Background/Purpose: SLE is marked by altered immune regulation linked to waxing and waning clinical disease activity. This study seeks to verify the alteration of inflammatory and regulatory mediators prior to clinical disease flare in a confirmatory cohort of SLE patients, seeking the best mediators to inform our Lupus Flare Prediction Index (LFPI).

Methods: Immune mediators were evaluated by xMAP assay in pre-flare (n=90) and “self” pre-nonflare (SNF) plasma samples (n=90) from the same female patients (49 European-, 41 African-American) with classified SLE (≥4 ACR criteria), as well as 48 race/age matched healthy controls (HC). Hybrid SLEDAI (hSLEDAI) scores, clinical features, medication usage, and the presence of SLE-associated autoantibody (AutoAb) specificities, including dsDNA, chromatin, Ro/SSA, La/SSB, Sm, SmRNP, and RNP, were also compared at pre-flare (100 ± 40 days prior to flare) vs. pre-SNF (96 ± 40 days prior to SNF) time points. A subset of 31, log-transformed, immune mediators were further evaluated using applied machine learning modeling approaches (random forest and XG Boost) to determine an optimal subset of mediators to inform the LFPI. The LFPI is the sum of log-transformed, standardized immune mediators, weighted by the Spearman r correlation coefficient of pre-flare/SNF immune mediator levels vs. hSLEDAI scores at disease flare/SNF.

Results: We did not observe differences with respect to hSLEDAI scores, clinical features, medication usage, or number and type of SLE-associated AutoAbs comparing pre-flare vs. pre-SNF time points, either in aggregate or by race. At follow-up, SLE patients experiencing a clinical disease flare (vs. SNF visit) had significantly elevated hSLEDAI scores (p< 0.0001), with an increased presence of arthritis (p< 0.0001), mucocutaneous (p< 0.0001), and serositis (p=0.0066) clinical features. Thirty-one immune mediators (Figure [A]) were altered in cases vs HC (p≤0.05), and altered in pre-flare vs. pre-SNF samples (p≤0.02 after adjusting for multiple comparison); inflammatory mediators were increased and regulatory mediators decreased in pre-flare samples. Variable importance of the 31 mediators was determined by random forest (Figure [A]) and confirmed by XG Boost. Forward selection/backward elimination drew 9-14 mediators that best differentiated pre-flare from pre-SNF (Table). A combination of the top 10 immune mediators best informed and maximized the performance of a newly refined LFPI (Figure [B]), achieving 87% sensitivity and 91% specificity. In addition to identifying impending clinical disease flare, the LFPI (and top mediators informing it) was significantly increased in SLE patients who went on to increased features of arthritis (p< 0.0001), mucocutaneous (p< 0.0001), or serositis (p=0.0003) clinical features at follow-up.

Conclusion: We verified the alteration of inflammatory and regulatory immune mediators with imminent lupus disease flare. A subset of mediators boosted the LFPI to identify SLE patients at risk of imminent flare who would benefit from early intervention strategies. Such an approach would be a game changer in prospective clinical trials and the management of lupus.

ACR Meeting Abstracts - https://acrabstracts.org/abstract/ability-of-inflammatory-and-regulatory-soluble-mediators-to-forecast-impending-clinical-disease-flare-and-inform-a-refined-lupus-flare-prediction-index-in-a-confirmatory-cohort-of-sle-patients/