Prediction of dementia using CT imaging in stroke (PRODUCTS)

A better understanding of who will develop dementia can inform patient care. Although MRI offers prognostic insights, access is limited globally, whereas CT-imaging is readily available in acute stroke. We explored the prognostic utility of acute CT-imaging for predicting dementia.

We included stroke or transient ischaemic attack (TIA) survivors from participating stroke centres in Scotland. Acute CT-scans were rated using ordinal scales for neurodegenerative and cerebrovascular changes (old infarcts, white matter lesions (WMLs), medial temporal lobe atrophy (MTA), and global atrophy (GA)) and combined together to a 'brain-frailty' score. Dementia status was established at 18-months following stroke or TIA.

Among 195 participants, 33% had dementia after 3 years of follow-up. High brain-frailty score (⩾2/4) correlated with higher risk of dementia (HR (95% CI) 6.02 (1.89-19.21)). As individual predictor, severe MTA was most strongly associated with dementia (adjusted HR (95% CI) 2.09 (1.07-4.08)). Other predictors associated with dementia included older age, higher prestroke morbidity (mRS), WMLs, and GA. Integrated in a prediction model with clinical parameters, prestroke mRS, cardiovascular disease, GA, MTA and Abbreviated-Mental-Test were the strongest predictors of dementia (c-statistic: 0.77).

Increased brain-frailty, and its individual components (WMLs, MTA, and GA) are associated with a higher risk of dementia in participants with stroke. Combining clinical and brain-frailty parameters created a moderate dementia prediction model but added little value over clinical parameters in combination with cognitive testing. CT-based brain-frailty may provide better prognostic insights when cognitive testing isn't feasible and for identifying highest-risk individuals for dementia prevention trials to increase trial efficiency.