To study the systemic release and kinetics of high mobility group box-1 protein (HMGB1) in relation to clinical features in a population of patients with severe sepsis or septic shock and to compare these with the kinetics of the cytokines interleukin-6, interleukin-8, interleukin-10, and tumor necrosis factor-alpha. Prospective study of two cohorts of patients. Intensive care unit and infectious disease clinic at Karolinska University Hospital Huddinge. Twenty-six patients with severe sepsis, 33 patients with septic shock, and a reference group of five patients with sepsis. None. Sixty-four patients were included, ten of whom died within 28 days. Cytokine levels were measured at five time points during the first week after admission and were correlated to Acute Physiology and Chronic Health Evaluation II and Sepsis-related Organ Failure Assessment scores. Two HMGB1 assays were used. Both demonstrated delayed kinetics for HMGB1 with high levels on inclusion that remained high throughout the study period. Serum concentration at 144 hrs, the last sampling point, was 300 times higher, 34,000 +/- 76,000 pg/mL (mean +/- sd), than any of the other cytokines. This study, however, found no predictable correlation between serum levels of HMGB1 and severity of infection. We did quite unexpectedly find significantly lower levels of HMGB1 in nonsurvivors compared with survivors as measured by our main assay, but the other showed no difference between the two groups. Levels of interleukin-6, interleukin-8, interleukin-10, and tumor necrosis factor-alpha correlated significantly with severity of disease, and all were significantly higher in patients with septic shock compared with those with severe sepsis. Neither of these comparisons showed significant correlations for HMGB1. This is the first prospective study assessing the release over time of HMGB1 in a population of patients with sepsis, severe sepsis, or septic shock. Levels remained high in the majority of patients up to 1 wk after admittance, indicating that the cytokine indeed is a downstream and late mediator of inflammation. Further studies are required to fully define the relationship of HMGB1 to severity of disease.