You are on page 1of 79

Contents

Introduction 3
VOCs in breath 3
Non-invasive biomarkers 3
Scope of this guide 6
Historical breath analysis 6
Breath analysis in the modern era 6
Established breath tests 8
Breath analysis in clinical trials 8
What are VOCs? 10
Where do breath VOCs originate from? 12
VOCs as biomarkers 16
Metabolomics vs. Genomics 16
Metabolomics for diagnosis and precision medicine 16
Whole body blood sampling via breath 18
Breath Biopsy vs. Liquid Biopsy 21
Applications of VOC biomarkers 23
Wide disease relevance 24
Novel Biomarkers 25
Precision Medicine applications 25
Clinical Trials patient matching 27
Drug metabolites and disease monitoring 27
Early detection and screening of cancer 27
Asthma and COPD 28
Oncology 32
Inflammatory diseases 39
Liver disease 42
Infectious disease 44
How to analyze VOC biomarkers 49
Overcoming challenges in breath sampling and analysis 50
Study design 51
Breath Analysis Workflow 54
Reliable collection of breath samples 54
VOC Biomarker analysis 57
Quality control 59
Data analysis 61
Beyond Breath: VOC Analysis in other Biospecimens 64
Biomarker Services 65
Breath Biopsy Products 70
Further reading 72
References 74

2
Introduction

VOCs in breath
Exhaled breath contains over 1,000 volatile organic compounds (VOCs).
Accessible by a completely non-invasive Breath Biopsy, these VOCs can be
used as disease biomarkers for early detection and screening, diagnosis, and
precision medicine including patient stratification for therapy selection and
treatment monitoring (Figure 1).

VOCs are products of metabolic activity taking place in the body, hence they
directly reflect the current state of cells, tissues and the microbiome - providing
a rich source of valuable information about the health of an individual.

Metabolic information from VOC biomarker analysis collected during a Breath


Biopsy offers researchers and medical doctors new possibilities beyond those
provided by purely genomic approaches. VOCs provide detail on current
disease activity rather than just on the blueprint for disease predisposition.

Non-invasive biomarkers
As exhaled breath sampling is also inherently non-invasive, Breath Biopsy is
preferable to patients over uncomfortable and potentially unpleasant
diagnostic procedures, such as stool sampling, blood sampling, and tissue
biopsies.

3
BREATH BIOPSY:
NON-INVASIVE VOC BIOMARKERS FOR DIAGNOSTICS
AND PRECISION MEDICINE

APPLICATIONS

CLINICAL PRECISION
DIAGNOSTICS MEDICINE

CANCER INFECTIOUS INFLAMMATORY


DISEASE DISEASE

Figure 1. Breath Biopsy is a completely non-invasive means to access valuable information about the health of individuals.
VOC biomarkers have numerous applications in clinical diagnostics and precision medicine across a wide range of
diseases.

4
Why breath?
Metabolites relate directly to current disease activity
Breath enables whole body blood sampling
VOC biomarkers in breath are relevant for a wide range of diseases
Breath biomarkers have wide-ranging applications in diagnostics and
precision medicine
Sampling is completely non-invasive and pain-free

Learn more at

owlstonemedical.com/breath-biopsy
www.

5
Scope of this guide
In this guide we introduce VOCs, and discuss the scientific basis for the
application of VOCs in breath as biomarkers for a wide range of diseases. We
provide an overview of current status of VOC biomarkers in the disease areas
of asthma, oncology, inflammatory and infectious disease.

We discuss some of the challenges that have previously limited the potential of
breath analysis, and describe how these challenges have now been overcome
with the introduction of new technology for breath sampling and analysis.

We will outline the workflows for breath collection and VOC analysis that have
introduced Breath Biopsy as a new diagnostic modality. Now, for the first time,
it is possible to perform robust biomarker discovery using breath as a sample
type, on a platform with the potential to transition to point-of-care.

Historical breath analysis


Clinicians have long recognised that certain breath odors are typical of some
diseases, though the biological origin of the smells has often been a mystery.
The practice of using this sensory information to identify disease was reported
by Hippocrates (460370 BC) who famously described fetor oris and fetor
hepaticus in writings about breath aroma and disease [1].

One example familiar to doctors for centuries is the fruity smell in exhaled
breath of patients with diabetic ketoacidosis. This symptom is due to ketones
produced when the body is unable to use glucose due to the lack of insulin.
Instead, it breaks down fat as an alternative source of fuel, producing pear
scented acidic ketones including acetone [2].

Nobel Prize-winning work of Linus Pauling


The expression of VOCs in breath that are linked to a patients disease state
clearly offers a powerful, non-invasive approach to identifying illness. The
promise of using breath borne VOCs directly as a tool for diagnosis has been
recognised for many years. Twentieth century attempts to quantitatively
identify VOCs in breath include those of Nobel Prize-winning chemist Linus
Pauling, whose results [3,4] highlighted that exhaled breath contains hundreds
of different volatile compounds.

During this era, the first discoveries relating to the origin of certain VOCs in
breath were made. This included the identification of small chain hydrocarbons,
which were found to be products of lipid peroxidation, possible biomarkers for
oxidative stress [5]. Another example is the abundant breath VOC isoprene
[68], which was established to be produced in the mevalonate pathway, and
subject to large increases in concentration during physical exertion [9,10].

Breath analysis in the modern era


Since the early discoveries, the field of breath research has benefitted from the
development of and improvements in technologies for the separation and
identification of gas phase compounds.

6
Technologies for breath analysis
Techniques well suited to the analysis of breath VOCs include gas
chromatography-mass spectrometry (GC-MS), proton transfer-reaction mass
spectrometry (PTR-MS) [11], selected ion flow tube-mass spectrometry
(SIFT-MS) [12], arrays of gas sensors (eNoses) and ion mobility spectrometry
(IMS) including field asymmetric ion mobility spectrometry (FAIMS) [1316].

Reliable breath collection is essential for robust analysis of


breath VOCs.
Owlstone Medicals ReCIVA Breath Sampler is the first breath sampler able to
provide an out-of-the-box means to reproducibly capture breath samples.
ReCIVA samples breath continuously over a few minutes, enabling the VOC
biomarkers originating from throughout the body to be accessed
non-invasively and with high sensitivity.

Learn more on page 54.

Rising number of publications on breath VOCs


The improvement in analytical technologies and the obvious promise of VOCs
as biomarkers for disease has led to increasing numbers of publications
relating to VOCs in breath [1], illustrated in Figure 2.

ANNUAL NUMBER OF BREATH VOC PUBLICATIONS


250
volatile organic
compound AND breath
200
NUMBER OF PUBLICATIONS

150

100

50

0
1980 1985 1990 1995 2000 2005 2010 2015
YEAR
Figure 2. Increase in annual publications in the area of VOCs and breath. Graph shows the results
of a Web of Knowledge search for publications including the terms volatile organic compound AND
breath.

We have been using FAIMS for almost five years and have found it able
to non-invasively detect a broad range of diseases, including cancer,
with high sensitivity and specificity.
Professor James Covington, University of Warwick

7
Established breath tests
Some breath tests that use volatile biomarkers have already become
established in clinical practice (Table 1 shows approved breath tests). However,
much of the potential shown by VOC tests has remained untapped due to a
lack of reproducibility between studies looking for disease biomarkers. The
variability in the results of breath research has stemmed largely from a lack of
high quality, standardised breath sampling technology [17,18].

APPROVED BREATH TESTS


Test VOC

Capnography CO2

C-urea breath test for detection of H. pylori infection [19]


13
CO2
13

Test for neonatal jaundice [20] CO

Disaccharide adsorption deficiency and small intestine bacterial H2, CH4


overgrowth syndromes [2124]

Monitoring asthma therapy [2529] NO

Breath test for heart transplant rejection [30] Alkanes

C-octanoic acid Gastric emptying breath test (GEBT) [31]


13
CO2
13

CO testing smokers [32] CO

Roadside intoxication tests EtOH

Table 1. Approved breath tests established in clinical practice [1].

Breath analysis in clinical trials


Analysis of all clinical trial data [33] shows that between 1997 and 2016 around
450 trials have used or are using some kind of breath analysis in their workflow
https://www.owlstonemedical.com/media/uploads/files/Breath-analysis-clinical-trials-March-2017.xlsx
(download Excel spreadsheet of breath analysis clinical trials). During this time
https://www.owlstonemedical.com/media/uploads/files/Breath-analysis-clinical-trials-March-2017.xlsx
period there has been a rise in the annual number of clinical trials starting that
used breath analysis, indicating the increasing interest in tests of this type
(Figure 3).

CUMULATIVE NUMBER OF CLINICAL TRIALS


INVOLVING BREATH ANALYSIS
500

450
CUMULATIVE NUMBER OF TRIALS

400

350

300

250

200

150

100

50

0
1995 2000 2005 2010 2015
YEAR
Figure 3. Increasing numbers of clinical trials using breath analysis are being registered. This
includes 75 trials analysing exhaled breath condensate. Data from https://clinicaltrials.gov/.
https://www.owlstonemedical.com/media/uploads/files/Breath-analysis-clinical-trials-March-2017.xlsx
Download the list of trials as Excel spreadsheet.
https://www.owlstonemedical.com/media/uploads/files/Breath-analysis-clinical-trials-March-2017.xlsx

8
The identified clinical trials cover a wide range of conditions, demonstrating
the breadth of applications where breath analysis is thought to be useful. A
breakdown of the generalized disease category of the trials is shown in Figure
4.

DISEASE CATEGORIZATION IN 448 CLINICAL TRIALS


USING BREATH ANALYSIS

CANCER

INFLAMMATORY

INFECTIOUS

OTHER (GASTROINTESTINAL)

OTHER (PULMONARY)

LIVER DISEASE

OTHER

Figure 4. Breakdown of disease category of 448 clinical trials using breath analysis in their
workflow. Data from https://clinicaltrials.gov/. https://www.owlstonemedical.com/media/uploads/files/Breath-analysis-clinical-trials-March-2017.xlsx
Download Excel spreadsheet including list of clinical
htps:/w w.owlstonemedical.com/ edia/uploads/files/Breath-an lysi -clincal-trials.xls
trials.

The worlds largest breath-based clinical trials


Owlstone Medicals breath sampling and analysis technology is being used in
large clinical trials relating to lung cancer [34], colorectal cancer [35] and
asthma [36]. It is also already used in more than 60 academic research groups
and clinical labs around the world.

Breath Biomarker Services


Identify novel non-invasive VOC biomarkers

owlstonemedical.com/services
www.

9
What are
VOCs?
Volatile organic compounds (VOCs) are gas phase chemicals that have a high
vapor pressure at room temperature. VOCs form part of the composition of
exhaled breath, and represent the volatile fraction of metabolites. These are
low molecular weight (<1 kDa) compounds produced by the bodys
metabolism, and represent the endpoint of gene, transcription and protein
expression (Figure 5).

Illnesses can lead to a loss of metabolic balance [37,38], hence monitoring


fluctuations in metabolites offers a powerful way to detect, diagnose and
monitor disease.

Thousands of breath borne metabolite VOCs have already been identified [39],
and they form a library of information about an individuals health. VOCs are
also emitted from other clinical samples types such as urine, feces and saliva
(see section on Beyond Breath on page 64).

Breath Biopsy offers researchers and clinicians relevant information about


disease activity because the identity and concentration of VOCs corresponds
to current metabolic processes in cells, tissues and the microbiome. This is in
contrast to genomic analyses, which can mostly only provide information
about an individuals starting blueprint.

METABOLITES: THE ENDPOINT OF GENE,


TRANSCRIPTION AND PROTEIN EXPRESSION

gene mRNA protein metabolite

GENOMICS TRANSCRIPTOMICS PROTEOMICS METABOLOMICS

Figure 5. Metabolites provide information about actual disease activity, in contrast to genomics
which only provides a starting blueprint.

10
O
O

O
OH
H H
N O
O
OH

OH
O
H N
H
HO 1
O O
OH H
H

HN OH
NH 1 H O

HO
NH
1

HN
H
H
O

H
O
H
H

H
HO
OH
H

O
O

HN
NH1

OH

O
HN

NH
HO

1
H

O
H

Whole body
blood sampling
Breath provides a window into a much larger volume of blood. By
sampling breath over a few minutes, VOC biomarkers originating
from all around the body are captured with high sensitivity.

Learn more on page 18.

11
Where do breath VOCs originate from?
The VOCs in exhaled breath have a number of potential sources. They can be
produced biochemically by an individuals cells and tissues (endogenous
VOCs) or be the result of external influences. These exogenous VOCs in breath
can be the product of metabolism by the bodys own resident microbiome, or
may be introduced from the environment, including downstream drug
metabolites from therapeutic interventions (Figure 6).

BREATH VOC SOURCES

>1,000 VOLATILE
BREATH
O
METABOLITES O

O
OH
H H
N O
O
OH

OH
O
H N
H
HO 1
O O
OH H
H

HN OH
NH 1 H O

HO
NH
1

HN
H
H
O

H
O
H
H

H
HO
OH
H

O
O

ENVIRONMENT

HN
NH1

OH

O
HN

NH
HO

1
H

O
H

DRUGS GENOMICS

MICROBIOME

LIFESTYLE/ENVIRONMENT PHENOTYPE GENOTYPE

External factors like diet, exercise, A snapshot of the current state Our genes can be markers of
medications, microbiota and even of health that can be used to predisposition or presence of disease,
where we live influence our prevent, diagnose and treat but they do not necessarily give
metabolic state. disease or improve health. information about disease activity.

Figure 6. Endogenous and exogenous sources of VOCs found in exhaled breath.

Endogenous VOC sources


Endogenous breath VOCs can arise from metabolic activity in lung and airway
tissue, or can have a systemic origin. Systemic metabolite VOCs can come from
anywhere in the body and are carried to the lungs by the circulatory system,
where there is a rapid exchange between the blood and air in the alveoli
(Figure 7). By measuring biomarkers from all these sources, Breath Biopsy can
provide a rapid and non-invasive means to gain detailed insight in the
functioning of the whole body.

12
ORIGINS OF VOCs IN BREATH

Figure 7. Schematic representation of detailing origins of VOCs in breath. 1. Local cellular


metabolites 2. Exogeneous compounds e.g. microbiota and 3. Compounds originating from
elsewhere in the body are transferred to gas in the alveoli [39].

Biochemical origins of VOCs


A variety of processes lead to alterations in the VOC profile. The origins of
disease related VOCs are complex, as they can have multiple potential sources.
They can originate from metabolic processes taking place within, for instance,
a tumour cell, as well as from metabolic changes in the surrounding tissues
reacting to the cancers presence.

Oxidative stress (OS) has been implicated in the initiation and development of
many systemic and organ-specific diseases, including cardiovascular,
pulmonary, autoimmunological, neurological, inflammatory diseases and
cancer [1]. OS is also known to play an important role in the production of VOC
biomarkers [40].

OS occurs when there is an imbalance between the rate of metabolic


production of reactive oxygen species (ROS) (Figure 8) and the rate at which
the body can detoxify the reactive intermediates they produce, and repair the
damage these compounds cause.

At the cellular level, higher than normal ROS levels damage key cellular
structures like phospholipid membranes, proteins and DNA. This damage
causes dysfunction, and can lead to apoptosis or programmed cell death.

Many VOCs are produced by these ROS interactions with cellular structures,
including alkanes ethane and pentane, nitric oxide and aldehyde
2-hydroxy-nonenal (Figure 8) [40].

As OS is implicated in so many conditions, and known to be important in


cancer and inflammatory diseases, OS related VOCs in breath are of particular
interest.

13
OXIDATIVE STRESS RESULTS IN VOC PRODUCTION

Reactive Oxygen
Species (ROS)

Heme Amino acid


iNOS Cell structure damage
Ox-1 metabolism

DNA damage Protein oxidation Lipid peroxidation

Volatile organic
compounds (VOCs)

Figure 8. Oxidative stress, which is implicated in many diseases, is linked to the production of VOCs
from ROS damage to cell structures and other interactions. Adapted from [40].

Exogenous VOC sources


In addition to VOCs biochemically produced by an individuals cells, breath
always contains exogenous VOCs that are present because of external factors
including environmental exposure, lifestyle, diet, the microbiome and
therapeutic interventions. Due to the importance of these factors in many
diseases, exogenous VOCs can provide important clinical information.

Environmental exposure
Compounds present in an individuals environment can alter the presence and
concentration of VOCs observed in breath. VOC biomarkers that typify
exposure have been referred to previously as the exposome [41], and have
been used to measure and assess exposure to certain chemicals in the
environment. For example, breath analysis has been used by the US military to
monitor exposure to volatile compounds in jet fuel - in this case, aromatic and
aliphatic hydrocarbons served as biomarkers of this type of exposure [4143].

Lifestyle, diet and the microbiome


An individuals lifestyle and diet also affects the VOCs present in their breath.
An good example of a lifestyle related exogeneous VOC is acetonitrile, which
appears in the breath of smokers, but is not produced endogenously [44]. Diet
influences breath composition through the production of VOCs from
carbohydrates and fatty acid metabolism. These processes produce acetyl
CoA, which via separate pathways yields the volatile compounds isoprene and
acetone. Isoprene is also an intermediate in cholesterol synthesis, and protein
metabolism generates ammonia and nitric oxide [45].

14
Metabolic processes of the bodys microbiota also generate VOCs. This can be
from bacterial fermentation of unadsorbed complex carbohydrates and
proteins in the gut. This generates gases including methane, hydrogen,
hydrogen sulphide and ammonia [45]. VOCs from the microbiome include
those produced by pathogenic microorganisms, such as C. difficile [46] or M.
tuberculosis [14], making them valuable for the diagnosis of these infections.

Drug metabolites
Breath analysis has been identified as a useful, non-invasive route to
monitoring certain pharmaceuticals (e.g. Propofol) during or after treatment.
Measuring specific drug related volatiles can provide a non-invasive means of
gaining insight into pharmocokinetics and pharmacodynamics. Breath analysis
has also been suggested for the detection of drugs of abuse, which could be
useful for roadside intoxication tests [47].

15
VOCs
as biomarkers
Metabolomics vs. Genomics
Genes act as the blueprint for an individuals biological functions. Despite years
of research focusing on the human genome, and a number of significant
advances, genomics has not yet delivered on its full potential to help us
understand and cure diseases.

In the last decade, hundreds of genome wide association studies (GWAS) have
been performed to identify regions of the genome associated with a range of
human disease. GWAS have failed to account for all the heritability of any
disease [48]. Only 10% of the variants that lead to common diseases are
located within coding DNA sequences. The remainder are located in
non-coding regions of the genome, or intergenic regions.

A recent paper in Science highlighted that only 5% of mutations leading to


cancer were attributable to inherited mutations, whereas a larger fraction was
thought to be attributable to random replication errors occurring during
normal DNA replication [49]. In addition, single mutations rarely result in
disease - usually multiple mutations are required.

Ultimately, biomarkers based on genomic data provide information about


disease predisposition, but can lack detail on current disease activity. In most
circumstances, disease is the result of multiple genetic and epigenetic
attributes interacting with environmental factors, diet and lifestyle choices, and
the microbiome - all of which combine to produce the phenotype, which is
most closely reflected by the metabolome.

Metabolomics for diagnosis


and precision medicine
An organisms metabolism is the set of biochemical processes that sustain life.
Metabolic activity produces a variety of small molecules known as
metabolites. The type and quantity of circulating metabolites is central to the
function of all living systems: this biochemical composition is also known as the
metabolome.

16
Metabolites make excellent disease biomarkers because alterations to an
organisms biological systems caused by disease, genetic mutations, the
microbiome or environmental factors change the profile of metabolites
produced. Hence, metabolomics has the closest biological proximity to the
phenotype of the observed individual, and allows rapid observations of
perturbations in this system.

Metabolites can provide early insight into developing disease states because
they reflect and magnify perturbations at lower genomic, transcriptomic and
proteomic levels several thousand fold [50,51]. As metabolic changes are
present at the earliest stages of disease, metabolites offer a route to earlier
diagnoses that can save lives.

Additionally, metabolites are ideal for differentiating between related disease


states, stratifying patients for therapy response, and monitoring toxicities and
other drug interactions.

Breath Biopsy provides an entirely non-invasive way to capture the volatile


fraction of metabolites in present in exhaled breath, and gain a rapid and
detailed insight into the functioning of the whole body, and the metabolic
changes associated with disease.

Breath Biopsy
A new diagnostic modality

www. owlstonemedical.com/breath-biopsy

17
Whole body blood sampling via breath
Traditional approaches to breath sampling have relied upon collecting a single
exhaled breath, providing a small snapshot of VOCs present in the body. New
advances in breath sampling technology allow breath to be collected over a
longer time period, enabling VOCs originating from the whole body to be
collected.

The ability to capture VOCs originating from throughout the body in a breath
sample is extremely powerful. Firstly, breath enables whole body sampling of
the volatile metabolome. Secondly, this makes VOCs in breath applicable as
biomarkers for a wide range of diseases, not just pulmonary conditions.

Whole body sampling of breath biomarkers for a wide range of


diseases
Each exhaled breath contains a rich spectrum of VOCs derived from the blood,
due to efficient exchange of chemicals between the pulmonary and circulatory
systems. A blood cell takes approximately 1 minute to circulate around the
entire body. Every minute, then, VOCs originating from sources all throughout
the body reach the alveoli in the lungs where they pass efficiently from the
blood into the alveolar air and are then exhaled.

By continuously sampling breath over a few minutes, the profile of VOCs


originating from throughout the body can be accessed in a way that is
completely non-invasive (Figure 9).

Breath provides a window into a much larger volume of blood,


enabling VOC biomarkers from all around the body to be
captured with high sensitivity.

Extended breath collection time for detection of low level VOCs


Some of the VOCs collected during even a 1 minute time period will be present
at extremely low concentrations. By collecting and pre-concentrating all the
VOCs exhaled over a more extended time period (e.g. 10 minutes), even low
concentration VOCs can be captured and quantified with accuracy.

Owlstone Medicals ReCIVA breath sampler collects and pre-concentrates


VOCs in breath onto sorbent tubes over extended sampling durations. This
approach makes Breath Biopsy sensitive to even low level VOC biomarkers,
such as those present at the earliest stages of disease.

18
WHOLE BODY BLOOD SAMPLING VIA BREATH

Collect and pre-concentrate


VOCs in sorbent tubes

1 min 10 mins
Sample Capture even
whole low level VOC
body biomarkers

1 min Rapid transfer of


=
1 complete VOCs from blood
blood to breath
circulation

Figure 9. The volatile metabolome of the entire body can be accessed by continuously sampling breath over a few
minutes. Capturing VOCs from exhaled breath onto ReCIVAs integrated sorbent tubes allows the pre-concentration of
VOC biomarkers originating from the entire body via the circulatory system.

19
Breath Biomarkers
for a wide range
of diseases
and applications

From breath, you can non-invasively access the volatile metabolome of the
entire body, making Breath Biopsy relevant for a wide range of conditions.

Volatile metabolites in breath are promising biomarkers with numerous


applications in clinical diagnostics and precision medicine.

Learn more on page 23.

20
Breath Biopsy vs. Liquid Biopsy
Breath Biopsy has inherent advantages over liquid biopsy techniques, which
look for circulating tumor cells (CTCs) or tumor-derived cell-free DNA/RNA
circulating in the blood. There has been much enthusiasm due to the less
invasive nature of liquid biopsies.

Liquid biopsy shows promise for therapy selection in late stage cancers, but
suffers from three fundamental challenges:
1. Circulating tumour DNA/RNA (ctDNA/RNA) is present only at limited
amounts, especially during the early stages of disease (Figure 10)
2. Only a limited amount of blood can be feasibly sampled and made available
for analysis, which limits the sensitivity of the liquid biopsy approach
3. Detection of a single mutation in ctDNA may not be a reliable indicator of
cancer, hence the liquid biopsy approach may suffer from low specificity for
early diagnosis of cancer.

Bettegowda et al. [52] reported that less than 50% of patients with Stage I
cancer had detectable ctDNA. Notably, the study was conducted using a
relatively large volume of plasma (5 ml). This highlights the limitations of liquid
biopsy, especially for early detection and screening applications.

LIMITATIONS OF LIQUID BIOPSY

Fraction of patients with detectable ctDNA in localized Fraction of patients with detectable ctDNA.
(stages I to III) and metastatic (stage IV) colorectal,
gastroesophageal, pancreatic, and breast cancers. 100
Frequency of cases with detecable ctDNA (%)
Frequency of cases with detecable ctDNA (%)

100
80

80
60

60

40
40

20
20

0 0
COLORECTAL GASTROESOPHAGEAL PANCREATIC BREAST STAGE I STAGE II STAGE III
STAGE IV
Localized Disease Metastatic Disease

Figure 10. Liquid biopsy shows promise for therapy selection in late stage cancer, but low
concentrations of ctDNA, and limitations in the amount of blood that can be feasibly sampled, limit
the sensitivity for early stage cancer detection (Adapted from [52]).

Sensitivity
Metabolite VOC biomarkers are altered at the earliest stages of disease. Breath
Biopsy can capture and detect these early warning VOCs that are present at
extremely low concentrations. This is possible because, unlike blood, breath
can be sampled on demand in effectively unlimited volumes.

By sampling for an extended period of time (~10 mins) using the ReCIVA Breath
Sampler, VOCs are concentrated onto sorbent tubes. Pre-concentration of VOC
biomarkers greatly increases the sensitivity of Breath Biopsy compared to
liquid biopsy techniques, enabling detection of VOC biomarkers present at the
onset of disease.

21
THE ADVANTAGE OF BREATH vs. BLOOD

BLOOD TEST BREATH TEST


Measures a fraction of your blood Measures VOCs from all of your blood
volume needing a needle volume completely non-invasively

1min

INVASIVE/ NON-INVASIVE/
PAINFUL PAIN-FREE

Figure 11. Breath enables higher sensitivity and whole body blood sampling as it captures VOCs
from the entire blood volume completely non-invasively. In contrast, a blood test samples only a
fraction of the blood volume.

Analytical advantages
Breath Biopsy also offers analytical advantages compared to liquid biopsy
techniques. The analysis of VOCs present in breath or other biospecimens can
be accomplished in a matter of minutes, using technology which has potential
to transition to point-of-care (see page 58). Hence, breath analysis is far more
rapid and cost-effective than the lengthy workflows needed for analysis of
nucleic acid by PCR or sequencing, which also typically require significant
sample preparation.

Did you know?


Exhaled breath isnt the only way to sample VOC biomarkers. There are a
variety of clinical sample types widely used to observe volatile metabolites:
Exhaled breath
Breath condensate
Urine
Blood, Serum, Plasma
Feces
Saliva
Mucus
Sputum
Sweat
In vitro cell cultures
Analysis of VOCs emitted from liquid and solid biospecimens is achieved by
headspace sampling, and metabolites in biofluids can also be analyzed
directly.

Learn more in the section on Beyond Breath on page 64.

22
Applications of
VOC biomarkers
Exhaled breath contains a vast untapped resource of chemical information in
the form of over 1,000 volatile organic compounds. As products of underlying
metabolic activity, VOCs directly reflect the state of cells and tissues, and
represent a goldmine of valuable non-invasive biomarkers.

The clinical applications for non-invasive, breath-based VOC biomarkers are


wide ranging, with numerous applications in precision medicine and
diagnostics.

Breath Biopsy Applications

Clinical diagnostics and prognostics


.(5 !0!0%+** /.!!*%*#
%/!/!)+*%0+.%*#
.!%/%+*)! %%*!
- Companion diagnostics
- Complementary diagnostics
- Patient stratification
- Therapy response biomarkers
- Toxicity prediction

23
Wide disease relevance
VOCs originating from all parts of the body are captured in breath, making
Breath Biopsy applicable to a wide range of diseases including:
t Cancer
t Inflammatory disease
t Infectious disease
t Metabolic disease
t Cardiovascular disease

Research case studies


Learn about how our academic and clinical partners
are applying VOC biomarkers in their research
across a range of disease areas.

www. owlstonemedical.com/case-studies

BREATH VOCs REPORTED FOR A WIDE RANGE OF DISEASES

Head and neck cancer Schizophrenia

Asthma and COPD Tuberculosis

Pulmonary embolism Mesothelioma

Breast cancer Lung cancer

Acute respiratory Cystic fibrosis


distress syndrome
Heart disease
Artherosclerosis
Gastric cancer
Liver cancer
H. pylori infection
Liver cirrhosis
Renal failure
Alcohol hepatitis
Carbohydrate
Non-alcohol fatty malabsorption
liver disease
Colorectal cancer
Diabetes
Rheumatoid arthritis
Inflammatory
bowel disease

Figure 12. VOC biomarkers in breath have been reported for a large number of diseases in the
literature [53,54], highlighting the wide disease relevance of Breath Biopsy.

24
Novel Biomarkers
Until relatively recently, the identification of robust VOC biomarkers in breath
has been hampered by the lack of standardized sampling and analytical
methods. Progress is now being accelerated by the advent of new technologies
for the reliable collection of patient breath samples (ReCIVA) and advances in
VOC analysis technology (FAIMS); when combined, these new technologies
have the potential to bring Breath Biopsy to use at point-of-care.
Breath and the metabolome are relatively under-explored in a clinical context,
so VOC biomarkers identified in breath are likely to be completely novel, which
has advantages for their utilization as companion or complementary
diagnostics.
The advances in breath sampling technology now enable Breath Biopsy to be
embedded into early stage clinical trials, allowing pharmaceutical companies
to perform biomarker discovery early on in the drug development process. This
enables the development of companion diagnostics which can be critical to
maximize the chance of achieving regulatory approval.

Precision Medicine applications


As the products of metabolism, VOCs are closely linked to disease activity, and
are therefore ideal for capturing the biological variability of disease. Hence,
VOC biomarkers are particularly useful for precision medicine applications that
depend on a detailed insight into the biological variability that underlies
disease processes.
The ability to measure disease activity by detecting metabolites in breath has
applications in treatment stratification, to match patients to the right treatment
regime. For example, VOCs in breath have already shown promising results in
stratifying asthma patients responsive to a particular therapy (see page 30),
and the NHS and Innovate UK have supported the early stage clinical trial
Stratification of Asthma Treatment by Breath Analysis (STRATA) to further
https://www.owlstonemedical.com/clinical-pipeline/strata/
https://www.owlstonemedical.com/clinical-pipeline/strata/

explore this.
Similarly, VOC biomarkers offer a non-invasive means to accurately stratify
inflammatory bowel disease patients to assist with therapy selection (see also
page 40).
VOC biomarkers that predict therapy response can improve treatment efficacy
within the intended population and also lead to development of an easily
deployable breath-based test for identification of responders.

Breath Biopsy Applications


in Precision Medicine
Download the Drug Discovery World article

www. owlstonemedical.com/ddw
25
Metabolites relate to
actual disease
activity
DNA provides the blueprint for disease predisposition, but metabolites give
information on the current state of tissues and cells, and the health of the
individual.
Altered metabolism occurs at the earliest stages of disease, and magnifies
changes at the genomic, transcriptomic and proteomic levels several thousand
fold.

Learn more on page 16.


26
Clinical Trials patient matching
VOC biomarkers can be used to develop companion diagnostics to screen and
stratify patients for inclusion in clinical trials. By selecting the patients most
likely to respond to the particular therapy, this approach enables smaller and
more cost effective clinical trials.
More importantly, stratifying patients for inclusion in clinical trials will lead
ultimately to more successful clinical trials, and maximize the chance of gaining
regulatory approval for therapies that would otherwise fail due to health
economic pressures.
There is evidence that VOCs can be used to classify asthma patients by
inflammatory subtype, highlighting the potential use of Breath Biopsy as a
companion diagnostic for therapy stratification (see also page 30).

Drug metabolites and disease monitoring


Breath Biopsy has potential applications in disease and treatment monitoring.
Comparing pre- and post-treatment breath tests could provide confirmation of
therapy efficacy by detecting changes in disease-related biochemical activity.

Tracking changes in the VOC profile in breath over time could provide an early
warning of changes in the underlying disease state.

Administration of drugs also results in additional metabolites which can be


accessed via a Breath Biopsy and used to evaluate the drugs
pharmacokinetics as well as assess drug compliance. There is some evidence
that metabolite changes may be associated with drug-induced toxicity [55],
hence VOC biomarkers could be explored as potential predictors of toxicity.

Early detection and screening of cancer


Breath is available in an effectively unlimited supply, which provides analytical
advantages compared to liquid biopsy techniques (see also page 21). Breath
Biopsy shows great promise for applications including early detection and
screening, where the goal is to detect biomarkers present at the earliest stage
of disease.

The ability to concentrate exhaled chemicals by sampling over a few minutes,


enables the detection of biomarkers at very low levels in a way that is not
possible with other sampling modes such as blood tests. This has the potential
to dramatically improve test sensitivity across a range of diseases.

The completely non-invasive nature of breath sampling will allow clinicians to


greatly increase compliance rates, which can be problematic in the case of
diagnostic procedures which are less pleasant, more invasive and carry risks to
patients.

Breath Biomarker Services


Let our experts add Breath Biopsy to your clinical trials

www. owlstonemedical.com/services
27
Asthma and COPD
As many as 334 million people globally have asthma and up to 14% of the
worlds children experience asthma symptoms [56]. Asthma management is
focused on achieving control of symptoms to minimize the risk of future
exacerbations.

Many patients do not respond sufficiently to treatment, however, so their


symptoms are not kept adequately under control. In the UK, 4.4% of patients
fail to respond to standard therapies even at high doses, and account for more
than half of asthma costs to the NHS.

Asthma symptoms are caused by chronic inflammation in the small airways.


Many different underlying mechanisms are responsible for this inflammation,
with different treatments required for different inflammation types. At present
there is no single diagnostic test for asthma in part because of its
heterogeneous underlying pathophysiology.

The difference in cell activity related to different inflammatory subtypes of


asthma is reflected by the metabolites these cells produce, and consequently
by the VOCs in exhaled breath. This is why asthma has become one of the most
studied diseases with regard to breath VOCs, and improvements in technology,
like ReCIVA, are accelerating progress in the field.

28
VOC Biomarkers for improved asthma diagnostics
One breath based biomarker, FENO (fractional exhaled nitric oxide), is already
being used to support asthma diagnosis. FENO is useful because it reflects a
protective biochemical pathway of the lungs, but it does lack specificity, as it is
a single biomarker affected by many processes other than asthma.

A study of VOCs in asthma found that they outperform FENO and lung function
tests in discriminating asthmatics and healthy controls. Furthermore,
combining VOCs with FENO gave an even higher diagnostic performance [57].

It has also been shown that VOCs can be used to discriminate with a high
degree of accuracy between asthmatic children and those with transient
wheezing [58]. In both this and two other studies [59,60], increases in
methylated alkanes were associated with asthma prevalence: these VOCs are
potential markers of oxidative stress which have been linked to airway
inflammation.

There are also promising initial results from Owlstone Medicals NHS funded
LuCID clinical trial [34], which aims to recruit 3,000 patients and uses the
ReCIVA Breath Sampler as well as FAIMS technology (Lonestar VOC Analyzer)
to analyze patient breath samples.

Data collected so far indicates that GC-MS combined with FAIMS can
discriminate between asthmatic and non-asthmatic patients (AUC 0.92, Figure
13) in a cohort of patients under clinical suspicion of lung cancer (a
heterogeneous population with a wide variety of medical conditions). Such an
approach allows identification of disease specific volatiles greatly facilitating
translation to clinical practice.

Breath VOCs in Asthma


Breath Biopsy applications in Asthma
diagnosis and treatment stratification

Download the whitepaper:

www. owlstonemedical.com/asthma-vocs
29
ASTHMATIC vs. NON-ASTHMATIC PATIENTS

1.0

0.8
True Positive Rate

0.6

Area Under Curve = 0.92


0.4

0.2

0.0

0.0 0.2 0.4 0.6 0.8 1.0


False Positive Rate
Figure 13. Discriminating patients with asthma in the LuCID clinical trial population using VOCs.

The existing evidence suggests that VOCs are strongly affected by the airway
inflammation that characterizes obstructive pulmonary diseases such as
asthma, and that they could be useful tools in non-invasive asthma diagnosis.

The non-invasive nature of breath VOC biomarkers is particularly valuable in


children, as other known biomarkers such as FENO or blood IgE levels require
exhalation at a continuous flow or invasive sampling, making them less suitable
for young children.

VOCs for personalised medicine and treatment stratification


Whilst the diagnostic power of breath VOCs could provide a valuable new tool
for clinicians, the most impactful application of an asthma Breath Biopsy is
likely to be in the phenotyping of the disease by characterization of underlying
inflammation.

The lack of stratifying diagnostics means that current guidelines advocate a


trial and error approach, which results in increased healthcare costs,
prolonged periods of poor disease control, and an increased risk of
exacerbations.

There is evidence that VOCs can be used to classify patients by asthma


phenotype with high accuracy e.g. [60]. VOCs will therefore allow different
phenotypes of asthma defined by different inflammatory responses to be
distinguished. This clearly points towards the use of Breath Biopsy as a
companion diagnostic for therapy stratification.

30
There is already evidence that shows VOCs can be used to predict steroid
responsiveness in mild to moderate asthma sufferers [61]. In this study, VOCs
were considerably more accurate at predicting responsiveness to steroid
treatment (AUC 0.88 0.16) than either FENO or measurements of eosinophil
cells from coughed up sputum samples (Figure 14).

PREDICTION OF STEROID RESPONSIVENESS

1.0
Sensitivity

0.5

Legend; valeus are (AUC 95% CI ; p-value)

VOCs (0.88 0.16 ; 0.008)

Eosinophil % (0.61 0.29 ; 0.441)

FENO (0.55 0.28 ; 0.751)

PD15 (0.60 0.29 ; 0.497)

0.0
0.0 0.5 1.0

1-specificity
Figure 14. Predicting steroid responsiveness in steroid naive patients [61].

A reliable tool to personalize treatment to a specific asthma phenotype would


be an important development. The last decade has seen the approval of
biological drugs targeted at specific pathways relevant to inflammatory
subtypes, yet their approval for clinical use has been delayed due to the high
cost of the treatment, combined with the difficulty of identifying patients with
the correct asthma phenotype who would benefit from the drug.

Using biomarkers to characterize asthma types has so far not found


widespread adoption due to the invasive nature of existing cell harvesting
techniques such as bronchoalveolar lavage, and the limited applicability of
FENO testing at identifying responders to new targeted biologic treatments
aimed at patients who remain uncontrolled under traditional asthma therapies.

Blood based biomarker IgE has shown reasonable positive predictive potential
for treatment response to the targeted anti-IgE treatment
XOLAIR/Omalizumab [62]. Therapy failure still occurs in a high percentage of
adult and pediatric patients [63], however, and a significant fraction of patients
lose responsiveness to XOLAIR over time.

Additionally, a large portion of patients that would benefit from XOLAIR are
currently not eligible based on the IgE entry criteria. Similar issues have arisen
with the use of periostin and blood eosinophils to identify potential responders
to other targeted asthma therapies.

31
Promising initial results from the U-BIOPRED (Unbiased BIOmarkers in
PREDiction of respiratory disease outcomes) consortium project show that
VOCs measured using a FAIMS device could be used to stratify asthmatic
patients into treatment sub-groups. For example, VOCs discriminated between
anti-IgE-treated XOLAIR and non-treated severe asthma patients with 83%
accuracy [64,65].

In the STRATA (Stratification of Asthma Treatment by Breath Analysis)


programme, Owlstone Medical seeks to replace the trial and error approach for
asthma treatment selection, and use our Breath Biopsy platform to stratify
asthma patients and match them to the correct existing treatments first time.
As a result, we will bring lasting benefit to patients, minimize costs for the NHS
and make a positive impact on the annual toll of 54,000 asthma-related
emergency hospital admissions and 1,167 asthma-related deaths.

STRATA Clinical Trial


Selecting the right treatment for severe asthma patients

www. owlstonemedical.com/strata

Oncology
Cancer has many complex and interacting causes, including genetic
predisposition, mutations in the genetic code, lifestyle, diet and many other
environmental factors. As metabolites are the conclusion of gene, transcription
and protein expression, they reflect gene-protein-environment interactions.
This unique position means they can provide an accurate representation of the
metabolic phenotypes of different cancers [66].

Tumor metabolomics
Cancer cells display altered metabolism from day one. The mechanisms by
which oncogenes reshape a range of metabolic pathways to meet the
nutritional demands of cancer cells is beginning to be elucidated [67].

One of the best described metabolic changes is the Warburg effect: cancer
cells have glycolytic rates up to 200 times above normal cells, even when
sufficient oxygen is present. This metabolic change is key to cancer survival as
it provides the building blocks a growing tumor requires. The Warburg effect is
a major contributor to a cancers volatile signature [68] and also affects the
profile of VOCs in breath [69].

Cancer specific metabolite VOCs, observable by Breath Biopsy, are likely to be


related to changes in cancer cell metabolic programming. In addition, changes
to VOC concentrations in the breath of cancer patients are likely to result from
other processes, such as local inflammation and increased oxidative stress (see
also the section Biochemical origins of VOCs on page 13), and knock-on
effects of the disease on the body.

32
Since altered metabolites are present in the first stages of disease, this enables
earlier detection and diagnosis. As disease prognosis is greatly dependent on
cancer stage and tumor type, these factors make metabolites ideal biomarkers
for cancer.

The search for VOC signatures associated with cancer has produced studies
looking at breath VOCs related to many cancers, including lung, breast,
colorectal, gastric, ovarian, liver, head and neck and malignant mesothelioma
[70].

ReCIVA Breath Sampler


For reliable, standardized breath collection

www. owlstonemedical.com/reciva

Lung cancer
Lung cancer is the most common cancer worldwide and also one of the
deadliest, with around 1.8m cases and 1.6m deaths each year. It has one of the
lowest 5-year survival rates of all cancers, at just 17.7% in the US.

However, early diagnosis can greatly improve the clinical outlook. Patients
diagnosed at Stage 1 in the US have a 55% 5-year survival rate, compared to
just 4.3% at stage IV [71].

Current diagnostic procedures, which include a chest x-ray or CT scan, suffer


high false positive rates which lead to unnecessary invasive procedures and
risks. As such, the potential benefits of a low-cost, non-invasive means of
providing early diagnoses are very clear.

Exhaled breath is an excellent source of VOC biomarkers originating from


airway and lung tissues. This makes breath analysis an ideal approach to the
detection of metabolites relating to cancer cells in the lungs.

33
A number of studies have already found evidence that lung cancers can be
detected based on VOCs present in breath [72], but a lack of sampling
standardization and underpowered clinical trials has so far prevented the
introduction of a breath test for lung cancer.

REPORTED VOC LUNG CANCER MARKERS

Volatile Candidate CAS number Approximate Concentration


Lung Cancer Markers concentration in Lung detected
Cancer Patient Breat using FAIMS

Butane-2,3-dione 431-03-8 30 - 50 ppb < 2 ppb

Isoprene 78-7-95 80 - 140 ppb < 0.4 ppb

Limonene 138-86-3 30 - 50 ppb < 2 ppb

Methanol 67-56-1 100 - 150 ppb < 2 ppb

Acetonitrile 75-05-8 2 - 60 ppb < 2 ppb

Acetone 67-64-1 450 - 650 ppb < 1.5 ppb

Acetamide 60-35-5 30 - 50 ppb < 3 ppb

Formaldehyde 50-00-0 30 - 50 ppb < 2 ppb

2-Butanone 78-93-3 30 - 50 ppb < 1 ppb

Benzaldehyde 100-52-7 30 - 50 ppb < 2 ppb

Acetophenone 98-86-2 30 - 50 ppb < 2 ppb

Methyl-propyl sulfide 3877-15-4 30 - 50 ppb < 1 ppb

Table 2. Lung cancer VOC biomarkers reported in the literature [44,7380], their approximate
concentrations in lung cancer patient breath, and the concentrations at which they have been
detected using the Lonestar VOC Analyzer (FAIMS).

A screening approach using the improved Breath Biopsy sampling and


analysis technology offers the promise of non-invasive early diagnosis of lung
cancers. At present, the scientific consensus is that no single marker can be
used to accurately detect lung cancer. Instead, combinations of lung
cancer-specific VOC biomarkers and changes in the concentrations of
compounds normally found in healthy breath make accurate detection of
malignant cells possible.

Owlstone Medicals microchip FAIMS technology is sensitive enough to detect


VOC biomarkers identified in previous studies [44,7380], at concentrations
relevant for human breath (Table 2).

To make lung cancer breath tests a reality, Owlstone Medicals LuCID (Lung
Cancer Indicator Detection) project is currently underway. LuCID includes
highly powered clinical trials for early detection of lung cancer, and is funded
by an SBRI Healthcare development contract.

LuCID Lung Cancer Clinical Trial


Worlds largest breath based clinical trial
for early cancer detection

www. owlstonemedical.com/lucid

34
Colorectal cancer
Colorectal cancer (also referred to as bowel cancer) is the fourth most common
form of cancer in the US, with around 135,000 new cases and 50,000 deaths
each year. It is the second most common cause of death from cancer in the US,
and has a mortality rate of around 35% within five years of diagnosis.

As with many other types of cancer, the stage at which colorectal cancer is
diagnosed has a huge impact on the patient's prognosis. Patients identified at
Stage I have around a 90% five-year relative survival rate, which drops to 13.5%
if the cancer has metastasized (stage IV) [81].

To increase the the number of early diagnoses, the way in which patients are
screened for colorectal cancer needs to be improved. Existing non-invasive
colorectal cancer screening methods such as gFOB (guaiac based fecal occult
blood) and the now more common FIT (Fecal Immunochemical Testing) show
relatively low sensitivities (FIT sensitivity = 66 - 88%, specificity = 87 - 96%). FIT
also lacks sensitivity to advanced adenoma (27 - 41%, specificity = 91 - 97%).

A major drawback of current screening techniques is that patients have to


provide stool samples. Globally, the uptake of colorectal cancer screening
utilising fecal samples varies from 60% in the UK to 8% in the US, and is as low
as 1 and 2% in China and India respectively (Figure 15).

35
PERCENTAGE OF POPULATION SCREENED
FOR COLORECTAL CANCER USING FECAL TESTING

USA
8% of 89 million population screened with FIT
Age range* 50-74 years old

UK 8%
SCREENED

60% of 10 million population screened with FIT 89M


Age range* 60-75 years old

FRANCE
42% of 19 million population screened with FIT
Age range* 50-74 years old

GERMANY
18% of 28 million population screened with FIT
Age range* >56 years old

ITALY 10M
60%
SCREENED
18%
SCREENED

28M
27% of 16 million population screened with FIT
Age range* 50-69 years old 42%
SCREENED

SPAIN 19M

27%
SCREENED

14% of 11 million population screened with FIT 14%


50-69 years old
SCREENED
Age range*
16M
11M

JAPAN
17% of 75 million population screened with FIT
Age range* >40 years old

CHINA 1%
SCREENED

1% of 564 million population screened with FIT


Age range* 40-70 years old 17%
564M SCREENED

INDIA 2%
SCREENED
75M

2% of 175 million population screened with FIT


Age range* 50-69 years old
175M
*Defined by the age range obtained from identified source for
national screening program or recommendations.

Figure 15. The percentage of the population screened for


colorectal cancer using FIT (Fecal Immunochemical Testing)
varies globally. Screening based on fecal testing typically
suffers from poor compliance due to poor patient
acceptability.

36
There is clear need for a new, non-invasive screening system with the potential
for high patient compliance and low cost. Studies have previously identified at
least 15 discriminating VOCs that can be used to identify between patients with
and without colorectal cancer (e.g. [82]). It has also been shown that breath
VOCs can be used to discriminate patients with colorectal cancer from those
with other cancers, specifically breast, lung and prostate [83].

There is also evidence that VOCs from urine can be used to discriminate
between patients with and without colorectal cancer [84]. Arasaradnam et al.
found that Owlstone Medicals FAIMS technology can be used to differentiate
between patients with colorectal cancer and healthy controls, with a sensitivity
of 88%, by rapid, non-invasive analysis of volatile organic compounds (VOCs)
in urine (Figure 16 and Figure 17).

The authors recruited 133 volunteers, 83 of whom had confirmed colorectal


cancer. VOC analysis was performed using an Owlstone Medical Lonestar VOC
Analyzer fitted with an ATLAS headspace sampling system.

These results show that detection of colorectal cancers VOC profile with
FAIMS from patient urine samples can prove valuable as part of a non-invasive
screening program for colorectal cancer.

FAIMS SPECTRUM OF VOCs IN URINE


100
0.001
0.01
0.1
1
80
Ion Current
(arbitary units)
DISPERSION FIELD (%)

60

40

20

0
-6 -4 -2 0 2 4
COMPENSATION VOLTAGE (V)

Figure 16. Lonestar VOC Analyzer FAIMS spectrum from urine sample of patient with colorectal
cancer.

37
FAIMS DISCRIMINATES BETWEEN PATIENTS
WITH COLORECTAL CANCER AND CONTROLS
0.006

CRC
CONTROL
0.004

DISCRIMINANT ANALYSIS

0.002

-0.000

-0.002

-0.004

-0.006
1 2
GROUP

Figure 17. To determine whether the patients with colorectal cancer could be distinguished from
the other healthy volunteers using the Lonestar VOC Analyzer (FAIMS), Arasaradnam et al. used
Fisher Discriminant Analysis on the FAIMS data collected from the urine samples of patients with
colorectal cancer and controls. They found that it was possible to use the entire VOC profile
measured with FAIMS to detect colorectal cancer cases with 88% sensitivity.

To investigate the performance of both breath and urine VOCs in a larger at risk
population of using FAIMS to screen for colorectal cancer, Owlstone Medical
have launched a 1,400 patient clinical trial in collaboration with Warwick
University and the University Hospital Coventry and Warwickshire NHS Trust.
Known as InTERCEPT, the trial aims to evaluate the accuracy of both breath
and urine VOC biomarkers in the detection of colorectal cancer at an early
stage, when patient survival rates are at their highest.

InTERCEPT
The worlds largest breath-based study for
early detection of colorectal cancer

www. owlstonemedical.com/intercept

38
Inflammatory diseases
The term inflammatory disease covers an array of disorders and conditions
that are characterized by the bodys complex inflammatory reaction to harmful
stimuli, including irritants, pathogens and damaged cells.

Acute inflammation is the bodys initial protective response, and involves


movement of leukocytes to the affected tissues. Prolonged, chronic
inflammation, leads to a progressive shift in the type of cells present, resulting
in simultaneous destruction and healing of tissues.

Inflammatory abnormalities underlie a vast variety of human diseases,


including asthma, many autoimmune diseases and are also implicated in the
development of cancer and heart conditions.

Inflammatory bowel disease


Inflammatory bowel disease (IBD) is a group of autoimmune diseases primarily
affecting the colon and small intestine. The principal types of IBD are Crohns
disease (CD) and ulcerative colitis (UC). IBD is common, and affects
approximately 1 - 1.3 million people in the US [85].

The cause of IBD is not entirely understood, but it is widely believed to be the
result of complex interactions between an individuals genetic susceptibility,
environmental triggers (e.g. diet, lifestyle, etc.) and the influence of an
individuals gastrointestinal bacterial colonies.

At present the diagnosis of IBD and the differentiation of disease types takes a
multi-modal approach that includes clinical, histological, serological,
radiological and endoscopic observations [86].

39
Accurate disease stratification is vital for determining the best treatment
strategy, but at present 10 - 15% of patients elude correct categorization.
Accurate phenotyping often requires endoscopy of the intestine, which high
cost, uncomfortable, has high non-compliance rates and can result in
complications such as perforation of the bowel [87,88].

Metabolic VOC biomarkers offer a non-invasive route to the accurate


stratification of IBD patients. They also offer a potential means of determining
which patients will develop more severe disease phenotypes, and who will
benefit most from expensive biological and/or immunomodulating agents
[89,90].

Breath Biopsy for IBD diagnosis and stratification


To determine whether breath VOCs offer an effective route to diagnosis and
stratification of patients with IBD, researchers have used Owlstone Medicals
FAIMS technology to see if patients breath VOC profile can be used to
distinguish IBD patients from healthy controls, and to also distinguish those
with Crohns disease (CD) and ulcerative colitis (UC) [16].

A total of 76 subjects were recruited for the study, 54 of which had


histologically confirmed IBD (29 UC and 25 CD) as well as 22 healthy controls.
The predictive performance of the data from FAIMS analysis of exhaled breath
samples was studied using a pipeline consisting of wavelet transformation,
feature selection and a sparse logistic regression classifier.

This was used to classify samples and calculate sensitivities and specificities as
part of a 10-fold cross-validation. This box plot (Figure 18) shows the predictive
power of IBD (UC and CD) vs. controls, UC vs. controls, CD vs. controls and
finally UC vs. CD.

FAIMS PREDICTIVE POWER


1.0
PREDICTIVE PROBABILITY

0.8
0.6
0.4
0.2
0.0

UC & CD V UC V CD V UC CD

Figure 18. Predictions of classifier to different combinations of diseases and controls (UC
ulcerative colitis, CD Crohns disease, V volunteer (healthy control)). The study used the
Lonestar VOC Analyzer (FAIMS) to analyze VOC biomarkers in breath [16].

40
The analysis showed that patients with IBD could be distinguished from control
patients using FAIMS analysis of VOCs in breath samples with a sensitivity of
74% (95% confidence intervals (CI): 0.650.82) and specificity of 75% (95% CI:
0.530.90), p-value 6.2 107. The AUC (area under curve) was 0.82 (95% CI:
0.74-0.89) (Figure 19).

FAIMS could distinguish those with UC from those with CD with a sensitivity of
67% (95% CI:0.540.79) and specificity of 67% (95% CI: 0.540.79), p-value
9.23 104. The AUC was 0.70 (95% CI: 0.600.80).

IBD (UC AND CD) vs HEALTHY CONTROLS

(AUC=0.82) (95% CI: 0.74-0.89)


1.0
0.8
SENSITIVITY

0.6
0.4
0.2
0.0

1.0 0.8 0.6 0.4 0.2 0.0

SPECIFICITY

Figure 19. Receiver operator curve (ROC) plot of IBD (UC and CD) vs. healthy controls.The study
used the Lonestar VOC Analyzer (FAIMS) to analyze VOC biomarkers in breath [16].

The study confirms the utility of FAIMS exhaled VOC analysis to distinguish IBD
from healthy controls, and UC from CD. It conforms to other studies using
different technology, whilst affirming exhaled VOCs as biomarkers for
diagnosing IBD.

Lonestar VOC Analyzer


Based on proven FAIMS technology

www. owlstonemedical.com/lonestar

41
Liver disease
The link between liver diseases and characteristic breath aromas has been
known to clinicians for thousands of years. Hippocrates (460370 BC)
famously described fetor hepaticus in his treatise on breath aroma and disease
[1].

Liver diseases including non-alcoholic fatty liver disease and viral hepatitis can
progress to cirrhosis and eventually hepatocellular carcinoma. Early diagnosis
and treatment of liver diseases is important because progression is common
and has a significant risk of mortality.

As a major organ involved in dealing with waste metabolites, the liver is a


significant source of metabolic VOCs that can be excreted in breath. Breath
Biopsy offers a non-invasive way to measure liver disease activity and gather
vital metabolic information for earlier diagnoses.

Hepatic encephalopathy
Owlstone Medicals FAIMS technology has already been utilised to analyse the
breath of patients with hepatic encephalopathy (HE). HE is a neuropsychiatric
condition which occurs when the liver cannot adequately remove toxins from
the blood. It can occur suddenly in people with acute liver failure but is more
often seen in those with chronic liver disease.

The prevalence of minimal HE (the mildest detectable class of HE under the


West Haven criteria) is reported in 30% 84% of patients with liver cirrhosis.
This has profound effects on daily functioning with nearly 50% of minimal HE
patients potentially unfit to maintain employment. The condition can be
treated effectively using Rifaximin, and earlier detection can prevent liver
disease progression and hospitalization.

42
In the pilot study [15], the chemical fingerprint of breath samples from patients
with HE and controls was analyzed using a Lonestar VOC Analyzer fitted with
an ultra violet ionisation source. 42 patients were recruited; 22 patients with HE
(13 covert and 9 with overt HE) and 20 healthy controls.

Extracted data was analysed using a pipeline based on a 2D wavelet transform


with thresholds to remove background noise. This was followed by feature
selection to identify key variables (using a Wilcoxon rank-sum test applied
separately to each feature), with the resultant feature set used to separately
train four classifiers (sparse logistic regression, Random Forest, Support Vector
Machine and Gaussian Process).

HE vs. CONTROLS

a glmnet (AUC=0.84)(95% CI: 0.75-0.93)


1.0
0.8
SENSITIVITY

0.6
0.4
0.2
0.0

1.0 0.8 0.6 0.4 0.2 0.0

SPECIFICITY

Figure 20. Receiver operator curve (ROC) plot for distinguishing HE from controls. The study used
the Lonestar VOC Analyzer (FAIMS) to analyze VOC biomarkers in breath [15].

FAIMS analysis of exhaled VOCs was able to classify HE patients from controls
with a sensitivity and specificity of 88% (0.730.95) and 68% (0.510.81)
respectively with an AUC of 0.84 (0.750.93).

This pilot study provides evidence that breath VOC biomarker analysis has
potential as a diagnostic aid in distinguishing HE of all grades from healthy
control subjects, and further suggests potential for VOC biomarkers as an aid
to distinguish between the covert and overt forms of HE.

ReCIVA Breath Sampler


For reliable, standardized breath collection

www. owlstonemedical.com/reciva

43
Infectious disease
Infectious disease mortality can be lowered by correctly targeting
antimicrobial drugs early on in the progress of an infection [91]. At present,
most treatment is untargeted due to a lack of adequate diagnostics.

Targeting requires knowledge of the causative bacterial, viral, or fungal


pathogen, but culturing has limited sensitivity and can take many days [92].
Other techniques like gram-staining and direct cellular examination are fast but
lack sensitivity and specificity, and dont provide information about exact strain
or antimicrobial targeting.

Improving diagnostic techniques to help better target infectious diseases has


the potential to save lives and help tackle antimicrobial resistance. One way of
doing this is to identify bacteria based on markers linked to their metabolisms,
which are distinct to those of their hosts.

Bacteria have been shown to produce VOCs not commonly generated by


human metabolic processes that could be used to identify bacterial strain,
improving the potential for drug targeting [9396]. These VOCs are present in
exhaled breath, making Breath Biopsy a useful, non-invasive approach to
infectious disease diagnosis.

Tuberculosis
There are around 8.6 million new cases, and 1.3 million deaths globally from
tuberculosis (TB) every year, a disease caused by the bacterium
Mycobacterium tuberculosis. Diagnosis is still mostly performed by trained
technicians examining sputum samples under a microscope [97].

Treatment and prevention of TB in low-income countries is particularly


hampered by the lack of cheap, simple and accurate diagnostic tests. The
sputum test only has a sensitivity of 62%, and isnt able to distinguish between
Mycobacterium species [98]. In particular, patients who are suspected of
having TB but who test negative using the sputum ZN-smear pose a diagnostic
problem.

Metabolites have been implicated as biomarkers for TB [99], and breath


analysis using technologies like FAIMS have great potential to provide a simple,
point of care diagnostic service that could revolutionise TB detection.

A recently published pilot study by Sahota et al. [14] found that Owlstone
Medical's Lonestar VOC Analyzer (FAIMS) was able to distinguish between
healthy controls and patients with either pulmonary or extra-pulmonary
tuberculosis (the latter type of the disease traditionally more difficult to
diagnose). They analyzed the FAIMS data from patient breath samples using a
wavelet feature extraction process followed by a random forest classifier, to
identify features in FAIMS spectra. This allowed discrimination of cases from
controls with a sensitivity of 81% and a specificity of 79%.

44
CLASSIFICATION OF CONTROL AND TB GROUPS

0.8

CONTROL
0.7
TB

0.6
PREDICTIVE PROBABILITY

0.5

0.4

0.6

0.2

0.1
1 2

GROUPS

Figure 21. Classification probabilities for the control and TB groups in pilot study using a Lonestar
VOC Analyzer by Sahota et al. Applying a Wilcoxon rank-sum test to the sets of classification
probabilities from the two groups, we get a p-value of 2.89E6, showing that there is a highly statis-
tically significant difference between the control and TB groups. The study used the Lonestar VOC
Analyzer (FAIMS) to analyze VOC biomarkers in breath [14].

TB vs. CONTROLS
1.0
0.8
SENSITIVITY

0.6
0.4
0.2
0.0

1.0 0.8 0.6 0.4 0.2 0.0

SPECIFICITY

Figure 22. Receiver Operator Curve (ROC) plot from Sahota et al. (AUC = 0.92; 95% CI: 084, 1) for
breath analysis of patients with TB. The study used the Lonestar VOC Analyzer (FAIMS) to analyze
VOC biomarkers in breath [14].

45
Detection of TB in patients with negative ZN-smear tests
A team from the Pulmonology Department at the Academic Medical Centre,
University of Amsterdam were interested to test whether breath biomarkers
could detect TB in patients suspected to have TB, but who displayed negative
ZN-smear tests.

Owlstone Medical's Lonestar VOC Analyzer (FAIMS) was included in a


proof-of-concept study. They showed that tuberculosis can be diagnosed
using this system via chemical markers in exhaled breath. Importantly, they also
found that TB could be detected in the patients with negative ZN-smear
results.

The results of these pioneering studies indicate that FAIMS, in particular, has
potential as a technology that can be applied to breath testing for infectious
diseases including TB.

Lonestar VOC Analyzer


Based on proven FAIMS technology

www. owlstonemedical.com/lonestar

Clostridium difficile
C. difficile is a bacterium that can infect the bowel, causing diarrhea. It
commonly affects the elderly and patients who have recently received
antibiotics. In recent years, the infection has become increasingly prevalent,
and more severe forms of the disease have emerged. More serious symptoms
include severe pseudomembranous colitis and toxic megacolon. Large hospital
outbreaks have required ward closures and extensive infection control
measures.

While the application of breath VOC biomarkers has not yet been investigated
for C. difficile diagnosis, there is evidence for the utility of VOCs emitted from
stool samples. VOC biomarkers emitted from stool samples could accurately
distinguish C. difficile-positive stool samples from healthy samples.

Both healthy and C. difficile-positive stool samples were analyzed using a


Lonestar VOC Analyzer (FAIMS) fitted with an ATLAS headspace sampler
(Figure 23, [46]. A total of 213 samples were analyzed, of which 71 were
confirmed as C. difficile-positive by microbiological analysis.

46
FAIMS SPECTRUM OF A C. DIFFICILE-POSITIVE
STOOL SAMPLE

100
0.001
0.01
0.1
1
80
Ion Current
(arbitary units)
DISPERSION FIELD (%)

60

40

20

0
-6 -4 -2 0 2 4
COMPENSATION VOLTAGE (V)

Figure 23. FAIMS spectrum of a C. difficile-positive stool sample. Stool samples were analyzed
using a Lonestar VOC Analyzer (FAIMS) fitted with an ATLAS headspace sampler [46].

Training and test samples (n=135) were used to identify which characteristics
discriminate between positive and negative samples, and to build machine
learning algorithms interpreting these characteristics.

The best performing algorithm was then validated on new, blinded validation
samples (n=78). The predicted probability of C. difficile (as calculated by the
algorithm) was compared with the microbiological test results (direct toxin test
and culture).

Using a Random Forest classification algorithm, FAIMS had a high


discriminatory ability on the training and test samples (AUC of 0.91 (95%
confidence interval (CI): 0.860.97)). When applied to the blinded validation
samples, the AUC was 0.86 (0.750.97). For samples analyzed 7 days of
collection (n=76), diagnostic accuracy was even higher (AUC of 0.93
(0.851.00), Figure 24).

A cutoff value of 0.32 for predicted probability corresponded with a sensitivity


of 92.3% (95% CI: 77.498.6%) and specificity of 86.0% (78.389.3%). For even
fresher samples, discriminatory ability further increased.

47
C. DIFFICILE POSITIVE vs. NEGATIVE SAMPLES

ROC curve

1.0
0.8
SENSITIVITY

0.6
0.4
0.2
0.0

0.0 0.2 0.4 0.6 0.8 1.0


1 - SPECIFICITY

Figure 24. The Receiver Operating Characteristic (ROC) curve for the (toxin and culture) negative
and (toxin and culture) positive samples. Samples analyzed within 7 days are included. Stool
samples were analyzed using a Lonestar VOC Analyzer (FAIMS) fitted with an ATLAS headspace
sampler [46]

This pilot study shows that FAIMS analysis of unprocessed stool samples can
differentiate between C. difficile-positive and negative samples with high
diagnostic accuracy. FAIMS analysis is also quick, easy to use, relatively cheap,
and can be used on the ward. FAIMS may be the ideal point-of-care test for C.
difficile in stool samples, reducing delays in diagnosis and treatment.

ATLAS Headspace Sampler


Combine with Lonestar VOC Analyzer to sample
liquid/solid biospecimens

www. owlstonemedical.com/atlas

48
How to analyze
VOC biomarkers
The era of modern breath testing started in 1971, with the work of Nobel Prize
winner Linus Pauling [3]. Since then, hundreds of scientific papers have been
published suggesting the presence of VOC biomarkers across a range of
diseases (Figure 12). There is limited agreement, however, in identified
biomarkers within a disease across these published studies.

The identification of robust VOC biomarkers in breath has been hampered by


sampling and analytical challenges [17]. In addition, small patient numbers in
pilot studies and the lack of blinded validation studies has impeded the
introduction of VOC biomarkers into the clinic.

At Owlstone Medical we have worked to understand and overcome the


challenges of breath analysis, and set up robust workflows for analysis of VOCs
in breath to enable breath biomarker studies within our clinical pipeline and
Biomarker Services.

Owlstone Medicals ReCIVA breath sampler and microchip chemical sensor


platform (FAIMS) have introduced breath as a new diagnostic modality. Now,
for the first time, it is possible to perform robust biomarker discovery using
breath as a sample type, on a platform with the potential to transition to
point-of-care.

49
HISTORICAL CHALLENGES

Modern breath testing commenced in 1971, with the work


of Nobel Prize winner Linus Pauling.
Hundreds of scientific papers published suggesting
the presence of VOC biomarkers across a range
of diseases.

SOME PUZZLING QUESTIONS

1 Why is there very little


agreement in identified
biomarkers within
2 Why is breath testing
not used routinely
in clinical setting?
a disease?

SOME HISTORICAL CHALLENGES


Maturity of breath sampling Study design and size -
hardware and protocols for small patient numbers
robust, repeatable sampling. in pilot studies and lack
of blinded validation
studies.
High end, expensive spectrometer
vs low performance enose.
Different analytical techniques WITHOUT SOLVING THESE YOU CAN'T HAVE
required in biomarker discovery CONFIDENCE IN INITIAL BIOMARKER
and clinical translation. DISCOVERY AND VALIDATION

Figure 25. Historical challenges that have hampered breath analysis, leading to the development of
new technologies that enable the identification of robust VOC biomarkers in breath.

Overcoming challenges in breath


sampling and analysis
Historically the robust and reproducible analysis of VOCs in breath has posed
many technical and analytical challenges which has led to limited agreement
between different studies.

Table 3 provides an overview of the historical challenges faced by breath


researchers, and the solutions now available through technological advances in
breath sampling and analysis.

50
CHALLENGES AND SOLUTIONS IN BREATH SAMPLING AND VOC ANALYSIS

Historical challenges Solutions

Breath Lack of standardized breath collection device ReCIVA Breath Sampler:


sampling Variation in volume of breath collected Reliable, reproducible breath collection
Contamination with exogenous VOCs from Collects a defined breath volume
environment Samples VOCs continuously over ~10 minutes
for increased sensitivity
Tedlar and other polymer storage bags: Collects breath samples onto stable sorbent
Limited sensitivity due to collection of single tubes
exhaled breath Hydrophobic sorbents in sorbent tubes
Significant losses of VOCs from bag minimize water interference
Water condensation on inner surface of bag CASPER Clean Air Supply ensures samples
adsorbs VOCs are not contaminated
Ingress of external VOCs through bag
Contamination with VOCs from bag
Carryover of VOCs when re-using bags

Breath High-end analytical instruments e.g MS: Microchip chemical sensor (FAIMS):
analysis Expensive and bulky Easier to operate than MS approaches
Inter-instrument variability Expertise required to Superior robustness and reproducibility
operate correctly Improved stability, sensitivity and selectivity
Limited throughput compared to eNose systems
Challenging to deploy at point-of-care Small size facilitates deployment at
point-of-care
Electronic Nose (eNose):
Limited temporal stability
Poor chemical selectivity

Table 3. Overview of historical challenges to the adoption of breath sampling, and solutions offered
by Owlstone Medical's Breath Biopsy technology.

Study design
As with any omics technique, the identification of robust VOC biomarkers in
breath involves broadly three stages:
1. Discovery
2. Verification
3. Validation

The above stages encompass the studies involved in discovering the optimal
classifier, verifying the classifier in a larger patient cohort, and performing an
external blinded validation in a larger patient cohort representative of the
intended use population, to validate both the VOC biomarker algorithm and
the analytical instrumentation that together comprise a diagnostic test ready
for regulatory approval.

Key in designing a study is to understand the intended clinical deployment


scenario. In principle, the inclusion and exclusion criteria for subjects should
match the clinical scenario as closely as possible to maximize external validity
of the identified biomarkers. However, some general concepts need to be taken
into consideration when designing a study aiming to identify breath
biomarkers.

51
Tips for a successful breath biomarker study
Use a robust, reproducible, standardized breath sampling protocol.
Select a patient cohort representative of the intended use population.
Record potential confounding factors in a Case Report Form (CRF)
accompanying the study, and assess systematic bias or confounders that
might impact the biomarker signal.

Pre-analytical variables and confounding factors


Challenges related to exhaled breath analysis include variability introduced by
variation in the breath sampling method (e.g. sampling total vs. alveolar breath,
or single-breath vs. fixed-time/-volume) effect of expiratory flow and breath
hold, type of sample collection materials, VOC recovery, and sample
pre-treatment.

It is vital that all breath samples in the study should be collected using a robust,
reproducible, standardized breath sampling protocol (see section on Reliable
collection of breath samples on page 54).

In addition, a number of factors have the potential to contribute to the profile


of VOCs in exhaled breath, including ambient inspired volatiles, effect of
humidity, food/medications, exercise, smoking, and comorbidities (Figure 26).

In the case of pilot or proof-of-concept studies, one approach is to attempt to


control these variables, to minimize their potential impact on the ability to
detect breath biomarkers. However this may not be the most useful approach,
for the following reasons:
1. Signals obtained in the pilot study may not be borne out in subsequent
verification studies including larger patient cohorts where the potential
variables may be less well controlled.
2. Specifying stringent conditions to minimize variation may pose challenges
for implementation of the test into clinical practice.
3. Many of the factors contributing to the VOC profile may be randomly
distributed across cases and controls, hence appropriate statistical
techniques can prevent them from having a big impact on diagnostic
performance.
4. Factors contributing to the VOC profile may actually form part of the signal,
and improve the clinical relevance of the biomarker profile.

There are three complementary approaches that can be taken to minimize the
impact of potential variables on your study outcome, and maximize the ability
to detect a biological signal:
1. Selection criteria for subjects should reflect the actual population in which
the test is to be used as closely as is practically feasible. This means potential
confounders are less likely to impact biomarker discovery.
2. It is crucial to assess systematic bias or confounders that are unevenly
distributed between case and control groups that might impact the
biomarker signal, hence potential confounding factors should be detailed in
a Case Report Form (CRF) accompanying the study. Clinical background
data can be used to assess the accuracy with which the breath biomarkers
can identify the presence of certain confounders.
3. Signal-to-noise ratio should be maximized by reducing background noise
(e.g. using a clean air supply), and using a well standardized and validated
breath collection method and analytical pipeline.

52
FACTORS THAT CAN CONTRIBUTE TO THE VOC PROFILE

GENDER

AGE BMI

SMOKING DRUGS

EXERCISE TIME SINCE EATING

TIME OF DAY TEETH BRUSHING

Figure 26. Factors that can contribute to the VOC profile [100]. It is crucial to assess systematic
bias or confounders that are unevenly distributed between case and control groups that might
impact the biomarker signal, hence potential confounding factors should be detailed in a Case
Report Form (CRF) accompanying the study.

Cohort size
Many of the breath analysis studies reported in the literature involve small
patient cohorts, which limits the statistical significance of the findings, and
makes it challenging to investigate the impact of possible confounding factors.

The sample size needed to reach statistical significance depends both on the
inter-sample variation, and on the expected amplitude of differences the study
is intended to show. In a cohort including a diverse set of human individuals,
the population is likely to be fairly heterogeneous with a high degree of
biological variation.

If the goal is to identify very small differences in VOC levels between


sub-groups of the cohort, then a large number of samples may be required. A
power analysis can be used to evaluate the number of samples needed to
detect an effect of a given size.

Patient population
Care should be taken to select not only the optimal cohort size, but also the
appropriate patient population. This patient population is likely to vary
according to the stage of the study.

53
Initial discovery or pilot studies may involve more clearly defined and
homogeneous groups. Whilst this increases chances of finding a signal, there is
a trade off with the external validity of these results which should be carefully
considered. Trying to focus on a population representative of the intended use
population is key to minimize this risk.

In some cases, e.g. screening, the target condition only occurs infrequently in
the population in which the breath test is to be used. In these cases, recruiting
an enriched population is often an effective way of running the discovery phase
of the study. However, in order to obtain the evidence required for regulatory
approval, the validation study will always need to be in the intended
population.

Ideally a study would focus on the intention to diagnose population during


early test development. This prevents embarking on follow-up studies that turn
out to be false negative as a consequence of population differences between
the studies.

Breath Analysis Workflow


At Owlstone Medical we have developed robust workflows for collection and
analysis of VOCs in breath (Figure 27).

Breath is collected continuously over a ~10 minute time period using the
ReCIVA Breath Sampler, and VOCs are pre-concentrated onto sorbent tubes
for analysis using the Lonestar VOC Analyzer which incorporates the proven
FAIMS microchip chemical sensor technology.

WORKFLOW FOR VOC BIOMARKER ANALYSIS IN BREATH

Clinical sample type Sample collection VOC Analysis

Breath ReCIVA GC-FAIMS


Breath Sampler (Lonestar VOC Analyzer -
and CASPER microchip chemical
Clean Air Supply sensor platform)
Pre-concentrate or GC - MS
breath samples
in sorbent tubes*

Figure 27. Workflow for VOC biomarker collection and analysis in breath. *Sorbent tubes must be
thermally desorbed prior to VOC analysis. Abbreviations: GC (Gas Chromatography), LC (Liquid
Chromatography), MS (Mass Spectrometry).

Reliable collection of breath samples


Reproducibility is key for any analytical technique, and standardization of
breath collection is key to enable the breath diagnostics field to perform robust
breath biomarker discovery for a range of medical diagnostics applications.

To address the challenges in sampling breath, Owlstone Medical formed the


Breathe Free Consortium, a group of 100 multi-disciplinary experts in breath
http://www.breathe-free.org/
http://www.breathe-free.org/
diagnostics, in order to jointly develop the ReCIVA Breath Sampler as the
industry standard for the reliable collection of breath samples.

54
The ReCIVA Breath Sampler provides a much-needed solution to the
challenges of breath using containers such as Tedlar bags which have been
shown to result in the loss of sampled VOCs [101]. Instead, the ReCIVA Breath
Sampler collects breath samples onto stable sorbent tubes, which can be
analyzed immediately or stored for later transport under ambient conditions to
an off-site laboratory for analysis. Sorbent tubes are used by National Institute
for Occupational Safety and Health (NIOSH), and are the accepted industry
standard.

Sampling breath using ReCIVA and sorbent tubes ensure reproducible


collection of VOCs (Figure 28 and Figure 29), unlike other commonly used
collection media, such as Tedlar bags [101].

REPRODUCIBLE BREATH COLLECTION USING ReCIVA


i)
BREATH COLLECT A
6000
SORBENT TUBE 1
5000 SORBENT TUBE 2
INTENSITY

4000

3000

2000

1000

0
22 23 24 25 26 27
RETENTION TIME (MINS)
ii)
BREATH COLLECT B
7000
SORBENT TUBE 1
6000 SORBENT TUBE 2
5000
INTENSITY

4000

3000

2000

1000

0
22 23 24 25 26 27
RETENTION TIME (MINS)
iii)
7000 BREATH COLLECT A
BREATH COLLECT B
6000

5000
INTENSITY

4000

3000

2000

1000

0
22 23 24 25 26 27
RETENTION TIME (MINS)

Figure 28. Chromatogram of VOCs in breath samples collected on independent sorbent tubes
using ReCIVA and analyzed using GC-FAIMS (Lonestar VOC Analyzer). High reproducibility is
observed within a single breath collect (i and ii). Small changes in the chromatogram are observed
between different breath collects performed on the same individual (iii), which is to be expected
because some VOCs in breath do vary over time.

55
REPRODUCIBLE MEASUREMENT OF VOCs
IN BREATH WITH ReCIVA AND FAIMS
12 12

10 10

D-LIMONENE PEAK AREA

D-LIMONENE PEAK AREA


8 8

6 6

4 4

2 2

0 0
1 2 SORBENT 1 2 SORBENT
TUBE TUBE
BREATH COLLECT A BREATH COLLECT B

Figure 29. D-limonene in breath samples collected on independent sorbent tubes (1 and 2) using
ReCIVA and analyzed using GC-FAIMS (Lonestar VOC Analyzer) shows high reproducibility within
a single breath collect. Small changes in D-limonene peak area are observed between different
breath collects (A and B, performed on the same individual), which is to be expected because some
VOCs in breath do vary over time.

How does ReCIVA


compare to Tedlar bags?
Find out here:
www. owlstonemedical.com/sorbent-vs-bags

Importantly, the ReCIVA Breath Sampler also allows VOCs in breath to be


sampled continuously over a ~10 minute time period. This enables the
pre-concentration of low abundance VOCs from much larger breath volumes
than can be practically captured using a bag.

Hence, ReCIVA facilitates the collection of VOCs originating from throughout


the body, and makes it possible to detect even low level VOC biomarkers, such
as those present at the earliest stage of disease (see also page 18).

Pressure and CO2 sensors in the ReCIVA Breath Sampler provide real-time
monitoring of patient breathing, allowing specific breath fractions to be
selected (e.g. enriched bronchial vs. end-tidal). ReCIVA is designed to be used
with the CASPER clean air supply, in order to ensure breath samples are not
contaminated by external VOCs.

Named Invention of the Year in the 2017 Top 50 in Digital Health Awards,
ReCIVA is the first breath sampler able to provide an out-of-the-box means to
reproducibly capture breath samples for analysis.

56
ReCIVA is CE marked and is already in use in over 60 academic research
groups and clinical labs around the world. In addition, a paediatric ReCIVA
breath sampler enables breath collection from children as young as 5.

ReCIVA is currently being deployed as part of the NHS funded lung cancer
detection project (LuCID) - the worlds largest breath-based study ever
https://www.owlstonemedical.com/clinical-pipeline/lucid/

https://www.owlstonemedical.com/clinical-pipeline/lucid/

undertaken for early cancer detection (read more on page 33).

ReCIVA Breath Sampler


For reliable, standardized breath collection

www. owlstonemedical.com/reciva

VOC Biomarker analysis


VOC biomarkers in breath can be analyzed using a range of techniques
including gas chromatography-mass spectrometry (GC-MS), proton
transfer-reaction MS (PTR-MS) [11] and selected ion flow tube MS (SIFT-MS)
[12]. However, these high-end analytical instruments are expensive, bulky and
require training and expertise to operate correctly.

The transition of mass spectrometry techniques to clinical scenarios has also


been hampered by challenges including limited throughput, automation, and
inter-instrument standardization. This has hindered the transition of VOCs from
research laboratory to deployable clinical tests.

A key recent advance in breath analysis is the miniaturization of technologies


which have the potential to be deployed at the point-of-care. Electronic Nose
(eNose) systems use arrays of non-selective chemical sensors to detect VOCs.
However, eNose systems often suffer from unwanted chemical interactions
between VOCs and the sensors, poor intra-device repeatability, limited
temporal stability and poor chemical selectivity.

Proven FAIMS technology


Field asymmetric ion mobility spectrometry (FAIMS) is now available in a
chip-based miniaturized form. FAIMS distinguishes ionized VOCs according to
differences in the speed that they move through a buffer gas under the
influence of an oscillating electric field [102].

Over $70 million has been invested in developing Owlstone Medicals


microchip chemical sensor platform (FAIMS), hence FAIMS been rigorously
validated and is already used in military and industrial applications globally.

Owlstone Medical has incorporated its FAIMS technology into the Lonestar
VOC Analyzer for medical applications, enabling improved stability and
sensitivity compared to eNose systems. In addition, the Lonestar VOC Analyzer
is easier to use than mass spectrometry approaches, and offers superior
robustness and reproducibility, due to its excellent intra-device stability (Figure
30) and ability to selectively detect VOCs in complex biological samples
containing thousands of chemical species.

57
FAIMS spectra are very information rich. This richness is due to the fact that the
technique directly measures the presence and concentration of distinct ionized
molecular species, as opposed to eNose chemical sensors which perform
pattern recognition based on chemical interactions between the sample and
sensor. Multivariate data analysis techniques allow for simple interpretation of
complex FAIMS data to classify different patient groups through highly
sensitive, highly specific detection of relative biomarker concentrations.

Clinical applications
The FAIMS chip is also programmable using powerful software which can
incorporate a range of clinical information. This makes the FAIMS platform
suitable for a wide range of medical applications, some of which are detailed in
a recent review [13].

The small size of the FAIMS microchip enables analysis of samples with
increased throughput, as required for applications in clinical diagnostics, and
will also facilitate development of a device for deployment at point-of-care.

FAIMS, along with the ReCIVA Breath Sampler, is being used in large clinical
trials relating to lung cancer (LuCID), colorectal cancer (InTERCEPT) and
https://www.owlstonemedical.com/clinical-pipeline/lucid/ https://www.owlstonemedical.com/clinical-pipeline/intercept/

https://www.owlstonemedical.com/clinical-pipeline/lucid/ https://www.owlstonemedical.com/clinical-pipeline/intercept/

asthma (STRATA).
https://www.owlstonemedical.com/clinical-pipeline/strata/
https://www.owlstonemedical.com/clinical-pipeline/strata/

EXCELLENT REPRODUCIBILITY OF FAIMS FOR VOC ANALYSIS


160

140

120

100
METHYL SALICYLATE
INTENSITY

D-LIMONENE
80

60

40

20

0
JANUARY FEBRUARY MARCH
TIME
Figure 30. FAIMS technology shows excellent reproducibility. Standard compounds were loaded
onto sorbent tubes at a defined amount and analyzed using GC-FAIMS (Lonestar VOC Analyzer).
The experiment was repeated regularly to assess the variation over time. Data for two compounds
over a three month period is shown. The relative standard deviation is 6.1 % for methyl salicylate and
3.4 % for D-limonene.

Lonestar VOC Analyzer


Based on proven FAIMS technology

www. owlstonemedical.com/lonestar
58
Quality Control
Consistency in the breath collection and VOC analysis workflow is key to
obtaining reliable data for biomarker discovery. In this section we provide an
overview of how we monitor the steps in the workflow to identify any outlier
samples and sources of technical variation.

Quality checks during breath collection


To ensure a successful breath collection using the ReCIVA Breath Sampler, it is
important to verify that the mask fits well by checking that there is no air
escaping from the mask around the nose or mouth.

The pressure and CO2 levels within the ReCIVA are monitored throughout
collection to warn the operator of leaks from the mask, hyperventilation,
hypoventilation, shallow or otherwise irregular breath profiles that negatively
impact VOC collection. Feedback is given in real-time to correct poor
collection technique. Breath profiles and sampling time are also stored along
with other patient metadata to inform subsequent data analysis.

Occasionally, it may not be possible collect the required volume of breath in the
collection time specified for the project. In this case, the breath sample can still
be analyzed, but this sample is flagged in case the reduced breath volume may
affect the results.

Quality control during VOC analysis


Successful VOC biomarker discovery requires a robust data set, where levels of
individual VOCs are measured with a high degree of accuracy. Some VOC
biomarkers may be present at extremely low levels, therefore it is important
that analytical techniques are highly sensitive, with low background levels.

At Owlstone Medical, we have implemented QC checks monitoring a range of


instrument parameters on a daily and weekly basis, to ensure that our VOC
analytical workflows are highly sensitive, accurate and reproducible (Figure
30).

Sorbent tubes are checked prior to breath collection, and blank samples are
regularly tested, to ensure that any background is below a threshold level
(Figure 32). We use a set of standard compounds with a range of
concentrations to check the linearity (Figure 31), limit of detection and
signal-to-noise ratio.

The VOC profile of each patient sample is also inspected for the presence of
compounds that we expect to be present above certain thresholds in a patient
breath sample. This QC serves as an initial check that the analysis has been
successful, before initiating data analysis.

My group have collaborated with Owlstone for several years in the


discovery of biomarkers for a range of diseases including colorectal
cancer, inflammatory bowel disease and tuberculosis. We are very
impressed with the capability of Owlstones new non-invasive
technology.
Professor Ramesh Arasaradnam, University of Warwick
and Consultant Gastroenterologist at UHCW

59
FAIMS SHOWS EXCELLENT REPRODUCIBILITY
AND LINEARITY DOWN TO 70 PPT

3.5

3.0

2.5

PEAK AREA 2.0

1.5

1.0

0.5

0.0

0 1 2 3 4 5 6
D-LIMONENE STANDARD ON-COLUMN MASS (ng)

Figure 31. Calibration plot of D-limonene standard applied to sorbent tubes at a range of amounts
and measured using GC-FAIMS (Lonestar VOC Analyzer). Excellent linearity is observed, enabling
accurate quantification down to 0.2 ng on-column mass which equates to around 70 parts per
trillion (ppt) in breath, with the potential to detect D-limonene down to 35 ppt in breath. Good
precision is demonstrated; data shown is the mean of measurements from 4 - 6 different sorbent
tubes analyzed on 4 different days (error bars represent 1 standard deviation).

MEASURING BACKGROUND VOC LEVELS


BREATH SAMPLE

7.28E+07

AMBIENT AIR COLLECT


ABUNDANCE (COUNTS)

5.46E+07

3.64E+07

BLANK COLLECT

1.82E+07

0
21.6539 21.7 21.8 21.9 22 22.1 22.2 22.277
RETENTION TIME (MINS)
Figure 32. Comparison of the abundance of D-limonene measured in a breath sample, an ambient
air sample or a blank collect sample. All samples were collected using the ReCIVA Breath Sampler
and analyzed using GC-MS. The breath and blank collect samples were collected using the CASPER
clean air supply (in the case of the blank collect sample, breath collect is mimicked). The ambient
air sample was collected by sampling ambient air without the CASPER clean air supply. This
demonstrates the importance of using a clean air supply to avoid contamination by VOCs in the
environment, and measuring the background VOC levels using a blank collect to ensure low level
VOCs are distinguishable from any system-related background.

60
Data analysis
In order to identify VOCs that are potential biomarkers for disease, the raw
spectrometry data must be processed and analyzed. To facilitate this process,
we have developed a data analysis pipeline that takes raw spectrometry data,
patient clinical and outcome data, develops classification algorithms, and
ultimately outputs a classification algorithm that uses multiple biomarkers to
accurately identify patients by disease presence, therapy response or
outcome (Figure 33).

VOC DATA ANALYSIS PIPELINE

Raw data
Build statistically comparable test and training data sets

Training Set Test Set

Data pre-processing Data pre-processing


Calibration and Calibration and
feature extraction feature extraction

Classification
algorithm development
using machine
learning approach

Apply classifiers to
Test Set

Classifier evaluation
Identify confounders
and any sources
of technical variation

Classifier refinement
Iterative validation and
refinement of features
and classifiers

Classifier evaluation
Sensitivity, specificity
and statistical significance

In silico verification
Classifiers are verified using
our proprietary database
of 1000s of patient
VOC profiles

Verification
Verification of the successful classifier in a larger cohort

Figure 33. The VOC data analysis pipeline.

61
Training and test data sets
The first step is to build training and test data sets that are statistically
comparable, where any confounding factors are represented equally. Both
training and test data sets then undergo unsupervised data pre-processing to
generate calibrated retention times and identify features (peaks in the FAIMS
or mass spectrum data, or absence thereof) that can be used for classifier
development. Any features that are system-related are excluded.

IDENTIFICATION OF FEATURES FROM RICH SPECTRAL FAIMS DATA

Acetone DCM
Isoprene
Acetonitrile
2-methyl
E-2-pentene
butane

Carbon siloxane Cyclohexane


Ethanol disulfide Acetic
2-butanone
acid
Pentane Vinyl
acetate Benzene

Figure 34.

Classifier development
Using the training data set, a machine learning approach is taken to train a
learning algorithm and develop classifiers using 10-fold stratified cross
validation [103]. The classifiers are applied to the test data set, and a
systematic study of different sources of bias is performed, in order to identify
any sources of technical variation and to verify that the classifier is robust to
confounding factors. The features and classifiers are then further refined
through an iterative process.

Classifier evaluation
In order to select the best performing classifier, classifier performance is
evaluated in terms of sensitivity, specificity and area under the ROC curve
(Figure 35). In the case of on-going studies, we also assess whether the
number of patients is sufficient for statistical significance, or whether the study
would benefit from recruitment of additional patients.

62
Did you know?
Our Biomarker Services include extensive data analysis including classifier
development and in silico verification using Owlstone Medicals database
containing thousands of patient VOC profiles.

Learn more: www.owlstonemedical.com/services

In silico verification
We use a proprietary database containing thousands of anonymized patient
VOC profiles to verify classification algorithms in silico. We assemble virtual
patient populations defined by demographics, comorbidity, and outcome, and
apply the classifier to gain early evidence of likely clinical performance, without
the need for additional samples or studies. This step can also identify
potentially confounding variables or relevant sub-populations that may not
have been adequately represented in the original trial population.

OPTIMIZATION OF PREDICTIVE VALUES

THRESHOLD
300 1.0
CONTROL DISEASE
250
True Positive Rate

0.8
g
Frequency

200 sin
es
0.6 gu
om
150 nd
ra
of
0.4 line
100 ce
en
fer
re
0.2
50

0 0.0
0 2 4 6 8 10 12 0 0.2 0.4 0.6 0.8 1.0
Signal False Positive Rate

Figure 35. Adjusting the threshold for discrimination between control and disease states enables
the positive and negative predictive values of the test to be optimized.

Verification
Analysis of candidate VOC biomarkers is performed in an independent patient
cohort to verify the performance of the classifier. In this situation, the same
data analysis pipeline will be applied to analyze the the list of candidate
biomarkers identified in the biomarker discovery phase, and further refine the
classifier.

Validation
Following successful verification of the classifier, the classifier is locked down
and an external blinded validation is performed in a larger patient cohort
representative of the intended use population, to validate both the VOC
biomarker algorithm and the analytical instrumentation that together comprise
a diagnostic test ready for regulatory approval.

63
Beyond Breath: VOC Analysis
in other Biospecimens
In addition to breath, Owlstone Medical have also developed robust workflows
for collection and analysis of VOCs in other biospecimens (Figure 36).

Analysis of VOCs emitted by in vitro cell cultures allows researchers to measure


which VOCs are produced by specific human cell types, including tumors. This
can provide valuable information regarding metabolic pathways and disease
mechanism, improving our understanding of the underlying biochemistry
relating to VOC production [104].

In the case of liquid or solid biospecimens, VOCs emitted by the biospecimen


can be sampled from the headspace using the ATLAS Headspace Sampler, and
transferred to Lonestar VOC Analyzer for analysis.

Alternatively, metabolites present in liquid biospecimens may be analyzed


directly from the biofluid by incorporating ultraFAIMS into an LC-MS workflow
for enhanced separation. For example, metabolites in serum may be analyzed
using just 100 l serum.

When analyzing archived samples, such as frozen urine samples, it is


recommended to standardize the sample collection procedures to avoid
technical variability that may be introduced through variation in storage time
or number of freeze-thaw cycles. In our experience, freezing soon after
collection preserves the volatiles in urine.

WORKFLOW FOR VOC BIOMARKER ANALYSIS


IN BIOFLUIDS AND OTHER BIOSPECIMENS

Clinical sample type Sample collection VOC Analysis

Headspace ATLAS GC-FAIMS


of liquid/solid Headspace (Lonestar VOC Analyzer -
biospecimens Sampler microchip chemical
e.g. urine, feces, Sample and transfer sensor platform)
in vitro cell cultures headspace gas or GC - MS
to Lonestar

Clinical sample type Sample collection VOC / Metabolite Analysis

Biofluids Standard LC-FAIMS-MS


e.g. urine, serum, collection (ultraFAIMS -
plasma, blood, saliva, Storage at microchip ion
pus, CSF -20 C mobility technology)

Figure 36. Workflow for VOC biomarker collection and analysis for biofluids and other
biospecimens. Abbreviations: GC (Gas Chromatography), LC (Liquid Chromatography), MS (Mass
Spectrometry).

ATLAS Headspace Sampler


Combine with Lonestar VOC Analyzer to sample
liquid/solid biospecimens

www. owlstonemedical.com/atlas
64
Biomarker
Services
Bring Breath Biopsy to your own research.

Discover novel non-invasive biomarkers for cancer,


inflammatory and infectious disease with our
comprehensive Breath Biomarker Services.

Get in touch with us to arrange a consultation to


discuss your project: owlstonemedical.com/contact
www.

Breath Biomarker Services


Let our experts add Breath Biopsy to your clinical trials

www. owlstonemedical.com/services

65
Outsource your Biomarker Discovery
Owlstone Medicals Biomarker Services enable academics, clinicians and
pharmaceutical companies to identify novel breath biomarkers for diagnostics
as well as precision medicine applications including therapeutic response,
patient stratification and outcome prediction.

We offer a complete set of services for analysis of volatile organic compound


(VOC) biomarkers in breath and other biospecimens. Our services include
everything from consultation on the optimal study design, to biomarker
discovery, verification and external blinded validation of the diagnostic test
algorithm and analytical instrumentation in the context of the clinical use case,
ready for regulatory approval (Figure 37).

We offer CRO capabilities to set up and run clinical trials, drawing on our
experience of running VOC biomarker clinical trials as part of our own clinical
pipeline.

Broadly, VOC biomarker discovery studies follow the stages outlined in Figure 37.

VOC BIOMARKERS:
FROM DISCOVERY TO VALIDATION
Discovery Verification Validation
Analyze full spectrum of Analyze candidate VOC Validation of defined VOC
VOCs in a small cohort biomarkers to verify biomarker algorithm and
and develop classifier. classifier performance analytical instrumentation
in a larger cohort. in large patient cohort
representative of the
intended use population.

In silico verification In silico verification Application for


using Owlstone Medicals using Owlstone Medicals regulatory approval
VOC profile database. VOC profile database.

Figure 37. Stages of a typical VOC biomarker discovery and validation project. During discovery
and verification, the classifier is further verified in silico using a population assembled from
Owlstones proprietary database of patient VOC profiles. Following successful verification of the
classifier, the classifier is locked down and an external blinded validation is performed in a larger
patient cohort representative of the intended use population, to validate both the VOC biomarker
algorithm and the analytical instrumentation that together comprise a diagnostic test ready for
regulatory approval.

BIOMARKER SERVICES WORKFLOW


Consult
Our team of clinical and technical experts advise on all aspects of study design and
sample collection.

Collect
We provide full guidance on optimal collection of breath and other biospecimens. We
install ReCIVA Breath Sampler in clinic and provide training and on-going support.

Analyze
Samples are analyzed using our proven, highly sensitive, FAIMS technology in
state-of-the-art laboratories, generating rich chemical spectral data.

Discover
Our experts identify discriminating features from spectral data and develop robust
classification algorithms using sophisticated machine learning approaches.

Report
Our in-depth report provides the results of our analysis as well as complete raw FAIMS
data. Our analysis includes in silico verification of the classifier using our proprietary
database containing thousands of VOC profiles.

Figure 38.

66
Study design
Prior to initiating a Biomarker Services project, we recommend arranging a
consultation with our team of clinical and technical experts to discuss the study
design. Taking advantage of the experience and knowledge we have developed
through our own clinical programs, we can provide advice on the optimal study
design to ensure that your project has the maximum chance of being
successful (read more on page 51).

We advise on optimal sample collection and recommend approaches to


minimize any pre-analytical variables or confounding factors. We also discuss
how to control for sources of biological variation, and ensure that your study is
sufficiently powered to deliver statistically significant results.

Breath collection
We ensure that breath samples are collected reliably and reproducibly by
installing the ReCIVA Breath Sampler at clinical sites, and providing training for
trial nurses to ensure optimal breath collection. Breath samples are
pre-concentrated on stable sorbent tubes and shipped to Owlstone Medicals
service laboratory for analysis.

Did you know?


As part of our Biomarker Services, we also analyze VOCs in a variety of other
clinical sample types, including:
Urine
Serum
Other biofluids e.g. plasma, blood, saliva, pus, CSF
Feces
In vitro cell cultures
Other sample types upon request

Contact us for more information: owlstonemedical.com/contact


www.

Rapid, high quality results


Owlstone Medical offers the largest laboratory for analysis of VOCs in breath
and other biospecimens. VOC Biomarkers are analyzed using our proven
FAIMS technology. Our exceptional analytical platforms, rigorous quality
standards and experienced staff deliver reliable, reproducible and rapid results.

67
Extensive QC checks are part of our Quality Management System, which covers
the entire analysis workflow starting with preparation of the sorbent tubes for
breath collection. We perform a number of daily and weekly QC checks to
monitor a range of instrument parameters, and use blanks and standard
compounds to monitor background, sensitivity, limit of detection and
signal-to-noise ratio.

Biomarker discovery
Our analysis experts develop robust classification algorithms from rich spectral
FAIMS data, using a sophisticated machine learning approach. Our clinical
experts are on hand to provide biological interpretation of the results in
relation to clinical outcomes. Our data analysis pipeline is designed to take raw
spectrometry data and output a classification algorithm that uses multiple
biomarkers to accurately identify patients by disease presence, therapy
response or outcome.

Classifiers are developed using a training set, and then applied to a test set.
Classifier development includes a systematic study of different sources of bias,
to identify any sources of technical variation and ensure that the classifier is
robust to confounding factors.

Using patient history, comorbidity and outcome information along with VOC
biomarker data we create tests to make actionable patient classifications that
can be deployed in clinic.

In silico verification
Classification algorithms are verified in silico using our proprietary database
containing thousands of patient VOC profiles, to provide early evidence of
likely clinical performance without the need for additional samples or studies.
Our proprietary database contains anonymized chemical spectra records from
all samples collected from trials in the Owlstone Medical clinical pipeline, as
well as Biomarker Service projects (subject to authorization).

We assemble virtual patient populations defined by demographics,


comorbidity, and outcome. Verifying the classification algorithm in these
populations in silico can also identify potentially confounding variables or
relevant sub-populations that may not have been adequately represented in
the original trial population.

68
Report
Our results are delivered in a format which is easily interpretable by a clinician.
We provide in-depth reports on the results of our statistical analysis including
principal component analysis, box plots (Figure 39) and receiver operating
characteristics curves (ROC). In addition, we provide the complete set of raw
FAIMS data for your research.

Our analysts, developers and clinical research staff will walk you through the
analysis and the results either in person or by phone. Our experts are on hand
to answer any questions that you may have about the current project or
possible next steps, including verification of the classifier, and validation of the
diagnostic test.

CLASSIFICATION PROBABILITIES

0.9
CONTROL
GROUP A
GROUP B
0.8
PREDICTIVE PROBABILITY

0.7

0.6

0.5

0.4

0.3

0.2

0.1

PC1 PC2
Figure 39.

Verification and Validation


Analysis of candidate VOC biomarkers is performed in a larger patient cohort
to verify the performance of the classifier, and further refine the classifier if
needed.

Following successful verification of the classifier, an external blinded validation


is performed in a larger patient cohort representative of the intended use
population, to validate both the VOC biomarker algorithm and the analytical
instrumentation that together comprise a diagnostic test ready for regulatory
approval.

Our experts are on hand to provide advice and guidance on the research
protocol, securing ethics, sample collection and storage and managing the
project to meet your individual requirements.

All data is blinded to our statisticians, and the blinding key is provided by an
independent clinician provides only after the test classification is reported,
allowing full confidence in the validity of the results.

69
Breath Biopsy
Products
Bring VOC Biomarker research to your own lab with our
complete range of products for breath collection and analysis.

ReCIVA Breath Sampler

ReCIVA provides an out-of-the-box means to collect breath samples reliably


and reproducibly.
Collect different breath fractions (enriched bronchial or end-tidal)
Pre-concentrate a large volume of breath onto sorbent tubes, enabling even
VOCs present at low levels to be detected
Developed in conjunction with a multi-disciplinary group of 100 breath
analysis experts

www. owlstonemedical.com/reciva
70
CASPER Clean Air Supply

CASPER provides a consistent, portable supply of clean air, ideal for use with
the ReCIVA Breath Sampler.
Ensure breath samples are not contaminated by external VOCs
Remove the need for a clean air supply line
Portable to wherever the patient is in clinic

www. owlstonemedical.com/casper
Lonestar VOC Analyzer

The Lonestar VOC Analyzer is an easy to use instrument for the detection of
VOC biomarkers in clinical samples such as breath and other biospecimens.
Rapid, sensitive, and selective VOC analyzer
Based on proven FAIMS technology
Biomarker discovery and targeted analysis

www. owlstonemedical.com/lonestar

Use the ATLAS Headspace Sampler to analyze VOCs emitted from liquid or
solid clinical samples on the Lonestar VOC Analyzer.

Learn more: owlstonemedical.com/atlas


www.

71
Further reading
Search for publications, posters etc. on the resources page:
https://www.owlstonemedical.com/resources/

Research case studies involving VOC biomarkers:


https://www.owlstonemedical.com/science-technology/research-case-studies/

Drug Discovery World Article: Breath based biomarker detection:


https://www.owlstonemedical.com/download/breath-based-biomarkers-precision-medicine/

Whitepaper: Application of Breath Volatile Organic Compounds (VOCs) in Asthma:


https://www.owlstonemedical.com/download/asthma-whitepaper/

Innovations in Pharmaceutical Technology - Breath Diagnostics:


https://www.owlstonemedical.com/media/uploads/files/IPT-Breath-Diagnostics-April-
2017.pdf

Breath Analysis Products and Services Overview:


https://www.owlstonemedical.com/media/uploads/files/Breath_Analysis_Products_Se
rvices_overview_screen.pdf

ReCIVA Breath Sampler flyer:


https://www.owlstonemedical.com/media/uploads/files/Reciva_slicksheet.pdf

ReCIVA Breath Sampler Manual:


https://support.owlstonenanotech.com/hc/en-us/articles/212371146-ReCIVA-Breath-Sa
mpler-Instructions-for-Use

ReCIVA Breath Sampler FAQs:


https://www.owlstonemedical.com/products/reciva-faq/

Application Note comparing sorbent tubes and Tedlar bags:


https://www.owlstonemedical.com/media/uploads/files/Reciva_sorbentvsTedlar.pdf

CASPER Clean Air Supply Manual:


https://www.owlstonemedical.com/media/uploads/files/CC-900583-TM-1A-90-0583_
CASPER_Air_Supply_-_Instructions_for_Use.pdf

Lonestar VOC Analyzer brochure:


https://www.owlstonemedical.com/media/uploads/files/Lonestar_Medical_Brochure_3
ITyl7e.pdf

Whitepaper: Headspace sampling:


https://www.owlstonemedical.com/media/uploads/files/Theory_of_headspace_sampli
ngcover.pdf

Ebook: Simulating human breath: The complete guide to generating VOC and
humidity calibration standards:
https://www.owlstonemedical.com/download/simulating-human-breath-complete-gui
de

FAIMS microchip chemical sensor technology:


https://www.owlstonemedical.com/science-technology/faims-technology/

LuCID Clinical Trial for early detection of lung cancer:


https://www.owlstonemedical.com/clinical-pipeline/lucid/

72
LuCID Infographic:
https://www.owlstonemedical.com/media/uploads/files/LuCID_Infographic.pdf

STRATA Clinical Trial for stratification of asthma treatment:


https://www.owlstonemedical.com/clinical-pipeline/strata/

STRATA Infographic:
https://www.owlstonemedical.com/media/uploads/files/STRATA-infographic.pdf

InTERCEPT Clinical Trial for early detection of colorectal cancer:


https://www.owlstonemedical.com/clinical-pipeline/intercept/

InTERCEPT Infographic:
https://www.owlstonemedical.com/media/uploads/files/INTERCEPT-infographic.pdf

73
References
https://www.owlstonemedical.com/media/uploads/files/Complete-guide-to-Breath-Biopsy.bib
Download .BIB file containing all references in this guide
https://www.owlstonemedical.com/media/uploads/files/Complete-guide-to-Breath-Biopsy.bib

[1] Amann et al., The human volatilome: volatile organic compounds (VOCs) in
exhaled breath, skin emanations, urine, feces and saliva, J. Breath Res. 8 (2014)
34001. http://dx.doi.org/10.1088/1752-7155/8/3/034001
[2] Laffel, Ketone bodies: a review of physiology, pathophysiology and application of
monitoring to diabetes., Diabetes. Metab. Res. Rev. 15 (1999) 412426.
http://www.ncbi.nlm.nih.gov/pubmed/10634967.
[3] Pauling et al., Quantitative analysis of urine vapor and breath by gas-liquid
partition chromatography., Proc. Natl. Acad. Sci. U. S. A. 68 (1971) 23742376.
http://www.ncbi.nlm.nih.gov/pubmed/5289873.
[4] Teranishi et al., Gas chromatography of volatiles from breath and urine., Anal.
Chem. 44 (1972) 1820. http://www.ncbi.nlm.nih.gov/pubmed/5006888.
[5] Riely et al., Ethane evolution: a new index of lipid peroxidation., Science. 183
(1974) 208210. http://www.ncbi.nlm.nih.gov/pubmed/4808857.
[6] Conkle et al., Trace composition of human respiratory gas., Arch. Environ. Health.
30 (1975) 290295. http://www.ncbi.nlm.nih.gov/pubmed/1137435.
[7] Jansson et al., Analysis of organic compounds in human breath by gas
chromatography-mass spectrometry., J. Lab. Clin. Med. 74 (1969) 961966.
http://www.ncbi.nlm.nih.gov/pubmed/5359667.
[8] Gelmont et al., Isoprene-the main hydrocarbon in human breath., Biochem.
Biophys. Res. Commun. 99 (1981) 14561460.
http://www.ncbi.nlm.nih.gov/pubmed/7259787.
[9] King et al., Isoprene and acetone concentration profiles during exercise on an
ergometer., J. Breath Res. 3 (2009) 27006.
http://dx.doi.org/10.1088/1752-7155/3/2/027006.
[10] Koc et al., The role of mathematical modeling in VOC analysis using isoprene as a
prototypic example., J. Breath Res. 5 (2011) 37102.
http://dx.doi.org/10.1088/1752-7155/5/3/037102.
[11] Taucher et al., Detection of isoprene in expired air from human subjects using
proton-transfer-reaction mass spectrometry., Rapid Commun. Mass Spectrom. 11
(1997) 12301234.
http://dx.doi.org/10.1002/(SICI)1097-0231(199707)11:11<1230::AID-RCM3>3.0.CO;2-Z.
[12] Spanel et al., Selected ion flow tube mass spectrometry analyses of stable
isotopes in water: isotopic composition of H3O+ and H3O+ (H2O)3 ions in
exchange reactions with water vapor, J. Am. Soc. Mass Spectrom. 11 (2000)
866875. http://www.ncbi.nlm.nih.gov/pubmed/11014448.
[13] Covington et al., The application of FAIMS gas analysis in medical diagnostics,
Analyst. 140 (2015) 67756781. http://dx.doi.org/10.1039/C5AN00868A.
[14] Sahota et al., A simple breath test for tuberculosis using ion mobility: A pilot
study., Tuberculosis (Edinb). 99 (2016) 143146.
http://dx.doi.org/10.1016/j.tube.2016.05.005.
[15] Arasaradnam et al., Breathomics--exhaled volatile organic compound analysis to
detect hepatic encephalopathy: a pilot study., J. Breath Res. 10 (2016) 16012.
http://dx.doi.org/10.1088/1752-7155/10/1/016012.
[16] Arasaradnam et al., Non-invasive exhaled volatile organic biomarker analysis to
detect inflammatory bowel disease (IBD)., Dig. Liver Dis. 48 (2016) 148153.
http://dx.doi.org/10.1016/j.dld.2015.10.013.
[17] Risby, Critical issues for breath analysis, J. Breath Res. 2 (2008) 30302.
http://dx.doi.org/10.1088/1752-7163/2/3/030302.
[18] Herbig et al., Towards standardization in the analysis of breath gas volatiles., J.
Breath Res. 8 (2014) 37101. http://dx.doi.org/10.1088/1752-7155/8/3/037101.
[19] Graham et al., CAMPYLOBACTER PYLORI DETECTED NONINVASIVELY BY THE
13C-UREA BREATH TEST, Lancet. 329 (1987) 11741177.
http://dx.doi.org/10.1016/S0140-6736(87)92145-3.
[20] Ryter et al., Carbon monoxide in exhaled breath testing and therapeutics, J.
Breath Res. 7 (2013) 17111. http://dx.doi.org/10.1088/1752-7155/7/1/017111.
[21] Eisenmann et al., Implementation and interpretation of hydrogen breath tests., J.
Breath Res. 2 (2008) 46002. http://dx.doi.org/10.1088/1752-7155/2/4/046002.

74
[22] Tuboly et al., Determination of endogenous methane formation by photoacoustic
spectroscopy, J. Breath Res. 7 (2013) 46004.
http://dx.doi.org/10.1088/1752-7155/7/4/046004.
[23] de Lacy Costello et al., The importance of methane breath testing: a review., J.
Breath Res. 7 (2013) 24001. http://dx.doi.org/10.1088/1752-7155/7/2/024001.
[24] Pimentel et al., Methane, a gas produced by enteric bacteria, slows intestinal
transit and augments small intestinal contractile activity., Am. J. Physiol.
Gastrointest. Liver Physiol. 290 (2006) G1089--95.
http://dx.doi.org/10.1152/ajpgi.00574.2004.
[25] Taylor, Advances in the clinical applications of exhaled nitric oxide
measurements, J. Breath Res. 6 (2012) 47102.
http://dx.doi.org/10.1088/1752-7155/6/4/047102.
[26] Hgman, Extended NO analysis in health and disease., J. Breath Res. 6 (2012)
47103. http://dx.doi.org/10.1088/1752-7155/6/4/047103.
[27] Gelb et al., Review of exhaled nitric oxide in chronic obstructive pulmonary
disease, J. Breath Res. 6 (2012) 47101.
http://dx.doi.org/10.1088/1752-7155/6/4/047101.
[28] Bucca et al., Exhaled nitric oxide (FENO) in non-pulmonary diseases., J. Breath
Res. 6 (2012) 27104. http://dx.doi.org/10.1088/1752-7155/6/2/027104.
[29] Piacentini et al., Exhaled nitric oxide in pediatrics: what is new for practice
purposes and clinical research in children?, J. Breath Res. 6 (2012) 27103.
http://dx.doi.org/10.1088/1752-7155/6/2/027103.
[30] Heartsbreath Test for Heart Transplant Rejection - Full Text View -
ClinicalTrials.gov, (n.d.). https://clinicaltrials.gov/show/NCT01397812.
[31] Health et al., Recently-Approved Devices - Gastric Emptying Breath Test (GEBT)
- P110015, (n.d.).
https://www.fda.gov/MedicalDevices/ProductsandMedicalProcedures/DeviceAp
provalsandClearances/Recently-ApprovedDevices/ucm443456.htm.
[32] EC50 MICRO+ SMOKERLYZER, (n.d.).
https://www.accessdata.fda.gov/scripts/cdrh/cfdocs/cfpmn/pmn.cfm?ID=K082315.
[33] ClinicalTrials.gov, (n.d.). https://clinicaltrials.gov/.
[34] LuCID - Lung Cancer Indicator Detection, (n.d.).
https://clinicaltrials.gov/ct2/show/NCT02612532.
[35] InTERCEPT, (n.d.). https://www.owlstonemedical.com/clinical-pipeline/intercept/.
[36] STRATA, (n.d.). https://www.owlstonemedical.com/clinical-pipeline/strata/.
[37] Serkova et al., The emerging field of quantitative blood metabolomics for
biomarker discovery in critical illnesses., Am. J. Respir. Crit. Care Med. 184 (2011)
647655. http://dx.doi.org/10.1164/rccm.201103-0474CI.
[38] Mastrangelo et al., Metabolomics as a tool for drug discovery and personalised
medicine. A review., Curr. Top. Med. Chem. 14 (2014) 26272636.
http://www.ncbi.nlm.nih.gov/pubmed/25515755.
[39] van der Schee et al., Breathomics in Lung Disease, Chest. 147 (2015).
http://dx.doi.org/10.1378/chest.14-0781.
[40] Calenic et al., Oxidative stress and volatile organic compounds: interplay in
pulmonary, cardio-vascular, digestive tract systems and cancer, Open Chem. 13
(2015). http://dx.doi.org/10.1515/chem-2015-0105.
[41] Pleil et al., Evolution of environmental exposure science: using breath-borne
biomarkers for discovery of the human exposome., Anal. Chem. 85 (2013)
99849990. http://dx.doi.org/10.1021/ac402306f.
[42] Tu et al., Human exposure to the jet fuel, JP-8., Aviat. Space. Environ. Med. 75
(2004) 4959. http://www.ncbi.nlm.nih.gov/pubmed/14736133.
[43] Egeghy et al., Benzene and naphthalene in air and breath as indicators of
exposure to jet fuel., Occup. Environ. Med. 60 (2003) 969976.
http://www.ncbi.nlm.nih.gov/pubmed/14634191.
[44] Bajtarevic et al., Noninvasive detection of lung cancer by analysis of exhaled
breath, BMC Cancer. 9 (2009) 348. http://dx.doi.org/10.1186/1471-2407-9-348.
[45] Ajibola et al., Effects of dietary nutrients on volatile breath metabolites, J. Nutr.
Sci. 2 (2013) e34. http://dx.doi.org/10.1017/jns.2013.26.
[46] Bomers et al., Rapid, accurate, and on-site detection of C. difficile in stool
samples., Am. J. Gastroenterol. 110 (2015) 588594.
http://dx.doi.org/10.1038/ajg.2015.90.
[47] Trefz et al., Drug detection in breath: non-invasive assessment of illicit or
pharmaceutical drugs, J. Breath Res. 11 (2017) 24001.
http://dx.doi.org/10.1088/1752-7163/aa61bf.
[48] Bowcock, Finding Genes for Common Diseases Using GWAS, Nat. Educ. 8 (2015) 5.
https://www.nature.com/scitable/nated/topicpage/finding-genes-for-common-di
seases-using-gwas-132732965.

75
[49] Tomasetti et al., Stem cell divisions, somatic mutations, cancer etiology, and
cancer prevention, Science (80-. ). 355 (2017) 13301334.
http://dx.doi.org/10.1126/science.aaf9011.
[50] Raamsdonk et al., A functional genomics strategy that uses metabolome data to
reveal the phenotype of silent mutations., Nat. Biotechnol. 19 (2001) 4550.
http://dx.doi.org/10.1038/83496.
[51] Wishart, Chapter 3: Small Molecules and Disease, PLoS Comput. Biol. 8 (2012)
e1002805. http://dx.doi.org/10.1371/journal.pcbi.1002805.
[52] Bettegowda et al., Detection of circulating tumor DNA in early- and late-stage
human malignancies., Sci. Transl. Med. 6 (2014) 224ra24.
http://dx.doi.org/10.1126/scitranslmed.3007094.
[53] Pereira et al., Breath analysis as a potential and non-invasive frontier in disease
diagnosis: an overview., Metabolites. 5 (2015) 355.
http://dx.doi.org/10.3390/metabo5010003.
[54] Wilson, Advances in electronic-nose technologies for the detection of volatile
biomarker metabolites in the human breath., Metabolites. 5 (2015) 140163.
http://dx.doi.org/10.3390/metabo5010140.
[55] Loizou, Animal-Free Chemical Safety Assessment, Front. Pharmacol. 7 (2016)
218. http://dx.doi.org/10.3389/fphar.2016.00218.
[56] The Global Asthma Report, (2014).
http://www.globalasthmareport.org/burden/burden.php.
[57] Montuschi et al., Diagnostic performance of an electronic nose, fractional exhaled
nitric oxide, and lung function testing in asthma, Chest. 137 (2010) 790796.
http://dx.doi.org/10.1378/chest.09-1836.
[58] Dallinga et al., Volatile organic compounds in exhaled breath as a diagnostic tool
for asthma in children., Clin. Exp. Allergy. 40 (2010) 6876.
http://dx.doi.org/10.1111/j.1365-2222.2009.03343.x.
[59] Dragonieri et al., An electronic nose in the discrimination of patients with asthma
and controls, J. Allergy Clin. Immunol. 120 (2007) 856862.
http://dx.doi.org/10.1016/j.jaci.2007.05.043.
[60] Ibrahim et al., Non-invasive phenotyping using exhaled volatile organic
compounds in asthma., Thorax. 66 (2011) 804809.
http://dx.doi.org/10.1136/thx.2010.156695.
[61] van der Schee et al., Predicting steroid responsiveness in patients with asthma
using exhaled breath profiling, Clin. Exp. Allergy. 43 (2013) 12171225.
http://dx.doi.org/10.1111/cea.12147.
[62] Westerhof et al., Biomarkers to identify sputum eosinophilia in different adult
asthma phenotypes, Eur. Respir. J. 46 (2015) 688 LP-- 696.
http://erj.ersjournals.com/content/46/3/688.abstract.
[63] Omalizumab for treating severe persistent allergic asthma | Guidance and
guidelines | NICE, (n.d.).
https://www.nice.org.uk/guidance/ta278?unlid=95300737020161031161546.
[64] Santini et al., Discrimination between oral corticosteroid-treated and oral
corticosteroid-non-treated severe asthma patients by an electronic nose
platform, Eur. Respir. J. 44 (2014).
http://erj.ersjournals.com/content/44/Suppl%7B_%7D58/P2054.abstract.
[65] Santini et al., Breathomics can discriminate between anti IgE-treated and
non-treated severe asthma adults, Eur. Respir. J. 46 (2015).
http://erj.ersjournals.com/content/46/suppl%7B_%7D59/OA1463.abstract.
[66] Wang et al., Metabolomics in Cancer Biomarker Research, Curr. Pharmacol.
Reports. 2 (2016) 293298. http://dx.doi.org/10.1007/s40495-016-0074-x.
[67] Nagarajan et al., Oncogene-Directed Alterations in Cancer Cell Metabolism,
Trends in Cancer. 2 (2016). http://dx.doi.org/10.1016/j.trecan.2016.06.002.
[68] Feinberg et al., Cancer metabolism: the volatile signature of glycolysis-in vitro
model in lung cancer cells., J. Breath Res. 11 (2017) 16008.
http://dx.doi.org/10.1088/1752-7163/aa51d6.
[69] Feinberg et al., Cancerous Glucose Metabolism in Lung Cancer-Evidence From
Exhaled Breath Analysis, J. Breath Res. 10 (2016).
http://dx.doi.org/10.1088/1752-7155/10/2/026012.
[70] Krilaviciute et al., Detection of cancer through exhaled breath: a systematic
review, Oncotarget. 6 (2015) 3864338657.
http://dx.doi.org/10.18632/oncotarget.5938.
[71] Cancer of the Lung and Bronchus - Cancer Stat Facts, (n.d.).
https://seer.cancer.gov/statfacts/html/lungb.html.
[72] Saalberg et al., VOC breath biomarkers in lung cancer, Clin. Chim. Acta. 459
(2016) 59. http://dx.doi.org/10.1016/j.cca.2016.05.013.

76
[73] Dragonieri et al., An electronic nose in the discrimination of patients with
non-small cell lung cancer and COPD., Lung Cancer. 64 (2009) 166170.
http://dx.doi.org/10.1016/j.lungcan.2008.08.008.
[74] Chen et al., A study of the volatile organic compounds exhaled by lung cancer
cells in vitro for breath diagnosis, Cancer. 110 (2007) 835844.
http://dx.doi.org/10.1002/cncr.22844.
[75] Fuchs et al., Breath gas aldehydes as biomarkers of lung cancer, Int. J. Cancer.
126 (2009) NA--NA. http://dx.doi.org/10.1002/ijc.24970.
[76] Gordon et al., Volatile organic compounds in exhaled air from patients with lung
cancer., Clin. Chem. 31 (1985) 12781282.
http://www.ncbi.nlm.nih.gov/pubmed/4017231.
[77] Hakim et al., Volatile Organic Compounds of Lung Cancer and Possible
Biochemical Pathways, Chem. Rev. 112 (2012) 59495966.
http://dx.doi.org/10.1021/cr300174a.
[78] Machado et al., Detection of lung cancer by sensor array analyses of exhaled
breath., Am. J. Respir. Crit. Care Med. 171 (2005) 12861291.
http://dx.doi.org/10.1164/rccm.200409-1184OC.
[79] ONeill et al., A computerized classification technique for screening for the
presence of breath biomarkers in lung cancer., Clin. Chem. 34 (1988) 16131618.
http://www.ncbi.nlm.nih.gov/pubmed/3042190.
[80] Phillips et al., Detection of lung cancer with volatile markers in the breath., Chest.
123 (2003) 21152123. http://www.ncbi.nlm.nih.gov/pubmed/12796197.
[81] Cancer of the Colon and Rectum - Cancer Stat Facts, (n.d.).
https://seer.cancer.gov/statfacts/html/colorect.html.
[82] Altomare et al., Exhaled volatile organic compounds identify patients with
colorectal cancer, Br. J. Surg. 100 (2013) 144150.
http://dx.doi.org/10.1002/bjs.8942.
[83] Peng et al., Detection of lung, breast, colorectal, and prostate cancers from
exhaled breath using a single array of nanosensors., Br. J. Cancer. 103 (2010)
542551. http://dx.doi.org/10.1038/sj.bjc.6605810.
[84] Arasaradnam et al., Detection of Colorectal Cancer (CRC) by Urinary Volatile
Organic Compound Analysis, PLoS One. 9 (2014) e108750.
http://dx.doi.org/10.1371/journal.pone.0108750.
[85] CDC - Epidemiology of the IBD - Inflammatory Bowel Disease, (n.d.).
https://www.cdc.gov/ibd/ibd-epidemiology.htm.
[86] Kurada et al., Review article: breath analysis in inflammatory bowel diseases,
Aliment. Pharmacol. {&} Ther. 41 (2015) 329341.
http://dx.doi.org/10.1111/apt.13050.
[87] Blotire et al., Perforations and haemorrhages after colonoscopy in 2010: a study
based on comprehensive French health insurance data (SNIIRAM)., Clin. Res.
Hepatol. Gastroenterol. 38 (2014) 112117.
http://dx.doi.org/10.1016/j.clinre.2013.10.005.
[88] Friedman et al., Factors that affect adherence to surveillance colonoscopy in
patients with inflammatory bowel disease., Inflamm. Bowel Dis. 19 (2013)
534539. http://dx.doi.org/10.1097/MIB.0b013e3182802a3c.
[89] Devlin et al., Evolving Inflammatory Bowel Disease Treatment Paradigms:
Top-Down Versus Step-Up, Med. Clin. North Am. 94 (2010) 118.
http://dx.doi.org/10.1016/j.mcna.2009.08.017.
[90] Yoshida et al., Diagnosis of gastroenterological diseases by metabolome analysis
using gas chromatographymass spectrometry, J. Gastroenterol. 47 (2012) 920.
http://dx.doi.org/10.1007/s00535-011-0493-8.
[91] Kumar et al., Duration of hypotension before initiation of effective antimicrobial
therapy is the critical determinant of survival in human septic shock., Crit. Care
Med. 34 (2006) 15891596.
http://dx.doi.org/10.1097/01.CCM.0000217961.75225.E9.
[92] Grace et al., Usefulness of Blood Culture for Hospitalized Patients Who Are
Receiving Antibiotic Therapy, Clin. Infect. Dis. 32 (2001) 16511655.
http://dx.doi.org/10.1086/320527.
[93] Bos et al., Volatile Metabolites of Pathogens: A Systematic Review, PLoS Pathog.
9 (2013) e1003311. http://dx.doi.org/10.1371/journal.ppat.1003311.
[94] Thorn et al., Multivariate analysis of bacterial volatile compound profiles for
discrimination between selected species and strains in vitro., J. Microbiol.
Methods. 84 (2011) 258264. http://dx.doi.org/10.1016/j.mimet.2010.12.001.
[95] Wilson et al., Advances in Electronic-Nose Technologies Developed for
Biomedical Applications, Sensors. 11 (2011) 11051176.
http://dx.doi.org/10.3390/s110101105.

77
[96] Schulz et al., Bacterial volatiles: the smell of small organisms., Nat. Prod. Rep. 24
(2007) 814842. http://dx.doi.org/10.1039/b507392h.
[97] WHO Tuberculosis Diagnostics Fact Sheet, (n.d.).
http://www.who.int/tb/publications/tbDiagnostics_factsheet1.pdf?ua=1.
[98] Dhingra et al., Validity and reliability of sputum smear examination as diagnostic
and screening test for tuberculosis, Indian J Allergy. (2003).
[99] du Preez et al., Metabolomics biomarkers for tuberculosis diagnostics: current
status and future objectives., Biomark. Med. 11 (2017) 179194.
http://dx.doi.org/10.2217/bmm-2016-0287.
[100] Blanchet et al., Factors that influence the volatile organic compound content in
human breath, J. Breath Res. 11 (2017).
http://dx.doi.org/0.1088/1752-7163/aa5cc5.
[101] Beauchamp et al., On the use of Tedlar bags for breath-gas sampling and
analysis, J. Breath Res. 2 (2008).
http://dx.doi.org/10.1088/1752-7155/2/4/046001.
[102] Wilks et al., Characterization of a miniature, ultra-high-field, ion mobility
spectrometer, Int. J. Ion Mobil. Spectrom. 15 (2012) 199222.
http://dx.doi.org/10.1007/s12127-012-0109-x.
[103] Kohavi, A Study of Cross-Validation and Bootstrap for Accuracy Estimation and
Model Selection, Proceeding 14th Int. Jt. Conf. Artif. Intell. 2 (1995) 11371143.
[104] Filipiak et al., A Compendium of Volatile Organic Compounds (VOCs) Released
By Human Cell Lines, Curr. Med. Chem. 23 (2016) 21122131.
http://dx.doi.org/0.2174/09 2986732366616031612505.

Legal statement
The document contains information about products that may or may not be available in any particular country or region of the world,
may be available under different trademarks in different countries and, where applicable, may be approved or cleared by a
government regulatory body for sale or use with different indications and restrictions in different countries. Readers are advised that
alternative products and services may be available from other manufacturers and service providers. Breath Biopsy and ReCIVA are
registered trademarks of Owlstone Medical Ltd. Tedlar is a registered trademark of E. I. du Pont de Nemours and Company.

78
OWLSTONE MEDICAL LTD

Cambridge, UK
162 Cambridge Science Park
Milton Road
Cambridge CB4 0GH

T: +44 (0)1223 428200


F: +44 (0)1223 428201

OWLSTONE MEDICAL LTD

London, UK
Unit 19
205 Richmond Road
London E8 3FF

T: +44 (0)2036 385656

OWLSTONE INC

Connecticut, USA
Owlstone Inc.
761 Main Avenue
Norwalk, CT 06851
OW-012042-MC - v1.0 - 04/2017

T: +1 203-908-4848
F: +1 203-908-4849

www.owlstonemedical.com