Breath Analysis Technology Achieves Breakthrough in Predicting Pneumonia Before Symptoms Appear
核心洞察
Breath Diagnostics (搜索)' OneBreath™ (搜索) technology demonstrated the ability to predict pneumonia (搜索) onset in cardiac surgery patients days before clinical symptoms emerge, achieving an AUROC of 0.833 in a study published in The Journal of Thoracic and Cardiovascular Surgery.
The prospective study of 75 cardiac surgery patients showed that volatile organic compounds (搜索) in exhaled breath can serve as reliable biomarkers for future disease, with 10 patients developing postoperative pneumonia (搜索).
This represents the first demonstration that breath analysis combined with machine learning can both diagnose and predict pneumonia (搜索), potentially transforming postoperative care and enabling earlier interventions.
Breath Diagnostics (搜索) has achieved a significant breakthrough in predictive medicine with its OneBreath™ (搜索) technology demonstrating the ability to predict pneumonia (搜索) onset in cardiac surgery patients days before clinical symptoms appear. A peer-reviewed study published in The Journal of Thoracic and Cardiovascular Surgery shows that volatile organic compounds (搜索) (VOCs) measured in exhaled breath can reliably serve as biomarkers for future disease.
Study Design and Clinical Performance
The prospective study enrolled 75 patients undergoing elective cardiac surgery at the University of Louisville. Breath samples were collected using a Tedlar bag system with microchip capture of carbonyl compounds, followed by ultra-high-performance liquid chromatography mass spectrometry for compound identification. Samples were taken preoperatively, within 24 hours postoperatively, and every three days during hospitalization.
Out of the 75 patients, 10 developed postoperative pneumonia (搜索). The diagnostic model achieved an AUROC of 0.833 and PRAUC of 0.818 on the test set. Remarkably, the predictive model using only baseline preoperative breath samples achieved the same performance metrics (AUROC of 0.833 and PRAUC of 0.818), identifying high-risk patients days before clinical symptoms emerged.
"These findings are highly significant," said Dr. Victor van Berkel, Chief Medical Officer and Co-Founder of Breath Diagnostics (搜索) and co-author of the study. "Our study demonstrated not only that pneumonia (搜索) can be diagnosed earlier than with current clinical methods, but also that its onset can be predicted before symptoms appear. This capability could transform postoperative care and outcomes for cardiac surgery patients, and potentially for those undergoing other major surgeries as well."
Addressing Critical Clinical Need
Pneumonia (搜索) develops in 6-20% of patients following elective cardiac surgery, making it one of the most frequent and serious postoperative complications. It is strongly associated with longer ICU stays, extended hospitalizations, higher costs, and increased mortality. Despite decades of research, preventive strategies such as prophylactic antibiotics have produced inconsistent results, largely because clinicians lack a reliable way to identify which patients are truly at risk.
Until now, there has been no reliable, non-invasive way to predict who will develop the condition. This study marks the first time that VOC breath analysis, paired with machine learning, has demonstrated the ability not only to diagnose but also to predict disease onset days in advance.
Technology Platform and Broader Applications
Carbonyl VOC profiles were analyzed using advanced machine learning workflows, including feature selection, random forest, and boosted generalized linear models. The diagnostic model was trained on perioperative VOC samples, while a separate predictive model was trained using only preoperative breath samples to assess risk before pneumonia (搜索) onset.
By detecting VOC signatures associated with inflammation, oxidative stress, and bacterial metabolism, this approach could be extended to a range of diseases from lung infections to cancer and inflammatory disorders. The company reports that its platform is among the most clinically validated breath technologies for lung cancer (搜索) detection, with 94% sensitivity and 85% specificity across multiple independent sites and stages of disease.
"Exhaled breath is an extraordinary diagnostic medium that has only begun to be tapped," said Ivan Lo, CEO of Breath Diagnostics (搜索). "This study demonstrates that with our OneBreath™ (搜索) technology, combined with advanced data science, breath can serve as a reliable biomarker for predicting disease before symptoms appear."
Development Status and Future Plans
The OneBreath™ (搜索) system remains in the research and development stage and has not been cleared or approved by the U.S. Food and Drug Administration or any other regulatory authority. Breath Diagnostics (搜索) will build on these results by expanding clinical validation into larger, multi-center patient cohorts to further confirm the predictive value of exhaled breath biomarkers.
The company is working with leading academic and clinical partners to refine study protocols, with the goal of pursuing regulatory clearances in the United States and select international markets. These efforts represent an important step toward the company's long-term goal of integrating predictive breath tests into hospital and surgical settings, where early detection may reduce complications and healthcare costs, and improve clinical decision-making.