B-cell depletion in SLE: clinical and trial experience with rituximab and ocrelizumab and implications for study design

B cells are believed to be central to the disease process in systemic lupus erythematosus (SLE), making them a target for new therapeutic intervention. In recent years there have been many publications regarding the experience in SLE of B-cell depletion utilising rituximab, an anti-CD20 mAb that temporarily depletes B cells, reporting promising results in uncontrolled open studies and in routine clinical use. However, the two large randomised controlled trials in extra-renal lupus (EXPLORER study) and lupus nephritis (LUNAR study) failed to achieve their primary endpoints. Based on the clinical experience with rituximab this failure was somewhat unexpected and raised a number of questions and concerns, not only into the true level of benefit of B-cell depletion in a broad population but also how to test the true level of effectiveness of an investigational agent as we seek to improve the design of therapeutic trials in SLE. A better understanding of what went wrong in these trials is essential to elucidate the underlying reasons for the disparate observations noted in open studies and controlled trials. In this review, we focus on various factors that may affect the ability to accurately and confidently establish the level of treatment effect of the investigational agent, in this case rituximab, in the two studies and explore hurdles faced in the randomised controlled trials investigating the efficacy of ocrelizumab, the humanised anti-CD20 mAb, in SLE. Further, based on the lessons learned from the clinical trials, we make suggestions that could be implemented in future clinical trial design to overcome the hurdles faced.


Background
B cells have been targeted in the treatment of systemic lupus erythematosus (SLE) and rheumatoid arthritis (RA) owing to the central role they play in the pathogenesis of these disorders. Th ese cells play a critical role in host defence through their maturation into antibody-secreting plasma cells, secretion of proinfl ammatory cytokines, antigen presentation and co-stimulatory support for T cells. However, dysfunctional recognition of self-antigens as nonself-antigens results in autoantibody production, sustained by plasma cells derived from the B-cell lineage that survive for prolonged periods in the lymphoid tissues. B cells also participate in infl ammatory reactions through antibody-independent mechanisms by acting as antigen-presenting cells and co-stimulation of T cells and other infl ammatory cell types, although as yet there are no validated biomarkers that distinguish pathogenic from protective B-cell subsets. Reagents that specifi cally target pathogenic B-cell subsets are therefore not likely to be available in the near future. Th is reality provides the rationale for targeting B cells in patients with SLE, RA and other autoimmune diseases [1][2][3][4][5].
B-cell-targeted immunotherapy was initially developed for the treatment of B-cell-related malignancies, which are associated with poor prognosis despite aggressive cyto toxic therapies. Of the many surface-expressed antigens on B cells studied as possible targets, CD20 -a trans membrane phosphoprotein expressed in normal B cells as well as 90% of lymphomas -is not shed or modulated, making it an attractive target. In 1994, Reff and colleagues reported a major (95%) and sustained (up to 90 days) B-cell depletion using a murine mAb (2B8) that targeted CD20 on B cells in nonhuman primates [6]. In 1997, a landmark study reported on both the safety and effi cacy of rituximab, a chimeric (mouse-human) mAb directed against CD20, for the treatment of relapsed, refractory low-grade or follicular lymphoma [7]. In November 1997, rituximab was licensed for this indication. Rituximab is now a part of the standard therapeutic regimen in the management of B-cell malignancies and remains among the most successful therapeutic Abstract B cells are believed to be central to the disease process in systemic lupus erythematosus (SLE), making them a target for new therapeutic intervention. In recent years there have been many publications regarding the experience in SLE of B-cell depletion utilising rituximab, an anti-CD20 mAb that temporarily depletes B cells, reporting promising results in uncontrolled open studies and in routine clinical use. However, the two large randomised controlled trials in extra-renal lupus (EXPLORER study) and lupus nephritis (LUNAR study) failed to achieve their primary endpoints. Based on the clinical experience with rituximab this failure was somewhat unexpected and raised a number of questions and concerns, not only into the true level of benefi t of B-cell depletion in a broad population but also how to test the true level of eff ectiveness of an investigational agent as we seek to improve the design of therapeutic trials in SLE. A better understanding of what went wrong in these trials is essential to elucidate the underlying reasons for the disparate observations noted in open studies and controlled trials. In this review, we focus on various factors that may aff ect the ability to accurately and confi dently establish the level of treatment eff ect of the investigational agent, in this case rituximab, in the two studies and explore hurdles faced in the randomised controlled trials investigating the effi cacy of ocrelizumab, the humanised anti-CD20 mAb, in SLE. Further, based on the lessons learned from the clinical trials, we make suggestions that could be implemented in future clinical trial design to overcome the hurdles faced.
mAbs. Interestingly, the response rate is variable amongst individuals with the same histological type of lymphoma as well as the overall response rate between diff erent histological types [8]. Th is suggests that B-cell depletion is not uniform across patients or indeed diseases for reasons yet to be fully understood, but Fcγ receptor function appears important with enhanced Fcγ receptor IIb expression being associated with reduced rituximab effi cacy in lymphoma [9]. Intriguingly, polymorphisms of this receptor are associated with SLE, although their precise role in the disease and potential for targeted therapeutic intervention is not understood.
In 1999, Professor Edwards' group at University College London treated a small number of patients with refractory RA using rituximab, having been encouraged by the safety and effi cacy profi le of induced transient depletion of B cells in haematological malignancies. Th is study and subsequent studies of rituximab in RA, including a large phase II randomised controlled trial, indicated that the treatment was potentially safe and eff ective [10][11][12][13]. Th e regimen in these studies utilised two doses (1,000 mg) of rituximab given 2 weeks apart, with premedication includ ing a single 100 mg intravenous dose of methylprednisolone and 10 mg chlorphenamine. In the original study, patients also received a course of high-dose prednisolone (60 mg for up to 3 weeks and then tapering over the next 3 to 4 weeks or maintaining at 5 mg a day). Responding patients were retreated at or just before predicted relapse. Initially, intravenous cyclophosphamide was used to accompany the rituximab [10]. Th e phase II study showed that cyclophosphamide could be replaced by methotrexate or rituximab on its own, although the response rates were better when rituximab was used in combination with methotrexate. Further, the assigned dose of prednisolone was reduced to 60 mg/day oral prednisone on day 2 and days 4 to 7 and 30 mg/day oral prednisone on days 8 to 14 [11].

Clinical experience of rituximab in SLE
Th e fi rst open, uncontrolled study of rituximab for patients with SLE, by Professor Isenberg's group at University College London, showed improvements in both clinical and laboratory features of disease following treatment with rituximab in refractory SLE [14]; these observations have been supported by the publication of many other similar open, nonrandomised studies [15][16][17][18][19] ( Table 1). Th e University College London regimen employed pre medication with 100 mg intravenous methylprednisolone in addition to 750 mg low-dose intravenous cyclo phos pha mide (for renal manifestations) 1 day prior to the fi rst of two doses of rituximab, given 2 weeks apart. More recently just one dose of cyclophosphamide has been used, and any subsequent need for immunosuppressive therapy is adjusted based on the merits of clinical response and disease manifestation activity that can be assessed using well-validated tools such as the British Isles Lupus Assessment Group (BILAG) 2004 index (for example, using the BLIPS computer software program; LIMATHON, Sheffi eld, UK).
Appreciating this robust clinical management focused on the individual patient -potentially involving multidisciplinary expert opinion, including rheumatologists, dermatologists and renal physicians -is important when comparing the results with those from large multicentre randomised controlled trials with variable quality observations in a broad population.
Worthy of note is that the indication for rituximab at Professor Isenberg's centre is a combination of active disease (renal or nonrenal) (assessed by the BILAG 2004 index) poorly controlled despite at least two standard immunosuppressive agents (not including cortico steroids) used for suffi cient time at optimal doses. To date, 100 patients have been treated at University College London with at least one cycle of rituximab and more than 30 patients have received repeated treatment. Although involving only small numbers, the observations from repeating the regimen showed that improvements in disease, including remission rates, were sustained in patients who responded to the initial treatment [20]. Th is same group has previously demonstrated following B-cell depletion therapy (BCDT) that anti-double-stranded DNA (anti-dsDNA) and anti-nucleosome antibodies reduce to 30 to 40% of baseline, whereas other autoantibodies such as anti-Ro and antibodies to pneumococcal poly saccharide (protective) remain unaltered. Th is observation would suggest that rapidly proliferating clones of B cells may give rise to short-lived plasma cells that produce these anti-dsDNA, anti-cardiolipin and anti-nucleosome antibodies and appear preferentially aff ected by BCDT [21], whereas other autoantibodies such as anti-Ro and anti-RNP or protective antibodies, which develop following immuni sa tion and are thought to be produced by long-lived plasma cells, remain unaltered.
In line with this experience, anti-dsDNA antibody levels tend to fall but not to normalise and these antibodies are probably produced by a combination of shortlived and long-lived plasma cells. Similar to these fi ndings, a post-hoc analysis of the EXPLORER trial focusing on the biological eff ects of rituximab revealed a signi ficant reduction in the levels of anti-dsDNA and anticardiolipin antibodies and a signifi cant increase in complement levels and serum BAFF in the rituximabtreated group versus placebo. Analysis of the repopu lation dynamics of subsets of B cells identifi ed naïve cells as the primary phenotype detected fi rst in circulation; however, the phenotype analysis was limited in that CD27memory cells were not examined in this study [22]. Th e changes in biological eff ects did not translate into clinical benefi ts at 1 year. Whether a long-term follow up with more detailed phenotype analysis at various time points would help predict response to rituximab therapy is not known. However, designing clinical trials to defi ne the precise relationship between the biological eff ects that occur following BCDT and the clinical response in the long term (typically, 2 to 5 years) would be met with the potential challenge of maintaining remission in the placebo group with conventional immunosuppressants alone. Th e eff ects extend to global disease control including an improvement in lipid profi le [23], but such benefi ts are not necessarily captured in randomised controlled trials with a short duration of follow-up.
Recently, following the approach by a group at Imperial College (see later) in a pilot study, eight patients with active disease were treated at diagnosis with rituximab in an attempt to avoid the use of corticosteroids. Using this approach it was possible to reduce the cumulative dose of steroids substantially in fi ve of the eight patients [24], a major long-term advantage.
A recent review of the rituximab experience in approximately 200 patients with refractory SLE, from open studies and real clinical experience, indicated that many would respond at least partially to B-cell depletion [25]. Diff erences in determining endpoints for these studies make it diffi cult to establish formal median and range of improvements. In a phase I/II dose-escalation trial of the safety and effi cacy of rituximab in addition to ongoing therapy in 18 patients with SLE, three dosing regimens of rituximab were studied as follows: six patients received a low dose, a single infusion of 100 mg/m 2 ; six patients received an intermediate dose, a single infusion of 375 mg/m 2 ; and fi ve patients received a high dose, four infusions of 375 mg/m 2 administered 1 week apart. Th ere was a signifi cant improvement in the disease activity, as measured by systemic lupus activity measure scores, in all patients by 2 months, which persisted at 12 months regardless of a change in anti-dsDNA anti body and complement levels. Six of 17 patients developed human anti-chimeric antibodies, resulting in reduced serum rituximab levels and ineffi cient B-cell depletion and less impressive effi cacy. Importantly, there were no signifi cant adverse events [26]. Th e UK-BIOGEAS registry study of 164 patients with refractory or relapsing lupus nephritis reported a 67% partial or complete response rate to rituxi mab using standardised response criteria [27].
Clinicians therefore continue to use rituximab for refrac tory lupus nephritis as well as nonrenal manifes tations including haematological, skin and central nervous system manifestations where clinically useful responses have been reported [28,29]. Th ere is thus extensive nonrandomised and retrospective experience of rituximab in the treatment of refractory SLE. A role for rituximab for this indication is supported by the consistency of the reports of improvement but diff erences in regimens, concomitant medications and endpoints remain, making it diffi cult to assess the extent of eff ectiveness of B-cell depletion accurately. Additionally, there is uncertainty as to how to reduce relapse risk after rituximab, and an unqualifi ed recommendation for rituximab in refractory SLE will require higher quality evidence.

Safety and effi cacy in clinical trials
To evaluate the safety and effi cacy of rituximab in SLE in a clinical trial setting, two double-blind, randomised, placebo-controlled trials (DBRCTs) investigating renal (LUNAR study) and nonrenal (EXPLORER study) manifes tations were undertaken ( Table 2). Both trials addressed the hypothesis that the addition of rituximab to the standard of care, corticosteroids and immuno suppres sants was superior to addition of placebo for the control of SLE activity.
In the EXPLORER study, the safety and effi cacy of rituximab in moderate-to-severe active nonrenal SLE was evaluated [30] (Figure 1). Th is study included 257 patients with ≥1 BILAG A score (>50% of patients at entry) or ≥2 BILAG B scores despite ongoing stable-dose immuno suppressant therapy with either azathioprine (100 to 250 mg/day), mycophenolate (1 to 4 g/day) or metho trexate (7.5 to 25 mg/week), which was continued during the trial. Background immunosuppressive therapy was evenly distributed. A key feature of treatment in this study was the additional course of high-dose corticosteroids patients received early in the study. Cortico steroids were given at initial doses of 0.5 mg/kg, 0.75 mg/kg or 1 mg/kg depending on severity (by BILAG score) and type of disease manifestations, followed by a taper regimen. Of the overall population, >50% were classed as steroid dependent, and ≥60% of patients received an average 45.9 ± 16.4 mg prednisolone and then attempted to reduce to a target dose of <10 mg/day over the 10-week taper period and ≤5 mg/day at week 52.
Patients were randomised at a ratio of 2:1 to receive rituximab (1,000 mg) or placebo. Eighty-eight patients received placebo and 169 patients received rituximab (two doses given 14 days apart) on days 1, 15, 168, and 182. Th e majority (≥50%) of patients in both groups had musculoskeletal and mucocutaneous disease.
Th e primary endpoint of the EXPLORER study was stringent, with complete and partial response defi nitions as follows.
To classify as a complete/major response, at week 24 an improvement in all organ systems with a BILAG C score or better was required. Further, this response was to be sustained at week 52, without experiencing a severe or moderate/severe fl are during the period to week 24 and week 52, respectively. A severe fl are was defi ned as a BILAG A score or as two new domains with BILAG B scores [31].
Patients were considered to have attained a partial response if: there was an improvement in all organ systems with a BILAG C score or better, which was sustained for 16 consecutive weeks; a BILAG B score in no more than one organ system at week 24 without a new BILAG A or BILAG B score to week 52 was achieved; and, at week 24, no more than two BILAG B scores were achieved without new BILAG A or BILAG B scores provided the baseline BILAG score was one A score plus ≥2 B scores, ≥2 A scores, or ≥4 B scores.
Th e secondary endpoints included the time-adjusted area under the curve minus the baseline BILAG score over 52 weeks, the proportion of patients who achieved a major and partial clinical response, the proportion of patients who achieved a BILAG C score in all organ systems at week 24, the time to the fi rst moderate to severe disease fl are, improvement in quality of life as measured by the Lupus Quality of Life, and the proportion of patients who achieved a major clinical response with a prednisolone dose <10 mg/day from week 24 to week 52. In addition, serological activity parameters including levels of autoantibodies, complement, immunoglobulins, T-cell and B-cell counts and human antichimeric antibody were monitored.
In the intent-to-treat analysis of 257 patients, approximately 70% of patients completed the study in both arms and the safety and tolerability was similar in both groups. Th ere was no diff erence between the addition of placebo and rituximab to the standard of care in the primary and secondary effi cacy endpoints, including the BILAGdefi ned response, in terms of both area under the curve and other analyses.
A preplanned subgroup analysis, however, detected a benefi cial eff ect of rituximab in the primary endpoint in the African American and Hispanic patients, a major clinical response in 13.8% and a partial response in 20% when compared with 9.4% and 6%, respectively. Notably, these patients had more active disease and more refractory disease as previously reported [32]. Th ere were signifi cant biological eff ects in the rituximab-treated group, with greater falls in anti-dsDNA levels and rises in complement levels compared with placebo. Interestingly, up to 9.5% of patients did not achieve complete B-cell depletion, but analysis without these patients did not change the primary outcome. Th is phenomenon has been observed in autoimmune prone mice [33,34]. A recent study investigating the role of highly sensitive fl ow cytometry detected a correlation between clinical response and B-cell numbers [35].
Th e LUNAR study investigated the safety and effi cacy of 2 × 1,000 mg rituximab, at both 0 and 6 months, as compared with placebo in addition to background therapy with high-dose glucocorticoids and mycophenolate mofetil 3 g/day in 144 patients with proliferative lupus nephritis, classes III and IV ( Figure 2).
Th e primary endpoint of the study was the proportion of patients with a complete or partial remission of nephritis at 12 months. Complete response was defi ned as, at week 52: serum creatinine improving from abnormal to normal level or from normal to ≤115% of baseline normal; a fall in the urine protein-creatinine ratio to <0.5; and urine sediment containing <5 red blood cells in a high-power fi eld without casts. Patients who did not meet complete response were considered to have achieved a partial response if: serum creatinine reduced to ≤115% of abnormal baseline; the number of red blood cells/high-power fi eld reduced to ≤50% baseline without red blood cell casts; and a reduction in urine proteincreatinine ratio from ≥3.0 to ≤3.0 or to <1 from ≤3.0.
Th e secondary endpoints were: complete renal response sustained from week 24 to week 52; time to fi rst complete renal response; and, at week 52, the urine protein-creatinine ratio improving from >3 to <1, the time-adjusted area under the curve minus the BILAG global score, and a change in the physical function of SF-36 health survey. As in the EXPLORER study, serological  indices, human anti-chimeric antibodies and B-cell depletion were monitored.
Th e response rates for rituximab and placebo were 26% and 30% for complete renal response and 30% and 15% for partial renal response, respectively. At week 52, more patients in the placebo arm (8 patients vs. 0 patients) received rescue cyclophosphamide therapy. Improvement in proteinuria was 32% and 9% for rituximab and placebo, respectively. Analogous to the fi ndings in the EXPLORER study and the ALMS trial, a greater proportion of black patients responded favourably, although this was not statistically signifi cant. Th ere was a greater reduction in anti-dsDNA levels in the rituximab-treated group. Whether the response noted in patients of African ancestry is attributable to the disease severity alone or whether there are potential diff erences in B-cell responsiveness to rituximab therapy in these patients is as yet unclear. In this respect, it is worth noting that ethnicity might infl uence the clinical response to treatment even with conventional immuno suppressants as noted in the ALMS study. Our own data (D Isenberg, unpublished observations) has not indicated a clearly diff erent outcome at 12 months post BCDT comparing Caucasians, Afro-Caribbean or Asian patients. Drawing any fi rm conclusions based on the disease severity alone would therefore be diffi cult.
However, overall this was a negative study in that there was no signifi cant diff erence between the rituximab group and the placebo group. Th e absolute diff erence in response was 11%, with 54% and 43% responding in the rituximab and placebo groups, respectively [36]. Th is value was less than the planned 23% diff erence, which in retrospect looks over-optimistic especially considering the analysis at only 12 months in this population. Again, diff erences in serological markers between groups were found and a subsequent analysis found greater falls of proteinuria in the rituximab group. More African patients in the rituximab group responded and cyclophosphamide rescue was required more frequently in the placebo group. Th erefore, despite some clear signals of effi cacy and safety, this study did not meet its primary or secondary endpoints.
Why did these two DBRCTs fail to meet their endpoints? As discussed earlier, there are several confounding factors that may have masked the ability to accurately quantify any signifi cant clinically meaningful benefi cial eff ects of rituximab (Table 3), perhaps the most important being the aggressive background immuno suppressive therapy in the placebo and rituximab-treated groups. High-dose corticosteroids, in particular, may have prevented the full extent of effi cacy of rituximab becom ing evident, a factor that warrants due consider ation in the design of future clinical trials for any investigational agent. Th e dilemma for trial designers is how rapidly to reduce glucocorticoid in patients with organ-threatening SLE. Trials with duration beyond 12 months would have greater chance of demonstrating the specifi c treatment eff ect that could be attributed to rituximab if corticosteroids are reduced to low levels during the fi rst 6 months. Corticosteroid dosing could also be included in the threshold for res ponse in trial endpoints. For example, standard treatment should allow low-dose prednisolone and the proportion of patients requiring >7.5 mg/day prednisolone could be classed as a failure. In the open studies, response was defi ned with such stringent criteria. Furthermore, apply ing such criteria would not detect organ-specifi c improve ment; for example, a signifi cant sustained improvement in a severe haematological abnormality but concurrent minor or moderate fl are in skin or muco cutaneous disease would be classifi ed as a failure.
Th e planned effi cacy margin in the LUNAR study was infl uenced by the 55% complete and partial response rate in the ALMS trial at 6 months using either myco phenolate or cyclophosphamide and corticosteroids. Th is suggested that 45% did not respond to standard of care; however, reasons for failure in the ALMS trial included death, severe adverse events, drug intolerance and patient/physician preference. One can estimate that true treatment failure was closer to 25% than 45%. A further factor in nephritis trials is the delayed response of the outcome measure, proteinuria, to reduction in histological activity in the kidney. Th e true time to remission of proteinuria is up to 2 years. Had the LUNAR trial aimed for a 12% effi cacy diff erence and involved a 2-year duration, the study may have met its endpoint despite a small sample size.
One should also note that to date there is insuffi cient evidence to support the routine use of rituximab therapy for patients with specifi c neuropsychiatric manifes tations. However, in a study of 10 patients with a range of neuropsychiatric manifestations (including cognitive dysfunc tion, psychosis and seizures) refractory to conventional immunosuppressants, including intravenous cyclophos phamide, there was a signifi cant improvement, measured by the Systemic Lupus Erythematosus Disease Activity Index score at 28 days after treatment with rituximab, in all patients -and in fi ve patients the response lasted for more than 1 year [28].
Th e other anti-CD20 mAb investigated in clinical trials for SLE is ocrelizumab (a humanised anti-CD20 mAb). In rheumatoid arthritis, ocrelizumab (two regimens used: 200 mg and 500 mg ×2 every 6 months) was eff ective in reducing signs and symptoms and joint damage when added to a stable dose of methotrexate [37,38]. However, a detailed analysis of results from four DBRCTs investigat ing the safety and effi cacy of ocrelizumab for RA indicated that an increase in serious infections associated with ocrelizumab compared with placebo were dose depen dent and occurred more frequently in Asia (particu larly Japan) [39].
Two simultaneous clinical trials were initiated to study the safety and effi cacy in lupus. Ocrelizumab was dosed diff erently from the RA and the rituximab SLE studies, at either 400 or 1,000 mg intravenously ×2 at entry with repeat, single dosing every 4 months. Th is regimen was designed to induce and maintain B-cell depletion throughout the trial periods. Th e BEGIN study for nonrenal SLE was can celled early. Th e BELONG study for proliferative lupus nephritis compared 1,000 mg or 400 mg ocrelizumab at 1 day and 15 days, then repeated with a single dose every 4 months on a background of high-dose glucocorticoids and either mycophenolate mofetil or cyclophosphamide dosed according to the EUROLUPUS protocol (Figure 1). Although the study was designed to continue for to at least 2 years, the primary endpoint was the proportion of patients achieving partial or complete nephritis remission at 48 weeks. A total of 381 patients were recruited before the trial was stopped early due to an imbalance in the rate of serious infections in the ocrelizumab patients receiving mycophenolate. Th e 221 patients who had passed the 32-week treatment point were assessed. Th e absolute diff erence in renal response was 12%, with 63% and 51% for the combined ocrelizumab and placebo groups prospectively. However, it is worth noting that in the subgroup analysis there was a greater treatment eff ect of ocrelizumab when combined with the EUROLUPUS cyclophosphamide regime (renal response of 65.7% for ocrelizumab vs. 42.9% for EUROLUPUS alone) than with mycophenolate mofetil (renal response of 67.9% for ocrelizumab vs. 61.7% for mycophenolate alone), which was largely ex plained by a higher response rate in general with myco phenolate mofetil whilst perhaps again refl ecting the outcome seen with rituximab in the LUNAR study [40] (Table 4). Effi cacy of the BCDT has also been demonstrated in another autoimmune condition, relapsing-remitting multiple sclerosis. A recent phase II randomised clinical trial investigating the safety and effi cacy of ocrelizumab (given together with pre-infusion steroids only) in multiple sclerosis showed a signifi cant reduction in neurological lesions compared with placebo as assessed by gadolinium-enhanced magnetic resonance images. Serious adverse events occurred in three of 55 patients receiving 2,000 mg ocrelizumab (one of 55 patients receiving 600 mg ocrelizumab, and two of 54 patients each in the placebo group and the IFNβ-1a group) [41]. Th ese results also support the notion that treatment regimens of BCDT continue to have the potential to be safe in the wider context of treatment for chronic refractory auto immune diseases.
Although not the principal focus of this review, it is notable in two trials involving >800 patients in each trial that belimumab (Benlysta), an anti-BLyS antibody, met its primary endpoint with a 10% and 14% absolute response diff erence over placebo [42,43]. Th e primary

Lessons learned so far and future clinical trial design -how to get it right?
Th e failure of clinical trials in SLE has introduced palpable uncertainty whilst providing some invaluable lessons regarding expectations for potential new thera pies, carefully planned trial designs and appropriate endpoints for the particular agent/regimen in question. It is relevant to note that most preliminary data used rituximab for refractory SLE when standard agents had failed. Th is is in contrast to the randomised trials, which added rituximab on top of standard therapy for non refractory patients. Several factors specifi c to SLE increase the complexity in designing successful trials. RA is a less heterogeneous disease and is much better under stood when compared with SLE and when arthritis is the main manifestation, despite the potential for other organs to be involved. Moreover, there exists a good deal more standardisation for clinical trials including validated endpoints -for example, Disease Activity Index, 28-joint Disease Activity Score. Conducting large-scale studies in a relatively short period of time is therefore possible -particularly as RA is more common and patient access is better, making statistically powered studies of relatively short duration feasible. For lupus, including nephritis, we are still some distance from achieving the same level of understanding and standardisation in the clinical trial setting.
In an attempt to improve the lupus patient's great unmet need, the European League Against Rheumatism has made a few suggestions to help researchers design successful trials [44]. Th e main points for the future design of clinical trials are to use strictly evaluated (a surrogate of therapeutic success against mortality or endorgan failure) outcome measures, including the disease activity indices, and to follow a standardised approach towards recording adverse events that could be used to measure benefi t-to-risk ratios from interventions, comparable between trials. Increasingly important in future trials, when comparing the interventional drugs, is the real diff erence there may be in their potential to cause harm in the long term.
Th e aims of randomised controlled trials are to be defi ned to test robust hypotheses generated based on the available evidence from the open studies and clinical experience. Further, careful attention needs be paid when considering important factors, patient selection and sample size, the therapeutic agent or regimen and its potential eff ectiveness (and meaningful treatment delta vs. control), the disease outcome measures and disease activity indices, adequate follow-up and the adverse events (Tables 5 and 6). Th ese variable factors contribute to a great element of uncertainty in predicting the probability of the success of clinical trial design in SLE. Firstly, the patient cohort in open studies and in clinic experience, at the time of rituximab treatment, had moderate-to-severe disease activity and most had failed conventional immunosuppressants (standard of care). In contrast, patients participating in the two DBRCTs (EXPLORER and LUNAR studies) had active disease, but patients who had failed conventional therapy (cyclo phosphamide and calcineurin inhibitors) were excluded. Further, patients with central nervous system manifes tations and severe organ-threatening conditions were excluded -situations in which rituximab has demonstrated a favourable record in the open studies [28,[45][46][47]. Capturing the variability in organ-specifi c outcomes for diff erent interventions tested is important. For example, rituximab may be a better choice than other conventional immunosuppressants when both renal and haematolo gical abnormalities co-exist. A favourable clinical response is more likely in seropositive patients. However, we have previously noted that anti-Sm positivity and/or a low C3 level at the time of treatment is associated with a reduction in the likelihood of sustained benefi t from Bcell depletion, and again suggest there is much work to be done to understand lupus disease and factors that may infl uence the design, population and, ultimately, the outcome of clinical trials [48].

The therapeutic agent and the regimen
Rituximab has been mainly been used to achieve B-cell depletion in two regimens, either as two doses of 1,000 mg given 2 weeks apart (two-dose regime, commonly used in SLE and RA) or as four doses of 375 mg/m 2 (fourdose regime, most common regime used in lymphoma, paediatric autoimmune diseases) given 1 week apart (ocrelizumab in SLE moved on from this to initial doses 2 weeks apart followed by a single infusion every 4 months to achieve and sustain B-cell depletion). Notably, a systematic review of the clinical experience of rituximab for the treatment of refractory SLE suggests that the lymphoma regimen (four doses, 375 mg/m 2 , given 1 week apart) may be more eff ective in achieving an improvement in disease than the two-dose regimen (two doses given 2 weeks apart) [49]. Based on this review alone, however, it is diffi cult to draw fi rm conclusions about the relative effi cacy of either regimen. Catapano and colleagues, using both regimens of rituximab for the treatment of refractory SLE, although not in a formal comparative setting, did not detect a signifi cant diff erence in either the degree of B-cell depletion or clinical outcomes [19]. Th e two-dose regimen, more convenient for patients requiring just the two hospital infusion visits, is therefore preferred.
Defi ning standard treatment used in the comparative arm is important, because not doing so would allow generous use of other immunosuppressants -particularly corticosteroids, which are highly eff ective but associated with unacceptable adverse eff ects in the long term, not necessarily identifi ed in clinical trials with short-term follow-up.
It would be interesting to take a treatment-to-target approach to achieve an adequate degree of B-cell depletion and clinical response. For example, evidence suggests that the effi cacy depends on the extent of B-cell depletion in RA [50]. Several research groups have noted that the degree of B-cell depletion is variable in SLE and that early repopulation is common in patients with a poor response to rituximab [35]. Th e underlying reasons for the variability in B-cell depletion remain elusive. A polymorphism in Fcγ receptor IIIa has been shown to be important in achieving an adequate degree of B-cell depletion, in favour of the high-affi nity genotypes Fcγ receptor IIIa V158F (V, valine; F, phenylalanine) [51]. Treatment-to-target would therefore seem a rational approach to take in an attempt to improve the major clinical response. However, some patients will probably require more frequent doses than others. One approach could be to counterbalance this variation using alterna tive dose regimes; for example, using two 500 mg doses given 2 weeks apart, as in a recent trial in RA that reported equal effi cacy, safety and tolerability between the two regimes using 500 mg or 1 g, provided adequate depletion was achieved [50,52]. Diff erent dosing regi mens could poten tially have considerable implications: fi rst, patient con venience, with a four-dose regimen requiring more hospital visits; second, a very-low dose regimen has been associated with the development of anti-drug antibodies in SLE while a medium dose (500 mg rituximab ×2) has been shown to be adequate in a number of patients with RA [50]; and, fi nally, cost-eff ectiveness of BCDT. In this respect, it has been noted that rituximab might be rapidly consumed in some patients, more frequently in SLE than RA [53]. Th is consumption would consequently reduce serum rituxi mab levels and may reduce clinical effi cacy.
Taking experience from ocrelizumab therapy in lupus, careful consideration is also necessary when designing studies to test the safety and effi cacy of B-cell-targeted approaches, including depletion in patients with active disease also taking mycophenolate. A combination of ocrelizumab and recently commenced mycophenolate does not appear to result in a meaningful additive response and results in an increased risk of infection adverse events (whether the combined impact on the B-cell population of anti-CD20 and mycophenolate was a contributory factor is not understood), whereas this was not the case when used in combination with the EUROLUPUS cyclophosphamide followed by azathio prine regimen. Defi ning the standard of care in the placebo arm is important to allow detection of the effi cacy for the intervention tested. For example, in the placebo arm a patient with disease activity requiring >7.5 mg prednisolone being classed as a failure will allow detecting the steroid-sparing eff ect of the intervention, a major advantage in the long term. Th e question has been raised as to whether to use rituximab in combination with cyclophosphamide, aza thio prine or mycophenolate, but there are some confl ict ing data [19,54]. Th e defi nitive answer is therefore awaited.
Another conundrum not yet fully resolved is whether there really is added benefi t in using repeated rituximab infusions on a regular basis (that is, maintenance therapy) or whether it is preferable to repeat B-cell depletion only when the patients relapse. A concern about repeated infusion is the potential occurrence of hypogammaglobulinaemia. Information from studies in patients with RA (J Edwards, personal communication) suggests that many patients begin to drop their IgG levels after annual rituximab infusions, particularly in patients with low baseline IgG levels [55]. Comparative data for SLE patients are awaited.
Clinical evidence for rituximab use -early disease or chronic refractory disease?
Limited evidence from two studies is worth considering. Firstly, as discussed, when used early in conventional immunosuppressive naïve disease, rituximab seems to be • Defi ning the adverse events The reporting of adverse events could be standardised adhering to the OMERACT-recommended guidance [63] eff ective and has a steroid-sparing eff ect [24]. Further, Pepper and colleagues have prospectively analysed the response to rituximab for biopsy-proven lupus nephritis, where a total of 14/18 (78%) patients achieved a complete or partial remission with a sustained response in 12/18 at 1 year (67%), with two patients having a relapse with an increase in proteinuria. Th ere was a reduction in predniso lone usage from a mean of 10 mg to 5 mg at 2 years, six patients stopped, six patients managed to reduce the dose and the remaining were maintained on the same dose. Five patients required a temporary increase for extra-renal manifestations [56].

Defi ning the outcome measures and clinical response
Clinical outcome measures are to be defi ned based on evidence, taking into account the probability of detecting change given the expected natural progress of the organspecifi c disease manifestations in an appropriate timeframe (potentially in contrast to the artifi cial time points used in clinical trials). In parallel, it is important to include the biomarkers that predict disease activity and outcomes in SLE. For example, there are a few validated outcome measures that predict end-stage renal disease; it has been shown that doubling of serum creatinine [57,58] and persistently elevated serum creatinine at 48 weeks [58] is predictive of end-stage renal disease. Another routinely available biomarker in clinical practice is urinary protein and an improvement in proteinuria at 1 year [59] and a decrease in serum creatinine or proteinuria at 6 months [60], whilst it may also be reasonably expected that renal response may continue to improve beyond the fi rst year of treatment and may be relevant to consider when identifying the maximal treatment diff erence for a clinical trial. However, there is limited evidence of reliable predictors of long-term outcome for nonrenal SLE. For reasons discussed earlier, steroid-sparing eff ect is an important factor when deciding the immuno sup pres sant of choice [56].

What disease assessment index to use?
Disease activity indices have been developed with a view to assess either disease activity or damage. Th e proposed SLE Responder Index, although used in the belimumab studies [61,62], has never been validated or shown to be reliable or sensitive to change or appropriate for wide use when evaluating effi cacy with other investigational agents. Th e key problem with global score indices is that they do not capture partial improvement and/or deterioration. Th e defi nitions of treatment failure and fl are remain variable between studies, which limit direct comparison of effi cacy of diff erent therapeutic agents. To facilitate a better comparison between studies, therefore, it is important to standardise the defi nition of a fl are and treatment failure.

Adequate follow-up period to detect signifi cant change in the disease activity and disease damage
Allowing an adequate follow-up period to detect clinically meaningful eff ects is very important. For example, haematological abnormalities such as anaemia and autoimmune thrombocytopaenia and skin changes such as vasculitic rash improve rapidly; in contrast, response in nephritis may take much longer to detect. Other important factors such as the eff ects of long-term accruement of organ damage and drug-related adverse eff ects could only be detected after many years.

Defi ning the adverse eff ects
Adverse events recorded in the clinical trials in SLE have not been adequately standardised to allow comparison between trials. In chronic disease such as SLE where a number of treatments have proved to have modest effi cacy, adverse eff ects associated with treatment have a signifi cant infl uence on the choice of treatment. As discussed, achieving primary and secondary endpoints of effi cacy at the expense of unacceptable adverse events has proven unfruitful in the case of the anti-CD20 (ocrelizumab) in RA [39] whilst the BELONG lupus nephritis trial was stopped early due to an imbalance of infectious adverse events. Th is fi nding does raise the question of whether the screening and monitoring criteria can be applied more stringently for the detection of risk or actual opportunistic infections prior to inclusion in the study, particularly when recruiting patients residing in areas endemic for opportunistic infections as mycobacteria or hepatitis. Also, another important question remaining unanswered is whether the adverse eff ects of biological agents are infl uenced by other identifi able factors such as disease history and treatment as well as a patient's immunology or indeed ethnicity. A robust defi nition of categories of adverse events therefore needs to be tested in clinical trials to understand and compare the safety of interventions in clinical trials. For example, is mycophenolate safe to use following rituximab induction therapy? Does the dose of mycophenolate need to be modifi ed to a low-dose regime or should an alternative less potent immunosuppressant such as azathioprine be used? Further, the dose of drug may be better adjusted based on patient characteristics; for example, a dose defi ned by the weight of the patient rather than a predefi ned dose (that is, 2 to 3 g). Th is factor is especially important when considering the use of mycophenolate in patients with low body mass index; for these patients, even 2 g may be a relatively high dose, especially when used in the maintenance regime following rituximab induction therapy. Th e recording of adverse events in clinical trials and open studies could be standardised adhering to rheumatology-specifi c criteria such as the OMERACT [63].

Key messages
• B-cell depletion with rituximab continues to be used in clinical practice for the treatment of refractory SLE, on the basis of a considerable number of publications describing the safety and effi cacy data from small open studies and clinical experience whilst noting that it has not been approved by health authorities for the treatment of lupus. • Contributing features that may have led to the failure of DBRCTs with anti-CD20-mediated B-cell depletion or at least identifying any true treatment eff ect size probably include concomitant use of high-dose steroids, stringent and nonorgan-specifi c clinical res ponse criteria, too short a follow-up, and, from a statistical perspective, the sample size. However, the trials confi rm the safety of repeated treatment with rituximab. • A better response to rituximab detected in patients of African-American and Hispanic ancestry highlights the importance of preplanned subgroup analysis and the need to better understand the potential disease drivers of a treatment eff ect when compared with a standard-of-care regimen in a trial setting. • Th e signifi cant biological eff ects seen with rituximab need to be monitored to assess clinical benefi t and risk in the long term. • Future clinical trial design in SLE and lupus nephritis may be guided by the key working groups of experts, including the European League Against Rheumatism task force, in order to achieve standardisation and to continually apply lessons from both clinical and trial experience.