayer, F., 2017. Com 15 cervejarias artesanais, Grande BH se consolida como o cintur~
ao
da cevada em MG. Disponível em: http://www.em.com.br/app/noticia/economia
For the food industry, before the application of QbD for the devel­ /2017/04/02/internas_economia,859102/cerveja-artesanal-cresce-producao-e
opment of a new product, a segmentation of the market is required, the m-minas-e-grande-bh.shtml. (Accessed 28 May 2018).

8
S.T. Estev~
ao et al. Journal of Food Engineering 289 (2021) 110182

Bento, C.V., Almeida, J.B., 2006. Elementos biotecnol� ogicos fundamentais no processo Curi, R.A., 2006. Produç~ ao de cerveja utilizando cevada como adjunto de malte. 2006.
cervejeiro: 1� parte–as leveduras. Revista Analytica 25, 36–42. 110f. Tese (Doutorado em Agronomia/Energia na Agricultura). Faculdade de
Brasil, 2019. Decreto n. 9.902, de 08 de julho de 2019. Altera o Anexo ao Decreto n� Ci^encias Agron^ omicas, Universidade Estadual Paulista, Botucatu.
6.871, de 4 de junho de 2009, que regulamenta a Lei n� 8.918, de 14 de julho de European Brewery Convention (EBC), 2005. Analytica - EBC., 5 ed. Brauerei-und
1994, que disp~ oe sobre a padronizaç~ ao, a classificaç~ao, o registro, a inspeç~ao, a Getr€ anke - Rundschau, Zurique.
produç~ ao e a fiscalizaç~ao de bebidas. Di�ario Oficial da Uni~ ao, Brasília, DF, 08 de jul. Fermentis, 2017. Dados T�ecnicos SafAle US-05. Disponível em: https://fermentis.com/
Disponível em. http://www.planalto.gov. wp-content/uploads/2017/10/SafAle-US-05-PT.pdf (Accessed 27 November 2018).
br/ccivil_03/_Ato2019-2022/2019/Decreto/D9902.htm#art2. (Accessed 29 August Food And Drug Administration. Guidance for Industry: Q8 (R2) Pharmaceutical.
2019). Fukuda, I.M., Pinto, C.F.F., Moreira, C.S., Saviano, A.M., Lourenço, F.R., 2018. Design of
Brasil, 2001. Instruç~ ao normativa n.54, de 5 nov. 2001. Adota o regulamento t�ecnico experiments (DoE) applied to pharmaceutical and analytical quality by design
MERCOSUL de produtos de cervejaria. Di� ario Oficial da Uni~ ao, Brasília, DF, 05 de (QbD). Brazilian Journal of Pharmaceutical Sciences 54 (Special).
nov. Disponível em: http://www.agricultura.gov.br/assuntos/vigilancia-agropecuar Gonçalves, A.Z.L., 2006. Produç~ ao de α-amilase e glucoamilase termoest� avel pelo fungo
ia/ivegetal/bebidas-arquivos/in-no-54-de-5-de-novembro-de-2001.doc/view, 2019 termofílico Thermomyces lanuginosus TO-03 por fermentaç~ ao submersa e em estado
29 mai.�. s�olido e caracterizaç~ ao das enzimas. Dissertaç~ao (Mestrado em Ci^ encias Biol�ogicas,
Brasil, 2009. Lei n. 6871, de 04 de junho de 2009. Regulamenta a Lei n. 8.918, de 14 de Area
� de Concentraç~ ao: Microbiologia Aplicada) - Instituto de Bioci^
encias,
julho de 1994, que disp~ oe sobre a padronizaç~ ao, a classificaç~ao, o registro, a Universidade Estadual Paulista, 76f Rio Claro, 2006.
inspeç~ao, a produç~ ao e a fiscalizaç~
ao de bebidas. Di� ario Oficial da Uni~ ao, Brasília, Leng, D., et al., 2017. Engineering of budesonide-loaded lipid-polymer hybrid
DF, 04 de jun. Disponível em: http://www.planalto.gov.br/ccivil_03/_Ato2007-2010 nanoparticles using a quality-by-design approach. Int. J. Pharm. (jul).
/2009/Decreto/D6871.html. (Accessed 12 January 2016). Likert, R., 1932. A Technique for the Measurement of Attitudes. Archives of Psychology,
Brasil, 2004. Resoluç~ ao RDC n� 216, de 15 de setembro de 2004. Disp~ oe sobre vol. 140. New York University, New York, p. 55p.
Regulamento T�ecnico de Boas Pr� aticas para Serviços de Alimentaç~ ao. Di�ario Oficial Lindstrom, L.A., Hansen, S.H., 2014. Frustrated Beer Brewer Created the pH Scale.
da Uni~ ao, Brasília, DF, 15 de set. Disponível em: http://portal.anvisa.gov.br/doc ESOF2014 Future Academy. Science Nordic. Disponível em: Acesso em. http://scien
uments/33916/388704/RESOLU%25C3%2587%25C3%2583O-RDC%2BN%2B2 cenordic.com/frustrated-beer-brewer-created-ph-scale. (Accessed 3 June 2019).
16%2BDE%2B15%2BDE%2BSETEMBRO%2BDE%2B2004.pdf/23701496-92 Mardegan, S.F., et al., 2013. Stable carbon isotopic composition of Brazilian beers-A
5d-4d4d-99aa-9d479b316c4b. (Accessed 28 November 2019). comparison between large- and small-scale breweries. J. Food Compos. Anal. 29 (1),
Briggs, D.E., Boulton, C.A., Brookes, P.A., Stevens, R., 2004. Brewing Science and 52–57.
Practice. Woodhead Publishing, Sawston. Medeiros, M.B., Aizemberg, R., Silva, T.A.O., Almeida E Silva, J.B., Produç~ ao de cerveja,
Candioti, L.V., et al., jun. 2014. Experimental design and multiple response optimization. 2018. In: Ribeiro. D. Microbiologia Industrial. v.2. Rio de Janeiro. In: Elsevier,
Using the desirability function in analytical methods development. Talanta 124, MENEZES, J.C., FLORES,^ A., GOUVEIA, F.F. (Eds.), Quality by Design (QbD): ci^encia
123–138. aplicada na manufatura e desenvolvimento de produtos químicos, vol. 23. Jornal do
Castro, O.M., 2014. Obtenç~ ao de cerveja super concentrada com utilizaç~ ao de xarope de Conselho Regional de Química, S~ ao Paulo, 129, set./out. 2014.
milho como adjunto de malte. 2014. 144f. Dissertaç~ ao (Mestrado em Biotecnologia Nascimento, A.P.S., Pinto, A.L., Freitas, J.M.S., Sousa, F.C., 2015. Efeitos da relaç~ ao
Industrial) – Escola de Engenharia de Lorena. Universidade de S~ ao Paulo, Lorena. tempo/temperatura e concentraç~ ao/pH na reaç~ ao de Maillard em diferentes
Cerri, C.F.F., 2012. Utilizaç~ao de arroz preto do tipo IAC-600 (Oryza sativa) como adjunto açúcares. Revista Brasileira de Agrotecnologia, Garanhus 5 (1), 1–8.
para produç~ ao de cerveja. 2012. 34f. Monografia (Graduaç~ ao em Engenharia Nunes, D.B., 1998. Rotinas para a otimizaç~ ao experimental de misturas. 1998. 163 f.
Industrial Química) – Escola de Engenharia de Lorena. Universidade de S~ ao Paulo, Dissertaç~ ao (Mestrado em Engenharia de Produç~ ao) – Escola de Engenharia.
Lorena. Universidade Federal do Rio Grande do Sul, Porto Alegre.
Cornell, J.A., 1973. Experiments with mixtures: a review. Technometrics 15 (3), 437. Olec, C. pH da cerveja: entenda como ele influencia o resultado final. Disponível em:�
Cornell, J.A., 1990. Experiments with Mixtures: Designs, Models and the Analysis of http://blog.casaolec.com.br/ph-da-cerveja-entenda-como-ele-influencia-o-resulta
Mixtures Data. Wiley, New York. do-final/>. Accessed on : 03 jun. 2019.
Coscione, A.R., Andrade, J. C. De, May, G.M., 2005. O modelamento estatístico de Politis, S., et al., 2017. Design of experiments (DoE) in pharmaceutical development.
misturas: experimento tutorial usando voltametria de redissoluç~ ao an�odica, 28. dez, Drug Dev. Ind. Pharm. 43 (6), 889–901, 3 jun.
pp. 1116–1122. Quím. Nova, 6. Rosa, N.A., Afonso, J.C., 2015. A Química da Cerveja. Quím. Nova 37 (2), 98–105 mai.
Costa, L.M.R., 2017. Produç~ ao de cerveja artesanal pela fermentaç~ ao de uma levedura da http://qnesc.sbq.org.br/online/qnesc37_2/05-QS-155-12.pdf. (Accessed 3 June
Jabuticaba: An� alise da cin�etica local de metab� olitos vol�ateis e dos efeitos das 2019). Disponível em:
vari�aveis no processo. 2017. 127f. Dissertaç~ ao (Mestrado em Ci^ encias em Yu, L.X., et al., 2014. Understanding pharmaceutical quality by design. AAPS J. 16 (4),
Engenharia Química) - Programa de Engenharia Química. Universidade Federal do 771–783, 23 jul.
Rio de Janeiro, Rio de Janeiro.

You might also like