Clinical Chemistry
309-3
12
42:2
Automation and
Analytical Techniques
(1996)
Qu antification of fecal carbohydrates by
near-infrared reflectance analysis
J URGEN
STEIN,*
BERND
PURSCHIAN,
STEFAN
WOLFGANG
F.
Established
methods
for quantitative
analysis of fecal carbohydrates
(CHO) are time consuming,
require
unpopular
sample handling,
and are therefore
rarely performed.
The
aim of this study was to evaluate the efficiency,
validity, and
practicability
of near-infrared
reflectance
analysis (NIRA),
compared
with standard methods for measurement
of fecal
TERMS:
malabsorption
.
nutritional
CASPARY
Materials
and Methods
NIIRA is based on the matrix- and substrate-specific
relation
between the reflectance
intensity diffused by the fecal sample
surface at a specific wavelength
and the composition
of the
sample. Each component
to be measured has typical functional
groups (e.g., CH, NH, OH) possessing
specific absorption
bands in the near-infrared
area (700-2500
nm). The spectroscopic response (reflectance)
from a fecal sample is related to the
concentration
of the compounds
(functional
groups) as follows
[8, 9]:
status
x
Z
-
f1 log Rf1
flog
-
Rf
-
... -
f,, log Rf
(1)
where xis the concentration
of the analyte, Rf, is the reflectance
for the filter n, f is the scaling factor for each filter, and Z is a
constant of bias correction.
The wavelength
and corresponding
spectral structure
for each filter are given in Table 1. Scaling
factors corresponding
to each selected filter were calculated by
multilinear
regression
[10, 11] of results obtained from a training set consisting
of samples analyzed by the routine chemical
method covering the whole concentration
range for fecal CHO.
The method based on this principle was described by Peuchant et al. [12]. In brief, the sum of squares of the residual
values (differences
between chemical method values and calculated values) must be as low as possible. The computer
assessed
every set of wavelength
combinations
from 1 to 12, through
spectroscopic
and statistical data, based on the use of a t-test
where t = f/SD of F.
Each combination
of wavelengths
was characterized
by the
coefficient of multiple correlation
(r) between the chemical and
calculated values:
of Internal Medicine, Johann7, D-60960
Frankfurt
70,
Germany.
Author for correspondence.
Received April
and
LEMBCKE,
PRINCIPLE
Carbohydrates
(CHO) are the major source of human energy
requirement.
Although ordinarily most of the ingested CHO are
completely
absorbed
before reaching the colon, several disorders can result in impairment
of absorption
from the small
intestine [1-3]. There being no quantitative
fecal CHO equivalent to the quantitative
estimation
of fecal fat, several laboratories use semiquantitative
screening
tests to confirm clinical
suspicion of CHO malabsorption.
Established
methods
for the
quantification
of fecal CHO [2, 3] require collecting stools for
24-72 h and unpopular
sample handling (e.g., extraction,
distillation). Therefore,
use of such tests has been mainly confined
to specialized laboratories.
Fat, water, and nitrogen in stools have recently been measured by near-infrared
reflectance
analysis (NTRA), a method
based on the measurement
of the scattered radiation in the near
infrared by the surface of a sample. The analysis is performed
in
Division of Gastroenterology,
Deparunent
Wolfgang-Goethe-University,
Theodor-Stern-Kai
BERNHARD
a very short time (<1 mm), without further processing
of the
feces and without the use of chemical reagents [4-7].
Because of the importance
of an accurate
and feasible
measurement
of the fecal CHO output in clinical gastroenterology, we have evaluated the efficiency of NIRA in comparison
with “wet” chemical methods for determining
fecal CHO.
CHO. Excretion
of fecal CHO was cross-validated
spectrophotometrically
with the anthrone
method.
Fecal CHO
concentrations
ranged from 2.7 to 24.5 g/kg wet weight.
Methods
comparison
showed
linear regression
over the
entire
range of diagnostic
relevance
with a correlation
coefficient
of 0.869 (S
±0.31).
Repeated
measurements
from the same stool collections
obviated
the need for
homogenization.
These results indicate that NIRA may be
a new, reliable, and accurate test in the diagnosis
of CHO
malabsorption.
INDEXING
ZEUZEM,
-
Fax 0049-69-6301-6448.
25, 1995; accepted October II, 1995.
r-
309
(1
-
S,,.(n
SD2(n-
-
k 1)
1)
(2)
310
Stein et al.: Fecal carbohydrates
Table 1. Spectral structure and results of calibration with
selected filters.
Specific
Filter
no.
corresponding
Wavelength, nm
1
2336
5
2208
7
9
10
11
Scaling
structure
-CH3;
>CH2
-CH-CH2
0
>C=CH2
2100
1940
1818
1680
-OH
-CH
>C=CH2;
-CH
factors
t-test
-15.48
15.38
-5.69
6.04
3.47
21.34
-19.48
-5.19
3.39
8.29
-7.05
-4.20
where S is the standard error of estimate, n is the number of
samples, and k is the number of filters; and by the constant of
Fisher (F-ratio), which indicates the quality of the regression:
F-ratio
r(n
=
k
-
(1
-
-
r2)k
1)
(3)
As the number
of wavelengths
in the combination
increased, r
approached
1 and the F-ratio increased to >50.
For the training set, 34 stool samples from patients with
various diseases and following various treatments
were selected.
APPARATUS
Experiments
were performed
with an Esetek infrared analyzer
(Model Fenir 8820, interference-filter-based
reflectance
spectrometer;
Stimotron,
Wendelstein,
Germany)
connected
to an
IBM-compatible
computer. The optical information
is provided
by a constantly
rotating
filter wheel fitted with 12 narrow
(10-nm) bandpass filters of fixed wavelength
specification
ranging from 1218 to 2345 nm (Table 1), which makes it possible to
measure reflectance (R) as the ratio of reflected energy from the
sample and the incident radiation.
R was measured at 1 to 12
wavelengths,
converted
into log R, and entered into the computer.
Because
near-infrared
reflectance
measurements
are
highly dependent
on the amount
of scattering
materials and
pathlength,
the 1940-nm water band was used to compensate
for
differences
in effective pathlengths
113].
The manufacturer’s
software program (Perten, Rome, Italy)
was used to select the best combination
of filters from a group
of filters that were highly correlated
spectroscopically:
Wavelengths with clearly assignable spectral responses were included
and those without were excluded. Wavelengths
with low constants and poor correlations
were excluded on statistical grounds
by use of Student’s
t-test. The program
functioned
with a
combination
of 1 to 12 wavelengths,
giving for each combination the values of constants
corresponding
to each filter,
values of the correlation
coefficient r, and the F-ratio.
the
quantified
gestion
creatitis
sprue).
by infrared
reflectance
(exocrine pancreatic
insufficiency
due to chronic panor cystic fibrosis, inflammatory
bowel disease, celiac
A subgroup
of extensively
investigated
patients with
functional
abdominal
pain, but with entirely normal gastrointestinal function, served as controls. In each case, final diagnoses
were established
after thorough investigation
according to generally accepted criteria. The study was in accordance
with the
ethical standards of the local committee.
ANALYSES
Stool was collected for 72 h in plastic containers
and analyzed
either on the same day or stored at -24 #{176}C
until analysis. For
NIRA, a stool sample of 2-3 g was homogenized
(UltraturaxBlender; Janke & Kunkel, Staufen, Germany),
then placed in a
disposable
cup and positioned
on the analyzing
site of the
infrared analyzer. Within a given spot sample, Ineasurements
were read at two distinct positions. While the selected filters in
turn
automatically
rotate into the light path, reflected spectral
energies from the sample are recorded by a microprocessor,
and
results (% substrate content) are displayed or printed in <1 mm.
Samples that had been frozen were thawed, homogenized
as
above, and centrifuged
at S000g for 10 mm. The supernatant
was filtered through
10-Mm (pore size) filters (Millipore,
Bedford, MA) before assay.
For comparison
we measured
fecal total CHO by using
anthrone
in a modification
of the spectrophotometric
method
[14, 15]. This method,
first described
by Dreywood
[16], is
based on the appearance
of a characteristic
blue-green
color
(A,,,,5 = 680 nm) due to formation
of a hexose-anthrone
complex in the presence of sulfuric acid. The method is highly
specific for all CHO (>95%),
but is not sensitive to organic
acids that do not include CHO [5J. The mean CV of this assay
after 10 repeated measurements
was 4.8%.
STATISTICS
Unless otherwise
SE. For statistical
stated, the results were expressed as mean ±
analysis, Student’s unpaired t-test was used.
Results
Table 1 shows the optimum set of wavelength
combinations
for
fecal carbohydrate
analysis.
Precision. On the same day the stool homogenate
was prepared,
10 separate aliquots of itwere analyzed.The resultingintraassay
CV obtained was 4.8%. Frozen aliquots from the same homogenate analyzed on 6 different days yielded an interassay CV of
6.9%. To determine differences between fecal CHO in homogenized stools and of spot samples from unhomogenized
stools,
we performed
measurements
on multiple readings from unhomogenized
stools. The mean difference decreased from 2.4 g/kg
wet weight (one spot sample) to 1.8 g/kg wet weight (five spot
samples) (Fig. 1).
PATIENTS
From January 1990 until December
1991 we studied 74 patients
(39 males, 35 females) of all ages (3-71, mean 32 years) who had
been admitted to our hospital (a) for the differential
diagnosis of
malahsorption
syndromes (history of diarrhea, weight loss) or (b)
with previously established
causes of malabsorption
or maldi-
Calibration and validation. Comparison
of the concentrations
measured by the anthrone
and the infrared method training set
in 34 stool samples (calibration
set) showed r = 0.962 and S =
1.65 g/kg wet weight (Fig. 2A). For the validation data, shown in
Fig. 2B, r = 0.869 and S,,. = 3.1 g/kg wet weight. The residual
Clinical Chemistry 42, No. 2, 1996
311
3
2
ci)
E
z
1
.00 0.75 1.50 2.25 3.00
nnn
n=1
n=2
n=number
n3
n4
Difference from mean value (glkg wet weight)
n5
of spot samples
Fig. 1. CHO measured by NIRA is as many as five spot samples
from
a singlestool,shown as the difference
between the mean valuefrom
the spotsamples and thevaluefrom the whole homogenized stool.
values (differences
between results obtained by the two methods) are shown in Fig. 3.
The mean CHO concentration
was 6.5 ± 3.4 g/kg wet
weight (range 2.7-24.5 g/day) by the NTRA and 5.3 ± 2.4 g/kg
wet weight (range 2.5-19.3 g/day) by the anthrone method, not
significantly
different (P >0.05).
Recovery. An aliquot of stool homogenate
content.
Three
different
amounts
was analyzed for CHO
of lactulose
solution
were
A
A
Fig. 3. Histogramofthe residual
valuesbetween resultsby NIRA and
by the anthronemethod forCHO assay.
added to the homogenate,
in amounts yielding CHO-to-stool
ratios of 3, 2, and 1 g of CHO per 1.0 g of stool. Samples were
then rehomogenized
and several aliquots of each homogenate
analyzed.
The recovery
of CHO was 98.7-102.5%
(n = 6)
regardless
of the amount of CHO added, suggesting
that the
method can be used to determine
both high and low quantities
of CHO excretion.
Stability of stool CHO. To
determine
the effect of bacterial
degradation
on stool CHO content in the time between passage
of stool and analysis, we added a known amount of lactulose
solution to fresh stool homogenate
and then rehomogenized
as
above. Freezing at -24 #{176}C
for 72 h was chosen to evaluate the
effect of storage during the collection period. Three aliquots of
stool, stored at this temperature
and analyzed every 72 h over a
2-month
period, yielded stable results (Fig. 4).
Discussion
10.0
Carbohydrates
20.0
(g/kg wet weight)-Anthrone
B
The importance
of quantitative
fecal CHO output measurement
in the diagnosis of malabsorption
syndromes
has been recognized in several previous
studies [1-3]. However,
traditional
methods
for quantifying
fecal CHO are complex
and time
consuming
and require unpleasant
sample handling.
Our data indicate that NIRA is a valuable practical alternative
to conventional
chemical methods
for measuring
fecat CHO.
The absence of any potentially
hazardous
reagent (therefore
242016
‘1)
12
10.0
20.0
Carbohydrates(glkg wet weight)-Anthrone
Fig. 2. Correlation
between fecalCHO concentration
(g/kg wet wt)
measured by NIRA and the anthrone method: (A) calibration data (y =
0.5+0.9x, r = 0.869);(B) validation data [y = 0.5 (±0.04) + 0.9
(±0.07)x,r = 0.87).
RangeofCHO concentrations 0.7-26.7 g/kg wet wt.; S = 1.6, ratio = 417.
#{149}
in 8, highvalues
frompatients
with lactulose-induced
diarrhea.
0
.0
Co
C)
-o
--O------O----O-
8
4.
-
-.------
0
10
20
30
U
40
50
60
days
Fig. 4. Stability ofthreedifferent
stoolsamples analyzedforcarbohydrate content during storage at -24 #{176}C
over 2 months.
312
Stein et al.: Fecal carbohydrates
excluding the need for associated precautions),
homogenization,
extraction,
distillation,
and other onerous procedures
allows this
method to be applied in any hospital laboratory,
not just in
specialized
centers. Moreover,
the short time required for the
analysis by NIRA (<1 mm) provides
immediately
available
results. Finally, comparison
of NIRA in homogenized
stools and
the traditional
anthrone
method
gives a precise correlation,
implying that NIRA is as accurate as the conventional
methods.
Repeated
analysis on different
portions
of the same stool
collection,
performed
in a few minutes, can avoid homogenization, usually the most unpleasant
part of the classical methods.
Five readings reduce the difference from the posthomogenized
value to <7%, but for clinical application
three readings may
suffice.
We emphasize that the use of NIRA is based on calibration
with a chemical method, which is the first and indispensable
step
when starting the measurements.
This means that the reflectometric results depend on the availability and accuracy of the
chemical methods used in the individual laboratory
as well as on
the choice of the biological matrix. Therefore,
NIRA cannot be
“more accurate” with respect to the true value than the reference
chemical methods are. Benini et at. f17J previously demonstrated
that changes of stool matrix (e.g., in liquid stool) may influence
the results of nitrogen analysis. In such cases, one should adjust
the calibration
with an adapted stool matrix or change the
chemical
method.
In this study, we did not focus on these
possibilities,
but also saw no indication from our patient groups
that this might be clinically relevant.
Further, as Benini and colleagues found for fat and nitrogen
/17], NIRA can underestimate
fecal CHO.
This is of little
consequence
for a clinical screening
method, but applying the
NIRA method to nutrient balance studies could lead to systemic
errors. The difference
between
reflectrometric
and chemical
values can also be partly explained
by the imprecision
of the
chemical methods. So Grimble et al. / 18] found that the Kjeldahl
method underestimated
fecal nitrogen because of acid digestion
of refractive nitrogen.
Similarly, the anthrone method can fail to
detect all CHO if the appropriate
extraction solvent is not used.
Thus, by using different
chemical
methods,
one can expect
better correspondence
between
NIRA values and chemical
values.
For routine purposes in a clinical setting, we consider NIRA
a valuable and precise alternative to the anthrone method.
Taking into account the simultaneous
measurement
of fat,
nitrogen, and water justifies the current instrumentation
price of
-$30 000 (US), which could be reasonable
according
to the
number of analyses performed;
also, only one trained analyst is
required.
We conclude that NIRA represents a useful replacement
for the
conventional
laboratory
methods in the investigation
of malabsorption.
In the future, NIRA may offer a new and powerful
diagnostic
approach
to rnalabsorption
syndromes,
by simultaneous determination
of several variables, such as dry weight,
fecal fat, and fecal nitrogen [6, 7], in addition to CHO reducing
quantified
by infrared
sugars, for evaluating
of the
reflectance
the digestive
and (or) absorptive
function
gut.
We thank the nursing staff of our department
for their helpful
cooperation
and C. Cardinale
for providing
us with the NIRA
instrument.
Some of these data were presented at the American
Gastroenterological
Assn meeting, Boston, May 11-16, 1993.
References
1. Caspary WF. Diarrhea associated with carbohydrate malabsorption. eon Gastroenterol
1985;9:631-55.
2. Ameen VZ, Powell GK, Jones A. Quantification of fecal carbohydrate excretion in patients with short bowel syndrome. Gastroenterology
1987;92:493-500.
3. Hammer HF, Fine KD, Santa Ana CA, PorterJL, SchillerLR,
FordtraniS.Carbohydratemalabsorption.
J ClinInvest1990;86:
1936-44.
4. Peuchant E,SallesC, Jensen R.Valueofa spectroscopic
“fecalogram”
in determining
the etiology
of steatorrhea.
Clin Chem
1988;34:5-8.
5. Benini L, Caliari
5, Guidi GC, Brentegani MT, Castellani G,
Sembenini C, et al. Near infrared spectroscopy for fecal fat
measurement:comparisonwithconventional
gravimetric
and titrimetric methods. Gut 1989;30:1344-7.
6. Stein J, Purschian B, Caspary WF, Lembcke B. Validation
of
near-infrared reflectance analysis (NIRA) for assessment of fecal
fat, nitrogen and water. A new approach to malabsorption syndromes [Abstractl. Gastroenterology 1992;102:243.
7. Stein J, Purschian B, Bieniek U, Caspary WF, Lembcke B. Nearinfrared reflectance analysis (NIRA): a new dimension in the
investigation of malabsorption syndromes. Eur J Gastroenterol
Hepatol1994;6:889-94.
8. Wetzel D. Near infraredreflectance
analysis.Sleeper among
spectroscopic
techniques.AnalChem 1983;55:1165-71A.
9. Nyquist RA, LeugersMA, McKelvy ML, Papenfuss RR, PutzigCL,
Yurga L.Infrared
spectroscopy.
AnalChem 1990;62:223-55R.
10. DraperNM, Smith H. Appliedregressionanalysis,
2nd ed. New
York: Wiley. 1981.
11. Mark H,Workmann J.Effect
ofrepackon calibrations produced for
near-infrared
reflectance
analysis.
AnalChem 1986;58:1454-9.
12. Peuchant E, Salles C, Jensen R. Determination of serum cholesterolby near-infrared reflectance spectroscopy. Anal Chem 1987;
59:1816-9.
13. GeladiP. MacDougall D, Martens H. Linearization
and scattercorrection
for near-infrared
reflectance
spectraof meat. Appl
Spectrosc1985;39:491-500.
14. MorrisDE. Determinationof dextranwith anthrone.Anal Chem
1953;25:1656-61.
15. Ameen VZ, PowellGK. A simplespectrophotometric
method for
quantitative fecalcarbohydratemeasurement. ClinChim Acta
1985;152:3-9.
16. Dreywood R. Qualitative test for CHO material. Ind Eng Chem (Anal
Ed) 1949:18:499-504.
17. Benini L, Caliari S. Bonfante F, Guidi GC, Brentegani MI, Castellani G, et al. Near infrared reflectance measurement of nitrogen
faecallosses.Gut 1989;33:749-52.
18. Grimble GK, West MF, Acuti AB, Rees RG, Hunian MK, Webster JD,
et al. Assessment of an automated chemiluminescence nitrogen
analyzer for routine use in clinical nutrition. J Parenter Enter Nutr
1988;12:100-6.