Optimisation of sampling parameters for standardised exhaled breath sampling

The lack of standardisation of breath sampling is a major contributing factor to the poor repeatability of results and hence represents a barrier to the adoption of breath tests in clinical practice. On-line and bag breath sampling have advantages but do not suit multicentre clinical studies whereas storage and robust transport are essential for the conduct of wide-scale studies. Several devices have been developed to control sampling parameters and to concentrate volatile organic compounds (VOCs) onto thermal desorption (TD) tubes and subsequently transport those tubes for laboratory analysis. We conducted three experiments to investigate (i) the fraction of breath sampled (whole versus lower expiratory exhaled breath); (ii) breath sample volume (125, 250, 500 and 1000 ml); and (iii) breath sample flow rate (400, 200, 100 and 50 ml min−1). The target VOCs were acetone and potential volatile biomarkers for oesophago-gastric cancer belonging to the aldehyde, fatty acids and phenol chemical classes. We also examined the collection execution time and the impact of environmental contamination. The experiments showed that the use of exhaled breath-sampling devices requires the selection of optimum sampling parameters. The increase in sample volume has improved the levels of VOCs detected. However, the influence of the fraction of exhaled breath and the flow rate depends on the target VOCs measured. The concentration of potential volatile biomarkers for oesophago-gastric cancer was not significantly different between the whole and lower airway exhaled breath. While the recovery of phenols and acetone from TD tubes was lower when breath sampling was performed at a higher flow rate, other VOCs were not affected. A dedicated ‘clean air supply’ reduces the contamination from ambient air, but the breath collection device itself can be a source of contaminants. In clinical studies using VOCs to elicit potential biomarkers of gastro-oesophageal cancer, the optimum parameters are 500 mls sample volume of whole breath with a flow rate of 200 ml min−1.

However, there has been a paucity of external validation studies where researchers have validated the findings in an independent population [17]. Currently, VOCs that are in routine clinical applications include exhaled nitric oxide in asthma [13,18,19], C urea breath testing for H. pylori [20] and hydrogen/ methane testing for small bowel intestinal overgrowth [21].
The lack of standardisation of breath sampling is a major contributing factor to the poor repeatability of results [22,23] and hence represents a major barrier to the adoption of breath tests in clinical practice. There is a lesson to be learned from the study of exhaled Any further distribution of this work must maintain attribution to the author(s) and the title of the work, journal citation and DOI. nitric oxide, as a biomarker for pulmonary inflammation. A turning point for the use of nitric oxide in the management of asthma was the development of international consensus guidelines (American Thoracic and European Respiratory Societies, 2005) [24] for the standardised measurement of exhaled nitric oxide that ultimately led to its utility as a diagnostic tool in clinical practice.
The critical importance of standardisation of breath analysis techniques for the identification and quantification of VOCs has been acknowledged and investigated in recent years [22,23,25]. Respiratory manoeuvres have been shown to influence VOC measurements [26]. The method of collecting breath samples also affects the level and profile of the VOCs measured. On-line sampling using direct injection methods such as proton transfer reaction-mass spectrometry (PTR-MS) [27,28] and selected ion flow tube-mass spectrometry (SIFT-MS) [29] reduce the effect of environmental contamination and loss of VOCs due to storage and transport. However on-line measurements using PTR-MS and SIFT-MS are challenging in a clinical environment. The utility of direct measurement has practical challenges as direct sampling on a wide-scale necessitates dedicated breath analysis laboratories/clinics with significant influence on work flow and economic consequences. Nalophan, Tedlar and inertised aluminium bags have been frequently used in clinical profiling studies due to their simplicity and low cost. However, on-line and bag sampling does not suit wide-scale multicentre clinical studies where storage and robust transport methods are essential for the conduct of those studies. Several devices and techniques have been developed to concentrate VOCs onto thermal desorption (TD) tubes and subsequently transport those tubes to the laboratory for analysis. Such an approach allows the control of breath sample volume, sample flow rate and the fraction of breath sampled (whole breath including mouth air versus lower respiratory exhaled breath). Those parameters were the subject of the current investigation with the aim to determine the optimum parameters for use in clinical studies.

Breath-sampling device
Exhaled breath samples were collected using a standardised breath-sampling device, 'Respiration Collector for In Vitro Analysis' (ReCIVA TM ) (Owlstone Medical, Cambridge, UK) in combination with a dedicated clean air supply 'Clean Air Supply Pump for ReCIVA' (CASPER) (Owlstone Medical, Cambridge, UK). For every sampling episode, the ReCIVA allows exhaled breath from the subject to be concentrated onto four Tenax/Carbograph-5TD TD tubes (Markes International Ltd, Llantrisant, UK). The device permits specific fractions of exhaled breath to be collected onto TD tubes through continuous monitoring of pressure and CO 2 levels within the mask during respiration with the pumps within the device being turned on in response to the appropriate phase of the respiratory cycle to allow a specific fraction of exhaled breath within the mask to be pumped onto the TD tube. The CASPER provides a continuous supply of room air at a flow rate of 40 l min −1 that has been passed through a scrubber containing Airpel ® (Desotec Ltd, Roeselare, Belgium) activated carbon to remove VOCs. Prior to sample collection all TD tubes were conditioned for 40 min at 330°C using a TC-20 tube conditioner (Markes International Ltd, Llantrisant, UK). The TD tubes were stored in an airtight container at room temperature and used for sample collection within one hour of conditioning. The four-piece TD tube assembly was inserted into a clean mask for each study participant and then attached to the ReCIVA device ensuring that the TD tube and mask assembly were seated correctly within the device. The ReCIVA device was connected to the controlling computer.

Participants
Ethical approval was obtained (REC 14/LO/1136). In the first experiment, 20 patients undergoing upper gastrointestinal endoscopy at Imperial College Healthcare NHS Trust were recruited. In the second and third experiments, healthy volunteers were invited in order to be able to cope with the demands of high sample volumes and flow rates. Academic staff of the Department of Surgery and Cancer, Imperial College London participated in those experiments. All subjects were required to be non-smokers above the age of 18 and without a history of systemic or metabolic disease.

Sampling process
Prior to participation in the study volunteers were required to be fasted for a minimum of 4 h and rested for 15 min. Participants were asked to hold the ReCIVA device whilst the head strap was attached to ensure a seal is formed between the ReCIVA mask and face. On commencing exhaled breath collection, the participant was asked to perform normal tidal respiration whilst seated at rest. Standard collection parameters as specified by the manufacturers were used during exhaled breath sample collection with the ReCIVA device unless otherwise specified as part of the experimental protocol. Following sample collection the mask was disposed of and the TD tubes were capped and prepared for analysis in the VOC laboratory, Division of Surgery, St Mary's Hospital, Imperial College London. Prior to analysis TD tubes were stored in an airtight container at room temperature and all TD tubes were analysed within 6 h of breath sample collection.

Control for environmental contamination
Prior to collecting exhaled breath samples, reference samples were collected to assess VOC contamination from the exhaled breath collection system including CASPER air supply, and the ReCIVA mask and tubing. This was performed by connecting the ReCIVA device to a glass head (AMP3 Ltd, Aldershot, UK) and setting the device sampling parameters to 'always on' to collect a 250 ml gas sample at a flow rate of 200 ml min −1 onto a single TD tube (with three blank tubes in the ReCIVA TD assembly) (figure 1). A clean mask was used for each reference sample collection and comparison was made with a 250 ml room air sample simultaneously collected onto a TD tube using a hand pump (SKC Ltd, Dorset, UK) at a flow rate of 200 ml min −1 . Comparison was made between TD tube samples collected from the ReCIVA and CASPER attached to a glass head and TD tube samples collected simultaneously from the room air (with identical sample volume and flow rate for each TD tube). This process was repeated 10 times (with a new clean mask used on each occasion).

• Experiment 1: Fraction of breath sampled
Pump 'A' within the ReCIVA device was set to collect two TD tubes of whole breath (including mouth air) and pump 'B' was simultaneously set to collect two TD tubes of lower airways exhaled breath. Each participant provided a breath sample at a volume set to 250 ml per TD tube and a sample flow rate of 400 ml min −1 .
• Experiment 2: Breath sample volume Each participant was asked to provide four sequential exhaled breath samples with the sample volume per tube being set at 125, 250, 500 and 1000 ml. Pumps A and B of the ReCIVA device were set to collect whole breath air (including mouth air) onto four TD tubes with the flow rate being set at 400 ml min −1 .
• Experiment 3: Breath sample flow rate Each participant was asked to provide four sequential exhaled breath samples and the sample flow rate per tube was sequentially decreased (400, 200, 100 and 50 ml min −1 ). Pumps A and B of the ReCIVA device were set to collect whole breath (including mouth air) onto four TD tubes and the sample volume was fixed to 500 ml.
Analysis with gas chromatography mass spectrometry (GC-MS) Exhaled breath samples concentrated onto TD tubes were analysed using GC-MS. The TD tubes were desorbed using a Markes TD-100 TD unit (Markes International Ltd, Llantrisant, UK) using a two stage desorption programme, applying a constant flow of helium at 50 ml min −1 . In the primary desorption stage, TD tubes were dry-purged for 3 min and heated at 280°C for 10 min. In the secondary desorption  A two-sided non-parametric Mann-Whitney Utest or Wilcoxon signed-rank test with a Bonferroni correction was applied for analysis. For the sample volume and sample flow rate experiments Friedman's two-way analysis of variance was used with a post-hoc Wilcoxon signed-rank test to compare the minimum and maximum concentrations for each VOC. For all tests a two-sided p value of 0.05 was deemed to be significant. All statistical analysis was conduced on SPSS Statistics software (IBM, version 24.0).

Experiment 1: Fraction of breath sampled
Ten of the participants were male with a median age of 58.5 years and ten of the participants were female with a median age of 57 years. None of the target VOCs from the aldehydes, fatty acids and phenols groups or acetone were significantly different between lower airways expiratory breath and whole expiratory breath (table 1).

Experiment 2: Breath sample volume
The participants were seven males with a median age of 32.0 years and three females with a median age of 30.0 years. The median time to collect a 125, 250, 500 and 1000 ml volume breath sample was 60.5; 99.5; 173.0 and 332.5 s respectively. Sixteen VOCs belonging to the chemical classes of interest (acetone, aldehydes, fatty acids and phenols) were detected. The concentrations of propanal, acetone, propanoic acid, pentanoic acid, undecanal and dodecanal were elevated with increasing volumes (table 2). Contamination from the collecting system GC-MS analysis identified eight VOCs that were present in higher concentrations in the samples collected from the ReCIVA and CASPER collection system compared to the room air samples (table 4). Of the eight VOC contaminants associated with the breath collection system, cyclopentane was elevated in whole breath samples compared to lower airway breath samples and the remaining seven VOCs were unchanged (supplementary table 2). There were no significant changes observed across the sample volumes in the concentration of the previously listed eight VOC contaminants associated with the breath collection system (supplementary table 3). Six of the eight VOC contaminants associated with the breath collection system were changed in concentration with altering the flow rate per tube (supplementary table 4).

Discussion
This study indicates that the use of exhaled breathsampling devices requires the selection of optimum sampling parameters. The increase in sample volume has improved the levels of VOCs detected. However, the influence of the fraction of exhaled breath and the flow rate depends on the target VOCs measured. While the concentration of potential volatile biomarkers for oesophago-gastric cancer was not significantly different between the whole and lower airway exhaled breath, the level of other VOCs varied. Also, the recovery of some VOCs such as phenols and acetone from TD tubes was lower when breath sampling performed at higher flow rate but the majority of other VOCs were not affected.
It is clearly established that an important consideration in exhaled breath collection is contamination from the ambient air [30][31][32][33]. This issue has been approached by using a dedicated 'clean air supply' where the inspired air has been passed through a carbon based scrubber to minimise the impact of environmental contamination on the exhaled breath sample. This will minimise the effect of environmental contamination from volatile compounds with rapid wash-out rates in the body but will not be the case for volatiles with longer retention time and from longterm environmental exposure [34,35]. The effect of environmental exposure on endogenous VOCs should be considered despite using a dedicated 'clean air supply' [35]. In addition, the breath collection device itself can be a source of some contaminants. Eight VOCs were present in higher concentrations in breath samples collected from the sampling equipment alone compared to paired room air samples suggesting that these VOCs are contaminants from the sampling equipment. Five VOCs out of the eight detected belonged to the chemical class of siloxanes, and thus are most likely originating from silicone-based tubing and mask materials of the ReCIVA and CASPER assembly.
In order to collect the lower airway exhaled breath, the exhaled breath collection system utilises CO 2 and pressure sensors with the expectation that VOCs transported via blood would be at higher concentrations compared to whole breath samples [36][37][38]. The experiments showed no significant difference in acetone or potential volatile biomarkers for oesophagogastric cancer between whole and lower airway exhaled breath. Other VOCs significantly varied between fractions of exhaled breath suggesting an influence of the oral cavity on VOC production.
This study has demonstrated that VOC concentrations are dependent on the exhaled breath collection volumes. Within examined volumes, there was no threshold at which no further changes in VOC concentration were observed. Moreover, using higher collection volumes did not introduce greater environmental contamination into the sample from the breath collection system. Consequently, the largest possible collection volume should be used in clinical studies with careful consideration given to the clinical condition and the time that it will take to collect the samples.
In terms of the flow rates used in this study to pump exhaled breath samples onto TD tubes, the majority of the VOCs investigated showed no reproducible relationship observed between flow rate and VOC concentration. However, recoveries of acetone and phenol were lower when breath sampling was performed at higher flow rates. This may be explained by VOC breakthrough at high flow rates of these VOCs. It is clear that the lowest flow rates require the greatest length of time to collect the breath sample (1140.5 s for a flow rate of 50 ml min −1 compared to 169 s for a flow rate of 400 ml min −1 ). In addition six of the VOCs associated with contamination from the collection system were present in high concentrations at the lowest flow rate investigated. The selection of a midrange flow rate (e.g. 200 ml min −1 ) would enable optimum collection of VOCs whilst minimising sample collection time and the impact of environmental contamination.
The study has limitations. The experiments have focused on VOCs relevant to oesophago-gastric cancer and therefore researchers are encouraged to examine target VOCs for specific diseases in future studies. We did not assess the impact of impact of additional sampling parameters such as breathing pattern, body position, and oral versus nasal respiration in this study and these    factors need to be considered to ensure that optimal and reproducible exhaled breath samples are collected in future studies [39][40][41]. We chose TD-GC-MS as the analytical method in this study as it represents the reference standard of VOC analysis, including exhaled breath. The type of chromatographic column and TD sorbent employed were chosen to investigate a wide spectrum of different compounds. However there is no single sampling or analytical method capable of capturing the whole spectrum of VOCs in breath and this should be considered in study design.

Conclusion
When an exhaled breath collection system is employed there is a significant effect of the sampling parameters on the measured VOCs. Also, some contaminants are produced from the breath-sampling device itself. The largest sample volume is recommended given careful attention to the clinical condition and the practicalities of collection times during a busy clinical practice. The influence of the fraction of exhaled breath and the flow rate depends on the target VOCs measured. While in profiling studies it is acceptable to use a midrange flow rate and either whole or lower airway exhaled breath, it is important to examine the effect of sampling parameters on target VOCs before embarking on validation experiments or large-scale clinical studies. For instance, for our clinical studies on the use of VOCs for the diagnosis of gastro-oesophageal cancer, we investigated our target VOCs in this study and consequently we intend to use 500 mls sample volume of whole breath with a flow rate of 200 ml min −1 . Grant support Imperial College Healthcare Charity and Imperial Confidence in Concept (ICiC) scheme.

Disclosures
The authors declare no conflict of interest.