Cross-sectional biomarker comparisons in asthma monitoring using a longitudinal design, the eNose premise
Aim:
- To examine the ability of eNose singnals in discriminating asthmatics from healthy controls at every time point before and after viral challenge.
- The ability of eNose signals to distinguish between previral challenge phase and every time point in postviral challenge phase for healthy and asthmatic cohorts separately.
Take home message: Cross-sectional study designs can identify differences between healthy and asthmatic states, longitudinal breath analysis is essential for understanding fluctuations driven by external triggers like viral infections. The study suggests that eNose technology could serve as a reliable, non-invasive tool for asthma monitoring and phenotyping, highlighting its potential use in real-time patient management without the need to identify individual VOCs.
Introduction
Biological responses in health and disease are dynamic, changing over time as the body adapts to external influences. Single biomarker snapshots in cross-sectional studies fail to capture these fluctuations, while longitudinal studies, despite being more informative, often face limitations due to the burden of repeated sampling.
Asthma is a complex and variable condition, with exacerbations triggered by factors like viral infections that significantly impact the disease’s clinical status. Exhaled breath volatile organic compounds (VOCs) measured by electronic nose (eNose) technology have shown promise for noninvasive asthma diagnosis and phenotyping. Unlike traditional methods, eNose allows for frequent, noninvasive breath sampling without major patient burden, offering a potential advantage for longitudinal assessments.
Methods
Twelve adult atopic asthma patients and twelve healthy, non-smoking volunteers were recruited for this study. Participants were monitored three times a week for two months before and one month after exposure to rhinovirus-16, totaling an average of 33 visits per participant. At each visit, real-time exhaled breath VOCs were measured using the SpiroNose®, which employs seven cross-reactive sensors to capture unique breath profiles for each individual.
Data Analysis:
The analysis involved three main approaches:
- An adaptive LASSO model was applied to the eNose sensor data with three-fold cross-validation to differentiate asthmatic patients from healthy controls, calculating AUC-ROC values.
- AUC-ROCs were also determined for individual sensors to identify which sensors contributed most to discrimination and to validate the LASSO findings.
- Median differences in sensor signals were analyzed to assess variability in breath profiles between asthmatics and controls.
Additionally, the same methods were used to compare pre- and post-viral challenge phases in both groups. The visit prior to viral challenge served as a reference point for detecting any viral-induced changes beyond normal daily variability. Linear mixed-effects models further assessed the relationship between sensor signals and viral exposure in both the asthma and control groups.
Results
The adaptive LASSO model achieved perfect discrimination (AUC-ROC of 1) between asthmatic patients and healthy controls at all study visits, both in the stable phase and after the viral challenge. The highest median differences in exhaled breath profiles were attributed mainly to sensors 5 and 6, with sensor 7 contributing less. After the viral challenge, sensor 5 showed the largest difference between groups.
Asthmatic patients showed higher AUC-ROC values when comparing pre- and post-viral challenge phases, starting from the day of the viral challenge. Sensor 5 demonstrated significant AUC-ROC values of 0.78 to 0.79 across visits 1, 3, and 7. In contrast, controls only showed significant discrimination later, at visit 7, with sensor 3 (AUC-ROC of 0.76). The adaptive LASSO model confirmed earlier and stronger discrimination in asthmatics compared to controls. The mixed-effects model also indicated that more sensors were associated with the viral challenge in asthmatics than in controls.
Discussion
This study explored whether a single snapshot of biomarkers can reliably differentiate between diseased and healthy individuals in both stable and unstable states. Exhaled breathprints successfully distinguished healthy participants from those with asthma at all time points, suggesting that certain VOCs detected by specific sensors (5, 6, and 7) provide consistent differentiation, despite day-to-day variability and viral exposure. This indicates the potential of the eNose as a robust diagnostic tool for asthma.
The study also found that during the viral challenge phase, changes in VOC patterns were more pronounced in asthmatic patients compared to healthy individuals. This suggests that while viral triggers affect both groups, the physiological response differs, making the eNose useful for monitoring and phenotyping asthma without needing to identify individual VOCs. However, further research using techniques like mass spectrometry is needed to identify the specific VOCs involved and understand the underlying mechanisms.
A limitation of the study was its small sample size, though this was partly mitigated by frequent sampling. Additionally, the study population may not fully represent real-world cases, as participants may have fewer comorbidities than typical asthma patients. In conclusion, while cross-sectional studies can identify biomarkers distinguishing asthma from healthy states, longitudinal designs are essential for capturing changes due to external triggers like viral infections. Overall, the eNose shows promise as a quick, non-invasive tool for asthma monitoring and diagnosis in clinical practice.
