The ability of an electronic nose to distinguish between complications in lung transplant recipients
Aim: To assess the ability of eNose technology to discriminate between complications after lung transplantation (LTx) in patients presenting with a clinical suspicion of an LTx complication.
Take home message: eNose technology shows high accuracy in non-invasively detecting bacterial colonization in lung transplant patients, offering a potential alternative to invasive diagnostics and enabling earlier intervention to improve patient outcomes.
Introduction
Complications like acute cellular rejection (ACR) and infection are known risk factors for the development of chronic lung allograft dysfunction, impacting long-term patient and graft survival after lung transplantation (LTx). Differentiating between complications remains challenging and time-sensitive, highlighting the need for accurate and rapid diagnostic modalities.
Methods
The ability of exhaled breath analysis with an electronic nose (eNose) to differentiate between acute cellular rejection (ACR), infection, and mechanical complications was evaluated in lung transplant recipients (LTR) presenting with suspected complications. LTR with a suspected complication and a subsequently confirmed diagnosis underwent breath analysis using the SpiroNose®. Supervised machine learning algorithms were applied to determine the eNose’s accuracy in distinguishing between various complications. Additionally, the incremental value of eNose measurements beyond standard clinical parameters was assessed.
Results
In 90 lung transplant recipients (LTR), a total of 161 measurements were conducted during suspected complications, resulting in 84 confirmed diagnoses. The eNose demonstrated an accuracy of 74% in distinguishing between acute cellular rejection (ACR), infections, and mechanical complications. When differentiating specifically between ACR and infections, the accuracy increased to 82%. Notably, combining eNose measurements with standard clinical parameters further enhanced diagnostic accuracy to 88% (P = .0139), achieving a sensitivity of 94% and a specificity of 80%.