Feasibility of an Electronic Nose to Aid Biochemical Assessment of Adrenal Lesions.

OBJECTIVES

Analysis of volatile organic compounds (VOCs) in exhaled breath has emerged as a promising noninvasive diagnostic tool for various diseases. The aim of this study was to evaluate the potential of an electronic nose to differentiate between functional adrenal lesions - pheochromocytoma (PHEO), primary hyperaldosteronism (PHA), and hypercortisolism (CS) - and nonfunctional adrenal lesions.

METHODS

A pilot study was conducted at 2 tertiary hospitals within the Netherlands. Patients with PHEO, PHA, CS, and nonfunctional adrenal lesions underwent breath testing with an electronic nose between May 2021 and June 2024. Each center employed a distinct electronic nose (device A and B). Comparability of data between the devices was assessed in a t-distributed stochastic neighbor embedding plot, and an artificial neural network was trained to classify breath patterns.

RESULTS

Data obtained from the 2 electronic noses were too heterogeneous for pooling and device B had an insufficient sample size for further analysis. Therefore, VOC patterns of 76 functional lesions (27 PHEO, 33 PHA, 16 CS) and 29 nonfunctional adrenal lesions measured exclusively with device A were analyzed. Moderate discriminative performance was observed in the training data: pooled functional lesions (area under the curve [AUC] 0.76), PHEO [AUC 0.76], PHA [AUC 0.72] and CS [AUC 0.58] versus nonfunctional lesions. However, model performance declined significantly when applying the model developed with training data on test data, with wide confidence intervals across all comparisons.

CONCLUSIONS

While slight differences in VOC patterns were detected between functional and nonfunctional adrenal lesions, the electronic nose demonstrated limited discriminative value for clinical practice.