| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
Special Report |
1 Department of Clinical Biochemistry, Royal Infirmary of Edinburgh, Edinburgh, UK;2 Department of Pathology and Laboratory Medicine, Mount Sinai Hospital, and Department of Laboratory Medicine and Pathobiology, University of Toronto, Ontario, Canada;3 Department of Pathology, Center for Biomarker Discovery, Johns Hopkins Medical Institutions, Baltimore, MD, USA;4 Tanjung Bungah, Penang, Malaysia (previously Nuclear Medicine Section, Division of Human Health, International Atomic Energy Agency, Vienna, Austria);5 Department of Pathology, LDS Hospital, Salt Lake City, UT, USA;6 Breast Oncology Program, University of Michigan Comprehensive Cancer Center, Ann Arbor, MI, USA;7 Center for Applied Proteomics and Molecular Medicine, College of Sciences, George Mason University, Manassas, VA, USA;8 Clinical Research Unit, Department of Obstetrics and Gynecology, Technical University of Munich, Munich, Germany;9 Department of Microbiology and Molecular Cell Biology, Center for Biomedical Proteomics, Eastern Virginia Medical School, Norfolk, VA, USA;10 Department of Clinical Biochemistry, Hillerød Hospital, Hillerød, Denmark.
aAddress correspondence to this author at: Department of Clinical Biochemistry, Royal Infirmary of Edinburgh, 51 Little France Crescent, Edinburgh EH16 4SA, UK. Fax 44-131-242 6882; e-mail c.sturgeon{at}ed.ac.uk.
 |
Abstract |
|---|
Methods: One subcommittee developed guidelines for analytical quality relevant to serum and tissue-based tumor markers in current clinical practice. Two other subcommittees formulated recommendations particularly relevant to the developing technologies of microarrays and mass spectrometry.
Results: Prerequisites for optimal use of tumor markers in routine practice include formulation of the correct clinical questions to ensure selection of the appropriate test, adherence to good clinical and laboratory practices (e.g., minimization of the risk of incorrect patient and/or specimen identification, tube type, or timing), use of internationally standardized and well-characterized methods, careful adherence to manufacturer instructions, and proactive and timely reactions to information derived from both internal QC and proficiency-testing specimens. Highly desirable procedures include those designed to minimize the risk of the reporting of erroneous results attributable to interferences such as heterophilic antibodies or hook effects, to facilitate the provision of informative clinical reports (e.g., cumulative and/or graphical reports, appropriately derived reference intervals, and interpretative comments), and when possible to integrate these reports with other patient information through electronic health records. Also mandatory is extensive validation encompassing all stages of analysis before introduction of new technologies such as microarrays and mass spectrometry. Provision of high-quality tumor marker services is facilitated by dialogue involving researchers, diagnostic companies, clinical and laboratory users, and regulatory agencies.
Conclusions: Implementation of these recommendations, adapted to local practice, should encourage optimization of the clinical use of tumor markers.
|
Introduction |
|---|
An explanation of the methods used for developing these guidelines will be provided (unpublished data). The disciplines of all authors and statements of conflicts of interest, declared according to NACB requirements, are provided in the Supplemental Data Disclosures Table in the online Data Supplement. The latter are also listed at the end of this manuscript. All comments received about these guidelines are also recorded in the online Data Supplement (see the Supplemental Data Comments Received Table), together with responses to the comments.
The preparation of these guidelines included review of the literature relevant to the use of tumor markers. Particular attention was given to review articles, including the few relevant systematic reviews, and to guidelines issued by expert panels. When possible, the consensus recommendations of the NACB panels were based on available evidence, i.e., were evidence based.
|
nacb quality requirements for the use of tumor markers |
|---|
This report also briefly addresses some of the issues relevant to enhancing the clinical utility of tumor marker testing, both now and in the future.
|
preanalytical quality requirements |
|---|
With the advent of electronic health records, the process of tumor-marker ordering should be linked with preanalytical precautions available through databases designed to provide support for clinical decision-making (11), such as the database recently made available through the AACC (12).
|
analytical quality requirements |
|---|
NACB recommendations for both IQC and PT are presented in Supplemental Table 3 in the online Data Supplement. Although most of these recommendations are applicable to all analytes, several have particular relevance to tumor markers. Specimens for both IQC and PT should always resemble clinical sera as closely as is feasible. When clinical decision points are employed, stable and consistent performance is essential, including the use of IQC specimens at concentrations close to decision-point concentrations. Such protocols are critical for screening tests performed on asymptomatic individuals, such as screening for prostate cancer by measuring prostate-specific antigen (PSA) (13), or for cases in which chemotherapy may be instituted on the basis of an increasing tumor-marker concentration in the absence of other scan evidence, such as monitoring of testicular cancer patients (14). Functional sensitivity (i.e., the lowest result that can be reliably reported, best defined as the concentration at which the day-to-day CV is <20%) is also very important for certain tumor marker applications, e.g., monitoring PSA in prostate cancer patients after radical prostatectomy. By repeatedly issuing specimens of the same low concentration pool, PT schemes can provide valuable information about the stability of results over time (15). Because tumor markers are often used to monitor cancer patients for long time periods, assessment of long-term assay stability at other analyte concentrations is also advisable.
Long-term monitoring presents major challenges, because patients may change hospitals or laboratories may change methods of tumor-marker measurement. Ideally results obtained with different methods would be fully interchangeable, but unfortunately this is not the case, with between-method CVs in excess of 20% observed in PT schemes for some tumor markers (16). Major causes of between-method variation for these complex analytes include poor calibration, differences in antibody specificity, and differences in method design (17).
Reasonably standardized and accurate calibration should be achievable, but only for those analytes for which a recognized international standard (IS) or international reference reagent (IRR) is available (see Supplemental Table 4 in the online Data Supplement) and universally adopted by manufacturers for primary calibration of their methods. There are no IS for any of the important CA series of tumor markers, a major gap that should be addressed urgently. When relevant IS or IRR are available, recovery experiments undertaken through PT schemes, together with linearity and stability studies, provide the independent validation of consensus target values that is essential for a well-designed PT scheme. Conveniently, because PT providers should be working toward improving between-method agreement, these experiments also permit assessment of the accuracy of calibration of individual methods, helping to identify methods requiring improvement (e.g., methods over- or underrecovering the relevant IS by more than 10%). Long-term PT data can also confirm the effect of successful introduction of a new IS. Data from one scheme, for example, suggests that mean CVs decreased from 21.9% in 1995, before the first PSA IRR was introduced, to 9.5% in 2004 (18).
IRR for isoforms of human chorionic gonadotropin (hCG) (19) and PSA (20) (developed under the auspices of the IFCC) provide additional tools for elucidating method-related differences associated with the second major cause of method-related differences, antibody specificity. Carefully designed experiments with the IRR for PSA and free PSA enable assessment of the calibration and equimolarity of assays for PSA, which are particularly critical in the context of prostate cancer screening. Similarly, experiments with the 6 recently established IRR for hCG isoforms should elucidate what is actually measured by available methods for hCG analysis (19)(21), an issue of major importance for oncology applications in which recognition of a broad spectrum of hCG-related molecules is recommended (22). Epitope-mapping projects such as those carried out under the auspices of the International Society for Oncology and Biomarkers may enable the formulation of broad recommendations regarding the most clinically appropriate antibody specificities for some tumor markers, as has already been achieved for hCG (22).
Such studies may lead to better understanding of optimal method design for complex tumor markers, thereby addressing the third major cause of method-related differences. Differences in method design are likely to contribute to numerical differences in results observed and to differences in method robustness attributable to clinically relevant interferences (see Supplemental Table 5 in the online Data Supplement). Maintaining vigilant awareness of the potential for the latter is essential. Ultimately, the most effective way to minimize the risk of unrecognized interference leading to serious clinical error is to promote regular dialogue between laboratory and clinical staff, thereby encouraging early discussion and investigation of any results not in accord with the clinical picture (14).
|
postanalytical quality requirements |
|---|
|
clinical issues that enhance the reliability and utility of tumor markers |
|---|
Tumor markers are only useful if 3 circumstances pertain:
These issues are interrelated. For example, studies of the prognostic value of a marker that do not address the manner in which the study populations were treated are not helpful to the clinician trying to decide whether to apply treatment.
Furthermore, although statistical analysis is important for estimating the reliability of marker-identified differences between groups, the P-value alone does not indicate clinical utility. If a study is sufficiently powered, a small difference in outcomes of 2 groups separated by marker results (positive vs negative) might be statistically significant. Too often an investigator will conclude that a marker is clinically useful because a derived P-value is <0.05. More important for clinical utility, however, is that one population (marker positive or negative) does extremely well while the other does very poorly, so that one group might accept the therapy of interest while the other would elect not to. In such cases it is imperative that the P-value suggests statistical significance, but this criterion is not the determining factor for clinical utility. Finally, a single study does not establish a scientific fact. Rather, secondary validation of the results of an interesting study in a subsequent data set is imperative, and the validation study should use the same assay and the same cut point(s). In addition, the patient population must be very similar to that of the preceding study. These requirements are among those highlighted in the excellent reporting recommendations for tumor marker prognostic studies recently developed and published as the REMARK guidelines (28), complementing previous broader statements on the Standards for Reporting of Diagnostic Accuracy (29).
Studies leading to the acceptance of a tumor marker for clinical utility must be carefully and thoughtfully designed to ensure results that are meaningful in the clinical setting. Unfortunately, most tumor marker investigations have been studies of convenience, using archived samples that happened to be available (26). Such studies, considered to be at level of evidence (LOE) III, are useful for generating hypotheses, but without follow-up investigations involving careful investigational planning and design the results cannot be accepted as fact. Indeed, LOE II studies, in which the marker is considered prospectively as a secondary objective in a clinical trial, or better yet, LOE I studies, in which the marker in question is the primary objective, are much more likely to yield clinically useful results. In other words, it is better to ask the question and get an answer than to get an answer and then ponder the question. Such evidence-based considerations are particularly important when patient lives are at stake and should be remembered whenever a tumor marker test is requested.
|
microarrays in cancer diagnostics |
|---|
Although still generally restricted to research use, microarrays—biochips, DNA-chips, protein-chips, or cell-chips—have major potential clinical applications in oncology, enabling parallel and simultaneous analysis of complex systems and pathways and providing a complex snapshot of biological properties of the cell, tissue, organ, or fluid under consideration. Microarrays relevant to cancer diagnostics have been commercially introduced and/or are being developed (see Supplemental Table 7 in the online Data Supplement).
principles of microarrays
A microarray is a compact device containing a large number of well-defined immobilized capture molecules (e.g., synthetic oligos, PCR products, proteins, or antibodies). The best-known microarrays, DNA-biochips, are miniature arrays of oligonucleotides attached to glass or plastic surfaces. These microarrays are used to examine gene activity (expression profiling) and identify gene mutations or single-nucleotide polymorphisms, by hybridization between the microarray sequences and a labeled probe (the sample of interest). There are 2 major methods for microarray fabrication: photolithography and mechanical deposition on materials such as glass slides or membranes (35).
Major potential advantages of microarray-based assays include high sensitivity, requirement for only small amounts of binding reagents, independence of sample volume, decreased incubation times, minimal reagent wastage, simultaneous access to many genes or proteins, and potentially, quantification. More detailed information on the subject is readily available in many specialized books and reviews.
tissue microarrays
High-throughput analysis of tissues is facilitated by new technologies such as multitissue northern blots, protein arrays, and real-time PCR (36). However, tissues are disintegrated before analysis, preventing identification of the cell types expressing the gene of interest. This limitation can be overcome by tissue microarrays, which consist of up to 1000 tiny cylindrical tissue samples assembled on a routine histology paraffin block, enabling simultaneous cost-effective analysis of up to 1000 tissue samples in a single experiment.
applications of microarrays
Microarrays have been successfully applied in a variety of settings, including gene expression profiling, detection of single-nucleotide polymorphisms, sequencing by hybridization, protein expression profiling, protein-protein interaction studies, and whole-genome biology experiments.
Although the potential of microarrays has yet to be fully realized, they have shown great promise as tools for deciphering complex diseases, including cancer (37) (see Supplemental Table 7 in the online Data Supplement).
limitations of microarrays
Microarray technologies are still evolving, presenting difficulties for standardization because no gold standards are yet available to facilitate comparison of data between laboratories and platforms. Recent reports suggest that microarray data are not reproducible (38) and may be biased (39).
key points: microarrays in cancer diagnostics
Microarrays will undoubtedly become routine diagnostic tools (see Supplemental Table 8 in the online Data Supplement), but first several variables must be optimized (e.g., capture molecules, hybridization protocols, data collection), QC systems must be established, and the appropriate level of analytical and clinical validation must be determined, as has been proposed (40)(41).
Evaluation of microarray expression data requires careful consideration of the validity and accuracy of results for the biological system and whether the data fundamentally describe the phenomenon being investigated, together with the possibility of the presence of artifacts, which can occur at any time. Validation requires experimental QC, independent confirmation of data, and universality of results. Automation is also essential to minimize variability and increase robustness. Currently, for most studies on the clinical application of microarrays (26), the LOE is only V. An exception to this is the 70-gene signature for predicting outcome in breast cancer (42), a microarray application that has undergone both internal (43) and external validation (44) and was cleared by the FDA in 2007.
Based on the information above, the NACB Panel has formulated the recommendations outlined in Table 1
.
|
ms in cancer diagnostics
The principles of MS as applied to cancer diagnostics are summarized here, together with recommendations for the use of MS in clinical practice, focusing on MALDI and related MS techniques such as SELDI for proteomic analysis.
principles of diagnostic ms
The typical mass spectrometer consists of an ion source, a mass analyzer that measures the mass-to-charge ratio (m/z) of the ionized analytes, and a detector that registers the number of ions at each m/z value. MS measurements are carried out in the gas phase of ionized species, often using electrospray ionization and MALDI to volatize and ionize the proteins or peptides. A variant of the latter is SELDI. The mass analyzer separates ionic species according to their m/z ratios. Four basic types of mass analyzers are commonly used in proteomic research: the ion trap, TOF, quadrupole, and Fourier transform ion cyclotron resonance; a potential fifth variant is the new OrbitrapTM mass spectrometer. These analyzers may be variously combined.
There are 2 approaches for biomarker discovery using MALDI/SELDI-TOF MS. One approach uses the differences between MS profiles of the disease and control specimens to generate a diagnostic model. A variation of this approach is to select several discriminate protein/peptide peaks and identify their nature. Diagnostics are based on multiplex immuno-MS or ELISA. The other approach is to degrade enzymatically the proteins to peptides, separate the peptides by techniques such as HPLC, and direct the eluted fractions into an ion source (electrospray ionization or MALDI), where they are converted into ionized species that enter the mass spectrometer, followed by algorithmic identification of the protein fragments comprising the mass spectra.
Protein identification is achieved through either peptide mass fingerprinting or peptide sequencing. In the former, peptide masses are compared with mass spectra of proteins listed in databases using appropriate software. Peptide sequencing is based on inducing random cleavage of peptide bonds between adjacent amino acid residues, e.g., by collision-induced dissociation, and determining the amino acid sequence of the resulting ion series.
application of ms in cancer diagnostics
MS is particularly well suited to serve as a diagnostic or biomarker discovery tool in cancer, because during cancer development, cancer cells and/or the surrounding microenvironment generate proteins and peptides of different types and in different concentrations than normal cells. These abnormal tissue distributions can be analyzed by imaging-based MS and patterns compared with controls to identify cancer-specific changes that may be clinically useful. Should leakage into the circulation occur from the tumor-host microenvironment, cancer-specific analytes may also be detectable in the blood (Fig. 1
). MS has been used to demonstrate many cancer-specific protein patterns, most often in blood and urine (see Supplemental Table 9 in the online Data Supplement).
|
Almost every published report of the profiles generated by MALDI-TOF MS has suggested that this method yields better diagnostic sensitivities and specificities than cancer biomarkers in current use, resulting in extensive publicity. However, initial enthusiasm has been tempered by other reports identifying potential problems with this approach and its clinical reliability (45), highlighting the importance of rigorous validation.
If MALDI-TOF profiling is to transition successfully from research technique to clinical laboratory, all sources of variation—preanalytical, analytical, and postanalytical—must be understood and controlled (45). The effects of sample storage and processing, sample type, patient selection, and demographic variables on test outcome must be clearly established, and analytical performance must be improved such that sensitivity, specificity, and dynamic range are comparable to those of established immunoassay techniques. Further investigations must address reproducibility of protein patterns across different batches of chips, different analysts, different sites, and different instrumentation. Robustness of the methodology is of concern, as are issues related to bioinformatic artifacts, data overfitting, and bias arising from experimental design. Some of these issues may relate to inappropriate use of publicly available mass spectral data sets, a consortium of investigators having recently succeeded in obtaining reproducible mass spectral signatures at multiple sites across time and instruments. This development is encouraging for those attempting to employ MALDI-TOF–type approaches for protein fingerprinting based diagnostics (46).
Current limitations and promises of MALDI-TOF, particularly as applied to clinical practice and cancer diagnostics, are addressed more fully in several recently published reviews (for references see the original unabridged manuscript in the online Data Supplement).
key points: mass spectrometry profiling in cancer diagnostics
Despite numerous publications describing diagnostic use of MALDI-TOF MS (see Supplemental Table 9 in the online Data Supplement), most are only LOE IV-V studies (i.e., either retrospective or small pilot studies) (26). According to published criteria (47), the stage of development of this technology is phase 1 (preclinical exploratory studies). NACB recommendations have been formulated on the basis of this information (Table 2
). MALDI-TOF MS approaches are promising for biomarker discovery and validation, but before clinical application can be instituted the issues discussed here must be addressed. Advantages of proteomic profiling include analysis without the need for a labeling molecule, potentially high specificity, multiparametric analysis, high throughput, very low sample volume requirements, and direct interface with computer algorithms. Important limitations include variable cross-platform reliability of signatures generated, major vulnerability to minor changes in sample handling and processing, and poor analytical sensitivity, particularly when the analyte is present in minute amounts in a highly complex mixture containing high-abundance molecules. However, there are inherent advantages to certain MALDI-TOF approaches. For instance, combining immune isolation with MALDI-TOF analysis allows for elimination of secondary antibodies and detection of multiple derivative analytes such as protein isoforms. In addition, exciting new research suggests that many low-abundance proteins and low molecular weight analytes exist in bound states in serum and are effectively amplified by carrier-protein–based sequestration. These low molecular weight analytes appear to have underpinned many past spectral fingerprints, indicating that many of these ions may be generated from low-abundance analytes. A list of these low molecular weight carrier protein–bound analytes has recently been provided for early stage ovarian cancer patients (48).
|
Until extensive validation studies are performed, MALDI-TOF MS fingerprinting approaches should not be used for cancer diagnosis. When MS fingerprinting is employed, independent validation sets including appropriate numbers of inflammatory, benign, and unaffected controls are essential, because specificity will determine success in the clinic. Encouraging developments include better appreciation of the influence of bias and variance and of the need for carefully defined operating procedures essential for validation of this technology and its introduction into clinical practice.
NACB Subcommittee members: Quality Requirements: Catharine M. Sturgeon, Chair; Soo-Ling Ch’ng, Elizabeth Hammond, Daniel F. Hayes, and Györg Sölétormos; Microarrays: Eleftherios P. Diamandis, Chair; Manfred Schmitt and Elena van der Merwe; Mass spectrometry: Daniel W. Chan, Chair; Eleftherios P. Diamandis, Lance A. Liotta, Emmanuel F. Petricoin; O. John Semmes, and Elena van der Merwe.
NACB Quality Requirements Subcommittee Members: Catharine M Sturgeon, Chair; Soo-Ling Ch’ng, Elizabeth Hammond, Daniel F. Hayes, and Györg Sölétormos. All comments received about the NACB Recommendations for Quality Requirements are included in the online Data Supplement. Dr Jean-Pierre Basayau, Professor Per Hyltoft Petersen, Professor Mathias Müller, and Professor Hans Schneider were invited expert reviewers.
NACB Microarray Subcommittee members: Eleftherios P. Diamandis, Chair; Manfred Schmitt and Da-elene van der Merwe. All comments received about the NACB Recommendations for Microarrays in Cancer Diagnostics are included in the online Data Supplement. Professor Roger Ekins was an invited expert reviewer.
NACB Mass Spectrometry Subcommittee members: Daniel W. Chan, Chair; O. John Semmes, Emmanuel F. Petricoin, Lance A. Liotta, Da-elene van der Merwe, and Eleftherios P. Diamandis. All comments received about the NACB Recommendations for Mass Spectrometry in Cancer Diagnostics are included in the on-line supplement. Professor William T. Morgan and Professor Roz Banks were invited expert reviewers.
|
Acknowledgments |
|---|
Financial Disclosures: C. Sturgeon has over the past ten years received speaker expenses and/or honoraria from Abbott Diagnostics, Bayer Diagnostics, Beckman Coulter, Becton Dickinson, Dako, DPC, Roche, Tosoh Biosciences, and Wallac Oy for lectures on quality aspects of tumor markers (with no product endorsements). D. Hayes has received expenses and/or honoraria as a consultant, advisory panel participant, or lecturer/speaker during the last 12 months from Abraxis, American Biosciences, AviaraDx, Cytogen, Monogram Bioscience, Pfizer, Precision Therapeutics, Predictive Biosciences, Siemens Medical Solutions Diagnostics, and StemCapter and during the last twelve months has been principal or coinvestigator of clinical research sponsored by GlaxoSmithKline, Pfizer, Novartis, and Wyeth Ayerst-Genetics Institute. D. Chan has received research funding as an investigator for Abbott Labs and Beckman Coulter, fees as a consultant for Tosoh Biosciences, and fees as a scientific advisory board member for Ciphergen and Roche Diagnostics. L. Liotta is a board member for and holds stocks in Theranostics but receives no honoraria or salary. E. Petricoin is an inventor and owner of US Government and university-assigned patent applications that cover aspects of the technologies discussed in this manuscript. M. Schmitt is a research collaborator with American Diagnostica and Epigenomics AG.
Acknowledgments: We thank the numerous scientists and clinicians who have contributed to this undertaking, especially the invited reviewers listed in the Appendix, Mrs. Hassima Omar Ali for her excellent assistance, David Bruns for agreeing to consider publishing these guidelines in Clinical Chemistry and for his highly valued advice, and of course the NACB and the AACC for their much appreciated support and encouragement.
|
Footnotes |
|---|
2 Nonstandard abbreviations: NACB, National Academy of Clinical Biochemistry; MS, mass spectrometry; IQC, internal QC; PT, proficiency testing; PSA, prostate-specific antigen; IS, international standard; IRR, international reference reagent; hCG, human chorionic gonadotropin; LOE, level of evidence.
|
References |
|---|
The following articles in journals at HighWire Press have cited this article:
|
|
 |
|
  D. F. Ransohoff and M. L. Gourlay Sources of Bias in Specimens for Research About Molecular Markers for Cancer J. Clin. Oncol., February 1, 2010; 28(4): 698 - 704. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 C M Sturgeon, L C Lai, and M J Duffy Serum tumour markers: how to order and interpret them BMJ, September 22, 2009; 339(sep22_1): b3527 - b3527. [Full Text]
|
|
|
|
 |
|
 M. J. Duffy, C. Sturgeon, R. Lamerz, C. Haglund, V. L. Holubec, R. Klapdor, A. Nicolini, O. Topolcan, and V. Heinemann Tumor markers in pancreatic cancer: a European Group on Tumor Markers (EGTM) status report Ann. Onc., September 8, 2009; (2009) mdp332v2. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 C. M. Sturgeon, M. J. Duffy, U.-H. Stenman, H. Lilja, N. Brunner, D. W. Chan, R. Babaian, R. C. Bast Jr., B. Dowell, F. J. Esteva, et al. National Academy of Clinical Biochemistry Laboratory Medicine Practice Guidelines for Use of Tumor Markers in Testicular, Prostate, Colorectal, Breast, and Ovarian Cancers Clin. Chem., December 1, 2008; 54(12): e11 - e79. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
E. P. Diamandis, B. R. Hoffman, and C. M. Sturgeon National Academy of Clinical Biochemistry Laboratory Medicine Practice Guidelines for the Use of Tumor Markers Clin. Chem., November 1, 2008; 54(11): 1935 - 1939. [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
Top
Abstract
Introduction