Improvement in the Prediction of Cerebrovascular Events With White Matter Hyperintensity

BACKGROUND: It remains unclear if white matter hyperintensity (WMH) on magnetic resonance imaging adds relevant cerebrovascular prognostic information beyond vascular risk factors and demographics alone. METHODS AND RESULTS: We performed a post hoc analysis of hypertensive individuals in SPRINT-MIND (Systolic Blood Pressure Intervention Trial-Memory and Cognition in Decreased Hypertension). The primary outcome was incident stroke or cognitive impairment (mild cognitive impairment or dementia). We fit logistic regression models with the predictors of Atherosclerotic Cardiovascular Disease Risk Score, age, sex, race, education, current cigarette smoking, and the SPRINT-MIND randomization arm. WMH was subsequently included in the model to determine if it improved area under the receiver operating curve using the DeLong test. We used a structural equation model to determine the indirect effect on the primary outcome mediated through WMH. We included 727 individuals (mean age at baseline 67.7 +/- 8.4 years, 61.1% were men, 62.6% were non-Hispanic White, and mean years of follow-up was 3.6 +/- 0.9). Of the 727 individuals, 67 (9.2%) developed incident stroke or cognitive decline. The area under the receiver operating curve of the baseline model (without WMH) was 0.75 (95% CI, 0.70-0.81), and after the addition of WMH it increased to 0.81 (95% CI, 0.76-0.86) (P=0.004 for difference). The mediation analysis showed that 26.3% of the vascular risk's effect on the primary outcome is indirectly mediated through WMH. CONCLUSIONS: In adult hypertensive individuals, we found that the addition of WMH to models predicting incident stroke or cognitive impairment improved the prognostic ability above vascular risk and demographics alone to a level consistent with excellent prediction.

Automated summary

It is unclear whether WMH on MRI provides relevant cerebrovascular prognostic information beyond traditional risk factors in hypertensive patients. A post hoc analysis of the SPRINT-MIND study showed that the addition of WMH to prediction models improved the prognostic ability above vascular risk and demographics alone, with 26.3% of the effect on the primary outcome mediated through WMH.