Introduction

Whenever a novel therapy is approved for any disease process, continued assessment of safety is paramount. DrotAA was approved for use in adult severe sepsis patients at high-risk of death based on results from the PROWESS trial completed in 2000 [1]. In that trial, patients on commercial heparin at baseline who received DrotAA had a higher 28-day mortality rate (25.9 vs. 22.6%). Consequently, the FDA required a trial in patients at high-risk of death to be treated with DrotAA and randomized to either heparin or placebo.

This large randomized safety study (identified as XPRESS) enrolled 2,000 severe sepsis patients at high-risk of death. Primary results from the XPRESS study, published recently [2], demonstrated no difference in serious bleeding in the heparin versus placebo arm and support the safety of coadministration of prophylactic heparin when infusing DrotAA for severe sepsis. In order to further explore differences between the heparin and placebo groups, we provide a more comprehensive analysis of the XPRESS safety data including subgroup analyses of adverse events performed for clinically relevant variables.

Materials and methods

Institutional/ethical review boards approved the protocol at each investigative site and informed consent was obtained from all patients or their authorized representatives.

Patients

From December 2002 to July 2005, adult patients with severe sepsis at high-risk of death were enrolled in a randomized, double-blind, placebo-controlled study conducted at 224 centers in 20 countries. Specific inclusion and exclusion criteria have been published [2]. Enrollment criteria were consistent with the approved indication for DrotAA in each country.

Intervention

DrotAA (Xigris, Eli Lilly and Company, Indianapolis, IN, USA) at 24 μg/(kg h) for a planned 96-h infusion was administered to each patient. Study drug was low-molecular-weight heparin (LMWH) (enoxaparin, 40 mg subcutaneously once per day), unfractionated heparin (UFH) (5,000 Units subcutaneously twice per day), or placebo in a 1:1:2 randomization scheme, so half of the patients received some form of heparin and half received placebo. A flow chart showing randomization and completion populations has been published [2].

Study drug was administered every 12 h, in accordance with the standard practice of the intensive care unit (ICU), during infusion of DrotAA. In the enoxaparin arm, a placebo was given 12 h later to preserve the blind nature of the experiment. The first study drug injection was given as soon as possible, but no later than 12 h, after start of the DrotAA infusion. All other patient care was at the discretion of the investigator.

Patients were excluded if they were contraindicated for treatment with DrotAA, prophylactic LMWH, or UFH, required a higher dose of prophylactic heparin than specified in the protocol, had acute or chronic renal failure with an estimated creatinine clearance <30 mL/min, had a concurrent need for other anticoagulant/antithrombotic medication during the DrotAA infusion, had a preexisting, uncorrectable medical condition, or their families were not committed to aggressive management of the patient’s severe sepsis.

Measurements and definitions

The number of organ dysfunctions present at baseline was assessed differently in this study from in previous studies of drotrecogin alfa (activated). In addition to cardiovascular, respiratory, renal, and hematologic dysfunction, and unexplained metabolic acidosis, information on the presence of central nervous system and liver dysfunction at baseline was also collected. Therefore, patients in this study could have had seven organ dysfunctions at baseline.

All patients were followed to 28 days for all-cause mortality. During the infusion period (Days 0–6) and the study period (Days 0–28), the following safety measures were assessed either by screening or on clinical occurrence:

  1. 1

    serious adverse events (SAEs) including fatal and nonfatal serious bleeding events (defined below);

  2. 2

    nonserious bleeding events requiring transfusion of packed red blood cells;

  3. 3

    study drug-related nonserious adverse events;

  4. 4

    adverse events that led to permanent discontinuation of study drug;

  5. 5

    venous thrombotic events;

  6. 6

    heparin-induced thrombocytopenia; and

  7. 7

    ischemic stroke.

Fatal bleeding events were overt bleeds considered the primary cause of death. Nonfatal serious bleeding events were: intracranial hemorrhage confirmed by brain imaging or autopsy, bleeding at a critical location (e.g., retinal hemorrhage, major hemarthrosis, or spinal hemorrhage), or any otherwise life-threatening bleed. Venous thrombotic events were defined as: pulmonary embolism (nonfatal or fatal); symptomatic or asymptomatic lower extremity deep-vein thrombosis (DVT) (detected by bilateral compression ultrasonography performed at the end of infusion between Days 4 and 6); and symptomatic central vein thrombosis. Heparin-induced thrombocytopenia (HIT) was defined as thrombocytopenia profile suggestive of HIT syndrome (i.e., severity, temporal evolution) while receiving heparin, and supported, at the discretion of the investigator, by laboratory testing (e.g., anti-heparin antibody, or relevant platelet aggregation studies). Ischemic stroke was defined as an acute onset of focal neurological defect persisting 24 h and without evidence on brain imaging of intracranial hemorrhage or nonvascular cause. A neurological exam was performed within 24 h of the end of the drotrecogin alfa (activated) infusion and at the time of discharge or on study day 28. Ischemic stroke was also reported on clinical occurrence.

Statistical analyses

All safety analyses were performed on all randomized patients. Statistical significance was based on a two-tailed test with an alpha level of 0.05 unless otherwise stated. Treatment groups were compared using analysis of variance (ANOVA) for continuous variables and Pearson’s chi-square test for categorical variables. Potential treatment-by-subgroup interactions were assessed by use of the Breslow–Day test for homogeneity of odds ratios across strata [3, 4], and interactions with a p value < 0.10 are noted as statistically significant.

Independent risk predictors for serious bleeding events during infusion and the 28-day study periods were assessed using a multivariate logistic regression model that included the baseline variables: age, gender, region, number of organ failure, baseline vasopressor usage, baseline ventilator usage, baseline organ dysfunction status, time to treat, Acute Physiology Age and Chronic Health Evaluation (APACHE) score, medical history (pulmonary embolism, hypertension, DVT, stroke, surgery), comorbidity (malignancy, congestive cardiomyopathy, chronic liver disease, chronic obstructive pulmonary disease), baseline heparin usage, racial origin, location prior to hospitalization, prime site of infection, source of infection, and type of infection. A STEPWISE selection criterion in SAS PROC LOGISTIC was used to determine which baseline measures were independent predictors of serious bleeding events. In the stepwise selection process, a p value cut-off of 0.05 was used as a criterion for both entry into the model and for potential removal from the model. Statistically significant factors were included in the logistic regression model with the treatment code to determine how heparin plus DrotAA compared with DrotAA alone (i.e., placebo patients) as a predictor of serious bleeding events.

Results

In this study, 1994 patients received DrotAA and were randomized: (990 placebo and 1,004 heparin). Within the heparin group, 493 patients received LMWH and 511 received UFH. Approximately half the patients in both groups (heparin, 51.7%; placebo, 49.8%) were exposed at baseline to some anticoagulant (LMWH and UFH, vitamin K antagonist, or any antiplatelet agent).

Mortality in this study has been reported [2]. Briefly, the 28-day mortality differed between the heparin and placebo groups, with a numerically lower rate in the heparin group (28.3 vs. 31.9%; p = 0.08). Subgroup analyses showed no difference in mortality between the heparin and placebo groups for patients not exposed to heparin at baseline (i.e., immediately prior to randomization) (29.5 vs. 28.9%). However, among patients on heparin at baseline, higher mortality was observed in patients randomized to placebo (35.6 vs. 26.9%; p = 0.005).

Non-serious adverse events

Incidence of non-serious adverse events considered by the investigator to be related to either study drug (heparin) or DrotAA in the combined heparin group and the placebo group during infusion were 10.0 and 7.2% (p = 0.026), respectively (odds ratio: 1.44, 95% CI 1.04–1.99). Incidence of non-serious adverse events over the 28-day study period in the combined heparin group and the placebo group were 10.1 and 7.9% (p = 0.089), respectively (odds ratio: 1.31, 95% CI 0.96–1.79). During the infusion period, there were numerically more heparin versus placebo patients who experienced non-serious bleeding events, particularly gastrointestinal (29 vs. 11 patients) or renal (eight vs. three patients).

Serious adverse events

During the 28-day study period there were 26 more SAEs in the placebo group than in the heparin group (15 vs. 12%; p = 0.058, odds ratio: 0.77, 95% CI 0.60–1.01). There was a statistically significant higher rate of ischemic stroke in the placebo group during the infusion period (placebo: 1.3% vs. heparin: 0.3%; p = 0.018, odds ratio: 0.24, 95% CI 0.07–0.87) and 28-day study period (1.8 vs. 0.5%; p = 0.009, odds ratio: 0.29, 95% CI 0.10–0.78). The ischemic stroke events in the heparin group occurred randomly during the study, but in the placebo group a cluster of events occurred during Days 2, 3, and 4, the time of DrotAA infusion (DrotAA/Heparin, five strokes, one stroke on each of Days 2, 4, 8, 16, and 19; DrotAA/Placebo, 17 strokes, three strokes on each of Days 2, 3, 4; two strokes on Day 13; and one stroke on each of Days 0, 7, 8, 10, 15, 24). Of the 22 patients who experienced an ischemic stroke, nine had a history of hypertension, one had a history of DVT, and two had a history of stroke. There were 12 more serious bleeding events in the placebo group than in the heparin group (4 vs. 5%, respectively, p = 0.163, odds ratio: 0.74, 95% CI 0.48–1.13).

Baseline heparin use was associated with a higher rate of SAEs in the placebo group. Heparin use was stopped in the placebo group, in accordance with the protocol, and among patients exposed to commercial heparin at baseline 18.0% of placebo patients and 11.6% of heparin patients experienced an SAE over the 28-day study period (Table 1: RR: 0.65, 95% CI 0.47–0.90. Placebo patients on commercial heparin at baseline, compared with heparin patients, had a higher number of cardiac events [4.1% (18/438) vs. 2.9% (9/461)], for example cardiac arrest (1.2 vs. 0.7%), myocardial infarction (0.9 vs. 0%), and atrial fibrillation (0.7 vs. 0.2%). Placebo patients on commercial heparin at baseline also had more gastrointestinal disorders (3.7 vs. 2.4%), for example gastrointestinal hemorrhage (0.9 vs. 0.2%), and venous thrombotic events (8.1 vs. 5.3%) than heparin patients. These differences were not statistically significant. With regard to ischemic stroke, all five of the heparin patients who experienced ischemic strokes were receiving heparin at baseline; only five of the 17 placebo patients who experienced ischemic strokes were receiving heparin at baseline. Among patients who were not on heparin at baseline, the percentage of patients who experienced an SAE was the same in both treatment groups for each time interval,

Table 1 Subgroup analyses of serious adverse events
Full size table

In additional subgroup analyses, treatment-by-subgroup interactions were statistically significant for patient location prior to hospitalization and baseline ventilator use (Breslow–Day p value < 0.10) but not for other subgroup categories, including age, race, primary site or type of infection, number of baseline organ dysfunctions, APACHE class, or time from first organ dysfunction to DrotAA infusion. Subgroup analyses of SAEs during the infusion period and the 28-day study period are provided in Table 1 for clinically important variables.

Outcomes in patients treated with low-molecular-weight heparin versus unfractionated heparin

Patients randomized to receive either LMWH or UFH had similar baseline characteristics (data not shown). The 28-day mortality rate for patients treated with LMWH and UFH was 27.3 and 29.3%, respectively (RR: 0.93, 95% CI 0.76–1.14). There were no statistically significant differences between individual heparin groups with regard to adverse events. There were numerically more SAEs in the LMWH group during Study Days 0 through 28 compared with unfractionated heparin patients (64 vs. 51 patients, p = 0.06).

Subgroup analyses of bleeding events

More bleeding events of any severity were observed in the heparin group than in the placebo group for patients who were not on heparin at baseline (Table 2). For serious bleeding events, analysis of clinically important subgroups identified no significant between-treatment differences for either the infusion period or the 28-day study period (Table 3).

Table 2 Subgroup analyses of any bleeding events
Full size table
Table 3 Subgroup analyses of serious bleeding events
Full size table

Predictors of serious bleeding events

Independent predictors of serious bleeding events were identified by use of a multivariate logistic regression model. The only independent predictors identified were source of infection and baseline number of organ dysfunctions during the infusion period (p = 0.005 and p = 0.008, respectively) and source of infection, baseline number of organ dysfunctions, and region during the 28-day study period (p = 0.017, p = 0.0034 and p = 0.04; ESM Table 3). After adjusting for the independent predictors for each time interval, the chances of having a serious bleeding event was not significantly increased with heparin plus DrotAA compared with DrotAA alone (placebo group) during the infusion period (odds ratio (OR): 0.898; OR 95% CI 0.499, 1.618) or the 28-day study period (OR: 0.722; OR 95% CI 0.467, 1.114; ESM Table 4).

Discussion

Coadministration of prophylactic heparin with DrotAA has an acceptable safety profile. There was no increase in SAEs or serious bleeding events in the heparin group compared with the placebo group. A similar or smaller percentage of heparin patients than placebo patients experienced an SAE or serious bleeding event during both the DrotAA infusion and the 28-day study period. A similar number of patients in both treatment groups experienced a CNS bleeding event or a study-drug-related SAE. Study-drug-related deaths were higher during infusion (p = 0.098) and during the 28-day period (p = 0.055) in the DrotAA/Placebo group than in the DrotAA/Heparin group.

Coadministration of heparin with DrotAA was associated with a statistically significant increase in nonserious study-drug-related adverse events and in any bleeding events that occurred during the infusion period. The differences were not significant for the 28-day study period.

Although this study enrolled patients in accordance with the DrotAA labeling that limits use to severe sepsis patients with APACHE scores >25 or ≥2 organ dysfunction, the serious bleeding event rate in the heparin group, 3.9%, was similar to that in the DrotAA groups of the PROWESS (3.5%) [1], ENHANCE (6.5%) [5], and ADDRESS (3.9%) [6] studies. The first two studies enrolled patients irrespective of APACHE score whereas the last enrolled severe sepsis patients with a lower risk of death (e.g., APACHE score <25).

However, in some observational studies, the number of bleeding events reported was higher than that seen in this study and other randomized trials There are several possible explanations of this difference. The definition of severe bleeding in the observational studies was not the same as the definition used in this study. The mortality rates were higher in the observational trials and the severity of the sepsis in these patients may also make them more prone to bleeding. Furthermore, randomized clinical trials have methods for prospectively monitoring adverse events and documenting source data verification (the “trial effect”) which retrospective web based observations lack. Finally some of these studies included patients treated off-label (e.g., patients with low platelet count) and this makes comparison with randomized clinical trials difficult [7–11].

There were no significant differences between heparin and placebo groups during the infusion period and the 28-day study period in fatal bleeding events, nonfatal serious bleeding events, and HIT. However, more placebo patients than heparin patients experienced an ischemic stroke. Studies have suggested an association between HIT and stroke (or other thrombotic diseases) [12–14], but analysis identified no such association in this study. This may be because of the small number of events or because most of the HIT diagnoses were clinical (i.e., heparin-platelet factor, antibody measurements were not always used) and thrombocytopenia commonly occurs in septic patients [15]. The incidence of ischemic stroke was lower in patients treated with heparin than in those who did not receive heparin.

Patients who experienced an ischemic stroke were more likely than the overall population to have a history of hypertension, to be on vasopressors, and to be on a ventilator. Stroke incidence in the placebo (DrotAA alone) group in this study was 1.8% which is similar to that seen in PROWESS in the DrotAA group (approximately 1%). By comparison, the stroke incidence in the placebo group of PROWESS was also approximately 1%. In addition, patients in PROWESS who experienced an ischemic stroke were likely to have a history of hypertension, diabetes, and COPD.

The significantly lower incidence of ischemic stroke in the heparin group in this study was unexpected. Activated protein C is reported to be a protective factor for ischemic stroke [16, 17]. To our knowledge, this is the first trial where prophylactic heparin use in severe sepsis patients appeared to prevent ischemic stroke. Most events occurred in the first hours after randomization, similar to the timing of the “rebound phenomenon” (i.e., thrombosis events after heparin cessation) described in the cardiology literature [18]. This suggests that among patients exposed to heparin at baseline, heparin cessation in accordance with the protocol for the placebo group may have led to the observed higher rate of ischemic stroke.

Placebo patients on heparin at baseline also experienced more cardiac events (myocardial infarction and atrial fibrillation), more gastrointestinal events, and more thrombotic events than heparin patients. It is not conclusive whether this increase in serious events was related to discontinuation of commercial heparin, but rebound thrombosis after heparin cessation has been described for both UFH [19, 20] and LMWH [21, 22]. Furthermore, the rebound effect has been linked to a rise in coagulation markers (e.g., thrombin generation) [23]. To our knowledge, this phenomenon of sudden heparin withdrawal has not been described in severe sepsis patients. Given the frequent use of heparin in septic patients for prevention of venous thrombotic events [24–26], and common occurrence of heparin cessation for a variety of reasons in critically ill patients, further study of heparin cessation in severe sepsis is warranted.

It is important to note that this study was powered to detect treatment differences in mortality. Although the safety endpoints included in the analyses reported here were prospectively defined, the observed events are so few in number that our results should only be regarded as hypothesis-generating.

In summary, coadministration of heparin and DrotAA was generally safe. Patients receiving heparin with DrotAA had significantly fewer ischemic strokes than placebo patients. For patients who are about to receive DrotAA for severe sepsis, prophylactic heparin should not be discontinued unless the clinical risks of heparin outweigh any potential benefit.