Octane in exhaled breath to diagnose acute respiratory distress syndrome in invasively ventilated intensive care unit patients

Study population

Inclusion criteria were an expected ventilation duration of at least 24 h and age ≥18 years. Patients were excluded when they received invasive ventilation for >48 h at any moment in the 7 days preceding the moment of inclusion or when they were expected to decease within 24 h. Tracheotomised patients were also excluded. Finally, patients were excluded if breath sampling was deemed clinically inappropriate (e.g. in patients with highly contagious conditions) or if consent was withdrawn by the patient or family.

Study procedures

Patients were recruited within the first 48 h after the start of invasive ventilation (supplementary figure S1) and two assessments were performed 24 h apart (time-points 1 and 2). Both assessments consisted of clinical data collection, breath sampling and lung ultrasound (LUS) examination (supplementary figure S1). The chest radiograph and worst arterial oxygen tension (P_aO2)/inspiratory oxygen fraction (F_IO2) ratio in the 24 h before breath sampling were collected and used for ARDS diagnosis. If available, a chest computed tomography (CT) scan was collected within 72 h of inclusion. In patients suspected of ARDS during time-points 1 and 2, additional breath samples were taken at approximately 48 h, 96 h and 12 days after inclusion (supplementary figure S1).

ARDS diagnosis: reference test

A panel of three experts, clinicians with expertise in the field of diagnosing ARDS, independently reviewed the available clinical variables and imaging to determine the ARDS diagnosis. This procedure of diagnosing ARDS was described previously [14]. In short, each expert scored the chest radiograph and chest CT scan on 1) the presence of bilateral opacities consistent with ARDS, 2) the most likely aetiology of the opacities and 3) if cardiogenic oedema was the most likely explanation for the abnormalities [3]. Thereafter the expert decided, based on all available information, whether the patient fulfilled the criteria of the Berlin definition for ARDS. The confidence of this diagnosis was scored on an 8-grade scale (supplementary figure S2a), leading to a grade ranging from 1 to 8, respectively corresponding to high confidence of “no ARDS” or high confidence of “ARDS” on both extremes [15].

Using an algorithm (supplementary figure S2b) that combined the confidence grades of the experts, for each imaging modality and day, the diagnosis was divided into three categories: 1) “no ARDS” if sufficient confidence was reached to exclude ARDS, 2) “ARDS” if sufficient confidence was reached to diagnose ARDS and 3) “uncertain diagnosis” in case of general uncertainty (none of the experts had strong confidence) or conflicting results. To reach a final diagnosis, we prioritised imaging with the most accurate information available for classification (supplementary figure S2c). For the “uncertain diagnosis” category, a consensus meeting was held to determine the final classification for the patient, resulting in two additional categories: “likely ARDS” and “likely no ARDS”. Additionally, the amount of pulmonary oedema was estimated using the Radiographic Assessment of Lung Oedema (RALE) and global LUS scores, as described in the supplementary material. Assessment of the reference test was blinded for any results from the index test.

Breath sampling

Exhaled breath was drawn for 6 min using a dedicated breath gas sampler and polytetrafluoroethylene tubing (Swagelok, Warrington, UK) [12], by connecting a three-way stopcock to the expiratory limb in the case of double-limb ventilation and after the heat–moisture exchange filter on the ventilator site in the case of single-limb ventilation (supplementary figure S3). This method has been shown to be reliable and reproducible [12, 16]. VOCs in exhaled breath were absorbed onto two sorbent tubes filled with 300 mg Carbograph 5TD (Markes International, Llantrisant, UK) and 90 mg Tenax GR (Sigma-Aldrich Chemie, Zwijndrecht, The Netherlands). Subsequently the sorbent tubes were refrigerated until analysis.

POC breath test: index test

The POC breath test was specifically designed and validated for the present study aiming to quantify the amount of (n-)octane present in exhaled breath samples of ICU patients [12]. The POC breath test consists of a thermal desorption unit, a GC column and a photo-ionisation detector. The retention time of octane and sensitivity of the detector were recalibrated at regular intervals using a standardised gas mixture. This was used to calculate the concentration of octane (ppb). The reader of the octane concentration was blinded to the ARDS diagnosis.

GC-MS: additional index test

To validate the amount of octane measured with the POC breath test, the other sorbent tube was analysed by means of GC-MS as described previously [17]. GC-MS analysis, de-noising, peak detection and alignment were performed using the R xcms package (Scripps Center for Metabolomics, La Jolla, CA, USA). The concentration of octane (ppb) was calculated with the m/z 85 peak using a calibration curve generated with a standardised gas mixture. Again, the reader was blinded to the ARDS diagnosis.

Study outcomes

The primary end-point of this study was the diagnostic accuracy of octane concentration in exhaled breath for ARDS, using the ARDS diagnosis by the panel of experts as the reference test. Secondary end-points were: 1) the added diagnostic value of the POC breath test on top of the Lung Injury Prediction Score (LIPS), 2) the association between the octane concentration and the certainty of the ARDS diagnosis (likely or certain diagnoses), and 3) changes in concentration of exhaled octane in patients with ARDS.

As a post-hoc analysis the correlations of the octane concentration with radiological findings of pulmonary oedema were assessed. This was performed for the global LUS score and the RALE score.

Sample size justification

With an expected sensitivity of 80% [8] and a minimally acceptable lower confidence limit of 65%, at least 52 patients with ARDS were required for the study [18]. At a predicted incidence of 10.4% [2], at least 500 patients were needed for the primary end-point.

Statistical analysis

All statistical analyses were performed in R version 4.0.3 (www.r-project.org) using the RStudio interface. For the primary outcome the area under the receiver operating characteristic curve (AUC) was calculated with the patients divided into ARDS (“likely ARDS” and “ARDS”) or no ARDS (“likely no ARDS” and “no ARDS”). The maximum octane concentration of the first two time-points was used. Data were summarised based on the type of data and the distribution of the values for continuous variables. Categorical variables were presented as number and percentage, and differences between the ARDS and no ARDS group and the ARDS categories were tested with a Chi-squared test. Continuous variables were presented as mean with standard deviation for variables with a normal distribution (based on histograms, before and after transformation) or with a presumed normal distribution (such as height) and differences between groups were tested with a t-test or one-way ANOVA, as appropriate. Continuous variables without a normal distribution were presented as median with interquartile range (IQR) and differences between groups were tested with a Mann–Whitney U-test or Kruskal–Wallis test, as appropriate. Receiver operating characteristic curve analysis was performed using the pROC package [19]. Correlation between two continuous variables was performed with the Pearson correlation for normally distributed variables and with the Spearman correlation for non-normally distributed variables. A p-value <0.05 was considered statistically significant.