Faculty Bibliography

Objective/UNASSIGNED:A common problem in clinical trials is missing data due to participant dropout and loss to follow-up, an issue which continues to receive considerable attention in the clinical research community. Our objective was to examine and compare current and alternative methods for handling missing data in SLE trials with a particular focus on multiple imputation, a flexible technique that has been applied in different disease settings but not to address missing data in the primary outcome of an SLE trial. Methods/UNASSIGNED:Data on 279 patients with SLE randomised to standard of care (SoC) and also receiving mycophenolate mofetil (MMF), azathioprine or methotrexate were obtained from the Lupus Foundation of America-Collective Data Analysis Initiative Database. Complete case analysis (CC), last observation carried forward (LOCF), non-responder imputation (NRI) and multiple imputation (MI) were applied to handle missing data in an analysis to assess differences in SLE Responder Index-5 (SRI-5) response rates at 52 weeks between patients on SoC treated with MMF versus other immunosuppressants (non-MMF). Results/UNASSIGNED:The rates of missing data were 32% in the MMF and 23% in the non-MMF groups. As expected, the NRI missing data approach yielded the lowest estimated response rates. The smallest and least significant estimates of differences between groups were observed with LOCF, and precision was lowest with the CC method. Estimated between-group differences were magnified with the MI approach, and imputing SRI-5 directly versus deriving SRI-5 after separately imputing its individual components yielded similar results. Conclusion/UNASSIGNED:The potential advantages of applying MI to address missing data in an SLE trial include reduced bias when estimating treatment effects, and measures of precision that properly reflect uncertainty in the imputations. However, results can vary depending on the imputation model used, and the underlying assumptions should be plausible. Sensitivity analysis should be conducted to demonstrate robustness of results, especially when missing data proportions are high.

Objective/UNASSIGNED:Existing methods for grading lupus flares or improvement require definition-based thresholds as increments of change. Visual analogue scales (VAS) allow rapid, continuous scaling of disease severity. We analysed the performance of the SELENA SLEDAI Physician's Global Assessment (SSPGA) and the Lupus Foundation of America-Rapid Evaluation of Activity in Lupus (LFA-REAL) as measures of improvement or worsening in SLE. Methods/UNASSIGNED:We evaluated the agreement between prospectively collected measures of lupus disease activity [SLE Disease Activity Index (SLEDAI), British Isles Lupus Assessment Group Index 2004 (BILAG 2004), Cutaneous Lupus Area and Severity Index (CLASI), SSPGA and LFA-REAL] and response [(SLE Responder Index (SRI)-4 and BILAG-Based Combined Lupus Assessment (BICLA)] in a clinical trial. Results/UNASSIGNED:Fifty patients (47 females, mean age 45 (±11.6) years) were assessed at 528 consecutive visits (average 10.6 (±4.1) visits/patient). Changes in disease activity compared with baseline were examined in 478 visit pairs. SSPGA and LFA-REAL correlated with each other (r=0.936), and with SLEDAI and BILAG (SSPGA: r=0.742 (SLEDAI), r=0.776 (BILAG); LFA-REAL: r=0.778 (SLEDAI), r=0.813 (BILAG); all p<0.0001). Changes (∆) in SSPGA and LFA-REAL compared with screening correlated with each other (r=0.857) and with changes in SLEDAI and BILAG (∆SSPGA: r=0.678 (∆SLEDAI), r=0.624 (∆BILAG); ∆LFA-REAL: r=0.686 (∆SLEDAI) and 0.700 (∆BILAG); all p<0.0001). Changes in SSPGA and LFA-REAL strongly correlated with SRI-4 and BICLA by receiver operating characteristic analysis (p<0.0001 for all). Additionally, LFA-REAL correlated to individual BILAG organ scores (musculoskeletal: r=0.842, mucocutaneous: r=0.826 (p<0.0001 for both)). Conclusion/UNASSIGNED:SSPGA and LFA-REAL are reliable surrogates of common SLE trial end points and could be used as continuous or dichotomous response measures. Additionally, LFA-REAL can provide individualised scoring at the symptom or organ level. Trial registration number/UNASSIGNED:NCT02270957.

Genetic variation within the major histocompatibility complex (MHC) contributes substantial risk for systemic lupus erythematosus, but high gene density, extreme polymorphism and extensive linkage disequilibrium (LD) have made fine mapping challenging. To address the problem, we compared two association techniques in two ancestrally diverse populations, African Americans (AAs) and Europeans (EURs). We observed a greater number of Human Leucocyte Antigen (HLA) alleles in AA consistent with the elevated level of recombination in this population. In EUR we observed 50 different A-C-B-DRB1-DQA-DQB multilocus haplotype sequences per hundred individuals; in the AA sample, these multilocus haplotypes were twice as common compared to Europeans. We also observed a strong narrow class II signal in AA as opposed to the long-range LD observed in EUR that includes class I alleles. We performed a Bayesian model choice of the classical HLA alleles and a frequentist analysis that combined both single nucleotide polymorphisms (SNPs) and classical HLA alleles. Both analyses converged on a similar subset of risk HLA alleles: in EUR HLA- B*08:01 + B*18:01 + (DRB1*15:01 frequentist only) + DQA*01:02 + DQB*02:01 + DRB3*02 and in AA HLA-C*17:01 + B*08:01 + DRB1*15:03 + (DQA*01:02 frequentist only) + DQA*02:01 + DQA*05:01+ DQA*05:05 + DQB*03:19 + DQB*02:02. We observed two additional independent SNP associations in both populations: EUR rs146903072 and rs501480; AA rs389883 and rs114118665. The DR2 serotype was best explained by DRB1*15:03 + DQA*01:02 in AA and by DRB1*15:01 + DQA*01:02 in EUR. The DR3 serotype was best explained by DQA*05:01 in AA and by DQB*02:01 in EUR. Despite some differences in underlying HLA allele risk models in EUR and AA, SNP signals across the extended MHC showed remarkable similarity and significant concordance in direction of effect for risk-associated variants.

OBJECTIVE:To determine the frequency, associations and outcomes of cerebrovascular events (CerVEs) in a multi-ethnic/racial, prospective, SLE disease inception cohort. METHODS:Patients were assessed annually for 19 neuropsychiatric (NP) events including 5 types of CerVEs: (i) Stroke; (ii) Transient ischemia; (iii) Chronic multifocal ischemia; (iv) Subarachnoid/intracranial hemorrhage; (v) Sinus thrombosis. Global disease activity (SLEDAI-2K), SLICC/ACR damage index (SDI) and SF-36 scores were collected. Time to event, linear and logistic regressions and multi-state models were used as appropriate. RESULTS:Of 1,826 SLE patients, 88.8% were female, 48.8% Caucasian, mean±SD age 35.1±13.3 years, disease duration 5.6±4.2 months and follow-up 6.6±4.1 years. CerVEs were the fourth most frequent NP event: 82/1,826 (4.5%) patients had 109 events, 103/109 (94.5%) were attributed to SLE and 44/109 (40.4%) were identified at enrollment. The predominant events were stroke [60/109 (55.0%)] and transient ischemia [28/109 (25.7%)]. CerVEs were associated with other NP events attributed to SLE (HR (95% CI): (3.16; 1.73-5.75) (p<0.001), non-SLE NP (2.60; 1.49-4.51) (p<0.001), African ancestry at US SLICC sites (2.04; 1.01-4.13) (p=0.047) and organ damage (p=0.041). Lupus anticoagulant increased the risk of first stroke and sinus thrombosis [2.23 (1.11, 4.45) p=.024] and TIA [3.01 (1.15, 7.90) p=0.025]. Physician assessment indicated resolution or improvement in the majority but patients reported sustained reduction in SF-36 summary and subscale scores following CerVEs (P<0.0001). CONCLUSION/CONCLUSIONS:CerVEs, the fourth most frequent NP event in SLE, are usually attributable to lupus. In contrast to good physician reported outcomes, patients report a sustained reduction in health-related quality of life following CerVEs.

Background/Purpose: Systemiclupus erythematosus (SLE) disproportionally affects women of childbearing age. Low disease activity for 6 months prior to conception leads to the best outcomes; however, there is little prospective data describing the relative frequency and predictors of flares during and after pregnancy under such conditions.
Method(s): Analyses used data from the PROMISSE study, a multicenter, prospective observational study (2003-2014) of 384 pregnant women meeting >=4 ACR SLE criteria. Subjects were enrolled <12 wks gestation and samples collected throughout pregnancy and post-partum. Exclusion criteria were multi-fetal pregnancy, active disease (prednisone >20 mg/ d), or renal disease (proteinuria >1 gm/24h, and creatinine > 1.2 mg/dL). Mild/moderate and severe flares were defined using the SELENA-SLEDAI Flare Index. Flares during pregnancy were assessed in all 384 patients, and post-partum flares in those with study visits 2-6 months post-partum. Logistic regression models were fit to the data to identify independent predictors of experiencing any type of flare during pregnancy and post-partum.
Result(s): Rates of Flare: 105 flares were recorded during pregnancy (3.8% 1st Trimester, 53.3% 2nd, 42.9% 3rd). Counting one flare per person, 100 of 384 patients (26%) flared at any point during pregnancy; 20.8% of patients had mild/ moderate flares and 6.25% had severe. 57 of 234 patients (24.4%) with a study visit 2-6 months post-partum flared; 22.7% had mild/moderate flares and 1.7% severe. Post-partum flares were mild, and 19 of 57 (33.3%) were treated; 13 with an increase in prednisone, 6 with NSAID or hydroxychloroquine, and 1 with mycophenolate mofetil. The rate of any type of flare was 0.39/person-year at any point during pregnancy and 0.84/person-year post-partum. The proportion of subjects who had any flares during and after pregnancy was similar, but the post-partum flares occurred over a shorter duration of follow-up. Correlates of Flare: Among baseline variables considered (Table) only age, ethnicity/race, low complement, and PGA were independently predictive of having any flare during pregnancy. Clinical features associated with adverse pregnancy outcome (platelet count, antihypertensive use, and LAC) were not predictive of flare. The mean time from the last visit during pregnancy to the post-partum visit was 20 weeks and ranged from 9 to 34 weeks. Neither baseline clinical variables nor clinical variables of the last visit during pregnancy were associated with occurrence of any post-partum flare.
Conclusion(s): Flares during pregnancy are correlated with clinical and serological activity during the first trimester. Flares during and after pregnancy are typically mild, infrequently require treatment, and occur at similar rates. (Table Presented)

BANK1 is a susceptibility gene for several systemic autoimmune diseases in several populations. Using the genome-wide association study (GWAS) data from Europeans (EUR) and African Americans (AA), we performed an extensive fine mapping of ankyrin repeats 1 (BANK1). To increase the SNP density, we used imputation followed by univariate and conditional analysis, combined with a haplotypic and expression quantitative trait locus (eQTL) analysis. The data from Europeans showed that the associated region was restricted to a minimal and dependent set of SNPs covering introns two and three, and exon two. In AA, the signal found in the Europeans was split into two independent effects. All of the major risk associated SNPs were eQTLs, and the risks were associated with an increased BANK1 gene expression. Functional annotation analysis revealed the enrichment of repressive B cell epigenomic marks (EZH2 and H3K27me3) and a strong enrichment of splice junctions. Furthermore, one eQTL located in intron two, rs13106926, was found within the binding site for RUNX3, a transcriptional activator. These results connect the local genome topography, chromatin structure, and the regulatory landscape of BANK1 with co-transcriptional splicing of exon two. Our data defines a minimal set of risk associated eQTLs predicted to be involved in the expression of BANK1 modulated through epigenetic regulation and splicing. These findings allow us to suggest that the increased expression of BANK1 will have an impact on B-cell mediated disease pathways.

Background Missing data due to drop-out and loss to followup is a common problem in SLE trials. The usual approachesfor handling this issue include analyzing only subjects withcomplete data (complete case analysis; CC), last observationcarried forward (LOCF), or imputing non-responses for missing outcomes (non-responder imputation; NRI). However, thevalidity of these methods depends on strong assumptionsabout the missing data mechanism. Multiple imputation (MI)is a flexible model-based technique that accounts for uncertainty in the imputation process by generating several possiblevalues for the missing data, resulting in multiple completedata sets. These are analyzed separately and results are combined. MI is being used more widely in different disease settings but has not been applied to analyze the primaryoutcome in a SLE trial. We explored the use of MI to addressmissing data in the composite outcome, SLE Responder Index(SRI)-5, using data from patients assigned to standard of care(SoC) in a 52 week trial.Methods Data on 279 SLE patients randomized to SoC for 52weeks who were receiving mycophenolate mofetil (MMF), azathioprine, or methotrexate at entry were obtained from theLupus Foundation of America-Collective Data Analysis Initiative database. Multiple imputation using chained equationswas applied to handle missing data in an analysis to evaluatedifferences in SRI-5 response rates at 52 weeks betweenpatients on MMF and the other immunosuppressants (nonMMF). Three different imputation models were consideredthat included various combinations of longitudinal measures ofdisease activity (both composite and individual measures) andpatient characteristics. Results were compared to estimatesusing the CC, LOCF, and NRI.Results Missing data rates were 32% in the MMF and 23%in the non-MMF groups. As expected, the NRI missing dataapproach yielded the lowest response rates; the smallest andleast significant estimates of between group differences wereobserved with LOCF (table 1). Group differences were magnified with all three MI models compared to results of othermethods. Imputing SRI-5 directly (MI-1) versus the individualcomponents (MI-2) yielded nearly identical results.Conclusions Given the limitations of conventional approachesfor handling missing data, the MI method should also be considered in SLE trials. However, results can vary depending onthe imputation model that is used, and the assumptionsrequired for validity of this and other missing data methodsmust be justified. Sensitivity analysis using different approachesis important to demonstrate robustness of results especiallywhen missing data rates are non-negligible

Background Little is known about the association of healthcare costs with damage accrual in SLE. We describe the costsassociated with damage progression using multi-statemodeling.Methods Patients fulfilling the revised ACR Classification Criteria for SLE from 32 centres in 11 countries were enrolledin the Systemic Lupus International Collaborating Clinics(SLICC) inception cohort within 15 months of diagnosis.Annual data on demographics, SLE disease activity (SLEDAI-2K), damage (SLICC/ACR Damage Index [SDI] if-6 monthsfrom diagnosis), hospitalizations, medications, dialysis, and utilization of selected medical/surgical procedures were collected.Annual health resource utilization was costed using 2017Canadian prices. Annual costs associated with SDI states wereobtained from multiple regressions adjusting for age, sex, race/ethnicity, and disease duration. As there were relatively fewtransitions to SDI states 5 11, these were merged into a singleSDI state. Five and 10 year cumulative costs were estimatedby multiplying annual costs associated with each SDI state bythe expected duration in each state, which was forecastedusing a multi-state model and longitudinal SDI data from theSLICC Inception Cohort (Bruce IN et al. Ann Rheum Dis2015;74:1706 13). Future costs were discounted at a yearlyrate of 3%.Results 1676 patients participated, 88.7% female, 49.2% Caucasian, mean age at diagnosis 34.6 years (SD 13.4), mean disease duration at enrollment 0.5 years (range 0 1.3 years), andmean follow up 7.8 years (range 0.6 16.9 years). Healthresource utilization and annual costs (after adjustment usingregression) were markedly higher in those with higher SDIs(SDI=0, annual costs $1847, 95% CI $1120 to $2574;SDI-5, annual costs $26 772, 95% CI $19 631 to $33 813).At SDI<=2, hospitalizations and medications accounted for97.1% of direct costs, whereas at SDI-3, dialysis was responsible for 55.0%.Five and 10 year cumulative costs stratified by baseline SDIwere calculated by multiplying the annual costs associatedwith each SDI by the expected duration in that state. Fiveand 10 year costs were greater in those with the highest SDIsat baseline (table 1).Conclusions Patients with the highest baseline SDIs incurannual costs and 10 year cumulative costs that are at least 10-fold higher than those with the lowest baseline SDI. By estimating the expected duration in each SDI state and incorporating annual costs, disease severity at presentation can beused to predict future healthcare costs, critical knowledge forcost-effectiveness evaluations of novel therapies

Systemic Lupus Erythematosus (SLE or lupus) (OMIM: 152700) is a chronic autoimmune disease with debilitating inflammation that affects multiple organ systems. The STAT1-STAT4 locus is one of the first and most highly-replicated genetic loci associated with lupus risk. We performed a fine-mapping study to identify plausible causal variants within the STAT1-STAT4 locus associated with increased lupus disease risk. Using complementary frequentist and Bayesian approaches in trans-ancestral Discovery and Replication cohorts, we found one variant whose association with lupus risk is supported across ancestries in both the Discovery and Replication cohorts: rs11889341. In B cell lines from patients with lupus and healthy controls, the lupus risk allele of rs11889341 was associated with increased STAT1 expression. We demonstrated that the transcription factor HMGA1, a member of the HMG transcription factor family with an AT-hook DNA-binding domain, has enriched binding to the risk allele compared to the non-risk allele of rs11889341. We identified a genotype-dependent repressive element in the DNA within the intron of STAT4 surrounding rs11889341. Consistent with expression quantitative trait locus (eQTL) analysis, the lupus risk allele of rs11889341 decreased the activity of this putative repressor. Altogether, we present a plausible molecular mechanism for increased lupus risk at the STAT1-STAT4 locus in which the risk allele of rs11889341, the most probable causal variant, leads to elevated STAT1 expression in B cells due to decreased repressor activity mediated by increased binding of HMGA1.

OBJECTIVE:Lupus disease measures such as the Systemic Lupus Erythematosus Disease Activity Index (SLEDAI) and the British Isles Lupus Assessment Group (BILAG) index are challenging to interpret. The Lupus Foundation of America-Rapid Evaluation of Activity in Lupus (LFA-REAL) is intended to provide an efficient application of anchored visual analog scores, each representing the individual severity of active symptoms, with the sum of individual scores deriving an overall disease activity assessment. Our objective was to compare the performance of LFA-REAL to systemic lupus erythematosus disease activity assessments and compare scores between trained lupus clinical investigators and clinicians. METHODS:Investigators scored the SLEDAI, BILAG, physician's global assessment (PGA), and LFA-REAL, while the clinicians scored the LFA-REAL. The level of agreement between physicians and instruments was determined. RESULTS:The study included 99 patients (93% women, 31% white, mean ± SD ages 43.4 ± 13.2 years). At the first visit, the mean ± SD SLEDAI score was 5.5 ± 4.5, BILAG score 6.7 ± 7.8, and PGA score 33.6 ± 24.5. The mean ± SD investigator LFA-REAL score was 46.2 ± 42.9, and clinician LFA-REAL score 56.1 ± 53.6. At the second visit, the mean ± SD investigator LFA-REAL score was 41.3 ± 36.7, and clinician LFA-REAL score 48.3 ± 42.6. Total LFA-REAL scores correlated positively with PGA, SLEDAI, and BILAG (ρ = 0.58-0.88, P < 0.001). LFA-REAL scores produced correlation coefficients of ρ > 0.7 for musculoskeletal, mucocutaneous, and renal BILAG domains. The intraclass correlation coefficient between the LFA-REAL scores of investigators and clinicians was 0.79 for visit 1 (P < 0.001) and 0.86 for visit 2 (P < 0.001). CONCLUSION:The LFA-REAL provides a reliable surrogate for more complicated disease activity measures when used by lupus clinical investigators or clinicians.