Low Levels of Protein C Are Associated With Poor Outcome in Severe Sepsis

clinical investigations in critical care Low Levels of Protein C Are Associated With Poor Outcome in Severe Sepsis* S. Betty Yan, PhD; Jeffrey D. Helterbrand, PhD; Daniel L. Hartman, MD, FCCP; Theressa J. Wright, MD; and Gordon R. Bernard, MD, FCCP

Study objective: To investigate whether protein C levels predict 30-day mortality rate, shock status, duration of ICU stay, and ventilator dependence in patients with sepsis. Design: Retrospective analysis of a subset of a previously published, prospective, randomized, double-blind, placebo-controlled trial (“Effects of Ibuprofen on the Physiology and Survival of Patients With Sepsis” [ISS]). Setting: A multicenter study performed in the United States and Canada (seven sites). Patients: Seventy hospitalized patients with acute severe sepsis and failure in one or more organs at entry into the ISS trial. Measurements and Main Results: Blood samples were obtained from all patients at baseline and at 20, 44, 72, and 120 h after the initiation of study drug (ibuprofen or placebo) infusion. Data obtained at these times included platelet count, prothrombin time, and partial thromboplastin time. The results described in this article are based on a subset of the total ISS population for whom additional coagulation assays were performed on the blood samples obtained at baseline and 44 h. These assays included protein C antigen, D-dimer, and fibrinogen levels. A total of 63 of the 70 patients (90%) studied in this report had acquired protein C deficiency at entry to the ISS trial (baseline). The presence and severity of acquired protein C deficiency were associated with poor clinical outcome, including lower survival rate, higher incidence of shock, and fewer ICU-free and ventilator-free days. Conclusions: Acquired protein C deficiency may be useful in predicting clinical outcome in patients with sepsis. Clinical studies are warranted to determine whether the replacement of protein C in sepsis patients may improve outcome. (CHEST 2001; 120:915–922) Key words: acquired protein C deficiency; protein C; septic shock; severe sepsis Abbreviations: APACHE ⫽ acute physiology and chronic health evaluation; DIC ⫽ disseminated intravascular coagulation; ISS ⫽ Ibuprofen in Sepsis Study; PAI ⫽ plasminogen activator inhibitor; PT ⫽ prothrombin time; PTT ⫽ partial thromboplastin time; TAFI ⫽ thrombin-activatable fibrinolysis inhibitor

is associated with endothelial cell activation, S epsis a hemostatic profile characterized by activation of the coagulation pathway, and subsequent activation of the fibrinolytic system, which then is followed by the inhibition of the fibrinolytic system.1– 6 The *From Eli Lilly and Company (Drs. Yan, Helterbrand, and Wright), Lilly Research Laboratories, Indianapolis, IN; Pfizer Global Research & Development (Dr. Hartman), Ann Arbor, MI; and the Department of Pathology (Dr. Bernard), Division of Allergy, Pulmonary and Critical Care Medicine, Vanderbilt University School of Medicine, Nashville, TN. Manuscript received August 1, 2000; revision accepted January 26, 2001. Drs. Yan, Helterbrand, and Wright are employees of Eli Lilly and Company. Dr. Hartman was an employee of Lilly during the collaboration of this study. This study is a scientific collaboration between Lilly and Dr. Bernard.

resulting procoagulant state may manifest clinically as a coagulation-dominant disseminated intravascular coagulation (DIC) producing multiple organ sysFor editorial comment see page 699 tem failure.5– 8 Alterations in the levels of various mediators of coagulation and fibrinolysis have been Funds needed for conducting the coagulation assays at the University of Vermont were provided by Eli Lilly and Co. The ibuprofen clinical trial was supported by grant HL43167 from National Heart, Lung, and Blood Institute of the National Institutes of Health. Correspondence to: S. Betty Yan, PhD, Eli Lilly and Company, Drop Code 0522, 307 E McCarty St, Indianapolis, IN 46285; e-mail: yan sau chi [email protected] CHEST / 120 / 3 / SEPTEMBER, 2001

915

reported to be associated with negative outcome in patients with sepsis.9,10 Given its pivotal role in the coagulation and fibrinolytic systems, protein C has been of particular interest as a factor in the hemostatic abnormalities of patients with sepsis. Protein C levels have been reported to be predictive of sepsis outcome.4,5,7,11–13 The possible role of the protein C pathway in sepsis has been reviewed extensively.14,15 The Ibuprofen in Sepsis Study (ISS) provided an opportunity to assess data on hemostatic variables in severe sepsis in a large patient population.16 We evaluated the association of protein C levels and other hemostatic variables with clinical outcomes, including mortality, in a subset of patients from the ISS.

Table 1—Brussels Table of Organ Dysfunction* Organ/Variable Cardiovascular/ systolic BP, mm Hg Pulmonary/ Pao2/Fio2 ratio Coagulation/ platelets, No./mm3† Renal/creatinine, mg/dL Hepatic/bilirubin, mg/dL

Moderate

Severe

Extreme

ⱕ 90; not fluid responsive

ⱕ 90

ⱕ 90

300–201

200–101

ⱕ 100

80–51

50–21

ⱕ 20

2.0–3.4

3.5–4.9

ⱖ 5.0

2.0–5.9

6.0–11.9

ⱖ 12

*Adapted from Bernard et al.16 Fio2 ⫽ fraction of inspired oxygen †Thousands of platelets per cubic microliter.

Materials and Methods The ISS was a randomized, double-blind, placebo-controlled multicenter study of IV ibuprofen in 455 patients with severe sepsis. The study was approved by the institutional review board at each of the seven centers in the United States and Canada, and consent was obtained from each patient or their next of kin. Patient characteristics and the main study results have been reported elsewhere.16 Patients were enrolled into the study from October 1989 to March 1995. Patients with a known or suspected site of serious infection had to meet all of the following criteria: core temperature, ⱖ 38.3°C or ⬍ 35.5°C; heart rate, ⱖ 90 beats/min in the absence of ␤-blocker treatment; and respiratory rate, ⱖ 20 breaths/min (or minute ventilation, ⬎ 10 L/min if the patient requires mechanical ventilation). In addition, patients had to exhibit dysfunction of at least one of the following organ systems: cardiovascular, renal, ARDS/pulmonary, or CNS. Investigational treatment consisted of IV ibuprofen (10 mg/kg q6h for eight doses) or placebo given in a similar fashion. Patients were observed for clinical outcomes, with 30-day all-cause mortality as the primary outcome measure. Secondary measures included duration of ICU stay, the presence or absence of mechanical ventilation, and the presence or absence of septic shock. Secondary measures were described in terms of ICU-free days, ventilator-free days, and shock-free days. This methodology has been described previously.17 Organ failure-free days were calculated over the 30-day study period. For each patient, the number of organ failure-free days was calculated as the number of days a patient was both alive and free of a particular organ failure. Definitions of organ failure are shown in Table 1.17 The study did not demonstrate a statistically significant effect of ibuprofen on ICU-free days, ventilator-free days, shock-free days, or survival.16 All patients enrolled in the parent study had an APACHE (acute physiology and chronic health evaluation) II score calculated at study entry (average, 11 h from sepsis onset). Blood samples were obtained from all patients at baseline and at 20, 44, 72, and 120 h after study entry. Data obtained at these times included platelet count, prothrombin time (PT), and partial thromboplastin time (PTT). The results described in this article are based on a subset of the total ISS population for whom additional coagulation assays were performed on the blood samples obtained at baseline and at 44 h. The samples were taken as a subset of the database without regard to the types of pathogen or any other etiology and included patients treated with both placebo and ibuprofen. For the purpose of selecting patients to be included in the coagula916

tion analyses, patients were categorized into one of the following eight groups based on these measures: (1) abnormal PTT only; (2) abnormal PT only; (3) abnormal platelets only; (4) abnormal PT and platelets; (5) abnormal PTT and platelets; (6) abnormal PT and PTT; (7) PT, PTT, and platelets normal (“all normal”); and (8) PT, PTT, and platelets abnormal (“all abnormal”). Abnormal values were defined as a platelet count of ⬍ 140,000/ mL, PT ⱖ 13.8 s, and PTT ⱖ 39 s, which are values that are generally consistent with normal range boundaries in local hospital clinical laboratories. Ten patients from each of the groups except for the all-abnormal group were selected randomly for inclusion in the coagulation variable analysis. Only six patients in the parent study exhibited abnormal values for all three measures; therefore, no patient sample from this group (group 8) was included in this study. The additional coagulation assays were for protein C antigen, D-dimer, and fibrinogen levels. For the purpose of this analysis, normal ranges for these coagulation variables were defined as follows: protein C antigen, 1.79 to 3.87 ␮g/mL18; D-dimer, 94 to 260 ng/mL; and fibrinogen, 190 to 400 mg/dL. Statistical Methods Baseline and 44-h summary statistics are presented for each coagulation measure (ie, levels of protein C antigen, D-dimer, fibrinogen, and platelets) by 30-day all-cause mortality status, which was the primary outcome measure for the ISS. To assess the relationship between baseline and 44-h protein C levels and mortality, threshold plots were constructed displaying mortality incidence rates in subpopulations defined by baseline and 44-h protein C antigen level thresholds. Logistic regression models were used to assess the relationship of each coagulation measure with 30-day all-cause mortality. A logistic regression model also was used to assess the relationship of each coagulation measure to the presence or absence of shock at baseline and at 44 h. To assess the relationship between each coagulation measure and ventilator-free days and ICU-free days, nonparametric Spearman rank correlations were calculated and statistical tests for nonzero correlation were performed. To assess correlations among coagulation measures, nonparametric Spearman rank analyses also were conducted. For all analyses, a two-sided p value of ⬍ 0.05 was considered to be evidence of an association. Protein C levels measured at 44 h after study entry were not available for three patients; their baseline measurements were imputed for the logistic regressions, Clinical Investigations in Critical Care

8,123 (29,537) 200 (104) 156 (121) 1.20 (0.46) 14.9 (6.6) 0.4 11.3 (12.0) 9.8 (10.6) 2,437 (1,734) 221 (117) 254 (146) 1.26 (0.38) 13.9 (4.7) 0.2 17.6 (14.2) 18 (10.3) 1,451 (1,670) 167 (96) 237 (155) 0.93 (0.32) 17.1 (8.2) 0.6 9.0 (11.1) 6.8 (10.3) 24,173 (64,898) 187 (77) 57 (20) 1.19 (0.48) 14.4 (6.9) 0.6 10.6 (11.3) 7.8 (10.0)

3,120 (2,504) 149 (60) 65 (18) 0.95 (0.37) 16.2 (6.3) 0.5 7.7 (12.8) 6.2 (10.9)

Table 2 displays baseline characteristics and results of the coagulation marker levels (ie, D-dimer, fibrinogen, protein C antigen, and platelets) of the seven subgroups of patients selected from the ISS trial as defined in the “Material and Methods” section. A total of 58 of the 70 patients (83%) included in this study received placebo, and the other 12 patients received ibuprofen in the ISS study. There were no statistically significant (ie, pⱕ 0.05) differences between the two treatment groups with respect to platelet levels, protein C levels, and fibrinogen levels at baseline and at 44 h after entry into the study. In addition, there were no statistically significant differences between the two treatment groups in ranked D-dimer levels at baseline and at 44 h.

21,030 (43,643) 194 (87) 77 (12) 1.33 (0.37) 13.6 (5.9) 0.4 10.3 (11.7) 9.7 (11.2)

Protein C Correlation With Clinical Outcome

D-dimer, ng/mL Fibrinogen, mg/dL Platelets, ⫻ 103 Protein C, ␮g/mL APACHE II Mortality Ventilation-free days ICU-free days

*Values given as mean (SD).

3,844 (6,273) 231 (118) 160 (54) 1.35 (0.41) 10.9 (6.6) 0.3 10.9 (11.7) 9.8 (10.1)

Measure

2,092 (2,945) 253 (140) 244 (101) 1.43 (0.9) 18.2 (5.7) 0.3 12.8 (12) 10.6 (10.2)

All Patients (n ⫽ 70) Normal PT, PTT, and Platelets (n ⫽ 10) Abnormal PT and PTT (n ⫽ 10) Abnormal PTT and Platelets (n ⫽ 10) Abnormal PT and Platelets (n ⫽ 10) Abnormal Platelets (n ⫽ 10) Abnormal PT (n ⫽ 10)

Results

Abnormal PTT (n ⫽ 10)

Table 2—Baseline Characteristics of Patients by Subgroup*

threshold plots, and correlation analyses with ventilator-free days and ICU-free days. Shock assessments measured at 44 h after study entry were unavailable for nine patients. Because each of these patients was in shock at baseline, and because shock status values after baseline were available, it was assumed that all were in shock at 44 h after study entry, and the last observed shock status value was used in the last-observation-carried-forward imputation method for shock status-related analyses. Similar weighted analyses were performed with patient weighting based on the proportion of patients in the ISS who fell into the same subgroup as the analyzed patient. In general, the conclusions drawn from the unweighted analyses and weighted analyses were consistent. The conclusions from the unweighted analyses are presented below.

Baseline protein C levels were below the lower limit of normal in 63 of the 70 patients (90%). A trend toward an increase in 30-day mortality with lower baseline protein C levels was observed (p ⫽ 0.19), and this same relationship was statistically significant when 30-day mortality was compared with 44-h protein C levels (p ⫽ 0.04) (Table 3). Threshold plots describing the relationship between protein C antigen levels at baseline and after 44 h and 30-day all-cause mortality are displayed in Figure 1. The thresholds were selected at 50%, 60%, 70%, 80%, 90%, and 100% of the lower limit of normal (ie, 1.79 ␮g/mL) of protein C. At both baseline (p ⫽ 0.01) and 44 h (p ⬍ 0.001), the presence of shock was correlated with lower baseline protein C levels (Table 4). The threshold plots describing the relationship between protein C antigen levels at baseline and 44 h and the presence or absence of shock are displayed in Figure 2. A reduction in the number of ventilator-free days was associated with lower protein C levels at both baseline (p ⫽ 0.02) and 44 h (p ⫽ 0.03) (Table 5). CHEST / 120 / 3 / SEPTEMBER, 2001

917

Table 3—Coagulation Marker Summary Statistics at Baseline and at 44 h by 30-Day All-Cause Mortality Status* Survivors Variables

Nonsurvivors

Logistic p Value

n

Mean

Median (IQR)

n

Mean

Median (IQR)

Baseline Protein C, ␮g/mL Fibrinogen, mg/dL Platelets, ⫻ 103 D-dimer, ng/mL

41 41 41 41

1.26 220 180 6,463

1.19 (.94–1.44) 201 (151–245) 139 (85–266) 1,254 (662–2,097)

29 29 29 28

1.12 173 123 10,553

1.09 (0.84–1.32) 170 (98–235) 84 (65–145) 2,416 (1,192–4,812)

0.19 0.054 0.04 0.57 0.02 (rank)

44-h Protein C, ␮g/mL† Fibrinogen, mg/dL Platelets, ⫻ 103 D-dimer, ng/mL

41 41 41 40

1.67 231 141 2,430

1.61 (1.26–2.07) 224 (173–258) 124 (71–184) 1,324 (923–2,477)

29 24 19 24

1.36 180 103 6240

1.20 (0.95–1.72) 171 (131–209) 88 (47–118) 2,350 (1,090–6,137)

0.04 0.02 0.11 0.02

*IQR ⫽ interquartile range (25th percentile to 75th percentile of distribution). †Includes imputation of baseline protein C antigen levels for three patients.

Fewer ICU-free days were associated with lower protein C levels at both baseline (p ⫽ 0.04) and 44 h (p ⫽ 0.02) (Table 5). D-Dimer, Fibrinogen, and Platelet Count Correlation With Clinical Outcome Baseline D-dimer levels were above the upper limit of normal in all patients. Thirty-six of the 70 patients (51%) had baseline fibrinogen levels below the lower limit of normal. Forty-one of the 70 patients (59%) had baseline platelet levels below the lower limit of normal. Based on a logistic regression model, lower baseline platelet levels (p ⫽ 0.04) and 44-h fibrinogen levels (p ⫽ 0.02) were associated with increased mortality (Table 3). Elevated levels of D-dimer at 44 h were associated with increased mortality (p ⫽ 0.02). Although baseline D-dimer levels were not significantly associated with mortality based on the logistic regression model, this finding was most likely due to the severe skewness in baseline D-dimer measurements. When baseline D-dimer levels were replaced by their ranks in the logistic model, the positive association between increased D-dimer levels and increased mortality did reach statistical significance (p ⫽ 0.02) (Table 3). Increased baseline D-dimer levels (p ⫽ 0.02) and decreased baseline fibrinogen levels (p ⫽ 0.03) were associated with the presence of shock at baseline (Table 4). Lower platelet levels at baseline were associated with fewer ICU-free days (p ⫽ 0.047) (Table 5). Higher D-dimer levels were associated with fewer ICU-free and ventilator-free days; however, significance was achieved only when 44-h levels were compared with ventilator-free days (p ⫽ 0.03) (Table 5). No association was demonstrated between fibrinogen levels and ICU-free or ventilator-free days. 918

Correlation of Coagulation Markers With Protein C The correlation between the coagulation markers and protein C levels was weak or nonexistent (Table 6). The sole significant positive correlation observed was between the 44-h platelet levels and protein C antigen levels (p ⫽ 0.03). A trend toward a positive association between 44-h fibrinogen levels and protein C antigen levels also was observed (p ⫽ 0.07). At baseline, there was also little correlation among PT, PTT, and platelet levels, and protein C levels, especially in the subgroup of patients that had normal baseline PT, PTT, and platelet levels (Table 2). The mean protein C levels in this subgroup was 1.26 ␮g/mL. A total of 9 of the 10 patients in this subgroup had baseline protein C levels below the lower limit of normal.

Discussion Activated protein C is a serine protease that inhibits coagulation factors Va and VIIIa, subsequently blocking the generation of thrombin. Both plasminogen activator inhibitor (PAI)-1 and activated thrombin-activatable fibrinolysis inhibitor (TAFI) are inhibitors of the endogenous fibrinolytic system. TAFI is activated by thrombin. Activated protein C also exhibits profibrinolytic activity by neutralizing PAI-1 or by limiting the activation of TAFI by limiting thrombin generation.19,20 Protein C circulates largely in the zymogen form in plasma and is activated via the complexing of thrombin with the endothelial cell surface protein thrombomodulin.21 In large blood vessels, the activation of protein C to activated protein C is facilitated by the endothelial protein C receptor in complex with thrombin-thromClinical Investigations in Critical Care

Figure 1. Baseline protein C levels vs 30-day all-cause mortality. The association of baseline (top) and 44-h (bottom) protein C antigen levels and 30-day mortality at thresholds of ⬍ 50%, ⬍ 60%, ⬍ 70%, ⬍ 80%, ⬍ 90%, and ⬍ 100% of the lower limit of normal (normal range, 1.79 to 3.87 ␮g/mL) of protein C antigen is shown.

bomodulin.14,22 In this study, we measured and analyzed four hemostatic markers (protein C, Ddimer, fibrinogen, and platelets) in a randomly selected subset of patients (n ⫽ 70) from the ISS trial. All ISS patients met the currently accepted defini-

tions of severe sepsis at study baseline.23 A total of 63 of the 70 patients (90%) studied in this report had protein C levels at baseline below the lower limit of normal. Thus, our study data suggest that almost all patients with severe sepsis, as selected using standard criteria,23 had acquired protein C deficiency. The threshold plots (Fig 1, 2) indicate that the lower the protein C levels, the higher the mortality and the higher the occurrence of shock. Protein C levels are lower in patients with sepsis, presumably as a result of associated consumptive coagulopathy,4,5,7,12,13,24,25 although decreased production also may play a role. Diagnostic kits are readily available to measure plasma protein C levels, and only citrated plasma samples are required. Thus, studies of protein C levels in many pathologic states, including sepsis, have been published in the literature. However, in order to measure plasma activated protein C levels, a special blood collection tube containing benzamidine, in addition to citrate, is required.26 So, samples for measuring levels of activated protein C levels have to be prospectively collected. In addition, to our knowledge, no diagnostic kit for measuring activated protein C is commercially available, and the methodology is complex.26 As such, the authors are only aware of a single published article with data on the levels of activated protein C in patients with sepsis.27 The data in that article indicated that only about 20% of patients with sepsis have elevated activated protein C levels beyond the normal range of 1 to 2 ng/mL to about 5 to 20 ng/mL. The increase in activated protein C in those 20% of patients appeared to be sporadic with no discernible pattern. The relationship between the protein C levels and the activated protein C levels in patients with sepsis is complex and not well understood, and it requires many more studies. Findings from this study of the association of protein C levels with mortality and the presence of

Table 4 —Coagulation Marker Summary Statistics at Baseline and 44 h by Shock Status* Shock Present Variables Baseline Protein C, ␮g/mL Fibrinogen, mg/dL Platelets, ⫻ 103 D-dimer, ng/mL 44-h Protein C, ␮g/mL† Fibrinogen, mg/dL Platelets, ⫻ 103 D-dimer, ng/mL

Shock Absent Median (IQR)

Logistic p Value

n

Mean

Median (IQR)

n

Mean

50 50 50 49

1.12 183 151 3,006

1.03 (0.84–1.33) 179 (107–235) 114 (68–177) 1,459 (799–4,476)

20 20 20 20

1.42 243 171 20,658

1.31 (1.19–1.54) 228 (146–290) 115 (81–233) 1,514 (651–3,048)

0.01 0.03 0.52 0.02

26 22 16 22

1.21 191 117 5,553

1.19 (0.89–1.49) 179 (156–251) 105 (49–156) 2,455 (1,076–4,377)

44 43 44 42

1.74 223 133 2,971

1.76 (1.26–2.13) 217 (159–274) 94 (73–180) 1,417 (913–2,460)

⬍ 0.001 0.15 0.54 0.16

*See Table 3 for abbreviations not used in the text. †Includes imputation of baseline protein C antigen levels for three patients. CHEST / 120 / 3 / SEPTEMBER, 2001

919

Figure 2. Protein C levels and association with severity of illness. The association of baseline (top) and 44-h (bottom) protein C antigen levels and the occurrence of shock at threshold protein C antigen levels of ⬍ 50%, ⬍ 60%, ⬍ 70%, ⬍ 80%, ⬍ 90%, and ⬍ 100% of the lower limit of normal (normal range, 1.79 to 3.87 ␮g/mL) of protein C antigen is shown.

shock in severe sepsis patients are consistent with those reported by others. Hesselvik et al12 reported that patients with septic shock had lower protein C levels than those without shock on the day following

Table 5—Coagulation Marker Correlation Analysis With Ventilator-Free and ICU-Free Days Ventilator-Free Days Variables Baseline Protein C Fibrinogen Platelets D-dimer 44-h Protein C* Fibrinogen Platelets D-dimer

ICU-Free Days

Rank Correlation

p Value

Rank Correlation

p Value

0.27 0.00 0.18 ⫺0.17

0.02 0.99 0.14 0.16

0.25 0.08 0.24 ⫺0.17

0.04 0.51 0.047 0.16

.26 .00 .14 ⫺.27

0.03 1.00 0.29 0.03

0.27 0.02 0.22 ⫺0.20

0.02 0.89 0.10 0.12

*Includes imputation of baseline protein C antigen levels for three patients. 920

initial hospitalization. Fourrier et al7 found that initial protein C levels were significantly lower in nonsurviving sepsis patients with DIC and documented a significant correlation between protein C levels in nonsurvivors. In a subgroup with sequential measurements, nonsurvivors exhibited a persistent deficiency in protein C activity. Lorente et al5 measured coagulation and fibrinolysis variables in sepsis patients on days 1, 4, and 7 after hospital admission. Their analysis showed that the median protein C level in nonsurvivors was below the lower limit of normal on day 1 after hospital admission. There was no significant difference between survivors and nonsurvivors in protein C measurements at the time of hospital admission. However, serial protein C measurements on subsequent days after hospital admission showed a significant difference between survivors and nonsurvivors. Only survivors exhibited a progressive normalization of protein C levels. The data from our study further support the suggestion from Lorente et al5 and others28 that serial protein C measurements provide better prognostic value than a single measurement in monitoring patients with sepsis. Recent data from Boldt et al25 also suggest that serial protein C measurements may be a helpful molecular marker to differentiate severe sepsis from other causes of critical illness. Among the four hemostatic markers studied in this report, protein C levels correlated best with the major outcome measures of severe sepsis (ie, mortality, presence of shock, ICU stay, and ventilator dependence). Fibrinogen level was not as consistently associated with the outcome measures, probably because it is known to be an acute-phase reactant.29 The fibrinogen level goes up in the early phase of severe infection, and fibrinogen subsequently is consumed during the severe coagulopathy phase of severe sepsis. In patients with sepsis, fibrinogen has been reported to be higher in nonsurvivors than survivors,5 lower in nonsurvivors than in survivors,13,30 –32 or not significantly different between survivors and nonsurvivors.2,33 Perhaps the seemingly contradictory data on fibrinogen levels in patients with sepsis are due to differences in sampling time and the acute-phase reactant nature of fibrinogen. Protein C level also was reported by others5,34 to be better

Table 6 —Coagulation Marker Correlation Analysis With Protein C Antigen Baseline

44-h

Variables

Rank Correlation

p Value

Rank Correlation

p Value

Fibrinogen Platelets D-dimer

0.16 0.08 0.06

0.18 0.50 0.62

0.23 0.28 ⫺0.08

0.07 0.03 0.51

Clinical Investigations in Critical Care

associated with sepsis outcome than D-dimer levels or platelet levels. Thrombocytosis rather than thrombocytopenia has been observed in some sepsis patients.35 It is also interesting to point out that in this study, 9 of the 10 patients in the subgroup who had normal baseline values for PT, PTT, and platelets had baseline protein C levels below the lower limit of normal. These data suggest that protein C may be an earlier and more sensitive marker than the more global coagulation markers of PT, PTT, and platelets that are currently more widely used in hospitals. The data from this and other studies5,7,24,25,35 suggest that protein C measurement, in addition to the DIC panel of markers, may offer a better prognostic value in sepsis. The importance of the protein C pathway in sepsis has been highlighted by findings in animal models. In Escherichia coli-induced septicemia in baboons, antibody blockade of the protein C pathway markedly increased the severity of endotoxin doses at concentrations for both the lethal dose in all exposed subjects and at 10% of the lethal dose.36 Our data reinforce the findings of other clinical and animal investigations that have indicated that endogenous protein C plays an important role in sepsis-associated coagulopathy. Other than the well-known antithrombotic activity of the protein C pathway, there is also evidence that protein C exerts direct anti-inflammatory effects apart from those mediated by thrombin inhibition. These may include the reduction of endotoxin-induced cytokine production by monocytes and activated protein C-endothelial protein C receptor interaction with activated neutrophils.15,37–39 PAI-1, an inhibitor for fibrinolysis, also is found to be elevated in sepsis and is positively correlated to mortality in most reported studies.20 Because activated protein C has profibrinolytic activity by neutralizing PAI-1, other investigators5,7,40,41 have suggested the need for randomized, controlled trials of activated protein C in patients with sepsis. Our finding that 90% of patients with severe sepsis have protein C levels less than the lower limit of normal gives further support to this notion. As such, a recombinant form of human activated protein C has been developed and is undergoing clinical study. In patients with severe sepsis, the mechanism for activating protein C in vivo (protein C converting to activated protein C by thrombin in a complex with thrombomodulin) may be operating suboptimally.42 Indeed, the most recent data from patients with sepsis support this hypothesis.43 Thus activated protein C would be expected to exert a more potent and predictable effect than protein C, the zymogen form. Activated protein C also may be more resistant to neutrophil elastase than is the zymogen form.44 In summary, we have found that lower protein C levels are common in patients with sepsis and are

associated with several severely negative clinical outcomes, including increased mortality and occurrence of shock, as well as increased duration of mechanical ventilation and ICU stay. These findings suggest that protein C levels can be used prognostically and that such agents as protein C, or preferably, activated protein C, may reverse the acquired protein C deficiency in patients with sepsis and improve outcome. A large, placebo-controlled, phase III trial using recombinant human activated protein C in patients with severe sepsis has been undertaken to test whether this experimental drug can improve the 28-day all-cause mortality rate of this disease. References 1 Suffredini AF, Harpel PC, Parrillo JE. Promotion and subsequent inhibition of plasminogen activation after administration of intravenous endotoxin to normal subjects. N Engl J Med 1989; 320:1165–1172 2 Hesselvik JF, Blomback M, Brodin B, et al. Coagulation, fibrinolysis, and kallikrein systems in sepsis: relation to outcome. Crit Care Med 1989; 17:724 –733 3 Pralong G, Calandra T, Glauser MP, et al. Plasminogen activator inhibitor 1: a new prognostic marker in septic shock. Thromb Haemost 1989; 61:459 – 462 4 Philippe J, Offner F, Declerck PJ, et al. Fibrinolysis and coagulation in patients with infectious disease and sepsis. Thromb Haemost 1991; 65:291–295 5 Lorente JA, Garcia-Frade LJ, Landin L, et al. Time course of hemostatic abnormalities in sepsis and its relation to outcome. Chest 1993; 103:1536 –1542 6 Carvalho AC, Freeman NJ. How coagulation defects alter outcome in sepsis: survival may depend on reversing procoagulant conditions. J Crit Illness 1994; 9:51–75 7 Fourrier F, Chopin C, Goudemand J, et al. Septic shock, multiple organ failure and disseminated intravascular coagulation: compare patterns of antithrombin III, protein C and protein S deficiencies. Chest 1992; 101:816 – 823 8 Mesters RM, Mannucci PM, Coppola R, et al. Factor VIIa and antithrombin III activity during severe sepsis and septic shock in neutropenic patients. Blood 1996; 88:881– 886 9 Bone RC. Modulators of coagulation: a critical appraisal of their role in sepsis. Arch Intern Med 1992; 152:1381–1389 10 Levi M, ten Cate H, van der Poll T, et al. Pathogenesis of disseminated intravascular coagulation in sepsis. JAMA 1993; 270:975–979 11 Brandtzaeg P, Mollnes TE, Kierulf P. Complement activation and endotoxin levels in systemic meningococcal disease. J Infect Dis 1989; 160:58 – 65 12 Hesselvik JF, Malm J, Dahlback B, et al. Protein C, protein S, and C4b-binding protein in severe infection and septic shock. Thromb Haemost 1991; 65:126 –129 13 Fijnvandraat K, Derkx B, Peters M, et al. Coagulation activation and tissue necrosis in meningococcal septic shock: severely reduced protein C levels predict a high mortality. Thromb Haemost 1995; 73:15–20 14 Esmon CT, Gu JM, Xu J, et al. Regulation and functions of the protein C anticoagulant pathway. Haematologica 1999; 84:363–368 15 Yan SB, Grinnell BW. Antithrombotic and anti-inflammatory agents of the protein C anticoagulant pathway. Ann Rep Med Chem 1994; 29:103–112 16 Bernard GR, Wheeler AP, Russell JA, et al. The effects of CHEST / 120 / 3 / SEPTEMBER, 2001

921

17 18 19 20 21 22

23 24

25

26 27

28 29 30

ibuprofen on the physiology and survival of patients with sepsis. N Engl J Med 1997; 336:912–918 Bernard GR, Wheeler AP, Arons MM, et al. A trial of antioxidants N-acetylcysteine and procysteine in ARDS. Chest 1997; 112:164 –172 Howard PR, Bovill EG, Mann KG, et al. A monoclonal antibody-based radioimmunoassay for measurement of protein C in plasma. Clin Chem 1988; 34:324 –330 Bajzar L, Nesheim M, Tracy PB. The profibrinolytic effect of activated protein C in clots formed from plasma is TAFIdependent. Blood 1996; 88:2093–2100 Vervloet MG, Thijs LG, Hack CE. Derangements of coagulation and fibrinolysis in critically ill patients with sepsis and septic shock. Semin Thromb Hemost 1998; 24:33– 44 Esmon CT. Molecular events that control the protein C anticoagulant pathway. Thromb Haemost 1993; 70:29 –35 Stearn-Kurosawa JJ, Kurosawa S, Mollica JS, et al. The endothelial cell protein C receptor augments protein C activation by thrombin-thrombomodulin complex. Proc Natl Acad Sci U S A 1996; 93:10212–10216 Bone RC. Toward an epidemiology and natural history of SIRS (systemic inflammatory response syndrome). JAMA 1992; 268:3452–3455 Brandtzaeg P, Sandset PM, Joo GB, et al. The quantitative association of plasma endotoxin, antithrombin, protein C, extrinsic pathway inhibitor and fibrinopeptide A in systemic meningococcal disease. Thromb Res 1989; 55:459 – 470 Boldt J, Papsdorf M, Rothe A, et al. Changes of the hemostatic network in critically ill patients: is there a difference between sepsis, trauma, and neurosurgery patients? Crit Care Med 2000; 28:445– 450 Gruber A, Griffin JH. Direct detection of activated protein C in blood from human subjects. Blood 1992; 79:2340 –2348 Mesters RM, Helterbrand J, Utterback BG, et al. Prognostic value of protein C levels in neutropenic patients at high risk of severe septic complications. Crit Care Med 2000; 28:2209 – 2216 Levi M, de Jonge E, van der Poll T, et al. Disseminated intravascular coagulation. Thromb Haemost 1999; 82:695– 705 Gabay C, Kushner I. Acute-phase proteins and other systemic responses to inflammation. N Engl J Med 1999; 340:448 – 454 Hazelzet JA, Risseeuw-Appel IM, Kornelisse RF, et al. Age-related differences in outcome and severity of DIC in children with septic shock and purpura. Thromb Haemost 1996; 76:932–938

922

31 Bonduel M, Hepner M, Sciuccati G, et al. Infectious purpura fulminans in children: a report on 20 patients from Argentina. Int J Pediatr Hematol Oncol 1998; 5:21–27 32 Leclerc F, Hazelzet J, Jude B, et al. Protein C and S deficiency in severe infectious purpura of children: a collaborative study of 40 cases. Intensive Care Med 1992; 18:202– 205 33 Massignon D, Lepape A, Bienvenu J, et al. Coagulation/ fibrinolysis balance in septic shock related to cytokines and clinical state. Haemostasis 1994; 24:36 – 48 34 Aronis S, Hatzis A, Platokouki H, et al. Severe protein C deficiency, predictor of lethal outcome in meningococcal septicemia [abstract]. Pediatr Res 1999; 45:765 35 Powars D, Larsen R, Johnson J, et al. Epidemic meningococcemia and purpura fulminans with induced protein C deficiency. Clin Infect Dis 1993; 17:254 –261 36 Taylor FB, Chank A, Esmon CT, et al. Protein C prevents the coagulopathic and lethal effects of E coli infusion in the baboon. J Clin Invest 1987; 79:918 –925 37 Esmon CT, Xu J, Gu JM, et al. Endothelial protein C receptor. Thromb Haemost 1999; 82:251–258 38 Esmon CT. Introduction: are natural anticoagulants candidates for modulating the inflammatory response to endotoxin? Blood 2000; 95:1113–1116 39 Hancock WW, Bach FH. Immunobiology and therapeutic applications of protein C/protein S/thrombomodulin in human and experimental allotransplantation and xenotransplantation. Trends Cardiovasc Med 1997; 7:174 –183 40 Corbeel LM, Proesmans W. Treatment of meningococcal purpura fulminans (Waterhouse-Friderichsen syndrome) with protein C concentrate. Eur J Pediatr 1997; 155:664 – 665 41 Duncan A. New therapies for severe meningococcal disease but better outcome? Lancet 1997; 350:1565–1566 42 Moore KL, Esmon CT, Esmon NL. Tumor necrosis factor leads to the internalization and degradation of thrombomodulin from the surface of bovine aortic endothelial cells in culture. Blood 1989; 73:159 –165 43 Faust SN, Heyderman RS, Harrison O, et al. Molecular mechanisms of thrombosis in meningococcal septicemia: the role of the protein C pathway in vivo [abstract]. Shock 2000; 13(suppl):29 44 Philapitsch A, Schwarz HP. The effect of leukocyte elastase on protein C and activated protein C [abstract]. Thromb Haemost 1993; 69:724

Clinical Investigations in Critical Care