FormalPara Key Summary Points

Why carry out this study?

Psoriasis patients value both rapid initial response and the magnitude and durability of long-term clinical response; however, a single measure encompassing both of these is not in common use.

What did the study ask?

What are the long-term cumulative clinical benefits, assessed as complete and almost-complete skin clearance based on area under the curve, for biologics in the treatment of moderate-to-severe psoriasis?

What was learned from the study?

In a network meta-analysis assessing area under the curve up to 52 weeks, ixekizumab showed the highest cumulative benefit for complete clearance, followed by risankizumab and other biologics.

For almost-complete clearance, risankizumab showed the highest cumulative benefit, followed by ixekizumab and other biologics.

Introduction

Psoriasis is a chronic inflammatory skin disease affecting approximately 2–3% of the population in the USA [1]. Many biologic therapies that target critical immune-mediated pathways involved in the pathogenesis of psoriasis have been approved, including first-generation tumor necrosis factor (TNF) inhibitors, followed by second-generation biologics such as interleukin (IL)-17, IL-12/23, and IL-23 inhibitors, which elevated the treatment goals for psoriasis. Since there are many biologic choices available to patients and practitioners, comparative data among biologics are needed to help inform treatment decisions.

Both rapid onset of treatment response and the magnitude of sustained response are important to patients affected by psoriasis [2,3,4]. However, in chronic inflammatory diseases generally, it is also important to assess cumulative disease impact as a marker of overall inflammatory burden; for example, in psoriatic and rheumatoid arthritis, structural damage serves as both an important endpoint and as a surrogate marker of cumulative disease impact. In patients suffering with chronic skin disease, including psoriasis, cumulative life course impairment (CLCI) [5] is a multidimensional tool designed to assess the cumulative burden of skin disease over time. Area under the curve (AUC) for clinical response provides a cumulative assessment of treatment effect, capturing not only the magnitude and speed of onset, but also the maintenance of clinical response over time [6,7,8]. Indeed, measurements of long-term cumulative benefit may provide more clinically meaningful data than measurements of response at a single timepoint.

Here, we present a network meta-analysis (NMA) comparing the cumulative clinical benefit after 1 year for ten biologic drugs at approved doses for psoriasis, across different classes of inhibitors of key cytokine-mediated pathways, including IL-17 (brodalumab, ixekizumab, secukinumab), IL-12/23 (ustekinumab), IL-23 (guselkumab, risankizumab), and TNF (adalimumab, certolizumab, etanercept, infliximab) based on the achievement of Psoriasis Area Severity Index 90% (PASI 90) and 100% (PASI 100) improvements. This analysis provides clinically meaningful and relevant comparative data on biologics that should better assist patients and practitioners in their treatment choices for moderate-to-severe psoriasis.

Methods

Data Source

Published phase 3 randomized clinical trial data of biologics given at approved doses for psoriasis were obtained from a systematic literature review as previously described (1990–2018) [6], with the literature search updated through to 30 September 2020 prior to this analysis (Supplementary Materials Fig. 1). The ongoing search is registered at PROSPERO (CRD42021244387). Searches were conducted across the Embase, PubMed, and Cochrane Library electronic databases, and titles and abstracts were reviewed independently by two researchers for inclusion and exclusion. Additional search strategies included conference abstracts, hand searches, clinical trial registries, and health technology websites, updated from the last search. A total of ten biologics were included in the PASI 90 and eight biologics in the PASI 100 analyses. These comprised TNF inhibitors: adalimumab [maintenance dosing 40 mg every 2 weeks (Q2W)], certolizumab (200 mg Q2W, PASI 90 only), etanercept (50 mg biweekly), and infliximab [5 mg/kg every 8 weeks (Q8W), PASI 90 only]; IL-17 antagonists: brodalumab (210 mg Q2W), ixekizumab [80 mg Q2W/every 4 weeks (Q4W)], and secukinumab (300 mg Q4W); IL-23 inhibitors: guselkumab (100 mg Q8W) and risankizumab [150 mg every 12 weeks (Q12W)]; and the IL-12/23 inhibitor ustekinumab (45/90 mg Q12W). AUC 52-week data were not available for tildrakizumab, an IL-23 inhibitor. Studies with rerandomization based on “responder” status without providing the visit-wise response rate on PASI 90/100 through to 52 weeks for each patient were excluded from the analysis; examples of studies excluded on this basis included the ixekizumab UNCOVER-1 [9], UNCOVER-2 [9], and IXORA-R [10] studies, and the tildrakizumab reSURFACE 1 [11] and reSURFACE 2 [11] studies. Phase 2 trials were also excluded from this analysis, as were trials for apremilast, or if AUC could not be calculated from published data available at 4-week intervals.

The 14 published studies of 18 clinical trials used for data extraction are presented in Table 1. Data extraction was quality checked by a second researcher not involved with the extraction; this quality check included verification of all extracted data against the source document.

Table 1 Overview of the studies for network meta-analysis
Full size table

This article is based on previously conducted studies and does not contain any new studies with human participants or animals performed by any of the authors.

Outcome Measures

Total cumulative benefit for ixekizumab and other biologic treatments from Week 0 to 52 at 4-week intervals was estimated using the AUC method as previously described [6,7,8]. The ratio of AUC to the maximum AUC during the 52 weeks (the normalized AUC) was calculated for all ten biologics. Specifically, the normalized AUC for each treatment was calculated as a proportion of maximum possible AUC. The normalized AUC was compared among biologics using a Bayesian NMA (BNMA) fixed effect model with a separate random effect baseline model [12, 13] to obtain stable estimates from the NMA and still account for the effects of the placebo arm on PASI 90/100 (NMA diagrams for PASI 90 and PASI 100, Supplementary Materials Fig. 2). Convergence for all models was assessed using trace plots as modified by Brooks and Gelman [14]. BNMA were performed in JAGS via R using the R2JAGS package and using R version 4.03 [15]. Cumulative days of PASI 90/100 responses were calculated by multiplying the normalized AUC for PASI 90/100 by the study duration (in days).

Data on PASI 90/100 used for AUC calculations for active treatments in this analysis were based, wherever possible, on nonresponder imputation (NRI) for missing data. Exceptions were for PASI 90 of one report for certolizumab [16] and two reports of studies assessing secukinumab versus ustekinumab [17, 18], and, for PASI 100, one report for secukinumab versus ustekinumab [18], all of which reported endpoints based on multiple imputation for missing data. For the ixekizumab UNCOVER-3 trial, in the present analyses, unpublished data based on the more conservative NRI method were used instead of the previously published modified NRI (mNRI) data. In addition, the placebo arms of the studies typically continued up to Week 12 (or 16); data after Week 12 (or 16) were imputed to Week 52 using last observation carried forward (LOCF). If the original study was of 48-weeks duration, LOCF was used to impute all missing values for PASI 90/100 to 52 weeks.

Sensitivity analysis was conducted by evaluating the normalized AUC at Week 52 for 8 of the 18 studies in the primary analysis that maintained randomization from baseline to 48 or 52 weeks: AMAGINE-2 [19], AMAGINE-3 [19], CLEAR [18], FIXTURE [20], IXORA-S [21], UltIMMa-1 [22], UltIMMa-2 [22], and ECLIPSE [23].

In all analyses, between-drug differences were considered significant where 95% credible intervals were nonoverlapping (normalized AUC and cumulative days) or did not overlap 1 (relative effects ratios).

Results

An overview of the clinical trials included in this NMA are listed in Table 1. Included for each trial is the duration of study, baseline PASI, and the response rates for PASI 90/100 at the last timepoint of the study through Week 52.

Normalized maximum AUC for complete (PASI 100) and almost-complete (PASI 90) skin clearance at Weeks 16 and 52 are shown in Fig. 1. The numerically highest cumulative clinical benefits [median AUC (95% credible interval)] for complete clearance over 52 weeks were observed for ixekizumab [0.436 (0.405, 0.467)], risankizumab [0.423 (0.397, 0.449)], and brodalumab [0.378 (0.326, 0.430)] (Fig. 1). For almost-complete skin clearance, the highest cumulative benefits were observed for risankizumab [0.671 (0.645, 0.697)], ixekizumab [0.642 (0.610, 0.673)], and brodalumab [0.632 (0.599, 0.656)]. Of note, ixekizumab and brodalumab, two IL-17 inhibitors, each showed significantly greater normalized AUC for almost-complete skin clearance versus secukinumab, another IL-17 inhibitor [0.578 (0.559, 0.597)], and ixekizumab also showed significantly greater normalized AUC for complete skin clearance versus secukinumab [0.324 (0.303, 0.344)]. Among IL-23 inhibitors, risankizumab, showed significantly greater normalized AUC versus guselkumab for both complete and almost-complete skin clearance [0.358 (0.330, 0.386) and 0.614 (0.587, 0.641), respectively]. Ixekizumab also showed significantly greater normalized AUC for complete skin clearance versus guselkumab, while risankizumab showed greater normalized AUC for complete and almost-complete skin clearance versus secukinumab. Results from the sensitivity analyses comparing absolute effects in the main NMA dataset versus 52-week data for the eight clinical trials randomized with active controls through 52 weeks were consistent with the overall findings (Table 2).

Fig. 1
figure 1

Placebo-adjusted normalized maximum AUC for PASI 100 and PASI 90 at 16 and 52 Weeks. Data displayed as median (95% credible interval). AUC area under the curve, PASI 100/90 100% or ≥ 90% improvement in Psoriasis Area and Severity Index

Full size image
Table 2 Normalized AUC at week 52 for the primary analysis based on all studies and the sensitivity analysis based on eight studies
Full size table

The cumulative benefits expressed as cumulative days of response over 52 weeks for PASI 100 and PASI 90 are shown in Fig. 2. Cumulative days [median days (95% credible interval)] at PASI 100 were highest for ixekizumab [158.7 (147.4, 170.0) days], followed by risankizumab [154.0 (144.9, 163.4) days], and brodalumab [138.0 (118.7, 156.9) days], with a range of 31.7 (23.7, 39.7) to 130.7 (120.5, 140.9) days for the other biologics in this analysis. Cumulative days at PASI 90 were highest for risankizumab [249.3 (239.5, 259.2) days], followed by ixekizumab [238.8 (227.1, 250.8) days], and brodalumab [235.1 (222.8, 247.5) days], with a range of 107.4 (93.9, 120.8) to 228.6 (218.4, 238.8) days for the other biologics in the analysis.

Fig. 2
figure 2

Cumulative days of response at PASI 100 and PASI 90 over 52 weeks. Data displayed as 100% maximum possible area under the curve and 95% credible interval. ADA adalimumab, BRO brodalumab, CER certolizumab, ETN etanercept, GUS guselkumab, INF infliximab, IXE ixekizumab, PASI 100/90, 100% or ≥ 90% improvement in Psoriasis Area and Severity Index, RIS risankizumab, SEC secukinumab, UST ustekinumab

Full size image

The relative cumulative benefits for PASI 100 and PASI 90 over 52 weeks by ranking among the biologics is shown in Fig. 3. For PASI 100, ixekizumab showed a relative benefit that was 3% higher than risankizumab and 15% higher than brodalumab (both not significant), and significantly higher than guselkumab (22%) and secukinumab (35%). For PASI 90, risankizumab showed a relative benefit that was 4% higher than ixekizumab and 6% higher than brodalumab (both not significant), and significantly higher than guselkumab (9%) and secukinumab (16%). All of the IL-17 and IL-23 inhibitors assessed showed significantly greater relative cumulative clinical benefit for both PASI 100 and PASI 90 at 52 weeks versus the TNF inhibitors and ustekinumab.

Fig. 3
figure 3

Ratio of relative risk between drugs for cumulative benefit based on PASI 100 or PASI 90 at Week 52. Relative benefits calculated as relative risk between drugs for achieving PASI 100 or PASI 90 based on mean (SE) for normalized AUC. Relative effects with credible intervals not overlapping 1 are shown in bold. ADA adalimumab, BRO, brodalumab, CER certolizumab, ETN etanercept, GUS guselkumab, INF infliximab, IXE ixekizumab, PASI 100/90 100% or ≥ 90% improvement in Psoriasis Area and Severity Index, RIS risankizumab, SEC secukinumab, UST ustekinumab

Full size image

Discussion

Many factors have to be considered when selecting the optimal therapy for psoriasis patients [24, 25]. Among them, speed, as well as magnitude and durability of response are important [2,3,4, 26, 27]. Results from our NMA suggest that ixekizumab followed by risankizumab provided the greatest cumulative benefits in complete skin clearance (PASI 100) over 52 weeks of treatment relative to guselkumab, secukinumab, and TNF inhibitors, with brodalumab intermediate between these tiers. For almost-complete skin clearance (PASI 90), risankizumab and then ixekizumab, brodalumab, and guselkumab provided the highest long-term cumulative benefit. As a class, TNF inhibitors provided the lowest cumulative benefits over 1 year of treatment. Findings obtained from the sensitivity analyses without inclusion of randomized clinical trials with a placebo control arm were consistent with the main results, indicating that the data from this NMA were robust.

Patients and clinicians are facing many options for biologic treatment for moderate-to-severe psoriasis. Determining the optimal choice is difficult because many factors must be considered simultaneously. One important factor is efficacy, but efficacy comparisons can be complicated without direct head-to-head studies, since comparisons across studies may be impacted by differences in patient populations, study investigators, methods for handling missing data, placebo response rates, and primary endpoint measurements (e.g., PASI 75 versus PASI 90 versus PASI 100) and timepoints (Week 12 versus Week 16 versus Week 52). These differences may be partially mitigated by network meta-analyses, which use common comparisons to normalize differences between studies. In addition, it has becoming increasingly clear that measuring efficacy responses to medications at single timepoints provides a limited view of therapeutic benefit to patients. The impact of a chronic disease such as psoriasis on patients is continuous, with each day being important in terms of clinical and quality of life improvements. Therefore, assessing the cumulative benefit of treatment, as estimated here using AUC analysis, takes into consideration both early rapid response and long-term maintenance of higher skin clearance levels by therapies such as ixekizumab and risankizumab, and provides a more complete picture of the efficacy of different treatments over time. In addition, estimating the cumulative days of a response provides a more stable measure than estimates of consistency or maintenance of response based on arbitrary definitions such as consecutive visits with a response.

Patients with psoriasis achieve the best quality of life with complete skin clearance (PASI 100), even when compared with almost-complete skin clearance (PASI 90) [26,27,28]. A study by Dey et al. also reported greater improvements in vascular inflammation and atherosclerosis in psoriasis patients with more complete versus partial skin clearance following treatment [29]. With the emergence of biologics capable of achieving high levels of clearance, such advantages have become more evident, and treatment goals have shifted to reflect the importance of complete skin clearance [30, 31]. In the current NMA, treatment of psoriasis with ixekizumab, an IL-17A inhibitor, led to the greatest amount of time with completely clear skin over the course of 1 year, corresponding to 23 weeks (44%) of the year. Risankizumab ranked second by this measure (22 weeks, 42%), whereas TNF inhibitors as a class demonstrated the lowest amount of time with completely clear skin over 1 year. These data for ixekizumab and risankizumab reflect a combination of excellent speed of onset and durability of response over time for both drugs.

Notably, while there were small numerical differences in cumulative benefit between ixekizumab and risankizumab for both complete and almost-complete skin clearance, these did not reach statistical significance. However, for both complete and almost-complete skin clearance, ixekizumab was superior to secukinumab, and risankizumab was superior to guselkumab. Thus, while the best performing biologics in both the IL-17 and IL-23 classes offered a high level of cumulative clinical benefit, significant differences were observed between biologics within both of these drug classes. In addition, each of the IL-17 and IL-23 inhibitors were superior to TNF and IL-12/23 inhibitors, indicating significant differences in cumulative clinical benefit between first-generation biologics and all agents in the anti-IL-17 and anti-IL-23 classes.

This analysis has limitations. As in all meta-analyses, heterogeneity in patient characteristics between trials may introduce bias in the comparisons even though only phase 3 randomized clinical trials with similar inclusion and exclusion criteria were included in this study. Missing data often present challenges to comparative effectiveness studies in general, as was the case for this study in particular, which required reporting of data at all visits from baseline to Week 52 to accurately calculate cumulative benefits. Studies that only reported response rates at limited timepoints, e.g., at Weeks 12, 16, or 52, were excluded from this analysis. In addition, we tried to eliminate data variations due to different missing data imputation methodology by selecting reports that used NRI for missing data imputation; however, a limited number of studies that used multiple imputation were included in the NMA due to data availability. Furthermore, imputation of missing values for the placebo arms beyond the 12/16 week induction phases could introduce variability to the estimated placebo-adjusted normalized AUC. A random effect baseline model was used in the NMA to address the variability, and sensitivity analysis using trials without the placebo arm were conducted to assess the consistency of the findings.

Another limitation in terms of informing clinical decision making is that this analysis includes only data from randomized controlled trials. One recent real-world study comparing the effectiveness of guselkumab versus risankizumab showed comparable efficacy through 44 weeks [32, 33], albeit with a relatively small sample size, and other studies have also assessed real-world efficacy outcomes for these drugs, although also in very limited numbers of patients [34,35,36,37]. Nevertheless, the use of real-world data should be considered for future research when comparing outcomes across treatments.

Conclusions

In summary, this study provides clinically meaningful and relevant comparative data on biologics that reflects a combined measurement of therapeutic onset and magnitude of sustained skin clearance over time, and should better assist clinicians in their treatment choices for patients with moderate-to-severe psoriasis.