The relative efficacy of monotherapy with Janus kinase inhibitors, dupilumab and apremilast in adults with alopecia areata: Network meta‐analyses of clinical trials

Janus kinase (JAK) inhibitors, biologics, and phosphodiesterase‐4 (PDE‐4) inhibitors are recent therapies for alopecia areata (AA)—albeit, knowledge gaps exist for these agents' relative efficacy.


| INTRODUC TI ON
Alopecia areata (AA) is estimated to have a global lifetime risk of 2%; the age of first onset is typically between 20 and 40 years. 1 This condition can affect the entire scalp (i.e., alopecia totalis) or the whole body (i.e., alopecia universalis). 1 Inhibitors of the Janus kinases (JAKs), biologics, and the phosphodiesterase-4 (PDE- 4) inhibitors are recent therapies for this condition. 2 The primary aim of our study was to determine the relative efficacy of JAK inhibitors, biologics and PDE4 inhibitors in individuals with AA.
Secondarily, we determined these agents' relative safety and examined whether patients' age and disease severity could modify these agents' efficacy. To our knowledge, this is the first network meta-analysis (NMA) study on the recent agents' relative efficacy.

| MATERIAL S AND ME THODS
Our work was conducted in accordance with the Preferred Reporting Items for Systematics Reviews and Meta-Analysis (PRISMA) guidelines; the protocol for the current study had been registered with the International Prospective Register for Systematic Reviews (INPLASY) under the ID 202220082.

| Identification of studies: literature review and data curation
Two authors (TW and AKG) systematically searched Cochrane Central Register of Controlled Trials, EMBASE (Ovid) and PubMed on July 25, 2022; a follow-up search was conducted on November 1, 2022. No language or date restrictions were applied. Items identified using the MeSH or Emtree term 'alopecia areata' were combined with results from the following search terms: 'alopecia areata', 'alopecia totalis', 'alopecia universalis', 'alopecia circumscripta'. Reference sections of previously published reviews, meta-analyses and articles retrieved in this study were screened for additional eligible studies. Our inclusion criteria were: randomized study design, study population diagnosed with AA above the age of 18, investigated agents classified as JAK inhibitors, biologics or PDE-4 inhibitors, and reported treatment efficacy using Severity of Alopecia Tool (SALT) scores. Two authors (TW and AKG) performed screening, identification and data extraction; any discrepancy was resolved through discussion with a third author (MAB).
Given the available data, our four measures of efficacy were 'the mean percentage reduction in SALT score at 24 weeks', and 'proportion of patients who achieved at least a 50%, 75% or 90% reduction in SALT score (SALT 50/75/90 ) at 24 weeks'. Our measure for safety was 'the number of reports for discontinuation due to any adverse event at 24 weeks'.

| Networks
For each of the five networks, the intention-to-treat (ITT) principle was applied wherever possible. Nodes represented regimens; a regimen was composite of an agent and its dosage. Edges represented direct evidence (i.e., of comparison of two nodes that were compared in an actual head-to-head trial); if a network was disconnected, arms of isolated studies were matched to the study with the most similar baseline characteristics. 3

| Statistical analyses
A Bayesian network meta-analysis (NMA) with vague priors was conducted for each network, using a random-and fixed-effects An intervention's SUCRA value, which ranges from 0% to 100% (inclusive), is an overall ranking metric; for our efficacy outcomes, higher SUCRA values correspond to higher efficacy. Regarding our outcome measure for safety, lower SUCRA values represent less safety. Relative effects, which we presented in league tables, correspond to comparative impact of every possible pair of interventions.
We investigated whether the efficacy of monotherapies as per likelihood of attaining SALT 50/75/90 could be modified by patients' age and disease severity through a multivariable ecological regression, whose generalized linear model (GLM) in Equation (1) is as follows: Y and ̂ correspond to the expected value of the outcome and regression coefficients, respectively. Patients' disease severity was quantified as the average SALT score at baseline. This multivariable regression model was updated, after data curation, to control for the effects of the various regimens.
The RStudio software was used; alpha (i.e., the cut-off for significance level) was set to 5%. 4

| RE SULTS
We examined the relative efficacy and safety of various therapeutic regimens for alopecia areata with data extracted from eight items, totalling nine randomized trials. [5][6][7][8][9][10][11][12] We found patients' age and baseline disease severity to be significant (p < 0.05) effect modifiers of regimens' efficacy (data not shown). Our study-level evaluation of evidence quality is presented in Figure S1.
(1) Y proportion who achieved SALTn =̂ age +̂ disease severity Our metric for 'safety' was discontinuation due to any adverse event-which was a secondary outcome. Thus, our findings for this outcome need to be verified with results of studies that investigated safety outcomes as a primary endpoint. Figure 1 summarizes our systematic search identifying eight articles, with results from nine randomized studies, eligible for data extraction; the study characteristics are detailed in Table S1.

| Search results, networks and assessment of evidence quality
The networks for our four efficacy outcomes of interest are presented in Figure 2 and Figures S2-S4; in our efficacy networks, the placebo and vehicle arms were amalgamated into one node, namely, 'placebo or vehicle'-which was the 'control' (i.e., reference) intervention; the placebo arm was the control node in our network for safety. The networks for efficacy were each disconnected, while that for safety was connected. Each of the five efficacy networks were disconnected because the interventions that Almutairi et al. (2019) investigated were not among the those which were investigated in the other remaining trials. 5 Given that each of the five networks had no closed loop, analysis of inconsistency between direct and indirect evidence was not feasible. 13 We used the Cochrane Collaboration's Risk of Bias Table to evaluate trials' quality of evidence ( Figure S1).

| Relative efficacy
We used a Kilim plot to present SUCRA values of our primary outcomes; such plots allow information to be presented in a visually intuitive manner ( Figure 3). 14 Each column in a Kilim plot represents a set of values, and the color of the cells therein is 'proportional' to the value thereof-such that a column of numerically ordered cells follow 'a colour gradient'. The closer the color gradient of other columns are to that of the column with the ordered cells, the greater the congruency between the sets of values. Table 1  No consistent statistically significant differences were observed between the top-ranking regimens. However, there were F I G U R E 1 Schematic of the search process. This flow chart depicts how the studies-whose data would be used for all quantitative analyses-were identified.

| Comparative safety
We used data from four studies to determine regimens' relative safety as per 'number of reports' of discontinuation due to any adverse events at 24 weeks. 6,7,9,11 The SUCRA values are presented in Table S2 where higher values correspond to more safety (i.e., less reporting); the column for safety is ordered and there is congruency in color gradient across all three outcomes-as per the overlapping regimens. The league table of relative effects is presented in Figure S9.

| Inconsistency analyses
Though inconsistency analyses were not feasible as there were no closed loops, we performed sensitivity analyses: we ran network meta-regressions where we conducted our network meta-analyses whilst adjusting for patients' age and/or baseline severity (data not shown). The results of our sensitivity analyses were, by and large, inconclusive-and this can be attributable to the fact that the evidence base is scant; hence, more informative sensitivity analyses would be feasible with an increase in empirical data.

| DISCUSS ION
In this network meta-analysis, results from nine randomized and controlled trials, totalling 1812 adult AA patients, were extracted for evidence synthesis. [5][6][7][8][9][10][11][12] We were able to assess the relative efficacy of JAK inhibitors, biologics and PDE-4 inhibitors using multiple clinically relevant endpoints measured in SALT scores. Overall, results from four networks were congruent in demonstrating the superior, dose-dependent, efficacy of oral JAK inhibitors in its higher probability of inducing hair regrowth at week 24 compared to dupilumab, apremilast and placebo (Figure 3). Oral JAK inhibitors also demonstrated decreased likelihood of treatment discontinuation due to adverse events (Table S2). greater disease severity at baseline were significantly associated (p < 0.05) with decreased probability of achieving improvements in SALT score. [5][6][7]9,11,12 A dose-dependent effect in efficacy was observed for patients treated with baricitinib and deuruxolitinib. Specifically, patients treated with deuruxolitinib 12 mg BID showed significantly higher probability of achieving ≥90% improvements in SALT scores compared to the 8 mg BID or the 4 mg BID dose (Figure 3, Figure S7). Similar results were seen for patients treated with baricitinib 4 mg QD compared to the 2 mg QD dose. In contrast, ruxolitinib 1.5% cream was significantly less efficacious compared to its oral formulation, as well as other oral JAK inhibitors (Figure 3, Figures S5,   S7 and S8). Furthermore, we found lower probabilities of treatment discontinuation due to adverse events following oral JAK inhibitor treatment compared to apremilast further supporting its use in adult AA patients (Table S2).
In agreement with the findings of previous studies, current data does not appear to show significant differences in efficacy between agents of the oral JAK inhibitors class; hence, further long-term studies with head-to-head comparisons are warranted as more agents become available. 15,16 Baricitinib has been approved by the U.S.

Food and Drug Administration (FDA) and the European Medicines
Agency for the management of severe AA. 17,18 Results from phase 3 studies on deuruxolitinib and ritlecitinib are expected to strengthen the therapeutic potential of oral JAK inhibitors (Clini calTr ials.gov; NCT04797650 and NCT03732807).
Clinically meaningful endpoints not assessed in the present study due to insufficient data include the proportion of patients achieving SALT ≤ 20 (≤20% residual disease or ≥80% scalp coverage). 19,20 Among the repertoire of metrics for quantitatively assessing al-  Extending baricitinib treatment in AA patients for ≥1 year did not result in increased incidence of adverse reactions, which were predominately mild or moderate in severity. 27 There is insufficient data in the current literature to compare the therapeutic options for AA with respect to long-term efficacy and the durability of treatment response; further clinical studies are warranted to better assist physicians as newer agents become available.
Given that our analyses excluded the pediatric (i.e., age < 18 years) population, our conclusions may not be generalizable to children; this exclusion is the reason our analysis did not include data from the ALLERGO trial (Clini calTr ials.gov; NCT02974868). Therefore, one extension of our work could be to determine whether our findings-as per efficacy, safety, and effect modification-would be replicated in a juvenile sample.
The population of patients that was included in our analyses were those whose AA only involved the scalp region; hence our findings are not generalizable for treatment non-scalp regions (e.g., eyebrows). Thus, our results support the conduct of more trials investigating the efficacy of newer AA agents for non-scalp AA.
As AA is a chronic condition that not only affects hair coverage but also causes impairments to the quality of life, further clinical studies incorporating patient-centered endpoints are warranted. 28

| CON CLUS IONS
The existing newer agents such as the JAK inhibitors, biologics and the PDE4 inhibitors are a promising development for the management of alopecia areata-which is a chronic condition where remissions and relapses are not uncommon. In this study, we found the oral JAK inhibitors to be more efficacious in inducing significant, and dose-dependent, hair regrowth at 24 weeks in adult AA patients measured by SALT scores; in addition, the JAK inhibitors have a lower likelihood of treatment discontinuation due to adverse events compared to apremilast. There is lim-

FU N D I N G I N FO R M ATI O N
There was no external funding.

CO N FLI C T O F I NTER E S T S TATEM ENT
The authors have no conflict of interest to declare.

DATA AVA I L A B I L I T Y S TAT E M E N T
The data that support the findings of this study are available from the corresponding author upon reasonable request.

E TH I C S S TATEM ENT
The authors declare the human ethics approval was not required for this study.