Non-Invasive Biomarker Models for Objective Severity Assessment and Detection of Subclinical Inflammation in Non-Lesional Atopic Dermatitis
Robert A Byers, Paul V Andrew, Sura Sahib, Oludolapo Katibi, Kirsty Brown, Anna Poyner, John Chittock, Samuel Williams, Laura Johnson, Fiona Wright, Britta C Martel, Petra Arlert, Ole E Sørensen, Jacob P Thyssen, Stephen J Matcher, Michael J Cork, Simon G DanbyAbstract
Background
Atopic dermatitis (AD) is a chronic, fluctuating inflammatory skin disease in which severity assessment relies on subjective clinical scoring. However, these approaches do not capture subclinical inflammation that may precede disease worsening or follow disease resolution. Furthermore, comprehensive research typically requires invasive methods such as skin biopsy, limiting routine use.
Objectives
This study developed and evaluated non-invasive, multimodal biomarker models to objectively assess AD severity and detect subclinical disease activity.
Methods
We conducted a cross-sectional observational study (NCT04295824) involving 80 participants aged 11–60, including healthy controls and individuals with mild to severe AD. Clinical severity and patient-reported outcomes were assessed, and 32 biomarkers spanning invasive to non-invasive modalities were collected from lesional and non-lesional skin. These included structural (e.g. optical coherence tomography), biophysical (e.g. transepidermal water loss), molecular (e.g. Fourier-transform infrared spectroscopy) and metabolite (biomarkers measured in skin cells and blood) markers. Lasso regression was used to build predictive models of AD severity and classify subclinical disease in clinically non-lesional skin
Results
A multivariable model comprised of non-invasive OCT imaging derived biomarkers predicted local AD severity outcomes with good accuracy (r=0.82). Global severity and patient reported outcomes could also be predicted with high accuracy (r=0.95,0.76 respectively), providing information regarding the extent of AD was provided to the model. Multimodal classification models distinguished healthy from clinically non-lesional AD skin with excellent performance (AUC = 0.94, 95% CI: 0.88-0.98), suggesting sensitivity to subclinical inflammation.
Conclusions
A comprehensive, patient-centered skin assessment approach can reliably determine AD severity and detect subclinical activity that traditional scoring methods may overlook. Robust OCT-derived metrics highlight the potential of integrating non-invasive, objective tools into clinical workflows to complement standard evaluations and improve treatment strategies. These methods also offer a more objective assessment of disease activity in clinical trials and practice.