The effect of bronchoconstriction on exhaled breathprints was examined in a cross-over study performed on asthma patients (n = 10) attending two visits with a mean time between visits of 7.5 days (range 4–14 days) at a similar time of the day (± 2 h). At Visit 1, methacholine (MCh) provocation and at Visit 2, a sham challenge with isotonic saline was performed. At Visit 1 exhaled breath was collected before MCh provocation (baseline), when at least a 20% drop in FEV₁ was achieved (post-methacholine) and after salbutamol inhalation (post-salbutamol). At Visit 2, exhaled air was collected before sham challenge (baseline), after the last inhalation of saline (post-saline) and after salbutamol inhalation (post-salbutamol). Subjects were not blinded to the procedures and visits were not randomized because of the need to match the number of inhaled doses between the visits.

The effect of methodological drifts in eNose signals was investigated in a control study performed in 10 volunteers (seven healthy, three asthmatics: see below). Breath sampling was performed three times following the same course as during the challenge procedures in the main study but without any intervention (0 min, 60 min and 90 min). Subjects did not eat and drank only water during and 3 hours prior to that period.

Ten adult patients with previously diagnosed asthma agreed to participate in the main study. The patients were never-smokers and had episodic chest tightness or wheezing with a pre-bronchodilator FEV₁ >65% predicted and documented airway responsiveness (PC₂₀ methacholine <8 mg/mL) or reversibility in FEV₁ predicted >12% after 400 μg inhaled salbutamol as established within 12 months prior to the study. Subjects with concurrent pulmonary disorders, diabetes mellitus, hypo- or hyper-thyroidism, severe cardiovascular disease, renal insufficiency, present cancer or cancer in the past 5 years, oral corticosteroid use, present parodontitis or recent dental treatment were excluded. Patients on inhaled medications other than short-acting or long-acting β₂-agonists and/or inhaled steroids or those who had a history of upper or lower respiratory tract infection in the four weeks before the measurements were excluded from the study.

Ten volunteers, including seven healthy non-smoking subjects with no previous history of airway diseases or other chronic diseases, and three asthmatics but otherwise healthy patients, without an upper or lower respiratory tract infection in the four weeks before the measurements, were recruited to participate in the control study for eNose drift analysis. The protocol was approved by the local medical ethics committee. Written informed consent was obtained from all subjects.

MCh challenge test was performed according to the 2-min tidal breathing method [23]. The patients inhaled methacholine with tidal breathing for 2 min, and FEV₁ was recorded 30 s and 90 s after the exposure (MasterscreenPneumo; Jaeger; Würzburg, Germany). Doubling doses of methacholine bromide ranging from 0.04 to 19.6 mg/mL were applied in 5-min intervals until PC₂₀ was achieved. PC₂₀ was calculated with linear interpolation. Subsequently, the patients inhaled 400 μg salbutamol per metered dose inhaler with a spacer, and after 10 min FEV₁ was measured. FEV₁ was considered to be restored if higher than 90% of baseline.

Sham challenge with 0.9% isotonic saline solution (154 mM NaCl) was performed in identical fashion with identical numbers of aerosol inhalations and spirometric maneuvers as performed when achieving PC₂₀.

Patients withheld long-acting β2 agonists and antihistamines for 48 hours, and short-acting β2 agonists and inhaled corticosteroids >8 hours before both challenge tests.

To reduce possible confounding effects, patients were asked not to eat and drink anything but water 3 h prior to breath collection and refrain from caffeine-containing beverages and peppermint exposure for 6 hours before the visits. Breath collection and sampling were performed using the 5-min tidal washin method [12]. Briefly, patients inhaled VOC-filtered air (A2, North Safety, Middelburg, Netherlands) and exhaled via a silica reservoir in a 2-way non-rebreathing valve (Hans Rudolph 2700, Hans Rudolph, Kansas City, MI, US). Following 5 min of equilibration with tidal breathing, patients performed a maximal inspiratory capacity maneuver and the full expiratory capacity volume was collected into a 10-L Tedlar bag. Bags were sampled within ten minutes followed by the parallel sampling of another Tedlar bag containing VOC-filtered room air as a reference.

Exhaled breath samples were analyzed at room temperature by the same handheld electronic nose (Cyranose 320; Smiths Detection, Pasadena, CA, US) with an array of 32 carbon black polymer sensors [8,9]. VOCs binding to a sensor cause a change in the electrical resistance of the sensor. The raw data of a breathprint compose of 32 values each corresponding to the relative change in electrical resistance of a sensor. According to the manufacturer’s instruction, the first measurement was disregarded at each session (first sniff effect).

Sensor data of the eNose were processed through Savitzky-Golay filtering and baseline correction. Offline analysis of raw data was performed using SPSS software (version 16.0). Principle component analysis was used to redistribute the variance of the original 32 sensors into a set of factors (four factors captured 95.6% of the variance within the data set in the main study and 97.3% of the variance at the control visit). The two factors (denoted as Factor 2 and Factor 3) which showed a significant response to methacholine and/or salbutamol inhalation were selected for further analysis and are being referred to as “the breathprint”.

A mixed model analysis followed by pairwise comparisons on factors was used to evaluate any change in breathprints when recorded repetitively during the visits. To compare the changes (deltas) in breathprints caused by nebulisation with methacholine and isotonic saline, corresponding baseline factors were subtracted from post-methacholine or post-saline factors, and these changes were compared with paired t-tests for each factor. To assess between-day variability of a breathprint, baseline factors at visits were compared with paired t-tests.

There are no previous data for calculation of the statistical power of studies with serial eNose measurements. However, previous parallel studies have shown adequate power at similar samples sizes [11,12]. Therefore, the current sample size was considered to be adequate for within-subject analysis.

FEV₁ values recorded within a visit were compared with mixed model linear followed by pairwise comparison, the percentage change in FEV₁ at the visits and baseline FEV₁ values between visits were analyzed with paired t-tests. Data were normally distributed (Kolmogorov-Smirnov test) and are expressed as mean ± SEM in the figures and as mean ± SD in the table and in the text. The level of significance was considered as p < 0.05.

Patient characteristics and baseline lung function parameters are presented in Table 1. The population was well-characterized, but we acknowledge that the number of subjects in this study was limited. Even though data for formal power calculations of serial eNose measurements are lacking, similar numbers of patients were successfully used in previous eNose studies [11,12] as well as in former studies on the effects of airway caliber on exhaled NO [18,20] based on adequate power calculations.

To investigate the effect of airway caliber on exhaled breathprints, we chose a controlled, cross-over challenge model with methacholine as well as sham (saline) provocation. This allowed the distinction in outcomes as produced by methacholine and inhaled aerosols or the procedure as such.

As expected, methacholine inhalation induced significant bronchoconstriction (baseline FEV₁ 3.33 ± 0.63 L vs. 2.30 ± 0.57 L post-methacholine, p < 0.001), and subsequently, airway caliber was restored by salbutamol inhalation [post-salbutamol FEV₁ 3.37 ± 0.72 L, p < 0.001, Figure 1(a)]. Breathprints, as demonstrated by Factor 2, were significantly changed after methacholine (p = 0.04) and also post-salbutamol (p = 0.006) when compared to the breathprints at baseline; however, the breathprints after salbutamol were not significantly different from the ones after methacholine [p = 0.34, Figure 1(b)].

Inhalation of isotonic saline did not change FEV₁ (baseline 3.40 ± 0.72 L vs. post-saline FEV₁ 3.27 ± 0.65 L, p = 0.13), but FEV₁ increased to 3.63 ± 0.73 L after salbutamol inhalation [post-salbutamol vs. baseline and post-saline: p = 0.008 and p < 0.001, Figure 2(a)]. Breathprints were significantly changed post-saline (Factor 3: p = 0.01) and post-salbutamol (Factor 2: p = 0.02, Factor 3: p = 0.03) when compared to the breathprints at baseline. The post-salbutamol breathprint was also altered as compared to post-saline [Factor 2: p = 0.02, Figure 2(b)].

Our primary aim was to study the effect of bronchoconstriction on breathprints, but we additionally analyzed the effect of acute bronchodilation induced by salbutamol after the provocation test. Our data show that an acute and marked increase in airway caliber after methacholine challenge does not modify exhaled breathprints. Nevertheless, we cannot exclude any carry-over effect of inhaled methacholine on the post-salbutamol breathprints. Notably, the small but significant increase in airway caliber by salbutamol following sham provocation unexpectedly altered the breathprint. We cannot explain this observation and its inconsistency with regard to giving salbutamol after methacholine challenge. It may suggest that salbutamol inhalation in the absence of methacholine challenge as such provided a signal on the breathprint similar to the sham challenge with inhaled saline.

As temporal drift in polymer sensors due to incomplete desorption of molecules from sensor surfaces has been reported [24], a control study was done, in which non-smoking volunteers followed the same course of three repeated eNose analyses without any intervention. We found no significant temporal change in the four factors analyzed (0 min vs. 60 min vs. 90 min; Factor 1: 0.12 ± 0.94 vs. −0.50 ± 0.97 vs. 0.38 ± 0.97, p = 1.00; Factor 2: 0.61 ± 0.91 vs. −0.07 ± 0.99 vs. −0.54 ± 0.82, p = 0.66; Factor 3: 0.52 ± 1.22 vs. −0.17 ± 0.84 vs. −0.34 ± 0.75, p = 0.16; Factor 4: −0.23 ± 1.22 vs. 0.01 ± 0.86 vs. 0.23 ± 0.93, p = 1.00). This suggests that the change in breathprints observed during the challenge procedures is not the result of sensor drifts in the eNose.

As expected, the decrease in FEV₁ by MCh provocation was more marked than after saline [−30.8 ± 10.5% vs. −3.7 ± 3.8%, p < 0.001, Figure 3(a)]. However, there was no significant difference between the changes (deltas) in breathprints as induced by MCh provocation and sham challenge [delta Factor 2: p = 0.27, delta Factor 3: p = 0.66, Figure 3(b)].

How can these results be explained? The similar changes in breathprint after challenges with methacholine and isotonic saline indicate that alterations in breathprint might be due to the nebulisation procedure itself rather than to the change in airway caliber. We used a previously validated sampling technique with filtering of VOCs in inspired air and drying of exhaled air [12] in order to minimize the influence of humidity, expiratory flow rate and environmental VOCs on the exhaled VOC mixture. Although exhaled air is fully saturated, even under physiological conditions, we cannot exclude an effect of humidity on the breathprints after inhaling nebulized solutions. The current findings suggest that the drying step in the breath collection procedure may need to be optimized when used after aqueous nebulization. This issue might be particularly relevant in future studies where patients are subjected to nebulized therapies. An alternative explanation is that isotonic saline (alone or as a diluent for MCh), having a chloride concentration of 154 mM that is higher than that of the epithelial lining fluid (115 mM), induces metabolic changes in the bronchial epithelium [25]. If so, this could affect exhaled VOC profile thereby modifying exhaled breathprints. This possibility might be examined by local bronchoscopic fluid installation and air sampling.

Baseline FEV₁ values were similar between the two visits (Visit 1: 3.33 ± 0.63 L vs. Visit 2: 3.40 ± 0.72 L, p = 0.34), and neither was there a significant difference in baseline breathprints (Factor 2: p = 0.92, Factor 3: p = 0.18, Figure 4). These findings demonstrate that breathprints do not show between-day variability in non-smoking asthma patients. This extends a previous study reporting no differences in breathprints in healthy smoking and non-smoking control subjects when recorded at two separate days [13]. As variability of breathprints is an important issue in the standardization of breath analysis, further large-scale studies should confirm these findings.