Daphne S Bakker1, Stefan Nierkens2, Edward F Knol3, Barbara Giovannone2, Eveline M Delemarre2, Jorien van der Schaft4, Femke van Wijk2, Marjolein S de Bruin-Weller4, Julia Drylewicz2, Judith L Thijs4. 1. National Expertise Center for Atopic Dermatitis, Department of Dermatology and Allergology, University Medical Center Utrecht, Utrecht, The Netherlands; Center for Translational Immunology, University Medical Center Utrecht, Utrecht, The Netherlands. Electronic address: d.s.bakker-4@umcutrecht.nl. 2. Center for Translational Immunology, University Medical Center Utrecht, Utrecht, The Netherlands. 3. National Expertise Center for Atopic Dermatitis, Department of Dermatology and Allergology, University Medical Center Utrecht, Utrecht, The Netherlands; Center for Translational Immunology, University Medical Center Utrecht, Utrecht, The Netherlands. 4. National Expertise Center for Atopic Dermatitis, Department of Dermatology and Allergology, University Medical Center Utrecht, Utrecht, The Netherlands.
Abstract
BACKGROUND: Atopic dermatitis (AD) is a highly heterogeneous disease, both clinically and biologically, whereas patients are still being treated according to a "one-size-fits-all" approach. Stratification of patients into biomarker-based endotypes is important for future development of personalized therapies. OBJECTIVE: Our aim was to confirm previously defined serum biomarker-based patient clusters in a new cohort of patients with AD. METHODS: A panel of 143 biomarkers was measured by using Luminex technology in serum samples from 146 patients with severe AD (median Eczema Area and Severity Index = 28.3; interquartile range = 25.2-35.3). Principal components analysis followed by unsupervised k-means cluster analysis of the biomarker data was used to identify patient clusters. A prediction model was built on the basis of a previous cohort to predict the 1 of the 4 previously identified clusters to which the patients of our new cohort would belong. RESULTS: Cluster analysis identified 4 serum biomarker-based clusters, 3 of which (clusters B, C, and D) were comparable to the previously identified clusters. Cluster A (33.6%) could be distinguished from the other clusters as being a "skin-homing chemokines/IL-1R1-dominant" cluster, whereas cluster B (18.5%) was a "TH1/TH2/TH17-dominant" cluster, cluster C (18.5%) was a "TH2/TH22/PARC-dominant" cluster, and cluster D (29.5%) was a "TH2/eosinophil-inferior" cluster. Additionally, by using a prediction model based on our previous cohort we accurately assigned the new cohort to the 4 previously identified clusters by including only 10 selected serum biomarkers. CONCLUSION: We confirmed that AD is heterogeneous at the immunopathologic level and identified 4 distinct biomarker-based clusters, 3 of which were comparable with previously identified clusters. Cluster membership could be predicted with a model including 10 serum biomarkers.
BACKGROUND:Atopic dermatitis (AD) is a highly heterogeneous disease, both clinically and biologically, whereas patients are still being treated according to a "one-size-fits-all" approach. Stratification of patients into biomarker-based endotypes is important for future development of personalized therapies. OBJECTIVE: Our aim was to confirm previously defined serum biomarker-based patient clusters in a new cohort of patients with AD. METHODS: A panel of 143 biomarkers was measured by using Luminex technology in serum samples from 146 patients with severe AD (median Eczema Area and Severity Index = 28.3; interquartile range = 25.2-35.3). Principal components analysis followed by unsupervised k-means cluster analysis of the biomarker data was used to identify patient clusters. A prediction model was built on the basis of a previous cohort to predict the 1 of the 4 previously identified clusters to which the patients of our new cohort would belong. RESULTS: Cluster analysis identified 4 serum biomarker-based clusters, 3 of which (clusters B, C, and D) were comparable to the previously identified clusters. Cluster A (33.6%) could be distinguished from the other clusters as being a "skin-homing chemokines/IL-1R1-dominant" cluster, whereas cluster B (18.5%) was a "TH1/TH2/TH17-dominant" cluster, cluster C (18.5%) was a "TH2/TH22/PARC-dominant" cluster, and cluster D (29.5%) was a "TH2/eosinophil-inferior" cluster. Additionally, by using a prediction model based on our previous cohort we accurately assigned the new cohort to the 4 previously identified clusters by including only 10 selected serum biomarkers. CONCLUSION: We confirmed that AD is heterogeneous at the immunopathologic level and identified 4 distinct biomarker-based clusters, 3 of which were comparable with previously identified clusters. Cluster membership could be predicted with a model including 10 serum biomarkers.