Impact of breath sample collection method and length of storage of breath samples in Tedlar bags on the level of selected volatiles assessed using gas chromatography-ion mobility spectrometry (GC-IMS)

The analysis of volatile organic compounds (VOCs) in exhaled air has attracted the interest of the scientific community because it provides the possibility of monitoring physiological and metabolic processes and non-invasive diagnostics of various diseases. However, this method remains underused in clinical practice as well as in research because of the lack of standardized procedures for the collection, storage and transport of breath samples, which would guarantee good reproducibility and comparability of results. The method of sampling, as well as the storage time of the breath samples in the polymer bags used for sample storage and transport, affect the composition and concentration of VOCs present in the breath samples. The aim of our study was to compare breath samples obtained using two methods with fully disposable equipment: a Haldane sampling tube intended for direct breath collection and breath samples exhaled into a transparent Tedlar bag. The second task was to monitor the stability of selected compounds of real breath samples stored in a Tedlar bag for 6 h. Gas chromatography coupled with ion mobility spectrometry (GC-IMS) implemented in the BreathSpec® device was used to analyse exhaled breath. Our results showed a significant difference in the signal intensity of some volatiles when taking a breath sample with a Haldane tube and a Tedlar bag. Due to its endogenous origin, acetone levels were significantly higher when the Haldane tube sampler was used while elevated levels of 2-propanol and unidentified VOC (designated as VOC 3) in the Tedlar bag samples likely originated from contamination of the Tedlar bags. The VOC stability study revealed compound-specific signal intensity changes of the selected VOCs with storage time in the Tedlar bags, with some volatiles showing increasing signal intensity during storage in Tedlar bags. This limits the use of Tedlar bags only for very limited time and carefully selected purpose. Our results highlight the importance of careful design and implementation of experiments and clinical protocols to obtain relevant and reliable results.


Introduction
Human breath is a complex mixture of gases consisting of nitrogen, oxygen, carbon dioxide, water vapour, inert gases and volatile organic compounds (VOCs).Analysis of VOCs in exhaled air is a very promising research area having great potential for diagnosis of diseases in non-invasive way (Das and Pal 2020).In addition to several other useful clinical applications, breath VOC analysis is used to detect lung (Handa et al 2014) and other cancers (Tyagi et al 2021), cystic fibrosis (Robroeks et al 2010), obstructive pulmonary disease ( Van Berkel et al 2010), diabetes (Das et al 2016), tuberculosis (Phillips et al 2007), nephropathy (Wang et al 2017), to predict heart and lung transplant rejection (Phillips et al 2004), and most recently to detect COVID-19 (Ruszkiewicz et al 2020).However, the lack of standardized procedures for collection, storage/transport and analysis of breath samples, has played a primary role in holding back progress and is one of the main reasons for the principal absence of breath tests in the clinical practice.(Beauchamp and Miekisch 2020).
The method of collection and storage of breath samples is partly determined by the analytical method and equipment available to the researcher (White and Fowler 2019).One of the newest methods of exhaled breath analysis is gas chromatography coupled with ion mobility spectrometry (GC-IMS) implemented in the BreathSpec ® device from G.A.S. (Dortmund, Germany).It combines high sensitivity (detection levels as low as ppb or even ppt), very fast analysis, and ease of use, offering a platform for a potential clinical application.This device allows online or pseudo-realtime sample analysis, where the patient exhales air directly into the device.However, offline techniques have gained much more popularity recently, especially during the emergence of the COVID-19 pandemic.The most commonly used offline sampling method applied in studies using BreathSpec is the use of a Haldane tube sampler.The breath sample is collected through a disposable breath sampling tube with an attached sampling syringe, the content of which is immediately injected into the device and analysed.This method requires the presence of the patient directly at the device.However, it is not always possible to bring a patient to a laboratory with a VOC analyser.In such cases, it is necessary to store and transport the breath sample to the laboratory (Ruszkiewicz et al 2022).Various types of containers can be used to collect and store breath samples, such as gas tight syringes, glass bulbs, passivated stainless steel canisters, and polymer collection bags.Syringes and glass bulbs are inexpensive, easy to use and clean, but they are also fragile with a limited volume.Stainless steel canisters are robust and provide optimal sample stability, but are relatively heavy, bulky and expensive.The most commonly used containers for taking breath samples are polymer sampling bags due to their inertness, material resistance, reusability and affordable price.Currently, Tedlar is one of the most popular and commonly accepted material for gas sampling in general and exhaled gas samples in particular (Mochalski et al 2009, White and Fowler 2019, Li et al 2021).
The sampling method and sample stability are the most important steps in the entire analytical procedure.The exhaled breath sampling procedure defines breath fractions (mixed expiratory, late expiratory, and end-tidal) to be analysed.When a Haldane sampling tube is used, the first exhaled portion of a tidal breath, which corresponds to the dead space (the volume of the air that does not take part in gas exchange), is omitted.The remaining part of breath used for analysis is composed of end-tidal or alveolar air.This portion of exhaled air contains a higher concentration of endogenous VOCs that originate from blood-gas exchange in the alveoli and thus more closely reflect metabolic conditions (Jia et al 2019, Santos et al 2021).When polymer bags are used, a mixed expiratory breath is usually collected.Mixed expiratory breath sampling encompasses the total exhaled breath which includes alveolar air and dead space air (filling the conducting zone of respiration made up of the mouth, nose, trachea, and bronchi) (Lawal et al 2017).In this method, the volatiles from the lung, mouth, nose, upper gastrointestinal tract, stomach and ambient air mix prior to sampling (Issitt et al 2022).In addition, there are some pitfalls when storing samples in polymer bags.Various phenomena, such as loss of compounds due to diffusion throughout the bag walls, penetration of volatiles from the external environment, adsorption of VOCs into the bag walls, release of contaminants from the polymer film, and chemical reactions between highly reactive sample components, can modify the original sample composition and subsequently lead to erroneous conclusions (Ghimenti et al 2015).The storage time in the sampling bag is therefore an important factor that can affect the composition and concentration of the analysed gases.Several authors have studied the stability of gas samples in Tedlar bags.In the studies of Steeghs et al (2007), Beauchamp et al (2008), Ghimenti et al (2015) and Mochalski et al (2013) a good stability of standard gas mixture compounds was demonstrated.However, there are currently only two studies analysing the stability of real breath samples stored in Tedlar bags, and the results are inconsistent.Steeghs et al (2007) found that most of the recoveries from breath compounds stored in black laminated Tedlar bags were more than 90% within 52 h and suggested the Tedlar bag as a reliable breath holder.In contrast, a study by Li et al (2021) showed that breath samples stored at ambient temperature can recover more than 80% of carbonyl compounds if the samples are processed within one hour, with a subsequent steep decline in recovery percentage.
Since the sampling method as well as storage time of breath samples in polymer bags influence the composition and concentrations of VOCs present in the breath samples, the aim of our study was (a) to find differences in the levels of selected VOCs in human breath samples taken with a Haldane sampling tube and in breath samples exhaled into a transparent Tedlar bag resulting from the collection of a different fraction of exhaled breath in these two methods, and to identify other potential factors leading to different results when using these two breath sampling methods.
(b) to monitor the stability of selected compounds of breath samples stored in transparent Tedlar bags for 6 h.Fast GC-IMS analysis enables results to be obtained in short intervals, allowing a thorough picture of changes in concentrations of selected VOCs to be obtained.
Both tasks in our study were intended to test the performance of two sampling methods in clinical application with an emphasis on their ease of use and high hygiene requirements after the Covid 19 pandemic.

Breath sampling using a disposable Haldane breath sampler and a Tedlar bag
Volunteers (N = 7, age = 35.7 ± 2.5 years, 5 female) provided two breath samples by exhaling through a disposable polypropylene Haldane breath sampler made from an Eppendorf tube with a oneway mouthpiece and a 4 mm hole drilled through the sidewall for a 5 cm 3 disposable polypropylene syringe (G.A.S. Dortmund, Germany).The subjects were asked to exhale slowly through the Haldane tube.In the late exhalation phase (excluding the first few seconds of exhalation), a sample of 5 cm 3 of their breath was drawn into the syringe.Subsequently, the syringe was removed from the sampler, and its content was injected into the GC-IMS device BreathSpec ® (G.A.S. Dortmund, Germany) for analysis.
To analyse the stability of breath samples stored in Tedlar bags, Tedlar bags with a mouthpiece (MediSense, Groningen, Netherlands) developed for the collection and storage of exhaled air samples were used.These bags are made from clear 2 mil (0.0508 mm) Tedlar ® film from DuPont ® .Tedlar is made of polyvinyl fluoride (PVF) film.Tedlar bags with a volume of 0.6 l (15.24 × 15.24 cm) were used in this study.After analysing two breath samples taken using a Haldane tube, the volunteers were asked to fill a Tedlar bag with a single exhaled breath.Subsequently, the septum located near the mouthpiece was punctured with a needle and 5 cm 3 of the sample was taken from the Tedlar bag with a syringe.The content of the syringe was then injected into the GC-IMS device for analysis.Breath samples were taken from the Tedlar bag immediately after filling (t = 0 min) and at times t = 15 min, 30 min, 45 min, 60 min, 75 min, 90 min, 105 min, 120 min, 150 min, 180 min, 240 min, 300 min and 360 min.A new Tedlar bag was used for each participant.Additionally, two bags were filled with laboratory air and two bags with nitrogen (produced by N 2 -Generator) to identify potential contamination originating from the Tedlar bags themselves or from the external environment.Changes in volatile intensity levels were monitored using the same protocol as for breath samples.
All breath samples were collected in the morning (before 9 AM) and samples from only one volunteer were analysed per day.On each measurement day, an ambient air sample was analysed for background VOC levels prior to breath sample measurement.
The study was approved by the Ethics Committee of the Jessenius Faculty of Medicine in Martin, Comenius University in Bratislava (decision of the Ethics Committee-EK 69/2022) and all volunteers gave written Informed Consent prior to providing a breath sample.The study protocol conforms to the ethical guidelines of the 1975 Declaration of Helsinki.

GC-IMS
In our study, the BreathSpec ® device (G.A.S. Gesellschaft für analytische Sensorsysteme mbH BioMedizinZentrumDortmund, Dortmund, Germany) was used for the analysis of exhaled breath samples.This device combines gas chromatography (GC) with ion mobility spectrometry (IMS).The double physical separation of volatile compounds together with the excellent sensitivity of IMS enables the detection and quantification of compounds with concentration levels down to ppb (parts per billion).A detailed description of the methods implemented in this device can be found elsewhere (Moura and Vassilenko 2022).Briefly, exhaled air samples are injected into the device using a stainless steel Luer-port adapter and then pass through a heated 6-way valve into the sample loop (T = 80 • C).The carrier gas (N 2 ) then transports the gas sample to the GC-column (MXT-WAX with a length of 30 m, an internal diameter of 0.53 mm and a thickness of 0.5 µm, T = 40 • C), where the isothermal separation of the components of the mixture takes place according to affinities of individual components to the surface or filling of the column.The separated volatiles are subsequently fed into the ionization part of the IMS (a radioactive β-emitter 3 H (100 MBq, 5.68 keV, T = 45 • C)).IMS separates ionized molecules based on the mobility of the molecules in the drift gas (we used N 2-produced by N 2 -Generator with nitrogen purity >99.999%) in a homogeneous electric field.The drift gas flow for the IMS is maintained by an electronic pressure control unit (EPC1) and the carrier gas flow for the GC is maintained by a second electronic pressure control unit (EPC2).In our study, we used the built-in program BREATH designed for the analysis of breath samples, with a total program duration of 10 min.The drift gas flow was constant at EPC1 = 200 ml min −1 and the carrier gas flow was constant at EPC2 = 5 ml min −1 during the first 30 s and then increased linearly up to 30 ml min −1 .
In order to minimize the contamination of the spectrometer, several procedures were applied: a built-in system cleaning protocol lasting at least 10 h the night before the measurement was run; drift gas flow and carrier gas flow was increased to their maximum values between measurements, blank samples using high-purity nitrogen were analysed at the beginning of each measurement day to test the cleanliness and performance of the device (no significant change in the instrument operation was observer over the course of the study).

Data analysis
Chromatograms of the breath and ambient air samples were analysed by VOCal 0.4.01 software (G.A.S., Dortmund, Germany).16 most significant spots were selected for subsequent analysis.Using VOCal's in-house GCxIMS database we were able to identify 12 spots/9 VOCs (ethanol M, ethanol D, methanol, acetone, 1-propanol M, 1-propanol D, 2-propanol M, 2-propanol D, 2-pentanone, acetonitrile, 1-hexanal, dimethyl sulfide; M-monomer, D-dimer).The four spots that could not be reliably identified were designated VOC 1M, VOC 1D, VOC 2 and VOC 3. Peak volumes of monomers and dimers were summed and analysed as a single VOC, so we analysed a total of 12 VOCs.For further analysis, signal intensities were calculated as volumes above the area minimum, where areas of interest were defined as rectangles bounding selected VOCs.Unfortunately, it was not possible to develop a calibration protocol to quantify the concentration of each VOC, so a semi-quantitative analysis was performed.
Before the experiments, the reproducibility of the breath sample measurement was tested using 6 replicates of breath samples taken by Haldane tube provided by three volunteers.Variability of signal intensity of selected VOCs was evaluated by the relative standard deviation (%RSD).
The Kruskal-Wallis test with the Conover-Inman post hoc test for pairwise comparisons when the Kruskal-Wallis test was significant was used to compare the signal intensities of selected VOCs of Haldane tube breath samples, Tedlar bag breath samples, and ambient laboratory air samples.A p value < 0.05 was considered as statistically significant.The statistical analysis was performed using SYSTAT software version 13 (Inpixon, USA).
The stability of the breath samples stored in the Tedlar bags was monitored as the percentage change in the signal intensities of selected VOCs at specified times compared to the signal intensities at the time of filling the Tedlar bag (t = 0): %c(A) t = (c(A) t-c(A) 0 ) /c(A) 0 (where c(A) t is the signal intensity of VOC A at time t and c(A) 0 is the signal intensity of VOC A at the time of the filling of the Tedlar bag t = 0).

Results
Representative chromatograms of laboratory air and breath sample from Haldane sampling tube and Tedlar bag at t = 0 chromatograms are shown in figure 1. Drift time (IMS) and retention time (GC measurement run [sec]) are presented in X-and Yaxis, respectively.Peak height enabling quantification of the compounds is illustrated by the colour scale.
The reproducibility test of the breath samples revealed that 9 of the 12 VOCs had a %RSD lower than 15% (see table 1).%RSD higher than 15% was found only for 1-propanol, 1-Hexanal and VOC 2.

Comparison of breath samples obtained using a Haldane sampling tube with a breath sample exhaled into a Tedlar bag
The impact of the sampling method (Haldane tube vs. Tedlar bag) on human breath composition was evaluated in the context of ambient laboratory air composition.We found no significant difference in any monitored VOC between the first and second Haldane tube samples (data not shown).Because participants spent more time in the laboratory before taking the second Haldane tube breath sample (20-30 min) and thus had more time for equilibration between the ambient air and the air in the lungs, we selected the second Haldane tube measurement for further analysis.
Box plots of 12 selected VOC signal intensities from ambient laboratory air, Haldane tube breaths, and samples taken from a Tedlar bag immediately after its filling are depicted in figure 2. The signal intensities of acetone, methanol, 1-propanol, 2pentanone and VOC 1 were significantly higher in both Haldene tube and Tedlar bag samples compared to ambient laboratory air.Significant differences between the Haldane tube samples and the Tedlar bag samples were found in acetone and 2propanol signal intensities.Acetone levels were higher when using a Haldane tube and 2-propanol when samples were taken from a freshly filled Tedlar bag.Differences, although insignificant, were also observed in 1-propanol and 2-pentanone with higher levels of 1-propanol in Tedlar bag samples and 2penatanone in Haldane tube samples.Volatile VOC 3 showed statistically significantly higher levels in exhaled samples than in ambient air, while being present in much higher levels in Tedlar bag samples.All other volatiles showed no differences between lab air, Haldane tube samples and Tedlar bag samples.

Stability of selected compounds of breath samples stored in a transparent Tedlar bag
The time courses of the percentage changes (%c(A) t ) of 12 selected VOCs stored in the Tedlar bags for 6 h are depicted in figure 3. Acetone levels increased  by up to 15.5 ± 6.9% within 6 h, with a sharp increase at the beginning of the experiment (in the first 45 min) followed by a slight decrease and relatively stable values until 120 min, followed by a gradual moderate increase until the end of the experiment.Methanol levels also increased rapidly by approx.20% from their initial values within the first 30 min and remained stable for up to 150 min.After that, the methanol signal intensities dropped significantly to values −27 ± 13.5% lower than the initial values.Ethanol showed relatively good stability of ±10% over 150 min of the experiment.After this time, the signal intensities of ethanol tended to increase and showed high intra-individual and interindividual variability.In propanol, we noticed a significant increase in signal intensity for both isomers (1propanol, 2-propanol) during the experiment.The signal intensities of the 1-propanol increased by up to about 50% during the first 60 min, followed by a slight decrease and subsequent return to levels about 50% higher than the initial values.1-propanol showed high intra-and inter-individual variability.A more pronounced increase in signal intensities was observed in 2-propanol.Its levels increased by more than 120% of the original values at the end of the experiment, with a gradual increase consistent in all individual samples.The levels of 2-pentanone also intensified with an average increase of 22% during the first 90 min and 50% thereafter.Acetonitrile signal intensity decreased by up to −25% during the first 120 min of the experiment with a subsequent increase.The levels of this VOC showed high intraand inter-individual variability throughout the experiment.1-Hexanal was shown to be relatively stable for the first 60 min but then showed unpredictable changes with a maximum increase around 90-120 min (up to 220% in one case) followed by a return to original values.Dimethyl sulfide remained stable in the Tedlar bag with a small increase (approx.8.5%) during the first 150 min, followed by a sudden decrease of −17 ± 6.5%.The unidentified volatile VOC 1 showed similar signal intensity changes to dimethyl sulfide and methanol.During the first 150 min its intensity rose approx.by 85 ± 43% above its initial value, followed by a decrease to approx.−62 ± 13% below its original levels.The levels of unidentified VOC 2 showed moderate increase within first 30 min of approximately 22 ± 13% with subsequent stabilisation and slight decrease after 150 min of the experiment.The most extreme increase in signal intensities was observed for VOC 3. Its values reached 300 ± 160% of their initial values after 6 h in a Tedlar bag.A gradual increase in signal intensity was recorded in all individual samples, but with high inter-individual variability.

Contamination tests
Galery plots of volatiles changing signal intensity during storage of laboratory air and nitrogen in Tedlar  experiments and its signal intensity increased significantly over time.

Discussion
Our research was focused on two tasks related to the collection method of human breath samples and to the stability of human breath stored in Tedlar bags.The main findings of our study are as follows: • The signal intensity of three of the monitored VOCs differ significantly when using the Haldane tube and Tedlar bag sampling method.Acetone levels were significantly higher when the Haldane tube sampler was used and 2-propanol and VOC 3 levels were higher when samples taken to the Tedlar bag were analysed.A breath sample measurement reproducibility test performed before the experiments revealed good reproducibility for most of the selected VOCs, thus confirming the applicability of GC-IMS for these types of tasks.Only 1-propanol, 1-hexanal, and VOC 2 showed higher variability, which may explain the higher inter-and intra-individual variability of these VOCs observed in our experiments.

Comparison of breath samples obtained using a Haldane sampling tube with a breath sample exhaled into a Tedlar bag
To the best of our knowledge, this is the first study comparing breath sampling methods using the Haldane sampling tube and the Tedlar bag.The difference between these two methods is in the portion of breath taken for the analysis.When a Haldane sampling tube is used the endtidal or alveolar air is used for the analysis and exhaled air contains a higher concentration of endogenous VOCs that originate from blood-gas exchange in the alveoli (Jia et al 2019).Acetone and 2pentanone are such volatile substances.Acetone is the most abundant metabolite in exhaled air.Its production is believed to originate from two pathways: decarboxylation of acetaldehyde (lipolysis or amino acid degradation) and the dehydrogenation of 2-propanol (by liver alcohol dehydrogenase) (Das et al 2016).A potential origin of 2-pentanone is considered to be the β-oxidation of fatty acids and gut microflora (Walker and Mills 2014).Several studies have shown that acetone and 2-pentanone are systemic and are not generated to a significant extent in the oral cavity.Wang et al found that acetone levels were equal in mouth-exhaled and nose-exhaled samples and were lower in the oral cavity (nonbreathed) and therefore concluded that acetone was entirely systemic (Wang et al 2008).Similarly, in the study of van den Velde et al (2007), levels of acetone and 2-pentanone were found to be significantly higher in alveolar air than in mouth air.Our results are consistent with these studies showing higher levels of acetone and 2-pentanone in exhaled breath when Haldane tubes were used compared to Tedlar bags, and confirm that Haldane tube sampling targets more alveolar air.
Breath sampling with the use of Tedlar bags usually does not differentiate particular breath fractions, so mixed expiratory air from all phases of exhalation is analysed.Of the volatiles selected for analysis in our study and reliably identified, 2-propanol and, to a lesser extent, 1-propanol show higher signal intensity in air exhaled into a Tedlar bag compared to air exhaled through a Haldane tube, suggesting their oral or exogenous origin.The oral cavity origin of these compounds was demonstrated in the studies by Wang et al (2008), Miekisch et al (2008) and van den Velde et al (2007).Wang et al (2008) found that propanol (both isomers) was present in the mouth breath and in the cavity air in approximately similar amounts, while nose breath levels were much lower.Miekisch et al (2008) showed that the ratio of alveolar and mixed concentrations of 2-propanol (C alv /C mixed ) from different sampling methods was <1, confirming its exogenous origin.A significant difference between alveolar and mouth air of 1-propanol was found in the study of van den Velde et al (2007).Van den Velde believes that the higher gradient of 1-propanol in the mouth is the result of bacterial fermentation of threonine in the mouth.In our study, the 2-propanol signal intensity levels of breath samples exhaled into Tedlar bags may be confounded by contamination from the Tedlar bags.The higher levels of 2-propanol in nitrogen-filled bags compared to laboratory air and its increasing signal intensity with storage time in both nitrogen and laboratory air tests indicate that 2propanol may be a contaminant of the Tedlar bags.
The origins of ethanol and methanol are not well understood.However, these VOCs are believed to originate from microbial fermentation of carbohydrates in the gastrointestinal tract (Das et al 2016).Our study revealed no significant difference in the levels of methanol and ethanol between Haldane tube and Tedlar bag breath samples.The signal intensity of methanol in ambient laboratory air was significantly lower than in exhaled breath samples, so this exogenous methanol cannot significantly bias its levels in breath, implying that methanol is of endogenous origin.In contrast, in some cases ethanol shows lower values in breath compared to ambient air in our study and in some cases vice versa.At the same time, we noted a large variability in the levels of ethanol in our laboratory air on different days, probably due to the use of disinfectants or different weather conditions.This variable background contamination makes it impossible to identify the origin of ethanol in our study.
Other monitored volatiles acetonitrile, 1-hexanal and dimethyl sulfide showed no significant differences in the signal intensities between breath samples taken by Haldane tube, Tedlar bag and ambient laboratory air.
Unfortunately, we were not able to identify the volatiles VOC 1, VOC 2, and VOC 3 using the internal GCxIMS VOCal database.Based on our results we can conclude that volatile VOC 1 is of endogenous origin, since its signal intensity is significantly higher in the samples of exhaled breath compared to ambient lab air.This suggests that VOC 1 may be isoprene, which is the second most abundant endogenous VOC in exhaled breath (Sukul et al 2023).However, we cannot confirm this in our study, as isoprene is not found in the current GC-IMS database.VOC 2 and VOC 3 are probably contaminants of Tedlar bags.The VOC 2 signal intensity shows slightly (statistically insignificant) higher values and VOC 3 significantly higher values in the air exhaled into the Tedlar bag compared to air exhaled through the Haldane tube.

Stability of breath samples stored in a transparent Tedlar bag
The most commonly used polymer bags designed for the collection, transport and storage of breath samples for the analysis of breath composition are Tedlar bags.Knowing the recovery percentages of the volatiles of interest stored in Tedlar bags over time is critical for the planning of the experiment and proper use of Tedlar bags in clinical practice.
The stability of VOCs in Tedlar bags has been studied by several authors.Steeghs et al (2007) showed good stability of standard gas mixture compounds; losses ranging from ∼10% for isoprene to ∼25% for styrene over 2 d of storage in a black laminated Tedlar bag.In Deng's study (2004) focusing exclusively on acetone, acetone was found to be stable in a standard Tedlar bag for 4 h, showing a significant decline after 6 h.A more detailed study on the ability of Tedlar bags to store gaseous compounds was carried out by Beauchamp et al (2008).They performed an experiment in which the concentration of 12 VOCs representing different compound classes (alcohols, nitriles, aldehydes, ketones, terpenes and aromatics) in the ppb range was monitored over a 70 h storage period.The concentrations of all compounds decreased with time, with compound-specific decay rates.After storage for 10 h, losses were less than 20% for all the analytes investigated, and it was advised that a sample should be analysed within 10 h of sample collection to provide acceptable recovery.Mochalski et al (2013) compared the storage capabilities of three types of polymer sampling bags (Tedlar, Kynar, and Flexfilm).Their findings yield evidence of the superiority of Tedlar bags over the remaining polymers in terms of background emission, species stability, and reusability.Based on their results, they recommend analysing samples within 6 h.A similar study comparing three types of bags (Nalophan, Tedlar and Cali-5-Bond) was performed by Ghimenti et al (2015).For the Tedlar bag, within 6 h, acetone, 2-propanol and hexanal showed a 20% loss.In this study, Nalophan bags were found to be the most suitable in terms of contaminants, good sample stability and costs.
All of the above studies used standardized gas mixtures with a limited number of species and reported relatively good stability of compounds stored in Tedlar bags for at least 6 h.There are currently only two studies on the stability of real breath components stored in Tedlar bags, with very different results.In the study of Steeghs et al (2007) the recoveries of sampled gas concentrations over time were found to be more than 90% 52 h after filling.Of the tentative identified compounds, only methanol and ethanol were found to decrease significantly, and the authors suggested Tedlar bags as suitable for human breath storage.In contrast, the results of the study by Li et al (2021) indicate a much shorter reliable storage time for holding human breath in Tedlar bags.In their study, the breath samples stored at ambient temperature recovered ∼80% of carbonyl compounds if the samples were processed within one hour.A decrease in recovery percentages followed.The results of this study also show that the degradation of the breath compounds was reduced and the appropriate storage time was prolonged to 2 h if the breath samples were stored at cold temperatures (4 • C).
In our study, we found compound-specific signal intensity changes in relation to the storage time of the breath sample in Tedlar bags.In contrast to previously published studies, where the authors describe the relative stability or loss of compounds stored in the Tedlar bag with time, we found increasing signal intensity of some volatiles with time of storage in the Tedlar bag.Such behavior is observed in acetone, 1propanol, 2-propanol and 2-pentanone in our study.We also noticed an initial increase in the signal intensity of methanol, dimethyl sulfide, and VOC 1 followed by a sudden decrease after 150 min of the experiment.Ethanol and acetonitrile showed relatively good stability over time, and 1-hexanal was stable for approximately 75 min with unpredictable signal intensity changes after that.The largest increases in signal intensity with storage time in Tedlar bags were recorded for unidentified volatile VOC 3. Our results are thus in line with Kevin Lamote's observations published in the Beauchamp's task force poll (Beauchamp 2015), where Lamote reports that 'certain VOCs can increase in concentration when multiple subsequent samples are taken from the same Tedlar bag' (no study has been published).
In this phase of our experiments, we aimed to describe the stability of breath volatiles in Tedlar bags, with no attempt to explain the causes of changes in the signal intensity of breath compounds.Further we will only speculate on possible explanations of our findings.
The penetration of volatiles from the external environment into the bag was not the reason for the increasing signal intensity of volatiles in the bag, since the concentrations of any VOCs were not significantly higher in the laboratory air.It has been suggested that there are hundreds, perhaps over a thousand of volatiles in the exhaled breath (Smith et al 2007), and the increase in the concentration of some compounds may be related to the reactive chemistry of a stored real breath sample.For example, the reaction between NO 2 and water leads to the formation of light-sensitive nitrous acid (HNO 2 ).When exposed to light, this compound undergoes photodissociation and produces hydroxyl radicals, which may oxidize compounds in the breath samples (Beauchamp et al 2008).
Another important consideration with bag sampling of the human breath is the humidity of the gas sample.Water vapour is a significant component in breath gas and can affect sample stability in bags (Beauchamp and Miekisch 2020).Water vapour from the breath can diffuse very quickly through the Tedlar bag (films) (Steeghs et al 2007, Beauchamp et al 2008), which can lead to the loss of chemical reactants and cause the concentrations of some volatiles to stabilize or even decrease.The sudden decrease in the signal intensity of methanol and two other volatiles at 150 min of the experiment can probably be attributed to the loss of humidity in the sample.It has been documented that methanol interacts with polyvinyl film via hydrogen bonds and easily and irreversibly adsorbs on the PVF bag material.This adsorption process works more efficiently under dry conditions than humid, and therefore can be accelerated by the loss of water vapour (McGarvey and Shorten 2000).Similar mechanisms could be at work for dimethyl sulfide and VOC 1.
A potential source of signal intensity distortion is also contamination from the Tedlar bags themselves.The most frequently reported Tedlar bag background contaminants are N,N-dimethyl acetamide, phenol and acetic acid.These pollutants are known to be present in the Tedlar manufacturing process (Trabue et al 2006).We did not capture these compounds in our study because the 10 min GC-IMS program we used, was too short for their detection due to their longer retention times.Other authors have reported several other contaminating compounds originating from the Tedlar bags.Mochalski et al (2013) found 9 Tedlar bag contaminants (three hydrocarbons (n-hexane, 2,4-dimethylheptane and 4 methyl octane), two volatile sulphur compounds (COS and CS 2 ), N,N-dimethylacetamide, phenol, acetonitrile and 1-methoxy-2-propyl acetate).Among the quantified species the highest levels were noted for acetonitrile (19 ppb).The results of our study do not confirm acetonitrile as a contaminant of Tedlar bags.The signal intensity levels of acetonitrile in the Tedlar bags were not significantly different compared to laboratory air and remained stable during the stability experiment.Even more Tedlar bag pollutants were identified by Hayes et al (2006) They reported 13 VOCs emitted very shortly (20 min) from bag material (ethanol, carbon disulfide, acetone, 2propanol, methylene chloride, 2-butanone, toluene, m,p-xylene, styrene, naphthalene, phenol, hydrocarbon C9, hydrocarbon C14).Furthermore, the recoveries of ethanol, carbon disulfide, acetone, 2propanol, and methylene chloride were found to be highly variable in their study, indicating inconsistent background contributions from each bag.The results of our study are partially consistent with this study.Our contamination tests revealed acetone, 2propanol, VOC 2 and VOC 3 as potential contaminants of Tedlar bags.Acetone showed a tendency to be released from the Tedlar bag after approximately 120 min, so longer storage may affect acetone levels.2-Propanol and VOC 3 were present in the Tedlar bags very soon after the bags were filled with nitrogen, and their signal intensity significantly increased with storage time in all experiments.2-Propanol appears to be a significant pollutant of the Tedlar bags used in our study.This limits its use as a potential lung cancer biomarker, as proposed by Koureas et al (2020) and Buszewski et al (2011), without strict control of sampling and measurement conditions.VOC 2 and VOC 3 could not be reliably determined, and therefore, we recommend that these volatiles be excluded from further analyses until they are reliably identified.No cleaning procedure was applied to the Tedlar bags before the experiments to simulate their practical use in clinical practice.Contaminants can therefore come from the manufacturing process.
The level of recovery from bags can also be affected by the volume/surface area ratio, chemical properties of the sample (molecular weight of the compounds), bag size, and film thickness of the bag walls (Mochalski et al 2013, Ghimenti et al 2015).In contrast to other studies that investigated the stability of real breath compounds in Tedlar bags (Steeghs et al 2007, Li et al 2021), we used smaller bags (0.6 L) filled with single breaths, whereas in these studies, the authors used 1 L and 5 L bags for multiple breaths.Moreover, Steegs et al used blackened bags (designed to prevent the degradation of photolabile species) with lower recovery rates (Sulyok et al 2001).These differences in Tedlar bags may also contribute to some degree of inconsistency in the results.

Study limitations
Concerns regarding the reproducibility of Haldane tube sampling may arise from the lack of monitoring breathing cycle phases.The quality of sampling may be enhanced by monitoring expired CO 2 (capnography) (Miekisch et al 2008).Currently, there are only three commercially available devices enabling sampling with the option of targeting the breath fraction: ReCIVA, Bio-VOC and Mistral.Very recently, Santos et al (2021) developed a novel instrumentation for selective breath sampling, enabling the precise collection of a predetermined portion of exhaled air using the AI (Machine Learning) algorithm.Similarly, there are attempts to sample end-tidal part of the breath into the polymer bags.A practice of exhaling the first part of the breath into the room, followed by exhalation of the remaining part into the bag, is applied.This, however, could reduce the reproducibility of the sampling techniques (Ruszkiewicz et al 2022).Researchers also use their own sampling prototypes, e.g.Miekisch et al (2008) or Mochalski et al (2013b) on CO 2 -controlled end-exhaled breath sampling into the Tedlar bags.Although the use of these devices can increase the accuracy of the sampling process, they can also contaminate the breath gas samples by emissions or absorption of volatiles from the sampling interface materials.(Pham et al 2023) Other sampling-specific parameters such as exhaling through the Haldane tube and filling the sampling bag with high or low rate of exhalation, in different volumes of exhalation, under different surrounding air conditions (temperature, humidity), and volunteers' diet also significantly affect the concentrations of exhaled VOCs (Thekedar et al 2011).These parameters and conditions were not controlled in our study.
It should also be noted that propanol and ethanol are common ingredients in disinfectants used in laboratory settings and therefore may be present in relatively high concentrations in the laboratory environment and may contaminate the measurement to some extent.
New Tedlar bags were used in our study without any cleaning procedure.The goal was to test the usability of Tedlar bags while maintaining their ease of use for clinical application.As we detected some contaminants such as acetone, 2-propanol, VOC 2 and VOC 3, the application of a cleaning procedure may be beneficial.However, no standard procedure has been established for the cleaning process, yet.Some authors use nitrogen, while others use argon to clean the bags, and no consensus is defined on whether bags should be kept heated during the cleaning process and on the number of times the bags should be cleaned.(Marzorati et al 2019).In addition, mechanical handling (filling and emptying) of sampling bags also leads to their ageing and causes damage due to mechanical stress, which modifies the structure of the polymer film (Mochalski et al 2009).As such, determining which cleaning technique is best is a challenging task.
Our results are limited to only one type of Tedlar bags-transparent with a polypropylene mouthpiece; volume 0.6 l (MediSense, Groningen, Netherlands) and polypropylene Haldane breath sampler made from an Eppendorf tube (G.A.S. Dortmund, Germany).
A limitation of our study is also its semiquantitative nature.However, quantitative analysis using GC-IMS is complicated by the non-linear dynamic range of spectrometers and are quite rare.Around the lower concentrations, the concentrationsignal relationship is more linear, whereas the signal becomes more plateaued as the concentration increases (manufacturer recommends Boltzman-type fit).This effect has been reported for many compounds (Zhu et al 2021).A precise quantitative analysis that would monitor concentrations of the selected volatiles, would require obtaining calibration curves of all the selected compounds, which was not possible in our study.Given the low concentrations of volatiles in human breath (ppb level), we do not expect significantly different results from our study using calibrated concentrations.However, this needs to be verified by further studies.

Conclusion
The chemical characterization of volatile compounds in human breath has the potential to provide a snapshot of the physiological or pathological processes taking place in the human body in a non-invasive way.The most critical steps in the entire analytical procedure are the selection of the sampling method and the stability of the samples in the storage containers (if they need to be used).
Our results confirm that the sampling method can significantly impact the results of breath analysis.
We found a significant difference in the signal intensity of some volatiles when taking a breath sample using a Haldane tube and a Tedlar bag, which can be attributed to the difference in the part of the breath taken for analysis (acetone) and the contamination of the Tedlar bags (2-propanol, VOC 3).
If the storage and/or transport of the breath samples is required the stability of a compound's concentrations over the time frame from sampling to the time of analysis is a critical requirement.Tedlar bags are the most frequently used breath storage containers.However, our study revealed compound-specific changes in signal intensity of selected VOCs with storage time in Tedlar bags.This limits the use of Tedlar bags only for very limited time (ideally within minutes, but certainly less than a few hours) and carefully selected purpose.Therefore, experiments and clinical protocols must be planned very carefully, and the time point of the analysis should be fixed and kept as close as possible to the time of sampling.

Figure 1 .
Figure 1.Representative chromatograms of laboratory air (left) and breath samples taken using a Haldane sampling tube (center) and Tedlar bag at t = 0 (right).

Figure 2 .
Figure 2. Box plots of signal intensities of selected VOCs from laboratory air and from breath samples taken using a Haldane tube and Tedlar bag at t = 0; the p values of the Kruskal-Wallis test and Conover-Inman post hoc test for pairwise comparisons are shown in the figure; a p value < 0.05 was considered statistically significant.

Figure 3 .
Figure 3.Time course of selected VOC signal intensities at defined times of experiment normalized to signal intensities at t = 0; the circle represents the median of 7 measurements and the error bars indicate the interquartile range.

Table 1 .
%RSD (mean ± SD) of signal intensity of selected VOCs calculated from 6 replicates of breath samples from three volunteers.