Systematic comparison of methods for offline breath sampling

Abstract

Harnessing the potential of exhaled breath analysis is an emerging frontier in medical diagnostics, given breath is a rich source of volatile organic compound (VOC) biomarkers for different medical conditions. A current downfall in this field, however, is the lack of standardized and widely available methods for offline sampling of exhaled VOCs. Herein, strides are taken toward the standardization of breath sampling in Tedlar bags by exploring several factors that can impact VOC heterogeneity, including tubing material, chemical composition of collection bags, breath fractionation, exhalation volume, and transfer flow rate. After bag-based sampling standardization, performance was benchmarked using two offline breath sampling methods, Tedlar bags and the Respiration Collector for In Vitro Analysis (ReCIVA). Three volunteers from the laboratory with no known respiratory diseases donated ≥ n = 5 samples collected onto adsorption tubes via each method, which were analyzed through thermal desorption (TD) coupled with gas chromatography-mass spectrometry (GC-MS). Data processing revealed a set of 15 highly reliable on-breath VOCs detected across volunteers, and most analytes (except indole) demonstrated higher sensitivity using Tedlar bags. Calculating relative standard deviation (RSD) values showed Tedlar bags were also significantly more reproducible compared to the ReCIVA (p < 0.03). Agreement between the two methods was demonstrated through correlating VOC signals with high statistical significance (R² = 0.70), indicating both devices are well situated for biomarker discovery applications.

Keywords: Exhaled breath; Gas chromatography-mass spectrometry (GC–MS); Method standardization; Thermal desorption; Volatile organic compounds (VOCs).

Fig. 1

A Chromatographic traces of ultra-high pure nitrogen samples collected in Tedlar bags and transferred using tubing of different chemical compositions show that Tygon tubing has a significantly elevated background. B Bar plots of 2,4-di-tert-butylphenol in the same samples show elevated levels in Tygon and PTFE tubing. C Bar plots show the signal of on-breath VOCs when transferred using the different tubing types, displaying no significant differences

Fig. 2

A GC–MS chromatograms of ultra-high pure nitrogen samples collected in Multi-Layer Foil and Tedlar bags, along with B bar plots illustrating the number of deconvoluted VOCs by sample type, show that Tedlar bags have a significantly reduced background and therefore are optimal for breath collection/analysis

Fig. 3

Bar plots for individual VOCs including A acetone and isoprene, along with B eucalyptol, pinene isomers, and other compounds, show fractionating lower airway breath is more sensitive relative to whole breath. C Box/whisker and D scatter plots displaying RSD values for untargeted VOCs show a higher degree of reproducibility in fractionated samples

Fig. 4

A First day of experiments regarding optimization of exhalation volume, showing the signals of on-breath VOCs are positively correlated with volume and 1.5 L is the most optimal condition. B The second day of experiments extends analyses to larger volumes, which showed further increases in sensitivity were observed

Fig. 5

Scatter plots showing A number of deconvoluted VOCs and B total integrated GC–MS signals as a function of transfer flow rate from the Tedlar bag indicate no correlation or impact. Scatter plots are also produced for C acetone, isoprene, limonene, D toluene, eucalyptol, and pinene, further illustrating transfer flow rate does not have a quantitative impact on exhaled breath VOC signals

Fig. 6

Hierarchical heatmap of the select 15 VOCs in background samples and breath samples collected from three different volunteers shows a more sensitive analysis through use of Tedlar bags relative to the ReCIVA

Fig. 7

Log₂ fold change (FC) values for all 15 VOCs across three volunteers, showing the majority of analytes are upregulated in Tedlar bags

Fig. 8

A Box/whisker plots for each method and volunteer regarding 12 VOCs for which RSD could be calculated. B Scatter plots show the average RSD for both methods across all three volunteers. Taken as a whole, they show that breath-based VOC analysis is significantly more reproducible using Tedlar bags

Fig. 9

Principal component analyses utilizing data from 15 VOCs which was autoscaled by A day of analysis to visualize methodological differences and B breath sampling method to determine concordance/agreement

Fig. 10

Scatter plot showing a statistically significant correlation in average log₂ VOC signals between ReCIVA and Tedlar bags across all three volunteers