Septic shock is a leading cause of death in critical patients. In patients with gram-negative septic shock, hemoperfusion with polymyxin B aims to remove endotoxins from plasma. We analyzed the clinical and biological response to hemoperfusion in patients with septic shock and acute kidney injury. This prospective case-control study in the medical-surgical intensive care unit of a university hospital included consecutive adults patients with septic shock and suspected gram-negative bacteria infection with elevated plasma endotoxin activity (EAA > 0.6 EU/ml) and acute kidney injury requiring continuous renal replacement therapy (CRRT). At onset of septic shock, half underwent CRRT plus hemoperfusion with polymyxin B for two hours a day during two consecutive days (hemoperfusion group) and half received only CRRT (control group). We measured clinical, physiological, and biological parameters (EAA, C-reactive protein, procalcitonin, and cytokines) daily during the first 5 days. We included 18 patients (male, 33%; mean age, 67.5; mean SOFA score, 11.3). Abdominal infections predominated (50% had peritonitis). At the beginning of CRRT, RIFLE classification was "failure" for 72% and "injury" for 28%. Baseline characteristics did not differ between groups. Patients in the hemoperfusion group required longer mechanical ventilation (12.4 vs. 9.4 days, p = 0.03) and CRRT (8.5 vs. 6 days, p = 0.01) than in the control group. Noradrenaline doses, lactate, procalcitonin, and C-reactive protein decreased in both groups. At day 5, EAA was significantly lower in the hemoperfusion group (0.58 EU/ml vs. 0.73 EU/ml in controls, p = 0.03). There were no significant differences between groups in other biomarkers or ICU mortality (33.3% in the treatment group vs. 44.4% in the control group, p = 0.5). No adverse effects of hemoperfusion were observed. Hemoperfusion with polymyxin B added to CRRT resulted in faster decrease in endotoxin levels, but we observed no improvements in clinical, physiological, or biological parameters. The online version of this article (10.1186/s13613-018-0465-8) contains supplementary material, which is available to authorized users