Objective: To compare the effectiveness of 100 mg of HA-1A and placebo in reducing the 14-day all-cause mortality rate in patients with septic shock and gram-negative bacteremia in the Centocor: HA-1A Efficacy in Septic Shock (CHESS) trial, and to assess the safety of 100 mg of HA-1A given to patients with septic shock who did not have gram-negative bacteremia. Design: Large, simple, group-sequential, randomized, double-blind, multicenter, placebo-controlled trial. Setting: 603 investigators at 513 community and university-affiliated hospitals in the United States. Patients: Within 6 h before enrolment, the patients had been in shock with a systolic blood pressure of less than 90 mmHg after adequate fluid challenge or had received vasopressors to maintain blood pressure. These episodes of shock began within 24 h of enrolment. A presumptive clinical diagnosis of gram-negative infection as the cause of the shock episode and a commitment from the patients' physicians to provide full supportive care were required. Measurements: Blood cultures were obtained within 48 h of enrolment, and death at day 14 after treatment was recorded. Adverse events occurring within 14 days after enrolment were also tabulated. Results: 2199 patients were enrolled; 621 (28.2%) met all enrolment criteria, received HA-1A or placebo, and had confirmed gram-negative bacteremia. Mortality rates in this group were as follows: placebo, 32% (95 of 293) and HA-1A, 33% (109 of 328) (p = 0.864, Fisher exact test, two-tailed; 95% CI for the difference, -6.2% to 8.6%). Mortality rates in the patients without gram-negative bacteremia were as follows: placebo, 37% (292 of 793) and HA-1A, 41% (318 of 785) (p = 0.073, Fisher exact test, one-tailed; CI, -0.8% to 8.8%). Conclusions: In this trial, HA-1A was not effective in reducing the 14-day mortality rate in patients with gram-negative bacteremia and septic shock. These data do not support using septic shock as an indication for HA-1A treatment. If HA-1A is effective in reducing the mortality rate in patients dying from endotoxemia, these patients must be identified using other treatment criteria.