Methods for the determination of HMF in honey: a comparison

M. Zappal
a a, B. Fallico a,*
, E. Arena a, A. Verzera b

a
Dipartimento di OrtoFloroArboricoltura e Tecnologie Agroalimentari (DOFATA), Facolt a di Agraria, Universit
a di Catania, Via S. Sofia, 98,
Catania 95128, Italy
b
a di Messina, Papardo, Messina 98168, Italy
Dipartimento di Chimica Organica e Biologica, Universit
Received 18 May 2003; received in revised form 3 March 2004; accepted 5 March 2004

Abstract
HMF (5-hydroxymethylfurfuraldehyde) is essential to evaluate the conformity of honey to the current legislation. Elevated
concentrations of HMF in honey provide an indication of overheating, storage in poor conditions or age of the honey. Both the
Codex Alimentarius Commission (Alinorm 01/25, 2000) and the European Union (Directive 110/2001) established that its con-
centration in honey usually should not exceed 80 or 40 mg/kg, respectively. The International Honey Commission recommends three
methods for the determination of HMF: two spectrophotometric methods, determination after White and after Winkler and a
HPLC method. These methods were recently tested by the International Honey Commission (1999). Aim of this research was to
compare HMF values in unifloral honeys measured by the three methods. From our data, HPLC and White methods usually give
similar values, except for eucalyptus honey; Winkler method gave for all honeys higher values than other two methods.

2004 Elsevier Ltd. All rights reserved.

Keywords: HMF (5-hydroxymethylfurfuraldehyde); Unifloral honey; Analytical methods

1. Introduction pean Union (EU Directive 110/2001) fixed a HMF limit

in honey of 40 mg/kg with the following exceptions: 80
HMF (5-hydroxymethylfurfuraldehyde) measure- mg/kg for honey coming from Countries or Regions
ment is used to evaluate the quality of honey; generally with tropical temperatures, 15 mg/kg for honey with low
not present in fresh honey, its content increases during enzymatic level (8-3 Schade Units).
conditioning and storage. Honey processing, requires The International Honey Commission (IHC, Stefan
heating both to reduce viscosity, and to prevent crys- Bogdanov, 1999, pp. 1–54) recommends three methods
tallisation or fermentation (Singh, Singh, Bawa, & Sek- for the determination of HMF. These methods include
hon, 1988) in air ventilated chambers, at 45–50 C for 4/7 two spectrophotometric methods widely used in routine
days or by immersion of honey drums in hot water. analysis, determination after White (1979) and after
Heating of unifloral honey leads to different HMF levels Winkler (1955), as well as the HPLC. The method de-
in honey (Fallico, Zappala, Arena, & Verzera, 2004). scribed by White involves measurement of UV absor-
HMF is formed during acid-catalysed dehydration of bance of clarified aqueous honey solutions with and
hexoses (Belitz & Grosch, 1999) and, it is connected to without bisulphite while that of Winkler involves mea-
the chemical properties of honey, like pH, total acidity, surement of the UV absorbance of honey solutions with
mineral content (Anam & Dart, 1995; Bath & Singh, added barbituric acid and p-toluidine (IHC, Stefan
1999; Hase, Suzuki, Odate, & Suzuki, 1973; Singh & Bogdanov, 1999, pp. 1–54).
Bath, 1997, 1998). The HPLC methods is according to Jeuring and
Codex Alimentarius (Alinorm 01/25 2000) established Kuppers (1980): honey is simply dissolved in water and,
that the HMF content of honey after processing and/or after a filtration, HMF is determined on a reverse phase
blending must not be higher than 80 mg/kg. The Euro- HPLC column by isocratic elution with water and
methanol as mobile phase. HPLC separates HMF from
other components and thus avoid interference in the
*
Corresponding author. Tel.: +39-957580214; fax: +39-957141960. determination. These interferences could be due to
E-mail address: [email protected] (B. Fallico). various aldehydes present in the honeys according to
0956-7135/$ - see front matter
2004 Elsevier Ltd. All rights reserved.
doi:10.1016/j.foodcont.2004.03.006
274 a et al. / Food Control 16 (2005) 273–277
M. Zappal

their floral origin or to products that appear during and the corresponding moisture content (%) was cal-
conditioning or storage (Wootton & Ryall, 1985). culated according to AOAC (1980);
However, the use of different analytical methods for • electrical conductivity was measured at 20 C in a
HMF determination and the use of inaccurate or inad- 20% (w/v) solution (dry matter basis) in deionised
equate procedures are actually a problem. The methods water (Loveaux, Pourtallier, & Vorwohl, 1973) by a
were recently tested by the International Honey Com- Delta Ohm HD 8706 conductivity meter;
mission (IHC, Stefan Bogdanov, 1999, pp. 1–54), the • ash was indirectly determined using the measured
methods yielded comparable values in collaborative electrical conductivity and applying the following
studies on three honey samples having different HMF equation: X1 ¼ ðX2
0:143Þ=1:743 were: X1 ¼ ash
content, to cover the main range of determination. Small value; X2 ¼ electrical conductivity in lS/cm at 20 C
differences between the methods resulted only at very (Piazza, Accorti, & Persano Oddo, 1991);
low levels, of no interest for assessing honey quality. • free acids, lactones, total acidity and pH were mea-
Aim of this research was to compare the HMF level in sured using a Mettler Toledo MP 220 pH meter
unifloral honeys measured by the three methods. according to Official Method (Repubblica Italiana:
GU. no. 282, 12/10/1984);
• diastase determinations were conducted by an enzy-
2. Material and methods matic-spectrophotometric method, using a kit Phade-
bas Amylase Test (Pharmacy & Upjohn Diagnostic
2.1. Samples AB).

Fourteen different honey samples, belonging to the 2.3. HMF determination

following floral origin: acacia, citrus, eucalyptus, chest-
nut and wildflower as declared in the label of the pro- 2.3.1. HPLC method
ducer, were purchased in some local shopping centres. Five grams of honey samples were diluted up to 50 ml
The unifloral honey samples were collected as follow: with distilled water, filtered on 0.45 lm filter and
two different samples of acacia, eucalyptus and chestnut immediately injected in a HPLC (Varian 9012Q)
honey and four different samples of citrus and wild- equipped with a Diode Array Detector (Varian, Star
flower honey, respectively (Table 1). Each honey sample 330). The HPLC column was a Merck Lichrospher, RP-
was purchased in duplicate in pot of 500 g. 18, 5 lm, 125 · 4 mm, fitted with a guard cartridge
packed with the same stationary phase (Merck, Milan).
2.2. Chemical analyses The HPLC conditions were the following: isocratic
mobile phase, 90% water at 1% of acetic acid and 10%
The following chemical determinations were carried methanol; flow rate, 0.7 ml/min; injection volume, 20 ll.
out on honey samples: All the solvents were HPLC grade (Merck, Milan). The
wavelength range was 220–660 nm and the chromato-
• moisture was determined measuring the refractive grams were monitored at 285 nm. HMF was identified
indices at 20 C by a Carl Zeiss 16531 refractometer by splitting the peak in honey with a standard HMF

Table 1
Characterisation of the different honey samples analysed
Samples Moisture (%) Ash (g%) Electrical pH Diastase Free acids Lactones Total acidity
conductivity activity (meq/kg) (meq/kg) (meq/kg)
(ms/cm) (Schade)
Acacia 1 17.4 ± 0.01 0.003 ± 0.001 0.15 ± 0.001 3.38 ± 0.02 7.7 ± 0.42 19.9 ± 0.48 4.2 ± 0.38 24.1 ± 0.51
Acacia 2 17.0 ± 0.01 n.d. 0.13 ± 0.002 3.55 ± 0.03 14.8 ± 0.38 13.3 ± 0.29 4.3 ± 0.82 17.6 ± 0.93
Citrus 1 16.6 ± 0.01 0.046 ± 0.002 0.22 ± 0.003 3.46 ± 0.01 7.8 ± 0.02 26.3 ± 0.35 4.2 ± 0.46 30.4 ± 0.11
Citrus 2 17.2 ± 0.01 0.120 ± 0.001 0.35 ± 0.001 3.46 ± 0.01 7.9 ± 0.30 29.8 ± 0.35 3.9 ± 0.92 33.6 ± 1.27
Citrus 3 19.1 ± 0.14 0.047 ± 0.001 0.23 ± 0.002 3.49 ± 0.03 10.0 ± 0.69 26.0 ± 0.01 5.0 ± 0.32 31.0 ± 0.91
Citrus 4 19.5 ± 0.01 0.050 ± 0.004 0.23 ± 0.008 3.43 ± 0.01 12.0 ± 0.93 27.0 ± 0.01 4.7 ± 0.82 31.7 ± 0.82
Eucalyptus 1 15.5 ± 0.17 0.209 ± 0.002 0.51 ± 0.004 3.68 ± 0.01 18.2 ± 0.65 29.3 ± 0.50 4.8 ± 0.84 34.1 ± 1.08
Eucalyptus 2 16.5 ± 0.35 0.235 ± 0.001 0.55 ± 0.002 3.66 ± 0.01 27.9 ± 1.18 34.1 ± 0.25 4.5 ± 0.53 38.6 ± 0.36
Chestnut 1 18.0 ± 0.12 0.688 ± 0.002 1.34 ± 0.004 4.98 ± 0.02 15.7 ± 0.63 17.3 ± 0.29 6.5 ± 0.53 23.7 ± 0.60
Chestnut 2 17.9 ± 0.12 0.929 ± 0.006 1.59 ± 0.002 5.84 ± 0.04 22.2 ± 0.94 11.4 ± 0.25 5.2 ± 0.53 16.5 ± 0.36
Wildflower 1 17.1 ± 0.10 0.129 ± 0.001 0.37 ± 0.001 3.92 ± 0.08 20.0 ± 0.31 24.1 ± 0.85 5.5 ± 0.84 29.6 ± 0.94
Wildflower 2 15.8 ± 0.01 0.061 ± 0.01 0.25 ± 0.001 3.70 ± 0.03 18.6 ± 0.29 22.9 ± 0.25 6.0 ± 2.08 28.8 ± 2.16
Wildflower 3 16.9 ± 0.06 0.550 ± 0.004 1.10 ± 0.008 4.38 ± 0.03 18.9 ± 0.49 32.5 ± 0.41 6.1 ± 0.84 38.6 ± 1.04
Wildflower 4 18.0 ± 0.10 0.178 ± 0.002 0.45 ± 0.004 3.76 ± 0.04 13.7 ± 0.49 38.3 ± 0.29 5.2 ± 0.53 43.4 ± 0.76
 a et al. / Food Control 16 (2005) 273–277
M. Zappal 275

(Sigma-Aldrich, Milan), and by comparison the spec- sults were expressed as expanded/uncertainty (U ) using
trum of HMF standard with that of honey samples. The 2 as coverage factor (95% C.L.), calculated as follows:
amount of HMF was determined using an external
U ¼ 2  Cx  RSDr
calibration curve, measuring the signal at k ¼ 285 nm.
Five grams of honey resulted the optimal weight where 2 is the coverage factor; Cx the concentration of
using the Ingamells e Switzer equation (Mannino, 2001). HMF (mg/kg of honey); RSDr is the relative repeat-
ability standard deviation calculated from duplicate
2.3.2. Spectrophotometric method (White) determination, it was calculated from 40 and 72 repli-
Five grams of honey were dissolved in 25 ml of water, cates for HPLC and spectrophotometric analyses,
transferred quantitatively into a 50 ml volumetric flask, respectively. The HPLC replicates were: 8 acacia, 12
added by 0.5 ml of Carrez solution I and 0.5 ml of citrus, 0 eucalyptus, 4 chestnut and 16 wildflower; the
Carrez II and make up to 50 ml with water. The solution spectrophotometric replicates were: 12 acacia, 18 citrus,
was filtered through paper rejecting the first 10 ml of the 12 eucalyptus, 6 chestnut and 24 wildflower.
filtrate. Aliquots of 5 ml were put in two test tubes; to Statgraphics plus software, version 5.0 was used to
one tube was added 5 ml of distilled water (sample perform statistical analyses of the HMF data obtained.
solution); to the second was added 5 ml of sodium bi- The multiple range tests were performed to evaluate the
sulphite solution 0.2% (reference solution). The absor- statistically significant difference between the HMF
bance of the solutions at 284 and 336 nm was concentration in honeys obtained with three methods.
determined using a VARIAN mod. Cary 1E UV–visible. The model elaborated shows a statistically significant
The quantitative value of HMF was determined both difference at the 95% confidence level.
by the external standard method (p 99% Sigma-Aldrich,
Milan) and by using the proposed formula for the
method reported by IHC (IHC, Stefan Bogdanov, 1999, 3. Results
pp. 1–54).
Table 1 reports the chemical parameters of the anal-
ysed honey: moisture, ash, electrical conductivity, pH,
2.3.3. Spectrophotometric method (Winkler)
diastase activity free acids, lactones and total acidity. All
Ten grams of honey were dissolved in 20 ml water
these data are in agreement with those reported in lit-
and transferred to a 50 volumetric flask. 2 ml of the
erature for each unifloral honey (Persano Oddo et al.,
solution and 5.0 ml of p-toluidine solution were put in
2000).
two different test tubes; to one tube was added 1 ml of
Table 2 reports the RSDr associated to each analyt-
distilled water (reference solution); to the second, 1 ml of
ical method for HMF determination in each unifloral
barbituric acid solution 0.5% (sample solution). The
honey and an average value for each analytical method
absorbance of the solutions at 550 nm was determined
including all honeys samples. Since the calculated RSDr
using a VARIAN mod. Cary 1E UV–visible. The
for chestnut honey by White method was very high
quantitative value of HMF was determined both by
(21.3%) it was not included in the calculation of the
the external standard method (p 99%, Sigma-Aldrich,
average RSDr. The lowest RSDr in honey analysis was
Milan) and by using the proposed formula for the
found with HPLC determination in chestnut honey (3%)
method (IHC, Stefan Bogdanov, 1999, pp. 1–54).
and an average value of 5.8%.
The two spectrophotometric methods show an aver-
2.4. Uncertainty estimation and statistical analyses age RSDr value of 6.0% and 8.6% using the White and
the Winkler methods, respectively.
The measured uncertainty for HMF analyses in Table 3 reports the HMF level in honey samples
unifloral honeys was estimated on the basis of the analysed by the three methods proposed by the IHC.
international laboratory (in-house) according to the Both, for acacia and citrus honeys the highest HMF
Nordic Committee on Food Analysis (Wood, Nilsson, values usually were those measured by spectrophoto-
& Wallin, 1998) procedure and Eurachem Guide. Re- metric analyses, the lowest values were that measured by

Table 2
Relative standard deviation RSDr in HMF determinations by different analytical methods
Analytical methods Average value Acacia Citrus Eucalyptus Chestnut Wildflower
HPLC 0.058 0.054 0.059 – 0.030 0.061
White 0.060a 0.075 0.044 0.060 0.213 0.064
Winkler 0.086 0.104 0.082 0.077 0.113 0.077
a
Average RSDr is calculated for each method using all determinations with the exception of Chestnut honey for White method.
276 a et al. / Food Control 16 (2005) 273–277
M. Zappal

Table 3
HMF level in commercial monofloral honey determined by HPLC and spectrophotometric methods (HMF ± U *) (mg/kg)
Samples HPLC White Winkler
Ext. st. Formula Ext. st. Formula Ex. St.
Acacia 1 16.2ab ± 1.89 18.4c ± 2.21 20.7d ± 2.50 17.5bc ± 3.03 15.7a ± 2.72
Acacia 2 8.4a ± 0.98 9.1ab ± 1.10 10.0b ± 1.20 11.9c ± 2.06 10.0b ± 1.73
Citrus 1 14.3a ± 1.67 16.4b ± 1.97 18.4c ± 2.21 18.2c ± 3.15 16.7bc ± 2.89
Citrus 2 45.2a ± 5.28 47.0a ± 5.64 54.2b ± 6.50 58.8c ± 10.17 51.5b ± 8.91
Citrus 3 9.4a ± 1.10 9.8a ± 1.18 10.8b ± 1.30 13.5c ± 2.34 11.4b ± 1.97
Citrus 4 8.1a ± 0.95 9.6b ± 1.15 10.3b ± 1.24 11.4c ± 1.97 9.5b ± 1.64
Eucalyptus 1 0.00a 27.7b ± 3.32 31.3c ± 3.76 52.4e ± 9.07 45.4d ± 7.85
Eucalyptus 2 0.00a 6.9b ± 0.83 7.3b ± 0.88 11.5d ± 1.99 9.7c ± 1.68
Chestnut 1 4.1a ± 0.48 4.0a ± 1.72 3.8a ± 1.63 10.4c ± 1.80 8.6b ± 1.49
Chestnut 2 0.00a 0.8b ± 0.34 0.00a 3.2d ± 0.55 2.2c ± 0.38
Wildflower 1 13.7a ± 1.60 14.2a ± 1.70 16.7b ± 2.00 19.1c ± 3.30 16.3b ± 2.82
Wildflower 2 14.2a ± 1.66 13.7a ± 1.64 14.5a ± 1.74 13.9a ± 2.40 11.7b ± 2.02
Wildflower 3 19.6a ± 2.29 17.0b ± 2.04 19.1a ± 2.29 27.1d ± 4.69 23.5c ± 4.07
Wildflower 4 85.5a ± 9.99 83.9a ± 10.1 97.5b ± 11.7 108.8c ± 18.82 99.5b ± 17.2
* U ¼ expanded uncertainty calculated using a cover of factor of 2 (95% confidence level).
a;b;c;d
Means in the same row followed by a different letter are significantly different at 95% C.L.

HPLC. The behaviour of eucalyptus honeys was com- As concerns the wildflower honeys, the 2nd sample
pletely different from all others (Table 3). Both samples, gave similar HMF values measured by HPLC or spec-
when analysed by HPLC, gave no measurable amount trophotometric methods; in wildflower 1, 3 and 4 the
of HMF; the HMF measured by White method, were HMF was the highest when measured by Winkler
27.7 and 31.3 mg/kg of honey for eucalyptus 1, 6.9 and method.
7.3 mg/kg of honey for eucalyptus 2, using the suggested Authors confirm suggestions given by International
formula (IHC, Stefan Bogdanov, 1999, pp. 1–54) and Commission of Honey (IHC, Stefan Bogdanov, 1999,
the external calibration, respectively. The HMF mea- pp. 1–54) to not use the Winkler method for determining
sured by Winkler method were 52.4 and 45.4 mg/kg of HMF in honey, because of carcinogenic of p-toluidine
honey (values out of the legal limit) for eucalyptus 1, and of the low precision of this method.
11.5 and 9.7 mg/kg of honey for eucalyptus 2. Thus, HPLC method seems to be the more appro-
At moment it is not possible to explain exactly the priate for HMF determination in honey, because the
reason of the disagreement among methods, but some presence of substances, probably derived by heat or
considerations may be done. During the first stages of storage damage, which interfere with the UV methods
heating some HMF precursors are formed: in fact, our did not reveal.
eucalyptus samples in HPLC analyses show another
peak with a maximum absorbance at 256 nm. Previ-
ously, studying the HMF kinetics in unifloral honeys, we References
evidentiated a different behaviour of eucalyptus honey
from all other honeys (Fallico et al., 2004). It showed a Anam, O. O., & Dart, R. K. (1995). Influence of metal ions on
longer lag-phase than other honeys. At the beginning of hydroxymethylfurfural formation in honey. Analytical Proceedings
heating treatment this honey gives the lowest HMF Including Analytical Communications, 32, 515–517.
level, at the end of heating it reaches the same levels AOAC (1980). Official Methods of Analysis (13th ed.). Washington,
DC: Association of Official Analytical Chemists, Method 31.111,
of other honey. p. 521.
HMF level in chestnut honey was almost the same Bath, P. K., & Singh, N. (1999). A comparison between Helianthus
both determined by HPLC or White method, it is annuus & Eucalyptus lanceolatus honey. Food Chemistry, 67, 389–
heavily over estimated by Winkler method. Although 397.
Belitz, H. D., & Grosch, W. (1999). Food chemistry. Berlin: Springer-
the first two methods give similar results, the uncertainty
Verlag.
associated with the spectrophotometric determination in Codex Alimentarius, Alinorm 01/25 (2000). Draft revised standard for
chestnut honey is very high, while RSDr by HPLC was honey at step 8 of the Codex procedure.
the lowest value (Table 2). EU Directive /110/2001 of 02/12/2001 (L 10/47).
As chestnut honeys form very low levels of HMF EURACHEM/CITAC Guide (2000). Quantifying uncertainty in
even after prolonged heating, the authors confirm their analytical measurement. Ellison, S.R.L., Rosslein, M., & Williams,
A. (Eds.).
doubts already showed in previous paper (Fallico et al., Fallico, B., Zappal
a, M., Arena, E., & Verzera, A. (2004). Effects of
2004), about using HMF level as index of thermal heating process on chemical composition and HMF levels in
damage for this unifloral honey. Sicilian monofloral honeys. Food Chemistry, 85, 305–313.
 a et al. / Food Control 16 (2005) 273–277
M. Zappal 277

Hase, S., Suzuki, O., Odate, M., & Suzuki, S. (1973). Changes in Repubblica Italiana: G.U. no. 282, 12/10/1984, Official methods of
quality of honey on heating and storage. I. Changes in hydroxy- honey analysis.
methylfurfural (HMF) content of honey. Journal of Food Science Singh, N., & Bath, P. K. (1997). Quality evaluation of different types of
and Technology, 20, 248–256. Indian honey. Food Chemistry, 58, 129–133.
Harmonised methods of the International Honey Commission, Singh, N., & Bath, P. K. (1998). Relationship between heating &
responsible for the methods: Stefan Bogdanov, 1999, pp. 1–54. hydroxymethylfurfural formation in different honey types. Journal
Jeuring, J., & Kuppers, F. (1980). High performance liquid chromato- of Food Science and Technology, 35, 154–156.
graphy of furfural and hydroxymethylfurfural in spirits and honey. Singh, N., Singh, S., Bawa, A. S., & Sekhon, K. S. (1988). Honey––its
Journal of the Association of Official Analytical Chemists, 63, 1215– food uses. Indian Food Packer, 42, 15–25.
1218. White, J. (1979). Spectrophotometric method for hydroxymethyl
Loveaux, J., Pourtallier, J., & Vorwohl, G. (1973). Methodes d’anal- furfural in honey. Journal of the Association of Official Analytical
yses des miels. Conductivite (Analytical methods for honey Chemists, 62, 509–514.
conductivity). Bull. Apic. Inf. Doc. Sci. Tech. Inf., 16, 7. Winkler, O. (1955). Beitrag zum Nachwals und zur Bestimmung von
Mannino, S. (2001). Verification of sample homogenity and calcula- Oxymethylfurfural in Honig und Kunsthonig. Zeitschrift fur
tion of optimum sample weight. In S. Mannino, & G. Schleining Lebensmittel Untersuchung und Forshung, 102(3), 161–167.
(Eds.), Practical tasks for food quality assurance (pp. 63). Food Wood, R., Nilsson, A., & Wallin, H. (1998). Procedure for the
Net-Project No 55792–CP–3–00–1FR–ERASMUS–ETN. estimation and expression of measurement uncertainty in chemical
Piazza, M. G., Accorti, M., & Persano Oddo, L. (1991). Electrical analysis developed by the Nordic Committee on Food analysis. In
conductivity, ash, colour and specific rotatory power in Italian Quality in the food analysis laboratory (pp. 125–147). UK: The
unifloral honeys. Apicoltura, 7, 51–63. Royal Society of Chemistry.
Persano Oddo, L., Sabatini, A. G., Accorti, M., Colombo, R., Wootton, M., & Ryall, L. (1985). A comparison of Codex Alimen-
Marcazzan, G., Piana, M. L., Piazza, M. G., & Pulcini, P. (2000). I tarius Commission and HPLC methods for 5-hydroxymethyl-2-
mieli uniflorali italiani nuove schede di caratterizzazione. Ministero furaldehyde determination in honey. Journal of Apicoltural
Politiche Agricole e Forestali, 29–67. Research, 24(2), 120–124.