Alopecia areata (AA) is the most common autoimmune cause of non-scarring hair loss, with a lifetime incidence of approximately 2%. Hair loss can range from small patches (patchy type) to complete loss of scalp (alopecia totalis) or body hair (alopecia universalis). Despite being non-scarring, AA often follows a chronic course, and spontaneous recovery in alopecia totalis occurs in only 16% of cases without treatment. Moreover, severe AA often requires long-term treatment, with frequent relapses during dose tapering.

Although the exact mechanisms are not fully understood, growing evidence points to a central role for immune-mediated pathways in its pathogenesis. Specifically, dysregulated cytotoxic T cells and natural killer (NK) cells appear to mediate an autoimmune attack on hair follicles, driven by an IFN-γ–dominant response. This leads to pronounced perifollicular inflammation, which disrupts the hair growth cycle and impairs follicle function without causing permanent damage or scarring (1). Despite this Th1-dominant profile, AA shows a notable association with atopy, particularly atopic dermatitis, which is the most common comorbidity. Interestingly, treatment responses also differ by atopic status: only patients with elevated IgE levels tend to respond well to dupilumab, with a SALT75 response rate of 50% after 72 weeks, compared to just 8% in those with low IgE levels. This highlights the complex and heterogeneous inflammatory signaling involved in AA pathogenesis (2, 3).

A key initiating factor in the autoimmune attack on hair follicles in AA is the collapse of hair follicle immune privilege (IP)—a local protective mechanism that normally shields follicles from immune recognition (4). In AA, this barrier breaks down, allowing immune cells to infiltrate and target the follicle. This infiltration is largely orchestrated by chemokines, small signaling proteins that guide immune cells to specific tissues, including the skin (5). With approximately 50 chemokines and 18 receptors identified, the chemokine system exhibits overlapping functions, making it a highly redundant and complex signaling network (6).

In AA, chemokines such as CXCL9, CXCL10, and CXCL11 are upregulated in lesional follicles under the influence of IFN-γ via the JAK–STAT pathway. These chemokines act through their common receptor CXCR3 to orchestrate directed recruitment of the cytotoxic Th1-type lymphocytes into the perifollicular region, amplifying local cytotoxic inflammation (2, 7). In murine AA models, pharmacological blockade of CXCR3 prevents disease onset, underscoring the pathogenic role of this axis (8).

Chemokines therefore form pivotal bridges between the inflammatory cytokine milieu (e.g. IFN-γ) and recruitment of effector and antigen-presenting cells to the hair follicle. They disrupt local immune privilege and initiate a feed-forward cycle of cytotoxic T-cell infiltration and follicular damage—providing mechanistic insight into how immune responses translate into clinical hair loss in AA.

Due to their rapid responsiveness to immune changes and higher circulating levels compared to cytokines, chemokines have emerged as valuable biomarkers in autoimmune diseases like AA. Their short half-life enables real-time monitoring of immune activity; however, this also makes them sensitive to transient fluctuations, such as those caused by infections, which can obscure the underlying chronic inflammation in autoimmune conditions (9, 10).

The large diversity of chemokines and their complex interactions pose significant challenges in understanding their precise role in AA. This scoping review aims to systematically map the current evidence regarding the involvement of chemokines in AA. By exploring their role in the disease pathogenesis, their potential as biomarkers, and the therapeutic implications, this review seeks to provide a comprehensive overview of existing research and identify gaps that warrant further investigation.

A systematic search was conducted in PubMed and Embase to identify all articles investigating chemokines in AA, with the aim of comparing levels between patients and healthy controls, as well as between lesional and non-lesional skin. All articles from inception to 15 November 2024 were screened for eligibility. The search strategy included the keywords “alopecia areata AND (chemokine OR chemokines)” in all fields. One article, published in February 2025, was added afterwards as it replaced an abstract that had been included during the initial search in November 2024, once the full text became available (11). Only human studies investigating chemokine levels in the blood and/or skin of AA patients were included, while animal models were excluded. Full-text articles, short manuscripts, letters and abstracts were all considered, and all languages were allowed. Duplicates were removed based on similar content and authorship. The following comparisons were extracted: comparison between AA and healthy controls and between lesional and non-lesional skin. Comparisons of non-lesional skin versus healthy skin were not taken into account. In vitro models were also excluded given their uncertain capacity to reflect the in vivo chemokine profile. The extracted data included the detected chemokine levels with their statistical significance, the number of patients, the method to measure chemokines (e.g., ELISA, RNA analysis,.). Meta-analysis was done in case at least 3 articles performed the same analysis at the protein or RNA level in comparable patient groups. The meta-analysis was carried out with Review Manager 5.4.1 (The Cochrane Collaboration, 2020) using an inverse variant random effects model with the standardized mean difference as an effect measure. The standardized mean difference was chosen instead of the mean difference due to substantial variability in baseline chemokine values among healthy controls, reflecting differences in laboratory kits and evaluation techniques. The mean chemokine concentrations, standard deviation, and number of patients were extracted from each publication. In case only the sample size, range, median, and/or interquartile range were mentioned, the mean was calculated by the formula of Luo et al., 2018 and the standard deviation was calculated based on Wan et al., 2014 (4, 5). If studies displayed the results in graphs without providing the exact values, data were extracted with GIMP 2.10.30 (GNU image manipulation program) using the methodology published by Van der Mierden et al., 2020. If mean values and standard deviations were not reported and no graphs were available, the data were calculated from the median and range using the formula by Hozo et al. (2005) (12).

A total of 309 records were identified through database searching, of which 46 met the inclusion criteria. The selection process is illustrated in Figure 1 .

Among the 46 included studies, 24 reported chemokine levels in blood (protein, RNA, or DNA), 21 in skin (protein or RNA), and one study reported on both (13).

Most results on chemokines in the circulation were found for CCL17 (n=7), CXCL9 (n=7), CXCL8 (n=6), CXCL10 (n=6), CCL5 (n=6), CCL4 (n=6), CCL7 (n=5), CXCL1 (n=5), and CCL2 (n=5) followed by CCL13 (n=4), CCL3 (n=4), CCL11 (n=3), and CCL20 (n=3). The remaining chemokines were only reported by one or two sources.

Similar results were found for chemokine reports in the skin: most results were found for CXCL10 (n=16), CXCL9 (n=12), CCL5 (n=10), CCL13 (n=8), CCL18 (n=8), CXCL11 (n=6), CCL2 (n=5), CXCL8 (n=4), CCL22 (n=4), CCL17 (n=4), CCL26 (n=4) CXCL1 (n=3), CXCL13 (n=3), CCL19 (n=3), CX3CL1 (n=3),CCL8 (n=3), and CCL20 (n=3).

35 out of the 46 studies directly compared chemokine levels in AA patients to healthy controls. Figure 2 provides an overview of these articles. Chemokines are listed from top to bottom, which reflects the strength of evidence: chemokines with the most reports appear at the top, and those with limited data appear lower down. The remaining 11 articles focused on comparisons between lesional and non-lesional skin, investigated associations with disease severity (SALT score), or assessed changes in chemokine levels following treatment.

Our review identified strong Th1 signaling in AA, marked by elevated expression of CXCL9, CXCL10, CXCL11, and CCL5 both locally and systemically. This Th1 dominance is further supported by meta-analysis and aligns with the immune patterns seen in other IFN-γ-driven diseases like vitiligo (67).

In contrast to the findings in vitiligo, Th2 signaling was prominent as well, with consistent upregulation of CCL13, CCL17, CCL22, and CX3CL1. This Th2 signature aligns with the previously mentioned frequent co-occurrence of atopy and the observed positive response to dupilumab in atopic AA patients. Moreover, levels of CCL13, CCL17, and CCL22 significantly decreased following effective dupilumab treatment while strongly correlating with reductions in SALT scores. Though these findings were more pronounced in atopic AA patients with elevated IgE levels, non-atopic patients also showed some improvements in SALT scores and reduced Th2 chemokine levels following dupilumab, although almost complete hair regrowth was rarely observed in the latter group (29). Interestingly, children have a more Th2-skewed immunity compared to adults and preliminary evidence suggests that even children with low IgE levels can respond to dupilumab, although a long-term treatment seems required before efficacy is observed (+/- 1 year) (68). Such an approach could be particularly appealing given the less favorable side effect profiles of other biological treatments, such as JAK inhibitors. Unfortunately, although published chemokine profiles in adults support this working mechanism, data on chemokine expression in children with AA remain limited.

CCL20, the signature chemokine for Th17 recruitment, was found to be elevated in the serum of AA patients, yet not in the skin (22, 25). There is moderate evidence for elevated levels of chemokines that recruit dendritic cells, monocytes, neutrophils, eosinophils, and B lymphocytes in AA, although their expression appears lower than the strong Th1 and Th2 signals and they have been less extensively studied. A surprising finding is the marked elevation of the neutrophil-attracting CXCL8 in both serum and lesional skin of AA patients, even though neutrophils are not typically involved in AA pathology and there are no clinical signs of neutrophilic inflammation. One possible explanation for the elevated CXCL8 levels is secondary upregulation in response to pro-inflammatory cytokines like IL-1β, IFN-γ, and TNF-α (69). The absence of significant neutrophil infiltration, despite elevated CXCL8, may be due to the concurrent increase in CXCL12, which retains leukocytes in the bone marrow, suggesting that their opposing effects may dampen neutrophil recruitment to the skin. Notably, beyond their opposing roles in neutrophil trafficking, CXCL8 and CXCL12 together may synergistically recruit T cells and monocytes (62). Another unexpected finding is the elevated CCL18 levels in AA, despite its typical downregulation by IFN-γ. These findings underscore the complex immune profile observed in AA, reflecting a multifaceted inflammatory environment.

8 studies examined the association between serum chemokine levels and disease severity, with variable results. Serum levels of CCL13, CCL7, CCL17, CCL27, and CXCL8 were associated with disease severity, while other chemokines were not (11, 16, 17, 19, 31, 39, 43, 55). Notably, these findings were based on single time point assessments and only one study reported dynamic changes in chemokine levels during disease progression - showing that CXCL9 and CCL5 increased during exacerbations and decreased during remission (19). In the skin, only 2 studies explored the relationship between chemokine expression and SALT scores, both reporting no correlation. However, following effective treatment with dupilumab, nearly all chemokines correlated with changes in SALT (29, 31).

Given their early involvement in disease pathogenesis, chemokines have also been studied as potential therapeutic targets. In murine models, chemokine-directed interventions have yielded encouraging results, including CXCL12-neutralizing antibodies, blockade of the CX3CR1/CX3CL1 fractalkine axis, CCR5 inhibition using maraviroc, and anti-CXCR3 antibodies (8, 28, 35, 70, 71).

Despite the overall consistency of results, some methodological limitations must be acknowledged. Regarding study quality, of the 46 included articles, 5 are abstracts and 3 are letters. Most of these focus on a limited number of chemokines, and in nearly all cases, their findings are supported by multiple full-length, peer-reviewed studies. Exceptions that warrant more cautious interpretation include CXCL12 (2 out of 3 sources are non–full articles), CCL23 (1 out of 2) and CXCL1 (1 out of 3). However, since these chemokines were reported by only a few studies, they received limited attention in our manuscript and did not influence the overall conclusions. A more detailed assessment of the potential impact of the abstracts and letters is provided in the Supplementary Material .

In the meta-analysis, 3 out of 10 articles were letters. As presented in the Supplementary Material , sensitivity analysis showed no significant changes, except for CXCL10. However, this was mainly due to the study of Wang et al. (2023), which shows contrasting results compared to other studies for several chemokines (including CXCL9, CCL5, CCL2, and CCL3). This study included COVID-induced alopecia areata which might explain differences in chemokine production. When this study was not taken into account CXCL10 was also significant without reports published as letters or short reports.

Regarding data heterogeneity of the meta-analysis, all ten studies used serum-based immunoassays and compared AA patients to healthy controls, yet considerable variability was present (I² = 83–97%). We explore potential sources of this below.

In terms of methodology, Geng shows the greatest variation, focusing on a pediatric population, while the other studies examined adults. Several studies explicitly excluded patients who had received systemic or topical treatments prior to blood sampling (e.g., Bilgic, Kuwano, Geng), whereas others did not report treatment status or applied no clear exclusion criteria (e.g., Maouia, Barahmani, Elzawhary). Some studies excluded individuals with comorbid autoimmune or inflammatory conditions, while others did not specify such exclusions. One study (Wang, 2023) included only new-onset AA cases with a suspected post-COVID vaccination etiology.

In terms of laboratory methods, all studies used some form of immunoassay; however, there was technical variation. 5 studies used classical ELISA (Bilgic, Maouia, Zainodini, Waśkiel-Burnat, Elzawahry), 4 used multiplex bead-based assays (Kuwano, Wang (2023), Geng, Barahmani), and 1 (Song) used an electrochemiluminescence (ECL) immunoassay.

Although these differences contribute to a degree of clinical heterogeneity, studies showed consistent directional effects in serum chemokine levels, suggesting a robust underlying biological signal. To correct for the high level of heterogeneity across studies, we chose a random-effects model over a fixed-effects model. This approach accounts for both within- and between-study variation and provides a more reliable estimate of the overall effect. Nevertheless, future meta-analyses may benefit from stratified analyses based on treatment status, disease chronicity, or age group, should more homogeneous datasets become available.

In conclusion, this review and meta-analysis clarifies the complex chemokine profile of AA. While a dominant Th1 signal is evident, Th2 and other immune pathways are also clearly involved. This multifaceted immune landscape may contribute to AA’s therapeutic resistance and highlights the potential of broad immunomodulatory approaches—including key regulators such as the Aryl Hydrocarbon Receptor (AhR) (72).

Given their elevated serum levels in AA patients compared to healthy controls, several chemokines identified in this review show potential as biomarkers for disease activity. However, longitudinal studies remain scarce. Future research should focus on tracking chemokine dynamics over time to develop predictive models that signal upcoming flares or remission—an especially valuable tool in AA, where clear clinical warning signs are lacking and disease unpredictability places a high psychological burden on patients.

Our findings further support the role of the Th2 axis in a subset of patients. IgE levels and Th2-related chemokines have been demonstrated to predict the response of alopecia areata to dupilumab (29). At our department, we already screen for elevated IgE levels in atopic AA patients to guide the use of dupilumab, and these results reinforce the rationale for such targeted approaches. Finally, chemokine profiling may serve as a powerful tool for large-scale immune stratification, potentially informing both prognosis and personalized therapy development.