Quantification of Silent Cerebral Infarction on High-Resolution FLAIR and Cognition in Sickle Cell Anemia

Abstract

Research in sickle cell anemia (SCA) has used, with limited race-matched control data, binary categorization of patients according to the presence or absence of silent cerebral infarction (SCI). SCI have primarily been identified using low-resolution MRI, with radiological definitions varying in lesion length and the requirement for abnormality on both fluid attenuated inversion recovery (FLAIR) and T1-weighted images. We aimed to assess the effect of published SCI definitions on global, regional, and lobar lesion metrics and their value in predicting cognition. One hundred and six patients with SCA and 48 controls aged 8–30 years underwent 3T MRI with a high-resolution FLAIR sequence and Wechsler cognitive assessment. Prevalence, number, and volume of lesions were calculated using a semi-automated pipeline for SCI defined as: (1) Liberal: any length (L-SCI); (2) Traditional: >3 mm in greatest dimension (T-SCI); (3) Restrictive; >3 mm in greatest dimension with a corresponding T1-weighted hypo-intensity (R-SCI). Globally, as hypothesized, there were large effects of SCI definition on lesion metrics in patients and controls, with prevalence varying from 24–42% in patients, and 4–23% in controls. However, contrary to hypotheses, there was no effect of any global metric on cognition. Regionally, there was a consistent distribution of SCI in frontal and parietal deep and juxta-cortical regions across definitions and metrics in patients, but no consistent distribution in controls. Effects of regional SCI metrics on cognitive performance were of small magnitude; some were paradoxical. These findings expose the challenges associated with the widespread use of SCI presence as a biomarker of white-matter injury and cognitive dysfunction in cross-sectional high-resolution MRI studies in patients with SCA. The findings indicate that with high-resolution MRI: (1) radiological definitions have a large effect on resulting lesion groups, numbers, and volumes; (2) there is a non-negligible prevalence of lesions in young healthy controls; and (3) at the group-level, there is no cross-sectional association between global lesion metrics and general cognitive impairment irrespective of lesion definition and metric. With high-resolution multi-modal MRI, the dichotomy of presence or absence of SCI does not appear to be a sensitive biomarker for the detection of functionally significant pathology; the search for appropriate endpoints for clinical treatment trials should continue.