Neuroglial Biomarkers for Risk Assessment of Ischemic Stroke and Other Cardiovascular Events in Patients With Atrial Fibrillation Not Receiving Oral Anticoagulation

Figures

• 

  Figure 1. Distribution of NFL, GFAP, tau, and UCHL1 in regard to heart rhythm at baseline and prior stroke/TIA. The shaded area represents the density and the lines represent the ECDF. Note the log‐scale. AF indicates atrial fibrillation; ECDF, empirical cumulative distribution function; GFAP, glial fibrillary acidic protein; NFL, neurofilament light chain; TIA, transient ischemic attack; and UCHL1, ubiquitin carboxy‐terminal hydrolase L1.

• 

  Figure 2. Smooth association between (A) NFL, (B) GFAP, (C) tau, and (D) UCHL1 and 1‐year risk of ischemic stroke. Cox regression model including ln(biomarker) as a restricted cubic spline with 3 knots. Dots indicate biomarker levels for patients with (top) and without (bottom) event. GFAP indicates glial fibrillary acidic protein; NFL, neurofilament light chain; and UCHL1, ubiquitin carboxy‐terminal hydrolase L1.

• 

  Figure 3. Summary of the models for NFL. The HRs correspond with a doubling of NFL. P indicates the P value for NFL's contribution in the respective models. C ref represents the C index for the corresponding model without NFL, and FNI is the fraction of new information provided by NFL (see Statistical Analysis). CV indicates cardiovascular; Hosp HF, hospitalization for heart failure; HR, hazard ratio; NFL, neurofilament light chain; and SE, systemic embolism.

• 

  Figure 4. Summary of the models for GFAP. The HRs correspond to a doubling of GFAP. For ischemic stroke, a nonlinear association was assumed. Thus, the HR corresponds to an arbitrary comparison for a doubling around the median. P represents the P value for GFAP's contribution in the respective models. The C index is for the specified model (ie, including all variables in the respective models). The C ref represents the C index for the corresponding model without GFAP, and FNI is the fraction of new information provided by GFAP (see Statistical Analysis). CV indicates cardiovascular; GFAP, glial fibrillary acidic protein; Hosp HF, hospitalization for heart failure; HR, hazard ratio; and SE, systemic embolism.

• 

  Figure 5. Summary of the models for tau. The HRs correspond to a doubling of tau. For Hosp HF, we assumed a nonlinear association, and the HR corresponds to a comparison for a doubling around the median. P represents the P value for tau's contribution in the respective models. The C index is for the specified model (ie, including all variables in the respective models). The C ref represents the C index for the corresponding model without tau, and FNI is the fraction of new information provided by tau (see Statistical Analysis). CV indicates cardiovascular; Hosp HF, hospitalization for heart failure; HR, hazard ratio; and SE, systemic embolism.

• 

  Figure 6. Summary of the models for UCHL1. The HR corresponds to a doubling of UCHL1. P represents the P value for UCHL1's contribution in the respective models. The C index is for the specified model (ie, including all variables in the respective models). The C ref represents the C index for the corresponding model without UCHL1, and FNI is the fraction of new information provided by UCHL1 (see Statistical Analysis). CV indicates cardiovascular; Hosp HF, hospitalization for heart failure; HR, hazard ratio; SE, systemic embolism; and UCHL1, ubiquitin carboxy‐terminal hydrolase L1.