Rapid detection of heart failure using a spectroscopic liquid biopsy

Abstract

Heart disease is growing annually across the globe with numbers expected to rise to 46% of the population by 2030. Early detection is vital for several reasons, firstly it improves the long-term prognosis of the patient by admitting them through the appropriate pathway faster, secondly it reduces healthcare costs by streamlining diagnosis and finally, in combination with management or treatment, it can prevent the progression of the disease which in turn improves the patient’s quality of life. Therefore, there lies an increasing need to develop assays which can rapidly detect heart disease at an early stage. The Dxcover® liquid biopsy platform employs infrared spectroscopy and artificial intelligence, to quickly analyse minute amounts of patient serum. In this study, discrimination between healthy controls and diseased patients was obtained with an area under the receiver operating characteristic curve (AUC) of 0.89. When assessing the heart failure vs all patients, which is most akin to what would be observed in a triage setting, the model when tuned to a minimum of 45% specificity yielded a sensitivity of 89% and an NPV of 0.996, conversely when sensitivity was set at a 45% minimum, the specificity was 96%, giving an NPV of 0.991 when using a 1.5% prevalence. Other models were assessed in parallel, but the performance of the ORFPLS model was overall superior to the other models tested. In this large scale (n = 404) proof-of-concept study, we have shown that the Dxcover liquid biopsy platform has the potential to be a viable triage tool in emergency and routine situations for the diagnosis of heart failure.