Hemodynamic Profiling Through Non-Invasive Estimation of Pulmonary Capillary Wedge Pressure

Background Elevated pulmonary capillary wedge pressure (PCWP) is an important marker of left ventricular filling pressure in heart failure and other cardiovascular diseases. Direct measurement requires right heart catheterization (RHC), while non invasive assessment remains limited. The objective of this study was to develop and validate a non-invasive, point-of-care, machine-learned algorithm for detecting clinically relevant PCWP elevation. Methods Orthogonal voltage gradient signals from thoracic electrodes and finger photoplethysmography signals were collected using a proprietary non-invasive acquisition device in prospective, multicenter studies across 19 U.S. sites. The algorithm was developed and tested in blinded, enrollment-gated validation using two prespecified populations: subjects with RHC-confirmed PCWP >18 mm Hg for sensitivity and symptomatic subjects with low likelihood of elevated PCWP for specificity, defined by guideline-based non-invasive criteria. Prespecified performance goals required lower confidence bounds exceeding 70% for sensitivity and 60% for specificity. Results In blinded validation, 108 subjects had RHC-confirmed PCWP >18 mm Hg and 147 symptomatic subjects met low-likelihood criteria. In the elevated-PCWP cohort, sensitivity was 82.4% (95% CI, 75.2%–89.6%), exceeding the prespecified lower confidence-bound goal of 70% (p=0.0005). In the low-likelihood cohort, specificity was 83.0% (95% CI, 76.9%–89.1%), exceeding the prespecified lower confidence-bound goal of 60% (p< 0.0001) The ROC-AUC was 0.912, with positive and negative likelihood ratios of 4.85 and 0.21, respectively. No significant performance differences were observed across prespecified subgroups, including age, sex, and race. Conclusions A non-invasive, point-of-care, machine-learned algorithm detected elevated PCWP with high sensitivity and specificity in a blinded, multicenter validation study.