Volatile Biomarkers: Non-Invasive Diagnosis in Physiology and Medicine
()
About this ebook
Volatile organic compounds (VOCs) in exhaled breath, sweat or urine carry much information on the state of human health. The role of VOCs in clinical diagnosis and therapeutic monitoring is expected to become increasingly significant due to recent advances in the field. Volatile Biomarkers: Non-Invasive Diagnosis in Physiology and Medicine includes the latest discoveries and applications for VOCs from the world's foremost scientists and clinicians working in this emerging analytic area.
- Appeals to a multidisciplinary audience, including scientists, researchers, and clinicians with an interest in breath analysis
- Features the latest scientific research and technical breakthroughs in the diagnostic and therapeutic aspects of volatile organic compounds
- Includes case presentations documenting applications in multiple areas of human health and safety
Related to Volatile Biomarkers
Related ebooks
Nanobiomaterials in Medical Imaging: Applications of Nanobiomaterials Rating: 0 out of 5 stars0 ratingsDrug Delivery Nanosystems for Biomedical Applications Rating: 0 out of 5 stars0 ratingsHandbook of Green Analytical Chemistry Rating: 5 out of 5 stars5/5Adverse Effects of Engineered Nanomaterials: Exposure, Toxicology, and Impact on Human Health Rating: 4 out of 5 stars4/5Toxicological Risk Assessment and Multi-System Health Impacts from Exposure Rating: 0 out of 5 stars0 ratingsThe Plasma Proteins: Isolation, Characterization, and Function Rating: 0 out of 5 stars0 ratingsMetabolic Inhibitors V3: A Comprehensive Treatise Rating: 0 out of 5 stars0 ratingsAdvances in Lipid Methodology Rating: 0 out of 5 stars0 ratingsNovel Developments in Pharmaceutical and Biomedical Analysis Rating: 0 out of 5 stars0 ratingsInformation Resources in Toxicology, Volume 2: The Global Arena Rating: 0 out of 5 stars0 ratingsEnvironmental Factors in Respiratory Disease Rating: 0 out of 5 stars0 ratingsComprehensive Quality by Design for Pharmaceutical Product Development and Manufacture Rating: 0 out of 5 stars0 ratingsEvaluating Water Quality to Prevent Future Disasters Rating: 0 out of 5 stars0 ratingsDialysis: a Memoir Rating: 0 out of 5 stars0 ratingsA Pharmacology Primer: Techniques for More Effective and Strategic Drug Discovery Rating: 1 out of 5 stars1/5Gregory's Pediatric Anesthesia Rating: 4 out of 5 stars4/5Stationary Phases in Gas Chromatography Rating: 0 out of 5 stars0 ratingsMonoclonal Antibody and Immunosensor Technology: <b>The production and application of rodent and human monoclonal antibodies</b> Rating: 0 out of 5 stars0 ratingsClinical Quality Management System A Complete Guide - 2020 Edition Rating: 0 out of 5 stars0 ratingsHaematology in Critical Care: A Practical Handbook Rating: 0 out of 5 stars0 ratingsPolyunsaturated Fatty Acid Metabolism Rating: 0 out of 5 stars0 ratingsStudies in Natural Products Chemistry: Stereoselective Synthesis (Part F) Rating: 0 out of 5 stars0 ratingsInfectious Diseases: Smart Study Guide for Medical Students, Residents, and Clinical Providers Rating: 0 out of 5 stars0 ratingsFormulation and Analytical Development for Low-Dose Oral Drug Products Rating: 0 out of 5 stars0 ratingsFrontiers in Medicinal Chemistry: Volume 9 Rating: 0 out of 5 stars0 ratingsHazardous Substances and Human Health: Exposure, Impact and External Cost Assessment at the European Scale Rating: 0 out of 5 stars0 ratingsYear One of Practice Transition: The Necessary Guide to Success Rating: 0 out of 5 stars0 ratingsLung Development Biological and Clinical Perspectives: Biochemistry and Physiology Rating: 0 out of 5 stars0 ratingsApplications of HPLC in Biochemistry Rating: 0 out of 5 stars0 ratingsPhysical Vascular Therapy - The Next Generation Of Medicine?: Microcirculation Of Blood - What Everyone Should Know About Rating: 0 out of 5 stars0 ratings
Chemistry For You
A to Z Magic Mushrooms Making Your Own for Total Beginners Rating: 0 out of 5 stars0 ratingsOrganic Chemistry I For Dummies Rating: 5 out of 5 stars5/5Chemistry: Concepts and Problems, A Self-Teaching Guide Rating: 5 out of 5 stars5/5Chemistry For Dummies Rating: 4 out of 5 stars4/5Biochemistry For Dummies Rating: 5 out of 5 stars5/5MCAT General Chemistry Review 2024-2025: Online + Book Rating: 0 out of 5 stars0 ratingsGeneral Chemistry Rating: 4 out of 5 stars4/5Chemistry: a QuickStudy Laminated Reference Guide Rating: 5 out of 5 stars5/5Fundamentals of Chemistry: A Modern Introduction Rating: 5 out of 5 stars5/5Organic Chemistry for Schools: Advanced Level and Senior High School Rating: 0 out of 5 stars0 ratingsThe Secrets of Alchemy Rating: 4 out of 5 stars4/5College Chemistry Rating: 4 out of 5 stars4/5An Introduction to the Periodic Table of Elements : Chemistry Textbook Grade 8 | Children's Chemistry Books Rating: 5 out of 5 stars5/5Handbook of Histopathological and Histochemical Techniques: Including Museum Techniques Rating: 4 out of 5 stars4/5Painless Chemistry Rating: 0 out of 5 stars0 ratingsThe Chemistry Book: From Gunpowder to Graphene, 250 Milestones in the History of Chemistry Rating: 5 out of 5 stars5/5Chemistry Workbook For Dummies with Online Practice Rating: 0 out of 5 stars0 ratingsMCAT Organic Chemistry Review 2024-2025: Online + Book Rating: 0 out of 5 stars0 ratingsStuff Matters: Exploring the Marvelous Materials That Shape Our Man-Made World Rating: 4 out of 5 stars4/5TIHKAL: The Continuation Rating: 4 out of 5 stars4/5Elementary: The Periodic Table Explained Rating: 0 out of 5 stars0 ratingsOrganic Chemistry II For Dummies Rating: 4 out of 5 stars4/5Chemistry All-in-One For Dummies (+ Chapter Quizzes Online) Rating: 0 out of 5 stars0 ratingsOrganic Chemistry I Essentials Rating: 4 out of 5 stars4/5Cannabis Alchemy: Art of Modern Hashmaking Rating: 0 out of 5 stars0 ratingsThe Nature of Drugs Vol. 1: History, Pharmacology, and Social Impact Rating: 5 out of 5 stars5/5PIHKAL: A Chemical Love Story Rating: 4 out of 5 stars4/5Chemistry for Breakfast: The Amazing Science of Everyday Life Rating: 4 out of 5 stars4/5AP Chemistry Flashcards, Fourth Edition: Up-to-Date Review and Practice Rating: 0 out of 5 stars0 ratingsMendeleyev's Dream Rating: 4 out of 5 stars4/5
Reviews for Volatile Biomarkers
0 ratings0 reviews
Book preview
Volatile Biomarkers - Cristina Davis
Volatile Biomarkers
Non-Invasive Diagnosis in Physiology and Medicine
Edited by
Anton Amann
Breath Research Institute Austrian Academy of Sciences Rathausplatz 4, A-6850 Dornbirn Austria
Univ.-Clinic for Anesthesia Innsbruck Medical University Anichstr 35, A-6020 Innsbruck Austria
ETH-Zentrum Laboratory of Physical Chemistry CHN E24, CH-8092 Zürich Switzerland
David Smith
Institute for Science and Technology in Medicine Keele University, Stoke-on-Trent, ST4 7QB UK
Table of Contents
Cover image
Title page
Copyright
List of Contributors
Foreword
PART A Interpretation of Breath Analysis Data
Chapter 1. Mathematical and Statistical Approaches for Interpreting Biomarker Compounds in Exhaled Human Breath
1.1 Introduction
1.2 Data interpretation
1.3 Conclusions and recommendations
Disclaimer
References
Chapter 2. Issues and Challenges in Human Breath Research: Perspectives from Our Experience
2.1 Introduction
2.2 Defining normal in clinical practice: the case of a routine liver blood test
2.3 Developing a breath test: can the blood assay be unseated?
2.4 Convincing clinicians
2.5 Breath markers
2.6 Conclusion
References
PART B Real-Time Analysis of Exhaled Breath
Chapter 3. Physiological Modeling for Analysis of Exhaled Breath
3.1 Introduction
3.2 Real-time measurements: experimental basics
3.3 Modeling
3.4 Concluding remarks
References
PART C Physiological and Clinical Studies
Chapter 4. Recent SIFT-MS Studies of Volatile Compounds in Physiology, Medicine and Cell Biology
4.1 Introduction
4.2 Direct breath analysis
4.3 VOC emission from skin; comparison with VOCs in breath
4.4 Exhaled breath condensate and broncoalveolar lavage
4.5 VOCs in urine headspace; ketones and ovulation; 3-HBA
4.6 Volatile biomarkers of cancer cells, in vitro and in vivo
4.7 Alcohol ingestion and detection and cannabis
4.8 Flowing afterglow mass spectrometry, FA-MS, and total body water
4.9 Summary remarks; future prospects for SIFT-MS and FA-MS in medicine
References
Chapter 5. The Analysis of Oral Air by Selected Ion Flow Tube Mass Spectrometry Using Indole and Methylindole as Examples
5.1 Introduction
5.2 Oral malodor
5.3 Oral air sampling considerations
5.4 Indoles
5.5 Summary and conclusion
References
Chapter 6. Smokers Breath as Seen by Proton-Transfer-Reaction Time-of-Flight Mass Spectrometry (PTR-TOF-MS)
6.1 Introduction
6.2 Materials and methods
6.3 Results and discussion
6.4 Conclusion
References
Chapter 7. Exhaled Breath Analysis in Occupational Medicine
7.1 Rationale for the use of exhaled breath analysis in occupational medicine
7.2 Exhaled nitric oxide
7.3 Exhaled breath condensate (EBC)
7.4 Exhaled volatile organic compounds
References
Chapter 8. Volatile Organic Compounds in Human Breath: Biogenic Origin and Point-of-Care Analysis Approaches
8.1 Biogenic origin of volatile compounds in human exhaled breath
8.2 Miniature mobile point-of-care diagnostic systems for VOC breath biomarkers: trends and future requirements
8.3 Advances in data analysis required for breath sensor technologies
References
Chapter 9. Breath Analysis in Critically Ill Patients—Potential and Limitations
9.1 Introduction
9.2 Technical aspects of breath analysis in critically ill patients
9.3 Methodological aspects
9.4 Clinical aspects of breath analysis in critically ill patients
9.5 Conclusions
References
Chapter 10. Analysis of Cancer Biomarkers in Exhaled Breath and Comparison with Sensory Indications by Dogs
10.1 Introduction
10.2 Experimental
10.3 Results and Discussion
10.4 Conclusions
References
PART D Nitric Oxide, NO, and Carbon Monoxide, CO
Chapter 11. Added Value with Extended NO Analysis
11.1 Background
11.2 A two compartment model
11.3 Different no models
11.4 Corrections for axial back diffusion
11.5 Limitations
11.6 Values from non-smoking healthy subjects
11.7 The usefulness of extended no analysis
11.8 Conclusions
References
Chapter 12. Carbon Monoxide as an Exhaled Biomarker of Pulmonary Diseases
12.1 Introduction
12.2 Chemical and biochemical properties of CO
12.3 Environmental sources of CO
12.4 Endogenous sources of CO: the heme oxygenase enzyme system
12.5 Signaling properties of CO
12.6 CYTO- and tissue-protective effects of CO
12.7 Methods for breath CO detection
12.8 Exhaled CO in human diseases
12.9 Conclusions
References
Chapter 13. Exhaled Nitric Oxide in Clinical Practice: Recent Advances and New Challenges
13.1 Introduction
13.2 Exhaled nitric oxide
13.3 Technical aspects of measurement
13.4 NO in clinical decision making
13.5 Conclusions and directions for future areas of research
References
PART E Clinical Breath Tests
Chapter 14. An Update on 13C-Breath Tests: The Transition to Acceptability into Clinical Practice
14.1 Introduction
14.2 History of 13C breath tests
14.3 Standardization of instrumentation and breath collection bags
14.4 Breath tests during the period 2005–2011
14.5 The future of 13C-breath tests
References
PART F Development and Use of Sensors
Chapter 15. Sensors for Exhaled Gas Analysis: An Analytical Review
15.1 Introduction. sensors as a prospective tool for implementation of fundamental findings in the area of exhaled gas analysis in the clinical setting
15.2 Sensory metrological performance essential for exhaled breath measurements
15.3 Types of sensors used for exhaled gas analysis
15.4 Detection principles and concepts involved in breath analysis using sensors
15.5 Medical applications of sensory breath analysis
15.6 Concluding remarks
References
Chapter 16. Arrays of Nanomaterial-Based Sensors for Breath Testing
16.1 Introduction
16.2 The design of the sensor array
16.3 Nanomaterials for sensor arrays
16.4 Chemiresistive MCNP films for sensor arrays
16.5 Single-walled carbon nanotubes (SWCNTS) for sensors arrays
16.6 Semiconducting nanowires for cross-reactive sensors
16.7 Summary and conclusions
References
Chapter 17. Smart Sensor Systems for Human Health Breath Monitoring Applications
17.1 Introduction
17.2 Smart sensor systems
17.3 Breath monitoring: smart sensor system development
17.4 Home asthma breath monitoring technology
17.5 Asthma monitoring system miniaturization
17.6 Summary and conclusion
References
Chapter 18. VOC Analysis by SIFT-MS, GC-MS, and Electronic Nose for Diagnosing and Monitoring Disease
18.1 Methodology and VOC analysis
18.2 Healthy volunteers
18.3 Acetone and diabetes
18.4 Acetone from skin
18.5 Tuberculosis and other infectious diseases
18.6 Gastro-intestinal illness
18.7 Bladder cancer
18.8 Concluding remarks
References
PART G Exhaled Breath Condensate (EBC) and Particulates
Chapter 19. Measurement of Biomarkers of Oxidative Stress and Airway Inflammation in Exhaled Breath Condensate: Methodology and Potential Applications in Patients with COPD and Healthy Smokers
19.1 Introduction
19.2 EBC analysis: methodology
19.3 Analysis of EBC from patients with COPD and healthy smokers
19.4 Advantages and limitations of the EBC technique
19.5 Future research
References
Chapter 20. Particles in Exhaled Air—A Novel Method of Sampling Non-Volatiles in Exhaled Air
20.1 Introduction
20.2 Number of particles in exhaled breath
20.3 Formation and origin of exhaled particles
20.4 Composition of exhaled particles
20.5 Sampling of exhaled particles in occupational medicine
20.6 Conclusions
References
PART H Volatiles of Microbial Origin: Urine, Stool and in vitro Cultures
Chapter 21. Challenges in the Investigation of Volatile Disease Biomarkers in Urine
21.1 Introduction
21.2 Distinctive odors and/or volatile profiles associated with diseases are present in urine and detected by animals and analytical instruments
21.3 Challenges in monitoring volatile disease biomarkers in urine
21.4 Discussion
References
Chapter 22. Volatile Organic Compounds (VOCs) Found in Urine and Stool
22.1 Introduction
22.2 Urine VOCs and disease
22.3 Bacteria present in human urine, current test methods, and VOC analyses
22.4 Stool VOCs and disease
22.5 A comparison of VOCs found in urine and stool
22.6 Summary
References
Chapter 23. Volatile Organic Compounds (VOCs) Released by Pathogenic Microorganisms in vitro: Potential Breath Biomarkers for Early-Stage Diagnosis of Disease
23.1 Introduction
23.2 Methodology
23.3 Results
23.4 Discussion
23.5 Summary
References
PART I Urban Search and Rescue Operations
Chapter 24. Potential Applications of Volatile Organic Compounds in Safety and Security
24.1 Introduction
24.2 Chemical analysis of breath
24.3 Analytical instrumentation; field technology
24.4 Factors affecting VOCs
24.5 Volatiles in safety and security applications
24.6 Urine as a potential source of markers of human presence
24.7 Evaluation of IMS-based portable technologies for the detection of urine-borne human scent constituents
24.8 SPATIO-temporal measurements of VOCs
24.9 Real-time measurement of exhaled breath and skin emanations
24.10 A hit list of compounds for urban search and rescue operations
24.11 Summary
References
Index
Copyright
Elsevier
The Boulevard, Langford Lane, Kidlington, Oxford OX5 1GB, UK
Radarweg 29, PO Box 211, 1000 AE Amsterdam, The Netherlands
225 Wyman Street, Waltham, MA 02451, USA
525 B Street, Suite 1900, San Diego, CA 92101-4495, USA
First edition 2013
Copyright © 2013 Elsevier B.V. All rights reserved.
No part of this publication may be reproduced, stored in a retrieval system or transmitted in any form or by any means electronic, mechanical, photocopying, recording or otherwise without the prior written permission of the publisher Permissions may be sought directly from Elsevier’s Science & Technology Rights Department in Oxford, UK: phone (+44) (0) 1865 843830; fax (+44) (0) 1865 853333; email: permissions@elsevier.com. Alternatively you can submit your request online by visiting the Elsevier web site at http://elsevier.com/locate/permissions, and selecting Obtaining permission to use Elsevier material
Notice
No responsibility is assumed by the publisher for any injury and/or damage to persons or property as a matter of products liability, negligence or otherwise, or from any use or operation of any methods, products, instructions or ideas contained in the material herein. Because of rapid advances in the medical sciences, in particular, independent verification of diagnoses and drug dosages should be made.
British Library Cataloguing in Publication Data
A catalogue record for this book is available from the British Library
Library of Congress Cataloging-in-Publication Data
A catalog record for this book is available from the Library of Congress
For information on all Elsevier publications visit our web site at store.elsevier.com
Printed and bound in China
13 14 15 16 17 10 9 8 7 6 5 4 3 2 1
ISBN: 978-0-44-462613-4
List of Contributors
Agapios Agapiou, National Technical University of Athens (NTUA), School of Chemical Engineering, Sector I, 9 Iroon Polytechniou Street, 15773 Athens, Greece
Clemens Ager, Breath Research Institute, Austrian Academy of Sciences, Rathausplatz 4, 6850 Dornbirn, Austria; University Clinic for Anesthesia, Innsbruck Medical University, Anichstraße 35, 6020 Innsbruck, Austria
Alexander A. Aksenov, Department of Mechanical and Aerospace Engineering, One Shields Avenue, University of California, Davis, CA 95616, USA
Anton Amann, Breath Research Institute, Austrian Academy of Sciences, Rathausplatz 4, 6850 Dornbirn, Austria; University Clinic for Anesthesia, Innsbruck Medical University, Anichstraße 35, 6020 Innsbruck, Austria
Hamzeh Bardaweel, Department of Mechanical and Aerospace Engineering, One Shields Avenue, University of California, Davis, CA 95616, USA
Peter J. Barnes, Section of Airway Disease, National Heart and Lung Institute, Imperial College London, London, UK
Maria Baur, Breath Research Institute, Austrian Academy of Sciences, Rathausplatz 4, 6850 Dornbirn, Austria; University Clinic for Anesthesia, Innsbruck Medical University, Anichstraße 35, 6020 Innsbruck, Austria
A.M. Biaggi-Labiosa, NASA Glenn Research Center, 21000 Brookpark Road, Cleveland, OH 44135, USA
Bogusław Buszewski, Department of Environmental Chemistry and Bioanalytics, Faculty of Chemistry, Nicolaus Copernicus University,7 Gagarin St., 87-100 Toruń, Poland
C.W. Chang, ASRC /NASA Glenn Research Center, 21000 Brookpark Road, Cleveland, OH 44135, USA
William H.K. Cheung, Department of Mechanical and Aerospace Engineering, One Shields Avenue, University of California, Davis, CA 95616, USA
Massimo Corradi, Department of Clinical and Experimental Medicine, Nephrology and Health Sciences, University of Parma, Via Gramsci 14, I-43100 Parma, Italy
Ben de Lacy Costello, Centre for Research in Biosciences, University of the West of England, Coldharbour Lane, Frenchay, Bristol BS16 1QY, UK
Martin Daniaux, Department of Radiology, Medical University Innsbruck, Anichstrasse 35, 6020 Innsbruck, Austria
Cristina E. Davis, Department of Mechanical and Aerospace Engineering, One Shields Avenue, University of California, Davis, CA 95616, USA; Clinical and Translational Sciences Center (CTSC), UC Davis Medical Center, Sacramento, CA 95817, USA
Jürgen Dunkl, Oncoytyrol-Center for Personalized Cancer Medicine GmbH, Eduard Bodem Gasse 3, 6020 Innsbruck, Austria; Institute of Ion Physics and Applied Physics, Leopold-Franzens University Innsbruck, Technikerstrasse 25, 6020 Innsbruck, Austria
Prabir K. Dutta, Ohio State University, Department of Chemistry, 100 West 18th Avenue, Columbus, OH 43210, USA
Raed A. Dweik, Cleveland Clinic, Respiratory Institute & Pathobiology/Lerner Research Institute, Desk A90, 9500 Euclid Ave., Cleveland, OH 44195, USA
Ashley Esarik, Northern Ontario School of Medicine and Department of Biology, Lakehead University, Thunder Bay, Ontario, Canada P7B5E1
Anna Filipiak, Breath Research Institute, Austrian Academy of Sciences, Rathausplatz 4, 6850 Dornbirn, Austria
Wojciech Filipiak, Breath Research Institute, Austrian Academy of Sciences, Rathausplatz 4, 6850 Dornbirn, Austria; University Clinic for Anesthesia, Innsbruck Medical University, Anichstraße 35, 6020 Innsbruck, Austria
Hossam Haick, The Department of Chemical Engineering and Russell Berrie Nanotechnology Institute, Technion-Israel Institute of Technology, Haifa 32000, Israel
Armin Hansel, Institute of Ion Physics and Applied Physics, Leopold-Franzens University Innsbruck, Technikerstrasse 25, 6020 Innsbruck, Austria
Jens Herbig, Oncoytyrol-Center for Personalized Cancer Medicine GmbH, Eduard Bodem Gasse 3, 6020 Innsbruck, Austria; Ionimed Analytik GmbH, Eduard Bodem Gasse 3, 6020 Innsbruck, Austria
Hartmann Hinterhuber, Univ.-Clinic for Psychiatry, Innsbruck Medical University, Anichstr. 35, 6020 Innsbruck, Austria
Marieann Högman, Department of Medical Sciences, Respiratory Medicine and Allergology, Uppsala University, SE 75105 Uppsala, Sweden; Centre for Research and Development, Uppsala University/County Council of Gävleborg, SE 801 88 Gvle, Sweden
Ildikó Horváth, Department of Pulmonology, Semmelweis University, Budapest, Hungary
Michael Hubalek, Department of Gynecology and Obstetrics, Medical University Innsbruck, Anichstrasse 35, 6020 Innsbruck, Austria
Gary W. Hunter, NASA Glenn Research Center, 21000 Brookpark Road Cleveland, OH 44135, USA
Tadeusz Jezierski, Institute of Genetics and Animal Breeding of Polish, Academy of Sciences, Department of Animal Behaviour, Jastrzębiec, 05-552 Wólka Kosowska, Poland
Gennadii Kamarchuk, B. Verkin Institute for Low Temperature Physics and Engineering, National Academy of Sciences of Ukraine, 47 Lenin Avenue, Kharkov 61103, Ukraine
Julian King, Breath Research Institute, Austrian Academy of Sciences, Rathausplatz 4, 6850 Dornbirn, Austria; University of Vienna, Faculty of Mathematics, Nordbergstr. 15, 1090 Wien, Austria
Helin Koc, Breath Research Institute, Austrian Academy of Sciences, Rathausplatz 4, 6850 Dornbirn, Austria; Vorarlberg University of Applied Sciences, Hochschulstr. 1, 6850 Dornbirn, Austria
Ingrid Kohl, Oncoytyrol-Center for Personalized Cancer Medicine GmbH, Eduard Bodem Gasse 3, 6020 Innsbruck, Austria; Ionimed Analytik GmbH, Eduard Bodem Gasse 3, 6020 Innsbruck, Austria
Ievgeniia Kushch, Institute for Children and Adolescents Health Care, National Academy of Medical Sciences of Ukraine, 52-A 50 let VLKSM Avenue, Kharkov 61153, Ukraine
Jae Kwak, Monell Chemical Senses Center, 3500 Market Street, Philadelphia, PA 19104, USA
Alice M. Kwan, Department of Mechanical and Aerospace Engineering, One Shields Avenue, University of California, Davis, CA 95616, USA
D. Laskowski, Cleveland Clinic, Respiratory Institute & Pathobiology/Lerner Research Institute, Desk A90, 9500 Euclid Ave., Cleveland, OH 44195, USA
Tomasz Ligor, Department of Environmental Chemistry and Bioanalytics, Faculty of Chemistry, Nicolaus Copernicus University, 7 Gagarin St., 87-100 Toruń, Poland
Chung-Chiun Liu, Department of Chemical Engineering, Case Western Reserve University, 10900 Euclid Ave., Cleveland, OH 44106, USA
D.B. Makel, Makel Engineering, Inc., 15505 Neo Parkway, Cleveland, OH 44128, USA
Rosa Margesin, Institute of Microbiology, Leopold-Franzens University of Innsbruck, Technikerstraße 25, 6020 Innsbruck, Austria
Pekka Meriläinen, Department of Medical Sciences, Respiratory Medicine and Allergology, Uppsala University, SE 75105 Uppsala, Sweden
Wolfram Miekisch, University of Rostock, Department of Anaesthesia and Intensive Care, Schillingallee 35, D-18057 Rostock, Germany
Pawel Mochalski, Breath Research Institute, Austrian Academy of Sciences, Rathausplatz 4, 6850 Dornbirn; Univ.-Clinic for Anesthesia, Innsbruck Medical University, Anichstr, 35, 6020 Innsbruck, Austria; Institute of Nuclear Physics PAN, Radzikowskiego 152, 31342 Kraków, Poland.
Anil S. Modak, Cambridge Isotope Laboratories Inc., 50 Frontage Road, Andover, MA 01810, USA
Suvra P. Mondal, Ohio State University, Department of Chemistry, 100 West 18th Avenue, Columbus, OH 43210, USA
Paolo Montuschi, Department of Pharmacology, Faculty of Medicine, Catholic University of the Sacred Heart, Largo F. Vito, 1, 00168 Rome, Italy
Antonio Mutti, Department of Clinical and Experimental Medicine, Nephrology and Health Sciences, University of Parma, Via Gramsci 14, I-43100 Parma, Italy
Markus Nagl, Department of Hygiene, Microbiology and Social Medicine, Division of Hygiene and Medical Microbiology, Innsbruck Medical University, Fritz-Pregl-Straße 3, 6020 Innsbruck, Austria
Anna-Carin Olin, Occupational and Environmental Medicine, Sahlgrenska Academy, Box 414, SE-405 30 Gothenburg, Sweden
Daniel J. Peirano, Department of Mechanical and Aerospace Engineering, One Shields Avenue, University of California, Davis, CA 95616, USA
Joachim D. Pleil, National Exposure Research Laboratory, Office of Research and Development, US Environmental Protection Agency, Research Triangle Park, NC 27711, USA
Alexander Pospelov, National Technical University Kharkov Polytechnic Institute
, 21 Frunze Str., Kharkov 61002, Ukraine
George Preti, Monell Chemical Senses Center, 3500 Market Street, Philadelphia, PA 19104, USA; Department of Dermatology, School of Medicine, University of Pennsylvania, PA 19104, USA
Norman M. Ratcliffe, Centre for Research in Biosciences, University of the West of England, Coldharbour Lane, Frenchay, Bristol BS16 1QY, UK
Terence H. Risby, Bloomberg School of Public Health, Johns Hopkins University, 615 North Wolfe Street, Baltimore, MD 21205, USA
Brian M. Ross, Northern Ontario School of Medicine and Department of Biology, Lakehead University, Thunder Bay, Ontario, Canada P7B5E1
Joanna Rudnicka, Department of Environmental Chemistry and Bioanalytics, Faculty of Chemistry, Nicolaus Copernicus University,7 Gagarin St., 87-100 Toruń, Poland
Stefan W. Ryter, Pulmonary and Critical Care Medicine, Brigham and Women’s Hospital, Harvard Medical School, 75 Francis Street, Boston, MA 02115, USA
Michael Schivo, Center for Comparative Respiratory Biology and Medicine, 451 Health Sciences Drive, #6517, Davis, CA 95616, USA; Clinical and Translational Sciences Center (CTSC), UC Davis Medical Center, Sacramento, CA 95817, USA
Alex Schmid, Breath Research Institute, Austrian Academy of Sciences, Rathausplatz 4, 6850 Dornbirn, Univ.-Clinic for Anesthesia, Innsbruck Medical University, Anichstr, 35, 6020 Innsbruck, Austria
Jochen K. Schubert, University of Rostock, Department of Anaesthesia and Intensive Care, Schillingallee 35, D-18057 Rostock, Germany
David Smith, Institute for Science and Technology in Medicine, School of Medicine, Keele University, Thornburrow Drive, Hartshill, Stoke-on-Trent ST4 7QB, UK
Jon R. Sobus, National Exposure Research Laboratory, Office of Research and Development, US Environmental Protection Agency, Research Triangle Park, NC 27711, USA
Steven F. Solga, St. Luke’s Hospital, 801 Ostrum Street, Bethlehem, PA 18015, USA School of Medicine, Johns Hopkins University, 733 North Broadway, Baltimore, MD 21205, USA
Patrik Španěl, J. Heyrovský Institute of Physical Chemistry, Academy of Sciences of the Czech Republic, Dolejškova 3, 182 23, Prague 8, Czech Republic
Andreas Sponring, Breath Research Institute, Austrian Academy of Sciences, Rathausplatz 4, 6850 Dornbirn, Austria; University Clinic for Anesthesia, Innsbruck Medical University, Anichstraße 35, 6020 Innsbruck, Austria
Gerald Teschl, University of Vienna, Faculty of Mathematics, Nordbergstr. 15, 1090 Wien, Austria
Susanne Teschl, University of Applied Sciences Technikum Wien, Höchstdtäplatz 5, 1200 Wien, Austria
Ulrike Tisch, The Department of Chemical Engineering, Technion-Israel Institute of Technology, Haifa 32000, Israel
Jakob Troppmair, Daniel-Swarovski Research Laboratory, Department of Visceral-, Transplant and Thoracic Surgery, Innsbruck Medical University, Innrain 66, 6020 Innsbruck, Austria
Claire Turner, The Open University, Department of Life, Health and Chemical Sciences, Walton Hall, Milton Keynes, MK7 6AA, UK
Karl Unterkofler, Breath Research Institute, Austrian Academy of Sciences, Rathausplatz 4, 6850 Dornbirn, Austria; Vorarlberg University of Applied Sciences, Hochschulstr. 1, 6850 Dornbirn, Austria
Marta Walczak, Institute of Genetics and Animal Breeding of Polish Academy of Sciences, Department of Animal Behaviour, Jastrzębiec, 05-552 Wólka Kosowska, Poland
Benjamin J. Ward, Makel Engineering, Inc., 15505 Neo Parkway, Cleveland, OH 44128, USA
Anna Wenda-Piesik, Department of Plant Growth Principles and Experimental Methodology, University of Technology and Life Sciences, 20 Kordeckiego St., 85-225 Bydgoszcz, Poland
Helmut Wiesenhofer, Breath Research Institute, Austrian Academy of Sciences, Rathausplatz 4, 6850 Dornbirn, Austria
Jennifer C. Xu, NASA Glenn Research Center, 21000 Brookpark Road, Cleveland, OH 44135, USA
Konstantin Zamuruyev, Department of Mechanical and Aerospace Engineering, One Shields Avenue, University of California, Davis, CA 95616, USA
Yuriy Zrodnikov, Department of Mechanical and Aerospace Engineering, One Shields Avenue, University of California, Davis, CA 95616, USA
Foreword
Volatile compounds, the majority being organic in nature, VOCs, are continuously generated in the human body and are partially emitted via exhaled breath and through the skin. VOCs are also released by blood, urine, and fecal samples. It stands to reason that in some way these VOCs carry information on the physiological and metabolic status of the individual and it is because of this that these VOCs, often referred to as volatile metabolites or biomarkers, have attracted the attention of physicians, physiologists, and surgeons as potential aids to clinical diagnosis and therapeutic monitoring. Hence, the detection, identification, and quantification of these volatile biomarkers have stimulated the development of sophisticated analytical methods and instrumentation by which individual biomarker molecules, or combinations of different biomarker molecules, can be analyzed to sufficient accuracy to be clinically useful. The expectation is that particular patho-physiological conditions can be signaled by individual volatile biomarker molecules, and this has been realized by the close association of exhaled NO with asthma, a much researched phenomenon that is now finding utility in clinical practice. However, this exemplar is, as yet, a rare circumstance. Much research is indicating that if complex and very variable conditions, such as cancer, are to be recognized by VOC detection in exhaled breath, they are much more likely to be signaled by a combination of VOCs, some specific to the condition, but also by unusual distributions of common breath metabolites that are present in the breath of all individuals.
The status of breath research up to the year 2005 was well documented and described in the book Breath Analysis for Clinical Diagnosis and Therapeutic Monitoring
edited by A. Amann and D. Smith (World Scientific Publishing Company, Singapore). The present text reports the continued progress of this topic, not least in the development of new instrumental analytical techniques. It expands the topic toward volatiles released by other biological fluids, including urine and stool, in the last of which bacterial action plays an important, even dominating role in the generation of volatiles, and also the release of volatiles from in vitro cell and bacterial cultures that can provide valuable pointers for in vivo studies.
This book comprises chapters written by authorities in particular areas of volatile biomarkers research. Some of the chapters are focused, others are more broadly based. There are chapters on the development of analytical instrumentation, including mass spectrometry-based devices and solid state sensors, on physiological and clinical studies via breath, urine, and stool VOC analyses, on volatile compounds emitted by pathogenic microorganisms and on breath condensate and exhaled particulates. Standardization of sampling methodologies for VOCs are described throughout the chapters, a topic that is more and more being realized as critical to breath analysis, since it will allow meaningful comparisons of results obtained in different laboratories. There is little point in attempting to subdivide, order, and categorize too rigorously the chapter topics in this book, but some generalizations are attempted, as can be seen by the sectioning of the contents list.
In one sense, we have put the cart before the horse
by including as the first chapters those topics that focus on interpretation and analysis of breath concentrations data, including mathematical and statistical approaches, the physiological models of the transport of volatiles from blood to breath, and other issues and challenges in human breath research. The rationale for this is the need for these important aspects to be considered when designing new techniques for breath analysis, notably of solid state sensors that must be very sensitive, yet quite small, low in costs and easy to use by both health professionals and patients alike. Such devices contrast with the larger mass spectrometry-based analytical instruments, including gas chromatography-mass spectrometry (GC-MS) proton transfer reaction mass spectrometry (PTR-MS) and selected ion flow tube mass spectrometry (SIFT-MS) instruments, which are more research tools than monitoring instruments and are generally unsuitable for the clinical environment and point-of-care application. The important advances made in these analytical methods and instrumentation for VOC analysis are impressive. It is becoming clear that combinations of these various analytical methods offer the most profitable approach to biomarker research and discovery. This is now contributing to the new field of metabolomics that is represented in several chapters in this book.
A group of chapters deal with specific physiological and clinical studies, both in the clinic/hospital environment and in the workplace, which exploit the well-established analytical methods referred to above. A major advance has been made in sampling methodology of exhaled breath and the headspace of other biological fluids. In this sense, those techniques like GC-MS that require sample collection using bags and traps, such as solid phase micro extraction (SPME) and adsorption/thermal desorption (ATD) are not ideal, and direct sampling techniques together with immediate analyses are most desirable. The direct sampling techniques include PTR-MS and SIFT-MS and some sensors, notably those specifically designed to detect and quantify the diatomic molecules NO and CO in exhaled breath. These direct sampling techniques are described in separate chapters, which also indicate how they are utilized to monitor and study wide-ranging conditions and diseases, including asthma.
Real-time analysis of exhaled breath has been developed to a high degree in that it is now possible to directly analyze single breath exhalations for several compounds and separately analyze the end-tidal portion of exhaled breath. Even so, the composition of exhaled breath passing along the airways and through the oral cavity can become modified, which can be a serious consideration when systemic biomarkers at the parts-per-billion by volume (ppbv) and lower are intended as indicators of disease. However, mouth-exhaled breath and nose-exhaled breath can now be separately analyzed for some VOCs in just a few seconds and so it can be determined which VOCs originate in the oral cavity (by bacterial and enzymatic activity) and those which largely originate at the alveolar interface and are most likely to be truly systemic. Similarly, these real-time analyses can equally be utilized to analyze volatile compounds in the humid gas/vapor headspace of urine, cell and bacterial cultures and stool samples. The results of such in vitro studies can assist in directing the researcher toward specific biomarkers generated in vivo by mammalian cells and bacterial cells, the last analyses, for example, greatly contributing to the identification of harmful respiratory and gut bacteria. Much effort has been given to the search for volatile cancer biomarkers and there are chapters dedicated to this topic, including a chapter that describes how dogs can sniff out
tumors. Progress has been limited, but the prize of discovery of a non-invasive biomarker of early-stage tumors is driving this research ever forward.
While strictly speaking not directly concerned with volatile biomarkers, exhaled breath condensate and particulates/aerosols are closely allied to breath volatiles research and discussions of these topics are properly included as separate chapters in this book. Knowledge of the types of VOCs emitted by human urine is being exploited in the detection and location of persons trapped in collapsed buildings following natural disasters such as earthquakes. A separate chapter is given to this important topic.
Some final remarks are in order here to indicate the way forward for VOC research and for breath research in particular. Accurate on-line, real-time measurements of the concentrations of trace metabolites in exhaled breath and other biological media can now be made down to ppbv concentration in the laboratory, but a major desire is to move direct analysis from the laboratory toward the clinical setting. This offers a real challenge, but direct analyses of exhaled breath of ventilated patients in the operating theatre and intensive care units are already being explored. Yet an even greater challenge lies ahead, which is the identification and accurate quantification in real time of new biomarkers of disease and infection, many, if not most, likely to be present in exhaled breath at sub-ppbv levels. This calls for greater analytical sensitivity than is currently available by on-line analytical techniques, and so further instrumentation developments are required. When these challenges in breath analysis have been met and overcome, the analyses of volatiles above other biological media important in clinical diagnosis (viz. urine, stool, blood, open infected wounds) should be straightforward. What is certain is that only thorough experimental work and innovation will promote this difficult and challenging topic such that trace VOC analysis will become a valuable diagnostic method that will enhance and extend conventional clinical diagnostics.
We are grateful to all the authors of the chapters in this book. Without their expert research work in advancing the topic of volatile biomarkers and exploring their relevance to physiology and medicine, coupled with their skill in reporting their research, this book could not have come into being. We are especially grateful to Hossam Haick for the depiction of an exhaling person that appears on the book cover, to Venelin Chernogorov for his editorial help and to Patrik Španěl for his invaluable comments in response to the many queries posed during the editing of the chapters.
Anton Amann
David Smith
PART A Interpretation of Breath Analysis Data
Chapter 1 Mathematical and Statistical Approaches for Interpreting Biomarker Compounds in Exhaled Human Breath
Chapter 2 Issues and Challenges in Human Breath Research: Perspectives from Our Experience
Chapter 1
Mathematical and Statistical Approaches for Interpreting Biomarker Compounds in Exhaled Human Breath
Joachim D. Pleil and Jon R. Sobus, National Exposure Research Laboratory, Office of Research and Development, US Environmental Protection Agency, Research Triangle Park, NC 27711, USA
Acknowledgments
The authors thank Terence Risby from Johns Hopkins University, Baltimore, MD, USA, Anton Amann from the Austrian Academy of Sciences, Innsbruck, Austria, Michael Phillips from Menssana Research Inc., Newark, NJ, USA, Wolfram Miekisch from Rostock University Hospital, Rostock, Germany, and Jens Herbig from Ionimed Analytik, Innsbruck Austria for many insightful discussion regarding the use of human breath biomonitoring data. We also thank Linda Sheldon, Tzipporah Kormos, Karen Oliver, and Myriam Medina-Vera of US EPA, and Matthew Stiegel from University of North Carolina, Chapel Hill for their expert advice.
1.1 Introduction
The various instrumental techniques, human studies, and diagnostic tests that produce data from samples of exhaled breath have one thing in common: they all need to be put into a context wherein a posed question can actually be answered. Exhaled breath contains numerous compounds; just the volatile organic fraction alone has been estimated to represent in excess of 500 different chemical species. In addition, the aerosol fraction contains proteins, signaling molecules, dissolved inorganic compounds, and even bacteria and viruses adding to the complexity of the total sample.
No single technique can detect everything in breath, in fact, even the most broadly designed breath measurements result in suites of compounds restricted by the methods used. For example, reactive oxygen species may be observed using real-time sensors or real-time mass spectrometry (MS), but not by gas chromatography-MS (GC-MS), whereas GC-MS can discriminate among a variety of hydrocarbons, alcohols, and ketones that may overlap completely in a real-time MS instrument without benefit of chromatographic separation. Furthermore, the fraction of the breath (gas-phase or aerosol phase) also determines what measurements can be made; for example proteins and signaling molecules could be detected in exhaled breath condensate via enzyme-linked immunosorbent assay (ELISA), nuclear magnetic resonance (NMR), or liquid chromatography (LC) MS, but not with any gas-phase instruments such as those based on optical spectroscopy or gas chromatography.
The first issue that the data analyst faces is that all data he or she sees, regardless how complex or detailed, is not comprehensive but always stratified (restricted) by the chemistry, instrumentation, and thermodynamics of the choices made for sampling and analysis. The second issue is that a suite of compounds measured in any given sample does not represent an independent set of variables; sub-groups of biomarker compounds often have appreciable covariance reflecting similar metabolic pathways or exogenous sources. The third issue relates to variability; any given sample is unique and it can never be taken for granted that the constituents of the breath are the same between people nor within one person over time. Finally, any individual compound in breath can have a wide range of concentrations that are all considered normal
or unremarkable
in the apparently healthy general population. This has implications for assessing health or exposure status based on just a few data points; under such constraints, the analyst can only interpret a measurement as a statistical probability that the concentration is probative.
1.2 Data interpretation
In this chapter, we describe a series of mathematical and statistical approaches geared specifically to the interpretation of volatile organics in exhaled breath that can be implemented to address the four issues outlined above. Although all methods interact, we have assigned them to five categories for the purposes of this discussion as follows:
1. Data visualization and summary statistics.
2. Variable independence and clustering.
3. Population statistics and variance components.
4. Stochastic models and meta-data.
5. Dynamic models and longitudinal data.
Each of these approaches can provide distinct information about a breath data set, generally increasing in detail in the order listed. We note that the broader interpretations gleaned from categories 1–3 feed the modeling processes in 4 and 5. We further note that not all of these procedures need to be performed; often it is sufficient to answer a question with a simple analysis based on a graph or a summary table. In the following, we describe these categories and show how they can progressively tell a story about acquired data. We assume that multiple compounds have been measured across a number of people, subjects may be grouped as cases and controls, host factor meta-data have been collected, and measurements may have been repeated, either with or without intervention or treatment.
1.2.1 Data visualization and summary statistics
The first step in any analysis of newly acquired data is to get a feel
for how the experiment turned out. Generally, we make graphs and calculate averages, standard deviations, data ranges, etc. We have found that it is very helpful to get all of the data into some form of visual representation, either as bar graphs, cluster plots, or heat maps, depending on complexity of the data structure. For example, in a recent publication, we described measurements of a series of polar volatile organic compounds (PVOCs) in exhaled breath condensate (EBC) made during an intervention study of diesel exhaust exposure.¹ Human subjects (3 males and 6 females) were each exposed for 2 h to a dilute diesel exhaust atmosphere and to a purified air atmosphere on separate occasions.² A variety of samples, including EBC, were collected immediately before and after the exposures, and again 24 h later; EBC means data for eight of the most prevalent compounds were plotted by nine subjects grouped by gender (Figure 1.1). We noted a slight gender bias for some compounds, but otherwise the results were unremarkable and we went on to assess the data from a statistical perspective. In a subsequent publication, however, we developed methodology for data visualization using heat map style graphics.³ Briefly, heat maps are visual representations of quantitative data on two axes; the x-axis reflects individual samples and the y-axis consists of groups of measured parameters. The field between the axes is comprised of an array contiguous boxes color coded to reflect quantitation. The term heat map
is derived from the convention that the higher levels tend toward red, the lower levels tend toward blue.
Figure 1.1 Bar graph visualization of balanced exhaled breath data from an intervention study of diesel exhaust exposure. Summary breath data for nine subjects and eight compounds.
We revisited the data set from Hubbard et al.¹ and found that heat map visualization was capable of showing more pattern detail. This technique had to be restricted to seven subjects with complete data to avoid blank spots on the heat map as we lost a few samples to follow-up. Figure 1.2 shows this alternate approach; note that we now have access to results from individual subjects’ samples, for all 16 measured PVOCs, and the flexibility for grouping samples by gender and longitudinal parameters. Here, the gender effect becomes obvious; males were expressing much higher levels of many of the measured PVOCs, especially 2-methyl propanal, 1-heptanol, butanal, pentanal, 1-hexanol, and 3-methyl-3-pentanol. Only hexanal and heptanal reverse this trend. We further see that there is no apparent treatment effect or longitudinal time effect, that is, there are no obvious pattern differences between diesel exposures and air exposures, nor pre- and post-exposure data. These simple visualization approaches, coupled with summary statistics, can provide hints as to how more detailed mathematical approaches could be implemented to quantify these observations. Also, we note that the visualization methods for complex data make the subsequent statistical calculations more accessible to the readership.
Figure 1.2 Heat map visualization of balanced exhaled breath data from an intervention study of diesel exhaust exposure. Individual samples grouped by longitudinal time frame (pre-exposure, post- exposure, and 24 h post-exposure) as well as by gender.
1.2.2 Variable independence and clustering
One of the most vexing problems encountered in data interpretation procedures is the determination of actual independence of what we generally denote as independent variables.
Subsequent statistical evaluations rely on the notion that measurements that are used in models are not overly correlated. For example, consider the height and weight parameters of human subjects as host-factor data in a complex breath experiment. We know that taller people tend to be heavier, and so there is significant correlation expected. If one were to treat height and weight as independent and place them both into a model for predicting a health outcome (along with a series of breath biomarkers), one can get completely different interpretations as to their respective importance depending upon which was entered first. This is why one often sees body-mass index (BMI) used as a composite parameter instead. Now consider unexpected correlations among variables, for example, among compounds measured in the exhaled breath of subjects in an intervention or case-control study. If certain compounds are repeatedly behaving the same way, then their inclusion in even simple multivariate regression models will result in mathematical instability.
We cannot know a priori if different analytes are tracking the same outcome, however, we can perform various correlation tests among presumed independent variables to assess the degree of independence. The most fundamental method is the correlation matrix wherein the regression between pairs of variables (Vi, Vk) is calculated as the Pearson correlation coefficient "r" which has possible values ranging from −1 to 1. Positive r-values indicate that a larger Vi is associated with a larger Vk, whereas a negative r indicates that a larger Vi is associated with a smaller Vk. The closer the r-value is to 0, the more independent the two variables. For example, in a study measuring height (H), weight (W), and percent body fat (BF), r = 0.486 for W vs. H, r = 0.074 for H vs. BF, and r = 0.613 for W vs. BF.⁴ Based on the results of this simple correlation matrix, one would probably not include both parameters H and W, nor both W and BF, but would keep H and BF as host factors in a resulting model.
The correlation matrix approach is considered a semi-quantitative measure for overall data independence interpretation because it treats variables two at a time; one cannot discern directly how multiple variables, or combinations of variables, correlate. A more sophisticated approach is available wherein Eigenvector projections are calculated for all variables in n-dimensional space. Using statistical software such as proc VARCLUS from SAS Inc. (Cary, NC, USA), it is possible to develop a dendrite
diagram that can be used to create clusters of variables that have certain levels of co-linearity. This is similar to principle components analysis (PCA), but rather than grouping samples, this approach groups variables. Variable clustering serves two purposes, it improves model stability by identifying/removing collinearity, and reduces the total number of variables for a more parsimonious model.⁵
As an example, consider the dendrite diagram (dendrogram) in Figure 1.3 where we show a generic example of both host factor and measurement data clustering.⁵ In this case, we use a hypothetical data structure with six host factors (gender, BMI, age, etc.) and seven environmental/biomarker variables (e.g. concentration measurements in blood or urine). This methodology has been applied to complex environmental dioxin congener data⁶ and to studies of jet fuel exposure in the US Air Force.⁷
Figure 1.3 Example dendrogram of variable cluster analysis. Starting with a total of m = 13 independent variables (p = 6 host factors and q83% of the variance, but now is difficult to interpret.
By collapsing the more highly correlated variables into clusters, we do not lose much explanatory power, but improve the interpretive power of subsequent models by increasing the ratio (n/m) between number of samples "n and the number of independent variables
m." We have observed that n/m 10 is a good rule of thumb for assessing the importance of individual variables.⁶ We caution that this procedure requires a certain amount of judgment on the part of the investigator. The first issue is choosing the variance level vs. number of clusters. This is very dependent on the eventual n/m parameter. The second issue is how to combine information from variables collapsed into clusters. There are a number of choices that depend on the nature of the variables; in the above example, we use a simple sum, but other strategies can be employed as well. For example, given highly correlated variables such as HF3 and HF4 (especially if they are not continuous), one can just discard one or the other. For measured concentrations, it is sometimes better to normalize each measurement to a total to avoid having a particular variable overwhelm the sum.
1.2.3 Population statistics and variance components
Traditional mathematical analyses of breath biomarkers, both for clinical and environmental research applications, utilize measurement statistics (e.g. mean and median) to evaluate differences between groups. For example, it is common to evaluate measurements statistics to compare cases vs. controls, exposed vs. unexposed subjects, or males vs. females (as shown in our earlier example of a diesel intervention study). Furthermore, it is common to employ stratified data analyses—that is, analyses using multiple levels of organization—to reduce the impacts of confounding variables. Consider the evaluation of smoking effects on breath biomarker levels; one could perform a single evaluation using all subjects (smokers vs. non-smokers), two evaluations after stratifying by sex (male smokers vs. male non-smokers; female smokers vs. female non-smokers), four evaluations after stratifying by sex and disease status (male smoker [case] vs. male smoker [control]; male non-smoker [case] vs. male non-smoker [control]; female smoker [case] vs. female smoker [control]; female non-smoker [case] vs. female non-smoker [control]), and so on. These stratified analyses can narrow down the potential origins of an observed effect, but require larger sample numbers with each level of stratification. Thus, an investigator must balance the desired outcome of a given analysis with costs required to achieve sufficient statistical power and sensitivity.⁶
Given adequate sample numbers for stratified group analyses, it is prudent to first investigate the underlying distribution(s) of the biomarker data in question. Our earlier examples for data analysis (i.e. data visualization and variable clustering) can be considered qualitative or semi-quantitative, and generally have no a priori conditions for data structure. Hypothesis testing on the other hand, is entirely quantitative, and in many cases relies on distributions of measurement data being approximately Gaussian, or normal.
Simple diagnostic procedures exist in most software packages to evaluate data distributions. Often times, simple histograms, normal probability plots, or quantile-quantile plots can be visually inspected to evaluate normality assumptions. More advanced statistical packages offer statistical tests (e.g. Shapiro-Wilk and Kolmogorov-Smirnov) to confirm results from visual inspection.
Measurements of analytes in biological media are often lognormally
distributed; that is, the distribution of the logged values is approximately normal. Lognormal data can be identified by a right-skewed distribution of the original (non-transformed) data, reflecting few values at exceedingly high levels, and many values near, but not below, a lower threshold (generally zero). Additional signs of lognormally distributed data include: (1) an observed arithmetic mean value that is larger than the median (reflecting the differential effects of extremely high values on these statistical parameters) and (2) an observed confidence interval that includes negative values (it is impossible to have negative amounts of a biomarker). When one encounters these signs, it is best to evaluate data transformation approaches before pursuing quantitative hypothesis testing.
Parametric
testing procedures are generally used to evaluate normally distributed data and lognormally distributed data that have been log-transformed. Commonly used parametric tests include the Student’s t-test, the paired t-test, and analysis of variance (ANOVA). A Student’s t-test evaluates differences in biomarker measurements across two groups (e.g. men vs. women); a paired t-test evaluates differences in biomarker measurements between paired samples from individuals (e.g. pre-intervention vs. post-intervention); and ANOVA compares biomarker measurements across multiple groups (e.g. children vs. adolescents vs. adults). In the event that biomarker data are not Gaussian, or cannot be mathematically transformed to approximate normality, non-parametric
methods exist for hypothesis testing. For example, the Mann-Whitney test, Wilcoxon test, and Kruskal-Wallis test are common substitutes for the Student’s t-test, paired t-test, and ANOVA, respectively. These tests generally approximate the results of the parametric tests given a large enough sample size. However, the exact data requirements (e.g. sample number, sample independence, random selection) for any given test should always be considered prior to hypothesis testing.
The statistical tests discussed thus far generally utilize a single observation for each subject. (While the paired t-test uses two observations per subject, the difference between the two values [i.e. a single value] is used for hypothesis testing.) Therefore, different approaches to data analysis are required when multiple measurements exist for each subject. At the most basic level, it is of interest to evaluate the spread of the data, or the variance,
between and within subjects. To do this, the total variance across all measurements is first partitioned into that which is observed across repeated measurements of individual subjects, known as within-person (intra-individual) variance,
and that which is observed across average levels of all subjects known as between-person (inter-individual) variance.
These between- and within-person variance components can be calculated using a number of techniques.⁸ The simplest of these methods is analysis of variance (ANOVA) which is suitable for balanced data sets where the same number of measurements exists for each subject. More complex methods, such as restricted maximum likelihood estimation (REML), may be required when working with unbalanced data sets; these methods are typically available only in more advanced statistical software packages.
Once the variance components are established, they become instrumental for identifying the origins of exposure or disease, and in turn, the best opportunities for mitigation or intervention. Large within-person variance and small between-person variance indicates little difference between subjects on average, but large differences over time for each subject. This result may point to a temporal event, random or otherwise, that affects each subject equally. Alternatively, small within-person variance and large between-person variance indicates little change over time, but marked differences between subjects. This result may point to a host-specific parameter (e.g. genotype, fitness-level) that dictates biomarker response. In the first example, a general intervention strategy, equally applicable to all subjects, might be suitable to reduce an exposure or limit a biological response. In the second example, a targeted strategy would likely be necessary to first identify the cause of elevated biomarker levels for certain individuals, and then for intervention.
1.2.4 Stochastic models and meta-data
After identifying group-based differences in biomarkers levels, and within- and between-subject variance components (for repeated-measures studies only), the next step is to develop statistical models using study meta-data. Statistical models serve two functions for breath research: (1) they allow investigators to simultaneously identify multiple significant predictors of breath biomarker levels and (2) they provide a platform for predicting breath biomarker levels in other studies where only meta-data exist.
An important decision for model development is identifying a dependent variable; this is not as easy as it sounds. Often, the dependent variable is obvious by design and represents the outcome
for a subject; it can be binomial, that is, cancer/not cancer, infected/not infected, or it can be a continuous health outcome variable such as cholesterol level, total urinary protein, pulmonary function (e.g. forced expiratory volume in 1 s (FEV1), forced vital capacity (FVC)), or DNA damage (e.g. sister chromatid exchange (SCE), strand breaks), among others.
However, there are many occasions when the choice of the dependent variable is not obvious, especially in environmental or cross-sectional public health studies for which the analyst was not consulted in sampling design. For example, suppose one has measured a suite of exogenous chemicals and metabolites in exhaled breath, and has acquired meta-data concerning recent activity, job type, height, weight, gender, ethnicity, etc., but no health effects information was collected or observed. What could the dependent variable be for interpretation purposes? In such cases, the first question to be addressed is: What do we want to know? Generally, we would like to explore the linkages between suspected exposure sources and the resulting internal dose in human subjects. Suppose that in the totality of all measurements, we observe benzene, toluene, ethyl-benzene, and xylenes (BTEX), plus many other organic compounds in breath. If one of the suspected sources for all exposures is automobile exhaust, we could sum the BTEX numbers into one composite parameter and use this to represent the dependent variable for overall fuels exposure. This is a bit of a bootstrapping approach, especially if we keep the individual BTEX compounds as independent continuous variables, but often this is the only way to build a stochastic model in the absence of a designed dependent variable.
Given that we have a number of independent variable
measurements and related meta-data and some consensus dependent (outcome) variable, then the next step is to determine which independent variables and data actually tell a story relating to the changes in the identified outcome variable. This requires some form of a modeling approach, and if the data include both host factors and continuous variables, the best approach is a generalized linear model, or a mixed effects
model.⁸,⁹
A general form of such a model is:
where Yhij is the value of some relevant biological parameter for the jth observation of the ith subject in the hare the values of the fixed effect variables such as environmental chemical concentrations (in air, water, food, dust, etc.), and host factors such as age, health state, gender, genetic polymorphisms, ethnicity, etc.; p is the total number of fixed effects (note: the host factors may be fixed for all j within a given ip h is the random effect for the hth group; bhi is the random effect for the ith subject from the hhij is the residual (unexplained) error for the jth observation of the ith subject from the hth group.
Software applications for this style of approach are commercially available (e.g. proc MIXED, SAS). Upon calculation, the coefficients and their p-values can be interpreted to determine the effect of including the particular fixed effect or random effect variable in the final model for explaining the variance in the biological parameters’ values. This can be done with iterative steps of forward addition or reverse elimination with the eventual objective being a parsimonious model without appreciable loss of modeling power. Once the final model is established, we can observe which exposure parameters and fixed effects are more likely to cause perturbations to the systems biology.⁵
In an earlier example, based on published results from a diesel exhaust intervention study, we demonstrated how visualization tools can be used to identify likely predictors of breath biomarker levels. Specifically, we used a 3D bar graph and a heat map to show the likely effect of gender, and the unlikely effects of the exposure intervention and sample time, on PVOC levels in EBC. Following from this qualitative work, we now demonstrate a statistical assessment of this data using mixed models. We note that these methodologies and results have been published alongside the aforementioned graphical work.¹
PVOC levels in breath were treated individually as model dependent variables, whereas model independent variables included fixed effects for the time of sample collection (pre-intervention vs. post-intervention vs. 24 h post-intervention), the type of intervention (diesel exhaust vs. purified air), and gender (male vs. female), as well as random effects for subject and residual error. Results of the models are shown in Table 1.1, and confirm a significant gender effect (p < 0.1) for 4 out of 9 modeled PVOCs, with men having higher levels than women in all cases. While not shown in Table 1.1, results also confirm no significant effects (p > 0.1) of collection time or exposure intervention on these PVOCs. These quantitative results corroborate those of the earlier qualitative analyses. Furthermore, these results include variance components estimates (listed under Random effects
in Table 1.1) that can be used to explore subject susceptibilities and possible intervention strategies.¹⁰,¹¹
Table 1.1
Mixed models results for PVOCs in EBC from subjects enrolled in a diesel exhaust intervention study.
: Estimated between-subject variance.
: Estimated within-subject variance.
NS: Not significant.
aParameter estimates for male subjects; female subjects were the reference group (i.e. Est = 0).
1.2.5 Dynamic models and longitudinal data
Dynamic models are used to interpret time dependence. In some study designs, the breath data structure investigates certain applied or observed external conditions. In diagnostic medicine, this could be a pre/post-drug treatment study, or a time-dependent study to monitor post-operative recovery status. In environmental studies, this could be a sample time series to determine the rates of uptake and elimination from different profiles of deliberate or incidental exposures to chemicals. Regardless of the exact design, the primary focus is temporally resolved data. Such studies are invaluable for deducing the time constants of absorption, distribution, metabolism, and elimination of exogenous compounds like pharmaceuticals and environmental pollutants, and for assessing different metabolic pathways. The underlying assumptions of such a study are that kinetic and physiological parameters measured under controlled conditions will be consistent for other exposure or treatment profiles. This allows us to model different scenarios to assess internal dose and metabolism.
For example, consider that we have gathered a series of breath samples pre-, during-, and post-exposure to a well-defined concentration of a specific compound.¹²–¹⁴ There are a number of ways of interpreting such data using classical pharmacokinetic principles. Researchers including Wallace from US Environmental Protection Agency (EPA), Weisel from Rutgers University, and Raymer from Research Triangle Institute constructed closed form solutions from dynamic measurements to assess the time constants and internal (hypothetical) compartments of the human body such as blood and lymphatic fluids, highly perfused tissues including organs and muscle, poorly perfused tissues including tendons, bone marrow, and connective tissues, and finally adipose tissue.¹⁵–¹⁷ These models were based on simple exponential uptake and decay behaviors with different time constants assigned empirically to the hypothetical compartments. This provided important information regarding the internal distribution of specific compounds, but did not provide organ-specific dose, nor did it allow generalizing to random intermittent exposures.
Subsequently, new methods were developed for interpreting observational kinetic data and estimating response to input functions for the general case. These included very complex physiologically based pharmacokinetic models (PBPK) that require a great deal of specific data, and also prior knowledge about organ dose kinetics, partition functions, and diffusion parameters that are generally only available from invasive animal studies.¹⁸,¹⁹ Although PBPK models are powerful tools, they are beyond the scope for this discussion.
A hybrid method was developed by Pleil et al.,¹² wherein the closed form solution was replaced with an iterative solution. This approach requires a starter set of empirical data linked precisely to well-characterized exposure data. Typically, this is generated in a human exposure chamber.²,²⁰ The methodology is based on conversion of differential equations to difference equations that can then be easily evaluated in standard spread sheet software as follows.
In the simplest form, consider that one has concentration measurements of the central compartment (blood), Cblood(t), and matching environmental air measurements, Cair(t) for some exogenous compound. One further assumes some stable (first order) rate constant for inhalation uptake (ku) and some aggregate first order elimination rate constant ke in 1/time units. Under the assumption of linear kinetics,
For a well-controlled air concentration profile (e.g. Cair(0) = 0 and Cair(t > 0) = C0), this differential equation is easily solved with the closed form solution:
when the boundary conditions are applied. This function can then be fitted to the real data of the proposed step function exposure as could be generated in an environmental chamber, and the rate constants ku and ke can be estimated. Regrettably, such closed form solutions are only possible in the simplest conceptual models. In general, there are multiple compartments (blood, highly perfused tissues, poorly perfused tissues, adipose (fatty) tissues, connective tissues, etc.) that all have different capacities and time constants. Each compartment may further exhibit different elimination pathways and time constants as well. As such, the only way to model behavior is via iterative computational methods. We have developed a difference equation approach to compute biomarker concentrations from essentially any complex system given initial conditions, some kinetic data, and a reasonable conceptual model of the compartmental structure.
Using the simple example above, one can rewrite the differential equation as a difference equation in the form of:
which can be expanded for any arbitrary time increment as:
and subsequently rearranged as:
which is a form wherein each future concentration can be calculated from the previous point for any complex exposure function Cair(t) as long as the rate constants and initial conditions are known. This iterative procedure is successful for more complex models for which we can draw a conceptual diagram because we never need to deduce a closed form solution. Although beyond the scope of this discussion, Pleil²¹ has presented a detailed discussion and application of these techniques for blood and breath biomarkers, multiple compartments, and human metabolites.
Figures 1.4a and 1.4b shows applications of this procedure to time-dependent blood and breath data for methyl tertiary butyl ether (MTBE) chamber exposures and the resulting phase-1 human metabolite tertiary butyl alcohol (TBA). Note that breath data have the same kinetic behavior as the blood data and that the TBA metabolite exhibits biological damping (delayed response) making it a longer time-frame marker for MTBE exposure.