Article Text
Abstract
Background In a recent study of a large cohort of systemic lupus erythematosus (SLE) patients, we evaluated a new algorithm to determine the attribution of neuropsychiatric (NP) events to SLE or other causes [1].
Objectives In the present study, we tested the performance of the algorithm in a new international multicenter cohort of patients, all of whom had one or more NP events as per the 1999 ACR case definitions.
Methods A similar methodology to that used in the first study, was adopted. A dedicated electronic chart was created, including the core set of items for classification. Four factors were considered for each NP event (i) the time of onset of the NP event; (ii) the presence of concurrent or confounding non-SLE factors (i.e. “associations” suggested in the glossary for the 1999 ACR case definition); (iii) the type of NP event (major vs minor or common according to what has been proposed by Ainiala et al. [2]; (iv) the presence of “favouring factors” (i.e. supporting attribution). Patients from each center satisfied classification criteria for SLE and had one or more NP events. To maintain blinding, thus avoiding circular thinking, all events were evaluated by two independent assessors from each center, each assigned different tasks: the first assessor provided an attribution diagnosis (related/uncertain/not related to SLE) on the basis of their clinical judgement utilizing all of the information available in the patient record; the second assessor applied the attribution algorithm using the same available information and provided an attribution score. The performance of the attribution score was compared to the attribution determined by clinical judgement, which was taken as the “gold standard”. Receiver Operating Curve (ROC) analysis was used to determine the area under the curve (AUC) and the calculation of sensitivity, specificity, positive and negative predictive values [PPV and NPV (CI 95%)] using dichotomous outcomes (related vs uncertain/not related to SLE).
Results There were 243 SLE patients (23 M, 220 F; mean (SD) age 47 (12.7) years with 336 NP events. Using a pre-defined attribution score cut-off ≥7 (derived by our previous national validation study of NP events related vs uncertain/not related to SLE) the AUC was 0.89 for NP events attributed to SLE with sensitivity of 88.5% (95% CI 77.7, 87.5), specificity of 74.6% (95% CI 67.9, 80.53), a PPV of 81.8% (95% CI 75.8, 86.8) and a NPV of 83.4% (95% CI 75.9, 89.3).
Conclusions The attribution algorithm for evaluating NP events in SLE patients compared favourably to the judgement of experienced clinicians for the majority of NP events. It is a valid tool and provides a standardized approach to determining the attribution of NP events in future studies of NPSLE.
A. Bortoluzzi et al. Rheumatology 2015;54:891–898.
H. Ainiala et al. Neurology 2001;57:496–500.
Disclosure of Interest None declared