Termination of the Department of Veterans Affairs Cooperative Study of Steroid Therapy for Systemic Sepsis Peter Peduzzi, PhD Cooperative Studies Program Coordinating Center, Department of Veterans Affairs Medical Center, West Haven, CT
ABSTRACT: The Department of Veterans Affairs Cooperative Study Program conducted a randomized, double-masked trial of steroid therapy versus placebo therapy for patients with systemic sepsis from 1983 to 1986. Treatment was initiated as soon as sepsis was recognized and before results of cultures confirmed infection. The original hypothesis was to test the effect of therapy on short-term (14-day) mortality in patients with gramnegative bacteremia. Because therapy had to begin before culture results were available, all septic patients had to be enrolled. Consequently, the study was modified to evaluate therapy in all patients with sepsis, and by post-stratification in those with gram-negative bacteremia. Patient enrollment was planned to continue for 3.5 years to achieve a sample size of 276 patients. After 223 patients were randomized, 14-day mortalities were 22% in the placebo-treated group versus 21% in the steroid-treated group (p = 0.97). In contrast, for the 51 patients with gram-negative bacteremia, mortalities were 27% placebo-treated versus 7% steroid-treated (p = 0.11). The Data Monitoring Board recommended continuation of the trial to evaluate what appeared to be an emerging gram-negative trend, but the Cooperative Studies Evaluation Committee decided to end the trial 12 months early because of the lack of efficacy in all septic patients. The reasons for the proposed extension and for the termination of the trial are presented. The more general problem of evaluating a biologically important subgroup imbedded in a large clinical trial is also discussed. KEY WORDS: Sequential monitoring, early termination, randomized clinical trials
INTRODUCTION In 1983, the Department of Veterans Affairs (VA) Cooperative Studies Program initiated a randomized clinical trial to evaluate the effectiveness of steroid therapy for the treatment of patients with systemic sepsis [1]. Use of steroids for the treatment of this condition has been and continues to be controversial. Animal studies have shown that steroids decrease morbidity and mortality, but results in humans have been less convincing. The most
Address reprint requests to: Peter Peduzzi, PhD, Cooperative Studies Program 151A, VA Medical Center, West Haven, CT 06516. Received April 20, 1990; revised November 13, 1990. Controlled ClinicalTrials 12:395-407(1991) © Elsevier SciencePublishing Co., Inc. 1991 655 Avenueof the Americas,New York,New York10010
395 0197-2456/91/$3.50
396
P. Peduzzi controversial trial of gram-negative sepsis was that of Schumer [2], who showed a 75% reduction in mortality with steroid therapy compared with placebo therapy. However, Schumer's trial was criticized [3] for using varying doses of two glucocorticoids, for lack of a uniform antimicrobial protocol, and for suboptimal hemodynamic support. The VA trial was designed to evaluate the role of early adjunctive steroid therapy in patients with systemic gram-negative sepsis. Since steroid therapy was expected to be most effective w h e n given as early as possible in the course of sepsis, treatment could not be delayed for blood culture results. Moreover, it was assumed that nearly all patients who satisfied the inclusion criteria for clinical sepsis would have gram-negative infections. However, during the design of the trial it became apparent that there were no reliable clinical means of distinguishing patients with gram-negative sepsis from those with infection from other causes. At the first Data Monitoring Board (DMB) meeting after initiation of the trial (9 months), 71 patients had been enrolled and of these 60% had gram-negative sepsis and only 23% had gram-negative bacteremia. Therefore, the DMB recommended that the study be modified to assess the effect of therapy in all patients with sepsis and in those with gramnegative infections. The DMB made this decision without knowledge of the observed effect of treatment. In 1985, after 2.5 years of patient intake, the sequential boundary of no difference in mortality between the two treatment groups was reached for all patients. On the other hand, a 75% reduction in mortality with steroid therapy was observed in the small subgroup of 51 patients with gram-negative bacteremia. At that time, the DMB recommended that the trial be continued to evaluate this emerging trend. However, at the scheduled midpoint review of the trial by the Cooperative Studies Evaluation Committee (CSEC), the Committee opted to terminate the trial because steroid therapy did not reduce mortality in all patients with sepsis. The CSEC also viewed the gram-negative trend as more apparent than real because it was based on a total of only eight deaths, six in the placebo-treated group versus two in the steroid-treated group. In this report, the contrasting arguments for extension and termination of the trial are summarized.
DESIGN OF THE VA TRIAL
The trial was designed as a randomized, double-masked, placebo-controlled trial of early, short-term, high-dose steroid therapy in patients with clinical signs of systemic sepsis and normal sensorium. All patients also received conventional therapy with antibiotics and intravenous fluids. Whenever possible, masked drug therapy was begun within 2 hours of the presumptive clinical diagnosis of sepsis. The trial protocol was approved by CSEC, by the H u m a n Rights Committee at the coordinating center and by the Institutional Review Boards at each participating medical center. The progress of the study was monitored annually by a DMB composed of outside experts not involved in the conduct of the study (see Appendix). The DMB requested to review masked data summaries at their annual meetings. Patients eligible for inclusion were those who had a clinical suspicion of
Early Termination of the VA Sepsis Trial
397
sepsis and who developed at least four of the following seven signs of sepsis within an 8-hour period: 1. Shaking chills, fever (>102°F) or hypothermia (28 breaths/min) or hypocapnia (partial pressure of arterial CO2 < 32 mm Hg). 3. Tachycardia (heart rate >100 beats/min) 4. Hypotension (systolic blood pressure
ABSTRACT: The Department of Veterans Affairs Cooperative Study Program conducted a randomized, double-masked trial of steroid therapy versus placebo therapy for patients with systemic sepsis from 1983 to 1986. Treatment was initiated as soon as sepsis was recognized and before results of cultures confirmed infection. The original hypothesis was to test the effect of therapy on short-term (14-day) mortality in patients with gramnegative bacteremia. Because therapy had to begin before culture results were available, all septic patients had to be enrolled. Consequently, the study was modified to evaluate therapy in all patients with sepsis, and by post-stratification in those with gram-negative bacteremia. Patient enrollment was planned to continue for 3.5 years to achieve a sample size of 276 patients. After 223 patients were randomized, 14-day mortalities were 22% in the placebo-treated group versus 21% in the steroid-treated group (p = 0.97). In contrast, for the 51 patients with gram-negative bacteremia, mortalities were 27% placebo-treated versus 7% steroid-treated (p = 0.11). The Data Monitoring Board recommended continuation of the trial to evaluate what appeared to be an emerging gram-negative trend, but the Cooperative Studies Evaluation Committee decided to end the trial 12 months early because of the lack of efficacy in all septic patients. The reasons for the proposed extension and for the termination of the trial are presented. The more general problem of evaluating a biologically important subgroup imbedded in a large clinical trial is also discussed. KEY WORDS: Sequential monitoring, early termination, randomized clinical trials
INTRODUCTION In 1983, the Department of Veterans Affairs (VA) Cooperative Studies Program initiated a randomized clinical trial to evaluate the effectiveness of steroid therapy for the treatment of patients with systemic sepsis [1]. Use of steroids for the treatment of this condition has been and continues to be controversial. Animal studies have shown that steroids decrease morbidity and mortality, but results in humans have been less convincing. The most
Address reprint requests to: Peter Peduzzi, PhD, Cooperative Studies Program 151A, VA Medical Center, West Haven, CT 06516. Received April 20, 1990; revised November 13, 1990. Controlled ClinicalTrials 12:395-407(1991) © Elsevier SciencePublishing Co., Inc. 1991 655 Avenueof the Americas,New York,New York10010
395 0197-2456/91/$3.50
396
P. Peduzzi controversial trial of gram-negative sepsis was that of Schumer [2], who showed a 75% reduction in mortality with steroid therapy compared with placebo therapy. However, Schumer's trial was criticized [3] for using varying doses of two glucocorticoids, for lack of a uniform antimicrobial protocol, and for suboptimal hemodynamic support. The VA trial was designed to evaluate the role of early adjunctive steroid therapy in patients with systemic gram-negative sepsis. Since steroid therapy was expected to be most effective w h e n given as early as possible in the course of sepsis, treatment could not be delayed for blood culture results. Moreover, it was assumed that nearly all patients who satisfied the inclusion criteria for clinical sepsis would have gram-negative infections. However, during the design of the trial it became apparent that there were no reliable clinical means of distinguishing patients with gram-negative sepsis from those with infection from other causes. At the first Data Monitoring Board (DMB) meeting after initiation of the trial (9 months), 71 patients had been enrolled and of these 60% had gram-negative sepsis and only 23% had gram-negative bacteremia. Therefore, the DMB recommended that the study be modified to assess the effect of therapy in all patients with sepsis and in those with gramnegative infections. The DMB made this decision without knowledge of the observed effect of treatment. In 1985, after 2.5 years of patient intake, the sequential boundary of no difference in mortality between the two treatment groups was reached for all patients. On the other hand, a 75% reduction in mortality with steroid therapy was observed in the small subgroup of 51 patients with gram-negative bacteremia. At that time, the DMB recommended that the trial be continued to evaluate this emerging trend. However, at the scheduled midpoint review of the trial by the Cooperative Studies Evaluation Committee (CSEC), the Committee opted to terminate the trial because steroid therapy did not reduce mortality in all patients with sepsis. The CSEC also viewed the gram-negative trend as more apparent than real because it was based on a total of only eight deaths, six in the placebo-treated group versus two in the steroid-treated group. In this report, the contrasting arguments for extension and termination of the trial are summarized.
DESIGN OF THE VA TRIAL
The trial was designed as a randomized, double-masked, placebo-controlled trial of early, short-term, high-dose steroid therapy in patients with clinical signs of systemic sepsis and normal sensorium. All patients also received conventional therapy with antibiotics and intravenous fluids. Whenever possible, masked drug therapy was begun within 2 hours of the presumptive clinical diagnosis of sepsis. The trial protocol was approved by CSEC, by the H u m a n Rights Committee at the coordinating center and by the Institutional Review Boards at each participating medical center. The progress of the study was monitored annually by a DMB composed of outside experts not involved in the conduct of the study (see Appendix). The DMB requested to review masked data summaries at their annual meetings. Patients eligible for inclusion were those who had a clinical suspicion of
Early Termination of the VA Sepsis Trial
397
sepsis and who developed at least four of the following seven signs of sepsis within an 8-hour period: 1. Shaking chills, fever (>102°F) or hypothermia (28 breaths/min) or hypocapnia (partial pressure of arterial CO2 < 32 mm Hg). 3. Tachycardia (heart rate >100 beats/min) 4. Hypotension (systolic blood pressure
