Sociodemographic and Clinical Predictors of Cognitive Performance in Lupus

Objectives To identify sociodemographic, biological, and clinical determinants influencing longitudinal cognitive performance in systemic lupus erythematosus (SLE), evaluated globally and across distinct cognitive domains. Methods Adults with systemic lupus erythematosus (2019 EULAR/ACR criteria) at a tertiary lupus clinic completed repeated neuropsychological assessments using the American College of Rheumatology (ACR) Neuropsychological Battery. Test-level outcomes were standardized to z-scores and analyzed to retain domain-specific detail. Candidate predictors included sociodemographic, disease-related, and treatment variables such as age, sex, education, disease duration, depression (BDI-II), anxiety (BAI), and current use of glucocorticoids, antimalarials, immunosuppressants, or biologics. Multilevel Bayesian models with random effects for patient and domain/test accounted for repeated measures and test hierarchy. A conventional model was compared with a prespecified model applying selective shrinkage for weak associations, and model fit was evaluated using leave-one-out cross-validation (LOOIC) to assess predictive performance. Results The dataset included 245 patients contributing 8465 visits with repeated cognitive assessments over 4 years. The selective shrinkage model showed superior predictive performance and is presented here. Overall cognitive performance improved modestly over time (+0.10 z-score units; 95% CrI 0.06-0.14). Higher anxiety was associated with worse cognition (−0.010 per BAI point; 95% CrI −0.015 to −0.005). Increasing age related to poorer performance (−0.011 per year; 95% CrI −0.020 to −0.002), while longer disease duration showed a slight positive trend (0.019 per year; 95% CrI 0.006-0.030). Higher education corresponded to better scores (0.20; 95% CrI −0.01-0.50). Medication classes and SDI were not consistently associated with outcomes (Figure 1). Domain- and test-level effects revealed variability across and within domains, highlighting heterogeneous cognitive functioning in lupus. Figure 1. Adjusted associations between patient factors and cognitive performance. Posterior means (point estimate) with 95% credible intervals. Positive values indicate better cognition (higher z-scores), negative values worse cognition. Conclusion In this longitudinal SLE cohort, anxiety, age, and disease duration were the most consistent correlates of cognitive performance, independent of damage and medication class. Clinically, identifying and managing anxiety may yield greater benefits for cognitive health than medication adjustments alone. Modeling performance at the test level revealed differences both between domains and within tasks, indicating that cognitive function in lupus is not uniform. These findings highlight the need for individualized, domain-specific assessment and targeted interventions to address specific cognitive challenges in lupus.