CSF Biomarkers Predict Gait Outcomes in Idiopathic Normal Pressure Hydrocephalus

Abstract

Background and objectives: The assessment of biomarkers in selecting patients with idiopathic normal pressure hydrocephalus (iNPH) for shunt surgery has been limited to small cohort studies and those with limited follow-up. We assessed the potential for CSF biomarkers in predicting immediate response to CSF tap test (TT) and long-term response after shunt surgery.

Methods: CSF was obtained from patients with iNPH referred for CSF TT after baseline assessment of cognition and gait. CSF neurofilament light (NfL), β-amyloid 42 (Aβ1-42), β-amyloid 40 (Aβ1-40), total tau (tTau), and phosphorylated tau 181 (pTau181) and leucine-rich alpha-2-glycoprotein-1 (LRG1) were measured by ELISA. The ability of these measures to predict immediate improvement following CSF TT and long-term improvement following shunt surgery was compared by univariate and adjusted multivariate regression.

Results: Lower NfL, pTau181, tTau, and Aβ1-40 were individually predictive of long-term improvement in gait outcomes after shunt surgery. A multivariate model of these biomarkers and MRI Evans index, adjusted for age, improved prediction (area under the receiver operating curve 0.76, 95% confidence interval 0.66-0.86). tTau, pTau181, and Aβ1-40 levels were statistically different in those whose gait improved after CSF TT compared with those who did not. Using a multivariate model, combining these markers with Evans index and transependymal flow did not significantly improve prediction of an immediate response to CSF TT.

Discussion: A combination of CSF biomarkers can predict improvement following shunt surgery for iNPH. However, these measures only modestly discriminate responders from nonresponders following CSF TT. The findings further suggest that abnormal CSF biomarkers in nonresponders may represent comorbid neurodegenerative pathology or a predegenerative phase that presents with an iNPH phenotype.

Copyright © 2022 The Author(s). Published by Wolters Kluwer Health, Inc. on behalf of the American Academy of Neurology.

Figures

Figure 1. The Plot Presents the ROC Curves Generated From the Univariate and the Multivariate Regression (Red) Models Listed in Table 2

The multivariate model uses LRG, pTau, Aβ1–40, Evans index, and transependymal flow as predictors and has the greatest AUC of 0.64 (95% CI 0.58–0.70). Aβ1–40 = β-amyloid 40; AUC = area under the receiver operating curve; CI = confidence interval; LRG = leucine-rich alpha-2-glycoprotein; pTau = phosphorylated tau.

Figure 2. The Plot Presents the ROC Curves Generated From the Univariate and Multivariate (Red) Regression Models Listed in Table 3

The multivariate model uses age, NfL, pTau, and LRG as predictors and has the greatest AUC of 0.76 (95% CI 0.66–0.86). Aβ1–40 = β-amyloid 40; AUC = area under the receiver operating curve; CI = confidence interval; LRG = leucine-rich alpha-2-glycoprotein; NfL = neurofilament light; pTau = phosphorylated tau; tTau = total tau.

Figure 3. Variable Importance Plot From Nonsupervised Random Forest Algorithm for Improvement After the TT Procedure

The mean decrease in accuracy attributed to a variable is determined during the classification error calculation phase. The more the accuracy of the random forest decreases due to the exclusion (or permutation) of a single variable, the more important that variable is deemed, and therefore, variables with a large mean decrease in accuracy are more important for classification of the outcome. The mean decrease in Gini coefficient is a measure of how each variable contributes to the homogeneity (purity) of the nodes and leaves in the resulting random forest. The Gini coefficient is a measure of homogeneity from 0 (homogeneous) to 1 (heterogeneous). The changes in Gini are summed for each variable and normalized at the end of the calculation. Variables that result in nodes with higher homogeneity have a higher decrease in Gini coefficient. Aβ1–40 = β-amyloid 40; Aβ1–42 = β-amyloid 42; LRG = leucine-rich alpha-2-glycoprotein; MoCA = Montreal Cognitive Assessment; NfL = neurofilament light; pTau = phosphorylated tau; TT = tap test; tTau = total tau.

Figure 4. Variable Importance Plot From Nonsupervised Random Forest Algorithm for Improvement After Shunt Surgery

Aβ1–40 = β-amyloid 40; Aβ1–42 = β-amyloid 42; LRG = leucine-rich alpha-2-glycoprotein; MoCA = Montreal Cognitive Assessment; NfL = neurofilament light; pTau = phosphorylated tau; TT = tap test; tTau = total tau; TUG = Timed Up and Go.