Why was the study undertaken?This systematic review was undertaken to identify clinically significant biomarkers that may predict how atopic dermatitis (AD) patients respond to systemic therapies and subpopulations who are more likely to develop treatment‐related adverse events. As systemic treatment options expand, identifying reliable predictive biomarkers is essential to inform precision medicine and to optimize patient outcomes.What does the study add?This systematic review synthesizes currently available evidence on biomarkers associated with treatment response and adverse events in AD. It demonstrates that commonly investigated biomarkers, such as IgE, total eosinophil count and LDH, exhibit inconsistent predictive value, with most data pertaining to dupilumab and its related ocular adverse effects, particularly conjunctivitis.What are the implications of this study for disease understanding and/or clinical care?Our findings underscore the complexity of identifying predictive biomarkers in AD due to its heterogeneous nature.
The lack of consistency across the included studies emphasizes the need for standardized methodologies to provide robust, validated biomarker data to facilitate personalized treatment selection and risk stratification.
INTRODUCTION
Atopic dermatitis (AD) affects around 20% of children and up to 10% of adults in high‐income countries.1 AD is associated with psychological distress and high economic burden, demonstrating the highest disease burden among skin disorders as measured by disability‐adjusted life years (DALYs).1, 2 Clinically, AD is complex and heterogeneous, with genetic, environmental and immune dysregulation factors involved in its pathogenesis.3 Despite varying phenotypes, therapeutic strategies are based on a ‘one‐size‐fits‐all’ management approach, rarely accounting for individual pathophysiological differences.4
Patient stratification using biomarkers has become clinically pertinent with the increasing availability of targeted systemic therapies for AD. Biomarkers, including serum IgE levels, lactate dehydrogenase (LDH), serum thymus and activation‐regulated chemokine (TARC), periostin and IL‐22, have been identified as potential predictors of disease response to therapy and/or treatment‐related adverse events (AEs).5
The BIOmarkers in Atopic Dermatitis and Psoriasis (BIOMAP) is a European consortium that aims to improve precision medicine and establish predictive biomarkers associated with AD disease trajectories, response to systemic therapies, treatment‐related AEs and the development of comorbidities.6 This systematic review aims to identify biomarkers that predict response to systemic therapies and treatment‐related AEs in patients with moderate‐to‐severe AD.
METHODS
This systematic review was registered in the International Prospective Register of Systematic Reviews (PROSPERO) (Response to systemic therapies: [CRD42020193265]; Adverse reactions: [CRD42020193274]) and Preferred Reporting Items for Systematic Reviews and Meta‐Analyses (PRISMA) guidance was followed throughout this review.7, 8, 9
The initial search was run on July 31, 2020 without date limits. Electronic searches were conducted in MEDLINE, Embase, CENTRAL, LILACS and the GREAT database supplemented by manually searching key reviews and relevant background literature. Searches in Embase and MEDLINE were re‐run on 23 September 2024, with date limits set to identify new evidence published since the original search.
Searches were quality‐assured and peer‐reviewed by a second information specialist using a process based on the Peer Review of Electronic Search Strategies (PRESS) checklist.10 Methods were aligned with those previously applied in related BIOMAP systematic reviews to ensure consistency across studies.6 The full search strategy is available in the Appendix S1.
Inclusion and exclusion criteria
Co‐authors independently screened all records identified in the initial search by title and abstract in EndNote and with the systematic review web platform Rayyan. Full texts of potentially relevant articles were reviewed for eligibility using protocols developed using the PICO (population, intervention, comparator and outcome) framework. Any discrepancies were resolved through discussion and by a senior author (CF).
Articles were included if they were randomized controlled trials (RCTs) or non‐randomized studies (prospective and retrospective cohort studies, case–control studies) with relevant outcome data for this review's primary outcomes.
Primary outcomes included clinician‐assessed and patient‐reported outcome measures, for example change in the Eczema Activity and Severity Index (EASI) or the Dermatology Life Quality Index (DLQI) over time, and incidence of treatment‐related adverse events. Case reports, case reviews, conference abstracts and non‐English studies were excluded. The restriction of non‐English studies in systematic reviews has been investigated previously, with no evidence of systematic bias associated with this procedure.11
Data extraction and risk of bias
Standardized data extraction forms were created in MS Word to collect key study information, including study design, country, study setting, biomarker(s) assessed and risk of bias (RoB) assessments. Data extraction forms were piloted before use.
Co‐authors extracted data from the initial search and the second search. For dichotomous outcomes, odds ratios, risk ratios, rate ratios or hazard ratios and their 95% confidence intervals were extracted for the independent effect of each biomarker. Sensitivity, specificity, area under the curve, negative and positive predictive values, and correlation coefficients for specific biomarkers were recorded when available.
Co‐authors conducted quality and RoB assessments for all outcomes included using the Quality in Prognostic Studies (QUIPS) checklist for prognostic studies.12 Due to significant heterogeneity, meta‐analysis was not performed.
RESULTS
The literature search identified 27,766 records. Following de‐duplication and conference abstract removal, 16,143 records remained. For therapeutic response, 19 additional records were manually identified, resulting in 16,162 records screened (Figure 1).
Of these, 15,892 were excluded based on predefined inclusion criteria (Appendix S2a). Twenty‐eight papers13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38, 39, 40 were included in the qualitative synthesis, reporting results from 15 prospective cohorts, 9 retrospective cohorts and 4 RCTs. Treatments investigated were dupilumab (n = 19), omalizumab (n = 3), ciclosporin (n = 2), mycophenolic acid (n = 1), tralokinumab (n = 1), upadacitinib (n = 1) and apremilast (n = 1).
Fifty‐two candidate biomarkers were investigated. The results are summarized in Table 1 (Appendices S4a and S5a).
For AEs, 6 additional records were identified manually. In total, 16,149 records were screened, and 16,092 were excluded based on our predefined inclusion criteria (Figure 2; Appendix S2b). Eleven papers13, 35, 36, 41, 42, 43, 44, 45, 46, 47, 48 were included in the qualitative synthesis, reporting results from six prospective cohorts, three retrospective cohorts and 5 RCTs.
All studies investigated associations between AEs and dupilumab. Six biomarkers were investigated. The results are summarized in Table 2 (Appendices S4b and S5b).
Full RoB assessments are included in Appendices S3a and S3b. Most studies exhibited low to moderate RoB across QUIPS domains; however, none achieved low RoB in all six domains. The primary source of bias was inadequate adjustment for confounders, which was often limited to single variables such as age or ethnicity.
Reporting of study participation was inconsistent, with two studies lacking recruitment details.26, 46 Moderate RoB studies generally reported participation adequately but omitted recruitment information. Attrition bias was also common. Studies with higher RoB showed greater attrition, rarely reported discontinuation reasons, and excluded missing data.
High RoB in the prognostic factor domain reflected insufficient reporting of biomarker measurements, whereas outcome measurement was largely scored low. Statistical analyses were generally appropriate.
Response to treatment
Nineteen papers reported results from 18 cohorts investigating the association between biomarkers and the response to dupilumab (Table 1).13, 14, 17, 18, 19, 20, 21, 24, 25, 27, 29, 30, 31, 34, 35, 36, 38, 40 Ten studies evaluated baseline total eosinophil count (TEC), and results were inconsistent.13, 14, 17, 18, 21, 24, 25, 29, 35, 36 Six studies reported a lack of association between baseline TEC and response to treatment. Three studies reported significant but conflicting associations between high baseline TEC and clinical improvement measured by EASI, DLQI, POEM and PP‐NRS. Ferruci et al. found that higher baseline TEC was associated with poorer EASI‐75 response which was significant at Week 4 but not significant at Week 16.18 Jang et al. observed no association with EASI at Week 16 or Week 52, but reported a significant positive association between higher baseline TEC and POEM improvement.24, 25 However, Beck et al.14 reported a weak positive correlation between higher baseline TEC and percentage improvement in EASI.
Eight studies assessed baseline total IgE;14, 17, 18, 19, 21, 25, 29, 30 two reported significant associations. Beck et al.14 observed weak positive correlations between baseline IgE and clinical improvement at 4 and 12 weeks. Whereas Olesen et al.30 linked elevated IgE (>2320 KU/L) with absence of EASI‐75 response at 4 weeks.
Baseline LDH levels were investigated in five studies;24, 25, 29, 30, 37 two studies reported a negative association between high baseline LDH and clinical improvement measured by EASI, POEM and Peak Pruritis Numerical Rating Scale (PP‐NRS) over time.25, 30 Seven studies evaluated baseline TARC,14, 27, 29, 35, 38, 40 with three reporting negative associations between high baseline TARC and clinical improvement measured by EASI, POEM, SCORAD (SCORing Atopic Dermatitis) and PP‐NRS.29, 38, 40
Two further studies reported significant findings. One reported an association between the number of CD4+ cells/mm2 of lesional skin and achievement of EASI‐75 at 16 weeks,20 One other reported a significant correlation between reductions in SCORAD Pruritis‐VAS (Visual Analog Scale) and serum CCL13 levels.31 Two studies measured serum dupilumab levels at 16 weeks but reported diverging results.15, 36 Finally, three studies reported results from an extended biomarker panel.27, 29, 34 Negative correlations were, for instance, reported between CCL18, CCL27, soluble interleukin‐2 receptor (sIL‐2R) levels and POEM and IFNγ levels and complete or partial response. Positive correlations were reported between IL‐22, IL‐4 and IL‐3 levels and complete or partial response.
One RCT investigated the association between therapeutic response to tralokinumab at 12 weeks and 4 candidate biomarkers: elevated baseline IgE, dipeptidyl peptidase‐4 (DPP‐4), TARC and periostin (Tables S1 and S7).39 No significant associations between any biomarker and treatment response were noted.
One retrospective cohort study evaluated the association between the treatment response to upadacitinib and four biomarkers (Table S2): TEC, total IgE, TARC and LDH (Tables S2 and S8).22 Low baseline LDH was associated with treatment response at 12 weeks; no other associations were reported.
Three studies investigated the association between biomarkers and conventional systemic therapies (Table S3); two investigated ciclosporin (CyA)26, 28 and one mycophenolic acid (MPA).37 Only one study reported a significant finding; Kim et al.28 found higher baseline eosinophil‐derived neurotoxin levels predicted relapse after CyA withdrawal.
Three studies examined biomarker associations with omalizumab16, 23, 31 and one with apremilast33 (Table S4). Two omalizumab studies assessed baseline IgE. Sheinkopf et al.33 reported that IgE levels >700 IU/mL correlated with greater IGA improvement at 6 and 9 months.
Chan et al.16 found that patients with more positive baseline skin prick tests (SPTs) showed significantly greater objective‐SCORAD improvement at 24 weeks. Hotze et al.23 observed reduced odds of SCORAD‐25 response in omalizumab‐treated patients carrying filaggrin (FLG) loss‐of‐function mutation. Additionally, baseline eosinophilia ≥0.39 × 109/L was associated with smaller improvement in EASI at 12 weeks, significant only in the high‐dose group.
Adverse events
All studies included investigated the association between biomarkers and dupilumab‐related conjunctivitis or ocular surface disease (DAOSD) (Tables 2 and S5). Six studies investigated the association between AEs and serum total IgE.41, 43, 45, 46, 47, 48 Three studies focused on conjunctivitis, and three reported on DAOSD. Akinlade et al.41 observed that higher levels of baseline total IgE and antigen‐specific IgE were positively associated with the event rate of conjunctivitis during dupilumab treatment.
Nettis et al.46 found that the risk of conjunctivitis was especially associated with very high total IgE at baseline (above the 75th percentile of study population).
Touhouche et al.48 reported a significant association between elevated IgE (>1000 U/mL) and DAOSD. Nahum et al.45 observed higher baseline IgE in patients who developed DAOSD at a mean 10.5‐month follow‐up, although the association was non‐significant (unadjusted OR 1.04, 95% CI 0.99–1.09). Shim et al.47 reported that high IgE (>2600 IU/mL) increased DAOSD risk (adjusted OR 2.70, 95% CI 0.89–8.15) over 4.5 years of follow‐up.
Seven studies investigated the association between AEs and baseline TEC.13, 35, 41, 43, 44, 47, 48 Three evaluated conjunctivitis, three DAOSD and one any AEs. Akinlade et al.41 found that higher baseline TEC was positively associated with the event rate of conjunctivitis during 16 weeks of treatment across all cohorts. However, two other studies reported no significant correlations.13, 43 Limão et al.44 reported higher median baseline TEC in patients who developed DAOSD over a year of treatment than those who did not.
However, three other papers reported no associations.35, 47, 48
Two studies investigated the association between TARC and AEs.35, 41 Akinlade et al.41 reported that higher levels of baseline TARC were associated with a higher event rate of conjunctivitis. Spekhorst et al.35 found no association between baseline TARC and dupilumab‐induced AEs.
Two other biomarkers were investigated in separate studies.42, 48 Touhouche et al.48 found no association between baseline fractional exhaled nitric oxide (FeNO) and those who developed DAOSD. Chiricozzi et al.43 reported that tear fluid IL‐33 levels were associated with developing DAOSD by Week 16.
DISCUSSION
We performed a systematic review of the literature regarding candidate biomarkers that predict therapeutic response and AEs to systemic treatment in AD patients. Most studies investigated the association between biomarkers and dupilumab. Studies reporting associations between AEs and biomarkers were limited to DAOSD.
While the development of dermatological biomarkers lags behind that of other fields, such as oncology, emerging evidence indicates that certain biomarkers, including TARC, FLG and IL‐13, may hold predictive potential.49, 50
Our findings provide early indications that LDH might be useful as a predictive biomarker for therapeutic response to both dupilumab and upadacitinib.22, 25 Our review identified an association between high baseline LDH levels and poorer response to treatment. Prior research has shown LDH to be a reliable predictor of AD severity.51 The results reported by Jang et al. are particularly valuable for understanding the long‐term effects, as their study evaluated the association of LDH levels with treatment response over a 52‐week period. In contrast, the results presented by Olesen et al. did not support these findings.
However, their analysis was limited to a 4‐week observation period, which may account for the discrepancy.
Our review identified inconsistent associations for IgE and TEC, two biomarkers commonly available in clinical practice. Elevated baseline IgE levels appeared predictive for improved response to dupilumab14, 17, 19, 21, 25 but associations with response to tralokinumab were conflicting,39 despite both agents targeting the IL‐13 pathway. Similarly, TEC exhibited variable predictive relevance across studies.
These inconsistencies likely stem from heterogeneity in study design, including differences in biomarker thresholds and assessment intervals, limiting the ability to establish their long‐term predictive validity.
DAOSD is the most reported dupilumab‐induced effect during treatment and is reported in up to 50% of AD patients.52, 53, 54 Total IgE levels appear to be a promising biomarker for the development of DAOSD. The mechanism linking dupilumab and OSD is poorly understood. Inhibition of Th2 pathways and skewing towards Th1 and Th17 responses may lead to a reduction in conjunctival goblet cells, contributing to epithelial barrier dysfunction.55, 56, 57 Dupilumab also negatively affects mucin production, further impacting ocular health.58
Strength and limitations
We performed a comprehensive literature search, used standardized quality and RoB assessments and have summarized all published studies until September 2024.
Limitations of this systematic review are primarily based on the data itself. Conventional systemic therapies such as ciclosporin (CyA) and methotrexate (MTX) remain widely used in clinical practice, but available biomarker data were scarce. Most studies included in this review investigated dupilumab, while there is a lack of research on newly available treatments.
Only one study each was identified for tralokinumab and upadacitinib, while no eligible studies were found for baricitinib, abrocitinib or lebrikizumab. Overall, both established and emerging therapies appear to be underrepresented in current biomarker research for AD. The available research is of variable quality and utilizes varying methodologies, which impacts direct comparability and precluded meta‐analysis.
The only AE investigated in the studies included in this review was DAOSD. This narrow focus substantially limits the clinical utility of our review and highlights significant gaps in the literature regarding AEs associated with other treatments.
Many studies only reported crude results, without adjusting for confounding factors. Furthermore, the studies included in this review did not provide data on patient subgroups, such as paediatric versus adult cohorts, or on differences between patients with and without allergic comorbidities.59 This lack of subgroup analyses limits the generalisability of the findings and underscores the need for research to evaluate biomarker utility across diverse patient populations.
Future research
A key limitation of the studies included in this review is the use of arbitrary thresholds for the biomarkers investigated. Although work is already underway to address this issue, a clear consensus has yet to be established.60 In many studies, biomarker thresholds were determined based on cohort‐specific medians; however, as cohorts vary in size and characteristics, this approach lacks external validity. This limitation is particularly pertinent for the most frequently investigated biomarkers, LDH, IgE, TEC and TARC, for which the establishment of uniform thresholds is essential.
Achieving this, however, is complicated by the heterogeneity of study cohorts, particularly with respect to varying levels of allergic comorbidities. Similarly, outcome definitions were not consistent across the studies included in this review, thereby reducing the validity of the findings and hindering meaningful comparisons between studies. This issue has also been highlighted in a previous systematic review, which reported substantial inconsistency in outcome definitions among clinical trials in AD.61
Further research is also needed on novel treatments and older conventional therapies, like MTX. The TREatment for Atopic dermatitis Trial (TREAT) found that MTX conferred longer disease control for moderate‐to‐severe AD compared to CyA, while also being more cost‐effective. The same study cohort recently demonstrated an association between natural moisturizer factor (NMF) expression and MTX exposure, even after treatment cessation, potentially explaining the better disease control seen with this agent off therapy in the trial compared to CyA.62 In many jurisdictions, health organizations require patients to trial and fail conventional systemic treatments before accessing novel, targeted treatments.
As there is very limited biomarker research focusing on conventional systemic treatments, it is not currently possible to predetermine which individuals will fail to respond. If a reliable biomarker was available to predict poor response to conventional treatment, it would save time in the patient's journey to the most effective treatment for them.
Future research should prioritize large prospective cohort studies with embedded biomarker measurements. This is best done through finding international consensus on core clinical and molecular outcomes first, to facilitate direct comparisons between studies. Such an approach would facilitate greater reproducibility of existing findings and enable more robust cross‐comparison of emerging datasets.
CONCLUSIONS
In summary, while there is limited preliminary evidence indicating that LDH might serve as a predictive biomarker for treatment response to dupilumab and upadacitinib, the clinical reliability of IgE and TEC remains uncertain due to inconsistent findings across studies. Currently, no biomarker demonstrates sufficient reliability to support its use in guiding systemic therapeutic decision‐making in AD. Furthermore, potential associations between specific biomarker profiles and adverse treatment outcomes, such as DAOSD, are not yet well characterized.
Robust, prospective studies are required to validate and standardize biomarker use in AD before they can be reliably implemented in routine clinical practice.
AUTHOR CONTRIBUTIONS
Eva Hilger: data curation (equal); investigation (equal); project administration (lead); supervision (supporting); writing—original draft preparation (lead); writing—review and editing (lead). Kaitlyn Chan: data curation (equal); investigation (equal); project administration (supporting); writing—original draft preparation (equal); writing—review and editing (equal). Alphonsus Yip: data curation (equal); investigation (equal); writing—original draft preparation (equal); writing—review and editing (equal).
Xingyue Maria Wang: data curation (equal); investigation (equal); writing—original draft preparation (equal); writing—review and editing (equal). Conor Broderick: data curation (equal); investigation (equal); writing—original draft preparation (supporting); writing—review and editing (supporting). Bernd Arents: conceptualisation (equal); writing—review and editing (supporting).
Karin van Bart: data curation (equal); investigation (equal). Kilian Eyerich: conceptualisation (equal); funding acquisition (equal); writing—review and editing (supporting). Joseph Rastrick: conceptualisation (equal).
Stephan Weidinger: conceptualisation (equal); funding acquisition (equal); writing—review and editing (supporting). Carsten Flohr: conceptualisation (equal); funding acquisition (lead); supervision (lead); writing—review and editing (supporting).
FUNDING INFORMATION
This project received funding from the Innovative Medicines Initiative 2 Joint Undertaking (JU) under grant agreement No 821511 (BIOMAP). The JU receives support from the European Union’s Horizon 2020 research and innovation programme and EFPIA. This publication reflects only the authors’ views, and the JU is not responsible for any use that may be made of the information it contains.
The granting agencies covered costs and played no role in the manuscript preparation. [Correction added on 16 February 2026, after first online publication: The funding information has been updated.]
CONFLICT OF INTEREST STATEMENT
CF is chief investigator of the UK‐Irish Atopic eczema Systemic Therapy Register (A‐STAR; ISRCTN11210918) and a principal investigator in the European Union (EU) Horizon 2020‐funded BIOMAP Consortium (http://www.biomap‐imi.eu/). He also leads the EU Trans‐Foods consortium. He has received compensation from the British Journal of Dermatology (reviewer and section editor) and EuroGuiDerm (guidelines lead) as well as Almirall, Apogee, Bioderma, Incyte, LEO, Pfizer and Sanofi for advisory and educational activities.
KE has received grants and personal fees from AbbVie, Almirall, Apogee, BMS, Boehringer Ingelheim, Lilly, Leo, Janssen, Pfizer, Sitrryx, Moonlake, Novartis, UCB and Sanofi. SW has received institutional research grants from Sanofi Deutschland GmbH, LEO Pharma and Pfizer, has performed consultancies for AbbVie, Almirall, Astria, Boehringer, Eli Lilly, Galderma, GSK, LEO Pharma, Pfizer, Sanofi‐Genzyme and Regeneron, he has also lectured at educational events sponsored by AbbVie, Eli Lilly, LEO Pharma, Sanofi‐Genzyme and Regeneron, and is involved in performing clinical trials with many pharmaceutical industries that manufacture drugs used for the treatment of psoriasis and atopic dermatitis. JR is an employee of UCB and holds shares in UCB.
All other authors have no conflicts of interest to declare.
ETHICAL APPROVAL
This study is a systematic review of previously published literature and does not involve the collection of primary data from human participants or animals. Therefore, ethical approval was not required.
ETHICS STATEMENT
Not applicable.
SYSTEMATIC REVIEW REGISTRATION
PROSPERO CRD42020193265.
PROSPERO CRD42020193274.
Supporting information