Hair Loss Profiles and Ritlecitinib Efficacy in Patients with Alopecia Areata: Post Hoc Analysis of the ALLEGRO Phase 2b/3 Study

The design and primary results of the international, randomized, double-blind, placebo-controlled, combined dose-ranging pivotal ALLEGRO phase 2b/3 trial (NCT03732807) have been previously described in detail [25]. Briefly, patients were randomized to receive ritlecitinib 50 mg or 30 mg once daily (QD) with or without a 200-mg QD loading dose for the initial 4 weeks, 10 mg QD, or placebo for 24 weeks. Randomization was stratified by severity (target: ≈ 40% of patients with AT/AU) and age (target: ≈ 15% of adolescents aged 12–17 years in each group).

Patients in ALLEGRO-2b/3 were aged ≥ 12 years with a diagnosis of AA and ≥ 50% scalp hair loss, including patients with AT or AU, and a current AA episode duration of 6 months to 10 years. Patients with other causes of alopecia, previous use of any JAK inhibitor, or clinically significant depression were excluded. This post hoc analysis included patients who received ritlecitinib 200/50, 200/30, 50, or 30 mg or placebo. The 10-mg dose group was included in the study for pharmacokinetic and safety dose-ranging purposes and was thus excluded from this analysis.

Baseline categorial variables, which were used as indicators for the LCA analysis described next, were based on extent of scalp, eyebrow, eyelash, and body hair loss and included extent of scalp hair loss as measured by Severity of Alopecia Tool (SALT) score (clinician reported); extent of eyebrow hair loss as measured through the Eyebrow Assessment (EBA) score (clinician reported); extent of eyelash hair loss as measured through the Eyelash Assessment (ELA) score (clinician reported); and extent of body hair loss as measured through the Alopecia Areata Patient Priorities Outcome (AAPPO) body assessment (patient reported). SALT assesses the amount of scalp hair loss with scores ranging from 0 (no scalp hair) to 100 (complete scalp hair loss) (Supplementary Table 2), and patients were categorized by baseline SALT score of < 80, ≥ 80 to < 100, or 100. EBA and ELA have 4-point scales ranging from 0 (none or no eyebrows/eyelashes) to 3 (normal eyebrows/eyelashes) (Supplementary Table 2), and patients were categorized by EBA or ELA score of 0 (none), 1–2 (minimal/moderate), or 3 (normal). The AAPPO assesses AA-related hair loss, emotional symptoms, and activity limitations. The AAPPO hair loss item subscale score is reported here. Patients describe the current amount of hair loss in different body areas (scalp, eyebrows, eyelashes, and body) using a 5-point response scale ranging from 0 (“no hair loss”) to 4 (“complete” hair loss), and improvements with a score of ≥ 2 from baseline are reported (Supplementary Table 2); patients were categorized by their responses for body hair loss: 0–1 (no or a little hair loss), 2–3 (moderate or a great deal of hair loss), or 4 (complete hair loss).

Patient-reported outcomes were assessed using the Patient Global Impression of Change (PGI-C) and Patient Satisfaction with Hair Growth (P-Sat) at Week 24. PGI-C is a single-item scale on which patients rate the improvement or worsening of AA compared with the start of the study, using a scale of 7 responses ranging from “greatly improved” to “greatly worsened” (Supplementary Table 2). PGI-C response was defined as “moderately improved” or “greatly improved.” P-Sat assesses patient satisfaction with hair regrowth since the start of the study in 3 domains: amount, quality, and overall (Supplementary Table 2). P-Sat response was defined as “moderately” or “very” satisfied for each domain.

LCA was used to identify hair loss profiles based on aforementioned scalp, eyebrow, eyelash, and body hair loss indicators at baseline. Latent class models with varying numbers of classes were estimated, and model selection was based on consideration of several criteria, including clinical interpretability; model fitness (log-likelihood, Akaike information criterion [AIC], Bayesian information criterion [BIC]), in which consistent AIC and the sample size–adjusted BIC were derived from AIC and BIC, respectively, by modifying the penalizing part of the metric (smaller values of AIC, BIC, consistent AIC, and sample size–adjusted BIC indicate better fit in the model); classification diagnostics (relative entropy); and smallest average latent class posterior probability, which assessed the average probability of each class model accurately predicting class membership for individuals.

After selection of the final LCA model, LCA class was assigned to each patient and included as a key covariate for the logistic regression analysis. Two logistic regression models evaluated the likelihood of achieving PGI-C score and P-Sat response for ritlecitinib (50 and 30 mg, with or without a 4-week 200-mg QD loading dose) vs placebo. An all-variable model considered all covariates listed above to adjust for the variables of interest. A stepwise model was used as a sensitivity analysis, in which nonsignificant covariates were excluded to increase precision for the independent variables of interest. Covariates included age (continuous), sex (male vs female), body mass index (continuous), prior pharmacological treatment for AA (yes or no), active hair shedding (yes or no), treatment arm (ritlecitinib 30 mg or 50 mg vs placebo), ritlecitinib 200-mg daily loading dose for 4 weeks (yes or no), and LCA class membership. Odds ratios were calculated for PGI-C response and P-Sat response for ritlecitinib 30 mg and 50 mg vs placebo. All analyses were implemented using R software: “glm” function of the “stats” package for logistic regressions and “poLCA” function of the “poLCA” package for LCA.

A total of 655 patients were randomized to and 653 were treated with ritlecitinib 200/50 mg, 200/30 mg, 50 mg, 30 mg, or placebo. Of the 655 randomized patients, mean age ranged from 32.4 to 34.5 years, 54.6% to 65.6% of patients were women, and 60.8% to 71.8% were white (Table 1). Of the 653 treated patients, 13.7% to 23.1% had baseline SALT scores of 50 to < 80, 30.8% to 40.5% had baseline SALT scores of ≥ 80 to < 100, and 45.8% to 46.5% had baseline SALT score of 100. Across the treatment groups at baseline, 44.3% to 50.4% of patients had no eyebrow hair and 15.2% to 18.5% had normal eyebrow hair, 37.2% to 43.1% had no eyelash hair and 22.1% to 26.9% had normal eyelash hair, and 13.0% to 16.2% reported no body hair loss, while 34.6% to 40.5% reported complete body hair loss.

A summary of the performance results (goodness-of-fit statistics and diagnostic statistics) for two- to nine-class latent class models is shown in Supplementary Table 3, showing that the three-, four-, and five-class models had the lowest BIC values. The distribution of variables used as indicators (extent of baseline scalp, eyebrow, eyelash, and body hair loss) within the classes identified by LCA for the three-, four-, and five-class models are shown in Supplementary Fig. 1. The five-class model identified distinct classes of patients with AT and AU (based on a baseline SALT score of 100) and was selected as the optimal model. In “class 4” of the five-class model, 100% of patients had a SALT score of 100 (complete scalp hair loss) and varying degrees of eyebrow, eyelash, and body hair loss, indicating that most of these patients had AT. In “class 5” of the five-class model, 91.9% of patients had a SALT score of 100 (complete scalp hair loss), 96.5% had no eyebrow hair, 89.4% had no eyelash hair, and 99.5% had no body hair, indicating that most of these patients had AU.

Based on baseline scalp, eyebrow, eyelash, and body hair loss categories, the classes from the five-class LCA model can be further described as (Fig. 1):

Among patients receiving ritlecitinib 30 mg and higher, 52.0% to 68.2%, 45.2% to 64.3%, 28.6% to 72.7%, 37.5% to 50.0%, and 15.8% to 37.8% of patients in classes 1, 2, 3, 4, and 5 achieved PGI-C response at Week 24, respectively, compared with 20.0%, 13.9%, 9.1%, 0.0%, and 2.2% receiving placebo (Table 2).

Adjusting for latent class membership, patients receiving ritlecitinib 30 or 50 mg (with or without a 200-mg daily loading dose for 4 weeks) were significantly more likely to achieve PGI-C response at Week 24 compared with those receiving placebo (Fig. 2). Adjusting for latent class membership, patients receiving ritlecitinib 30 or 50 mg were also significantly more likely to be satisfied with amount of, quality of, and overall hair regrowth based on P-Sat response at Week 24 compared with those receiving placebo (Fig. 3).