Angus DC, Birmingham MC, Balk RA, Scannon PJ, Collins D, Kruse JA, Graham DR, Dedhia HV, Homann S, MacIntyre N, for the E5 Study Investigators . E5 Murine Monoclonal Antiendotoxin Antibody in Gram-Negative SepsisA Randomized Controlled Trial. JAMA. 2000;283(13):1723–1730. doi:10.1001/jama.283.13.1723
Author Affiliations: Department of Anesthesiology and Critical Care Medicine, Critical Care Medicine Division, and Center for Research on Health Care, University of Pittsburgh, Pittsburgh, Pa (Dr Angus); Kaleida Health/Millard Fillmore Hospital and the State University of New York, Buffalo (Dr Birmingham and Mr Collins); Rush-Presbyterian-St Luke's Medical Center, Chicago, Ill (Dr Balk); XOMA Corporation, Berkeley, Calif (Dr Scannon); Detroit Receiving Hospital, Detroit, Mich (Dr Kruse); St John's Hospital-Springfield Clinic, Springfield, Ill (Dr Graham); West Virginia University Health Science Center, Morgantown (Dr Dedhia); Swedish-American Hospital, Rockford, Ill (Dr Homann); and Duke University Medical Center, Durham, NC (Dr MacIntyre).
Caring for the Critically Ill Patient Section Editor: Deborah J. Cook, MD, Consulting Editor, JAMA.
Context Knowledge and understanding of gram-negative sepsis have grown over
the past 20 years, but the ability to treat severe sepsis successfully has
Objective To assess the efficacy and safety of E5 in the treatment of patients
with severe gram-negative sepsis.
Design A multicenter, double-blind, randomized, placebo-controlled trial conducted
at 136 US medical centers from April 1993 to April 1997, designed with 90%
power to detect a 25% relative risk reduction, incorporating 2 planned interim
Setting Intensive care units at university medical centers, Veterans Affairs
medical centers, and community hospitals.
Patients Adults aged 18 years or older, with signs and symptoms consistent with
severe sepsis and documented or probable gram-negative infection.
Intervention Patients were assigned to receive 2 doses of either E5, a murine monoclonal
antibody directed against endotoxin (n = 550; 2 mg/kg per day by intravenous
infusion 24 hours apart) or placebo (n = 552).
Main Outcome Measures The primary end point was mortality at day 14; secondary end points
were mortality at day 28, adverse event rates, and 14-day and 28-day mortality
in the subgroup without shock at presentation.
Results The trial was stopped after the second interim analysis. A total of
1090 patients received study medication and 915 had gram-negative infection
confirmed by culture. There were no statistically significant differences
in mortality between the E5 and placebo groups at either day 14 (29.7% vs
31.1%; P = .67) or day 28 (38.5% vs 40.3%; P = .56). Patients presenting without shock had a slightly
lower mortality when treated with E5 but the difference was not significant
(28.9% vs 33.0% for the E5 and placebo groups, respectively, at day 28; P = .32). There was a similar profile of adverse event
rates between E5 and placebo.
Conclusions Despite adequate sample size and high enrollment of patients with confirmed
gram-negative sepsis, E5 did not improve short-term survival. Current study
rationale and designs should be carefully reviewed before further large-scale
studies of patients with sepsis are conducted.
While knowledge and understanding of gram-negative sepsis have grown
over the last 20 years, the ability to treat it successfully has not changed
substantially.1 Despite the introduction of
more potent antibiotics and more sophisticated life-support technology, an
estimated 200,000 Americans develop gram-negative sepsis each year with reported
mortality rates of 30% to 65%.2- 5
The prime initiator of gram-negative sepsis is endotoxin, the lipopolysaccharide
component of the bacterial outer membrane.6- 8
Endotoxin triggers the production of proinflammatory monokines (eg, tumor
necrosis factor) which in turn stimulate a variety of proinflammatory and
anti-inflammatory mediator cascades that result in the systemic signs and
organ dysfunction that characterize clinical sepsis.9
Accordingly, it has been hypothesized that agents that bind endotoxin may
mitigate the subsequent cascade, resulting in a decrease in the clinical manifestations
of sepsis and improvement in outcome.1,10,11
A murine monoclonal antibody of the IgM class, E5 (XOMA Corp, Berkeley,
Calif) was developed by immunizing mice to endotoxin from the J5 mutant of Escherichia coli.12,13
Although endotoxins vary across gram-negative species, E5 has been demonstrated
in vitro to react with the core elements of endotoxin, implying specific binding
to a broad range of gram-negative endotoxins.12,14
In several animal models of sepsis, including neutropenic rats inoculated
rats who underwent cecal perforation,16 and
mice who were exposed to lipopolysaccharide,17
E5 therapy administered after the onset of clinical signs significantly reduced
mortality when compared with controls.
Prior human studies include 2 small studies (N = 9 and N = 39)18,19 in which safety and dosage were assessed
in septic patients and 2 large efficacy trials (N = 486 and N = 847). Both
efficacy trials were randomized, double-blind multicenter trials of patients
with known or suspected gram-negative sepsis in which the primary end point
was 30-day mortality. The first trial failed to detect a reduction in mortality
overall but post-hoc analysis of patients with gram-negative infection, who
were not in refractory shock at enrollment (n = 137), showed improved resolution
of organ dysfunction and survival.10 The subsequent
trial excluded patients in refractory shock and demonstrated improved resolution
of organ failure but failed to demonstrate a significant reduction in mortality.20 In both trials, E5 appeared to be well tolerated
with no significant increase in adverse events.
Based on these observations, we undertook a third efficacy trial of
E5 to determine whether the administration of E5 enhances survival in patients
with severe sepsis due to documented or probable gram-negative infection.
This trial was designed to have greater power than the 2 previous efficacy
trials, to exclude patients with refractory shock, and to analyze the subgroup
without nonrefractory shock. However, the study was discontinued after the
second interim analysis because of insufficient effect per a prespecified
This prospective, double-blind, randomized, placebo-controlled trial
was conducted at 136 medical centers in the United States from April 1993
to April 1997. To be eligible for the study, patients had to have documented
or probable gram-negative infection, meet criteria for severe sepsis, be 18
years or older, require care in an intensive care unit, and not be improving
despite standard supportive therapy and appropriate antibiotic coverage. Patients
were entered into the study only after review and approval of entry criteria
by a screening authorization committee on call 24 hours per day.
Gram-negative infection was considered documented by a positive culture
or convincing gram stain no more than 2 calendar days prior to enrollment.
A patient could be enrolled without a documented gram-negative infection in
the setting of postsurgical intra-abdominal sepsis in which the intestinal
tract had either been perforated or required partial resection for ischemia
(provided that >1 day but <7 days had elapsed since surgery). Where frank
spillage of stool into the abdomen resulted from these events, the patient
could qualify without a gram stain or culture and without the 24-hour waiting
Sepsis, severe sepsis, and septic shock were defined as per the recommendations
of the American College of Chest Physicians/Society of Critical Care Medicine
Consensus Conference.21 Specifically, sepsis
was defined as the presence of any 2 of the following: temperature higher
than 38°C or lower than 36°C; heart rate greater than 90/min, respiratory
rate greater than 20/min, minute ventilation greater than 10 L/min, or PaCO2 of less than 32 mm Hg; and white blood cell count greater than 12
× 109/L or less than 4 × 109/L, or greater
than 10% immature forms (bands).
Severe sepsis was defined as sepsis and either hypotension or evidence
of hypoperfusion and organ dysfunction. Hypotension was defined either as:
systolic blood pressure of less than 90 mm Hg or a reduction of greater than
40 mm Hg from baseline despite administration of greater than 500-mL crystalloid
solution or greater than 250-mL colloid suspension, or vasopressor therapy
to maintain a systolic blood pressure of greater than 90 mm Hg despite an
appropriate fluid challenge.
Hypoperfusion or organ system dysfunction was defined as one of the
following: blood lactate level greater than 2 mmol/L; renal dysfunction (>44.2
µmol/L [>0.5 mg/dL] increase in serum creatinine or urine output <0.5
mL/kg for >2 hours despite hydration in patients with normal baseline renal
function or >76.26 µmol/L [>1 mg/dL] increase in serum creatinine in
patients with chronic renal impairment); disseminated intravascular coagulation
(abnormally low or acutely decreased [>25%] platelet count with elevated prothrombin
time, elevated partial thromboplastin time, or clinical bleeding, and absence
of confounding factors such as liver failure or anticoagulant therapy); hepatobiliary
dysfunction (serum bilirubin of 34.2 µmol/L [>2.0 mg/dL] with elevation
of alkaline phosphatase, γ-glutamyl transferase, or serum transaminases
beyond twice the upper limit of normal and no evidence of preexisting hepatobiliary
disease); acute lung injury (PaO2<70 mm Hg on room air or ratio
of PaO2 to fraction of inspired oxygen [FIO2] <280
mm Hg in the absence of congestive heart failure or primary pulmonary pathology
such as lower respiratory tract infection, pulmonary embolism, or interstitial
lung disease). In patients presenting with pneumonia, pulmonary dysfunction
could not be used as the sole criterion for organ dysfunction. Septic shock
was defined as sepsis with hypotension plus hypoperfusion and organ dysfunction.
Patients were excluded from participation if any of the following conditions
were present: granulocyte count of less than 1 × 109/L prior
to the onset of sepsis; infections associated with burns, pregnancy, or lactation;
prior therapy with, or known allergy to, murine antibodies; human immunodeficiency
virus infection; documented or suspected recent acute myocardial infarction;
refractory shock (systolic blood pressure <90 mm Hg despite maximum vasopressor
therapy); treatment with another investigational therapy; or lack of commitment
to full life-support measures by the primary physician.
Patients were required to meet all entry criteria (except gram stain
and culture documentation) within the 12 hours prior to commencing study medication.
The study protocol was approved by each study center's institutional review
board. Informed consent was obtained from the patient, next of kin, or legally
authorized representative prior to enrollment.
The formulation of E5 that was used in the study contained a clear solution
of 2-mg murine monoclonal antibody per milliliter of 5 mmol/L of sodium phosphate
per 0.15 mmol/L of sodium chloride buffer, containing 0.01% polysorbate 80.
Placebo consisted of a 0.1 mg/mL human serum albumin solution that was identical
in appearance to E5. The dose of study medication was 2 mg/kg administered
by intravenous infusion over 1 hour as 1 mL/kg of E5 solution. The full dose
was administered twice, first on study day 1, and again 24 hours later on
study day 2. This dosage regimen was based on prior pharmacokinetic studies
in septic patients and was the same as that used in the 2 prior efficacy trials.10,20
Clinical data were collected on standardized case report forms and included
specific entry criteria, date and time of study drug infusion, and relevant
data concerning adverse clinical events and efficacy end points. Each case
report form was verified against the medical record by a study monitor and
reviewed by a medical monitor, both of whom were blinded to treatment allocation.
Adverse events were independently reviewed by an independent data safety and
Mortality at day 14 was the primary study end point. However, survival
was determined through day 28, and adverse events were monitored until day
32. Patients discharged before day 28 were contacted by telephone. In the
event that a patient was inaccessible at day 28, at least 2 additional telephone
calls were made, and a telegram was sent within the next 2 weeks to ascertain
Study medication was discontinued and the patient was withdrawn from
the study if a serious adverse event occurred; the patient was determined
to be ineligible for the study following assignment of study medication; or
if the investigator, patient, or patient's representative decided to withdraw
For sample size calculation, the 14-day, all-cause mortality for documented,
severe gram-negative sepsis was assumed to be 28% for placebo patients. A
sample size of 1700 was calculated as adequate to show a 25% or greater relative
risk reduction in 14-day all-cause mortality with at least 90% power at the
4.5% significance level. Because prior studies suggested patients in severe
sepsis without shock were more likely to benefit from E5, we also decided
prospectively to analyze this subgroup. In a previous study, 43% of the patients
with documented gram-negative sepsis were not in shock at time of treatment.
Assuming a similar proportion, we estimated 730 patients in the nonshock subset,
providing 80% power to detect a reduction in mortality from 25% to 14.5% with α
The planned interim analyses were based on the approach of Lan and DeMets.22 The stopping rules were derived using the method
of Pampallona, Tsiatis, and Kim, an extension of the Lan and DeMets approach
to control for statistical power, which preserves both type I error at .0225
(under the null hypothesis of no treatment difference) and type II error at
.0225 (under the alternative hypothesis of a differential mortality rate of
0.0909 [placebo − E5]). The interim analyses were reviewed by the data
safety and monitoring board.
Data for patients were considered evaluable and included in the efficacy
and safety analyses if any amount of study medication was administered. Baseline
characteristics and mortality rates were compared between treatment groups
using the 2 sample t tests and Pearson χ2 tests as appropriate. Patients who received study medication, who
had incomplete dose administration, or who did not receive study medication
were summarized by reason. Patient mortality was also reviewed by individual
sites for each treatment group, and Kaplan-Meier survival probabilities were
determined and compared by the log-rank test. In addition to the primary analyses
of all patients who received treatment and the subgroup who presented without
shock, we also conducted post-hoc analyses using intention-to-treat and in
subgroups classified by the presence or absence of major comorbidities, organ
failure, site of infection, and infectious origin. The incidence of serious
adverse events among randomized and treated patients was summarized by treatment
group using the World Health Organization dictionary.23
Significance for the primary end point was based on the stopping rule
above. Significance for distribution of baseline characteristics, adverse
events, and post-hoc analyses was assumed at P<.05.
A total of 1102 patients were randomly assigned to receive either E5
or placebo (Figure 1). Of those,
1090 (98.9%) received some amount of study medication and were therefore considered
to have been treated. For the 12 patients who did not receive treatment (8
in the placebo arm and 4 in the E5 arm), reasons cited included the development
of a serious adverse event between enrollment and initiation (6 in the placebo
arm and 3 in the E5 arm), error in intravenous administration (1 in the placebo
arm), and exclusion criteria not met (1 in the placebo arm). No reason was
given for failure to administer therapy for 1 patient in the E5 arm.
The mean number of patients enrolled per site was 8 (range, 1-50). Baseline
characteristics of the cohort are described in Table 1. Patients ranged in age from 18 to 95 years with a mean
age of 60.7 years. Hypotension or need for vasopressors was present in three
quarters of the patients at presentation, shock and renal dysfunction were
each present in half, pulmonary dysfunction and lactic acidosis were each
present in a third, and disseminated intravascular coagulation was present
in a quarter. Hepatobiliary dysfunction was only noted in 5% of the patients
at presentation. Most patients (84%) had acute dysfunction of at least 1 organ
system at enrollment.
Most patients presented with primary bacteremia, intra-abdominal infection,
pneumonia, or urinary tract infection. Gram-negative infection was confirmed
by culture in 915 patients (83.9%). In 736 cases (67.5%), a single gram-negative
organism was isolated. E coli was isolated most commonly
(39.0% of cases), followed by Klebsiella (18.6%), Pseudomonas (13.7%), Enterobacter
(11.7%), Bacteroides (5.2%), Proteus (3.8%), Citrobacter (3.4%), Serratia (3.4%), Acinetobacter (2.7%), Haemophilus (2.3%), and Xanthomonas
(2.1%) species. The microbiologic origin was generally similar across sites
of infection though Pseudomonas species were the
most common isolate in patients presenting with respiratory tract infection
There were no significant differences in baseline characteristics between
the E5 group and the placebo group with the exception of lactic acidosis,
which was more common in the E5 group (P = .04).
A total of 546 patients received E5 and 544 patients received placebo. Approximately
90% of patients in each treatment group completed both doses of study medication.
For 61 patients in the E5 group (11.1%) and 46 patients in the placebo group
(8.3%), dose administration was incomplete. Four patients in the E5 group
and 8 patients in the placebo group had no infusion of study medication. The
most common reason for incomplete infusion was death, which occurred in 5.5%
of randomized patients in the E5 group and 5.4% of patients in the placebo
group. Among patients who received no infusion, the most common reason was
the occurrence of a serious adverse event for 0.5% of patients in the E5 group
and 1.1% of patients in the placebo group.
Mortality rates are shown in Table
2. No patients were lost to follow-up. There were no observed differences
between E5- and placebo-treated patients at 14 or 28 days. Mortality at 14
days was 29.7% (162 deaths) and 31.1% (169 deaths) in the E5 and placebo groups,
respectively (P = .67). This net difference (1.4%)
fell within the formal stopping boundaries and hence led to termination of
the study by the data safety and monitoring board. Mortality at 28 days was
38.5% (210 deaths) and 40.3% (219 deaths) in the E5 and placebo groups, respectively
(P = .56). Kaplan-Meier analysis confirmed the lack
of a statistically significant difference in survival during the first 28
days (Figure 2).
Not surprisingly, patients without shock (n = 469) had a lower mortality
rate than those with shock (n = 621); (30.9% vs 45.7% on day 28, respectively; P<.001). In the prospectively defined post-hoc analysis
of those without shock, mortality was lower in those treated with E5 but the
difference was not significant (mortality at 28 days was 28.9% in the E5 group
and 32.6% in the placebo group, P = .32; Table 2 and
Figure 3). Other post-hoc analyses, including intention-to-treat
analysis and subgroup analyses based on comorbidity, organ failure, site of
infection, and organism, failed to demonstrate any significant difference
between treatment groups (Table 2).
Both serious (54.1%) and nonserious (56.7%) adverse events were commonly
reported in the study. However, there were no statistically significant differences
between treatment arms and most events appeared to be complications of sepsis.
In particular, allergic reactions to E5 were not apparent in this cohort.
Serious adverse events occurring with an incidence greater than 1% are presented
in Table 3.
In this third phase 3 study of the E5 monoclonal antiendotoxin antibody,
we failed to demonstrate that E5 reduces mortality in patients with gram-negative
sepsis. Furthermore, though mortality in those without shock was lower when
treated with E5, as had been hypothesized previously, neither this difference
nor any other post-hoc analysis was statistically significant. These results
are disappointing and raise important issues regarding the role of antiendotoxin
therapies, our understanding of sepsis, and the evaluation of potential antisepsis
There have been several large clinical efficacy trials of antisepsis
strategies that have failed to demonstrate a beneficial effect on mortality.24,25 The many reasons cited for the failure
of prior trials can be categorized under 3 domains: choice of therapy, study
design,26 and the prevailing conceptual model
Problems with choice of therapy include insufficient knowledge regarding
the biologic effects and the appropriate dosage and duration of therapy. With
the exception of E5, HA-1A antiendotoxin monoclonal antibody,11
and taurolidine,28 most antisepsis strategies
tested in the last decade have targeted endogenous inflammatory mediators.24 However, these mediators have multiple effects and
the degree of suppression or enhancement required to optimize survival is
complex and poorly understood.1,9
In contrast, endotoxin is an exogenous, toxic stimulus and its neutralization
in the acute setting is likely to be beneficial to the host. Unlike some antiendotoxin
agents, E5 was shown in preclinical studies to not only bind endotoxin12- 14,29- 31
but to neutralize it with subsequent attenuated inflammatory response and
improved outcome.15- 17,30- 34
The considerable prior experience with E5 has also led to a good understanding
of its underlying pharmacokinetic and pharmacodynamic properties, and 2 prior
large clinical trials have used E5 at the same dosage with encouraging results
in similar populations.10,20 We
do not, however, know the optimal duration of therapy. One might hypothesize
that E5 should have been administered over a longer period since sepsis and
organ dysfunction can often last for many days, during which recurrent endotoxemia
Attempting to learn from potential design flaws in previous trials,
we undertook several steps in the design and conduct of this trial to minimize
the chance of failure due to study design. First, we shortened the time window
during which patients had to present with severe sepsis from 24 to 12 hours.
Second, only patients failing to improve during the 12-hour period could be
enrolled. Third, we targeted only patients with strong evidence of gram-negative
infection and sepsis (ie, those most likely to benefit from an antiendotoxin
strategy). Fourth, patients who appeared too sick to benefit (ie, were experiencing
refractory shock or deemed unlikely to survive) were excluded.
Each of the above steps was monitored and tracked prospectively through
the use of a screening authorization center staffed 24 hours per day by experts
familiar with sepsis. The screening center authorized enrollment only after
discussion with the clinical site investigators to determine suitability.
The entire enrollment process was then reviewed separately by a patient evaluation
and review committee to ensure compliance with study goals. In addition, the
2 prior phase 3 studies provided valuable information regarding potentially
important subgroups in which benefit might be maximal. We used this information
to construct entry criteria that would maximize enrollment of particular subgroups,
to determine sample size, and to specify prospective subgroups for analysis.
A potential limitation of this study was our ability to detect a treatment
effect in the subgroup of patients with gram-negative sepsis without shock,
who were identified in previous studies as the group most likely to benefit.
Even if we had completed enrollment, we would only have been able to detect
a significant effect in this subgroup if mortality were reduced by almost
half. Because not all patients had confirmed gram-negative infection and the
trial was stopped after the second interim analysis, we had little statistical
power to determine whether the observed decrease in mortality with E5 was
significant. Another potential limitation of our study was the large number
of sites and long enrollment period. Both of these factors may have increased
the variability in management and outcome of sepsis and organ dysfunction.
However, analyses of time trends and interinstitutional comparisons failed
to demonstrate significant differences in outcome (data not shown). Thus,
we believe our study design was generally sound with regard to most prior
criticisms of sepsis trial design. Therefore, we conclude E5 either has no
effect or has an effect too small to be detected using our standard approach
to sepsis trial design and sample size.
If there is an important effect of E5, or of other unsuccessful antisepsis
strategies, we may have to consider a more radical change in sepsis study
design. For example, we may wish to study more homogeneous patient groups.
As with other sepsis studies, our cohort varied widely with respect to site
of infection, microorganism, and number and type of organ failures. It is
possible that E5, or other antisepsis strategies, may work in particular subgroups
in which mortality is most prominently driven by the septic process. To isolate
this effect would likely require more focused enrollment, higher sample size,
or both. Successful cardiology trials, for example, have been targeted at
a relatively focused group of patients with acute myocardial infarction and
have been powered to detect much smaller changes in mortality than most sepsis
trials. If we had powered our study to find the same absolute risk reductions
in 28-day mortality as the Global Utilization of Streptokinase and Tissue
Plasminogen Activator for Occluded Coronary Arteries (GUSTO I) study of thrombolytic
therapy in acute myocardial infarction,35 we
would have required 76,000 patients. Other design modifications might include
the study of combination therapies, or therapies titrated to particular elements
of the inflammatory response that characterizes sepsis and are documented
in all enrolled patients. Finally, choosing end points other than all-cause,
short-term mortality may allow easier determination of therapies that have
an important effect on septic morbidity.
In conclusion, despite prior encouraging results regarding the potential
effects of both E5 and other antiendotoxin antibodies, and despite our efforts
to conduct a study of a large cohort of carefully selected patients, we have
failed to prove that E5 is beneficial in the treatment of sepsis. This trial
concludes, at least for now, more than 20 years of antiendotoxin research
without providing any definitive answers about mortality reduction despite
a nagging tendency to show beneficial effects in post-hoc analyses. If we
are to avoid such an indeterminate outcome from future endeavors, it behooves
us not only to look to better and more realistic discovery programs and laboratory
research but also to focus on improved clinical study design.