Recent controlled studies that evaluated the efficacy of an adjuvant antithrombin (AT) III therapy in severe sepsis used a uniform AT-III dose and duration of therapy and did not adjust to the actual AT-III deficit. It was the aim of the present study to explore if surgical patients with severe sepsis might have a treatment benefit from an activity-guided AT-III therapy. We performed a retrospective cohort analysis using an intensive care unit (ICU) database. To examine the effect of AT-III on outcome and on red cell transfusion rate, multivariate generalized additive models (GAMs), Cox-type additive hazard regression models, and propensity score adjustments were used. Five hundred forty-five postoperative surgical patients requiring ICU therapy because of severe sepsis were analyzed. Antithrombin III was given to those patients believed to be at a high risk of dying. Antithrombin III therapy was guided by the individual AT-III activity and aimed at the maintenance of an activity of at least 100%. Antithrombin III supplementation was discontinued after the plasma AT-III activity had been persistently normal without simultaneous AT-III infusion. We found that patients receiving additional AT-III (n = 230) were sicker than those on standard therapy (n = 315; admission Acute physiology and chronic health evaluation II score, 19.8 ± 7.3 vs. 17.9 ± 7.1 [mean ± SD]; P < 0.005). Correspondingly, 28-day mortality was higher in patients who had an additional AT-III therapy than in those on standard therapy (46.3% vs. 36.9%; P < 0.03), as was the number of red cell units transfused during ICU stay (21.5 ± 26.7 vs. 9.3 ± 12.1; P < 0.001). At multivariate analysis, there was no significant effect of AT-III therapy on 28-day mortality (GAM: odds ratio, 1.012; 95% confidence interval [CI], 0.651 - 1.573; P = 0.957) and 90-day survival time (Cox-type additive hazard regression: hazard ratio, 1.034; 95% CI, 0.779 - 1.387; P = 0.794). However, AT-III therapy was associated with a significantly higher frequency of red cell unit transfusion (GAM/zero-inflated Poisson: estimate, 1.26; 95% CI, 1.15 - 139; P < 0.001). Our results suggest that there seems to be no relevant effect of an activity-guided AT-III therapy on the prognosis of surgical patients with severe sepsis. However, transfusion frequency rises by AT-III therapy.