Baseline characteristics of the study participants
The current study analyzed 63 CADASIL patients with genetically confirmed NOTCH3 mutation. The average age of the patients at inclusion was 58.9 ± 9.3 years, and 40 (63.5%) patients were men. Sixteen (25.4%) patients did not have clinically documented stroke events at baseline (no stroke group), 26 (41.3%) had only IS (IS group), and 21 (33.3%) had at least one symptomatic hemorrhagic stroke (ICH group) at baseline. Hypertension was more prevalent in patients with stroke, especially in the ICH group (p = 0.001). Other clinical history was not significantly different among the three groups. Of the neuroimaging features, the ICH group had more severe WMH, higher numbers of CMBs, and more mixed location of CMBs (Table 1).
The natural log-transformed plasma biomarkers levels of the control and each stroke subgroup of the CADASIL patients are plotted in Figure 1. In patients with history of stroke, the time intervals between the latest stroke and blood sampling were 415.5 ± 899.8 days. A non-significant trend of negative correlation between the intervals and levels of NfL (Spearman = −0.25, 95% CI = -0.51–0.05) and GFAP ( = −0.10, 95% CI = -0.39–0.20) were observed. The age-adjusted mean plasma levels of the biomarkers were generally higher in the CADASIL patients than in the control group (NfL: 3.30 ± 0.28 vs 2.08 ± 0.56 natural log pg/mL; GFAP: 5.21 ± 0.30 vs 4.04 ± 0.60 natural log pg/mL; UCHL1: 2.64 ± 0.30 vs 1.61 ± 0.65 natural log pg/mL; all p < 0.01), except for those of tau (-0.30 ± 0.29 vs 0.22 ± 0.56 natural log pg/mL, p = 0.11). Among CADASIL patients, the plasma NfL level was higher in the IS (3.53 ± 0.42 natural log pg/mL) and ICH (3.44 ± 0.47 natural log pg/mL) groups than in the no stroke group (2.68 ± 0.56 natural log pg/mL, p = 0.01 and p = 0.04, respectively), whereas the GFAP level was elevated in the ICH group (5.87 ± 0.49 natural log pg/mL) than in the IS (5.01 ± 0.44 natural log pg/mL, p = 0.01) or no stroke group (4.62 ± 0.57 natural log pg/mL, p = 0.001).
Relationship between plasma biomarkers and clinical/neuroimaging features
The numbers of CMBs was positively correlated with the plasma levels of both NfL ( = 0.32, 95% CI 0.03–0.56, p = 0.03) and GFAP ( = 0.37, 95% CI=0.08–0.60, p = 0.01). In the multiple linear regression analysis, similar finding was observed that higher plasma NfL was associated with a higher number of CMBs (β = 0.16, 95% CI 0.02–0.30, p = 0.02). No significant associations were observed between plasma biomarkers and other neuroimaging markers (Table 2).
Impacts of plasma biomarkers on clinical outcome
The multivariable logistic regression analysis differentiated the CADASIL patients from the control group based on higher plasma levels of NfL (aOR 12.4, 95% CI 2.87–53.1, p = 0.001), GFAP (aOR 27.9, 95% CI 2.94–265.4, p = 0.004), and UCHL1 (aOR 4.08, 95% CI 1.57–10.6, p = 0.004). We investigated which plasma biomarkers were associated with the history of stroke (n = 47) and ICH (n = 21) at baseline in the CADASIL patients. For baseline stroke, the diagnosis performance was highest NfL (AUC 0.735, p = 0.007) followed by GFAP (AUC 0.674, p = 0.046), while for baseline ICH, GFAP had the best performance (AUC 0.745, p = 0.003; Figure 2). After adjusting for age, sex, and hypertension, the higher NfL level was found to be associated with any stroke at baseline (aOR 2.02, 95% CI 1.06–3.87, p = 0.03), whereas the GFAP level had a borderline association (aOR 2.51, 95% CI 0.92–6.89, p = 0.07). By contrast, only the plasma GFAP level was found to be associated with ICH at baseline (aOR 2.06, 95% CI 1.26–3.35, p = 0.004; Table 3).
During a mean follow-up period of 3.1 ± 2.1 years, 10 patients (16%) had at least 1 incident stroke and 6 of them were ICH. The annual stroke and ICH rates were 8.1 and 4.8 per 100 person-years, respectively. When plotting the ROC curve, plasma NfL and GFAP remained the only two plasma biomarkers that had ability in predicting incident stroke (AUC = 0.783 and 0.742 for NfL and GFAP, respectively, both p < 0.05) or ICH (AUC = 0.798 and 0.782 for NfL and GFAP, respectively, both p < 0.05) (Figure 2). In Cox regression analysis, NfL predicted the incident stroke (HR 1.93 per 1-unit increase in natural log-transformed NfL, 95% CI 1.19–3.13, p = 0.01). The proportional hazards assumption was not violated. In addition, both higher NfL (HR 2.39, 95% CI 1.21–4.70, p = 0.01) and GFAP (HR 2.80, 95% CI 1.21–6.53, p = 0.02; Table 3) levels were predictive of incident ICH during follow-up. The results remained consistent when the plasma biomarkers were analyzed in their original scale (HR 1.004 per 1-unit increase in the absolute NfL level, 95% CI 1.000–1.008, p = 0.03; HR 1.001 per 1-unit increase in the absolute GFAP level, 95% CI 1.000–1.002, p = 0.048). Other plasma biomarkers did not predict recurrent stroke.
In the Kaplan–Meier plot (Figure 3), patients with the highest NfL tertile (>39 pg/mL), were found to be associated with higher risks of incident stroke (log rank p = 0.01) and ICH (p = 0.01). Similarly, patients with the highest GFAP tertile (>140 pg/mL) were at a risk of both incident stroke (p = 0.01) and ICH (p = 0.009).