Structural MRI profiles and tau correlates of atrophy in autopsy-confirmed CTE

Table 2 Summary of regression models comparing brain donors with CTE and participants with normal cognition on visual rating scales

Brain region	Est. marginal mean diff.	95% CI	FDR-adjusted P-value
Orbital-frontal cortex	1.29	0.52–2.06	0.009
Dorsolateral frontal cortex	1.31	0.42–2.19	0.013
Superior frontal cortex	1.05	0.15–1.96	0.046
Anterior temporal lobes	1.57	0.68–2.46	0.009
Parietal-occipital lobes	0.54	−0.48 to 1.57	0.375
Medial temporal lobes	1.60	0.25–2.95	0.046
Lateral ventricles	1.72	0.62–2.82	0.013
Third ventricle	0.80	0.26–1.35	0.013
Fourth ventricle	0.03	−0.05 to 0.10	0.501
Corpus callosum (genu + body + splenium)	1.13	−0.13 to 2.40	0.122
Periventricular white matter hyperintensities	0.28	−0.19 to 0.74	0.330
Deep white matter hyperintensities	0.14	−0.32 to 0.59	0.553
Total number of microbleeds	0.21	−0.25 to 0.67	0.428
Brain region	OR	95% CI	FDR-adjusted P-value
Cavum septum pellucidum	6.69	1.46-50.09	0.049

The majority consensus score among the raters was used; in the absence of a majority, the median was used. Sample sizes vary across regions due to missing data as result of missing sequences from the MRI scans (see Table 1). Orbital-frontal cortex, dorsolateral frontal cortex, superior frontal cortex, anterior temporal lobes, parietal-occipital lobes, medial temporal lobes, and lateral ventricles are a summary composite of left and right hemisphere 0 (none)–4 (severe) ratings (possible range 0–8). Each region of the corpus callosum (genu, body, splenium) were separately rated on the 0–4 scale and summed. Periventricular and deep white matter hyperintensities were rated on the 0–4 scale. Absence/presence of anterior and posterior cavum septum pellucidum were rated and combined into a single variable. Total number of microbleeds is a summary composite of microbleeds in all lobes. Linear regression models were used to compare brain donors with CTE and participants with normal cognition on each visually rated region with the exception of the CSP for which binary logistic regression was used. A separate model was performed for each region and all models controlled for age at the time of the MRI scan. The estimated marginal mean differences are differences between brain donors with CTE and participants with normal cognition for the given outcome adjusted for age at MRI scan. P-values were false discovery rate (FDR) adjusted using the Benjamini-Hochberg Procedure

ISSN: 1758-9193