Background/Purpose: Cell bound complement activation products (CB-CAPs) including erythrocyte-bound C4d (EC4d) and B-lymphocyte C4d (BC4d) demonstrate increased diagnostic accuracy compared to conventional SLE markers (anti-dsDNA, anti-Smith and C3/C4). Clinical studies have validated a multianalyte panel (MAP) algorithm that integrates CB-CAPs with conventional autoantibodies (Ab), demonstrating balanced sensitivity and specificity for SLE. This study tests the clinical validity of CB-CAPs and MAP in a large cohort by evaluating their unique contribution to SLE sensitivity beyond conventional markers.

Methods: A pooled analysis was performed using biomarker data collected from 12 single and multi-center clinical validation trials involving 36 investigators across the U.S. A total of 1034 unique adult SLE subjects meeting the 1997 ACR criteria were analyzed and compared to 1489 controls (apparently healthy volunteers [AHV], N = 458; other autoimmune rheumatic diseases [OARD], including RA, N = 439; primary Sjögrens, N = 166; other rheumatic disease, N = 138; chronic pain disorders, including fibromyalgia, N = 235; chronic localized pain, N = 52). Flow cytometry was performed on erythrocytes and lymphocytes isolated from whole blood samples (EC4d & BC4d). Abs were measured by ELIA and C3/C4 were measured by immunoturbidimetry. Machine learning algorithms were employed to build predictive models for distinguishing SLE from a combined group of OARD and AHV. A Gradient Boosting Classifier model, chosen for superior LogLoss in 10-fold cross-validation, served as the base model (CB-CAPs, conventional Ab and C3/C4). A logistic regression model using MAP scores was developed. Both models underwent 10-fold cross-validation, with a 20% test set used for validation. Biomarker sensitivity for SLE was compared using chi-square analysis.

Results: The SLE cohort was majority female (82%), white (36%), mean (SD) age of 40 (± 10) with mean years since diagnosis of 10 (± 9) (Table 1). The combination of EC4d, BC4d, Abs and C3/C4 captured 67% of SLE with EC4d, BC4d, conventional markers contributing 4%, 7% and 17%, respectively, to the overall sensitivity for SLE (Fig.1A). Isolated positive CB-CAPs were more sensitive than isolated low C3/C4 (17% vs. 3%; p < 0.001) (Fig. 1B). The addition of MAP to CB-CAPs and conventional markers captured 76% of SLE (Fig. 1C). MAP was isolated positive in 25% of SLE cases, significantly (p < 0.001) higher than Abs (1%) and C3/C4 (3%), respectively. ROC analysis of MAP and a base model combining CB-CAPs, Abs and C3/C4 yielded similar AUCs on training (88% vs. 86%) and test (90% and 88%), respectively (Fig. 2A). BC4d had the highest relative feature importance (37%), compared to anti-dsDNA (19%) and EC4d (16%), with C3 & C4 at < 10% (Fig. 2B). The base model was highly specific for SLE, except vs. Sjögrens (Fig. 2C & 2D).

Conclusion: Integrating CB-CAPs with conventional markers significantly improves SLE diagnostic accuracy. The combination of MAP, conventional Ab and C3/C4 captured 76% of SLE cases, while MAP alone identified 25% of cases that were missed by conventional markers. This approach enhances diagnostic accuracy, highlighting the vital contribution of CB-CAPs to SLE sensitivity beyond conventional markers.

« Back to ACR Convergence 2024

ACR Meeting Abstracts - https://acrabstracts.org/abstract/systematic-analysis-demonstrates-the-added-value-of-cb-caps-to-sle-diagnosis-in-a-large-validation-cohort/