Abstract Body

Older persons with HIV (PWH) undergo pathological changes to cerebral white matter, detectable on diffusion-weighted MRI. These changes can be quantified using the brain age gap (BAG), the difference between true age and neuroimaging-predicted ‘brain age.’ It is hypothesized that white matter aging, i.e. progressive accumulation of microstructural damage, is accelerated in PWH with detectable viral load. However, other risk factors for white matter aging remain uncertain.

290 PWH (age=48.2±13.7 yr.; 23% female; 65% African-American; 85% virally suppressed) and 165 HIV- controls (age=37.5±16.3 yr.; 47% F; 56% A.A.) provided informed consent and were imaged at a single site. Age, race, and sex were co-variates in all analyses. Diffusion MRI was performed on a 3-Tesla Siemens scanner (TR/TE=9100/104ms, resolution=2x2x2mm, gradient directions=30). Potential correlates of HIV severity and white matter aging were quantified: plasma viral load, current CD4+ lymphocytes, nadir CD4+, hepatitis C, body-mass index, 10-year Framingham cardiovascular risk score, and global cognitive z-score. A Gaussian process regression model was trained to predict age from diffusion MRI scalars (diffusivity and fractional anisotropy) using 624 publicly available healthy controls from the Cambridge Center for Aging and Neuroscience. The trained model was then applied to this cohort to quantify BAG ([b]Panel A[/b]). To test for accelerated white matter aging, BAG was modeled as an interaction between viral load and age. Potential risk factors for white matter aging were identified using generalized linear regression with backwards variable selection and a stop criterion of p<0.10.

The trained machine learning model predicted age in previously unseen participants with a mean absolute error of 5.3 years after linear de-trending (R2=0.48; p<0.001). Age and viral load (>50 copies/mL) had a significant interactive effect on BAG, such that PWH with detectable viral load accumulated +1.5 years of additional BAG per decade vs. HIV- controls (p=0.020; Panel B). Backwards selection identified two features associated with advanced white matter aging: greater 10-year Framingham risk score (p=0.002) and lower global cognitive z-score (p=0.043).

Aging with detectable plasma HIV and elevated cardiovascular risk are significant correlates and potential risk factors for white matter pathology, and may contribute to the etiology of cognitive impairment in PWH.