Phenotyping Asthma using an Electronic Nose

Abstract

Rationale: Asthma is a chronic respiratory condition affecting many millions of people worldwide. The recommended long-term preventative treatment for asthma is corticosteroid medication. Improvement in asthma symptoms in response to corticosteroids is not universal, and prescription to non-responsive patients is costly and potentially dangerous. The assessment of whether patients will respond to corticosteroid treatment is an important clinical problem.

This study investigated the utility of a gas sensor array, the ‘Electronic nose’, in predicting response to steroid among asthmatics and in differentiating asthmatics from healthy controls using exhaled breath. The anti-inflammatory actions of steroids and the biochemical basis of steroid response are complex and may be better quantified by the electronic nose than by single biomarkers.

In parallel with the assessment of steroid response a study was conducted on the ability of the electronic nose to predict sputum eosinophil counts among asthmatics. Eosinophil count is a predictor of steroid response and can be used as a guide to asthma treatment.

Statement of problems: Can steroid-responsive and non-steroid responsive asthmatics be distinguished using electronic nose analysis of exhaled breath?

Is sputum eosinophil count able to be predicted by electronic nose analysis of exhaled breath?

Methods: 47 patients (27 asthmatics, 20 healthy controls) participated in the study.

Asthmatic patients completed a two-week trial of oral prednisone. Asthmatics were classified as steroid responsive if FEV1 improved by 15%, PC20AMP improved by >300%, or ACQ (asthma control questionnaire) improved by >0.5 points over the steroid course. 16 asthmatics were steroid responsive and 11 asthmatics were steroid unresponsive.

Breath samples were taken before and after the steroid course. A sputum sample was taken prior to the steroid course. Asthmatics were defined as eosinophilic if their sputum cell count contained more than 3% eosinophils. Healthy controls provided a single breath sample and sputum sample.

Main results: Steroid responsive and non-steroid responsive asthmatics were unable to be distinguished either before or after the steroid course (no significant principal component differences, Mdistances<2, all p>0.2). Healthy controls were significantly differentiated from pre-steroid asthmatics (PC2, PC4, PC6, p=0.0090, 0.000060, 0.0090 respectively, M-distance=4.66,p=0.031) and less differentiated from post-steroid asthmatics (PC6, p=0.0016, Mdistance=3.80,p=0.16). A multilayer perceptron based predictive model for the comparison was associated with a cross-validation value (CVV) of 83% between controls and pre-steroid samples and 70% between controls and post-steroid samples. Improvement in ACQ (PC6, p=0.0041) and improvement in FEV1 (PC6, p=0.045) could be detected based on differences in samples before and after the steroid course.

A principal component from pre-steroid asthmatics was strongly correlated with sputum eosinophil counts (coefficient=0.615, p=0.0082). Breathprints from eosinophilic and noneosinophilic asthmatics were significantly differentiated (PC4, p=0.0033, M-distance=2.00, p=0.0020). A predictive model for the comparison was associated with a CVV of 77%.

Conclusions: Steroid responsive and non-steroid responsive asthmatics cannot be differentiated on the basis of exhaled breath analysis by electronic nose. Healthy controls and asthmatics can be significantly differentiated on this basis. Electronic nose readings are correlated with sputum eosinophil counts and eosinophilic and non-eosinophilic asthma can be significantly differentiated.