Ruling out SARS-CoV-2 infection using exhaled breath analysis by electronic nose in a public health setting

Publication: R. de Vries, R.M. Vigeveno, S. Mulder, N. Farzan, D.R. Vintges, J.J. Goeman, S. Bruisten, B. van den Corput, J.J. Miranda Geelhoed, L.G. Visser, M. van der Lubben, P.J. Sterk, J.C.C.M. in ’t Veen, G.H. Groeneveld. Ruling out SARS-CoV-2 infection using exhaled breath analysis by electronic nose in a public health setting. MEDRXIV. doi: https://doi.org/10.1101/2021.02.14.21251712

Aim: To assess the accuracy of exhaled breath analysis by electronic nose (eNose) for the discrimination between individuals with and without a SARS-CoV-2 infection.

Take home message: The eNose can rapidly and accurately rule out SARS-CoV-2 infection, achieving high sensitivity (98-100%) in multiple validation cohorts. This non-invasive approach can reduce the need for follow-up tests, ease diagnostic burdens, and streamline screening in public health settings, demonstrating its potential as an effective tool for respiratory infection detection.

Introduction

The study explored the use of exhaled breath analysis via an electronic nose (eNose) to detect SARS-CoV-2 infection. Traditional diagnostic tests, such as RT-PCR, are resource-intensive and often face logistical challenges. eNose technology, which detects volatile organic compounds (VOCs) in exhaled breath, offers a rapid, non-invasive alternative that could expand testing capacity and speed up results. This study aimed to evaluate the diagnostic accuracy of the eNose in distinguishing between individuals with and without SARS-CoV-2 infection.

Methods

A prospective study was conducted in public health test facilities in the Netherlands. Participants were adults who presented with COVID-19 symptoms or had been exposed to a confirmed case. The study was divided into four cohorts: a training set, a validation set, a replication set, and an asymptomatic set. Breath samples were collected using the SpiroNose® device, which employs cross-reactive metal oxide sensors to capture the breath profile. Data were processed and analyzed using machine learning algorithms. The primary outcome was the diagnostic accuracy of the eNose, validated across the cohorts using ROC analysis.

Results

For the analysis 4510 individuals were available. In the training set (35 individuals with; 869 without SARS-CoV-2), the eNose sensors were combined into a composite
biomarker with a ROC-AUC of 0.947 (CI:0.928-0.967). These results were confirmed in the validation set (0.957; CI:0.942-0.971, n=904) and externally validated in the replication set (0.937; CI:0.926-0.947, n=1948) and the asymptomatic set (0.909; CI:0.879-0.938, n=754). Selecting a cut-off value of 0.30 in the training set resulted in a sensitivity/specificity of 100/78, >99/84, 98/82% in the validation, replication and asymptomatic set, respectively.

Discussion

The findings demonstrate that eNose technology can effectively differentiate between individuals with and without SARS-CoV-2 infection. This study, the first large-scale population-based analysis, supports the potential of eNose as a screening tool in public health settings. By rapidly identifying individuals who are unlikely to be infected, the eNose could alleviate the burden on traditional diagnostic methods, reduce costs, and streamline the testing process.