444
H. M. Heise
glucose assays with different measurement techniques based on skin spectroscopy
[6, 7].
Some more information is allowed that the aqueous glucose spectrum was
obtained with scaled water absorbance subtraction, which leads to an incomplete
compensation due to the disturbance of the water hydrogen-bonding network by the
sugar molecules, evident from the spectral dips. Interestingly, the spectral features
of glassy sugars, as obtained from syrup samples after careful water evaporation,
show the same wavelength dependencies as the aqueous phase spectrum. Using this
technique, the otherwise opaque spectral intervals with strong water absorption are
now accessible, despite of some still existing uncompensated water absorptions in
the regions of the strong water bands. A spectrum of fructose is also displayed, which
may be relevant for diabetics in non-invasive glucose measurement after consumption of the other sugar, illustrating discrimination limitations based on the spectral
interval used.
For many years, our group has been involved in the development of in vitro
reagent-free multi-analyte systems for blood analysis based on vibrational spectroscopy and two selected papers, shedding light on the potential of near-infrared
spectroscopy for clinical chemistry, need to be discussed [8, 9]. In this context, also
various important chemometric aspects such as spectral variable selection or different
validation strategies for multivariate calibration modeling have been investigated. In
nearly all studies, partial least squares (PLS) calibration has been utilized, but also
science-based calibration (SBC) modeling was successfully tested for in vitro analysis. The latter approach uses explicitly the quantitative analyte spectral signatures,
while other spectral contributions are estimated statistically. For more details, the
reader is referred to our book chapter [9].
The usual strategy, applied by many research groups, has been used to demonstrate the spectroscopic assay capability for in vitro analysis by having aqueous
mixtures of a few compounds such as albumin, glucose, lactate and others. It is
certainly advantageous that an elaborate experimental design can be followed to
cover an adequate physiological composition with a reasonable spread of differing
concentrations, which can be most accurately prepared, for example, by gravimetry.
Samples from patients are more complex in composition and with a distribution of
analyte concentrations that is usually defined by the natural spread. Such specimens
can be selected, for example, from a population of healthy and/or diabetic subjects
if blood glucose assays have to be tested and the assay will focus on such kinds
of subjects. In some cases, also spiking of samples has been taken into account for
increasing the concentration levels or the distribution of concentration values. The
assays for obtaining reference concentrations are often limited in precision and accuracy, unless very special effort is undertaken for improved clinical analytics. Usually,
such testing scenarios are carried out under special experimental conditions with a
tightly thermostated sample cell of constant thickness, which is important especially
for near-infrared spectroscopy of aqueous biofluids because of the hydrogen-bonding
network of the water molecules that is very sensitive to temperature changes (see
again Fig. 20.1).
H. M. Heise
glucose assays with different measurement techniques based on skin spectroscopy
[6, 7].
Some more information is allowed that the aqueous glucose spectrum was
obtained with scaled water absorbance subtraction, which leads to an incomplete
compensation due to the disturbance of the water hydrogen-bonding network by the
sugar molecules, evident from the spectral dips. Interestingly, the spectral features
of glassy sugars, as obtained from syrup samples after careful water evaporation,
show the same wavelength dependencies as the aqueous phase spectrum. Using this
technique, the otherwise opaque spectral intervals with strong water absorption are
now accessible, despite of some still existing uncompensated water absorptions in
the regions of the strong water bands. A spectrum of fructose is also displayed, which
may be relevant for diabetics in non-invasive glucose measurement after consumption of the other sugar, illustrating discrimination limitations based on the spectral
interval used.
For many years, our group has been involved in the development of in vitro
reagent-free multi-analyte systems for blood analysis based on vibrational spectroscopy and two selected papers, shedding light on the potential of near-infrared
spectroscopy for clinical chemistry, need to be discussed [8, 9]. In this context, also
various important chemometric aspects such as spectral variable selection or different
validation strategies for multivariate calibration modeling have been investigated. In
nearly all studies, partial least squares (PLS) calibration has been utilized, but also
science-based calibration (SBC) modeling was successfully tested for in vitro analysis. The latter approach uses explicitly the quantitative analyte spectral signatures,
while other spectral contributions are estimated statistically. For more details, the
reader is referred to our book chapter [9].
The usual strategy, applied by many research groups, has been used to demonstrate the spectroscopic assay capability for in vitro analysis by having aqueous
mixtures of a few compounds such as albumin, glucose, lactate and others. It is
certainly advantageous that an elaborate experimental design can be followed to
cover an adequate physiological composition with a reasonable spread of differing
concentrations, which can be most accurately prepared, for example, by gravimetry.
Samples from patients are more complex in composition and with a distribution of
analyte concentrations that is usually defined by the natural spread. Such specimens
can be selected, for example, from a population of healthy and/or diabetic subjects
if blood glucose assays have to be tested and the assay will focus on such kinds
of subjects. In some cases, also spiking of samples has been taken into account for
increasing the concentration levels or the distribution of concentration values. The
assays for obtaining reference concentrations are often limited in precision and accuracy, unless very special effort is undertaken for improved clinical analytics. Usually,
such testing scenarios are carried out under special experimental conditions with a
tightly thermostated sample cell of constant thickness, which is important especially
for near-infrared spectroscopy of aqueous biofluids because of the hydrogen-bonding
network of the water molecules that is very sensitive to temperature changes (see
again Fig. 20.1).