cSLE-T2T-GLOBAL: Childhood Systemic Lupus Erythematosus Treat-to- Target Analysis across Global Registries


Childhood-onset systemic lupus erythematosus (cSLE) is a rare autoimmune disease with a worse prognosis than adult-onset SLE, yet it remains understudied, leading to a lack of evidence-based therapeutic strategies. Most treatment approaches for cSLE draw from adult SLE research, making it difficult to determine optimal treatment strategies while minimizing damage. A collaborative international initiative, supported by several organizations including PReS and CARRA, aims to validate pediatric-specific treatment targets for cSLE. Recent collaborative work mapped data fields across the three largest cSLE cohorts, creating the largest cSLE dataset globally. A task force led by Dr. Eve Smith has developed consensus-based definitions for lupus low disease activity and clinical remission, requiring clinical validation. Leveraging successful collaboration and existing infrastructures, the initiative aims to validate cSLE treatment targets and assess their impact on patient outcomes through a prospective trial. The study will explore personalized predictions of target attainment using both conventional statistical methods and machine learning approaches.

Facts and Figures

Project Leads
MD, PhD Eve Smith
University of Liverpool

MD, PhD Rebecca Sadun
Duke University
FOREUM research grant: € 199.998

Meet the Team

Project Lead

MD, PhD Eve Smith
University of Liverpool
MD, PhD Rebecca Sadun
Duke University
MD, PharmD Jeniffer Cooper
University of Colorado
MD, MS Emily Smitherman
University of Alabama
MD, PhD Alexandre Belot
University of Lyon
MD, PhD Michael Beresford
University of Liverpool
MD, MS Laura Lewandowski
National Institutes of Health


Aim 1: To evaluate across the combined dataset the impact of paediatric-specific T2T target attainment on damage accrual (via the SLICC Damage Index) and key secondary outcomes.

  • Determine the proportion of patients who reach cLLDAS and/or clinical remission on or off steroids at any timepoint within each cohort, assessing both maximum continuous (uninterrupted) time in target and the total proportion of observed time spent in these targets.
  • Use Prentice, Williams, and Peterson gap time models19-21 to understand the association between attainment of the three aforementioned targets (at any timepoint during follow-up, continuously, or as a proportion of observed time) and the risk of new damage accrual and key secondary outcomes longitudinally.
  • Hypothesis: The risk of SLICC damage accrual will be associated with achievement of T2T targets, with those demonstrating a longer continuous period/higher total proportion of observed time in target seeing greater protection from damage.

Aim 2: To conduct sensitivity analyses to optimize the cLLDAS and clinical remission definitions across the combined cohort, refining the existing consensus-derived definitions as necessary.

  • Evaluate each criterion in the definitions of cLLDAS and clinical remission on/off steroids to eliminate redundancy and determine the simplest definition (i.e., establish whether any criteria can be removed).
  • Re-fit the PWP gap time models utilised for Aim 1.2 using simplified (e.g. omitting criterion from the aforementioned targets) and/or transformed targets (e.g. varying the thresholds for SLEDAI, PGA, corticosteroid doses), to compare protection against damage accrual and key secondary outcomes between the resulting models, and if/how existing consensus-derived target definitions can be improved.
  • Hypothesis: Consensus-based definitions of cLLDAS and clinical remission on/off steroids may need to be further refined, yielding data-driven target definitions that are superior to current consensus-based target definitions.

Aim 3: To investigate the potential for personalised predictions of target attainment, comparing conventional statistical methods to Machine Learning approaches.

  • In accordance with TRIPOD guidelines22, use conventional statistical analyses to predict the timelines for cLLDAS and remission target attainment based on patient characteristics and disease severity at baseline.
  • In accordance with TRIPOD-ML guidelines23, use machine learning approaches to predict the timelines for cLLDAS and remission target attainment, and create a risk score for prediction of target attainment at 3, 6, 12, or 24 months based on baseline characteristics. This will be evaluated using a training and validation set.
  • Compare the performance of conventional statistical methods and machine learning approaches, using a concordance index and receiver operating characteristic curves (AUC-ROC) to determine the predictive capabilities of the different approaches.
  • Hypothesis: Machine learning approaches will enhance our understanding of how to use baseline characteristics (clinical, demographic, laboratory, disease severity, and initial therapy) to predict T2T target attainment, indicating a personalised timeline for escalation of therapy to attain targets.


Our project is carefully structured into three distinct phases. Initially, we will focus on establishing the necessary data elements, formalizing contracts, and agreements for data sharing. Concurrently, the data will undergo rigorous cleaning to ensure alignment and integrability, a process anticipated to take approximately three months. The University of Liverpool team will then lead the execution of the first two aims, each spanning six months. Simultaneously, the harmonized dataset will be shared with the Duke University team to lead analyses for Aim 3. This phase will exploit both conventional statistical methods and machine learning to enhance the personalization of treatment strategies, aiming to compare the insights derived from these analytical approaches. The full study group will meet at least every two weeks to support the interpretation and discussion of findings and provide methodological support. To document and disseminate our findings, a manuscript will be produced for each aim, outlining the progress and outcomes achieved.

Patient Voice

Our project prioritizes incorporating the views of patients in the development of T2T treatment strategies for cSLE. We will engage our six patient research partners (three from Europe and three from the US) throughout the project to ensure their experiences directly inform our research design and implementation. This will include regular feedback sessions, enhancing our understanding of the results, and ensuring that we are responsiveness to their needs. Our patient partners have worked with us on previous projects and will be responsible for developing the public, patient involvement strategy for this study, guiding us as to how to use their expertise best. Such input is crucial for refining our global T2T strategies, ensuring that they align with patient expectations and contribute to improved outcomes.

Project Map