Upadacitinib Is a Better Choice than Abrocitinib for Patients with Moderate-to-Severe Atopic Dermatitis: An Updated Meta-Analysis

,


Introduction
Atopic dermatitis (AD) is a chronic recurrent infammatory skin disease characterized by severe pruritus and afects up to 10% of adults and 25% of children and adolescents [1,2]. AD has complex pathophysiology involving environmental factors, impaired skin barrier function, genetic susceptibility, and immune imbalance [1]. Tese pathophysiological changes result in the loss of transepidermal water, xerosis, and eczema, leading to the destruction of skin barriers in about 80% of AD patients [3][4][5]. Individuals with AD are also at high risk of asthma, allergic rhinitis, and food allergy, which could also increase the hazards of relevant health and psychosocial outcomes [6].
Although topical anti-infammatory agents and emollients are the primary treatment for AD, they are not effective enough for patients with moderate-to-severe AD [7]. Topical corticosteroids (TCS) are widely regarded as the frst-line choice for moderate-to-severe AD, but side efects of TCS could be hardly ignored, including skin atrophy, hypothalamic-pituitary-adrenal axis suppression, and acneiform or rosacea-like eruptions [7]. Charman et al. had reported a phenomenon called TCS phobia: they included 200 patients with atopic eczema for the questionnaire, and the outcome demonstrated that 145 patients (72.5%) worried about TCS treatment, while 48 patients (24%) admitted noncompliance because of safety concerns [8]. As for other options, systemic corticosteroids are not recommended due to side efects and rebound [6,9]. Phototherapy is also constrained by adverse events like actinic damage, local erythema, burning, and stinging [9]. Terefore, a new efcient therapeutic agent for AD that avoids most of the side efects mentioned above is needed.
Recently, biological blocking agents for immune cytokine pathways were reported as an optional treatment for moderate-to-severe AD [6]. In recent years, the US Food and Drug Administration approval had been given to two oral JAK inhibitors (abrocitinib and upadacitinib) [10]. Eight randomized clinical trials (RCTs) of JAK1 inhibitors in AD have been published in the past three years [11][12][13][14][15][16][17]. JAK1 inhibitors regulate signal transduction and relieve pruritus through IL-4, IL-13, and other cytokines such as IL-31, IL-22, and thymic stromal lymphopoietin [18]. Meanwhile, JAK1 inhibitors could avoid potential risks of neutropenia and anemia caused by JAK2 inhibition [19]. Terefore, we conducted this meta-analysis to evaluate the efcacy and safety of JAK1 inhibitors, especially upadacitinib versus abrocitinib, for the treatment of moderate-to-severe AD. Both upadacitinib and abrocitinib are oral selective JAK inhibitors with greater inhibitory potency for JAK1 than JAK2, JAK3, or tyrosine kinase 2 (TYK2).

Search Strategy.
Our review was registered in PROS-PERO (registration number CRD42021244435) before the literature search. We followed the guidelines of Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) ( Table S1). Two independent reviewers (RL and PH) searched Web of Science, PubMed, Embase, and Cochrane databases updated on Mar 21st, 2021 for RCTs (we processed another search at the end of the study on Apr 11th, 2023). Te search strategy used for the PubMed database is available as supplementary material (Table S2). To expand the search range, the keywords were "atopic dermatitis," "Janus kinase inhibitor," or "JAK inhibitor." Te link https:// clinicaltrials.gov was searched for completed but unpublished RCTs. Two researchers (YZ and SR) independently screened the titles and abstracts. Tey only reviewed full-text articles which met the inclusion criteria. Reference lists of eligible reviews and trials were searched for additional citations.

Selection Criteria.
Tere was no restriction on sex, age, nationality, and race. Oral placebo treatment with an identical appearance was regarded as a comparison. Tere was no restriction on the dosage of JAK1 inhibitors. We only discussed data from the phase two and three RCTs. We only included data from patients with moderate-to-severe AD in the meta-analysis. Patients were permitted to use oral antihistamines and nonmedicated emollient. Patients with acute or chronic medical or psychiatric conditions, laboratory abnormalities, infectious diseases, coagulation disorders, receiving other therapies (individual explanation in diferent trials) before randomization, or having prior exposure to any JAK1 inhibitor were excluded. Concomitant use of topical (corticosteroids, calcineurin inhibitors, phosphodiesterase inhibitors, tars, antibiotic creams, or topical antihistamines) or other systemic therapies for AD or rescue medication were also prohibited.

Data Extraction.
Two researchers (YZ and PH) independently extracted data from eligible articles. Te extracted data included characteristics of the study, characteristics of the patient, baseline, and outcome data. Decisions were made by consulting another reviewer SR when YZ and PH met disagreements and failed consensus. Tey would also contact the corresponding author by email to send additional information when data were incomplete. Outcomes were classifed as primary outcomes and secondary outcomes. Primary outcomes included the proportion of IGA responders (IGA ≤ 1 or achieving a ≥2-point improvement from baseline) and the proportion of EASI-75 responders (improvement ≥75% in EASI from baseline). All assessment tools, in included RCTs, are shown in Table S3.

Quality Assessment. Two researchers (YZ and RL) used
Cochrane risk of bias assessment tool (CROBAT) to assess the quality of included studies independently. CROBAT included "allocation concealment," "random sequence generation," "blinding of participants and personnel," "blinding of outcome assessment," "incomplete outcome data," "selective reporting," and "other bias" (Table S4). Each question had 3 answers: "low risk," "moderate risk," and "high risk." According to the published information, researchers would assess the risk level of RCTs. Another reviewer BL would make decision when YZ and RL met disagreements and failed consensus. Small-study efects that led to potential reporting or publication bias could be calculated by Egger's test [20]. Publication bias was evaluated by funnel plots and P ≤ 0.05 was considered a statistically signifcant risk of bias. We used the Grading of Recommendations, Assessment, Development, and Evaluation (GRADE) tool to evaluate the quality of evidence for each outcome. Te GRADE tool classifed evidence of outcomes into "high," "moderate," "low," and "very low." Each assessment could reduce or promote the level of quality. Specifc rules are explained in Table S4.

Statistical
Analysis. STATA 16.0 and Review Manager 5.3 were used in our study. All forest plots were produced by Review Manager 5.3 with inverse variance as the statistical method. Continuous data using diferent scales would be measured by standard mean diference (SMD) with 95% confdence intervals (CI), while using the same scale would be summarized by weighted mean diference (WMD) with 95% CI. Dichotomous data would be calculated by odds ratio (OR) with 95% CIs [21]. Heterogeneity in the result of the meta-analysis was assessed using Cochrane Q and I 2 statistics with appropriate analysis models. All statistical tests were two-tailed, and P ≤ 0.05 was regarded as a statistically signifcant diference. Egger's tests were performed by STATA 16.0.
Subgroup analysis would be carried out when detailed data were available. It would be based on the type, dosage, and treatment time of JAK1 inhibitors. Diferent doses of the same drug were compared directly by the forest plot, where P value of the test for overall efect would demonstrate the signifcance. Besides, P value of the test for subgroup difference demonstrated the diference between diferent drugs because we regarded two drugs as two subgroups. Sensitivity analysis was performed in the meta-analysis by excluding each study once at a time to check whether the efectiveness of the outcome was determined by individual studies. Figure 1 shows detailed steps of the literature search, in which 404 studies were reviewed: 314 studies were excluded after screening titles and abstracts, remaining 90 studies were reviewed in full text. After excluding 82 studies according to selection criteria, eight RCTs with 4634 moderate-to-severe AD patients were included in our meta-analysis [11][12][13][14][15][16][17]. Table 1, all RCTs were multicentre trials: 6 (75%) were phase III clinical trials, and 2 (25%) were phase II clinical trials. Te sample size of included studies ranged from 167 to 901. Tere was no signifcant diference between the JAK1 inhibitor group and placebo group in general information (age, sex, disease duration, IGA point, and EASI). Figure S1 demonstrates the risk of bias summary. All included RCTs met compliance with the selection criteria. All RCTs were blind and followed the rules of randomization and allocation concealment. Tree RCTs without results posted on https://clinicaltrials. gov had risks of incomplete outcome data. Detailed assessment outcomes of CROBAT are shown in Table S5.
Moreover, the JAK1 inhibitor group resulted in more pruritus numerical rating scale (NRS) responders (defned as a 4-point or greater improvement from baseline in NRS score) than the placebo group (OR � 5.74 95% CI 4.18-7.88), and upadacitinib group displayed signifcant diferences compared with the abrocitinib group in Table S6.2 (OR � 9.21 95% CI 7.74-10.95; subgroup diference: P � 0.0001). JAK1 inhibitors also showed signifcant differences in scoring atopic dermatitis (SCORAD) and decreased lesion area, but the levels of GRADE assessment were not high due to high heterogeneity and risk of publication bias.
In our supplemental tables, we have presented various outcomes: PRISMA checklist (Table S1), detailed search strategy (Table S2), diferent assessment tools in included RCTs (Table S3) and our meta-analysis (Table S4), and outcome of CROBAT (Table S5). Table S6 shows the secondary outcomes including EASI-50/90/100 responders and other outcomes such as Pruritus NRS responder, SCORAD responder, and BSA change. We also provide a recommendation of infammatory diseases with JAK inhibitors treatment in Table S7.

Discussion
Our meta-analysis included eight high-quality multicentre RCTs with 4634 patients. We confrmed the efects of JAK1 inhibitors on moderate-to-severe AD both at EOT and after 2 weeks of treatment (Table 2). However, both upadacitinib Journal of Clinical Pharmacy and Terapeutics and abrocitinib groups demonstrated signifcant diferences in acne and headache. Besides, upadacitinib had signifcantly higher risks of upper respiratory tract infection and nasopharyngitis, and abrocitinib had signifcantly higher risks of nausea ( Table 3). As for subgroup analysis, we observed that both upadacitinib and abrocitinib demonstrated signifcant therapeutic efects and dose-dependent relationships (30 mg upadacitinib group and 200 mg abrocitinib group showed better outcomes than lower dose groups). Rapid response was displayed on JAK1 inhibitors after 2 weeks of treatment, both upadacitinib and abrocitinib demonstrated signifcant diferences in EASI-75 responder compared with the placebo ( Based on the current understanding of immunological mechanisms, various inhibitors targeting cytokines and interfering with signaling pathways have been developed. JAK1, 2, 3, and TYK2 signal pathways modulate the infammatory processes by activating intracytoplasmic transcription factors, namely signal transducer and activator of transcription (STAT) [22]. Activated proteins form dimers and translocate into the nucleus to modulate the expression of genes and fnally regulate type-2 diferentiation, T-cell activation, innate immunity, and epidermal diferentiation complexes [6,23]. JAK1 inhibitor regulates signal transduction and relieves pruritus through IL-4, IL-13, and other cytokines such as IL-31, IL-22, and thymic stromal Studies included in the meta-analysis (n=8)         [24]. Moreover, JAK2 participates in signal transduction by erythropoietin and other colony-stimulating factors, which results in frequent neutropenia and anemia events in JAK2 inhibitors [25]. Terefore, compared with other JAK inhibitors, JAK1 inhibitors had the potential advantage of improving pruritus and avoiding hematologic adverse events [19]. Apart from the general mechanism of the JAK1 inhibitor mentioned before, upadacitinib also decreases the production of proinfammatory mediators induced by IL-6, IL-15, IFN-α, and IFN-c [26]. As for pharmacology, upadacitinib demonstrated an oral bioavailability of 76% with rapid absorption (plasma concentrations peak at around 1 to 2 hours after administration) and a 4-hour half-life, while abrocitinib also had rapidly absorbed after oral administration (reaching plasma concentrations peak within 1 h) and a 5-hour half-life [27,28].
A new recommendation for immune-mediated infammatory diseases with JAK inhibitors treatment was formulated by an expert committee comprising 29 multinational and experienced clinicians [29]. High levels of agreement were voted on for every point (10-point scale). It recommended that severe infections, severe organ dysfunction, pregnancy, and lactation were contraindications of JAK inhibitors (vote 100%, 9.9 points) and consider dosage should be adjusted in patients with higher age (>70 years) and signifcantly impaired renal or hepatic function (vote 100%, 9.9 points). As for the adverse events, serious infections, particularly opportunistic infections, including herpes zoster, received general consent. Te risk of infection could be lowered by reducing or eliminating concomitant glucocorticoid use (vote 100%, 9.9 points). Lymphopenia, thrombocytopenia, neutropenia, and anemia may also occur (vote 100%, 9.8 points). More detailed recommendations are displayed in Table S7.
Various adverse events were reported, in included RCTs. Both abrocitinib and upadacitinib demonstrated risks of acne and headache in our study. Although most of these symptoms were not serious, they might exacerbate the manifestation of AD, resulting in lower quality of life. Besides, 28 (1.5%) patients in the upadacitinib group and 13 (1.0%) patients in the abrocitinib group had reported herpes zoster, but the diference in incidence rate was insignifcant, and all cases of herpes zoster infection were nonserious, and none led to discontinuation of RCTs. Furthermore, JAK1 inhibitors might suppress platelet production by inhibiting Ashwell-Morrell receptors and downstream inhibition of thrombopoietin production [30]. Dose-related decreases in median platelet count were reported in all RCTs of abrocitinib, but the minimum level of platelet occurred in the 4th week and gradually returned to the baseline value with the ongoing administration of the drug. Besides, no plateletrelated event was reported in RCTs of upadacitinib. Terefore, it seems platelet count is reversible in most patients, and no clinically important event was observed (such as hemorrhage associated with the decreased platelet counts. A recent meta-analysis included 42 studies that discussed the venous thromboembolism (VTE) risk with JAK inhibitors; no evidence could support the current warnings of VTE risk for JAK inhibitors [31]. Another meta-analysis included 82 studies had also concluded that JAK inhibitors demonstrated no increased risk of malignancy or serious infections on multiple immune-mediated diseases [32]. Other side efects, including acne, nasopharyngitis, headache, and nausea, were frequently reported in included RCTs, but most side efects were not severe and relieved after drug withdrawal.
Several previous meta-analyses discussed the JAK inhibitors treatment on AD. Among these studies, only one network meta-analysis showed a similar design and conclusion to our study [33]. However, it only included one RCTof abrocitinib and one RCTof upadacitinib, and adverse events were not included in the data analysis. Our study focused on evaluating JAK1 inhibitors and comparing the efcacy and safety of upadacitinib and abrocitinib for moderate-to-severe AD. In summary, our study demonstrated rapid response and dose-dependent response on both JAK1 inhibitors and corroborated the advantages of upadacitinib. Still, our study has certain limitations. Firstly, three RCTs of upadacitinib had no 12-week data, and three RCTs of abrocitinib had no 16-week data. Terefore, we chose to analyse the results at EOT time rather than a specifc same time period. Because both 12-week and 16-week RCTs showed signifcant efects in their endpoint outcomes, and the diferences in our analysis were also signifcant, thus the heterogeneity led by diferent timing was acceptable without compromising our conclusion. Besides, adverse events with complicated causes demonstrated negative signifcant differences in JAK1 inhibitors, but it was difcult to explain the mechanism and provide recommendations. Close monitoring is required for high dose JAK1 inhibitors treatment, and low dosage regimen might be an alternative in case of adverse events. Moreover, diferent sample sizes between phase two and three RCTs might afect the veracity potentially.

Conclusion
JAK1 inhibitors demonstrate promising efcacy in AD with rapid response and dose-dependent response and signifcantly higher risks of acne and headache. Based on existing data, oral 30 mg upadacitinib QD has better outcome than oral 200 mg abrocitinib QD and is a recommended dosage regimen for moderate-to-severe AD patients. Oral 15 mg upadacitinib QD might be an alternative dosage regimen in case of treatment-emergent adverse events.

JAK:
Janus kinase AD: Atopic dermatitis TCS: Topical corticosteroids TYK: Tyrosine kinase PRISMA: Preferred Reporting Items for Systematic Reviews and Meta-Analyses IGA: Investigator's Global Assessment EASI: Eczema Area and Severity Index BSA: Body surface area SCORAD: Scoring atopic dermatitis NRS: Numerical rating scale DLQI: Dermatology life quality index POEM: Patient-oriented eczema measure CROBAT: Cochrane risk of bias assessment tool GRADE: Grading of Recommendations, Assessment, Development, and Evaluation RCTs: Randomized controlled trials WMD: Weighted mean diference SMD: Standard mean diference CIs: Confdence intervals OR: Odds ratio EOT: End of treatment.

Data Availability
All data generated or analysed during this study are extracted from RCTs searched from online databases (Medline, Embase, Web of Science, Cochrane database, and https://clinicaltrial.gov).

Ethical Approval
Ethical approval was not required, given that analyses were conducted on deidentifed, secondary data derived from published studies. Our review followed the guidelines of Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA), and the protocol was registered in PROSPERO (registration number CRD42021244435). Included studies must be in accordance with the Declaration of Helsinki and International Council for Harmonization Good Clinical Practice Guidelines and approved by respective ethics committees. Written informed consent of patients was also required.

Disclosure
Ying Zhang is the frst author of the manuscript.

Conflicts of Interest
Te authors declare that they have no conficts of interest.

Authors' Contributions
BL was in charge of the main idea and was the guarantor of integrity of the entire clinical study; RL and YZ were in charge of the study concepts, design, manuscript preparation, and editing; RL and PH searched databases independently. YZ and SR screened the titles and abstracts, articles meeting inclusion criteria; YZ and PH independently extracted data from eligible articles and conducted data analysis. YZ and RL independently assessed the quality of included studies; PH and SR were in charge of language polishing and grammar revision. Figure S1: risk of bias summary. Table S1: PRISMA checklist. Table S2: search strategy in Pubmed. Table S3: assessment standards in RCTs. Table S4: assessment tools in metaanalysis. Table S5: assessment result of CROBAT. Table  S6: result of secondary outcomes.