Intraindividual cognitive function course over time in patients with systemic lupus erythematosus Authors: Oshrat E Tayer-Shifman, Jiandong Su, Kathleen Bingham, Mahta Kakvan, Maria Carmela Tartaglia, Lesley Ruttan, Sherief Marzouk, Joan Wither, May Y Choi, Dennisse Bonilla, Simone Appenzeller, Dorcas Beaton, Patricia Katz, Robin Green, Dafna D Gladman, Zahi Touma Published: 2025 ABSTRACT AbstractBackgroundCognitive impairment (CI) is one of the most common manifestations of neuropsychiatric systemic lupus erythematosus (SLE). This study aimed to characterise the course of CI over a 1-year period in patients with SLE and its associated factors.Methods175 adult SLE patients from the University of Toronto Lupus Clinic were assessed at baseline, 6 months and 12 months using the American College of Rheumatology Neuropsychological Battery. CI was classified based on standardised z-scores in cognitive domains. Patients were categorised as persistent-CI (CI at all three time-points; T0, T1 and T2), never-CI (no CI at any time-point) or fluctuating-CI (CI at 1–2 assessments). Sociodemographic, clinical, laboratory and medication data were collected at each visit. Patients with persistent-CI were compared with never-CI patients. CI severity was determined based on the mean z-score of tests across all six domains.ResultsOver 1 year, 46% of patients experienced CI, with 17% showing persistent-CI, 29% fluctuating-CI and 54% never-CI. Persistent-CI patients exhibited more severe CI compared with fluctuating-CI. The most frequently affected cognitive domains were learning and memory, simple attention and processing speed, and visual-spatial construction. Factors associated with CI persistence over 1 year included Black race, older age at SLE diagnosis, divorced/separated status at T0 and higher disease-related damage at T0.ConclusionThis study highlights the variable nature of CI in SLE patients, with most exhibiting a stable course over 1 year. Factors such as sociodemographic characteristics and comorbidities may influence CI persistence. WHAT IS ALREADY KNOWN ON THIS TOPIC Cognitive impairment (CI) is common in systemic lupus erythematosus (SLE), but its course varies across studies, and data on factors associated with its persistence remain limited. Cognitive impairment (CI) is common in systemic lupus erythematosus (SLE), but its course varies across studies, and data on factors associated with its persistence remain limited. WHAT THIS STUDY ADDS Over 1 year, the majority of SLE patients maintained stable cognitive function; among them, a minority experienced persistent impairment, and others retained normal cognitive function. About one-third fluctuated between impaired and non-impaired states. Over 1 year, the majority of SLE patients maintained stable cognitive function; among them, a minority experienced persistent impairment, and others retained normal cognitive function. About one-third fluctuated between impaired and non-impaired states. Patients with fluctuating cognitive function had milder impairment than those with persistent-CI. Patients with fluctuating cognitive function had milder impairment than those with persistent-CI. Older age at SLE diagnosis, Black race, divorced/separated status and higher disease-related damage were associated with the persistence of CI. Older age at SLE diagnosis, Black race, divorced/separated status and higher disease-related damage were associated with the persistence of CI. HOW THIS STUDY MIGHT AFFECT RESEARCH, PRACTICE OR POLICY Cognitive assessments should be incorporated into clinical practice for SLE patients, especially those at higher risk of persistent impairment, to improve early detection and management. Cognitive assessments should be incorporated into clinical practice for SLE patients, especially those at higher risk of persistent impairment, to improve early detection and management. Introduction Systemic lupus erythematosus (SLE) is a complex chronic autoimmune multisystem disease, which is highly heterogeneous in its pattern of organ involvement and prognosis.1 The nervous system is commonly affected by SLE and includes 19 manifestations that are defined as neuropsychiatric SLE (NPSLE).2 3 Cognitive impairment (CI) is one of the most common manifestations of NPSLE, with reported frequencies ranging from 20% to 80%.46 Differences in study populations, the neuropsychological assessment tools used, and the definition of CI all contribute to inconsistencies in the prevalence of CI across studies.4 In our systematic review from 2018, we found a pooled CI prevalence of 38% (95% confidence interval 33% to 43%).4 CI may occur in the absence of active lupus or other neuropsychiatric manifestations79 and significantly affects patients’ everyday functioning, employment status and health-related quality of life.1014 The most severe impairment is found in the cognitive domains of attention, memory, psychomotor speed and executive function, while the domains of language and visual-spatial formation are involved to a lesser extent.111517 The course of CI in lupus patients may be variable and has been described as improving, fluctuating in some patients and persistent in others. In the majority of patients, CI does not appear to deteriorate over time.4 18 19 However, the available information about the course of CI in SLE patients, the domains affected over time and the factors associated with CI persistence is limited. The majority of these results emerged from studies with small numbers of patients. The aim of this study was to characterise the course of CI over a 1-year period in patients with SLE, using the American College of Rheumatology Neuropsychological Battery (ACR-NB). We hypothesised that CI does not follow the same trajectory in all patients and would be associated with specific clinical characteristics. We will describe the intraindividual over time and between groups course of CI, its severity, the affected domains and the factors associated with CI persistence over a 1-year period using the ACR-NB. Methods Study population 175 adult SLE patients who attended the University of Toronto Lupus Clinic between July 2016 and February 2022 participated in the study. Inclusion criteria were: (a) fulfilment of the revised ACR criteria for SLE classification or three criteria and a supportive biopsy,1 (b) age between 18 and 65 and (c) ability to give informed consent. Exclusion criteria were: (a) mental or physical disability preventing participation in the study, and (b) lack of fluency in English or inability to complete the verbal items of the cognitive assessment.20 The project was approved by the University Health Network Research Ethics Board. Procedures and outcome measures All patients were administered the modified ACR-NB at baseline (T0), 6 months (T1) and 12 months (T2). As previously described,20 the modified battery consists of 11 cognitive tests representative of six cognitive domains: manual motor speed, simple attention and processing speed, visual-spatial construction, language processing, learning and memory, and executive functioning. In a significant portion of participants (n=43, 14.8%), data for the manual motor speed domain were unavailable as a result of difficulty in performing the tests involving finger tapping caused by hand pain or established joint deformities. As a result, the scores for the manual motor speed domain were excluded from the subsequent analysis. Patients’ scores on the ACR-NB were compared with normative data stratified for age and sex, yielding standardised z-scores that were used in analyses. The list of tests and corresponding references, which include the normative datasets, is provided in online supplemental Table S1. A domain was defined as impaired if a z-score of ≤−1.5 was reached in at least one test in the following domains: simple attention and processing speed, visual-spatial construction, and language processing or a z-score of ≤−1.5 in two or more tests in the following domains: learning and memory and executive functioning. Patients were classified on the ACR-NB as cognitively impaired at each time point if two or more domains were impaired. The patients were categorised into three groups according to their CI course over the study period: (a) persistent-CI if they had CI across all three time-points (T0, T1 and T2) over 1 year; (b) never-CI, if they did not have CI in any time-point and (c) fluctuating-CI if they had CI at only 1–2 time-points. This classification was adapted from the longitudinal framework proposed by Hanly et al,18 who described five categories of cognitive status (‘never impaired’, ‘stable impaired’, ‘resolved’, ‘emerged’ and ‘fluctuated’). We combined the latter three transient categories into a single fluctuating-CI group to reflect any non-persistent CI. The severity of CI was determined based on the mean z-scores of all tests within the six domains. Disease activity was measured by the SLE Disease Activity Index 2000 (SLEDAI-2K) 30 days21 at T0 and as adjusted mean SLEDAI-2K from the first clinic visit to the baseline visit of this study. Cumulative SLE disease manifestations (by SLEDAI-2K organ system) were collected from the first clinic visit to the baseline visit of this study. Damage was measured at baseline (T0) by the Systemic Lupus International Collaborating Clinics/ACR Damage Index (SDI), excluding the CI item. Depression was measured at T0 by the Beck Depression Inventory-II, with a cut-off score of ≥18 used to define depression,22 and anxiety at T0 by the Beck Anxiety Inventory, with a cut-off score of ≥22 used to define anxiety.22 Fatigue was measured at T0 by the Fatigue Severity Score and pain at T0 by the transformed scale of the bodily pain domain of the SF-36. Available brain imaging, including CT, MRI and brain scan/Single Photon Emission Computed Tomography, was reviewed and categorised as normal or abnormal. Statistical analyses Demographic and clinical characteristics were summarised, and statistical significance was set at an alpha level of p<0.05. Patients with persistent-CI were compared with never-CI patients using a t-test, χ2 test and Cochran-Armitage trend test for categorical variables in more than two classes. Patients with fluctuating-CI were excluded from the comparison to focus on two well-defined groups: never-CI and persistent-CI. Results A total of 175 patients completed all three neuropsychological assessments at T0, T1 and T2 over a 1-year period. The demographics and clinical characteristics of the cohort are presented in table 1. Table 1Baseline demographics of patientsTotal, N=175Female, sex, n (%)156 (89.1%)Age at enrolment, mean±SD, years41.2±12.6Age at SLE diagnosis, years27.2±10.8Disease duration at enrolment, mean±SD, years14.0±10.4Race, n (%) White103 (58.9%) Black35 (20.0%) Chinese18 (10.2%) Others19 (10.9%)SLE, systemic lupus erythematosus. Baseline demographics of patients SLE, systemic lupus erythematosus. Prevalence of CI over time: CI was present in 46% of patients (81/175) in at least one assessment over 1 year. The prevalence of CI varied between assessments, with 35% at T0, 28% at T1 and 29% at T2. Course of CI over time: 30 patients (17%) demonstrated persistent-CI over 1 year, 51 patients (29%) demonstrated fluctuating-CI, and 94 patients (54%) never experienced CI over the three assessments and thus were defined as never-CI (figure 1). CI course over three time points in 1 year, n=175. CI, cognitive impairment. Severity of CI over time: the persistent-CI patients as a group (figure 2a) displayed lower mean z-scores (and non-overlapping standard errors), implying more severe CI than the fluctuating-CI patients, and both groups had worse cognitive function than the never-CI patients at each time point: T0, T1 and T2. The average z-scores of the six cognitive domains for every individual patient, demonstrating almost complete separation of the patients according to the different groups, are shown in figure 2b. Average z-scores of the six domains over three time points according to CI course group: persistent-CI, fluctuating-CI and never-CI by (a) group with SE bars and by (b) individual patient. CI, cognitive impairment. Domains affected over time: in the persistent-CI group, the most affected domains over time were learning and memory, impaired in 70% to 77% of patients across the three time points, and visual-spatial construction, impaired in 67% to 80% of patients. The domain of simple attention and processing was impaired in 50% to 60% of patients. Executive functioning and language processing were not considerably affected, with impairment only in 30% to 40% and 30% to 33% of patients, respectively. There was no marked change over time in the domains involved in the persistent-CI group, and the affected domains were very similar across T0, T1 and T2 (figure 3a). A similar pattern of the impaired domains was seen in the fluctuating-CI group (figure 3b), as the most affected domains were learning and memory (affected in 37% to 61% of patients, and visual-spatial construction (affected in 41% to 55%), followed by simple attention and processing (affected in 20% to 24%). Similar to the persistent-CI group, the domains of executive functioning and language processing were not considerably affected (impaired in 8% to 12% and 14% to 22% of patients, respectively). There was more variation in the degree of impairment of the involved domains over time as compared with those with persistent-CI, as can be seen, for example, in the learning and memory domain (with 61% of impairment in T0 as opposed to 37% in T1). Impaired domains were also observed in the never-CI group (figure 3c), mostly in learning and memory and visual-spatial construction domains (impaired in up to 19% and 20%, respectively), as impairment in one domain was not sufficient for CI classification. Prevalence of domain impairment and CI in patients according to time point for the three CI course groups: (a) persistent-CI, (b) fluctuating-CI and (c) never-CI. CI, cognitive impairment. Factors associated with CI persistence: demographics (table 2) and clinical characteristics (table 3 and online supplemental tables S2–S5) were compared between persistent-CI (n=30) and never-CI patients (n=94). A significantly higher proportion of patients of Black race (37% vs 14%, p=0.04) and divorced/separated patients (21% vs 5%, p=0.036) were found in the persistent-CI group compared with never-CI group. There was also a trend for lower education level (27% vs 16%, p=0.198) and almost twice as many patients who were disabled or on sick leave (30% vs 17%, p=0.309) among patients with persistent-CI compared with never-CI. Even though the age at T0 was comparable and did not differ significantly statistically, patients with persistent-CI were significantly older at SLE diagnosis (31±15 vs 26±9, p=0.02). Patients with persistent-CI also had a significantly higher level of SLE damage other than the CI, as measured by the SDI (1.6±1.7 vs 0.8±1.3, p=0.017). Examining the specific SDI domains revealed that patients with persistent-CI had significantly higher levels of damage only in the neuropsychiatric domain (10 (33.3%) vs 14 (14.9%), p=0.026) compared with those with never-CI (online supplemental Table S3). Additionally, a significantly higher proportion of cardiovascular events (23% vs 5%, p=0.004) and abnormal brain imaging (30% vs 16%, p=0.009) were observed, although a small number of patients were in each group. There was a trend for a higher prevalence of anxiety among patients in the persistent-CI group. There were no differences in disease activity (measured either as SLEDAI-2K at study baseline or as adjusted mean SLEDAI-2K), medical treatment (including glucocorticoids, antimalarials, immunosuppression, anticoagulation, psychotropic and pain medications), SLE manifestations (except for skin involvement that was more prevalent among persistent-CI patients), other comorbidities, or immunological profile (online supplemental Tables S2-S5). A multivariate analysis did not identify any variables as independently associated with CI persistence. Table 2Demographics of persistent-CI versus never-CI patientsNever-CIPersistent-CIP value(N=94)(N=30)Sex, n (%) 0.48 Female83 (88.3%)25 (83.3%) Male11 (11.7%)5 (16.7%) Age at T0 (first cognitive assessment), years, mean±SD40.2±12.042.9±12.20.278Race 0.04 White59 (64.1%)15 (50.0%) Black13 (14.1%)11 (36.7%) Chinese9 (9.8%)3 (10.0%) Others11 (12.0%)1 (3.3%) Highest education level, n (%) 0.198 Grade eight or high school15 (16.1%)8 (26.7%) College or university78 (83.9%)22 (73.3%) Employment, n (%) 0.309 Employed55 (58.5%)12 (40.0%) Disabled/sick leave16 (17.0%)9 (30.0%) Student10 (10.6%)3 (10.0%) Retired/homemaker/looking for work/other13 (13.8%)6 (20.0%) Marital status, n (%) 0.036 Single45 (47.9%)13 (44.8%) Married/common law44 (46.8%)10 (34.5%) Widowed/divorced/separated5 (5.3%)6 (20.7%) Bold values indicate statistically significant differences (p<0.05). Demographics of persistent-CI versus never-CI patients Bold values indicate statistically significant differences (p<0.05). Table 3Clinical characteristics of persistent-CI versus never-CI patients at T0Never-CI(n=94)Persistent-CI(n=30)P valueAge at SLE diagnosis, years, mean±SD25.8±9.031.1±14.7 0.02 SLE disease duration, years, mean±SD14.4±11.011.8±8.70.251Number of ACR criteria, mean±SD6.3±1.36.3±1.40.825SDI score,* mean±SD0.8±1.31.6±1.7 0.017 SLEDAI-2K, mean±SD3.1±3.43.6±3.20.545Adjusted mean SLEDAI-2K,† mean±SD4.1±3.34.4±3.60.665Comorbidities Depression (BDI-II), n (%)23 (29.1)9 (36.0)0.516 Mean score±SD14.6±12.217.5±14.50.324Anxiety (BAI), n (%)19 (23.5)9 (36.0)0.214 Mean score±SD14.5±10.819.6±14.40.056FSS, mean±SD4.2±1.84.8±1.70.113Pain score,‡ mean±SD63.4±26.152.5±29.50.086Fibromyalgia, ever diagnosed, n (%)30 (31.9)10 (33.3)0.885Antiphospholipid syndrome, n (%)11 (11.7)3 (10.0)0.798CVE, n (%) ever5 (5.3)7 (23.3) 0.004 Diabetes, n (%)0 (0)0 (0)N/AHypertension, ever, n (%)38 (40.4)15 (50.0)0.356Dyslipidaemia, ever, n (%)56 (59.6)19 (63.3)0.714Smoking, past 5 years, n (%)15 (16.0)4 (13.3)0.728Abnormal brain imaging, ever, n (%)15 (16.0)9 (30.0)0.09 Head CT scan2 (2.1)5 (16.7) 0.003 Brain MRI11 (11.7)7 (23.3)0.115 Brain scan or SPECT7 (7.4)2 (6.7)0.886Bold values indicate statistically significant differences (p<0.05).All variables measured at study baseline unless stated otherwise.Hypertension is defined as blood pressure of >140/90 or antihypertensive treatment; dyslipidaemia is defined as abnormal total cholesterol or on specific treatment.BAI cut-off score of ≥22 was used to define anxiety.BDI-II cut-off score of ≥18 was used to define depression.CVE including stroke, transient ischaemic attack, myocardial infarction or atherosclerotic congestive heart failure.SDI excluding the cognition component.*From the initial clinic visit to the baseline visit of this study.†Excluding CI component.‡Transformed score from the bodily pain domain of the 36-Item Short Form Survey, a low score corresponds to higher pain reported.ACR, American College of Rheumatology; BAI, Beck Anxiety Inventory; BDI-II, Beck Depression Inventory-II; CI, cognitive impairment; CVE, cardiovascular event; FSS, fatigue severity score; SDI, Systemic Lupus International Collaborating Clinics/American College of Rheumatology Damage Index; SLE, systemic lupus erythematosus; SLEDAI-2K, SLE Disease Activity Index 2000; SPECT, Single Photon Emission Computed Tomography. Clinical characteristics of persistent-CI versus never-CI patients at T0 Bold values indicate statistically significant differences (p<0.05). All variables measured at study baseline unless stated otherwise. Hypertension is defined as blood pressure of >140/90 or antihypertensive treatment; dyslipidaemia is defined as abnormal total cholesterol or on specific treatment. BAI cut-off score of ≥22 was used to define anxiety. BDI-II cut-off score of ≥18 was used to define depression. CVE including stroke, transient ischaemic attack, myocardial infarction or atherosclerotic congestive heart failure. SDI excluding the cognition component. From the initial clinic visit to the baseline visit of this study. Excluding CI component. Transformed score from the bodily pain domain of the 36-Item Short Form Survey, a low score corresponds to higher pain reported. ACR, American College of Rheumatology; BAI, Beck Anxiety Inventory; BDI-II, Beck Depression Inventory-II; CI, cognitive impairment; CVE, cardiovascular event; FSS, fatigue severity score; SDI, Systemic Lupus International Collaborating Clinics/American College of Rheumatology Damage Index; SLE, systemic lupus erythematosus; SLEDAI-2K, SLE Disease Activity Index 2000; SPECT, Single Photon Emission Computed Tomography. Discussion This study is one of the largest and most comprehensive studies to date examining the course of CI in patients with SLE using the ACR-NB, the domains impaired over time and the factors associated with CI persistence. When assessed over three time points in 1 year, in a cohort of SLE patients with an average of 14.0±10.4 years of disease, 46% of SLE patients exhibited CI at some point. More than two-thirds of the patients had a stable cognitive function over 1 year, including 54% who never experienced CI and 17% with persistent-CI. Less than one-third of patients were fluctuating between states. Previous longitudinal studies on the course of CI in SLE patients have shown inconsistent results. These studies were limited by small sample size and, more importantly, by different neuropsychological assessment tools used for CI measurement and definition, which vary in specificity and sensitivity, as well as the cognitive domains assessed; thus, a direct comparison between the results of these studies is not possible.23 It is important to emphasise that, in contrast to our study, none of these studies (except one small study involving 15 patients)24 employed the recommended ACR-NB for CI evaluation. Several small studies showed resolution or fluctuation of CI over time. Hanly et al reported CI in 21% of 70 SLE patients, with persistent-CI in only 5% of 59 SLE patients (3 patients) who were re-evaluated after 1 year.19 In a 5-year follow-up study of the same cohort,18 CI persisted in only 4% of CI patients (2 patients) of 47 patients who completed all three assessments. Ceccarelli et al25 followed up 43 SLE patients with a baseline CI prevalence of 21%, reporting improvement in cognitive function in 50% of them after 10 years, while only 10% of patients experienced cognitive decline at follow-up. Hay et al26 reported that only one of nine patients with CI at baseline was still impaired at a second interview, with a tendency for CI improvement in patients whose psychiatric disorders had resolved. In contrast, several other studies were consistent with our study findings, reporting stable cognitive function over time. A small study by Kozora et al24 studying 15 SLE patients with high CI prevalence of 78% at baseline showed that most patients had stable cognitive function, while only 27% were fluctuating. Carlomagno et al27 followed up 51 SLE patients with a CI prevalence of 30% at baseline and showed that most CI patients showed persistence of CI (27% of patients in the cohort) and only 10% fluctuated over time. Waterloo et al28 reported stable cognitive function in 28 SLE patients after 5 years.1 Langensee et al29 reported stable cognitive function in 52 SLE patients re-evaluated after 50 months. Perera et al,30 who studied 1281 SLE patients, reported that most participants demonstrated stability in cognition on repeated assessments over a 7‐year assessment period. This large study used only two tests, the Hopkins Verbal Learning Test-Revised and the Controlled Oral Word Association Test. These studies are in agreement with our results, which show that only 27% of patients fluctuate over time. This study provides insight into the degree of CI and its association with cognitive status change/stability over time. Our results show that fluctuating-CI patients displayed less severe CI, compared with persistent-CI patients, as measured by the mean z-scores of all tests within the six domains. It seems that some of the fluctuation reflects less severe or borderline CI with small variations over time in cognitive function around the cut-off for the definition of CI. In keeping with previous studies, the domains most affected in our patients were learning and memory and simple attention and processing. Visual-spatial construction was also commonly impaired, although in other studies, it was involved less often.11 15 16 The same domains: learning and memory, visual-spatial construction and simple attention and processing were predominantly affected in the fluctuating- and persistent-CI groups. However, fluctuating-CI patients displayed more variation in the affected domains over time compared with persistent-CI patients. The factors contributing to the variation were not evaluated in this study; however, changes in disease activity, comorbidities such as depression and anxiety, and treatment may be associated with cognitive function and specific cognitive domains. A previous study, which assessed patients with inactive disease, showed that disease activity (not neuropsychiatric) at presentation was related to CI.31 Our previous study on the longitudinal relationships between depression and anxiety symptoms in SLE and cognitive function found that depression and anxiety were stable across time and most strongly related to cognitive domains of verbal and visual memory and working memory, with higher depression and anxiety scores associated with poorer test performance.32 In a cross-sectional study, higher levels of depressive symptoms have been found to be associated with lower performance on the cognitive domains of attention and memory,33 and a 2-year prospective study in 49 SLE patients26 found that resolution of depression or anxiety disorders was associated with improvement in memory and verbal fluency. Patients of Black race and divorced/separated status were factors associated with CI persistence in our study, as was older age at SLE diagnosis. A trend of association of persistent-CI with lower education level and being disabled or on sick leave was also observed. Previous observational studies reported an association between sociodemographic characteristics and cognitive function, with a higher prevalence of CI among Blacks and Hispanics,23 34 35 and those with lower income,35 lower education level23 30 36 and unemployment status,11 37 similar to our results. The reasons behind the increased rates of CI persistence among Black patients remain unclear.17 Whether it is mediated by socioeconomic factors or the quality of received healthcare that was not accounted for in this study, or if it is linked to genetic predisposition affecting the severity of the disease itself, these findings warrant cautious interpretation. The multivariate analysis did not identify any sociodemographic or clinical variables as significantly associated with CI persistence. This is likely because the development of CI persistence is a complex, multifaceted process that cannot be attributed to just a few factors. We showed that patients with persistent-CI had a significantly higher overall level of SLE damage (other than the CI), and particularly, within the neuropsychiatric domains. An association between CI in SLE and overall damage was found in previous studies.25 35 38 39 Interestingly, two studies38 40 identified the cardiovascular domain of the SDI to be associated with severe CI. However, the small number of patients with cardiovascular damage in each group in our study limits the ability to draw definitive conclusions. The pathogenesis of NPSLE, including CI, is complex and involves both inflammatory/immune-mediated and thrombotic/ischaemic pathways.3 7 The present study does not attempt to link the course of CI to specific underlying mechanisms. Structural neuroimaging and neuropathological studies suggest that CI may be associated with microvascular brain injury, including diffuse microinfarcts, microhaemorrhages, vasculopathy and endothelial hyperplasia, much of which remains subclinical and detectable only through imaging.41 In this context, it is noteworthy that patients with persistent-CI in our cohort exhibited a higher proportion of abnormal brain imaging findings, despite a similar frequency of clinically evident CNS involvement. To further explore potential factors contributing to fluctuations in cognitive performance, we performed exploratory linear mixed-model analyses (data not shown). These analyses examined whether changes in cognitive function over time were associated with disease activity, symptom measures (depression, anxiety, fatigue and pain), serological markers or treatment exposure. No consistent associations were identified between cognitive changes and SLE activity or serologic indices. Modest relationships were observed between CI and fatigue and depressive symptoms in univariable models, which were not sustained after multivariable adjustment. In domain-specific analyses, depression was negatively associated, and immunosuppressive use was positively associated with performance in the learning and memory domain, while psychotropic medication use and corticosteroid dose were positively associated with visual-spatial construction scores. These findings likely reflect the multifactorial and heterogeneous nature of CI in SLE and underscore the complexity of mechanisms underlying cognitive function fluctuations, which are presumably influenced by lupus-specific, psychological and treatment-related factors. Although cognitive status was assessed at three time points, a limitation of our study includes a relatively short follow-up of only 1 year, limiting the ability to draw conclusions on the course of cognitive function in SLE patients for a longer period. Additionally, it is conceivable that some of the patients who were classified as never-CI patients may have had CI in the past. Three cognitive assessments conducted 6 months apart enabled us to capture the fluctuating nature of CI in SLE patients. However, it is possible that some practice effects might have been responsible for the increase in the average z-scores over time among patients from the three CI groups (figure 2a). An additional limitation is the small size of the persistent-CI group, which restricts our ability to detect statistically significant differences from the never-CI group and restricts the ability of the multivariate analysis to determine the independent factors associated with CI persistence. The focus of our study was on the intraindividual course of cognitive function over time in patients with SLE. Although including a healthy or disease control group would have strengthened the evaluation of the modified ACR-NB’s precision, we addressed this by using population-based normative data to calculate z-scores. In addition, the close similarity between the modified ACR-NB and the standard ACR-NB supports the validity of our assessment approach in patients with SLE.20 Summary More than half of patients with SLE experienced CI at some point over 1 year. Most of the patients exhibit a stable course of cognitive function over time, with only 29% of patients fluctuating between CI and non-CI states. The persistent-CI patients group had more severe CI, while the fluctuating patients had milder CI. The same domains are affected in both fluctuating-CI and persistent-CI groups. Disease severity, as manifested by higher disease-related damage, sociodemographic and comorbidities, may all contribute to CI persistence. Our results underscore the need for further research into the factors associated with CI persistence in SLE patients to better target those at greatest risk and develop tailored treatments. Supplementary material