Journal of Dairy Research (2003) 70, 85–90.

f Proprietors of Journal of Dairy Research 2003 85

DOI: 10.1017/S0022029902005897 Printed in the United Kingdom

Kinetics of hydroxymethylfurfural, lactulose and furosine

formation in milk with different fat content
Wendie L Claeys, Ann M Van Loey and Marc E Hendrickx*
Department of Food and Microbial Technology, Faculty of Agricultural and Applied Biological Sciences, Katholieke Universiteit Leuven,
Kasteelpark Arenberg, 22, B-3001 Heverlee, Belgium

Received 29 October 2001 and accepted for publication 27 August 2002

In the context of the general applicability of hydroxymethylfurfural (HMF), lactulose and

furosine as time-temperature integrators (TTIs) for thermal processing of milk, the influence of
milk fat content was studied. Formation kinetics were analysed for milk with fat content of
4.0 ± < 0.1 %. In previous experiments, it was observed that, under isothermal and non-isothermal
heating conditions, formation of the three chemical compounds could be described by pseudo-
zero order kinetics. Since the kinetic model was known, the experimental design could be
simplified. Data were analysed by a non-linear regression procedure and results were evaluated
by construction of joint confidence regions and temperature time tolerance (TTT-) diagrams.
Formation kinetics of HMF and lactulose was not affected by milk fat content. Regarding
furosine, significant differences were observed between kinetic parameters in whole, semi-
skimmed and skimmed milk. The observed differences however were negligible in the context
of process impact evaluation.

Keywords: Milk fat, hydroxymethylfurfural, lactulose, furosine, kinetics.

Thermal processing of milk not only guarantees a micro- the reduction of a product quality parameter without any
biologically safe and shelf-stable product, but can also knowledge of the actual time-temperature profile of the
cause substantial changes in the nutritional, organoleptic heating process. As the definition implies, identification of
and/or technological properties, of which, many have been TTIs requires an extensive study of their kinetics (Taoukis
extensively reviewed in literature (Burton, 1988; Schaafsma, & Labuza, 1989 ; Hendrickx et al. 1994; Van Loey, 1996).
1989). To evaluate the adequacy of the wide range of heat In this context we analysed formation kinetics of
processes used in the dairy industry, knowledge of the hydroxymethylfurfural (HMF), lactulose and furosine in
process impact (in terms of food safety and quality) on milk detail (Claeys et al. 2001). Kinetics could be described by
is required. Moreover, knowledge of the impact could also a pseudo-zero order model, i.e. concentration increased
provide a reliable basis for any legislative proposal on milk linearly as a function of time (up until a certain time). A first
quality and for controlling milk authenticity in terms of assessment of kinetic parameters was made by isothermal
processing. experiments. Next, estimated parameters were evaluated
Due to limitations of in situ methods, it is current practice under dynamic temperature conditions since industrial
to quantify the effect of a heat treatment on food products thermal processes are generally characterized by a non-
by use of physical mathematical methods or by use of time- isothermal heating up and cooling down phase. However, if
temperature integrators (TTIs) (Van Loey et al. 1996). The TTIs are to be used as general markers for heat processed
latter are heat sensitive components extrinsic or intrinsic milk, kinetics need to be evaluated for variability associated
to the food product, which show a time- and temperature- with the raw material as well.
dependent, easily measurable and irreversible change that Among factors that influence thermally induced chemi-
can be correlated to the changes of a target attribute of a cal changes in milk are: (i) protein and sugar concentration,
food (a safety or quality parameter) undergoing the same e.g. formation of furosine is, in contrast to lactulose, highly
temperature exposure. One of the major advantages of TTIs dependent on protein concentration, and is therefore ex-
is that they allow calculation of the process lethality or pressed as mg/100 g protein (Montilla & Olano, 1997;
Rattray et al. 1997) ; (ii) salt concentration and pH. Citrate
*For correspondence; e-mail: marc.hendrickx@agr.kuleuven.ac.be and phosphate were found to catalyse the formation of
86 WL Claeys and others

lactulose, presumably by acting as bases (Andrews & Experimental Approach

Prasad, 1987) ; (iii) process conditions or the heat load
Generally, when determining kinetics under isothermal
applied. It has been demonstrated that direct ultra-high-
conditions, at each temperature the experimental domain
temperature (UHT) heating by steam injection generally
is divided in different preset time intervals. However, when
causes a smaller degree of lysine damage compared with an
the kinetic model is known, most reliable parameters
indirect UHT heating system, probably due to dilution of
(smallest confidence domains) are obtained when the
the milk during direct UHT with consequently a reduced
number of samples (n) is equal to the number of parameters
concentration of reactants and additionally different heat
(p). For n larger than p, the optimum is reached by repeating
transfer conditions (Nangpal et al. 1990; Dhen-Müller et al.
the data points of the standard design (n = p). Slope and
1991).
intercept of a curve for example, will be more accurate
Concerning the effect of milk fat on formation of HMF,
when replicates are made at the two end points of the
lactulose and furosine, reported results are scarce and con-
observation domain. The number of replicates per measure-
tradictory (de Koning et al. 1990; Berg, 1993 ; Pellegrino,
ment should be at least 5 to make a reasonable estimation
1994; Morales & Jiménez-Pérez, 1999). Therefore, the
of the variance possible (Fedorov, 1972; Duggleby, 1981;
objective of this work was to determine to what extent milk
Endrenyi, 1981; Atkinson & Bogacka, 1997).
fat content influences their formation kinetics. If so, milk fat
content should be taken into account when HMF, lactulose
and furosine are used as intrinsic TTIs for evaluating thermal Data-analysis : Kinetic Parameter Estimation
processed milk.
In previously performed experiments, it was observed that
HMF, lactulose and furosine formation reached a plateau
Materials and Methods upon prolonged heating, and could as such be described by
a fractional conversion model (Claeys et al. 2001). Taking
Milk
only the first phase of the model into consideration, a linear
Pooled whole and skimmed milk were purchased at the relation between concentration and treatment time was
Society for Milk Quality (Lier, Belgium). Semi-skimmed observed, and kinetics could be simplified to a pseudo-zero
milk was obtained by mixing equal volumes of whole and order model:
skimmed milk. The milk was divided into small portions   
(50 ml) and stored frozen at – 40 8C prior to thermal treat- Ea 1 1
C = C0 + kref exp -- t (1)
ment. R Tref T

with C0 the initial concentration, C the concentration of

Thermal treatment the chemical compound formed at treatment time t, Ea
Samples of milk in closed test tubes (pyrex, 16r160 mm) (J/mol) the activation energy quantifying the effect of tem-
were immersed in a thermostated oil bath at a constant perature on the reaction rate, R the universal gas constant
temperature. At different preset times, samples were taken (8.314 J/mol,K) and kref the reaction rate constant at refer-
from the oil bath and immediately cooled in ice water to ence temperature Tref. Based on eqn (1), data were analysed
stop further formation of the chemical compound studied. by means of a global non-linear regression procedure on
response values. Accuracy and precision that could be
Analytical Procedures associated with the estimated parameters were evaluated
by constructing 90 % joint confidence regions (Draper &
Total HMF content was quantified spectrophotometrically Smith, 1981 ; Johnson, 1992).
at 443 nm (Pharmacia LBK-Biochrom) using thiobarbituric All statistical procedures were carried out using the
acid (TBA) as a substrate following the method described by SAS software package 6.12 (SAS Institute, Cary, NC 27513,
Keeney & Bassette (1959). Concentration of lactulose was USA).
determined using a D-glucose/D-fructose test combination
of Boehringer Mannheim (1995). Formation of furosine was
measured by RP-HPLC (Amersham Pharmacia Biotech;
Results and Discussion
reversed-phase furosine dedicated column, 250r4.6 mm,
Metal-Free, Alltech) essentially according to the method Since the kinetic model of HMF, lactulose and furosine
of Resmini et al. (1990), with the exception that 100 ml of formation was known, sampling could be simplified as
hydrolysate was injected instead of 10 ml in order to stated in the Experimental Approach section. Doing so,
increase reproducibility and accuracy of measurements. reliability and accuracy of results could be increased and
Milk fat and protein content of whole and skimmed amount of experimental work reduced (Fedorov, 1972;
milk were analysed by the Society for Milk Quality (Lier, Duggleby, 1981; Endrenyi, 1981). To allow for the calcu-
Belgium) based on the Röse–Gottlieb gravimetric principle lation of Ea, two temperatures were chosen, corresponding
(Belgian Norm, 1988) and the Kjeldahl method (Inter- to the two extremes of the temperature domain analysed
national Dairy Federation, 1993) respectively. during former experiments (Claeys et al. 2001). Formation
 Milk fat content: HMF, lactulose and furosine 87

Table 1. Kinetic parameters kref (reaction rate constant at reference temperature, Tref ) and Ea (activation energy) for the formation of
hydroxymethylfurfural, lactulose and furosine in heated whole, semi-skimmed and skimmed milk. (See Data Analysis section for
details)
Values for kref and Ea ± standard error for n = 28

Whole milk Semi-skimmed milk Skimmed milk

Milk composition
Fat content, g/l 40.2 50/50 mix 0.5
Protein content, g/l 33.6 a 35.7
Hydroxymethylfurfural ; T-domain : 90–120 8C ; Tref : 105 8C
kref, mmol/l per min 0.771 ± 0.021 0.794 ± 0.01 0.815 ± 0.022
Ea, kJ/mol 116.4 ± 2.2 113.4 ± 1.1 110.5 ± 2.2
Lactulose ; T-domain : 90–120 8C ; Tref : 105 8C
kref, mg/l per min 31.7 ± 1.2 31.2 ± 1.0 28.2 ± 0.8
Ea, kJ/mol 109.2 ± 1.6 110.8 ± 1.5 113.6 ± 1.5
Furosine ; T-domain : 90–130 8C ; Tref : 110 8C
kref, mg/100 g protein per min 9.64 ± 0.14 8.95 ± 0.11 8.61 ± 0.07
Ea, kJ/mol 88.4 ± 0.9 92.8 ± 0.8 91.3 ± 0.5

of HMF and lactulose was studied at 90 8C and 120 8C, correlated. In Fig. 1a, 90 % joint confidence regions clearly
formation of furosine at 90 8C and 130 8C. At each tem- overlap indicating similar formation kinetics for HMF in
perature, sampling proceeded at two end points of the three types of milk. Seemingly, milk fat content has no
observation domain and was repeated seven times. To influence on HMF formation.
exclude an initial phase of non-isothermal heating as a This can also be illustrated by a practical-oriented
consequence of warming up of the milk in the vials, the first example. Morales et al. (2000) measuring average HMF
sampling point was taken at 3 min when temperature in amounts in on-line heat-treated milk samples, observed
the vials had reached a constant value (as verified exper- a mean HMF value of 2.5 mmol/l in pasteurized milk, 5.6
imentally). The mean of the concentrations measured in the and 8.7 mmol/l in directly and indirectly UHT-heated milk
corresponding milk samples was used as reference (C0 in respectively, and 22 mmol/l in sterilized milk. Considering
eqn (1)). the latter amount, corresponding time temperature combi-
Kinetic parameters kref and Ea were computed by one nations are presented by the TTT-diagram in Fig. 2a. From
step non-linear regression and are summarized in Table 1. this diagram it is clear that when the same processing tem-
Differences in kinetics between whole, semi-skimmed and peratures are applied to whole, semi-skimmed and skim-
skimmed milk were evaluated by means of 90% joint med milk, similar processing times are needed to attain the
confidence regions (Fig. 1). These are a more accurate rep- same level of HMF formation.
resentation of the statistical confidence associated with the Similar conclusions were drawn by Berg (1993), who
model parameters than the commonly applied individual examined HMF formation in UHT-milk with 0.1, 1.5, 3.0,
confidence intervals, because they take the possible cor- 4.2 and 4.6 % fat. In contrast however, are the observations
relation between the simultaneously estimated parameters of Morales & Jiménez-Pérez (1999). Studying HMF forma-
into account. When correlation between parameters is high, tion in lactose/caseinate model systems and in milk with
a parameter pair may be well within the rectangular domain different fat contents, they observed a negative effect of
defined by the separate 95 % confidence intervals, but may milk fat content on total HMF formation in both systems,
be very unlikely to occur since it is very far outside of after pasteurization as well as after direct UHT and ster-
the 90 % joint confidence ellipse. Individual confidence ilization. A different behaviour was found for formation of
intervals are suitable only for describing the limits of a free HMF, which was enhanced at higher fat concentra-
single parameter without regard of the other parameter tions. Total and free HMF are produced through different
(Haralampu et al. 1985; Van Loey, 1996). Next, tempera- pathways. While free HMF is formed only through
ture time tolerance (TTT-) diagrams were constructed, in decomposition of lactulosyllysine in the Maillard reaction,
which curves connect those time temperature conditions total HMF can also be formed by degradation of lactose.
leading to the same level of HMF, lactulose or furosine Formation of free HMF is small compared to formation
formation (Fig. 2). of total HMF, and probably not a reliable measure for
Based on a 95 % individual confidence level, no sig- the occurrence of the Maillard reaction. In spite of the
nificant differences were observed for HMF formation kin- protective effect of fat on total HMF formation observed
etics in whole, semi-skimmed and skimmed milk (Table 1). by Morales & Jiménez-Pérez (1999), Ea values of total HMF
However, 90 % joint confidence regions are better tools formation did not differ for varying fat contents, which is in
of comparison since kinetic parameter values were highly line with our observations.
88 WL Claeys and others

(a) 5·5 (a)

0·95

4·5

In(t) (min)
kref (
mol/l, min)

0·85
3·5
3+
+2
+1 2·5
0·75

1·5
85 95 105 115 125
0·65 T (°C)
102000 112 000 122 000
Ea (J/mol) 5 (b)
(b)
0·040
4

In(t) (min)
0·035
3
kref (g/l, min)

+1
+2
0·030 2
+3
1
0·025 80 90 100 110 120 130
T (°C)

0·020
102000 107 000 112 000 117 000 2
(c)
Ea (J/mol)
(c)
10·5 1
In(t) (min)
kref (mg/100g protein, min)

10·0 0
+1
9·5 –1

9·0 +2 –2
+3 85 95 105 115 125 135
8·5 T (°C)

8·0 Fig. 2. Time temperature tolerance diagram for formation of (a)

80000 85 000 90 000 95 000 100 000 22 mmol/l hydroxymethylfurfural as lower heating limit for
Ea (J/mol) sterilization, (b) 600 mg/l lactulose as upper heating limit for
UHT, and (c) 8 mg/100g protein furosine as upper heating limit
Fig. 1. Individual 95 % confidence intervals (- -) and 90 % joint for high-pasteurization of whole (—) semi-skimmed (- - -) and
confidence regions (indicated by ellipses) for the simultaneously skimmed (…) milk.
estimated kinetic parameters kref (reaction rate constant at
reference temperature Tref ) and Ea (activation energy) analysed
in heat-treated whole (1) semi-skimmed (2) and skimmed (3) This can be verified by the TTT-diagram in Fig. 2b. In
milk for (a) hydroxymethylfurfural (Tref = 105 8C ; correlation the application of lactulose as a heat marker, it is suggested
between kref and Ea : (1) 0.929, (2) 0.918, (3) 0.906), (b) lactulose that lactulose concentration should be below detection
(Tref = 105 8C ; correlation between kref and Ea : (1) 0.892, (2) limit in pasteurized milk and may not exceed 50 mg/l in
0.906, (3) 0.858), and (c) furosine (Tref = 110 8C; correlation highly pasteurized milk. However, these criteria are under
between kref and Ea : (1) 0.005, (2) 0.222, (3) 0.175). For details, discussion due to limitations in sensitivity of the detection
see Data Analysis section.
methods. Under UHT conditions an upper limit of 600 mg
lactulose/l milk has been proposed (Schlimme et al. 1994;
Wilbey, 1996). Time temperature combinations leading to
Like HMF, formation kinetics of lactulose were com- such lactulose concentration in whole, semi-skimmed and
parable in whole, semi-skimmed and skimmed milk (Table 1) skimmed milk are presented by the curves in the TTT-dia-
as indicated by the 90 % joint confidence regions in gram. In order to achieve the same lactulose level in three
Fig. 1b. Consequently, it seems that milk fat content does types of milk, associated processing temperatures differ
not have to be taken into account when using lactulose as less than 1 deg C when the same processing time is con-
a heat marker. sidered.
 Milk fat content: HMF, lactulose and furosine 89

Results reported in literature on the effect of milk fat skimmed milk. In case of lactulose formation, it is mainly
content on lactulose formation are rather contradictory and the salt system that is responsible for most of the lactulose
contain no explicit kinetic data. In agreement with our formed during thermal processing of milk and not so much
results are those of Berg (1993), Andrews (1984) and Geier the protein fraction (Montilla & Olano, 1997).
& Klostermeyer (1983), who observed no effect of milk Although significant differences were observed between
fat content on the formation of lactulose during heating. kinetic parameter values for furosine formation in whole,
Contrary are the observations made by de Koning et al. semi-skimmed and skimmed milk, in practice they seem to
(1990), who reported substantially higher lactulose forma- be negligibly small as can be illustrated by the TTT-diagram
tion (40–50 %) in milk with 3 % fat compared to milk with in Fig. 2c. Furosine concentrations found in samples of
1.5 % fat subjected to the same UHT treatment. Pellegrino Belgian consumption milk ranged from 4 to 7 mg/100 g
(1994) on the other hand, observed a protective effect of fat protein for pasteurized milk, from 35 to 109 mg/100 g
against heat-induced formation of lactulose. The difference protein for direct UHT-heated milk, from 168 to 180 mg/
between lactulose formed in whole and in skimmed milk 100 g protein for indirect UHT-heated milk, and from 220
was more apparent after a direct UHT process (19–31 %) to 372 mg/100 g protein for sterilized milk (Van Renter-
than after an indirect UHT process (3–5%), making the ghem & De Block, 1996). Although differences in process-
author suggest that – next to the major role of the type of ing conditions are clearly reflected by the furosine content,
process – the effect of fat will be more effective against low an overlap in furosine content is observed between the
heat damage. different heat classes, except for pasteurized and UHT-
With regard to furosine formation, based on a 95 % heated milk. The TTT-diagram in Fig. 2c represents those
individual confidence kinetic parameter values differed temperature time combinations leading to a furosine con-
significantly in whole, semi-skimmed and skimmed milk tent of 8 mg/100 g protein, a level that has been discussed
(Table 1). As depicted in Fig. 1c, 90% joint confidence by the European Union as a limit for peroxidase-positively
regions of simultaneously estimated parameters kref and Ea pasteurized milk (Clawin-Rädecker & Schlimme, 1995).
are separated. Since correlation between kref and Ea was From this diagram it is clear that the influence of fat on
low, 90 % joint confidence ellipses coincide approximately furosine formation is fairly insignificant ; at an equal pro-
with the rectangular domain formed by the individual 95 % cessing time processing temperatures differ less than 1.5–2
confidence intervals associated with each parameter. Fur- deg C for formation of the same amount furosine in whole,
osine formation seemed to proceed faster (higher k110 8C semi-skimmed and skimmed milk, which is an acceptable
value) and to be less temperature sensitive (lower Ea) in milk deviation in the context of process impact quantification.
with a higher fat content. Since Ea values differ, Arrhenius Pellegrino (1994) however, observed a similar protective
curves will intersect at a point where the effect of milk effect of milk fat on furosine formation as she did for lac-
fat on furosine formation is inverted. In the temperature tulose formation. The protective effect was explained by a
region studied, there appears to be a positive correlation difference in heat load between whole and skimmed milk.
between fat content and furosine formation, while above Fat content would affect viscosity of milk and so hinder
that region ( > 130 8C) an opposite trend is expected the transfer of heat. This explanation does not agree with
(assuming extrapolation is justified). our observations, since fat content did not have any affect
A possible explanation for the fact that an effect of fat whatever on HMF or lactulose formation kinetics, which
was observed on furosine and not on HMF or lactulose have a higher temperature sensitivity than furosine.
formation kinetics is the difference in the pathway involved. The main objective of this study was to evaluate to what
In heated milk lactulose is mainly formed by lactose iso- extent milk fat content jeopardizes the applicability of
merization (Lobry de Bruyn – Alberda van Ekenstein or LA HMF, lactulose and furosine as intrinsic TTIs. To diminish
transformation), while furosine is formed during acid as much as possible other factors next to fat content, a single
hydrolysis of lactulosyllysine, an Amadori product of the batch of milk was used. Based on data from literature no
Maillard reaction. HMF is a result of both the Maillard unambiguous conclusion could be drawn ; milk fat content
reaction and the isomerization reaction. Although, in milk has been reported to have a synergetic as well as antag-
the route of HMF formation is mostly the LA transformation onistic effect, or no effect at all on the formation of the
(Berg, 1993; Morales et al. 1997). The early Maillard chemical compounds studied in this paper. From our results
reaction and lactose isomerization are affected differently milk fat content appeared to have no effect on HMF or
by milk composition and heating conditions (Berg, 1993 ; lactulose formation. Although a significant difference was
van Boekel, 1998). During heating, denatured serum pro- observed for furosine formation kinetics in whole, semi-
teins initiate binding to micellar k-casein via disulphide skimmed and skimmed milk, the difference was too small to
bonds. Native fat-globule proteins contain cysteine and be relevant. HMF, lactulose and furosine thus seem gen-
can therefore interact with denatured serum protein. When erally applicable as intrinsic TTIs for the evaluation of
milk is heated, the amount of proteins associated with the consumption milk.
fat globules increases (Morales et al. 1995). This kind of However, given that milk composition is subjected to
interaction could explain the somewhat lower temperature seasonal variation and that fat content is not the only vari-
sensitivity of furosine formation in whole compared with able, variations observed in this study need to be evaluated
90 WL Claeys and others

in a broader framework. Further experimental work on this Haralampu S, Saguy I & Karel M 1985 Estimation of Arrhenius model
topic will be undertaken. parameters using three least squares methods. Journal of Food Pro-
cessing and Preservation 9 129–143
Hendrickx M, Maesmans G, De Cordt S, Noronha J, Van Loey A, Willocx F
This research has been supported by the Office of the Federal & Tobback P 1994 Advances in process modelling and assessment : the
Office for Scientific, Technical and Cultural Affaires as a part physical mathematical approach and product history integrators. In :
of project NP/01/034. Minimal processing of foods and process optimisation, pp. 315–336
(Eds R Single & F Oliveira). Boca Raton, USA : CRC Press
Erratum : In Claeys W, Ludikhuyze L & Hendrickx M 2001 For- International Dairy Federation 1993 Milk – Determination of nitrogen
mation kinetics of hydroxymethylfurfural, lactulose and furosine content Brussels : IDF (FIL-IDF Standard 20 B)
under isothermal and non-isothermal conditions. Journal of Dairy Johnson M 1992 Review : why, when and how biochemists should use
Research 68 287–301 the units of lactulose in Figs 2 & 3 should least squares. Analytical Biochemistry 206 215–225
read g/l rather than mg/l. Keeney M & Bassette R 1959 Detection of intermediate compounds in the
early stages of browning reaction in milk products. Journal of Dairy
Science 42 945–960
Montilla A & Olano A 1997 Effect of the protein content and dilution on
References the lactulose/furosine ratio in heat-treated milk. Milchwissenschaft 52
Atkinson A & Bogacka B 1997 Compound D- and Ds-optimum designs 506–507
for determining the order of a chemical reaction. Technometrics 39 Morales F-J, Romero C & Jiménez-Pérez S 2000 Characterization of
347–356 industrial processed milk by analysis of heat-induced changes. Inter-
Andrews G & Prasad K 1987 Effect of the protein, citrate and phosphate national Journal of Food Science and Technology 35 193–200
content of milk on formation of lactulose during heat treatment. Journal Morales F & Jiménez-Pérez S 1999 HMF formation during heat-treatment
of Dairy Research 54 207–218 of milk-type products as related to milkfat content. Journal of Food
Andrews G 1984 Distinguishing pasteurized, UHT and sterilized milks by Science 64 855–859
their lactulose content. Journal of the Society of Dairy Technology 37 Morales F-J, Romero C & Jiménez-Pérez S 1997 Chromatographic
92–95 determination of bound hydroxymethylfurfural as an index of milk
Arrhenius S 1889 About the reaction rate of the inversion of non-refined protein glycosylation. Journal of Agriculture and Food Chemistry 45
sugar at souring. Zeitschrift für Physikalische Chemie 4 226–248 1570–1573
Belgian Norm 1988 Milk and milk products – Determination of fat con- Morales F, Romero C & Jiménez-Pérez S 1995 New methodologies for
tent – Röse-Gottlieb gravimetric method (reference method). NBN V kinetic study of 5-(hydroxymethyl)-furfural formation and reactive
21–001 lysine blockage in heat-treated milk and model systems. Journal of Food
Berg H 1993 Reactions of lactose during heat treatment of milk: a Protection 58 310–315
quantitative study. PhD thesis, Wageningen Agricultural University. Nangpal A, Reuter H, Dhen-Müller B & Erbersdobbler H 1990 Formation
Wageningen, The Netherlands of furosine during UHT treatment of milk – comparison between direct
Boehringer, Mannheim 1995 Lactulose in milk. Enzymatic Bioanalysis and indirect heating. Kieler Milchwirtschaftliche Forschungsberichte 42
Food Analysis. Cat. No 139 106 43–51
Burton H 1984 Reviews of the progress of dairy science: The bacterio- Pellegrino L 1994 Influence of fat content on some heat-induced changes in
logical, chemical, biochemical and physical changes that occur in milk milk and cream. The Netherlands Milk & Dairy Journal 48 71–80
at temperatures of 100–150 8C. Journal of Dairy Research 51 341–363 Rattrey W, Gallmann P & Jelen P 1997 Influence of protein standardization
Claeys W, Ludikhuyze L & Hendrickx M 2001 Formation kinetics of and UHT heating on the furosine value and freezing point of milk. Lait
hydroxymethylfurfural, lactulose and furosine under isothermal and 77 297–305
non-isothermal conditions. Journal of Dairy Research 68 287–301 Resmini P, Pellegrino L & Battelli G 1990 Accurate quantification of
Clawin-Rädecker I & Schlimme E 1995 Determination of furosine in furosine in milk and dairy products by a direct HPLC method. Italian
pasteurised milk by use of ion-pair reversed-phase liquid chromato- Journal of Food Science 2 173–183
graphy. Kieler Milchwirtschaftliche Forschungsberichte 47 169–175 Schaafsma G 1989 Effects of heat treatment on the nutritional value of milk.
Dehn-Müller B, Müller B & Erbersdobler H 1991 Studies on protein International Dairy Federation Bulletin no. 238 68–70
damage in UHT milk. Milchwissenschaft 46 431–434 Schlimme E, Buchheim W & Heeschen W 1994 Evaluation of different
de Koning P, Badings H, van der Pol J, Kaper J & Vos-Klompmaker E heat treatments and heat indicators for consumption milk. DMZ-
1990 Effect of heat treatment and fat content on UHT-milk. Lebensmittelindustrie und Milchwirtschaft 115 64–69
Voedingsmiddelentechnologie 1 11–14 Taoukis P & Labuza T 1989 Applicability of time temperature integrators
Draper N & Smith H 1981 Applied regression analysis. New York : John as shelf life monitors of food products. Journal of Food Science 54
Wiley & Sons 783–788
Duggleby R 1981 Experimental designs for the distribution-free analysis Van Boekel M 1998 Effect of heating on Maillard reactions in milk. Food
of enzyme kinetic data. In : Kinetic data analysis, design and analy- Chemistry 62 403–414
sis of enzyme and pharmacokinetic experiments, pp. 169–179 (Ed. Van Loey A 1996 Enzymic time temperature integrators for the quantifi-
L Endrenyi). NY–London : Plenum Press cation of thermal processes in terms of food safety. PhD thesis, K.U.
Endrenyi L 1981 Design of experiments for estimating enzyme and Leuven, Laboratory of Food Technology, Leuven, Belgium
pharmacokinetic parameters. In : Kinetic data analysis, design and Van Loey A, Hendrickx M, De Cordt S & Tobback P 1996 Quantitative
analysis of enzyme and pharmacokinetic experiment, pp. 137–167 evaluation of thermal processes using time temperature integrators.
(Ed. L Endrenyi). NY–London : Plenum Press Trends in Food Science and Technology 7 16–26
Fedorov V 1972 Theory of optimal experiments. Academic Press, Van Renterghem R & De Block J 1996 Furosine in consumption milk and
NY–London milk powders. International Dairy Journal 6 371–382
Geier H & Klostermeyer H 1983 Formation of lactulose during heat Wilbey A 1996 Estimating the degree of heat treatment given to milk.
treatment of milk. Milchwissenschaft 38 475–477 Journal of the Society of Dairy Technology 49 109–112