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.