Response to placebo in non-renal, non-neuropsychiatric systemic lupus erythematosus: a systematic review and pooled analysis

Abstract

Objectives

Most randomized controlled trials (RCTs) in SLE have failed to reach their respective end points, with the rates of response to placebo (plus standard-of-care treatment) being unexpectedly high. The aim of this systematic review was to quantify the response to placebo in non-renal, non-neuropsychiatric lupus.

Methods

The PubMed database was searched (from 2000 to December 2019) for phase II/III RCTs assessing the efficacy and safety of biologics in non-renal, non-neuropsychiatric SLE. Data on the efficacy and safety of the placebo-treated patients were collected in a pre-established data retrieval form. Descriptive statistics were used.

Results

A total of 24 RCTs (n = 11128 in total) were included. Placebo-treated patients (n = 3899) were mostly females (93.5%), Caucasians (60.2%), of mean age 39.7 years, and having a mean disease duration of 7.4 years. Their mean initial SLEDAI 2000 was 10.4, whereas 60.5% had positive anti-dsDNA antibodies, 41.9% low C3 and 35.6% low C4 at randomization. Standard-of-care treatment included glucocorticosteroids in 85.9%, antimalarials in 72.8% and immunosuppressives in 48.5%. The response to placebo was 36.2% for the primary end point (as defined in each study), 39.8% for the SLE Responder Index-4 (SRI-4), 29.2% for SRI-5, 28.4% for SRI-6 and 30.9% for BILAG-based Combined Lupus Assessment response. Regarding safety, there were serious adverse events in 16.3% of patients, serious infections in 5.5% and malignancies in 0.3%, and death occurred in 0.56% of patients.

Conclusion

More than one-third of the placebo-treated patients achieved their respective primary end points in RCTs with biologics in non-renal, non-neuropsychiatric SLE. The response rate was higher for certain end points, such as the SRI-4, while it decreased with more stringent end points.

SLE, randomized controlled trials, placebo

Rheumatology key messages

• The rates of response to placebo (plus standard treatment) in lupus randomized controlled trials (RCTs) are unexpectedly high.

• The pooled (24 RCTs) response to placebo in non-renal, non-neuropsychiatric SLE was 36.2% (primary end points).

• This high bar calls for changes to the design of future SLE trials.

Introduction

Most randomized controlled trials (RCTs) with biologics in SLE have failed to achieve their primary end point, in striking contrast to the revolution witnessed in other systemic autoimmune diseases, such as RA and seronegative spondylarthritides [1–3]. The reasons behind this failure remain elusive, despite the progress made in the understanding of disease pathogenesis and the development of more targeted therapies. Most investigators agree that SLE is unlikely to arise from a single pathogenetic mechanism implicating a specific cytokine or cellular subpopulation, or even a distinct autoimmune pathway [4]. Hence, CNS involvement may be vastly different from kidney disease or cutaneous inflammation. Even within the context of a single organ involvement (e.g. kidney), existing evidence suggests the presence of significant differences in terms not only of the pathological findings but also of the response to treatment and long-term prognosis [5]. Similar conclusions can be drawn from the heterogeneity of the CNS involvement (with 19 possible distinct clinical syndromes identified from the ACR in 1999 and many more added over time) [6] and of the skin [7]. Therefore, it seems unrealistic that every new biologic drug will be effective against this diversity of pathogenetic mechanisms and clinical manifestations.

Aside from disease heterogeneity and trial design factors, Touma and Gladman [8] recently suggested that the standard-of-care (SoC) treatment leads to there being a ‘very high bar to reach’ in order for a given biologic to demonstrate superiority against placebo. This bar varies significantly for different end points, thus creating confusion in the assessment of the response to placebo and the design of subsequent trials. The aim of this systematic review was to quantify the response to placebo for different end points in the RCTs with biologics and small molecules in non-renal, non-neuropsychiatric lupus patients.

Methods

This systematic review was conducted according to the 2009 PRISMA statement [9]. The PubMed database was searched with the Medical Subject Heading (MeSH) term ‘systemic lupus erythematosus’. The applied filters included the type of article [RCTs], publication date [from 1 January 2000 to 31 December 2019], species [humans] and language of the text [English]. Two independent reviewers (K.T. and L.W.) performed the literature search and screened all retrieved articles by title and/or abstract. Phase II/III RCTs that assessed the efficacy and safety of biologic drugs in human non-renal, non-neuropsychiatric SLE were identified. Phase I trials, exploratory studies for the pharmacokinetics of each drug and post hoc analyses as well as RCTs on exclusive populations (e.g. Japanese patients only) and those with non-biologic interventions were excluded. In the case of disagreement between the reviewers, a third author (D.D.G. or M.B.U.) acted as referee.

Data on the efficacy (primary and secondary end points, mainly composite) and safety (serious adverse events, serious infections, malignancies and deaths) of the placebo-treated patients were collected in a pre-established data retrieval form. Additional data collected were the demographics of the placebo-treated patients (age, sex, race/ethnicity), the baseline disease characteristics (disease duration, disease activity, cumulative damage, organ/system affected), the serologic activity (anti-dsDNA, C3, C4) as well as the concomitant SLE-specific medications (glucocorticosteroids and mean daily prednisone dose, antimalarials, immunosuppressives) that comprised the SoC treatment. Data on end points that had been analysed in three or more RCTs are reported.

Statistical analysis was conducted with the SAS 9.4 software using descriptive statistics.

Results

A total of 350 articles were initially identified and screened by title and/or abstract. Of the 350, 36 articles were selected and assessed for eligibility based on the full text; of these, 22 were deemed eligible for this analysis. These articles reported on 24 RCTs, 15 assessing the efficacy and safety of different biologics at 48 or 52 weeks [10–23, cited in chronological order of publication] and 9 assessing these outcomes at 12 or 24 weeks [24–31, cited in chronological order of publication]. The RCT of anti-CD-154 (CD40L) [32] and the phase II study of belimumab [33] were excluded because they only reported on the numerical reduction of the SLEDAI 2000 (SLEDAI-2K) or the Safety of Estrogen in Lupus National Assessment (SELENA)-SLEDAI and not on composite end points.

Analysis of RCTs with outcomes assessed at 48 or 52 weeks

There were 13 RCTs that reported outcomes at 48 or 52 weeks (Supplementary Table S1, available at Rheumatology online) and enrolled cumulatively 6017 lupus patients in the active arms and 3184 in the placebo arms. The demographic, clinical and therapeutic characteristics of the placebo (plus SoC) -treated patients are shown in Table 1. The response rates (and range) to placebo (plus SoC therapy) for different end points are shown in Table 2. The expected difference between the active drug and the placebo was 15.9% (average in 12 studies, range 12–24%), with a predicted power of 84–99%. The actual difference was 7.2% on average (range –4.2 to 17.9%). Details on the SoC treatment and primary end points for each RCT are provided in Supplementary Table S3, available at Rheumatology online.

Table 1

Characteristics of the placebo-treated patients in the included RCTs

Variable	RCTs reporting outcomes at 48 or 52 weeks (n = 3184)	RCTs reporting outcomes at 12 or 24 weeks (n = 715)
Females, n (%)	2988 (93.8)	561/610 (92)
Caucasians, n (%)	1997 (62.7)	286/610 (46.9)
Mean age at enrolment, mean (s.d.), years	39.6 (2.1)	40.3 (2.7)
Mean disease duration, mean (s.d.), years	7.5 (2)	6.7 (1.6)
Mean SLEDAIa, mean (s.d.)	10.4 (1.2)	10.3 (1.2)
Musculoskeletal manifestation, n (%)	1870/2394 (78.1)	207/267 (77.5)
Mucocutaneous manifestations, n (%)	1760/2394 (73.5)	200/267 (74.9)
Cytopenias, n (%)	199/1685 (11.8)	18/267 (6.7)
Renal manifestations, n (%)	172/1882 (9.1)	20/267 (7.5)
Serositis, n (%)	114/1685 (6.8)	18/267 (6.7)
Vasculitis, n (%)	88/1410 (6.2)	10/267 (3.7)
CNS manifestations, n (%)	30/1437 (2.1)	5/267 (1.9)
Anti-dsDNA positive, n (%)	1278/2096 (61)	313/535 (58.5)
Low C3, n (%)	946/2223 (42.6)	89/246 (36.2)
Low C4, n (%)	811/2223 (36.5)	68/246 (27.6)
Glucocorticosteroids, n (%)	2432/2830 (85.9)	526/615 (85.5)
Mean daily prednisone dose, mean ± s.d., mg	11.5 (3.1)	9.7 (4.2)
Daily prednisone dose ≥7.5 mg, n (%)	1062/1917 (55.4)	327/635 (51.5)
Antimalarials, n (%)	2008/2742 (73.2)	482/678 (71.1)
Immunosuppressives, n (%)	1329/2742 (48.5)	349/715 (48.8)
AZA	590/2560 (23)	55/326 (16.9)
MTX	422/2560 (16.5)	49/326 (15)
MMF	314/2560 (12.3)	38/326 (11.7)

^aSLEDAI-2K or SELENA-SLEDAI.

RCT: randomized controlled trial.

Table 2

Response rates at 48–52 weeks for placebo-treated patients

Outcome	Patients (responders/all)	Response rate (%)	Range of response (%)
Primary end pointa	1154/3184	36.2	17.5–48.6
SRI-4	984/2503	39.3	32.3–48.4
SRI-5	400/1424	28.1	20.4–39.3
SRI-6	389/1424	27.3	18.9–37.4
BICLA	344/1088	31.6	25.7–36.1
Patients reducing prednisone to ≤7.5 mg/day	141/919	15.3	12–26.6
Normalization of anti-dsDNA antibodies	83/1055	7.9	5.7–11.9
Normalization of C3	126/616	20.5	14–22.4
Normalization of C4	59/347	17	7.7–21.9

^aAs defined by each study

SRI-4: SLE Responder Index 4 (reduction of ≥4 in SELENA-SLEDAI or SLEDAI-2K, no worsening in PGA, no new A or two new B scores in the British Isles Lupus Assessment Group, BILAG)

SRI-5: as above with ≥5 reduction in SELENA-SLEDAI or SLEDAI-2K

SRI-6: as above with ≥6 reduction in SELENA-SLEDAI or SLEDAI-2K

BICLA: (BILAG-based Combined Lupus Assessment) improvement from baseline in the BILAG-2004 score, no worsening in SLEDAI-2K and PGA, no unpermitted changes in concomitant medications.

Analysis of RCTs with outcomes assessed at 12 or 24 weeks

Nine RCTs reported outcomes at 12 or 24 weeks (Supplementary Table S2, available at Rheumatology online) for 1212 lupus patients in the active arms and 715 in the placebo arms. The demographic, clinical and therapeutic characteristics of the placebo (plus SoC) -treated patients are shown in Table 1. The response rates (and range) to placebo (plus SoC therapy) for different end points are shown in Table 3. The expected difference between the active drug and the placebo was 20% (average in six studies, range 15–25%) for a predicted power of 80–90.5%. The actual difference was 9.9% on average (range 1.9–29%). Details on the SoC treatment and primary end points for each RCT are provided in Supplementary Table S3, available at Rheumatology online.

Table 3

Response rates at 12–24 weeks for placebo-treated patients

Outcome	Patients (responders/all)	Response rate (%)	Range of response (%)
Primary end pointa	266/715	37.2	21.1–44.8
SRI-4	160/371	43.1	33–48
SRI-5	129/390	33.1	21–35.3
SRI-6	155/490	31.6	19–34.6
BICLA	42/162	25.9	21.1–33

^aAs defined by each study

SRI-4: SLE Responder Index 4 (reduction of ≥4 in SELENA-SLEDAI or SLEDAI-2K, no worsening in PGA, no new A or two new B scores in the British Isles Lupus Assessment Group, BILAG)

SRI-5: as above with ≥5 reduction in SELENA-SLEDAI or SLEDAI-2K

SRI-6: as above with ≥6 reduction in SELENA-SLEDAI or SLEDAI-2K

BICLA: (BILAG-based Combined Lupus Assessment) improvement from baseline in the BILAG-2004 score, no worsening in SLEDAI-2K and PGA, no unpermitted changes in concomitant medications.

Safety analysis

In all 3899 placebo-treated patients who were enrolled in 24 RCTs, there were 635/3899 (16.3%) serious adverse events, 163/2980 (5.5%) serious infections, 7/2391 (0.3%) malignancies and 22/3899 (0.56%) deaths. Side effects were comparable for patients in the active arms [serious adverse events in 1100/7229 (15.2%), serious infections in 272/5497 (4.9%), malignancies in 14/3620 (0.4%) and deaths in 49/7229 (0.68%)].

Discussion

In this pooled analysis of RCTs conducted in non-renal, non-neuropsychiatric SLE, the response to placebo (plus SoC treatment) exceeded 36% for all primary end points combined. That was particularly high for the most commonly used end point [SLE Responder Index-4 (SRI-4)], reaching or even surpassing 40% for trials reporting outcomes at 48–52 weeks and 12–24 weeks, respectively. These figures could have been even higher considering the relatively low cumulative rate of immunosuppressive (48.5%) and antimalarial use (72.8%). The actual difference between the active arms and placebo was only 7.2% and 9.9%, respectively, whereas the expected difference (as reported in each study) ranged between 12 and 25%.

The considerable response to placebo in the SLE trials contrasts with the commonly observed rates in RA: a recent meta-analysis of 32 RCTs reported a median ACR20 response to placebo of ∼35% (9–46%), an ACR50 of 15% (3–29%) and an ACR70 of 5% (0–16%) [34]. The authors found that the response to placebo was significantly increased over time (e.g. for example the ACR20 response was increased from ∼24% in the trials that were published during 2000–2010 to 35% in the most recent trials). They attributed this to various causes, including improved adherence of patients enrolled in developing countries (where the study sponsor usually covers the expenses of the concomitant DMARDs), enrolment of patients in the early stages of the disease (when the chance of response is greater) as well as expectation bias. Poor adherence is an important factor for treatment failure in SLE, in which the percentage of non-adherent patients ranged from 43% to 75%, with most studies consistently reporting rates of >50% in a recent systematic review [35]. The difference between the placebo effect in RA and SLE trials may be attributed to the more comprehensive SoC treatment in lupus patients (combination of glucocorticosteroids, antimalarials and immunosuppressives) compared with RA, in which MTX is the usual comparator.

What are the reasons for the relatively small difference in the efficacy of the active drugs and the placebo in lupus trials? The answer is probably multifactorial and involves the efficacy of the SoC treatment, the nature of the disease itself, the selection of the outcomes and the study design, among others. The SoC treatment (presumably received by all patients in the included RCTs) achieved a 36.2% response in all primary end points, while the addition of a biologic led to a 7–10% (on average) improvement in response rates. The clinical heterogeneity of SLE reflects the differences in the underlying pathogenetic mechanisms, making it unlikely for a certain biologic to suppress the autoimmune inflammation in all cases. Even within the context of a single organ involvement (e.g. skin), different pathogenetic pathways may lead to the expression of distinct clinical phenotypes (acute, subacute or discoid lesions) [7, 36]. Detailed information on the type of skin disease is lacking for the patients enrolled in the lupus RCTs. Furthermore, topical treatments (particularly topical glucocorticosteroids and calcineurin inhibitors), that are not regularly monitored in RCTs, have been moderately beneficial for acLE and scLE [37]. In the same line, spontaneous remissions have been reported for cutaneous lupus; in a recent RCT comparing HCQ and placebo, >30% of the placebo-treated patients achieved improvement after 16 weeks [38].

On further consideration of the nature of the disease, it is apparent that most of the enrolled patients were in late phases of the disease (mean disease duration of 7.4 years at enrolment). This may have a significant impact on the fluctuations of disease activity, because it has been shown that disease severity subsides considerably over time [39]. Consequently, minor (musculoskeletal and mucocutaneous) flares may be more easily controlled with SoC treatment. Moreover, the timing of the randomization of each patient is not known. Conceivably, most of the enrolled patients have followed a relapsing–remitting or a persistently active disease course [40]. These patients might have entered a period of increased disease activity (flare) a few months before randomization and eventually responded in a short period of time after baseline (apparently due to the effect of the SoC therapy). Other patients who entered a flare shortly before randomization may have required more time to respond and thus needed sufficient time for the active drug to act. In addition, permitted changes in the SoC medications shortly before or after baseline might have impacted the outcomes. The tempo of response may raise another significant point for consideration. In this context, we have shown that treatment with MMF leads to resolution of the initial non-renal clinical manifestation in >50% of the patients after 6 months and is accompanied by a significant (∼33%) reduction in the concomitant prednisone dose [41]. Renal involvement may require up to 5 years for complete resolution [42]; such studies have not been performed for the articular and cutaneous manifestations of SLE, although the common belief is that such manifestations usually respond in a considerably shorter time.

The currently used end points constitute another major issue in lupus RCTs. The most commonly used outcome (SRI-4) includes a reduction of 4 points in SLEDAI-2K (or SELENA-SLEDAI) and no new activity in any other organ system and no deterioration in the Physician’s Global Assessment (PGA). Given that most of the enrolled patients have active musculoskeletal and/or mucocutaneous disease, an SRI-4 can only be achieved through the complete resolution of arthritis. In the case of mucocutaneous involvement, the responder should completely resolve both manifestations (inflammatory rash and alopecia for example) or one manifestation and one immunologic marker (normalization of the anti-dsDNA titres or the complement levels). Thus, this represents an ACR100 (for one or more disease features included in the SLEDAI-2K), a target that may be unrealistic. In this context, other end points such as the SLEDAI-2K Responder Index-50 (SLEDAI-2K RI-50), which considers the severity of the initial manifestation and requires a 50% improvement, may be more relevant [43]. In addition, a new activity index that controls for the concomitant use of glucocorticosteroids [44], SLEDAI-2KG may be applied to combine two distinct end points (SLEDAI-2K response and proportion of patients who reduced their prednisone dose) in a single outcome. Of note, the BILAG-based Combined Lupus Assessment (BICLA) response was the first to be utilized in a successful lupus RCT after the BLISS studies [23].

Furthermore, other end points that consider the factor of time may be more relevant in a chronic and, primarily, relapsing–remitting disease like SLE. Although such measures have not been validated, the time spent in a state of clinical remission (taking into account the lack of agreement on its definition) or even in a low disease activity state could be of value in the context of RCTs. This is further supported by recent studies that uniformly showed that clinical remission and low disease activity confer similar hard outcomes (such as damage accrual and mortality) over time [45]. In this context, Morand et al. [46] showed that attainment of lupus low disease activity state could discriminate responders and was associated with improved outcomes in a post-hoc analysis of the MUSE trial.

What strategies should be implemented to enhance the chance of success for future lupus trials? It seems reasonable that the enrolment of a homogeneous patient group should become a priority. This should apply not only to seemingly similar manifestations (e.g. mucocutaneous disease) but extend further to the categories of cutaneous lupus. Patient randomization should consider the timing of the current flare and that of the concomitant treatment. Finally, the application of different end points (such as SLEDAI-2K RI-50 or SLEDAI-2KG in the context of the composite end points BICLA or SRI-4 or lupus low disease activity state and time-related outcomes) may offer a clear differentiation between the active and the placebo arms. If none of the aforementioned study design factors are modified, then the sample sizes should be reconsidered. With the average difference between the active arm and the placebo arm being between 7 and 10% (the average of the concluded RCTs at 48–52 and 16–24 weeks, respectively), the sample size needed to obtain a statistically significant difference of 0.05 with a power of 90% would be 1036–2104 patients (for a 1:1 ratio between the active arm and placebo arm) [47].

The main limitations of the present study are based on the lack of access to the raw data from each study. Therefore, it is impossible to assess the characteristics of the placebo-treated patients who achieved their respective end point as opposed to the non-responders. It would be interesting to know the detailed clinical phenotype of such patients (mild arthritis e.g. 2–3 affected joints or a mild acLE) as well as the SoC therapy (if these were the patients treated with antimalarials and immunosuppressives or not). In addition, the outcomes that were used for the composite scores differed between studies. For example, some studies used the SLEDAI-2K for the SRI-4, while some others utilized the SELENA-SLEDAI version. However, the differences between these indices of disease activity are small; hence, the chances of this significantly impacting the outcome are slim. The SRI-4 based on SELENA-SLEDAI was used as a primary end point in three studies reporting outcomes at 52 weeks [12, 13, 20], and the response in the placebo-treated patients was 353/842 (41.9%). The SRI-4 based on the SLEDAI-2K was used as a primary end point in another three such trials [17, 19, 22], and the response rate in the placebo-treated patients was 164/394 (41.6%). Moreover, this study is not a formal meta-analysis, since the active comparator includes multiple drugs with different mechanisms of action and different pathogenetic targets. As such, measures of inconsistency, publication bias and critical appraisal of the included studies are not reported. However, this is a pooled analysis of all the published phase II/III RCTs in SLE that have assessed the most widely used composite end points and it may inform the design of future clinical trials.

Conclusion

In conclusion, the response to placebo (plus SoC treatment) in RCTs conducted in non-renal, non-neuropsychiatric SLE patients exceeds 36% for all primary end points combined, setting a very high bar for future trials. Multiple factors such as the nature of the disease, study design and SoC therapy are contributing to this phenomenon and should be considered in the development of new drugs for SLE.

Funding: No specific funding was received from any funding bodies in the public, commercial or not-for-profit sectors to carry out the work described in this manuscript.

Disclosure statement: The authors have declared no conflicts of interest.

Data availability statement

Data are available upon reasonable request by any qualified researchers who engage in rigorous, independent scientific research, and will be provided following review and approval of a research proposal and Statistical Analysis Plan (SAP) and execution of a Data Sharing Agreement (DSA). All data relevant to the study are included in the article.

Supplementary data

Supplementary data are available at Rheumatology online.

Acknowledgements

The Toronto Lupus Clinic Research Program is supported financially by the University Health Network, the Lou Rocca Family and the Lupus Foundation of Ontario.

All authors were involved in the study conception and design, acquisition of data, and analysis and interpretation of data. 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. M.U. 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.