Early drotrecogin alpha (activated) administration in severe sepsis is associated with lower mortality: a retrospective analysis of the Canadian ENHANCE cohort

Descriptive statistics, including mean, standard deviation, count, and percentages, were performed on baseline characteristics and disease severity markers for the Canadian sample. Baseline demographics, disease severity, and clinical variables that were associated with mortality were compared in a bivariate analysis between survivors and non-survivors. One-sided Wilcoxon rank sums test were used for continuous variables and Chi-square or Fisher's exact tests were used for categorical variables.

We wished also to explore whether there was a relation between the patient's apparent clinical status (ie. stable, worsening, or improving), the use of DrotAA, and outcome. However, because there is no consensus on these clinical definitions of stability and because different physicians view stable as either good (i.e. same category as improving) or bad (same category as worsening), we chose to consider a single composite clinical status category, namely stable or worsening, as there is an emerging clinical consensus that stable portends a worse outcome than improving [12].

Separate analyses were performed for time to treatment and mortality rate on the stable/worsening clinical status subgroup. Twenty-eight-day mortality was compared between patients who received early treatment (defined as less than 24 hours from first sepsis-induced organ dysfunction to start of infusion of DrotAA) and those who received late treatment (24 hours or more to start of infusion of DrotAA). The analysis was repeated for each category of clinical status using Cochran-Mantel-Haenszel statistics. The Breslow-Day test was used to examine time to treatment and age interactions.

To identify possible predictors of mortality, multivariate logistic regression was used while adjusting for multiple co-variates. Variables identified to be statistically significant in the bivariate analysis were fitted into a logistic regression model. Correlations between the variables that were retained in the model using the stepwise method with a significance level of 0.05 were examined for multicollinearity.

We also sought to explore whether the timing by clinically important intervals would stand as an independent predictor of survival. Following extensive exploratory analysis, we identified that a six hour cutoff increment appeared to make a difference in terms of survival, and so the variable for time from first sepsis-induced organ dysfunction to start of study drug was divided by six in order to determine if there were differences in outcome by six-hour intervals from onset of organ dysfunction to treatment. Age was classified into four groups (≤49, 50 to 64, 65 to 74, and ≥75 years of age) to aid in clinical interpretation.

All statistical tests were performed at an alpha level of 0.05. There was no adjustment for multiple comparisons.

The Canadian cohort from the international ENHANCE study contained 305 patients enrolled at 21 sites across Canada (see list of investigators), 141 (46%) of whom received DrotAA within the first 24 hours and 164 (54%) of whom received DrotAA after 24 hours. Patient demographics and disease severity are summarized in Table 1. Half of the patients (50.2%) had organ dysfunction in each of the cardiovascular, renal, and respiratory systems, indicating that this was a severely ill cohort.

Exploratory analysis of the differences between survivors (n = 236) and non-survivors (n = 69) for the Canadian cohort revealed the significant or clinically important variables listed in Table 2. These variables were then used in a bivariate analysis of survivors and non-survivors and the results are presented in Table 3. Age, an increase in serum creatinine from baseline to day 1, a decrease in platelet count from baseline to day 1, the number of organ dysfunctions at baseline, and pre-infusion Acute Physiology and Chronic Health Evaluation (APACHE) II score were significantly associated with 28-day mortality. Thus, non-survivors at 28 days were on average older, had more severe disease at baseline and had greater early progression of renal and hematological dysfunction.

Although time to treatment was not statistically significant in this bivariate analysis, the variable was retained in subsequent analyses based on results from the full ENHANCE trial, which demonstrated that time to treatment within 24 hours from first organ failure was significantly associated with lower mortality. In addition, we felt that timing was an important clinical issue, and that there might be other factors confounding the association between mortality and treatment.

A number of significant mortality predictors were identified by the bivariate analysis: age, baseline Sequential Organ Failure Assessment (SOFA) respiratory score, baseline prothrombin time, number of organ failures at baseline, pre-infusion APACHE II score, baseline SOFA cardiovascular score, change in platelets, and change in creatinine.

The interaction between age and early versus late treatment with DrotAA was examined in the predefined clinical subgroup stable or worsening (n = 242). For these patients, early treatment with DrotAA was associated with lower mortality regardless of age (P = 0.0409; Table 4).

Table 5 displays the odds ratio estimates of dying for those variables found to be significant independent predictors of 28-day mortality. The results suggest that when treated with DrotAA, patients aged 49 years and younger have lower odds of mortality than those 65 years and older (odds ratio: 0.25 to 0.27), and that each six-hour delay in starting DrotAA after the first sepsis-induced organ dysfunction was associated with a 23% higher odds of mortality (odds ratio: 1.23). A rise in serum creatinine of 50 μmol/L from baseline to day 1 increased the odds of mortality by 32% (odds ratio: 1.32), whereas an increase in platelet count of 50 × 10⁹/L from baseline to day 1 decreased the odds of mortality by 34% (odds ratio: 0.66).

Bleeding was the only serious adverse event related to DrotAA administration. Serious bleeding events were defined as life-threatening or disabling events, intracranial hemorrhage, or bleeding that required transfusions for two consecutive days. A descriptive analysis comparing the Canadian and non-Canadian cohorts revealed that in the Canadian cohort, bleeding events were relatively few and less frequent compared with non-Canadian sites. For example, during the 96-hour infusion period, 2% of Canadian patients had serious hemorrhage (versus 3.8% of non-Canadian patients), and 0.3% had intracranial hemorrhage (versus 0.7% of non-Canadian patients). In the 28-day study period (which included the 96-hour infusion), serious bleeds occurred in 3.9% of Canadian patients (6.9% of non-Canadian patients).

Our initial analysis revealed that in patients administered DrotAA, two baseline measures of severity – mean number of organ dysfunctions and APACHE II score – and two dynamic laboratory measures – serum creatinine and platelet count – were associated with 28-day mortality. Following a logistic regression for independent predictors, only a subset of the variables remained statistically significant: age, first day change in serum creatinine, first day change in platelet count, and time to treatment.

Most clinical studies have utilized 'static' characteristics obtained at baseline to predict mortality. The finding in the current analysis that early changes in serum creatinine and platelet count were independent predictors of mortality has direct clinical implications. In the rapidly changing environment of severe sepsis, it is helpful to identify predictors that reflect evolving rather than simply baseline status. The idea that identification of early dynamic predictors of mortality could be very useful in guiding therapy has recently been suggested from an analysis of dynamic coagulation changes in sepsis [18]. Dhainaut and colleagues found that continued or worsening coagulopathy during the first day of severe sepsis was associated with new organ failure and increased mortality at 28 days [18]. Data from an integrated sepsis database also provides evidence that serum creatinine change within the first post-baseline day is predictive of outcome [8]. Other outcome predictors have been identified in severe sepsis [7, 25]. Micek and colleagues observed that the number of organ dysfunctions (adjusted odds ratio: 2.30) and inappropriate antimicrobial treatments (adjusted odds ratio: 15.5) were independent predictors of mortality [7]. In addition, Johnston found that baseline platelet counts less than 80 × 10⁹/L carried an odds ratio of death of 2.05 and a baseline prothrombin time greater than 30 was associated with an odds ratio of death of 2.88 in severe sepsis [25].

We observed that the risk of mortality based on odds ratios increased by 23% with each six-hour delay in receiving DrotAA. Because this was seen only in the retrospective subgroup analysis, it could be merely reflective of chance. On the other hand, there could be subsets of patients who will benefit more from earlier treatment. This possibility is supported by others who also found value in early treatment [4, 26]. In an analysis of the full ENHANCE sample, the adjusted odds of death were 21.8% higher for patients treated later than 24 hours after the first sepsis-induced organ dysfunction, compared with treatment within 24 hours, although the association was not statistically significant when adjusted for age [6]. In the current analysis, when a further breakdown of age was used and the time to treatment effect was based on odds ratios for death, time to treatment remained statistically significant after adjusting for age. However, as this is a post hoc, exploratory analysis, this should be considered as hypothesis-generating only, and interpretation should be made with caution.

Others have also observed value in early treatment of severe sepsis with DrotAA. A recent meta-analysis of five clinical trials using DrotAA found that patients treated within 24 hours of first sepsis-induced organ dysfunction had significantly higher 28-day survival compared with those treated later than 24 hours (76.4% versus 73.5%) [27]. These investigators also found that in their adjusted model, logistic regression analysis suggested that treatment with DrotAA within 24 hours of organ dysfunction was associated with lower odds of death (23%), compared with treatment after 24 hours. In a prospective study of mortality predictors by Micek and colleagues, the time to treatment with DrotAA in survivors was significantly earlier (16.9 hours, standard deviation (SD) 11.9 hours) compared with non-survivors (25.5 hours, SD 28.0 hours) [7]. Most recently, Kanji and colleagues, in a Canadian observational study, also found that early treatment with DrotAA (within 12 hours of first organ dysfunction) was associated with a lower risk of death [28].

International guidelines for the management of severe sepsis suggest the use of DrotAA for patients at high risk of death [13]. Furthermore, results from the current exploratory evaluation suggest that treatment with DrotAA should be initiated as soon as possible in all patients with sepsis-induced organ dysfunction who meet criteria for its administration. However, delays in initiating treatment with DrotAA in Canada are common – in a registry of 4087 Canadian intensive care patients (1269 with severe sepsis), the average time to treatment with DrotAA after identification of organ dysfunction was 1.3 days [29]. Others have highlighted potentially harmful consequences of delayed therapy with DrotAA [27]. A meta-analysis in 4459 patients demonstrated that delays in treatment were directly correlated with the number of organ dysfunctions, the need for mechanical ventilation or vasopressors, or recent surgery [27]. In addition, that analysis found an inverse relationship between baseline APACHE II score and time to treatment [27]. Some practice guidelines [20] recommend that DrotAA be withheld when the clinical course is uncertain, which may further contribute to worsening clinical status and delays in therapy. Some clinicians may thus wait for an APACHE II score to exceed 25 before starting DrotAA, as suggested by the Food and Drug Administration [25], despite the fact that APACHE scoring has been validated as a static baseline measure only in populations and not in individual patients, and despite the fact that multiple organ dysfunction may be improving or worsening in the face of no or minor changes in APACHE score. Inherent in such delays in escalating therapy may be the impression held by some that apparent clinical stability is a sign that the patient with severe sepsis may soon begin to improve with the current level of therapy.

The concept of a false sense of security due to apparent clinical stability may be particularly important when one considers that the clinician's impression of what constitutes clinical stability is usually based on 'macroscopic' observations (e.g. blood pressure, urine output, need for vasopressors etc), whereas the underlying pathophysiology of severe sepsis is best defined in terms of a pathologic microcirculation [17, 30]. For example, it has been demonstrated that microvascular dysfunction can persist in at least 40% of patients who appear clinically stable [11]. Apparent clinical stability may thus not reflect underlying pathophysiology nor predict survival in severe sepsis. Deficits in microvascular function are present at the earliest stages of sepsis and improved tissue oxygen delivery relative to tissue oxygen demand may prevent progression of sepsis [22]. Treatment addressing 'hidden' microvascular dysfunction may be one potential explanation of why clinically stable patients appear to benefit from early treatment with DrotAA [31]. Treatment with DrotAA may reduce microvascular inflammation and coagulation, and thus improve function and tissue oxygenation.

Ideally therefore, aggressive medical intervention for the reversal of organ dysfunction should be initiated by addressing microvascular dysfunction, inflammatory response, and coagulopathy. However, escalation of therapy is usually based on clinical status and is thus often reserved for patients who are overtly deteriorating or at immediate risk of death. Clinicians may therefore hold off intervention with DrotAA or other aggressive resuscitation measures if they interpret that a patient is 'stable', and may wait a significant amount of time in order to determine a 'clear' increased or increasing risk of death.

This retrospective subset analysis of the open-label ENHANCE trial has a number of other limitations because of the post hoc and subgroup analyses. A sample size of 305 homogeneous patients may have biased the results because of a lack of power. Furthermore, predictor variables were collected during the administration of DrotAA, which may confound the prognostic validity of mortality predictors with the effects of treatment with DrotAA. In addition, the observed improvement in survival associated with early use of DrotAA may have been a marker for patients who received earlier and more aggressive resuscitation in the first place (i.e. early goal-directed therapy [11]). An analysis of the original global ENHANCE data suggests that this is unlikely, as well as the fact that this was a relatively new concept during the enrolment period. In the global ENHANCE study, patients treated later (>24 hours) were more likely to be male and older and to have had recent surgery. Patients treated later most likely had more severe disease as suggested by a greater need for vasopressors and mechanical ventilation, and a greater number of organ dysfunctions, and total SOFA score [6]. Patients treated earlier had significantly lower 28-day all-cause mortality (22.9%) than those treated later (27.4%) [6].