A Longitudinal Analysis of Change in Lupus Nephritis in an International Inception Cohort Using a Multistate Markov Model Approach

Abstract

Background/Purpose: Patients with lupus nephritis (LN) may have improvement or deterioration in renal status over time. To capture bidirectional change we used a reversible multistate Markov model to study transitions in glomerular filtration rate (GFR) and proteinuria (PrU) in a prospective, international, inception cohort of SLE patients receiving standard of care.

Methods: Patients were evaluated at enrollment and annually. LN was identified from the ACR classification criteria and/or renal biopsy. Data included medications, estimated GFR (eGFR) and PrU (ePrU), disease activity (SLEDAI-2K), organ damage [(SLICC)/ACR damage index (SDI)] and lupus autoantibodies including IgG anticardiolipin (aCL) and lupus anticoagulant (LAC). GFR states were defined: state 1 (eGFR: >60 ml/min); state 2 (eGFR: 30–60 mL/min); and state 3 (eGFR: <30 ml/min). Similarly, PrU states were defined: state 1 (ePrU: <0.25 gr/day); state 2 (ePrU: 0.25–3.0 gr/day); and state 3 (ePrU: >3.0 gr/day). Multistate models were used to provide estimates of relative transition rates and state occupancy probabilities for various time periods.

Results: Of 1,826 SLE patients, 89% were female, 49.2% Caucasian with mean±SD age 35.1±13.3 years. The mean disease duration at enrollment was 0.5±0.3 years and follow-up was 4.6±3.4 years. LN occurred in 700/1,826 (38.3%) patients. There was no observed change in eGRF state for 2303/2430 (94.8%), 136/255 (53.3%) and 26/55 (47.3%) visits when the previous eGFR state was 1, 2 and 3 respectively. The corresponding values for ePrU states were 1167/1460 (79.9%), 547/983 (55.6%) and 59/205 (28.8%). For both outcomes the likelihood of improvement (states 2 to 1 and 3 to 2) was greater than deterioration (states 1 to 2 and 2 to 3). The transition from state 3 to ESRD was more likely with eGFR (54%) than ePrU (9%). Probability estimates of transitioning between eGFR and ePrU states, ESRD and death at 1, 2 and 5 years were determined. At year one the highest probability was for patients to remain in the initial eGFR (95%, 55%, 42%) or ePrU (81%, 56%, 31%) state. Following 2 and 5 years, the estimated probability for improvement in either eGFR or ePrU was higher than deterioration. Multivariate analysis identified older age (p<0.001), race/ethnicity (Hispanic, Asian and African ancestry) (p<0.001), higher ePrU state (p<0.001), higher renal biopsy chronicity score (p=0.013) and baseline aCL antibodies (p=0.039) as predictors for deterioration in eGFR states and male sex (p=0.04) for improvement. For ePrU, multivariate analysis identified race/ethnicity (Hispanic, Asian and African ancestry) (p=0.009), corticosteroid use (p=0.031), higher eGFR state (p=0.011) and higher renal biopsy chronicity score (p=0.015) as predictors for deterioration. Positive LAC (p=0.006) and ISN/RPN class V nephritis (p=0.013) were associated with lower improvement rate.

Conclusion: Multistate modeling in patients with LN generates probability estimates of transitions between disease states that reflect improvement or deterioration in renal outcomes. This approach can identify predictors of change in renal status and can inform clinical trial design by providing minimum expectations for benefit from new therapeutic interventions for LN.