Evaluating the use of recombinant human activated protein C in adult severe sepsis: Results of the Surviving Sepsis Campaign*

Transcription

1 Evaluating the use of recombinant human activated protein C in adult severe sepsis: Results of the Surviving Sepsis Campaign* Brian Casserly, MD; Herwig Gerlach, MD, PhD; Gary S. Phillips, MAS; John C. Marshall, MD; Stanley Lemeshow, PhD; Mitchell M. Levy, MD Objective: The Surviving Sepsis Campaign developed guidelines for the administration of recombinant human activated protein C in adult severe sepsis. However, it is not clear how these impacted clinical practice or patient outcome. Design and Setting: The Surviving Sepsis Campaign has developed an extensive database to assess the efficacy of the overall effect of its guidelines on clinical practice and patient outcome. From data submitted to the Surviving Sepsis Campaign database from January 2005 through March 2008, we evaluated data regarding the administration of recombinant human activated protein C in adult severe sepsis. Subjects: Data from 15,022 subjects at 165 sites were analyzed. Measurements and Main Results: Of patients with severe sepsis in the database, 1,009 of 15,022 (8%) received recombinant human activated protein C. Recombinant human activated protein C was administered within 24 hrs of the onset of sepsis in 76% (771 of 1009) of patients. Patients in North America (7.1%) and Europe (6.8%) were more likely to receive recombinant human activated protein C than patients in South America (4.2%, p <.001). After adjusting for covariates, the group that received recombinant human activated protein C had a significantly reduced associated hospital mortality (odds ratio 0.76, 95% confidence interval , p <.001). Comparing all the patients who received recombinant human activated protein C to those who did not receive recombinant human activated protein C, the reduction in the adjusted hospital mortality was only statistically significant in patients who had multiorgan dysfunction (odds ratio 0.82, 95% confidence interval , p =.027) vs. those who only had single organ dysfunction (odds ratio 0.78, 95% confidence interval , p =.072). However, in patients who received recombinant human activated protein C before 24 hrs there was a reduction in adjusted hospital mortality in patients with only one organ dysfunction (odds ratio 0.70, 95% confidence interval , p =.03) as well as patients with multiorgan dysfunction (odds ratio 0.78, 95% confidence interval p =.012). There was a statistically significant increase over time in the percentage compliance with the institution of a recombinant human activated protein C administration policy from the first, second, and eighth quarters (47.4%, 46.2%, and 60.7%, respectively) (p <.001). There was also a statistically significant increase in the actual administration rates of recombinant human activated protein C over the same timeline (p <.001), with administration rates of recombinant human activated protein C reaching 9.2% in the last quarter. Conclusions: Recombinant human activated protein C use was associated with a significant improvement in hospital mortality in patients who participated in the Surviving Sepsis Campaign. (Crit Care Med 2012; 40: ) Key Words: activated protein C; mortality; sepsis; Surviving Sepsis Campaign Severe sepsis and septic shock are major healthcare problems, affecting millions of individuals around the world each year (1, 2). Patients with severe sepsis requiring intensive care unit (ICU) admission have high rates of ICU and overall hospital mortality, with estimates ranging *See also p From the Division of Pulmonary, Critical Care and Sleep Medicine (BC), Memorial Hospital of Rhode Island, The Brown Alpert Medical School (BC), Brown University, and Division of Critical Care (MML), Rhode Island Hospital, Providence, RI; Division of Anesthesiology and Critical Care (HG), Vivantes - Klinikum Neukoelln, Berlin, Germany; Center for Biostatistics (GSP), College of Public Health (SL), The Ohio State University, Columbus, OH; and Li Ka Shing Knowledge Institute (JCM), St. Michael s Hospital, University of Toronto, Toronto, Canada. The Surviving Sepsis Campaign was funded, in part, by unrestricted educational grants from Eli Lilly and Edwards Lifesciences. from 18% to 50% (3 5). Protein C, an endogenous anticoagulant, is produced by the liver and activated in the circulation, where it exerts anticoagulant activity through the cleavage and inhibition of factors Va and VIIIA (6). Activated protein C (APC) also appears to exert an indirect anti-inflammatory action by inhibition Dr. Marshall consulted for Eisai, Bayer, Spectral Diagnostics, Eli Lilly, and Daiichi Sankyo, and received honoraria/speaking fees from Biomerieux for travel costs for sepsis lecture. The remaining authors have not disclosed any potential conflicts of interest. For information regarding this article, Brian_Casserly@brown.edu Copyright 2012 by the Society of Critical Care Medicine and Lippincott Williams & Wilkins DOI: /CCM.0b013e31823e9f45 of thrombin generation, and it can also modulate inflammation through interactions with a dedicated cell surface receptor (7). Conversely, levels of endogenous APC are reduced in sepsis and predict poor outcome (8). Recombinant human APC (rhapc) has been approved for the treatment of patients with severe sepsis (9). The Surviving Sepsis Campaign (SSC) guidelines suggest that adult patients with sepsis-induced organ dysfunction associated with the clinical assessment of high risk of death, many of whom will have an Acute Physiology and Chronic Health Evaluation (APACHE II) 25 or multiorgan failure, receive rhapc if there are no contraindications (10). Furthermore, they recommend that adult patients with severe sepsis and a low risk of death, many of whom will have APACHE II <24, do not receive Crit Care Med 2012 Vol. 40, No

2 rhapc. The evidence on which these recommendations are based is primarily derived from two randomized control trials. In 2001, Protein C Worldwide Evaluation in Severe Sepsis (PROWESS), the first double-blind, placebo-controlled trial with rhapc, showed a significant reduction in overall 28-day mortality (6.4%), with a relative reduction of approximately 20% (11), particularly in patients at greater risk of death reflected either in higher APACHE II scores or a greater number of organ failures (12). The ADDRESS trial recruited 2,613 patients judged to have a low risk of death at the time of enrollment. This study was subsequently stopped for futility since rhapc failed to show a reduction in 20-day mortality rates (13). The discovery of effective therapies for patients with severe sepsis has proven challenging, and many currently recommended treatments, such as rhapc, remain the subject of controversy and ongoing study. Although rhapc has been shown to demonstrate a favorable risk/benefit profile in some large clinical studies (14, 15), other studies in different populations failed to confirm this benefit (16, 17). In fact, uptake figures from National Institute for Health and Clinical Excellence show a reduction in the use of rhapc in clinical practice (18). Furthermore, there appears to be a large discrepancy between the protocols used in clinical trials and clinical practice patterns (19). While awaiting additional evidence from ongoing trials (20, 21), we examined the SSC database to analyze the association between treatment with rhapc and outcomes among patients with severe sepsis. METHODS Sites and Patient Selection The process of participation in the SSC is described in detail elsewhere (22). Eligible subjects were those having a suspected site of infection, two or more systemic inflammatory response syndrome criteria (23), and one or more organ dysfunction criteria (22). Clinical and demographic characteristics and time of presentation with severe sepsis criteria were collected for analysis of timebased measures. Time of presentation was determined through chart review and defined in instructions to site data collectors on the Campaign Web site and in educational materials. For patients enrolled from the emergency department, the time of presentation was defined as the time of triage. For patients admitted to the ICU from the medical and surgical wards and for patients in the ICU at the time of diagnosis, the time of presentation was determined by chart review for the diagnosis of severe sepsis. Data Collection Data were entered into the SSC database locally at individual hospitals into pre-established, unmodifiable fields documenting performance data and the time of specific actions and findings. Data stripped of private health information were submitted every 30 days to the secure master SSC server at the Society of Critical Care Medicine (Mount Prospect, IL) via file transfer protocol or as comma-delimited text files attached to submitted to the Campaign s server. Institutional Review Board Approval The global SSC improvement initiative was approved by the Cooper University Hospital Institutional Review Board (Camden, NJ) as meeting criteria for exempt status. The U.S. Department of Health and Human Services Office for Human Research Protections reiterated that quality improvement activities, such as the SSC, often qualify for institutional review board exemption and do not require individual informed consent (24). Analysis Set Construction The analysis set was constructed from the subjects entered into the SSC database from January 2005 through March Inclusion was limited to sites with at least 20 subjects and at least 3 months of subject enrollment. Analysis presented here only includes the first 2 yrs of subjects at each site. Subject age and gender were not collected in deference to country-specific privacy laws. Data were organized by quarter through 2 yrs. The first 3 months that a site entered subjects into the database was defined as the first quarter regardless of when those months occurred from January 2005 through March All baseline characteristics present in the database were included in the risk-adjustment models. Statistical Analysis We compared demographic and clinical characteristics of patients who either received or did not received rhapc. Differences across groups were determined using Pearson s chi-square test of association for categorical variables. A chi-square test for trend was used to determine whether rhapc administration policy and the actual administration increased over quarters 1, 4, and 8. Since the study s goal was not to predict hospital mortality, but rather to identify the role of rhapc administration on survival, we used a risk-factor modeling approach to determine which covariates to add to the random-effects logistic regression model. Only covariates that acted either as a confounder or as an effect modifier were included. A confounder was identified when its addition to the model changed the odds ratio (OR) associated with the rhapc administration by >10% in either direction, without considering statistical significance. A covariate that had a statistically significant interaction (p.05) with rhapc administration was considered to be an effect modifier. Random-effect logistic regression was used since patients are nested within a particular ICU. This method takes into account the variability within and between ICUs and uses this inherent correlation when estimating the standard errors that are used to test model coefficients. The hierarchical nature of the SSC data lends itself to this type of analysis. A sensitivity analysis used a propensity-matched dataset to compare the unadjusted mortality between those subjects who received and did not receive rhapc using random-effects logistic regression. We generated a propensity score (logit) using logistic regression on whether or not the patient received rhapc and 44 predictor variables. Patients that received rhapc were matched 1:1 with patients that did not receive rhapc using a random seed, nearest neighbor (caliper 0.001), and without replacement. All analyses were run using Stata 11.1, Stata Corporation, College Station, TX. RESULTS Between January 2005 and March 2008, 15,775 subjects at 252 qualifying sites were entered into the SSC database. Excluding hospitals that contributed fewer than 20 subjects, the final sample consisted of 15,022 patients at 165 hospitals (a median of 57 and a range of subjects per hospital). Data from up to eight quarters were analyzed from each site. Hospitals contributed data for a mean duration of 15.6 months (median of 14 months). Patient Characteristics and Prescribing Patterns Over the 2-yr period, 6.7% (1,009 of 15,022) of the patients in the database received rhapc (Table 1). Due to the limited clinical information on the patients within the database, it is not known how many patients had a contraindication to the administration of rhapc and were therefore ineligible to receive rhapc. However, patients were seven times more likely to receive low-dose steroids than rhapc. Unadjusted hospital mortality was 1418 Crit Care Med 2012 Vol. 40, No. 5

3 Table 1. Patient characteristics by administration of recombinant human activated protein C Patient Characteristics 37.5% and 34.6% in those did and did not receive rhapc, respectively, and was not statistically significant at the 0.05 level (p =.065). In the rhapc group, a higher percentage of patients had pneumonia (50.6% vs. 44%, p <.001) or abdominal source of infection (27% vs. 20.7%, p <.001) compared to the patients that did not receive rhapc. In the rhapc group, a lower percentage of patients had urinary tract infections as a source of their sepsis (15.4% vs. 21.2%, p <.001) (Table 1). All forms of baseline organ failure were more common in the rhapc group, including hepatic dysfunction (12.4% vs. 10.1%, p =.018). Patients receiving rhapc were less likely to have single organ dysfunction (29.4% vs. 42.8%, p <.001) but No Recombinant Human Activated Protein C Administered, % Recombinant Human Activated Protein C Administered, % Count, n (%) 14,013 (93.3) 1,009 (6.7) Hospital mortality, n (%) 4,848 (34.6) 378 (37.5).065 Location when sepsis identified a, % <.001 Emergency department Ward Intensive care unit Site of infection, % Pneumonia <.001 Urinary tract infection <.001 Abdominal <.001 Meningitis Skin Bone Wound Catheter Endocarditis Device Other infection Baseline acute organ dysfunction, % Cardiovascular <.001 Pulmonary <.001 Renal <.001 Hepatic Hematologic <.001 Number of acute organ dysfunctions, % < Mechanical ventilation, % <.001 Cardiovascular dysfunction (presence of hypotension), % No shock <.001 Shock Lactate > Vasopressors Lactate >4.0 and vasopressors a p value based on Pearson chi-square test of association; b it should be noted that patients in the database received the bundle support in the intensive care unit and not in the location where the sepsis was discovered. more likely to have multiorgan failure (p <.001) (Table 1). In addition, these patients were more likely to be mechanically ventilated (79.2% vs. 50.5%, p <.001) and in shock (89.9% vs. 70.1%, p <.001). However, the way shock was defined appeared to influence the prescribing of rhapc. A smaller percentage of patients in the rhapc group had shock as defined by a lactate >4 (3.7% vs. 5.6%, p <.001). Given the suggestion in the literature that the administration of rhapc within 24 hrs improved patient outcome, we also examined the timing of the administration of rhapc: 76% (771 of 1009) of rhapc was administered within 24 hrs of the onset of sepsis (Table 2). Patients whose sepsis was identified in the emergency department (5.9%) p a were less likely to receive rhapc than patients whose sepsis was identified in the ICU (7.1%) or on the ward (7.7%, p <.001) (Table 2). Next, we examined the effect of the geographic location of the patient on the likelihood of receiving rhapc. Patients in North America (7.1%) and Europe (6.8%) were more likely to receive rhapc than patients in South America (4.2%, p <.001) (Table 2). Changes in Hospital Mortality After adjusting for possible confounders, including shock (no shock, lactate >4 only, vasopressors only, or lactate >4 and vasopressors), pulmonary infection, cardiovascular organ failure, pulmonary organ failure, mechanical ventilation, the interaction between pulmonary organ failure and mechanical ventilation, hepatic organ failure, renal organ failure, measured lactate, blood culture before antibiotics, broad-spectrum antibiotics, glucose control, pneumonia, urinary tract infection, abdominal infection, source of sepsis (emergency department, ward, or ICU), region (North America, Europe, or South America), and median glucose within 24 hrs of time of presentation, patients who received rhapc had a significantly reduced associated hospital mortality (OR 0.76, 95% confidence interval [CI] , p <.001) (Table 3). The reduction in adjusted hospital mortality was restricted to patients who received rhapc within 24 hrs (OR 0.71, 95% CI , p <.001). It was not seen in the patients who received rhapc after 24 hrs (OR 0.95, 95% CI , p =.737). A post-hoc power analysis indicated that approximately 272,000 subjects would have 80% power to detect an OR of 0.95 based on probability of hospital mortality of (adjusted based on regression analysis) for those subjects that did not receive rhapc, and a probability of hospital mortality of (based on the OR of 0.95) for those that received rhapc post 24 hrs. This also assumes that the proportion of subjects that receive rhapc post 24 hrs is 4% while 96% do not receive rhapc. The significance level was set at 0.05 (type 1 error). The absolute difference in hospital mortality was only 1.2% and, as expected, a very large sample size would be required to declare statistical significance. There was no statistical difference between ORs for those who received rhapc within 24 hrs and those after 24 hrs (interaction p =.068, Table 3). Nine hundred twenty (920) Crit Care Med 2012 Vol. 40, No

4 Table 2. Clinical characteristics of the administration of recombinant human activated protein C Count Percent p Timing of administration of rhapc No rhapc 14, Yes, 24 hrs Yes, >24 hrs Total 15, Administration of rhapc by sepsis origin Was rhapc administered, % yes Emergency department 7, <.001 Ward 5, Intensive care unit 1, Administration of rhapc by geographic region North America 8, <.001 Europe 4, South America 1, rhapc, recombinant human activated protein C. Table 3. Hospital mortality and the administration of recombinant human activated protein C using random-effects logistic regression on recombinant human activated protein C Group a Mortality Odds Ratio b,c 95% Confidence Interval p d p e rhapc administered <.001 Not applicable rhapc administered 24 hrs < rhapc administered >24 hrs rhapc, recombinant human activated protein C. a In the first regression model, the risk factor is no rhapc (n = 14,013) vs. administered rhapc (n = 1,009). In the second regression model the risk factor is no rhapc (n = 14,013), rhapc administered 24 hrs (n = 771), vs. rhapc administered >24 hrs (n = 238); b referent group is no rhapc administered; c rhapc logistic regression model odds ratio adjusted for the following variables: shock set (no shock, lactate >4 only, vasopressors only, or lactate >4 and vasopressors), pulmonary infection, cardiovascular organ failure, pulmonary organ failure, mechanical ventilation, the interaction between pulmonary organ failure and mechanical ventilation, hepatic organ failure, renal organ failure, measured lactate, blood culture before antibiotics, broad-spectrum antibiotics, glucose control, pneumonia, urinary tract infection, abdominal infection, source of sepsis (emergency department, ward, or intensive care unit), region (North America, Europe, or South America), and median glucose within 24 hrs of time of presentation; d p value test if the odds ratio is less significantly different from 1.0; e p value test if the two odds ratios differ from each other more likely to receive rhapc than those with no coagulopathy (crude OR 1.57, p <.001). Patients with coagulopathy who received rhapc experienced a reduction in the adjusted risk of hospital mortality (OR 0.68, 95% CI , p =.017). This was also true if they received rhapc within 24 hrs (OR 0.61, 95% CI , p =.009). In the database used for the current study, pulmonary organ failure was defined as bilateral pulmonary infiltrates with a ratio of Pao 2 to Fio 2 of <300 mm Hg. In these patients who received rhapc and had pulmonary organ failure, there was a reduction in the adjusted hospital mortality compared to the patients who did not receive rhapc (OR 0.75, 95% CI , p =.009). This was also true for patients with pulmonary organ failure who received rhapc within 24 hrs (0.70, 95% CI , p =.009). Change in Achievement of the Target of the rhapc of the Management Bundle and its Effects on Administration of rhapc Over Time There was a statistically significant increase in compliance over time (p <.001) with hospital policy for rhapc administration, from 47.4% in the first quarter to 60.7% in the eighth quarter. By comparing the actual administration rates of rhapc over the same timeline, there was a statistically significant increase over time (p =.022), with administration rates of rhapc reaching 9.2% in the last quarter. rhapc subjects were matched 1:1 to 920 non-rhapc subjects using their propensity score. Unadjusted random-effects logistic regression indicated that receiving rhapc reduced hospital mortality by 25% (OR 0.75, 95% CI , p =.005) and is very similar to rhapc administered results shown at the top of Table 3. Comprehensive results are shown in Appendix A. Hospital mortality was reduced in patients with both single (OR 0.72, 95% CI , p =.023) and multiple organ (OR 0.78, 95% CI , p =.005) dysfunction who received rhapc (Table 4). Similarly, in patients who received rhapc before 24 hrs, there was a significant reduction in adjusted hospital mortality in patients with either one organ dysfunction (OR 0.65, 95% CI , p =.009) or multiorgan dysfunction (OR 0.73, 95% CI , p =.002) (Table 4). We also examined the effects of an increased bleeding risk as identified by a thrombocytopenia defined as platelet count of < /µl, or presence of a coagulopathy defined as an international normalized ratio >1.5, or a partial thromboplastin time >60 secs on the administration of rhapc and the effect it had on adjusted hospital mortality. Patients having thrombocytopenia were just as likely to receive rhapc as those who did not have thrombocytopenia (crude OR 1.08, p =.472). Furthermore, patients who had thrombocytopenia and received rhapc had a reduction in their adjusted hospital mortality compared to those who did not receive rhapc (OR 0.51, 95% CI , p <.001). This was also true for patients who received rhapc within 24 hrs (OR 0.46, 95% CI , p <.001). Patients with coagulopathy were DISCUSSION The efficacy of rhapc in serious sepsis, which had been controversial since the initial release of the PROWESS trial results, has now been resolved with the recent PROWESS-SHOCK trial, which was negative (40). This current study was completed and the manuscript accepted for publication before the release of the results of the PROWESS-SHOCK trial. According to news reports, there was no significant difference in 28-day all-cause mortality for rhapc (26.4%) vs. placebo (24.2%) in the 1,696 patients enrolled in this randomized, controlled trial in patients with septic shock. These results are consistent with those found in the ADDRESS trial, which did not show any mortality advantage in patients with a low risk of death, a negative pediatric study, and the 1420 Crit Care Med 2012 Vol. 40, No. 5

5 Table 4. Hospital mortality and the administration of recombinant human activated protein C in subjects with either one or more organ failures using random-effects logistic regression models Subset: Number of Organ Failures a rhapc vs. No rhapc rhapc 24 hrs vs. No rhapc n Odds Ratio for rhapc b most recent negative phase II trial in acute lung injury (13, 16, 17), which have all served to add weight to the argument against use of this agent. The PROWESS study generated controversies regarding changes in study protocol, the effects of early trial stoppage, and the subsequent subset analysis (12). In light of the results of the PROWESS-SHOCK trials, the drug was withdrawn from the market by Eli Lilly. Even before the results of the last trial were made public, the impact of the controversy surrounding rhapc clearly influenced clinical decision making since the use of rhapc worldwide is low (19, 25 27). Clinicians have been left with a mistrust of the data accrued from preclinical models and little sense of the true clinical role of rhapc in severe sepsis. This is further compounded by a lack of a clear understanding of the exact mechanism of the therapeutic action of rhapc (28). The low rates of administration of rhapc (6.7%) in clinical practice are an interesting contrast with the administration of low-dose steroids, which was prescribed to 53% of the eligible population in the same SSC database used in analysis for this present study, despite the fact that similar inconsistencies are seen in data from randomized controlled trials (29). It is interesting that, contrary to the results of the PROWESS-SHOCK trial, in our study, the administration of rhapc in patients with severe sepsis was associated with lower adjusted hospital mortality (Table 3). Our results are similar to recently 95% Confidence Interval p c Condition 15, organ failures 1 organ failure 15, vs. 1 organ failure 14,784 14, organ failures 14, organ failure 14, vs. 1 organ failure rhapc, recombinant human activated protein C. a Referent group is no rhapc administered. b rhapc logistic regression model odds ratio adjusted for the following variables: shock set (no shock, lactate >4 only, vasopressors only, or lactate >4 and vasopressors), pulmonary infection, cardiovascular organ failure, pulmonary organ failure, mechanical ventilation, the interaction between pulmonary organ failure and mechanical ventilation, hepatic organ failure, renal organ failure, measured lactate, blood culture before antibiotics, broad spectrum antibiotics, glucose control, received 20 ml/kg of crystalloid, pneumonia, urinary tract infection, abdominal infection, source of sepsis (emergency department, ward, or intensive care unit), region (North America, Europe, or South America), and median glucose within 24 hrs of time of presentation. c The p value tests if the odds ratio for the 2 organ failure is significantly different from 1 organ failure odds ratio. published observational studies (27, 30). In total, these three trials represent a combined analysis of over 2,700 patients who received rhapc for severe sepsis. These findings represent a substantial accumulation of observational data suggesting a beneficial effect of rhapc in routine clinical practice. These observational data stand in opposition to the negative results of several randomized controlled trials, now in both high- and low-risk patients with severe sepsis and septic shock. Interestingly, in contrast to a recent Italian observational study, the number of patients in our study that received the drug outside of the 48-hr window was very small (19). Comparing uncontrolled clinical experience to the original PROWESS trial is also confounded by differences in the way rhapc is administered. Studies suggest that, in clinical practice, the administration of rhapc is often delayed beyond the first 24 hrs and administered to patients with increased bleeding risk (19, 26). Our study suggests that patients in the SSC database are more likely to receive rhapc within 24 hrs compared to these observational studies examining real-life clinical practice (Table 2). Furthermore, there is a suggestion in the literature that the maximal benefit of rhapc is derived from administration of the drug within 24 hrs of the identification of severe sepsis (15) and consequently, the European Medicines Agency amended its indication for APC to start treatment within 24 hrs (31). The benefit of administration of APC within 24 hrs was also demonstrated in our study (Table 3). This may have contributed to improve adjusted hospital mortality with rhapc compared to other observational studies (19, 26). Patients with increased bleeding risk are usually excluded from randomized controlled trials, but they constitute a significant portion of the clinical population with severe sepsis in medical ICUs. However, this hampers the extrapolation of results from randomized controlled trials to the usual ICU population, which typically has higher mortality and bleeding rates. In an Italian observational study (19), almost 5% of patients with platelet counts below 30,000 mm 3 received rhapc. In our study, patients with thrombocytopenia or evidence of a coagulopathy had higher crude ORs of receiving rhapc than if they did not have thrombocytopenia or coagulopathy. Therefore, patients with thrombocytopenia or coagulopathy were more likely to receive rhapc than patients who had no evidence of thrombocytopenia or coagulopathy. It should be reiterated that the criteria used in our study to define thrombocytopenia and coagulopathy <100,000 mm 3 and International Normalized Ratio 1.5, respectively, do not represent absolute contraindications to the administration of rhapc. In fact, in our study patients with thrombocytopenia who received rhapc had a reduced adjusted hospital mortality compared to patients who had thrombocytopenia but did not receive rhapc. Similarly, patients with a coagulopathy who received rhapc had a reduced adjusted hospital mortality compared to patients with a coagulopathy who did not receive rhapc. These findings are in keeping with a subset analysis of the PROWESS study that revealed a survival advantage in patients with disseminated intravascular coagulation (32). The improved adjusted hospital mortality rate in patients who received rhapc was restricted to patients with multiorgan dysfunction, i.e., two or more (Table 4). The effect of severity of illness was probably a key factor in the reduced adjusted mortality seen in patients with acute lung injury in our study. This finding contrasts with the recent randomized controlled trial demonstrating no survival advantage in patients receiving APC with acute lung injury with APACHE II scores <25 (16). It should be noted that a survival advantage was not demonstrated in patients with only one organ failure, although the p value was.07, and it was demonstrated in patients with only one organ failure Crit Care Med 2012 Vol. 40, No

6 that received rhapc within 24 hrs (Table 4). The proportion of patients receiving rhapc with only one organ dysfunction was high (Table 4), and this replicates the results of the Italian observational study (5.7%) (19). Several observational studies have much higher mortality rates in their treatment arms compared to PROWESS, even after stratifying patients based on the number of organ dysfunctions (19, 26, 27). Our study is clearly limited by its observational design. This is exacerbated by the fact that our assessment was strictly based on the variables within the SSC database and patient demographic data, including sex and age, were not available. Due to the lack of patient demographic data like age and sex, it was not possible to calculate an APACHE II score for individual patients. Other observational studies suggest that rhapc is more commonly administered to younger patients (27, 33). This means our results are vulnerable to confounding by indication. In our study, the possibility of confounding by indication is based on the assumption that physicians may withhold an expensive therapy in patients they thought would be less likely to benefit, e.g., the elderly (34). Another limitation of our study is that data about bleeding complications are not available in the database. Higher rates of bleeding have been identified in routine clinical practice compared to the rates identified in randomized controlled trials (19, 26). In the presence of a higher incidence of bleeding, the mortality reduction effect is more difficult to demonstrate (19, 26). In a recently published Canadian study, the incidence of severe bleeding (10%) was higher than any of the large multicenter trials (PROWESS, 5.3%; ENHANCE, 6.5%; and ADDRESS, 3.9%) as was the overall mortality rate (45.2%) (26). However, in the Canadian study, the use of rhapc was still associated with the mortality reduction if used within 24 hrs of severe sepsis being identified. This finding was similar to our study. A relative contraindication to rhapc was predictive of a serious adverse bleeding event in the Canadian study. Furthermore, in our study, rhapc was prescribed in patients with relative contraindications, like coagulopathy and/or thrombocytopenia, and rhapc was still associated with improved adjusted hospital mortality. All of the weakness stated above may account for the difference in results between our current analysis and the recently released randomized controlled trial. It is worth noting, however, that the hospital mortality rates in our study were 34.6% (rhapc administered) and 37.5% (rhapc not administered). These rates are 10% higher than those reported for both groups in the PROWESS-SHOCK trial, raising the possibility that the PROWESS-SHOCK trial may have underpowered as a result of the low mortality rate seen in the placebo group. Given the withdrawal of the drug from the market, this speculation will remain just that, and rhapc will be added to the long list of failed immunomodulatory agents for critically ill patients with sepsis (35). The discrepancy between the rigors of clinical trial and real-life clinical practice is well known (36, 37). Secondary analysis, such as this study, of the individual elements of the resuscitation and management bundles in the SSC may provide insight into the value of the various recommendations that constitute this performance-improvement program. The SCC implementation program does appear to successfully drive a change in clinical practice (22). This is substantiated by the finding in our study of increasing compliance being demonstrated over time with the rhapc metric of the management bundle, which also resulted in significant increase in the administration of rhapc over time. In conclusion, the SSC performanceimprovement initiative resulted in a significant increase in the administration of rhapc, highlighting the effect of the Campaign on clinical practice patterns. The administration of rhapc, although infrequent, was associated with a significant improvement in adjusted hospital mortality. REFERENCES 1. Angus DC, Linde-Zwirble WT, Lidicker J, et al: Epidemiology of severe sepsis in the United States: Analysis of incidence, outcome, and associated costs of care. Crit Care Med 2001; 29: Martin GS, Mannino DM, Eaton S, et al: The epidemiology of sepsis in the United States from 1979 through N Engl J Med 2003; 348: Alberti C, Brun-Buisson C, Burchardi H, et al: Epidemiology of sepsis and infection in ICU patients from an international multicentre cohort study. Intensive Care Med 2002; 28: Brun-Buisson C, Meshaka P, Pinton P, et al: EPISEPSIS: A reappraisal of the epidemiology and outcome of severe sepsis in French intensive care units. Intensive Care Med 2004; 30: Finfer S, Bellomo R, Lipman J, et al: Adultpopulation incidence of severe sepsis in Australian and New Zealand intensive care units. Intensive Care Med 2004; 30: Castellino FJ, Ploplis VA: The protein C pathway and pathologic processes. J Thromb Haemost 2009; 7: Mosnier LO, Zlokovic BV, Griffin JH: The cytoprotective protein C pathway. Blood 2007; 109: Shorr AF, Bernard GR, Dhainaut JF, et al: Protein C concentrations in severe sepsis: An early directional change in plasma levels predicts outcome. Crit Care 2006; 10:R92 9. de Jonge E: Drotrecogin alfa Eli Lilly. IDrugs 2002; 5: Dellinger RP, Levy MM, Carlet JM, et al: Surviving Sepsis Campaign: International guidelines for management of severe sepsis and septic shock: Crit Care Med 2008; 36: Bernard GR, Vincent JL, Laterre PF, et al: Efficacy and safety of recombinant human activated protein C for severe sepsis. N Engl J Med 2001; 344: Warren HS, Suffredini AF, Eichacker PQ, et al: Risks and benefits of activated protein C treatment for severe sepsis. N Engl J Med 2002; 347: Abraham E, Laterre PF, Garg R, et al: Drotrecogin alfa (activated) for adults with severe sepsis and a low risk of death. N Engl J Med 2005; 353: Benson K, Hartz AJ: A comparison of observational studies and randomized, controlled trials. N Engl J Med 2000; 342: Vincent JL, Bernard GR, Beale R, et al: Drotrecogin alfa (activated) treatment in severe sepsis from the global open-label trial ENHANCE: Further evidence for survival and safety and implications for early treatment. Crit Care Med 2005; 33: Liu KD, Levitt J, Zhuo H, et al: Randomized clinical trial of activated protein C for the treatment of acute lung injury. Am J Respir Crit Care Med 2008; 178: Nadel S, Goldstein B, Williams MD, et al: Drotrecogin alfa (activated) in children with severe sepsis: A multicentre phase III randomised controlled trial. Lancet 2007; 369: Eve RL, Duffy MR: Should we be using activated protein C to treat severe sepsis? Br J Hosp Med (Lond) 2009; 70: Bertolini G, Rossi C, Anghileri A, et al: Use of Drotrecogin alfa (activated) in Italian intensive care units: The results of a nationwide survey. Intensive Care Med 2007; 33: University of Versailles: Activated protein C and corticosteroids for human septic shock (APROCCHS). Available at: clinicaltrials.gov/ct2/show/nct Accessed May 10, Finfer S, Ranieri VM, Thompson BT, et al: Design, conduct, analysis and reporting of 1422 Crit Care Med 2012 Vol. 40, No. 5

7 a multi-national placebo-controlled trial of activated protein C for persistent septic shock. Intensive Care Med 2008; 34: Levy MM, Dellinger RP, Townsend SR, et al: The Surviving Sepsis Campaign: Results of an international guidelinebased performance improvement program targeting severe sepsis. Intensive Care Med 2010; 36: Levy MM, Fink MP, Marshall JC, et al: 2001 SCCM/ESICM/ACCP/ATS/SIS International Sepsis Definitions Conference. Crit Care Med 2003; 31: Peer Education & Evaluation Resource Center: Building blocks to peer program success: 7.6 Evaluating peer programs: Protection of human subjects and evaluation confidentiality and creating boundaries in the workplace. Available at: files/7humansubjectsprotectionandevaluation. pdf. Accessed May 10, Ferrer R, Artigas A, Levy MM, et al: Improvement in process of care and outcome after a multicenter severe sepsis educational program in Spain. JAMA 2008; 299: Kanji S, Perreault MM, Chant C, et al: Evaluating the use of Drotrecogin alfa (activated) in adult severe sepsis: A Canadian multicenter observational study. Intensive Care Med 2007; 33: Lindenauer PK, Rothberg MB, Nathanson BH, et al: Activated protein C and hospital mortality in septic shock: A propensitymatched analysis. Crit Care Med 2010; 38: Levi M: Antithrombin in sepsis revisited. Crit Care 2005; 9: Snijders D, Daniels JM, de Graaff CS, et al: Efficacy of corticosteroids in communityacquired pneumonia: A randomized doubleblinded clinical trial. Am J Respir Crit Care Med 2010; 181: Ferrer R, Artigas A, Suarez D, et al: Effectiveness of treatments for severe sepsis: A prospective, multicenter, observational study. Am J Respir Crit Care Med 2009; 180: European Medicines Agency: Final Scientific Discussion Xigris. EMEA/H/C/396/0013. April 21, Available at: ema.europa.eu/ema/index.jsp?curl=pages/ medicines/human/medicines/000396/ human_med_ jsp&mid=wc0b01 ac058001d124&jsenabled=true. Accessed March 7, Dhainaut JF, Yan SB, Joyce DE, et al: Treatment effects of drotrecogin alfa (activated) in patients with severe sepsis with or without overt disseminated intravascular coagulation. J Thromb Haemost 2004; 2: Steingrub JS, Cheatham ML, Woodward B, et al: A prospective, observational study of Xigris Use in the United States (XEUS). J Crit Care 2010; 25:660.e9 660.e Kalil AC: Does recombinant activated protein C work in patients with severe sepsis? Crit Care Med 2010; 38: Federal Drug Administration: FDA Drug Safety Communication: Voluntary market withdrawal of Xigris [drotrecogin alfa (activated)] due to failure to show a survival benefit. Available at: Drugs/DrugSafety/ucm htm. Accessed November 3, Reade MC, Huang DT, Bell D, et al: Variability in management of early severe sepsis. Emerg Med J 2010; 27: Rothwell PM: External validity of randomised controlled trials: To whom do the results of this trial apply?. Lancet 2005; 365:82 93 Appendix A: Propensity score matching on whether the patient received rhapc Propensity score matching was used to validate the adjusted analysis using multivariable random-effects logistic regression as a way to judge the sensitivity of our primary analysis. The propensity to receive rhapc was developed using the 44 variables listed in Appendix Table 1. We generated a propensity score (logit) using logistic regression matching 1:1 patients that receive rhapc to those who did not using a random seed, nearest neighbor (caliper 0.001), and without replacement (39). Appendix Table 1 illustrates the mean of the variables within the two groups (rhapc vs. no rhapc) both before and after matching along with the reduction in the absolute standardized bias along with a t test and p value. If a particular value of a variable is not mentioned in the table it is due to it being the referent group. For example Table 1 mentions location where sepsis was identified as either ward or ICU and not ED. ED is the referent group for this variable and is thus not included in the table. The overall reduction in bias decreases from a median of 12.3 to 1.8 and is summarized in Appendix Table 2. Nine hundred twenty out of 1,009 patients that received rhapc were matched to 920 out of the 14,012 subjects that did not receive rhapc (Appendix Table 3 and Appendix Fig. 1). Random-effects logistic regression results are shown in Appendix Table 4 for 3 scenarios. The first is the unadjusted results on all 15,022 subjects and shows for those 1,009 subjects that received rhapc, the odds of mortality are 21% higher than those that did not receive rhapc (p =.009). The second model is the same as the first but is adjusted by the covariates shown in footnote 3 of the table. Here the odds of mortality are reduced by 24% if the patient receives rhapc relative to not receiving rhapc (p <.001). The third model is an unadjusted analysis using only the 1,840 subjects (920 in each group) that were propensity score matched. Here the odds of mortality are reduced by 25% if the patient receives rhapc relative to not receiving rhapc (OR 0.75, 95% CI ; p =.005). The results of the adjusted multivariable model and the propensity matched data are almost identical to each other thus validating the risk adjusted model. Analyses run using Stata 11.1 (Stata Corporation, College Station, TX). Crit Care Med 2012 Vol. 40, No

8 Appendix Table 1. Propensity score variables before and after matching and reduction in bias Variable Sample matching rhapc Mean No rhapc Standardized bias, % Reduct sd bias t statistic t test p Ward Before <.001 After Intensive care unit Before After Europe Before After South America Before <.001 After Cardiovascular organ fail, yes Before <.001 After Lactate >4 in 6 hrs Before <.001 After Hypotensive, yes Before <.001 After Mean arterial pressure 65 mm Hg, with pressors Before After Mean arterial pressure 65 mm Hg, N/A Before <.001 After Received steroid with 24 hrs Before <.001 After Received steroid after 24 hrs Before <.001 After Received steroid, N/A Before <.001 After Lactate > 4 hrs, only Before After Hypotensive, only Before <.001 After Lactate > 4 hrs and hypotensive Before <.001 After Median glucose Before After Pneumonia, yes Before <.001 After Urinary tract infection, yes Before <.001 After Abdominal, yes Before <.001 After Meningitis, yes Before After Catheter, yes Before After Device, yes Before After Other infection, yes Before After Renal organ failure, yes Before <.001 After Hepatic organ failure, yes Before After Hematologic organ fail, yes Before <.001 After Mechanical ventilation, yes Before <.001 After Pulmonary organ fail, yes Before <.001 After Interaction: Pulmonary organ failure and mechanical ventilation Before <.001 After Two baseline organ failures Before After Three baseline organ failures Before <.001 After >.999 Four baseline organ failures Before <.001 After Five baseline organ failures Before <.001 After >.999 Hyperthermia, yes Before After Crit Care Med 2012 Vol. 40, No. 5

9 Appendix Table 1. Propensity score variables before and after matching and reduction in bias (Continued) Variable Sample matching rhapc Mean No rhapc Standardized bias, % Reduct sd bias t statistic t test p Chills with rigors, yes Before After >.999 Tachycardia, yes Before <.001 After Tachypnea, yes Before After Leukopenia, yes Before <.001 After Hyperglycemia, yes Before After Measure lactate with 6 hrs, yes Before <.001 After Blood cultures before antibiotics, yes Before After Broad spectrum antibiotic, yes Before After Glucose control, yes Before After N/A, not applicable; rhapc, recombinant human activated protein C. Appendix Table 2. Summary statistics of the reduction in the absolute value of the bias Sample Mean sd Median Minimum Maximum Before matching After matching Appendix Table 3. Propensity score 1:1 matching without replacement using a caliper < Received rhapc Unmatched Matched Total No 13, ,012 Yes ,009 Total 13,181 1,840 15,021 rhapc, recombinant human activated protein C. Distribution of Propensity Scores Unmatched rhapc [N = 89] Unmatched no rhapc [N = 13092] Matched rhapc [N = 920] Matched no rhapc [N = 920] Propensity Score Appendix Figure 1. Distribution of the propensity score by matched and unmatched patients that either received or did not receive recombinant human activated protein C (rhapc). Crit Care Med 2012 Vol. 40, No

10 Appendix Table 4. Hospital mortality random-effects logistic regression on rhapc rhapc administered N Mortality Odds Ratioa 95% Confidence Interval p b Unadjusted model 15, Adjusted modelc 15, <.001 Unadjusted model using 1:1 propensity score matched observations 1, rhapc, recombinant human activated protein C. a Referent group is no rhapc administered; b p value test if the odds ratio is less significantly different from 1.0; c rhapc logistic regression model odds ratio adjusted for the following variables: shock set (no shock, lactate > 4 only, vasopressors only, or lactate > 4 and vasopressors), pulmonary infection, cardiovascular organ failure, pulmonary organ failure, mechanical ventilation, the interaction between pulmonary organ failure and mechanical ventilation, hepatic organ failure, renal organ failure, measured lactate, blood culture before antibiotics, broad spectrum antibiotics, glucose control, pneumonia, urinary tract infection, abdominal infection, source of sepsis (emergency department, ward, or intensive care unit), region (North America, Europe, or South America), and median glucose within 24 hrs of time of presentation Crit Care Med 2012 Vol. 40, No. 5