C24-09 Sepsis Smell Pilot Case-Control Study Showing Differentiation Between Adults With Sepsis Versus Control

Abstract Rationale Early diagnosis of sepsis continues to be a clinical challenge despite guideline-based early and serial assessments and integration of clinical and laboratory indices. Exhaled volatile organic compounds (VOC) have been shown to provide a rapid and non-invasive bedside diagnostic aid based on metabolism related to infection and the immune response. However, there is heterogeneous accuracy across different technological platforms. We hypothesized that a handheld electrochemical breath device could identify a VOC signature that would differentiate patients with sepsis from healthy controls. Methods We evaluated a handheld, non-invasive, bedside device that captures VOC signal in exhaled breath. We collected data from 20 patients with known sepsis secondary to pneumonia based on benchmark clinical and laboratory data, and 20 age- and sex-matched healthy controls in a pilot case-control design. The interactions between VOCs and the electrochemical sensor array within the device were measured over the course of one minute of breath collection from each subject. In this time, voltage was swept across the array periodically, with the capacitance of the sensor array changing with voltage and the VOC interactions. Several metrics were derived from the resulting capacitance-voltage curves, and ten were selected based on importance to be used as inputs into several unsupervised and supervised statistical learning methods. One of these methods was Principal Component Analysis (PCA), which is a dimensionality reduction technique that creates variables by linearly transforming the original inputs. These variables are known as components, and in this study, the majority of the original information was contained in the first two. These components were visualized in a 2-dimensional scatter plot, shown below (Figure 1). Results The generated PCA scatter plot showed distinct separation between sepsis and control subjects along the x-axis, or Principal Component 1 (Figure 1). Conclusion In this case-control pilot, a handheld electrochemical breath device showed biologically plausible separation between VOC signatures of patients with sepsis and healthy controls. More definitive diagnostic results would require a large cohort with sepsis mimics and head-to-head comparison with the standard of care (laboratory work-up and clinical scoring system) versus a hybrid approach of both. This abstract is funded by: milewski foundation