Модели на основе оптической когерентной томографии выявляют скрытое воспаление при атопическом дерматите

TOPLINE Composite non-invasive biomarker models, especially those based on optical coherence tomography (OCT), accurately predicted atopic dermatitis (AD) severity and distinguished healthy skin from clinically non-lesional AD skin with high accuracy, suggesting sensitivity to subclinical inflammation that traditional clinical scoring methods may overlook. METHODOLOGY - Researchers conducted a cross-sectional observational study involving 80 participants aged 11-60 years, including healthy control individuals (n = 20) and individuals with at least one active AD lesion on cubital fossae (n = 60) diagnosed according to the UK Working Party diagnostic criteria. - A total of 32 biomarkers spanning invasive to non-invasive modalities were collected from lesional and non-lesional skin sites at opposing cubital fossae; these biomarkers included structural markers via OCT, biophysical markers such as transepidermal water loss (TEWL), molecular markers via Fourier transform infrared (FTIR) spectroscopy, and metabolite markers from skin cells and blood. - Clinical severity was assessed using both local and global Eczema Area and Severity Index (EASI) scores, along with patient-reported outcomes via Patient-Oriented Eczema Measure (POEM), with measurements conducted in a climate-controlled room between December 2020 and July 2022. - Lasso-regularised generalised linear modelling with nested 5 × 5-fold participant-level grouped cross-validation was utilised to build predictive models for AD severity and to classify healthy vs non-lesional AD skin. - Performance metrics included the correlation coefficient (r) and root mean square error for severity prediction models and area under the receiver operating characteristic curve (AUC) for classification models distinguishing healthy from non-lesional AD skin. TAKEAWAY - A multivariable model using only non-invasive OCT-derived biomarkers predicted local AD severity (r, 0.82), global EASI (r, 0.95), and POEM (r, 0.76) when body surface area information was included. - A model using four key biomarkers distinguished healthy from clinically non-lesional AD skin (AUC, 0.94; accuracy, 88.0%), indicating sensitivity to subclinical inflammation. - Among individual non-invasive biomarkers, OCT epidermal thickness (r, 0.78), TEWL (r, 0.75), and the dermal attenuation coefficient (r, -0.72) showed the highest correlations with local EASI scores (P < .001 for all). - OCT-derived epidermal thickness demonstrated the strongest individual classification performance between healthy and non-lesional skin (AUC, 0.93), followed by the FTIR-derived amide ratio (AUC, 0.87), carboxylate signal (AUC, 0.86), and blood vessel orientation (AUC, 0.82). IN PRACTICE "This study demonstrates that composite models built from non-invasive biomarkers, particularly those derived from OCT, can accurately and objectively predict AD severity and, importantly, differentiate seemingly non-lesional skin with subclinical disease from healthy skin. The outputs form the basis of a simple-to-use bedside tool which could provide an immediate 'severity score' for AD management and clinical trials, with the potential to guide individualized treatment decisions and assess disease resolution beyond the visible surface," the authors of the study wrote. SOURCE The study was led by Robert A. Byers, Sheffield Dermatology Research, School of Medicine & Population Health, The University of Sheffield Medical School, Sheffield, England. It was published online on August 20, 2026, in the British Journal of Dermatology. LIMITATIONS The study cohort was relatively small and came from a single centre. The cohort was mostly White, which may have limited generalisability. Washout periods were shorter for participants with severe disease, so residual treatment effects may have biased biomarker-severity relationships in that group. Some biomarker groups were not included because of technical limits. DISCLOSURES This study was funded by LEO Pharma A/S. Some authors reported having prior collaboration with LEO Pharma. Four authors reported being employees and shareholders of LEO Pharma. Full disclosures are noted in the original article. This article was created using several editorial tools, including AI, as part of the process. Human editors reviewed this content before publication.