Diagnosis of Primary Neuropsychiatric Systemic Lupus Erythematosus: Attribution Models Versus Physician Judgment

Antonis Fanouriakis^1,2, Cristina Pamfil³, Simona Rednic³, Prodromos Sidiropoulos¹, George Bertsias^4,5 and Dimitrios Boumpas^5,6,7, ¹Rheumatology, Clinical Immunology, and Allergy, University of Crete, Medical School, University Hospital, Heraklion, Greece, ²Foundation of Research and Technology, University of Molecular Biology and Biotechnology, Heraklion, Greece, ³Rheumatology, University of Medicine and Pharmacy, Cluj-Napoca, Romania, ⁴Rheumatology, Clinical Immunology, and Allergy, University of Crete, Heraklion, Greece, ⁵Institute of Molecular Biology and Biotechnology, Foundation of Research and Technology-Hellas, Heraklion, Crete, Greece, ⁶Biomedical Research Foundation, Academy of Athens, Athens, Greece, ⁷Clinical Immunology and Rheumatology, National And Kapodistrian University of Athens, Medical School, Athens, Greece

Meeting: 2015 ACR/ARHP Annual Meeting

Date of first publication: September 29, 2015

Keywords: Diagnostic Tests and nervous system lupus

Session Information

Date: Tuesday, November 10, 2015

Title: Systemic Lupus Erythematosus - Clinical Aspects and Treatment Poster Session III

Session Type: ACR Poster Session C

Session Time: 9:00AM-11:00AM

Background/Purpose: Attribution of neuropsychiatric manifestations to systemic lupus erythematosus per se (“primary NPSLE”) is challenging and depends largely on physician judgment based on clinical and neuroimaging results. To facilitate physicians, attribution models have been proposed but their diagnostic performance has not been validated. We compared attribution of neuropsychiatric events as per physician judgment against different attribution models.

Methods: SLE patients with neuropsychiatric involvement were retrospectively identified. All neuropsychiatric manifestations were classified as attributed to SLE, not attributed to SLE, or uncertain, according to physician judgment. Results were compared against three published attribution models: the Systemic Lupus International Collaborating Clinics (SLICC) models A (most stringent) and B (less stringent), and a recently proposed algorithm by the Italian study group on NPSLE (total score 0-10). Sensitivity and specificity for each model was calculated against physician judgment. For the Italian algorithm, a receiver operating curve (ROC) analysis was also performed to calculate the area under the curve (AUC) using dichotomous outcome (attributed vs. not attributed/uncertain).

Results: A total of 242 neuropsychiatric manifestations, experienced by 191 patients, were included in the study. According to physician judgment, 136 events were attributed and 96 not attributed to SLE; for 10 events, attribution was considered uncertain. Using the SLICC models, only a small proportion of the events were attributed to SLE [34 events (14.0%) with model A, 67 events (27.7%) with model B]. Consequently, when compared with physician judgment, both models showed high specificity, but poor sensitivity (Table 1). Exclusion of so-called ‘minor’ neuropsychiatric manifestations (ie. cases of headache, mild depression, mild cognitive dysfunction and polyneuropathy without electrophysiologic confirmation), which are a priori not attributed to SLE in the SLICC models, led to moderate increases in sensitivity (27.7% and 42.0% for models A and B, respectively), but specificity was reduced especially in SLICC model B (94.8% and 65.5% for models A and B, respectively).

The AUC of the ROC curve for the Italian study group algorithm was 0.859, indicating good reliability as a diagnostic test for NPSLE attribution. When serial cut-offs of total score were tested, values ≥ 7 showed the best combination of sensitivity and specificity (82.4% and 82.9%, respectively [Table 1]).

Conclusion: Attribution models show varying levels of sensitivity and specificity, related to physician judgment. A recently proposed attribution score performs well and may be suitable for use in real-life clinical settings. SLICC model A shows the highest specificity and can be used for confirmation of doubtful cases.

Table 1. Sensitivity and specificity of different attribution models for neuropsychiatric SLE

Model	Sensitivity	Specificity	Positive predictive value	Negative predictive value
Italian model Score ≥ 8 Score ≥ 7 Score ≥ 6	42.6% 82.4% 87.5%	93.3% 82.9% 68.6%	89.2% 86.1% 78.3%	55.7% 78.4% 80.9%
SLICC model A (stringent)	22.8%	97.2%	91.2%	49.5%
SLICC model B (less stringent)	34.6%	80.6%	79.1%	48.2%

Disclosure: A. Fanouriakis, None; C. Pamfil, None; S. Rednic, None; P. Sidiropoulos, None; G. Bertsias, None; D. Boumpas, None.

To cite this abstract in AMA style:

Fanouriakis A, Pamfil C, Rednic S, Sidiropoulos P, Bertsias G, Boumpas D. Diagnosis of Primary Neuropsychiatric Systemic Lupus Erythematosus: Attribution Models Versus Physician Judgment [abstract]. Arthritis Rheumatol. 2015; 67 (suppl 10). https://acrabstracts.org/abstract/diagnosis-of-primary-neuropsychiatric-systemic-lupus-erythematosus-attribution-models-versus-physician-judgment/. Accessed .

« Back to 2015 ACR/ARHP Annual Meeting

ACR Meeting Abstracts - https://acrabstracts.org/abstract/diagnosis-of-primary-neuropsychiatric-systemic-lupus-erythematosus-attribution-models-versus-physician-judgment/