The Potential for Troponin to Inform Prognosis in Patients With Stable Coronary Artery Disease

Enhancing the diagnosis of functionally relevant coronary artery disease with machine learning

Functionally relevant coronary artery disease (fCAD) can result in premature death or nonfatal acute myocardial infarction. Its early detection is a fundamentally important task in medicine. Classical detection approaches suffer from limited diagnostic accuracy or expose patients to possibly harmful radiation. Here we show how machine learning (ML) can outperform cardiologists in predicting the presence of stress-induced fCAD in terms of area under the receiver operating characteristic (AUROC: 0.71 vs. 0.64, p = 4.0E-13). We present two ML approaches, the first using eight static clinical variables, whereas the second leverages electrocardiogram signals from exercise stress testing. At a target post-test probability for fCAD of <15%, ML facilitates a potential reduction of imaging procedures by 15-17% compared to the cardiologist's judgement. Predictive performance is validated on an internal temporal data split as well as externally. We also show that combining clinical judgement with conventional ML and deep learning using logistic regression results in a mean AUROC of 0.74.

Diagnostic and prognostic value of H-ficolin for functionally relevant coronary artery disease

BACKGROUND

We aimed to test the diagnostic and prognostic ability of H-ficolin, an initiator of the lectin pathway of the complement system, for functionally relevant coronary artery disease (fCAD), and explore its determinants.

METHODS

The presence of fCAD was adjudicated using myocardial perfusion imaging single-photon emission tomography and coronary angiography. H-ficolin levels were measured by a sandwich-type immunoassay at rest, peak stress-test, and 2 h after stress-test. Cardiovascular death and non-fatal myocardial infarction were assessed during 5-year follow-up.

RESULTS

Among 1,571 patients (32.3 % women), fCAD was detected in 462 patients (29.4 %). H-ficolin concentration at rest was 18.6 (15.3-21.8) µg/ml in patients with fCAD versus 17.8 (15.4-21.5) µg/ml, p = 0.33, in patients without fCAD, resulting in an AUC of 0.53 (95 %CI 0.48-0.56). During follow-up, 107 patients (6.8 %) had non-fatal myocardial infarction and 99 patients (6.3 %) experienced cardiovascular death. In Cox regression analysis, H-ficolin was not a predictor of events in the overall cohort. Subgroup analysis suggested a potential link between H-ficolin and non-fatal myocardial infarction in patients without fCAD (adjusted HR 1.03, 95 % CI 1.02-1.15, p = 0.005). H-ficolin concentration showed a weak positive correlation with systolic (r = 0.069, p < 0.001) and diastolic blood pressure (r = 0.111, p < 0.001).

CONCLUSION

H-ficolin concentration did not have diagnostic and/or prognostic value in patients referred for fCAD work-up.

Serum neurofilament light chain in functionally relevant coronary artery disease and adverse cardiovascular outcomes

Background: Functionally relevant coronary artery disease (fCAD), causing symptoms of myocardial ischemia, can currently only be reliably detected with advanced cardiac imaging. Serum neurofilament light chain (sNfL) is a biomarker for neuro-axonal injury known to be elevated by cardiovascular (CV) risk factors and cerebrovascular small-vessel diseases. Due to their pathophysiological similarities with fCAD and the link to CV risk factors, we hypothesised that sNfL may have diagnostic and prognostic value for fCAD and adverse cardiovascular outcomes.Methods: Of the large prospective Basel VIII study (NCT01838148), 4'016 consecutive patients undergoing cardiac work-up for suspected fCAD were included (median age 68 years, 32.5% women, 46.9% with history of CAD). The presence of fCAD was adjudicated using myocardial perfusion imaging single-photon emission tomography (MPI-SPECT) and coronary angiography. sNfL was measured using a high-sensitive single-molecule array assay. All-cause and cardiovascular death, myocardial infarction (MI), and stroke/transient ischaemic attack (TIA) during 5-year follow-up were the prognostic endpoints.Results: The diagnostic accuracy of sNfL for fCAD as quantified by the area under the curve (AUC) was low (0.58, 95%CI 0.56-0.60). sNfL was strongly associated with age, renal dysfunction, and body mass index and was a strong and independent predictor of all-cause death, cardiovascular death, and stroke/TIA but not MI. Time-dependent AUC for cardiovascular-death at 1-year was 0.85, 95%CI 0.80-0.89, and 0.81, 95%CI 0.77-0.86 at 2-years.Conclusion: While sNfL concentrations did not show a diagnostic role for fCAD, in contrast, sNfL was a strong and independent predictor of cardiovascular outcomes, including all-cause death, cardiovascular death and stroke/TIA.

Using High-Sensitivity Cardiac Troponin for the Exclusion of Inducible Myocardial Ischemia in Symptomatic Patients: A Cohort Study

BACKGROUND

The optimal noninvasive method for surveillance in symptomatic patients with stable coronary artery disease (CAD) is unknown.

OBJECTIVE

To apply a novel approach using very low concentrations of high-sensitivity cardiac troponin I (hs-cTnI) for exclusion of inducible myocardial ischemia in symptomatic patients with CAD.

DESIGN

Prospective diagnostic cohort study. (ClinicalTrials.gov: NCT01838148).

SETTING

University hospital.

PATIENTS

1896 consecutive patients with CAD referred with symptoms possibly related to inducible myocardial ischemia.

MEASUREMENTS

Presence of inducible myocardial ischemia was adjudicated using myocardial perfusion imaging with single-photon emission computed tomography, as well as coronary angiography and fractional flow reserve measurements where available. Staff blinded to adjudication measured circulating hs-cTn concentrations. An hs-cTnI cutoff of 2.5 ng/L, derived previously in mostly asymptomatic patients with CAD, was assessed. Predefined target performance criteria were at least 90% negative predictive value (NPV) and at least 90% sensitivity for exclusion of inducible myocardial ischemia. Sensitivity analyses were based on measurements with an hs-cTnT assay and an alternative hs-cTnI assay with even higher analytic sensitivity (limit of detection, 0.1 ng/L).

RESULTS

Overall, 865 patients (46%) had inducible myocardial ischemia. The hs-cTnI cutoff of 2.5 ng/L provided an NPV of 70% (95% CI, 64% to 75%) and a sensitivity of 90% (CI, 88% to 92%) for exclusion of inducible myocardial ischemia. No hs-cTnI cutoff reached both performance characteristics predefined as targets. Similarly, using the alternative assays for hs-cTnI or hs-cTnT, no cutoff achieved the target performance: hs-cTnT concentrations less than 5 ng/L yielded an NPV of 66% (CI, 59% to 72%), and hs-cTnI concentrations less than 2 ng/L yielded an NPV of 68% (CI, 62% to 74%).

LIMITATION

Data were generated in a large single-center diagnostic study using central adjudication.

CONCLUSION

In symptomatic patients with CAD, very low hs-cTn concentrations, including hs-cTnI concentrations less than 2.5 ng/L, do not generally allow users to safely exclude inducible myocardial ischemia.

PRIMARY FUNDING SOURCE

European Union, Swiss National Science Foundation, Kommission für Technologie und Innovation (Innosuisse), Swiss Heart Foundation, Cardiovascular Research Foundation Basel, University of Basel, University Hospital Basel, Roche, Abbott, and Singulex.