JOURNAL OF FOOD COMPOSITION AND ANALYSIS (2001) 14, 93}100

doi:10.1006/jfca.2000.0972
Available online at http://www.idealibrary.com on

ORIGINAL ARTICLE

Comparative Study on Total Lipid Determination using Soxhlet,

Roese-Gottlieb, Bligh & Dyer, and Modified Bligh & Dyer
Extraction Methods
P. Manirakiza, A. Covaci, and P. Schepens
Toxicological Centre, University of Antwerp, Universiteitsplein 1, B-2610 Wilrijk, Belgium

Received May 30, 2000, and in revised form November 8, 2000

Soxhlet method was used in hot and standard extraction mode and compared with liquid}liquid
extraction methods for lipid determination in food samples. Several solvent mixtures were
investigated and Soxhlet extraction time was optimized. Soxhlet extraction was convenient only
for solid samples. Except Roese}Gottlieb method, other liquid}liquid extraction methods were
suitable for lipid extraction from all samples analysed. Even if the precision of each method was
good (R.S.D.(3.5%), the comparison of extracted lipid amounts showed that, depending on the
type of sample, some methods present serious limitations.  2001 Academic Press

Key Words: lipids; extraction; liquid}liquid; Soxhlet.

INTRODUCTION

Lipids are a diverse group of biological substances made up primarily of non-polar

compounds (triglycerides, diglycerides, monoglycerides and sterols) and more polar
compounds (free fatty acids, phospholipids and sphingolipids). They bind covalently
to carbohydrates and proteins to form glycolipids and lipoproteins, respectively. The
possibility of lipids to bind to other molecules and the ability of di!erent solvent
mixtures to solubilize lipid classes has led to the concept of &&total lipid extract'' and
&&extractable lipid''. Solvents used for lipid extraction should have a high solubility for
all lipid compounds and be su$ciently polar to remove them from their binding sites
with cell membranes, lipoproteins and glycolipids (Smedes and Askland, 1999).
The knowledge of lipid content in food or other tissues is important from two
points of view. First, in nutrition, equilibrated alimentation requires knowledge on
lipid, protein and sugar content of the food. Second, in environmental studies or
monitoring programs, concentrations of persistent organic contaminants (dioxins,
PCBs, organochlorine pesticides) are often expressed on a lipid content (Randall et al.,
1991; Baily et al., 1994; Delbeke et al., 1995; Ramos et al., 1998).
Several methods have been developed for total lipid extraction (Bligh and Dyer,
1959; Christie, 1976; Gardner et al., 1985; de Boer, 1988; Booij and Van den Berg,
1994; Smedes, 1999). Most of them use liquid}liquid extraction (LLE) methods, which
are time consuming, and imply a lot of manual operations.

 To whom correspondence and reprint requests should be addressed. Tel.: #32-3-820-2743. Fax:
#32-3-820-2722. E-mail: [email protected]

0889}1575/01/010093#08 $35.00/0  2001 Academic Press

94 MANIRAKIZA E¹ A¸.

It has been demonstrated (de Boer, 1988; Randall et al., 1991) that the use of
di!erent methods results in di!erent lipid recovery. Indeed, results varied widely due
to di!erences in extraction methodology, as con"rmed by an intercomparative study
including &&the extractable lipid'' determination (Baily et al., 1994). It was also shown
that, when used by di!erent laboratories, the Bligh & Dyer method led to high
variability in the results.
This study aims to evaluate the e$ciency of hot Soxhlet extraction method and to
compare it with other methods for lipid determination by applying them to di!erent
matrices. Time involved, amount and toxicity of solvents and repeatability of each
method are discussed.

MATERIALS AND METHODS

Sampling

Margarine, eggs, full-fat powdered chocolate, full-fat liquid milk and skimmed pow-
dered milk were purchased from a local supermarket, while chicken feed and "sh #our
were obtained from a producer. The lipid content for each sample was provided on the
label by the manufacturer. Pooled human serum was obtained from the University
Hospital of Antwerp and its lipid composition was determined by enzymatic methods
(Elitech Diagnostics kit, Brussels, Belgium). Liquid samples were stored at #43C
before use and solid ones were kept at room temperature. All samples were used
without any pre-treatment.

Chemicals

Anhydrous sodium sulfate (p.a.), ammonia solution 25% (p.a.) and n-hexane (Supra-
Solv) were from Merck (Darmstadt, Germany). Diethyl ether (p.a.), chloroform (p.a.),
petroleum ether (g.p.r.), acetone (p.a.) and dichloromethane (p.a.) were from Lab-Scan
(Dublin, Ireland). Methanol (p.a.), iso-propanol (p.a.) and cychlohexane (p.a.) were
purchased from Accros Organics (Geel, Belgium).

Instrumentation

Soxhlet extraction was performed on a BuK chi Universal Extraction System B - 811
(Flawil, Switzerland) with thimbles provided by Schleicher & Schuell (Dassel, Ger-
many). Centrifugation of extracts was done on IEC Centra-3E centrifuge (Bedford-
shire, England) and the samples were concentrated on a rotary vacuum evaporator,
Heidolph VV 2000 (Analis, Namur, Belgium). The concentrated extracts were com-
pletely dried in an oven (Belgolabo, Overijse, Belgium) and "nally weighed with an
analytical balance (Mettler-Toledo, Switzerland).

Procedure

Soxhlet extraction. The BuK chi extraction system o!ers four modes of extraction
(hot extraction, warm extraction, standard or classical Soxhlet and continuous mode).
Each of them consists mainly of three steps: extraction, rinsing and concentration,
which can be performed in stages for gradual solvent evaporation until near dryness.
Only hot extraction and standard Soxhlet have been used.
 COMPARATIVE STUDY ON TOTAL LIPID DETERMINATION 95

During hot extraction mode, the solvent is distilled into the extraction chamber,
while the upper heating element is turned on. The temperature of the extraction
chamber depends on the solvent mixture and, as a rule, should be 203C lower than the
boiling point of the mixture. The solvent is always kept above a "xed level by means of
an optical sensor, ensuring equilibrium between the rate of fresh solvent entering the
extraction chamber and solvent leaving the chamber. In this way, the sample is
permanently in contact with hot, but not boiling solvent. For comparison of methods,
time has been used instead of the usual cycles for standard Soxhlet system.
Five grams of solid sample were weighed in a thimble and subjected to Soxhlet
extraction with 80 mL of solvent; 5 mL of liquid sample were mixed with 10 g anhydrous
sodium sulfate before extraction. After extraction and evaporation of solvent, the beakers
were further dried at 1043C for 30 min, after which the sample weight was constant.

Bligh & Dyer extraction (Bligh and Dyer, 1959). To a 100 mL conical #ask contain-
ing 5 g or 5 mL of sample, 20 mL methanol (MeOH) and 10 mL chloroform (CHCl )

were added and the mixture was vortexed for 2 min. Ten millilitres CHCl were added

a second time and the mixture was shaken vigorously for 2 min. Eighteen millilitres of
distilled water were added and the mixture was vortexed again for 2 min. The layers
were separated by centrifugation for 10 min at 2000 rpm. The lower layer was
transferred to a pear-shaped #ask with a Pasteur pipette. A second extraction was
done with 20 mL 10% (v/v) MeOH in CHCl by vortexing for 2 min. After centrifu-

gation, the CHCl phase was added to the "rst extract. Evaporation was done with

rotavapor and the residue was further dried at 1043C for 1 h. Volume of solvents could
be adjusted to the sample size.

Modi,ed Bligh & Dyer extraction. This method is a modi"cation of the Bligh &
Dyer procedure in which MeOH is replaced by propan-2-ol and CHCl by cyclo-

hexane, respectively (Smedes, 1999). One key advantage of cyclohexane over CHCl

was its lower density, which consequently separated on top of the extraction mixture.
A mixture of water : propan-2-ol : cyclohexane (11 : 8 : 10) was used as solvent and
volumes could be adjusted to the sample size. A second extraction with 20 mL 10%
(v/v) propan-2-ol in cyclohexane was done. After centrifugation, the cyclohexane
phase was added to the "rst extract. Evaporation was done with rotavapor and the
residue was further dried at 1043C for 1 h.

Roese-Gottlieb extraction. This method consists of total lipid extraction in petro-

leum ether after hydrolysis of the bounded lipids (AOAC, 1990). To 1 g sample, 6 mL
of boiling water were added and the mixture was vortexed for 1 min and allowed to
cool. One millilitre of 25% ammonia solution was added and vortexed for 2 min, then
7.5 mL of MeOH were added to the mixture and vortexed for 2 min. To this last
mixture, 17 mL diethyl ether were added, shaken vigorously and "nally, 17 mL
petroleum ether were added and shaken strongly for extraction. The mixture was
centrifuged and the upper layer was transferred to a tarred beaker. Concentration was
performed with rotavapor and further drying at 1043C for 1 h.

RESULTS AND DISCUSSION

Optimization of Soxhlet

Extraction time. The Soxhlet system has been applied in hot and standard extrac-
tion mode for samples with high lipid content (e.g. margarine) to determine the
96 MANIRAKIZA E¹ A¸.

FIGURE 1. In#uence of time on standard and hot Soxhlet extraction e$ciency for margarine using
acetone}hexane (1 : 4) mixture: *䉬*, Hot Soxhlet; *E*, Standard Soxhlet.

optimum extraction time (Fig. 1). Two hours was a convenient extraction time to
obtain high lipid yields, while longer extraction times led to an insigni"cant increase of
the extracted material. Hot Soxhlet (upper curve) leads to higher lipid contents than
standard Soxhlet (lower curve) for the same extraction time (Fig. 1). After 2 h, the
extracted lipid amounts for margarine were 361 and 354 mg lipid/g with hot and
standard Soxhlet, respectively. The results are in good agreement with the value of
362 mg/g mentioned by the producer. Similar results were obtained for other samples
tested (chicken feed and "sh #our).

Extraction solvents. The yield of Soxhlet extraction is in#uenced by solvent com-

position (Fig. 2). Lipid content as determined by extraction with methanol (152 mg/g)
was more than two times higher than the content determined by extraction with less
polar solvents (60 and 79 mg/g for hexane and acetone : hexane"1 : 4, respectively).
This is most likely because non-lipids were extracted too. The amount of extracted
lipids as well as the amount of non-lipid material extracted increases with the polarity
of solvents used (de Boer, 1988). Higher results for lipid extraction using polar solvents
have been also observed by Campbell and Wells (1994).
Dichloromethane}hexane (1 : 4) and acetone}hexane (1 : 4) were the best mixtures
for lipid extraction by Soxhlet. Acetone}hexane (1 : 4) was preferred for safety reasons
and because this mixture led to an easy evaporation step through azeotropic distilla-
tion (Smedes and de Boer, 1997). However, more water is extracted from water-rich
samples, as already observed by de Boer (1988). This may explain the di!erence
between the results obtained for eggs when Soxhlet extraction was applied using
acetone}hexane (10% lipid recovery) and dichloromethane}hexane (30% lipid recov-
ery). Furthermore, both results were still very low, together with results for other
liquid samples (e.g., milk and serum), suggesting a decrease of the extraction e$ciency
by protein denaturation or a low extraction capability for bounded lipids.
Smedes (1999) worked on wet samples ("sh tissues) and, when using 30% acetone in
hexane as extracting solvent, he obtained poor yields ((75%) for samples with low
lipid content. Same "ndings were made by de Boer (1988). When using Soxhlet
extraction, de Boer (1988) advised to work on dried samples.
 COMPARATIVE STUDY ON TOTAL LIPID DETERMINATION 97

FIGURE 2. Hot Soxhlet extraction e$ciency of di!erent solvent mixtures for "sh #our (producer value of
75 mg lipid/g).

Evaluation of hot Soxhlet extraction mode. The use of hot Soxhlet allowed an
extraction time of 2 h for an e$cient lipid extraction for all solid matrices (Table 1).
When compared with the standard Soxhlet mode, no signi"cant increase in lipid
content was observed for liquid samples (eggs, milk and serum).
Furthermore, the e$ciency of Soxhlet extraction was higher for samples with a high
lipid content (de Boer, 1988). Indeed, hot Soxhlet extracted more lipids from pow-
dered milk (96%) than from liquid milk (48%), when compared with lipid contents
mentioned by the food producers. When compared with standard Soxhlet, hot
Soxhlet shows higher extraction e$ciencies (Fig. 2). The permanent contact with hot
solvent may be the reason of this success. The Soxhlet system used in this study
(BuK chi) has the advantage that the sample is always in contact with fresh solvent, while
in other systems (e.g., Foss Tecator), the thimble is directly immersed into the
extraction vessel containing the same boiling solvent. For "ne powdered samples (e.g.,
"sh #our), glass-"bre thimbles are needed because small particles pass the material
when using cellulose thimbles (the most commonly used thimbles).
By programming the whole procedure from the extraction step to drying, the
exposure of analysts to solvents is reduced when compared to liquid extraction
techniques. The total solvent volume used was not larger than 80 mL for Soxhlet or
liquid}liquid extraction methods.

Comparison of Di+erent Extraction Methods

Precision of each method was evaluated by the relative standard (x the average result
obtained from n determinations). The precision of all tested methods was good
(R.S.D.(3.5%), although accuracy of some methods was low for certain samples.
Extraction yields were calculated based on the food producer lipid contents, con-
sidered only as indicative value, although no information on the method used was
available.
The Roese}Gottlieb method was suitable for powdered milk (82% lipid recovery),
but not for liquid milk (30% recovery), while Soxhlet extraction was not found
suitable for water-rich matrices, showing lipid recoveries less than 50% (Fig. 3). Bligh
& Dyer and modi"ed Bligh & Dyer methods showed good results for all liquid
 98
TABLE 1
Comparison of lipid extraction parameters for the "ve investigated methods

Chocolate Milk Liquid Margarine Eggs Chicken Fish Human

powder powder milk feed #our serum

MANIRAKIZA E¹ A¸.
Reference 32 5 36 362 98 85 75 6
Methods value (mg/g)
Bligh & Dyer Mean 33 3 27 317 95.5 78.5 72 6
R.S.D. (%) 2.0 2.2 3.3 0.4 2.6 1.1 2.0 1.5
Modi"ed Mean 33.4 4.9 32 282 91.5 77.5 74 4.5
Bligh & Dyer R.S.D. (%) 2.3 0.4 0.5 0.4 1.1 1.6 1.2 1.2
Roese}Gottlieb Mean 25.6 4.1 11.2 337 41 73.5 62 1.4
R.S.D. (%) 2.5 1.1 1.4 0.7 2.4 1.6 0.5 1.3
Standard Mean 31 4.1 14.5 349 11 30 51 79 2.1
Soxhlet R.S.D. (%) 1.4 2.7 1.5 0.5 1.7 1.5 2.4 2.7 2.5
Mean 34.7 4.8 17.4 362 12.5 39 82.5 82 3.6
Hot Sohxlet R.S.D. (%) 2.1 1.7 2.3 0.5 1.2 2.0 1.1 2.4 2.3

Mentioned by the producer.

n"8.
Acetone : hexane"1 : 4.
DCM : hexane"1 : 4.
 COMPARATIVE STUDY ON TOTAL LIPID DETERMINATION 99

FIGURE 3. Extraction e$ciencies for each of the "ve investigated methods calculated based on the lipid
content provided by the producer: , Hot Soxhlet, , Standard Soxhlet; , Bligh & Dyer; , Modifed
Bligh & Dyer; , Roese-Gottlieb.

samples (milk, eggs, human serum), because MeOH and CHCl is a suitable mixture

for total lipid determination (neutral and polar lipids) (Smedes and Askland, 1999). By
using cyclohexane, the modi"ed Bligh and Dyer method reduces the solvent toxicity
and presents the advantage that the organic layer is separated at the top. This method
gave similar results as Bligh and Dyer method except for lower lipid recoveries from
human serum (75%). The di!erence in lipid extraction from human serum may be
explained by the fact that serum contains more polar lipids (phospholipids) and
propan-2-ol is less polar and has less solvating properties than MeOH.

CONCLUSIONS

Because of the inherent di$culties in the phase separation and the high number of
manipulations used in liquid}liquid extraction, the hot Soxhlet extraction may be
advised as a convenient automated method for lipid extraction from solid samples.
However, without speciation of lipids and non-lipids extracted from the samples, the
results should be interpreted with care.

ACKNOWLEDGEMENTS

The valuable suggestions of Dr Foppe Smedes during the preparation of the manuscript are greatly
appreciated.
100 MANIRAKIZA E¹ A¸.

REFERENCES

AOAC (1990). O.cial Methods of Analysis of the Association of O.cial Analytical Chemists (W. Horwitz,
Ed.), 15th edn., Method 989.05. Association of O$cial Analytical Chemists, Arlington, U.S.A.
Baily, S. K., Wells, D. E., de Boer, J., and Delbeke, K. (1994). The measurements of lipids as a co-factor for
organic contaminants in biota, QUASIMENE Project, Aberdeen, U.K., p. 1.
Bligh, E. G. and Dyer, W. J. (1959). A rapid method of total lipid extraction and puri"cation. Can. J.
Biochem. Physiol. 37, 911}917.
Booij, K. and Van den Berg, C. (1994). Comparison of techniques for the extraction of lipids and PCBs from
benthic invertebrates. Bull. Environ. Contam. ¹oxicol. 53, 71}76.
Campbell, D. and Wells, D. E. (1994). The measurements of lipids as a co-factor for organic contaminants in
biota, QUASIMENE Project, Aberdeen, U.K., p. 21.
Christie, W. W. (1976). ¸ipids Analysis. Pergamon Press, New York.
de Boer, J. (1988). Chlorobiphenyls in bound and non-bounded lipids of "shes; comparison of di!erent
extraction methods. Chemosphere 17, 1803}1810.
Delbeke, K., Teklemariam, T., de la Cruz, E., and P. Sorgeloos, P. (1995). Reducing the variability in
pollution data: The use of lipid classes for normalization of pollution data in marine data. Int. J. Environ.
Anal. Chem. 55, 147}162.
Gardner, W. S., Cichocki, E. A., and Parrish, C. C. (1985). Micromethod for lipids in aquatic invertebrates.
¸imnol. Oceanogr. 30, 1099}1105.
Ramos, L., Hernandez, L. M., and Gonzalez, M. J. (1998). Comparative study of methodologies for the
analysis of organochlorinated compounds in powdered full-fat milk. I. PCBs. Int. J. Environ. Anal. Chem.
71, 119}136.
Randall, R. C., Lee, H., Ozretich, R. J., Lake, J. L., and Pruell, R. J. (1991). Evaluation of selected lipid
methods for normalizing pollutant bioaccumulation. Environ. ¹oxicol. Chem. 10, 1431}1436.
Smedes, F. and de Boer, J. (1997). Determination of chlorobiphenyls in sediments}analytical methods.
¹rends Anal. Chem. 16, 503}517.
Smedes, F. and Askland, T. K. (1999). Revisiting the development of the Bligh and Dyer total lipid
determination. Mar. Poll. Bull. 38, 193}201.
Smedes, F. (1999). Determination of total lipid using non-chlorinated solvents. Analyst 124, 1711}1718.