Baseline Predictors of Systemic Lupus Erythematosus Flares

Transcription

1 ARTHRITIS & RHEUMATISM Vol. 65, No. 8, August 2013, pp DOI /art , American College of Rheumatology Baseline Predictors of Systemic Lupus Erythematosus Flares Data From the Combined Placebo Groups in the Phase III Belimumab Trials Michelle A. Petri, 1 Ronald F. van Vollenhoven, 2 Jill Buyon, 3 Roger A. Levy, 4 Sandra V. Navarra, 5 Ricard Cervera, 6 Z. John Zhong, 7 and William W. Freimuth, 7 on behalf of the BLISS-52 and BLISS-76 Study Groups Objective. To identify predictors of moderate-tosevere systemic lupus erythematosus (SLE) flare in 562 patients treated with standard therapy alone in phase III belimumab trials, and to evaluate the impact of standard therapies on preventing flares. Methods. Post hoc analysis assessed baseline ClinicalTrials.gov identifiers: NCT and NCT Supported by GlaxoSmithKline and Human Genome Sciences, Inc. 1 Michelle A. Petri, MD, MPH: Johns Hopkins University School of Medicine, Baltimore, Maryland; 2 Ronald F. van Vollenhoven, MD, PhD: Karolinska Institute, Stockholm, Sweden; 3 Jill Buyon, MD: New York University Langone Medical Center, New York, New York; 4 Roger A. Levy, MD, PhD: Hospital Universitário Pedro Ernesto Universidade do Estado do Rio de Janeiro, Rio de Janeiro, Brazil; 5 Sandra V. Navarra, MD: University of Santo Tomas Hospital, Manila, Philippines; 6 Ricard Cervera, MD, PhD, FRCP: Hospital Clínic, Barcelona, Spain; 7 Z. John Zhong, PhD, William W. Freimuth, MD, PhD: Human Genome Sciences, Inc., Rockville, Maryland. Drs. Petri and Navarra have received consulting fees, speaking fees, and/or honoraria from GlaxoSmithKline and Human Genome Sciences, Inc. (less than $10,000 each) and were investigators in the BLISS-52 and BLISS-76 trials. Dr. van Vollenhoven has received consulting fees, speaking fees, and/or honoraria from Abbott, Bristol- Myers Squibb, GlaxoSmithKline, Human Genome Sciences, Inc., MSD, Pfizer, Roche, and UCB (less than $10,000 each) and was an investigator in the BLISS-52 and BLISS-76 trials. Dr. Buyon has received consulting fees from GlaxoSmithKline and Human Genome Sciences, Inc. (more than $10,000 each) and was an investigator in the BLISS-52 and BLISS-76 trials. Dr. Levy has received consulting fees, speaking fees, and/or honoraria from GlaxoSmithKline and Human Genome Sciences, Inc. (more than $10,000 each) and was an investigator in the BLISS-52 and BLISS-76 trials. Dr. Cervera has received consulting fees, speaking fees, and/or honoraria from Glaxo- SmithKline, Human Genome Sciences, Inc., UCB, Roche, MedImmune, and Cephalon (less than $10,000 each) and was an investigator in the BLISS-52 and BLISS-76 trials. Drs. Zhong and Freimuth own stock or stock options in Human Genome Sciences, Inc. Address correspondence to Michelle A. Petri, MD, MPH, Johns Hopkins University School of Medicine, Division of Rheumatology, 1830 East Monument Street, Suite 7500, Baltimore, MD mpetri@jhmi.edu. Submitted for publication November 14, 2012; accepted in revised form April 24, demographics, disease activity, and biomarkers in patients with and those without flare at treatment weeks 24 and 52. Severe flare was defined by the modified SLE Flare Index (SFI) and the development of any new British Isles Lupus Assessment Group (BILAG) A domain score. Severe and moderate flare was defined by development of 1 new BILAG A domain score or 2 new BILAG B domain scores. Baseline characteristics associated with a >10% absolute difference or a >50% increase in flare rates were considered predictive. Results. Frequencies of flares over 52 weeks according to the SFI, any new BILAG A domain score, and 1 new BILAG A domain score or 2 new BILAG B domain scores were 23.7%, 23.1%, and 32.0%, respectively. Flare predictors by univariate analysis on all 3 indices at weeks 24 and 52 were a score >12 on the Safety of Estrogens in Lupus Erythematosus National Assessment version of the SLE Disease Activity Index (SELENA SLEDAI); anti double-stranded DNA (antidsdna) positivity; proteinuria (>0.5 gm/24 hours); BILAG renal, vasculitic, and hematologic scores; elevated C-reactive protein levels; and B lymphocyte stimulator (BLyS) levels >2 ng/ml. Independent predictors by multivariate analysis at week 52 were SELENA SLEDAI and/or BILAG renal involvement and antidsdna >200 IU/ml (on all 3 indices); SELENA SLEDAI and/or BILAG neurologic and vasculitic involvement (on 2 indices: any new BILAG A domain score and 1 new BILAG A domain score or 2 new BILAG B domain scores); BLyS levels >2 ng/ml (on 2 indices: the SFI and 1 new BILAG A domain score or 2 new BILAG B domain scores); and low C3 level (on the SFI). Baseline medications did not significantly decrease or increase moderate-to-severe SLE flare risk. Conclusion. Patients who were receiving standard SLE therapy and had renal, neurologic, or vasculitic 2143

2 2144 PETRI ET AL involvement, elevated anti-dsdna or BLyS levels, or low C3 had increased risk of clinically meaningful flare over 1 year. Hydroxychloroquine use was not predictive. The clinical course of systemic lupus erythematosus (SLE) is highly variable (1). Flare (mild, moderate, or severe) is common, occurring in 65 70% of patients with SLE within 1 year (2 5). Many investigators have attempted to identify predictors of SLE flare, with inconsistent results. The absence of an accepted definition of any SLE flare complicates the interpretation and comparison of findings (5). There is, however, good agreement on what constitutes severe flare, which is more clinically meaningful as it is usually associated with a change in treatment (2,4). Identification of patients at increased risk of moderate-to-severe flares may also improve efficiency of clinical trials and patient clinical management. Studies from the Hopkins Lupus Cohort found that decreased levels of anti double-stranded DNA (anti-dsdna) were most likely to occur on the day of a flare (6,7). In a prospective study from the same group, African American race and low C4 level at baseline independently predicted flare over the course of 1 year (5). Anti-dsDNA positivity and low C3 level at baseline predicted 1 British Isles Lupus Assessment Group (BILAG) (8) A domain score or 2 B domain scores, but neither was a predictor in a multiple regression model. Similarly, in another prospective study, significant elevations in anti-dsdna titers preceded 24 of 33 flares by 8 10 weeks (9). In another study, baseline erythrocyte sedimentation rate, anemia, lymphopenia, and antinuclear antibody (ANA) titer predicted SLE flare ( 3-point increase in SLE Disease Activity Index [SLEDAI] [10] score) over the course of 1 year (3). Neither C4 nor anti-dsdna was significantly associated with flare. Further, when evaluating flares ( 4-point increase in SLEDAI 2000 [SLEDAI-2K] [11] score) over 1 3 years, only the adjusted mean SLEDAI-2K score was a modest predictor of flare (12,13). It has been observed that patients with active disease are more likely to have a clinically meaningful flare in the most commonly affected organ domains (musculoskeletal, mucocutaneous, renal, and neurologic), and predictors can vary across organ domain flares (2 5,12,14,15). In patients from the Hopkins Lupus Cohort, low baseline SLE disease activity (score 4 on the Safety of Estrogens in Lupus Erythematosus National Assessment version of the SLEDAI [SELENA SLEDAI] [16]) and concomitant high baseline levels of interferon-regulated chemokines were found to predict flare (64% with a high chemokine score had flares versus 28% with a low chemokine score; P 0.001) (17). Potential predictors of flare were assessed in patients with serologically active but clinically quiescent SLE (global BILAG scores 6 and anti-dsdna levels 50 units/ml on 2 occasions), and 81% experienced flare over the course of 5 years (18). Antinucleosome antibody positivity and markedly elevated anti-dsdna antibody levels were associated with significantly shorter times to first flare, and higher absolute antinucleosome titers were associated with multiple episodes of flare. Studies that examined whether medications affect SLE flare risk produced mixed results. Discontinuation of hydroxychloroquine (HCQ) therapy was associated with a higher flare rate in 1 study (19), while another study found that baseline HCQ use did not affect the frequency of BILAG A or B domain scores (5). In another trial, patients with stable SLE that became serologically active were then treated with 1 2 weeks of placebo or prednisone (20). Over the subsequent 3 months, severe flare occurred in 30% of patients treated with placebo versus 0% of patients treated with prednisone. Clinical flare is a common cause of hospitalization (21). The total number of flares per provider visit has been associated with increased risk of organ damage and death (22). Flare also results in higher healthcare costs (23). In the 2 phase III trials of belimumab (a human recombinant IgG1 monoclonal antibody that specifically inhibits the biologic activity of soluble B lymphocyte stimulator [BLyS]), the Study of Belimumab in Subjects with SLE 52-week trial (BLISS-52) and the BLISS 76-week trial (BLISS-76) (ClinicalTrials.gov identifiers NCT and NCT , respectively), patients with autoantibody-positive SLE were randomized to treatment with placebo or belimumab, while also receiving standard SLE therapy (24,25). The SLE Responder Index (SRI) response rate at week 52 was significantly greater with both doses of belimumab (1 mg/kg and 10 mg/kg) versus placebo in BLISS-52, and with 10 mg/kg belimumab in BLISS-76. The present retrospective post hoc combined analysis of baseline demographics, SLE disease activity, and laboratory biomarkers was conducted to identify predictors of moderate-to-severe flare at treatment weeks 24 and 52 in patients randomized to receive standard therapy alone in the BLISS trials (n 562). Two different definitions of severe SLE flare and a third definition evaluating both severe and moderate SLE flares were examined by univariate and multivariate analyses. A

3 PREDICTORS OF FLARE IN SLE 2145 Table 1. trials* Baseline characteristics of the patients treated with standard SLE therapy alone in the BLISS BLISS-52 (n 287) BLISS-76 (n 275) All (n 562) Demographics Women Age, mean SD years Race White Asian African descent Disease characteristics SLE disease duration, mean SD years SELENA SLEDAI Mean SD score Score Proteinuria Vasculitis Neurologic Low complement Increased DNA binding BILAG involvement 1 A domain score or 2 B domain scores A domain score Renal domains A C Vasculitis domain A or B Hematologic domain A or B Neurologic domain A or B Medication Corticosteroid Prednisone or equivalent 7.5 mg/day Dosage, mean SD mg/day Other immunosuppressive agents Mycophenolate mofetil Azathioprine Methotrexate Biomarkers ANA titer 1: Anti-dsDNA 30 IU/ml Anti-Sm 15 units/ml Low C3 ( 900 mg/liter) Low C4 ( 160 mg/liter) CRP positive ( 3 mg/liter) BLyS 0.5 ng/ml Proteinuria 0.5 gm/24 hours * Except where indicated otherwise, values are the percent of patients. SLE systemic lupus erythematosus; BLISS-52 Study of Belimumab in Subjects with SLE 52-week trial; BLISS-76 BLISS 76-week trial; SELENA SLEDAI Safety of Estrogens in Lupus Erythematosus National Assessment version of the SLE Disease Activity Index; BILAG British Isles Lupus Assessment Group; ANA antinuclear antibody; anti-dsdna anti double-stranded DNA; CRP C-reactive protein; BLyS B lymphocyte stimulator. second objective was to evaluate the impact of individual standard therapies on these 3 moderate-to-severe flare rates over time. PATIENTS AND METHODS Study design. BLISS-52 (n 865) and BLISS-76 (n 819) were multicenter, multinational, placebo-controlled trials in which patients with active SLE (SELENA SLEDAI score 6 at screening) were randomized to standard therapy plus either placebo or belimumab at 1 or 10 mg/kg (24,25). In addition to active SLE, other inclusion criteria included autoantibody positivity (ANA titer 1:80 or anti-dsdna level 30 IU/ml), stable standard SLE therapy for 30 days, no severe active lupus nephritis or severe active central nervous system lupus, no intravenous cyclophosphamide within 6 months of enrollment, and no intravenous immunoglobulin or prednisone ( 100 mg/day) within 3 months. Changes to standard therapy (immunosuppressive and antimalarial drugs)

4 2146 PETRI ET AL Figure 1. Baseline disease activity in patients with and patients without flare as defined by the modified Systemic Lupus Erythematosus (SLE) Flare Index at week 52 (A) and week 24 (B), by 1 new British Isles Lupus Assessment Group (BILAG) A domain score or 2 new BILAG B domain scores at week 52 (C), and by any new BILAG A domain score at week 52 (D). * P 0.05; P 0.01; # P versus no flare. SS Safety of Estrogens in Lupus Erythematosus National Assessment version of the SLE Disease Activity Index. were restricted after 16 weeks of treatment. The prednisone dose was unrestricted for the first 24 weeks, but thereafter the dose had to be within 25% or 5 mg of the baseline dose, with no further increases for the remainder of the study. Addition of an angiotensin-converting enzyme inhibitor after 4 months, or statins after 6 months, was prohibited. Patients received study treatment for 52 weeks in BLISS-52 and for 76 weeks in BLISS-76 (24,25). The primary efficacy end point was the SRI response rate at week 52, defined as a 4-point reduction in SELENA SLEDAI score, no new BILAG A domain score and 1 new B domain score, and no worsening in physician s global assessment score compared with baseline. In both BLISS trials, each site was required to obtain ethics committee/institutional review board approval of the final study protocol (24,25). Patients rights, safety, and wellbeing were protected based on the principles of the Declaration of Helsinki. Informed consent was obtained from each patient prior to study screening. SLE flare definitions. This post hoc analysis used the following 3 definitions of SLE flare: severe flare as defined by the SLE Flare Index (SFI), modified to exclude the single criterion of an increase in the SELENA SLEDAI score to 12 (4); development of any new BILAG A domain score (considered severe flare) (2); development of 1 new BILAG A domain score or 2 new BILAG B domain scores (an inclusive definition that comprises patients with severe flares [1 new BILAG A domain score] and patients with moderate flares [2 new BILAG B domain scores]) (2). Baseline demographic variables and disease activity measures. Demographic variables evaluated as possible predictors of moderate-to-severe SLE flare included race, Hispanic/Latino origin, age, sex, height, weight, and body mass index. Baseline values of the following measures of SLE disease activity were also considered as possible predictors of moderate-to-severe flare: SLE duration; SELENA SLEDAI (category [0 3, 4 9, 10 11, 12], mean score, range, and scores for the 24 individual items reported when there were 30 patients per group and for the 9 organ systems [see Supporting Appendix A for the guidelines used for scoring proteinuria for the SELENA SLEDAI, available on the Arthritis & Rheumatism web site at doi/ /art.37995/abstract]) (10); physician s global assessment (category, mean and median scores, range); BILAG A E domain scores for the 8 organ domains (26); Systemic Lupus International Collaborating Clinics/American College of Rheumatology (ACR) Damage Index (SDI) (27) (mean and median score range); and ACR SLE classification criteria (28) (presence/absence of each criterion; SLE is diagnosed when 4 criteria are present, serially or simultaneously). Baseline abnormal laboratory biomarker values. The proportions of patients with abnormal baseline values were

5 PREDICTORS OF FLARE IN SLE 2147 Figure 2. Baseline laboratory values in patients with and patients without flare as defined by the modified SLE Flare Index (SFI) at week 52 (A) and week 24 (B), by 1 new BILAG A domain score or 2 new BILAG B domain scores at week 52 (C), and by any new BILAG A domain score at week 52 (D). P 0.05; P 0.01; # P versus no flare. CRP C-reactive protein; anti-dsdna anti double-stranded DNA; BLyS B lymphocyte stimulator (see Figure 1 for other definitions). calculated for the following biomarkers to evaluate whether there was a threshold effect of different markers of biologic activity: ANA titer ( 1:80, 1:80 1:1,280, 1:1,280), antidsdna level ( 30, , 200 IU/ml), anti-sm antibody (negative/positive [ 15 units/ml]), C3 (high/low, 1,800/ 900 mg/liter), C4 (high/low, 47/ 16 mg/dl), C-reactive protein (CRP) levels, proteinuria ( 0.5, , , 2.0 gm/24 hours), IgG, IgA, and IgM levels, and BLyS quartile levels. The analysis also considered the predictive value of baseline measures of B and T cell subsets and ratios. Data for B and T cell subsets were obtained from BLISS-76 only. The numbers of dual-positive CD3 CD4 and CD3 CD8 T cells and the CD4 :CD8 ratio (10 9 /liter) were determined. Baseline counts per mm 3 of the following B cell subsets were assessed for predictive value: CD19, CD20, CD20 / CD27 (memory), and CD20 /CD27 (naive). In addition, the following counts of subsets per ml were evaluated using the formula for rare subset count (rare subset count/ml rare event count/cd19 event count CD19 count/mm 3 1,000): CD20 /CD69 (activated B cells), CD27 high /CD20 (short-lived plasma cells), CD138 /CD20 (CD138 plasma cells), CD138 /CD20 (plasmacytoid cells), and CD38 high / CD27 high (SLE subset). Baseline standard and adjunctive SLE therapies. Baseline use of the following medications was evaluated to identify predictors of moderate-to-severe SLE flare: corticosteroids (mean dosage, patients receiving 7.5 mg/day versus patients receiving 7.5 mg/day), antimalarial therapy, immunosuppressive therapy, mycophenolate mofetil, azathioprine, methotrexate, angiotensin pathway antihypertensives, aspirin, nonsteroidal antiinflammatory drugs (NSAIDs), and hydroxymethylglutaryl-coenzyme A reductase inhibitors. Statistical analysis. Rates of SLE flares observed during the first 24 weeks and 52 weeks of the studies were calculated using each of the 3 flare indices for groups of patients categorized by each demographic variable, baseline disease activity measure, and laboratory biomarker examined. A clinically meaningful difference in flare rate at week 24 or 52 was prospectively defined prior to the conduct of the analysis as a 10% absolute difference or a 50% increase associated with the presence or absence of a baseline variable. Laboratory biomarkers were defined as categorical variables (present or absent at baseline). Univariate comparisons of flare rates between patient groups were performed using a likelihood ratio test and were reported for baseline disease activity characteristics, with a minimum of 20 subjects. Univariate analysis for neurologic organ system (n 17) was not performed; however, these data were included in the multivariate analysis. A multivariate Cox proportional hazards model was used to determine baseline variables that were predictive of SLE flare based on a stepwise forward-selection process using a selection criterion at an alpha level of The impact of changes over time in any of the variables was not examined.

6 2148 PETRI ET AL Table 2. Baseline disease activity and biomarkers in the 562 patients with flare by treatment week 52* Category Total number of patients in category Modified SFI for severe flare (n 133) One new BILAG A domain score or 2 new BILAG B domain scores (n 180) Any new BILAG A domain score (n 130) SELENA SLEDAI score (25.0) 1 (25.0) 1 (25.0) (17.8) 66 (25.5) 46 (17.8) (22.6) 43 (31.4) 32 (23.4) (34.0) 70 (43.2) 51 (31.5) BILAG score No A or B domain 45 8 (17.8) 15 (33.3) 7 (15.6) 1 B domain (24.2) 165 (31.9) 123 (23.8) 1 A domain or 2 B domains (26.1) 108 (30.6) 93 (26.3) 1 A domain (28.1) 21 (23.6) 19 (21.3) Proteinuria, gm/24 hours (20.3) 127 (28.7) 88 (19.9) (31.8) 20 (45.5) 16 (36.4) (41.9) 23 (53.5) 17 (39.5) (34.4) 10 (31.3) 9 (28.1) ANA titer 1: (20.0) 14 (31.1) 11 (24.4) 1:80 1:1, (21.3) 76 (31.7) 59 (24.6) 1:1, (26.4) 90 (32.5) 60 (21.7) Anti-dsDNA, IU/ml (16.4) 45 (24.6) 35 (19.1) (19.3) 55 (28.6) 35 (18.2) (35.3) 80 (42.8) 60 (32.1) BLyS quartile Quartile 1 ( ng/ml) (16.8) 36 (26.3) 27 (19.7) Quartile 2 ( to ng/ml) (18.6) 38 (27.1) 25 (17.9) Quartile 3 ( to ng/ml) (20.3) 39 (28.3) 27 (19.6) Quartile 4 ( ng/ml) (39.6) 63 (45.3) 48 (34.5) * Values are the number (%) of patients of the total number of patients within a category. SFI SLE Flare Index (see Table 1 for other definitions). Enrollment was based on an initial screening SELENA SLEDAI score of 6; in some patients, however, disease activity declined during the 35-day window from screening to day 0/baseline. RESULTS Baseline characteristics. Table 1 summarizes baseline characteristics of patients from the BLISS trials randomized to receive standard SLE therapy alone. SLE flare rates. The SLE flare rates at week 24 according to each of the 3 indices used were as follows: 13.7% of patients (n 77) had a severe flare according to the SFI (severe SFI flare), 19.8% of patients (n 111) had 1 new BILAG A domain score or 2 new BILAG B domain scores, and 12.6% (n 71) had any new BILAG A domain score. By treatment week 52, 23.7% of patients (n 133) had a severe SFI flare, 32.0% (n 180) had 1 new BILAG A domain score or 2 new BILAG B domain scores, and 23.1% (n 130) had any new BILAG A domain score. Baseline disease activity. Five baseline disease activity measures predicted SLE flares on all 3 indices at both weeks 24 and 52 according to univariate analysis: a SELENA SLEDAI score 12, moderate-to-severe disease activity involving SELENA SLEDAI defined proteinuria, renal involvement (BILAG domains A, B, or C), vasculitic involvement (BILAG domains A or B), and hematologic involvement (BILAG domains A or B) (Figure 1) (week 24 data are depicted for only the modified SFI; the results using the other 2 indices were similar and are not shown). The SELENA SLEDAI serologic markers of low C3/C4 and anti-dsdna titers at baseline predicted severe SFI flare at week 52 and 1 new BILAG A domain score or 2 new BILAG B domain scores, but not any new BILAG A domain score. Baseline disease activity measures not predictive of flare were SLE duration; physician s global assessment; SDI; ACR classification criteria; BILAG general, mucocutaneous, neuropsychiatric, musculoskeletal, and cardiorespiratory involvement; and SELENA SLEDAI elements other than score 12 and proteinuria. Data for variables that were not predictive on all 3 indices and at both weeks 24 and 52 are not shown. Abnormal levels of laboratory biomarkers at baseline. According to univariate analysis, abnormal baseline measures of 4 laboratory biomarkers elevated CRP level, proteinuria 0.5 gm/24 hours, anti-dsdna

7 PREDICTORS OF FLARE IN SLE 2149 Figure 3. Baseline SLE medication use in patients with and patients without flare as defined by the modified SLE Flare Index at week 52 (A) and week 24 (B), by 1 new BILAG A domain score or 2 new BILAG B domain scores at week 52 (C), and by any new BILAG A domain score at week 52 (D). P 0.05 versus no flare. MMF mycophenolate mofetil; AZA azathioprine; MTX methotrexate (see Figure 1 for other definitions). positivity, and serum BLyS level 2 ng/ml predicted SLE flare on all 3 indices at both week 24 and week 52 (Figure 2) (week 24 data are depicted only for the modified SFI; the results with the other 2 indices were similar and are not shown). Low C3 and C4 levels predicted flare on 2 indices (severe SFI flare and 1 new BILAG A domain score or 2 new BILAG B domain scores) at week 52 and on all 3 indices at week 24. Table 3. Predictors of flare over 52 weeks by multivariate analysis* Modified SFI for severe flare Renal organ involvement by SELENA SLEDAI and/ or BILAG HR (95% CI) One new BILAG A domain score or 2 new BILAG B domain scores 1.93 ( ) Renal organ involvement by SELENA SLEDAI and/or BILAG HR (95% CI) Any new BILAG A domain score 1.98 ( ) Renal organ involvement by SELENA SLEDAI and/or BILAG HR (95% CI) 2.36 ( ) Anti-dsDNA 200 IU/ml 1.73 ( ) Anti-dsDNA 200 IU/ml 1.63 ( ) Anti-dsDNA 200 IU/ml 1.83 ( ) Baseline BLyS 2 ng/ml 1.86 ( ) Baseline BLyS 2 ng/ml 1.45 ( ) Neurologic organ involvement by SELENA SLEDAI and/ or BILAG 2.47 ( ) Neurologic organ involvement by SELENA SLEDAI and/ or BILAG 3.11 ( ) Vasculitic organ involvement by SELENA SLEDAI and/ or BILAG 1.73 ( ) Vasculitic organ involvement by SELENA SLEDAI and/ or BILAG 1.88 ( ) Low C ( ) * The statistical model included demographic characteristics, baseline disease activity characteristics, and baseline laboratory values simultaneously. BILAG organ domain involvement was defined as an A or B score at baseline, with the exception of renal involvement, which was defined as an A, B, or C score at baseline. SFI SLE Flare Index; HR hazard ratio; 95% CI 95% confidence interval (see Table 1 for other definitions).

8 2150 PETRI ET AL Anti-Sm positivity ( 15 units/ml) predicted flare at week 24 based on severe SFI flare and 1 new BILAG A domain score or 2 new BILAG B domain scores; however, it was not predictive of flare on any index at week 52. Baseline biomarkers not predictive of moderate-tosevere flares at weeks 24 and 52 on all 3 indices included ANA titer, immunoglobulins, and B and T cell subsets (data not shown). Baseline levels of predictors. An analysis of SLE flare rates according to baseline values at specific cut points was performed to identify potential threshold effects where elevated levels of disease activity or a biomarker were predictive of an SLE flare. Threshold effects leading to increased flare rates were found to correlate with higher baseline levels for the following variables: SELENA SLEDAI scores 12, proteinuria levels 0.5 gm/24 hours, anti-dsdna titer 200 IU/ml, and BLyS levels 2 ng/ml (Table 2). Predictive value of baseline standard and adjunctive SLE therapies. None of the baseline medications evaluated, including corticosteroids, immunosuppressives (individually and as a group), antimalarials, and other concomitant medications, predicted flare on any index at either week 24 or week 52, with the exception of any steroid on the SFI at week 24 (Figure 3). Multivariate analysis. The presence of renal organ involvement (as measured by the SELENA SLEDAI and/or BILAG) and anti-dsdna titer 200 IU/ml at baseline were independent predictors of moderate-to-severe SLE flare at week 52 (Table 3). This finding was consistent across all 3 flare indices on multivariate analysis. Neurologic and vasculitic organ involvement at baseline (as measured by the SELENA SLEDAI and/or BILAG) each independently predicted flare at week 52 according to 2 indices (1 new BILAG A domain score or 2 new BILAG B domain scores and any new BILAG A domain score). Baseline BLyS level 2 ng/ml independently predicted flare at week 52 on 2 indices (SFI and 1 new BILAG A domain score or 2 new BILAG B domain scores). Low C3 level was an independent predictor of a severe SFI flare at week 52. DISCUSSION In this study, multiple baseline disease activity characteristics involving organ systems most likely to be associated with severe or life-threatening SLE manifestations, serologic activity, and markedly elevated BLyS levels were found to be predictors of moderate-to-severe SLE flare over 1 year. This analysis has several strengths: inclusion of data from a large number of patients, use of 3 SLE flare indices focusing on severe or moderate and severe flares in the absence of an accepted definition of flare, and application of multivariate analysis. The 3 flare indices demonstrated high concordance regarding predictors. Nearly all variables (except low C3/C4 and anti-dsdna positivity) that were significant predictors on univariate analysis using 1 flare index were also significant predictors with the other 2 indices at both week 24 and week 52. There was a difference in rates of flares across the flare indices. By treatment week 52, the index consisting of 1 new BILAG A domain score or 2 new BILAG B domain scores had the greatest flare rate (32% of patients; 23.1% of patients had any new BILAG A domain score), as it was more inclusive (it identified moderate flare if at least 2 new B organ domain scores were identified) and is the same criterion used to show clinically meaningful disease worsening in the SRI efficacy end point (24,25). The modified SFI and any new BILAG A domain score flare indices had nearly identical frequencies of severe flares (23 24% of patients). In multivariate analysis, severe flares by these 2 indices had many similar characteristics (common independent predictors were renal disease and anti-dsdna titer 200 IU/ml), but the modified SFI identified baseline low C3 and BLyS level 2 ng/ml, and any new BILAG A domain score flare index identified baseline vasculitic and neurologic disease activity as independent predictors. The modified SFI defined severe flare requires triggering disease manifestations to be accompanied by significant changes in immunosuppressive agents or by hospitalization (4), whereas any new BILAG A domain score is usually associated with a change in therapy (2), although this is not required. This study provides data that help us to identify patients at greater risk of severe SLE flare; these data should also help us to select the most appropriate SLE flare index in clinical trials assessing the ability of new therapies to reduce rates of moderate-to-severe lupus flare. Trial design would depend on patient population and sample size (power calculations), organ disease manifestations, serologic activity, overall primary efficacy end point, disease activity indices used, and desired trial duration. In the multivariate analysis, 2 variables antidsdna titer 200 IU/ml and BILAG and/or SELENA SLEDAI renal involvement independently predicted moderate-to-severe SLE flare using all 3 indices. BILAG and/or SELENA SLEDAI neurologic and vasculitic involvement were independent predictors of flares according to any new BILAG A domain score index and

9 PREDICTORS OF FLARE IN SLE 2151 according to the 1 new BILAG A domain score or 2 new BILAG B domain scores index. Baseline BLyS levels 2 ng/ml were an independent predictor of flares according to the modified SFI and according to the 1 new BILAG A domain score or 2 new BILAG B domain scores index. Low C3 level independently predicted flare, using the modified SFI. BLyS levels have been shown to correlate with SLE disease activity over time and with autoantibody levels (29 31). Of note, BLyS levels in the BLISS trials were not predictive of an SRI response with belimumab treatment, which suggests that BLyS correlated more strongly with variations in SLE disease activity, such as that observed with moderate-tosevere SLE flare (32). In addition, early normalization of complement levels or anti-dsdna titers by week 8 was found to be associated with a reduced risk of severe flare in all patients, irrespective of treatment group. Comparison of results from the BLISS trials with those of previous studies is complicated by the differing definitions of SLE flare, study designs, variables, baseline patient characteristics, sample size, and followup periods. One study identified anti-dsdna positivity as predicting flare over the course of 1 year (5). In another study, increased anti-dsdna levels preceded flare by 8 10 weeks (9). In a third trial, anti-dsdna levels 5 times the upper limit of normal were associated with a shorter time to first flare in patients with serologically active but clinically quiescent disease (18). Other investigators reported no consistent association between antidsdna positivity or low complement levels and risk of flare (3,5 7,12). The Toronto Lupus Cohort (average SLEDAI-2K score 3.7) evaluated the determinants of any flare and persistently active disease over 1 3-year periods and identified modest predictors of SLE flare (SLEDAI-2K score increase 4 points from previous visit) such as the adjusted mean SLEDAI-2K score over 2 3 years and steroid use in the first year (12). Different SLE flare definitions focusing on different or lesser degrees of severity with or without required therapeutic changes have yielded different results. The present study did not identify any significant demographic predictor of flare. An analysis from the Hopkins Lupus Cohort showed that African American race independently predicted risk of flare over the course of 1 year (5). The differing proportions of patients of African descent in the present analysis and the Hopkins study (9% and 37%, respectively) may account for this divergence. Evidence is conflicting regarding whether any drug therapy, given either as maintenance therapy or in response to serologic indicators of increased disease activity, can affect risk of SLE flare. Baseline use of medication generally was not associated with a lower risk of subsequent flare in the present study. The exception was that use of any corticosteroid at baseline was associated with a higher flare rate at week 24 according to the SFI, although this was not confirmed by multivariate analysis. An earlier Canadian study, however, showed that maintaining HCQ therapy was associated with reduced risk of flare short term (19) and long term (33). Patients were randomized to withdrawal or continuation of the previously prescribed HCQ. Those in whom the drug therapy was stopped had a 2.5-fold higher rate of flare (mild-to-severe) over the course of 6 months and an 2-fold increase in major flares (flare definition was to increase prednisone by 10 mg/day or start a new immunosuppressive agent) over 3.5 years (33). It should be noted that the Canadian study comprised patients with clinically quiescent SLE who had continued HCQ therapy for 6 months, while the BLISS trials examined patients with active seropositive disease. The placebo group in the Canadian study included 10-fold fewer patients than the BLISS trials, and the Canadian study evaluated the effect of withdrawing a maintenance treatment, while the present analysis assessed the impact of stable SLE therapies (19,24,25,33). Patients in the BLISS trials receiving antimalarials did have numerically lower flare rates at week 52 according to 2 of the 3 flare indices and at week 24 on all 3 indices. Standard therapy in the Canadian study included only prednisone and NSAIDs, while in the BLISS trials it also included immunosuppressive agents such as methotrexate, mycophenolate mofetil, and azathioprine. In the Plaquenil LUpus Systemic study, attempted optimization of HCQ dose over 7 months did not alter flare rate but, importantly, lower flare rates were observed in those who actually did achieve target levels of HCQ during the second 7 months of followup (34). It is also worth noting that in the Hopkins Lupus Cohort, baseline use of prednisone or immunosuppressive agents was associated with an increased risk of BILAG A domain and B domain scores, while a protective effect did not hold true for HCQ use (5). Finally, using a different approach, Tseng et al (20) found that initiating prednisone therapy based on serologic disease activity markers in clinically stable patients reduced severe flare risk. The present study has the limitations of any retrospective post hoc analysis. The definition of a clinically significant difference in SLE flare rates used in this investigation requires prospective evaluation. The

10 2152 PETRI ET AL disease activity measure and laboratory biomarker predictors identified herein require validation in a prospective clinical trial. Although the number of neurologic events that occurred was small, these events were found to be independent predictors of flare in the multivariate analysis. This may be a result of the severity of these manifestations, but future analysis of greater numbers of these events is needed to confirm these findings. This analysis evaluated only baseline variables; it does not describe clinical (organ-specific) or biomarker changes immediately preceding the diagnosis of a new flare. These are topics for future analyses. In summary, anti-dsdna titers 200 IU/ml and renal organ involvement at baseline were independent predictors of moderate-to-severe SLE flare at week 52 on all 3 flare indices. Neurologic and vasculitic organ involvement and levels of BLyS 2 ng/ml at baseline each independently predicted flare at week 52 on 2 of the indices. Low C3 level was an independent predictor of severe SFI flare at week 52. Flare risks were generally similar across standard SLE therapies, including immunosuppressives, antimalarials, and corticosteroids. Close monitoring of patients with disease activity or biomarkers predictive of SLE flare may improve their care and long-term outcomes. Identification of flare predictors may also improve the efficiency of clinical trials by enabling researchers to balance distribution of predictors among subjects at baseline, adjust for predictors in statistical analysis, and recruit patients at higher or lower risk of flare as appropriate to the objective. ACKNOWLEDGMENTS Editorial support was provided by Eileen McCaffrey and Eleanore Gross of BioScience Communications, New York, NY, and was funded by GlaxoSmithKline and Human Genome Sciences, Inc. AUTHOR CONTRIBUTIONS All authors were involved in drafting the article or revising it critically for important intellectual content, and all authors approved the final version to be published. Dr. Petri had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis. Study conception and design. Petri, van Vollenhoven, Zhong, Freimuth. Acquisition of data. Petri, van Vollenhoven, Levy, Navarra, Cervera, Zhong, Freimuth. Analysis and interpretation of data. Petri, van Vollenhoven, Buyon, Levy, Navarra, Cervera, Zhong, Freimuth. ROLE OF THE STUDY SPONSORS GlaxoSmithKline and Human Genome Sciences, Inc. facilitated the study design, provided editorial support for the manuscript, and reviewed and approved the manuscript prior to submission. The authors independently collected the data, interpreted the results, and had the final decision to submit the manuscript for publication. Publication of this article was not contingent upon approval by GlaxoSmithKline and Human Genome Sciences, Inc. REFERENCES 1. Bertsias G, Ioannidis JP, Boletis J, Bombardieri S, Cervera R, Dostal C, et al. EULAR recommendations for the management of systemic lupus erythematosus: report of a Task Force of the EULAR Standing Committee for International Clinical Studies Including Therapeutics. Ann Rheum Dis 2008;67: Gordon C, Sutcliffe N, Skan J, Stoll T, Isenberg DA. Definition and treatment of lupus flares measured by the BILAG index. Rheumatology (Oxford) 2003;42: Mirzayan MJ, Schmidt RE, Witte T. Prognostic parameters for flare in systemic lupus erythematosus. Rheumatology (Oxford) 2000;39: Petri M, Buyon J, Kim M. Classification and definition of major flares in SLE clinical trials. Lupus 1999;8: Petri M, Singh S, Tesfasyone H, Malik A. Prevalence of flare and influence of demographic and serologic factors on flare risk in systemic lupus erythematosus: a prospective study. J Rheumatol 2009;36: Ho A, Barr SA, Magder LC, Petri M. A decrease in complement is associated with increased renal and hematologic activity in patients with systemic lupus erythematosus. Arthritis Rheum 2001;44: Ho A, Magder LS, Barr SG, Petri M. Decreases in anti doublestranded DNA levels are associated with concurrent flares in patients with systemic lupus erythematosus. Arthritis Rheum 2001;44: Symmons DP, Coppock JS, Bacon PA, Bresnihan B, Isenberg DA, Maddison P, et al, and Members of the British Isles Lupus Assessment Group (BILAG). Development and assessment of a computerized index of clinical disease activity in systemic lupus erythematosus. Q J Med 1988;69: Ter Borg EJ, Horst G, Hummel EJ, Limburg PC, Kallenberg CG. Measurement of increases in anti double-stranded DNA antibody levels as a predictor of disease exacerbation in systemic lupus erythematosus: a long-term, prospective study. Arthritis Rheum 1990;33: Bombardier C, Gladman DD, Urowitz MB, Caron D, Chang CH, and the Committee on Prognosis Studies in SLE. Derivation of the SLEDAI: a disease activity index for lupus patients. Arthritis Rheum 1992;35: Gladman DD, Ibanez D, Urowitz MB. Systemic Lupus Erythematosus Disease Activity Index J Rheumatol 2002;29: Nikpour M, Urowitz MB, Ibanez D, Gladman DD. Frequency and determinants of flare and persistently active disease in systemic lupus erythematosus. Arthritis Rheum 2009;61: Ibanez D, Gladman DD, Urowitz MB. Adjusted mean Systemic Lupus Erythematosus Disease Activity Index-2K is a predictor of outcome in SLE. J Rheumatol 2005;32: Illei GG, Takada K, Parkin D, Austin HA, Crane M, Yarboro CH, et al. Renal flares are common in patients with severe proliferative lupus nephritis treated with pulse immunosuppressive therapy: long-term followup of a cohort of 145 patients participating in randomized controlled studies. Arthritis Rheum 2002;46: Borowoy AM, Pope JE, Silverman E, Fortin PR, Pineau C, Smith CD, et al. Neuropsychiatric lupus: the prevalence and autoantibody associations depend on the definition: results from the 1000 face of lupus cohort. Semin Arthritis Rheum 2012;42: Petri M, Kim MY, Kalunian KC, Grossman J, Hahn BH,

11 PREDICTORS OF FLARE IN SLE 2153 Sammaritano LR, et al, for the OC-SELENA Trial. Combined oral contraceptives in women with systemic lupus erythematosus. N Engl J Med 2005;353: Bauer JW, Petri M, Batliwalla FM, Koeuth T, Wilson J, Slattery C, et al. Interferon-regulated chemokines as biomarkers of systemic lupus erythematosus disease activity: a validation study. Arthritis Rheum 2009;60: Ng KP, Manson JJ, Rahman A, Isenberg DA. Association of antinucleosome antibodies with disease flare in serologically active clinically quiescent patients with systemic lupus erythematosus. Arthritis Rheum 2006;55: Canadian Hydroxychloroquine Study Group. A randomized study of the effect of withdrawing hydroxychloroquine sulfate in systemic lupus erythematosus. N Engl J Med 1991;324: Tseng CE, Buyon JP, Kim M, Belmont HM, Mackay M, Diamond B, et al. The effect of moderate-dose corticosteroids in preventing severe flares in patients with serologically active, but clinically stable, systemic lupus erythematosus: findings of a prospective, randomized, double-blind, placebo-controlled trial. Arthritis Rheum 2006;54: Edwards CJ, Lian TY, Badsha H, Teh CL, Arden N, Chng HH. Hospitalization of individuals with systemic lupus erythematosus: characteristics and predictors of outcome. Lupus 2003;12: Stoll T, Sutcliffe N, Mach J, Klaghofer R, Isenberg DA. Analysis of the relationship between disease activity and damage in patients with systemic lupus erythematosus a 5-yr prospective study. Rheumatology (Oxford) 2004;43: Zhu TY, Tam LS, Lee VW, Lee KK, Li EK. The impact of flare on disease costs of patients with systemic lupus erythematosus. Arthritis Rheum 2009;61: Furie R, Petri M, Zamani O, Cervera R, Wallace DJ, Tegzova D, et al, for the BLISS-76 Study Group. A phase III, randomized, placebo-controlled study of belimumab, a monoclonal antibody that inhibits B lymphocyte stimulator, in patients with systemic lupus erythematosus. Arthritis Rheum 2011;63: Navarra SV, Guzman RM, Gallacher AE, Hall S, Levy RA, Jimenez RE, et al, for the BLISS-52 Study Group. Efficacy and safety of belimumab in patients with active systemic lupus erythematosus: a randomised, placebo-controlled, phase 3 trial. Lancet 2011;377: Isenberg DA, Gordon C, BILAG Group, British Isles Lupus Assessment Group. From BILAG to BLIPS disease activity assessment in lupus past, present and future. Lupus 2000;9: Gladman D, Ginzler E, Goldsmith C, Fortin P, Liang M, Urowitz M, et al. The development and initial validation of the Systemic Lupus International Collaborating Clinics/American College of Rheumatology Damage Index for systemic lupus erythematosus. Arthritis Rheum 1996;39: Hochberg MC, for the Diagnostic and Therapeutic Criteria Committee of the American College of Rheumatology. Updating the American College of Rheumatology revised criteria for the classification of systemic lupus erythematosus [letter]. Arthritis Rheum 1997;40: Cheema GS, Roschke V, Hilbert DM, Stohl W. Elevated serum B lymphocyte stimulator levels in patients with systemic immune based rheumatic diseases. Arthritis Rheum 2001;44: Petri M, Stohl W, Chatham W, McCune WJ, Chevrier M, Ryel J, et al. Association of plasma B lymphocyte stimulator levels and disease activity in systemic lupus erythematosus. Arthritis Rheum 2008;58: Zhang J, Roschke V, Baker KP, Wang Z, Alarcon GS, Fessler BJ, et al. A role for B lymphocyte stimulator in systemic lupus erythematosus. J Immunol 2001;66: Stohl W, Hiepe F, Latinis KM, Thomas M, Scheinberg MA, Clarke A, et al, on behalf of the BLISS-52 and BLISS-76 Study Groups. Belimumab reduces autoantibodies, normalizes low complement levels, and reduces select B cell populations in patients with systemic lupus erythematosus. Arthritis Rheum 2012;64: Tsakonas E, Joseph L, Esdaile JM, Choquette D, Senecal JL, Cividino A, et al, the Canadian Hydroxychloroquine Study Group. A long-term study of hydroxychloroquine withdrawal on exacerbations in systemic lupus erythematosus. Lupus 1998;7: Costedoat-Chalumeau N, Galicier L, Aumaitre O, Frances C, Le Guern V, Liote F, et al, on behalf of Group PLUS. Hydroxychloroquine in systemic lupus erythematosus: results of a French multicentre controlled trial (PLUS Study). Ann Rheum Dis E-pub ahead of print.