Pathologically based criteria to distinguish essential tremor from controls: analyses of the human cerebellum

Abstract

Objective: Essential tremor is among the most prevalent neurological diseases. Diagnosis is based entirely on neurological evaluation. Historically, there were few postmortem brain studies, hindering attempts to develop pathologically based criteria to distinguish essential tremor from control brains. However, an intensive effort to bank essential tremor brains over recent years has resulted in postmortem studies involving >200 brains, which have identified numerous degenerative changes in the essential tremor cerebellar cortex. Although essential tremor and controls have been compared with respect to individual metrics of pathology, there has been no overarching analysis to derive a combination of metrics to distinguish essential tremor from controls. We asked whether there is a constellation of pathological findings that separates essential tremor from controls, and how well that constellation performs.

Methods: Analyses included 100 essential tremor brains from the essential tremor centralized brain repository and 50 control brains. A standard tissue block from the cerebellar cortex was used to quantify 11 metrics of pathological change. Three supervised classification algorithms were investigated, with data divided into training and validation samples.

Results: Using three different algorithms, we illustrate the ability to correctly predict a diagnosis of essential tremor, with sensitivity and specificity >87%, and in the majority of situations, >90%. We also provide a web-based application that uses these metric values, and based on specified cutoffs, determines the likely diagnosis.

Interpretation: These analyses set the stage for use of pathologically based criteria to distinguish clinically diagnosed essential tremor cases from controls, at the time of postmortem.

Figure 1

Cerebellar pathology in control versus ET brains characterized by 11 morphological metrics. (A, B) Purkinje cell (PC) count (per mm) in (A) control versus (B) ET, LH&E stain. The linear density of PCs (arrows) is lower in ET, indicating PC loss. (C, D) CB‐GAD % empty baskets in (C) control versus (D) ET. Dual immunostain for calbindinD_28k (CB) and glutamic acid decarboxylase (GAD) identifies basket cell plexuses that lack a PC soma (“empty baskets,” arrows), an indirect measure of PC loss. (E, F) Basket cell rating, Bielschowsky stain (E, control, basket cell rating = 1), (F, ET, basket cell rating = 3). The basket cell plexus is hypertrophic in ET with coarse and thickened processes. (G) Torpedo in ET, LH&E stain, an oval swelling of the PC axon (arrow). (H) Heterotopic PC in ET, LH&E stain, with a PC soma displaced into the molecular layer (arrow). (I) PC dendritic swelling in ET (arrow), a focal swelling of the PC dendrite in the molecular layer, Bielshowsky stain. (J, K) Vesicular glutamate transporter type 2 (VGLut2) labeled CFs in outer 20% of molecular layer (arrows; outer 20% demarcated by line) are increased in (K) ET versus (J) controls. (L, M) VGlut2 CF synaptic density (arrows, synaptic puncta) is decreased in (M) ET versus (L) controls. (N–P) CB immunostain in 100 μm thick tissue sections. In ET (O, P), CB torpedoes (arrows), CB torpedo with recurrent axon (black carets), and CB PC thickened axons (arrowheads) are more commonly present than in (N) control brains. Scale bar = 100 μm in A, B, J, K, N–P; 50 μm in C–I; 25 μm in L, M.

Figure 2

Principal component analysis (PCA) of cerebellar morphologic metrics in ET versus controls. (A, B) PCA of the z‐scored raw data using (A) 11 pathological metrics (including calbindinD_28k [CB] in thick tissue sections) or (B) 8 pathological metrics (paraffin tissue analyses only) demonstrates segregation of ET cases (red circles, n = 100) from controls (blue circles, n = 50). (C, D) Mean PC1 and PC2 values with standard error of the mean for control and ET cases from the (C) 11 metric (panel A) and (D) 8 metric (panel B) PCA demonstrate significantly different PCA values (Kruskal–Wallis test) among ET and controls. CTL, control; PC1, principal component 1; PC2, principal component 2. **p < 0.01; ****p < 0.0001.

Figure 3

Logistic regression with a ridge penalty algorithm. Standardized coefficient estimates for each metric are shown for (A) the total sample and (B) the age‐matched sample, using the 11 or 8 metrics of cerebellar pathology. The magnitude of the horizontal bars (from zero) depicts the impact of each metric on the odds of an ET diagnosis. Positive bars indicate an increase in the odds of an ET diagnosis for greater values of the metric and negative bars indicate an increase in the odds of an ET diagnosis for lower values of the metric.