Background Phase III randomised controlled trials (BLISS 52 and 76) in patients with systemic lupus erythematosus (SLE) have established the efficacy and safety of belimumab plus standard SLE therapy (SoC) vs SoC alone. As with many continuation studies, the long-term extensions (LTE) of the BLISS trials (BEL112233/NCT00724867 and BEL112234/NCT00712933) were single-arm, open-label studies; therefore, comparative long-term efficacy of belimumab and SoC vs SoC was not assessed. Existing observational databases could provide a valuable source of comparison data.

Objectives This study (BEL206347) explores the use of propensity score (PS) adjustment to enable a post-hoc comparison of LTE patients to patients from an existing SLE cohort, assessing long-term efficacy of belimumab plus SoC vs SoC using the SLICC/ACR Damage Index (SDI) as the outcome.

Methods Potential PS model criteria were identified by a systematic literature review for factors predicting organ damage. PS model criteria were restricted to baseline (at index [first recorded disease activity of SLE Disease Activity Index {SLEDAI} ≥6] for SoC cohort) characteristics available for both cohorts, including age, gender, race, disease duration, clinical and laboratory characteristics, SLEDAI over time, baseline SLEDAI and SDI and SoC (e.g. corticosteroid use/dose). Patients with active severe lupus nephritis/central nervous system lupus, those who had received B-cell-targeted therapy or belimumab or those with an index date before 1990 were excluded from the SoC cohort. The primary PS procedure used 1:1 matching with a 20% calliper (PSM); alternatively, inverse PS weighting (IPSW) was used to preserve the LTE sample. The balance in baseline factors across cohorts pre-/post-PS-adjustment was assessed using standardised distance measures.

Results The literature review identified the Toronto Lupus Cohort (TLC) as an external registry with a comparable patient cohort. Without PS-adjustment, the standardised distance across cohorts exceeded 25% for 13 of 17 baseline factors in the PSM model. After PSM, standardised distance was <10% for all factors, but less than half of the LTE patients were matched. Using IPSW, data from all patients could be used but with a weighted standardised distance of <10% for 8 and <25% for 15 of the 17 PSM model factors.

Conclusions The TLC was identified as an existing lupus registry with the endpoints of interest and detailed clinical characteristics. PS matching further maximised the comparability of the LTE and TLC cohorts. The final predictors used in the matching will be validated by two SLE clinicians. The preliminary results indicate a statistically adequate balance in clinical characteristics is attainable. PS matching will allow comparison of long-term organ damage in patients with SLE receiving belimumab plus SoC vs SoC alone. This approach can be applied to other extension studies to assess long-term treatment effectiveness.

Acknowledgements Study funded by GSK. Katie White, PhD, Fishawack Indicia Ltd, provided editorial assistance funded by GSK.

Disclosure of Interest M. Urowitz Grant/research support from: GSK, UCB and Eli Lilly, Consultant for: GSK, UCB and Eli Lilly, R. Wielage Consultant for: GSK, K. Kelton Consultant for: GSK, R. Ohsfeldt Consultant for: GSK, Amgen and AstraZeneca Hygieia, Y. Asukai Shareholder of: GSK, Employee of: GSK, S. Ramachandran Shareholder of: GSK, Employee of: GSK