Originally Published IVD Technology November/December 2003
DETECTION TECHNOLOGIES
Reagent-free clinical analysis and diagnostics: Laboratory medicine in a new lightInfrared spectroscopy technology enables clinical analytical labs to carry out common serum and urine assays without reagents.
R. Anthony Shaw, Sarah Low Ying, Angela Man, and Kan-Zhi Liu
The March issue of IVD Technology led off with an editorial entitled, "Staying Committed to R&D," in which Richard Park highlighted the need for a company's research and development to stay abreast or ahead of the competition. He observed that innovation often arises not by looking within the industry, but by paying close attention to developments from less familiar territory. In this article, such an advance is described.
) |
| Technology designed for microtiter plate processing provides the foundation for IR-based reagent-free clinical laboratory analyses. (click to enlarge) |
A reagent-free technology is available that offers the possibility of accurate, cost-effective analytical and diagnostic testing. Infrared (IR) spectroscopy has long been established as a chemical identification technique; every molecular species provides a unique IR spectroscopic fingerprint. In contrast, visible spectroscopy has traditionally been the method of choice for quantitative tests. As a result, many analytical methods rely upon reagents to serve two functions. The first, molecular recognition, requires that a reagent recognize and react with only the analyte of interest. The second, quantitation, calls for the production of color that is proportionate to the concentration of the analyte.
Capitalizing on the sensitive and stable IR spectrometers currently available, a growing wave of analytical methods uses the specificity of the IR spectrum as the basis for analysis. A sample's absorption profile is a superposition of the spectra of the sample's components. Therefore, even the spectra of very complex mixtures provide the basis for quantitative analysis of the major constituents of a sample. Analytical IR spectroscopy has grown and flourished as an industry unto itself. It has been used to quantify alcohol in beer, determine active ingredient levels in pharmaceutical preparations, and monitor chemical reactions in real time.
This article summarizes the roles that IR spectroscopy can play in the clinical laboratory, namely performing many common serum and urine assays without the need for reagents. Both research and development activities are discussed; the former demonstrating the viability of the technique, and the latter clearing the way for practical implementation of the technique both in high-volume and niche applications (e.g., point-of-care testing or cardiac risk panel). Finally, in light of evidence that the IR spectra of serum, blood, cells, and tissues may serve directly as the basis for diagnostic testing, examples are introduced, and the future prospects discussed.
Making Light Work
IR spectroscopy grew out of the discovery of "invisible light" by Sir Frederick William Herschel in 1800. With a prism set at his window to refract sunlight, Herschel moved the bulb of a mercury thermometer through the visible spectrum produced by the prism. As the bulb passed through the spectrum, Herschel recorded the thermometer readings. As an astronomer, his aim was a practical one—to determine which spectral region produced the most heat, and hence discomfort to the astronomer's eye—but curiosity provoked him to measure the temperature at a position beyond the red end of the visible spectrum. To his surprise, the temperature was even higher there than it was at any position in the visible spectrum. He coined the term, "calorific rays" for this portion of the spectrum, which we now refer to as the IR region.
Most samples routinely encountered are somewhat cooler than and generate less energy than the sun. Therefore, it is more convenient to transmit IR light through samples and measure their characteristic spectrum of absorptions rather than their emissions. This spectrum provides a basis for identifying the sample and for quantifying its constituents, if it is a mixture.
) |
| Figure 1. Infrared absorption spectra of a dry serum film and three serum constituents. Each major serum constituent yields a unique IR spectral fingerprint that contributes to the overall IR absorption profile of the mixture containing that component, and hence provides the basis for its quantitation. (click to enlarge) |
To shed light on the principles underlying IR spectroscopy, the IR absorption spectra of a human serum sample and some of its constituents are presented (see Figure 1).
Since serum protein levels are typically around 70 g/L, and the next most concentrated species (i.e., glucose, triglycerides, cholesterol, and urea) appear in the range of 0.5–4 g/L, the spectroscopic contributions of the latter analytes are superimposed on the dominant profile of the protein constituents. All proteins produce similar IR absorption patterns because the patterns arise from the repeating amide linkage within the proteins. The serum spectrum and albumin spectrum are therefore very similar. The spectrum of albumin was chosen for comparison because it is the most abundant serum protein, and the most pronounced absorptions are representative of those contributed by all serum proteins.
The retrieval of quantitative analytical information from IR spectroscopy requires both good spectra and the means to convert the spectroscopic information into analytical results. These requirements have been met by two technical developments, namely the use of spectrometers that are stable enough to provide high-quality and reproducible spectra, and the implementation of widely available software and computing power that allow for the routine development of quantitation algorithms. Metaphorically speaking, these algorithms serve as the black boxes that yield analyte levels from spectra.
IR spectroscopy offers many benefits when used as an analytical method. One advantage is that several analyte levels can be recovered from a single IR spectrum. No reagents are required, so the method saves users the cost of reagents and reagent storage. Another plus is that linearity is routinely available over the full range of analyte levels. In addition, the method is suitable for automation.
The strengths of this analytical approach are most fully realized in high-throughput applications where reagent costs can add up, or in specialized assays that otherwise require significant labor or other investments. The following summarizes proof-of-concept examples that establish a potential role for IR spectroscopy in both clinical analytical and diagnostic work.1,2
Clinical Analysis
) |
| Figure 2. The analytical performance for eight serum assays. Each scatterplot compares quantitative analyses, as provided by IR spectroscopy, to the "true" values, as provided by standard clinical analytical methods. All eight analyses are available from a single measurement via the IR spectrum, carried out using approximately 10 ml of serum, and without any reagents.3,4 a. Data in mmol/L. b. Data in g/L. (click to enlarge) |
IR spectroscopy can be employed to perform multiple quantitative clinical analyses on one sample simultaneously, without reagents (see Figure 2).3,4 The use of IR spectroscopy has proven suitable for key serum, urine, and whole blood assays, as well as for certain niche applications such as fetal lung maturity testing via amniotic fluid analysis.5–7 A summary of proven IR-based methods is provided (see Table I).
When surveying proven methods of IR-based spectroscopy for biofluid analysis, the most obvious feature shared by the serum analytes is that they are the most abundant species in serum; they are also among the most commonly analyzed via other methods. Therefore, one potential role for IR spectroscopy is its ability to carry out high-volume, repetitive serum analysis at a fraction of the present cost, while maintaining the accuracy available from the routine methods currently in use.
A second prospective clinical role for IR spectroscopy is in point-of-care testing. For example, since both glucose and urea levels are available from the IR spectrum of whole blood, IR spectroscopy could be used routinely in various hospital settings, e.g., in the emergency room, operating theatre, dialysis unit, or on wards where glucose tests are common.6 One possibility for implementation is in the neonatal ward, where the option of low-sample-volume testing is particularly attractive.
| Sample | Measure |
| Serum | Urea, glucose, total protein, albumin, triglycerides, total cholesterol, HDL cholesterol, LDL cholesterol. |
| Urine | Urea, creatine, protein. |
| Whole blood | Urea, glucose. |
| Amniotic fluid | Lecithin/sphingomyelin ration and surfactant/albumin ration (both fetal markers of lung maturity), lactate, glucose. |
| Table I. Viable analytical methods based upon IR spectroscopy. | |
The amniotic fluid example illustrates a case in which the sensitivity to lipid levels offered by IR spectroscopy proves to be particularly advantageous. Ongoing studies at the Institute for Biodiagnostics in the National Research Council of Canada (Winnipeg, MB, Canada) are aimed at capitalizing on this advantage by broadening the range of serum lipid assays. One goal of this research is to establish an IR-based LDL cholesterol method that is effectively a direct analysis, thereby providing accurate levels in those cases for which the Friedewald approximation fails.
A key characteristic of IR-based analysis is that while the measurements of several analyte levels may be available from the same IR serum spectrum, each requires a different quantitation algorithm. Another distinguishing feature is that IR-based analysis is a secondary method; an established analytical technique is required to provide the true analyte levels that provide the basis to train the IR-based method(s).1 An overview of the procedure used to derive a new IR-based analytical method is summarized (see Sidebar).
Method development begins by acquiring spectra for a set of approximately 200–300 samples, together with the corresponding quantitative analyses for the component(s) of interest (as provided by established analytical methods). The samples are then divided into a training set comprising approximately two thirds of the available samples and a test set made up of the remaining third of the samples. For each analyte of interest, a quantitation algorithm is derived via a regression analysis routine, using a partial least-squares approach that requires the set of training spectra and corresponding analytical levels as input. As a final gauge of its accuracy, the newly derived algorithm is used to predict analyte levels for the set of test samples. The predicted analyte levels are then compared to the true values (see Figure 2).
This analytical approach allows for a unique advantage. The IR spectrum is not measured for the native serum (or for native urine or other samples), but rather for a residual film that remains after evaporation of the constituent water. Therefore, samples may be dried and then shipped to a remote location for measurement. This process minimizes the shipping costs and concerns associated with native fluids.
Diagnostic Testing
Quantitative clinical analytical tests are carried out in order to provide diagnostic information. The IR spectrum of serum, or of any other biological sample, must be explicitly converted to the language of quantitative chemical analysis in order to generate such information.
A number of recent studies have explored the possibility of recovering specific diagnosis directly from the spectroscopic fingerprints. For example, evidence suggests that a classification algorithm can be used to generate diagnostic patterns that distinguish serum specimens of diabetic subjects from those of nondiabetic subjects. This approach to translating data derived from the spectra may yield useful diagnostic classifiers for a variety of disorders.
For instance, a classification scheme has been developed to distinguish type 1 diabetics, type 2 diabetics, and healthy donors based upon the IR spectra of their serum specimens.8
The approach used to develop these diagnostic tests parallels that used to develop analytical methods (see Sidebar). First, the spectra of a large number of samples are accumulated for each of the disease categories of interest. An appropriate algorithm is then employed to identify unique attributes of each group of spectra, thereby distinguishing each diagnostic category.
One such scheme is the genetic algorithm/optimal region selection approach. This approach was originally developed and implemented at the Institute for Biodiagnostics for the diagnostic classification of magnetic resonance spectra, and has since been adopted for the classification of IR spectra.9 Following this scheme, each spectrum is condensed to a set of 5–15 regions that, when presented together, provide the basis for diagnostic classification.
) |
| Figure 3. Spectral regions optimal to distinguish the serum IR spectra of type 1 diabetics from the spectra of control sera. The spectral information carried within these restricted regions provides greater diagnostic power than is available from the spectra as a whole. The discovery of diagnostic methods based upon IR spectroscopy (or indeed other forms of spectroscopy) capitalizes on the judicious use of software packages to identify diagnostic patterns that rise above the natural background variability among spectra.8 (click to enlarge) |
The result of this process is graphically illustrated (see Figure 3). This illustration highlights the seven spectral segments that emerged as the basis for discriminating, with sensitivity and specificity of approximately 80%, between specimens from type 1 diabetics and those from control donors.8 It is interesting to note that none of these spectral regions coincides with major glucose absorptions. The same general approach has been used to identify rheumatoid arthritis patients based upon serum IR spectra, to diagnose arthritis using spectra of synovial fluid, and to gauge kidney rejection status from the spectra of urine specimens.10–12
Parallel research efforts are being directed toward the development of IR-based cytology and histology tests. While conventional cytology and histology are largely grounded in the detection of abnormalities in cellular shape, IR spectroscopy yields complementary diagnostic indicators originating from the chemical make up of tissues and cells.
For example, one study revealed spectroscopic features that are characteristic of chronic lymphocytic leukemia cells.13 In that instance, the DNA absorption patterns were altered in a characteristic way compared with those of normal lymphocytes. In addition, spectra of brain tissue have been shown to carry signatures characteristic of plaques from patients with multiple sclerosis and patients with Alzheimer's disease.14,15 In both cases, the hallmark signatures originated with characteristic changes in the secondary structure of certain brain proteins. Brain cancer (astrocytoma) leaves a unique imprint on the spectra, originating from changes in the nature or relative amounts of brain lipids.16
While research yields both analytical and diagnostic methods, as well as the promise of more new methods to come, a parallel development effort has paved the way to the implementation of these methods in the clinical laboratory.
Implementation
The adoption of IR-based clinical analytical and diagnostic methods for routine use in the clinical arena depends on the availability of practical sampling methods and hardware that are well suited for routine use. The method used at the Institute for Biodiagnostics involves first drying the sample of interest to convert the dissolved constituents to a residual film, and then acquiring the IR transmission spectrum of that film.
This approach eliminates the strong absorptions of water that can otherwise make spectral measurement cumbersome. One drawback to the drying approach is that the water-resistant barium fluoride or calcium fluoride IR-transparent substrates onto which the sample is dried tend to be very expensive. As an alternative to these specialty materials, the use of silicon wafers intended for (but rejected by) the semiconductor industry has been explored. These wafers are both transparent to IR light and very inexpensive. Another practical benefit is that they can be cut to any size desired, from small slides suitable for a single sample to larger plates accommodating a hundred samples or more.
These wafers can be masked to demarcate wells of consistent size and shape, thereby ensuring that each sample spreads over an equal area. The methodology used in the Institute for Biodiagnostics' proof-of-concept instrumentation incorporates wafers that are cut to match the size of microtiter plates, and then masked to provide wells corresponding to those making up a 96-well (12 x 8 in.) microtiter plate. A commercial liquid-handling system then transfers samples onto the plate. Bruker Optics (Billerica, MA), a manufacturer of IR spectrometers, and Pike Technologies (Madison, WI), an accessory supplier, provide hardware with which spectra can be acquired sequentially for samples at each of the 96 sample positions.
The integration of liquid handling and sample placement systems into one unit has cleared the path to high-throughput automated analysis. The innovations that opened the door to this advancement can be applied to various other embodiments. Spectroscopic instruments have been optimized for analytical work in various disciplines, and in preparation for the distinct possibility of new IR-based medical devices on the horizon, hardware providers are miniaturizing their instrumentation in anticipation of clinical diagnostic applications.
The precise form that these biomedical IR devices will take will depend very much on their intended applications. Clearly, high-throughput applications require either the integration of a liquid-handling device, or integration with handling components already in place on high-throughput analytical instrumentation. On the other hand, units intended for the doctor's office, point-of-care, ER, or operating theater would be compact stand-alone devices that use sampling and measurement procedures tailored for routine use by nonspecialists. The route to such instrumentation has been mapped out, and the development of end-user products is being pursued.
Outlook
IR spectroscopy has the potential to be used in a wide variety of biomedical analyses. A variety of established proof-of-concept applications have been introduced and summarized (see Table I). These applications include fetal lung maturity testing; whole blood glucose testing and point-of-care testing; urea analysis at the point of care; and cardiovascular risk panel in which total, HDL, and LDL cholesterol, triglycerides, and glucose are all recovered from the same IR spectrum. Possibilities for additional applications are likely to be conceptualized and developed in the near future.
Specialized quantitative assays are, in principle, straightforward to develop for many fluids that are sampled less frequently than serum and urine. For example, synovial fluid, cerebrospinal fluid, saliva, and crevicular fluid may all be sampled and their spectra measured in the procedure developed at the Institute for Biodiagnostics. To the extent that the major constituents of these samples are of diagnostic interest, they may be available from IR spectra. Further expansion of the range of clinical analyses available for serum and urine will hinge on the development of simple preconcentration methods. The development of such methods is being explored.
The diagnostic potential of IR spectroscopy is open to exploration, and the experience to date suggests that there is large untapped potential in this approach to diagnostics. One closely related diagnostic approach is the detection of ovarian cancer based upon mass spectrometry of donor serum.17,18 This test is founded on principles and methods analogous to those outlined here, namely the use of pattern recognition to identify diagnostic patterns revealed in spectra.
The diagnostic utility of testing via mass spectrography has been demonstrated, yet its biochemical foundation is not yet fully understood. While the gap between practical utility and understanding has ignited a spirited debate in the scientific community, the fact is that the data-mining approach apparently yields diagnostic tests that work. The only feature distinguishing these spectrometric tests from traditional diagnostic methods is the order of discovery; tests based upon pattern recognition will lead to new insights into the biochemical basis of disease, rather than vice versa.
While the mass spectrometry community seeks to replicate the ovarian cancer success story for other diseases, the IR spectroscopy community is building several tests simultaneously. Automated sampling and measurement technology opens the door to very rapid evaluation of prospective IR-based tests, and now is the best time to seek out and capitalize on the possibilities this new diagnostic approach has to offer.
References
1. RA Shaw and HH Mantsch, "Infrared Spectroscopy in Clinical and Diagnostic Analyses," Encyclopedia of Analytical Chemistry, ed. RA Meyers. (Chichester, UK: Wiley, 2000), 83–102.
2. RA Shaw and HH Mantsch, "Vibrational Spectroscopy Applications in Clinical Chemistry," Handbook of Vibrational Spectroscopy, ed. JM Chalmers and PR Griffiths (Sussex, UK: Wiley, 2002), 3295–3307.
3. RA Shaw et al., "Multianalyte Serum Analysis Using Mid-Infrared Spectroscopy," Annals of Clinical Biochemistry 35 (1998): 624–632.
4. KZ Liu et al., "Simultaneous Determination of Serum Cholesterol in High- and Low-Density Lipoproteins Using Infrared Spectroscopy," Clinical Chemistry 48 (2002): 499–506.
5. RA Shaw et al., "Toward Reagent-free Clinical Analysis: Quantitation of Urine Urea, Creatinine and Total Protein From the Mid-Infrared Spectra of Dried Urine Films," Clinical Chemistry 47 (2000): 1493–1495.
6. S Low Ying et al., "Quantitation of Glucose and Urea in Whole Blood by Mid-Infrared Spectroscopy of Dry Films," Vibrational Spectroscopy 28 (2002): 111–116.
7. KZ Liu et al., "Comparison of Infrared Spectroscopic and Fluorescence Depolarization Assays for Fetal Lung Maturity," American Journal of Obstetrics and Gynecology 183 (2000): 181–187.
8. W Petrich et al., "Disease Pattern Recognition in Infrared Spectra of Human Sera Using Diabetes Mellitus as an Example," Applied Optics 39 (2000): 3372–3379.
9. A Staib et al., "Disease Pattern Recognition Testing for Rheumatoid Arthritis Using Infrared Spectra of Human Sera," Clinica Chimica Acta 308 (2001): 79–89.
10. AE Nikulin et al., "Near-Optimal Region Selection for Feature Space Reduction: Novel Preprocessing Methods for Classifying MR Spectra," NMR in Biomedicine 11 (1998): 209–216.
11. HH Eysel, et al. "A Novel Diagnostic Test for Arthritis: Multivariate Analysis of Infrared Spectra of Synovial Fluid," Biospectroscopy 3 (1997): 161–167.
12. RL Somorjai, et al. "Distinguishing Normal from Rejecting Renal Allografts: Application of a Three-Stage Classification Strategy to MR and IR Spectra of Urine," Vibrational Spectroscopy 28 (2002): 97–102.
13. CP Schultz et al., "Study of Chronic Lymphocytic Leukemia Cells by FT-IR Spectroscopy and Cluster Analysis," Leukemia Research 20 (1996): 649–655.
14. LP Choo et al., "Infrared Spectroscopic Characterisation of Multiple Sclerosis Plaques in the Human Central Nervous System," Biochimica Biophysica Acta 1182 (1993): 333–337.
15. LP Choo et al., "In-situ Characterization of b-amyloid in Alzheimer's Diseased Tissue by Synchrotron Fourier Transform Infrared Microspectroscopy," Biophysical Journal 71 (1996): 1672–1679.
16. G Steiner et al., "Distinguishing and Grading Human Gliomas by Infrared Spectroscopy," Biospectroscopy. In press.
17. EF Petricoin et al., "Use of Proteomic Patterns in Serum to Identify Ovarian Cancer," Lancet 359 (2002): 572–577.
18. Correlogic Systems Inc., "Correlogic Systems Licenses Ovarian Cancer Diagnostic Test To Quest Diagnostics and LabCorp," (2002) [accessed 10 October 2003]; available from Internet: www.correlogic.com/ questlabcorp_final.htm.
PHOTO CREDIT NATIONAL RESEARCH COUNCIL OF CANADA
Anthony Shaw, PhD, is a research officer, Sarah Low Ying and Angela Man are technical officers, and Kan-Zhi Liu, PhD, is a research officer at the Institute for Biodiagnostics, National Research Council of Canada (Winnipeg, MB, Canada) They can be contacted via e-mail at anthony.shaw@nrc-cnrc.gc.ca, sarah.low-ying@nrc-cnrc.gc.ca, angela.man@nrc-cnrc.gc.ca, and kan-zhi.liu@nrc-cnrc.gc.ca, respectively.
Copyright ©2003 IVD Technology
