Attribution in the assessment of nervous system disease in SLE

Giving credit where credit is due This editorial refers to Development and validation of a new algorithm for attribution of neuropsychiatric events in systemic lupus erythematosus, by Alessandra Bortoluzzi et al., doi: 10.1093/rheumatology/keu384, pages 891–8. SLE is a prototypic autoimmune disease with the potential to target any organ system. It can mimic both organ-specific and non-organ-specific systemic illnesses. Thus, evaluation of clinical and laboratory abnormalities in SLE patients includes determination of their attribution to either SLE or non-SLE causes. This is critical, but often challenging, particularly for the nervous system events. For example, the optimal management of SLE patients who present with cerebrovascular disease, seizures or psychosis will depend in part on correct attribution of the cause of events. In this issue, Bortoluzzi et al. [ 1 ] address this important clinical conundrum. Although well recognized, the prevalence of neurological and psychiatric disease, neuropsychiatric SLE (NPSLE varies between 21% and 95% of SLE patients [ 2 ]. The wide variance is due in part to lack of clear definitions for NP events and partly to failure to determine their attribution. In 1999, the ACR produced a standard nomenclature and case definitions for 19 NP syndromes [ 3 ] that can occur in SLE. These include common events such as headache and rare events such as psychosis. For each of the NP syndromes, potential causes other than SLE were identified as either excluding SLE causes or recognizing non-SLE variables as having contributed in part to the event. This component of the ACR classification, in combination with other clinical, laboratory and neuroimaging data, has provided a framework for attributing NP events to SLE, non-SLE causes or both [ 4 ]. Data from animal and human studies suggest two separate and potentially complementary primary immunopathogenic pathways for NPSLE [ 5 ]: (i) vascular injury involving large and small calibre vessels mediated by aPLs, immune complexes and leucoagglutination. Clinical sequelae include focal NP events such as stroke and diffuse NP events such as cognitive dysfunction; (ii) autoimmune inflammation-mediated injury with increased permeability of the blood–brain barrier, intrathecal formation of immune complexes, and production of IFN-α and other inflammatory mediators. Clinical sequelae include diffuse NP manifestations such as psychosis and acute confusion. The study by Bortoluzzi et al. [ 1 ] involved specialists in rheumatology, neurology, psychiatry and neuroradiology, with support from the Italian Society of Rheumatology. It consisted of three phases: (i) a literature review and a consensus-driven process to identify and give weightage to potential attribution variables; (ii) a retrospective data-driven process using an established SLE cohort at a single academic centre to confirm and refine the attribution variables; and (iii) a validation exercise for attribution rules using retrospective data on SLE patients followed up at seven Italian rheumatology centres over a 5-year period. All patients had at least one of the 19 NP events as per the ACR case definitions with a minimum of 1 year of follow-up. Recurrent NP events were also recorded, and potential non-SLE causes were identified. The final attribution of NP events was made by the onsite multidisciplinary team and graded as related, uncertain or unrelated to SLE. The attribution model addressed four themes: (i) the temporal relationship of NP events to the diagnosis of SLE; (ii) identification of minor or common NP events; (iii) recognition of confounding factors (i.e. alternative aetiologies or non-SLE contributing factors derived from the ACR case definitions for the 19 NP syndromes); and (iv) favouring factors [i.e. clinical and non-clinical variables that support the attribution to SLE derived from European League Against Rheumatism (EULAR) recommendations on NPSLE [ 6 ] and the expert panel]. The first two items applied to all NP events; for the latter two items, lists of variables specific for each NP event were generated. Statistical modelling was used to derive and validate a weighted attribution score. The area under the receiver operating characteristic curve, using dichotomous outcomes (related vs uncertain/not related), was 0.866 in the derivation dataset and 0.816 in the validation dataset. In the combined dataset, a cut-off score of ≥6 yielded a positive predictive value (PPV) of 78.8% (95% CI 73.4%, 83.6%) and a negative predictive value (NPV) of 76.1% (95% CI 69.1%, 82.2%). Adjusting the attribution score to >7 provided a PPV of 86.3% (95% CI 76.2%, 93.2%), and a score <3 provided an NPV of 85.7% (95% CI 63.7%, 97%) in the validation dataset. This study by Bortoluzzi et al. is commendable and timely, and addresses an area of need. As acknowledged, components of the attribution model have been proposed and used by other investigators before [ 4 ], but this study has formalized, expanded and validated previous work. Over 400 individual NP events were available for analysis in the derivation and validation datasets. The frequency and characteristics of NP events and their attribution to SLE and non-SLE causes are comparable to those of epidemiological studies [ 7–9 ]. Both inclusion of a training dataset for initial derivation of the attribution variables and the subsequent recruitment of a validation dataset are methodologically robust. The use of a combined dataset to determine cut-off scores may have biased some of the findings through inclusion of the derivation dataset. This study involved multiple academic centres, albeit from the same European country, in which a shared medical culture may have favoured a homogeneous diagnostic approach. For pragmatic reasons, the study design was retrospective. Plexopathy was the only NP event not available for study, and for five additional NP manifestations, there were no more than five events for each syndrome in the combined datasets. The correct attribution of NP events is important for care of individual patients and also for studies of both pathogenic mechanisms and therapeutic interventions. In practice, the decision to use more potent immunosuppression is unnecessary and potentially risky for the treatment of NP events attributed to non-SLE causes, for which symptomatic therapies may be effective and safe. The wrong attribution of NP events (e.g. seizures, mood disorders or headaches) to SLE when scoring disease activity measures (e.g. SLEDAI-2K [ 10 ] and BILAG [ 11 ]) may result in erroneously high global DASs. This is a particular problem with NP variables in view of their heavy weighting. The availability of a validated and reproducible attribution score is a significant advancement in the study and treatment of nervous system diseases in lupus patients. Additional studies involving patients of different ethnicity, managed in both non-academic and academic centres and monitored prospectively with well-defined clinical outcomes, are needed before we can be confident of accurately attributing causation to SLE and/or non-SLE factors for NP events in SLE patients. J.G.H.’s work is supported by an operating grant (MOP-86526) from the Canadian Institutes of Health Research grant. Disclosure statement : The author has declared no conflicts of interest.

Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.