Can conventional brain MRI support the attribution process in neuropsychiatric SLE? A multicentre retrospective study

Objectives: We aimed to investigate which elementary lesions, identified through conventional brain MRI, correlated with the attribution of neuropsychiatric (NP) manifestations of SLE as determined by clinical judgement (CJ) and a validated attribution algorithm (AA). Methods: We conducted a multicentre, retrospective cohort study of patients with SLE (1999-2018) from four tertiary SLE centres. Patients were assessed using American College of Rheumatology nomenclature and underwent MRI at their first NP event. NP manifestations were attributed to SLE using CJ and the AA. Elementary lesions were classified as follows: large infarcts, parenchymal haemorrhages, subarachnoid haemorrhages, inflammatory-type lesions, myelopathy, T2/fluid-attenuating inversion recovery (FLAIR) hyperintense lesions, lacunes, cerebral atrophy and microbleeds. Statistical analyses were performed using χ2 and Fisher's exact tests. Univariable and multivariable logistic regression models were performed. A sensitivity analysis was performed using a revised AA, which excluded the item 'presence of abnormal MRI' from the list of favouring factors. Results: Among 154 patients, 88 (57%) had NP events attributed to SLE by CJ and 85 (55%) by AA. MRI was normal in 57/154 (37%) cases, while T2/FLAIR hyperintense lesions were the most frequent findings (71/154, 46%). A normal MRI was more common in non-attributed NP events per CJ and AA (OR 0.42, 95% CI 0.21 to 0.82 and 0.27, 95% CI 0.13 to 0.52, respectively). Cerebral atrophy was more frequent in non-attributed events per CJ (adjusted OR 0.06, 95% CI 0.01 to 0.35), while inflammatory-type lesions were more prevalent in SLE-attributed events according to AA (OR 3.91, 95% CI 1.15 to 18.1), with no significant change in sensitivity analyses. Conclusions: Our study elucidates the role of conventional MRI findings in the attribution process in NPSLE. The presence of selected elementary lesions or, instead, their absence could have a relevant weight in assessing NP events. These findings may assist clinicians in achieving a more accurate attribution of NP manifestations.