A double-blind, randomized, placebo-controlled, phase II trial of baricitinib for systemic lupus erythematosus: how to optimize lupus trials to examine effects on cutaneous lupus erythematosus

Results of a phase II trial of baricitinib for systemic lupus were recently published in the Lancet.¹ We are delighted to see evidence of the effectiveness of a Janus kinase (JAK)1/2 inhibitor for lupus. Since this mechanism may have key relevance for cutaneous lupus erythematosus (CLE),² it is also important that the trial specifically examined skin at three end points. However, no evidence for specific efficacy in skin was observed.

A coprimary outcome examined complete resolution of rash, using the 2000 Systemic Lupus Erythematosus Disease Activity Index (SLEDAI-2K). A statistically significant increase in the proportion of patients with resolution of either arthritis or rash was seen but the data for proportion with resolution of rash alone were not given. It is mentioned in the Discussion that there was no difference in the proportion of patients ‘achieving the mucocutaneous organ domain score’ but how that was measured is not explained. In an exploratory end point, the Cutaneous Lupus Erythematosus Disease Area and Severity Index (CLASI) was evaluated, finding no improvement on the CLASI activity score with baricitinib.

Both the SLEDAI-2K and the CLASI are validated instruments, but they are quite different. By design, the SLEDAI-2K records improvement in rash only if there is complete clearance of lesions, rendering it a stringent, but insensitive, measurement for improvement. Stringent end points have lower overall response rates, which may have limited the power to detect a difference in what was, by systemic lupus erythematosus (SLE) standards, a relatively modest trial size. As most available therapies do not completely (or even almost completely) clear skin lesions in a given patient, the CLASI has an advantage over the SLEDAI by capturing improvements that may fall short of total resolution, but remain meaningful for the patient. The CLASI provides versatile options for evaluation, capable of detecting varying degrees of change, depending on cut-offs used to define improvement. In this analysis, the mean change from baseline was examined. This may be an extremely sensitive detector for change, but risks loss of discriminatory capacity if there are large numbers of clinically insignificant incremental changes in one group vs. another.

Unlike the SLEDAI, the CLASI has been fully validated as a cutaneous-specific end point for SLE and successfully used in a number of phase II clinical trials for SLE and CLE.^3–7 Biomarker studies have shown excellent correlation with cross-sectional data and change in the CLASI after therapeutic intervention.^8–10 At a recent international CLE meeting, based on the numerous validation studies and successful discrimination of skin improvement in clinical trials, it was unanimously agreed that the CLASI should be used as the skin outcome for lupus trials examining responsiveness in the skin.¹¹ After extensive consultation with the key leaders in the lupus community there was a similar recommendation for use of the CLASI for measuring lupus skin disease.¹² However, the CLASI, like any other outcome measure for lupus, needs to be used in a manner that is optimal for the conditions of the trial design.

SLE trials have traditionally added novel biologics to a potpourri of steroids, antimalarials and various immunosuppressants and this trial was no exception. In a trial design such as this, too much sensitivity to change in an outcome measure was bound to risk high placebo-group response rates, as was demonstrated in all the end points of this report. On the other hand, smaller phase II trials that decrease sensitivity to change by raising the bar for defining improvement too much may lower all response rates so far that the power to detect a difference is lost. A major advantage of the CLASI over the SLEDAI is that it collects data in a manner enabling flexible analytic approaches. This allows the small size of the trial vs. the active treatments given to the placebo group to inform the choice of an end point, optimizing these conflicting vectors of sensitivity to change, clinical significance of response and discriminatory capacity of the end point.

Previous studies have demonstrated the minimal clinically significant improvement in CLASI is a change of CLASI activity of four or 20% and there is a refined, linear correlation of the CLASI with quality of life scores down to a CLASI activity score of three. However, a higher cut-off of 50% improvement in CLASI activity has been determined to be optimal for phase II trials in lupus, and has demonstrated large differences between drug and placebo.^5–7 This 50% drop is also meaningful to patients as a 50% improvement in CLASI activity, if scores at entry to the study are eight or higher, demonstrates a meaningful 10-point drop in emotion and function scores on the Skindex, a validated quality of life measures for skin disease (unpublished data). Clearly, though, if scores at entry are very low, a 50% improvement may not mean as much.

In the study of baricitinib, all patients were grouped together for the analysis regardless of their baseline CLASI score, which led to a mean entry score in the assessed population of four. It is known that moderate-to-severe skin disease includes patients with CLASI score of 10 or higher.⁴ Since there is a high probability of floor effects when the amount of disease activity is too low at the start of the trial, any demonstration of improvement in the skin is impeded when the amount of disease activity is too low at entry.

To determine whether baricitinib is effective for skin, it would be important to specifically perform a subanalysis of patients with severe enough activity to demonstrate meaningful improvement. We are not yet convinced, between the very high bar set for SLEDAI improvement and the minimal disease included in the CLASI analysis, that CLE has yet been properly evaluated in the baricitinib study.

When potential new therapies are finally successful in running the gauntlet of clinical trials and U.S. Food and Drug Administration approval, it would be helpful to have the drug approved for patients with CLE who do not meet criteria for SLE if improvement is demonstrated for lupus skin disease. We know that 25% of patients with moderate-to-severe discoid lupus erythematosus do not meet criteria for SLE, but would benefit from new therapies. This recommendation was agreed upon in the white paper co-authored by many prominent experts within the lupus community.¹² Automatic disenfranchisement of these patients from treatments approved only for SLE underscores the importance of determining whether or not treatments efficacious for ‘rash or arthritis’ are actually working for CLE.

Of course, there are various forms of CLE. In the context of an SLE trial, the subset of CLE for each patient is not currently being reported. Hopefully this can change or as a minimum baseline pictures might help later affirm the subtype for subsequent analysis of the trial. On the other hand, recent studies demonstrate that even within a subtype of CLE there is heterogeneity that may correlate with clinical response to therapy.¹³ There is a need for more basic and translational investigation into the heterogeneity of CLE to allow for optimal selection of potential therapies and inclusion criteria for studies.

Lastly, given the cost of studies and need for better effective and safer treatments for patients with CLE, whether or not they have SLE, it would be optimal to have dermatologists included in the design and interpretation of outcomes to learn as much as possible about the effects of novel therapies on the skin in lupus.