Current atopic dermatitis (AD) models link epidermal abnormalities in lesional skin to cytokine activation. However, there is evolving evidence of systemic immune activation and detectable abnormalities in nonlesional skin. Because some of the best single correlations with severity (Scoring of AD, or SCORAD) are detected not only in lesional but also nonlesional skin and blood, more complex biomarker models of AD are needed. We thus performed extensive biomarker measures in these compartments using univariate and multivariate approaches to correlate disease biomarkers with SCORAD and with a combined hyperplasia score [thickness and keratin 16 (K16) mRNA] at baseline and after cyclosporine A treatment in 25 moderate to severe AD patients. Increases in serum cytokines and chemokines (IL-13, IL-22, CCL17) were found in AD versus healthy individuals and were reduced with treatment. SCORAD correlated with immune (IL-13, IL-22) and epidermal (thickness, K16) measures in lesional and, even more strongly, in nonlesional AD. Serum cytokines also had higher correlations with nonlesional markers at baseline and with treatment. Multivariate U statistics improved baseline and treatment-response SCORAD correlations. Nonlesional models showed the strongest correlations, with further improvement upon integration of serum markers. Even better correlations were obtained between biomarkers and the hyperplasia score. Larger cohorts are needed to confirm these preliminary data.