INTRODUCTION — Sepsis is a clinical syndrome characterized by systemic inflammation due to infection. There is a continuum of severity ranging from sepsis to severe sepsis and septic shock. Over 750,000 cases of sepsis occur in the United States each year, resulting in approximately 200,000 fatalities [1]. Even with optimal treatment, mortality due to severe sepsis or septic shock is approximately 40 percent and can exceed 50 percent in the sickest patients [2-5].

Numerous interventions exist that decrease mortality due to sepsis. In this topic review, the management of severe sepsis and septic shock is discussed. Definitions, diagnosis, pathophysiology, and investigational therapies are reviewed separately. (See "Sepsis and the systemic inflammatory response syndrome: Definitions, epidemiology, and prognosis" and "Pathophysiology of sepsis" and "Investigational and ineffective therapies for sepsis".)

THERAPEUTIC PRIORITIES — Therapeutic priorities for patients with severe sepsis or septic shock include:

• Early initiation of supportive care to correct physiologic abnormalities, such as hypoxemia and hypotension [6-9].
• Distinguishing sepsis from systemic inflammatory response syndrome (SIRS) (table 1 and table 2) because, if an infection exists, it must be identified and treated as soon as possible (table 3). This may require a surgical procedure (eg, drainage), as well as appropriate antibiotics.

EARLY MANAGEMENT — The first priority in any patient with severe sepsis or septic shock is stabilization of their airway and breathing. Next, perfusion to the peripheral tissues should be restored [7,10].

Stabilize respiration — Supplemental oxygen should be supplied to all patients with sepsis and oxygenation should be monitored continuously with pulse oximetry. Intubation and mechanical ventilation may be required to support the increased work of breathing that typically accompanies sepsis, or for airway protection since encephalopathy and a depressed level of consciousness frequently complicate sepsis [11,12].

Sedative and induction agents (eg, etomidate) used to intubate patients with severe sepsis or septic shock are discussed separately. Other aspects of intubation and mechanical ventilation are similarly described elsewhere. (See "Sedation or induction agents for rapid sequence intubation in adults" and "Advanced airway management in adults" and "Rapid sequence intubation in adults" and "The decision to intubate" and "The difficult airway in adults".)

Chest radiographs and arterial blood analysis should be obtained following initial stabilization. These studies are used in combination with other clinical parameters to diagnose acute lung injury (ALI) or acute respiratory distress syndrome (ARDS), which frequently complicate sepsis. (See "Acute respiratory distress syndrome: Definition; epidemiology; diagnosis; and etiology" and "Mechanical ventilation in acute respiratory distress syndrome".)

Assess perfusion — Once the patient's respiratory status has been stabilized, the adequacy of perfusion should be assessed. Hypotension is the most common indicator that perfusion is inadequate. Therefore, it is important that the blood pressure be assessed early and often. An arterial catheter may be inserted if blood pressure is labile or restoration of arterial perfusion pressures is expected to be a protracted process, because a sphygmomanometer may be unreliable in hypotensive patients [8]. Attempts to insert an arterial line should not be allowed to delay the prompt management of shock. (See "Arterial catheterization".)

Critical hypoperfusion can also occur in the absence of hypotension, especially during early sepsis. Thus, clinical evidence of impaired perfusion should be sought in all patients with sepsis.

Common signs of hypoperfusion include cool, vasoconstricted skin due to redirection of blood flow to core organs (although warm, flushed skin may be present in the early phases of sepsis), obtundation or restlessness, oliguria or anuria, and lactic acidosis. These findings may be modified by preexisting disease or medications. As an example, elderly patients, diabetic patients, and patients who take beta-blockers may not exhibit an appropriate tachycardia as blood pressure falls. Patients with chronic hypertension may develop critical hypoperfusion at a higher blood pressure than healthy patients (ie, relative hypotension).

Catheters — After initial assessment, a central venous catheter (CVC) should be inserted in most patients with severe sepsis or septic shock. A CVC can be used to infuse intravenous fluids, infuse medications, infuse blood products, and draw blood. In addition, it can be used for hemodynamic monitoring by measuring the central venous pressure (CVP) and the central venous oxyhemoglobin saturation (ScvO2). In one clinical trial, treatment of septic shock guided by the ScvO2 reduced mortality [13]. (See "Indications for and complications of central venous catheters".)

We believe that pulmonary artery catheters (PACs) should not be used in the routine management of patients with severe sepsis or septic shock. PACs can measure the pulmonary artery occlusion pressure (PAOP) and mixed venous oxyhemoglobin saturation (SvO2). In theory, this may be helpful to guide circulatory resuscitation. However, the PAOP has proven to be a poor predictor of fluid responsiveness in sepsis and the SvO2 is similar to the ScvO2, which can be obtained from a CVC [14]. PACs increase complications and have not been shown to improve outcome [15-17]. (See "Swan-Ganz catheterization: Indications and complications".)

Respiratory changes in the radial artery pulse pressure, aortic blood flow peak velocity, and brachial artery blood flow velocity are considered dynamic hemodynamic measures, whereas CVP and PAOP are considered static hemodynamic measures [18,19]. There is increasing evidence that dynamic measures are more accurate predictors of fluid responsiveness than static measures, as long as the patients are in sinus rhythm and passively ventilated with a sufficient tidal volume [14,20,21]. It seems likely that dynamic measures will become more common and be used to identify patients who are likely to increase organ perfusion in response to intravenous fluids.

Restore perfusion — Once it has been established that hypoperfusion exists, early restoration of perfusion is necessary to prevent or limit multiple organ dysfunction, as well as reduce mortality. Hypoperfusion results from loss of plasma volume into the interstitial space, decreased vascular tone, and myocardial depression. The increase in the cardiac output that is necessary to compensate for the diminished vascular tone may be limited by the myocardial depression.

Resuscitation of the circulation should target a ScvO2 or SvO2 ≥70 percent [7,13]. Other reasonable goals include a central venous pressure 8 to 12 mmHg, a mean arterial pressure (MAP) ≥65 mmHg, and a urine output ≥0.5 mL/kg per hour.

These goals derive from a clinical trial in which 263 patients with severe sepsis or septic shock were randomly assigned to therapy targeting a ScvO2 ≥70 percent, or conventional therapy that did not target a ScvO2 [13]. Both groups initiated therapy within six hours of presentation and targeted the same CVP, MAP, and urine output. Mortality was lower in the group that targeted a ScvO2 ≥70 percent (31 versus 47 percent). This approach is known as "early goal-directed therapy" (ie, administered within the first six hours of presentation) (algorithm 1).

Earlier studies of critically ill patients that used similar targets (SvO2 ≥70 percent) found no mortality benefit [22]. This is probably because these studies were not conducted during the crucial initial hours. This is supported by a systemic review that compared resuscitation targeting specific physiologic endpoints to standard resuscitation [23]. In a meta-analysis of randomized trials initiated within 24 hours of the onset of sepsis (6 trials, 740 patients), resuscitation targeting specific physiologic endpoints improved mortality (39 versus 57 percent for standard resuscitation, odds ratio 0.50, 95% CI 0.37-0.69). In contrast, a meta-analysis of randomized trials initiated more than 24 hours after the onset of sepsis (3 trials, 261 patients) found that resuscitation targeting specific physiologic endpoints did not improve mortality (64 versus 58 percent for standard resuscitation, odds ratio 1.16, 95% CI 0.60-2.22).

In our clinical practice, we adhere to the principles of early goal-directed therapy; that is, we initiate aggressive therapy early to restore perfusion and we target a ScvO2 ≥70 percent. However, we consider the numeric goals for CVP, MAP, and urine output guidelines and always consider additional clinical signs of hypoperfusion when assessing the patient's response to a therapy and need for more of a therapy.

Intravenous fluids — Relative intravascular hypovolemia is typical and may be severe. As an example, early goal-directed therapy required a mean infusion volume of approximately five liters within the initial six hours of therapy in the trial described above [13]. As a result, rapid, large volume infusions of intravenous fluids are indicated as initial therapy for severe sepsis or septic shock, unless there is coexisting clinical or radiographic evidence of heart failure.

Fluid therapy should be administered in well-defined (eg, 500 mL), rapidly infused boluses [8,9]. Volume status, tissue perfusion, blood pressure, and the presence or absence of pulmonary edema must be assessed before and after each bolus. Intravenous fluid challenges can be repeated until blood pressure is acceptable, tissue perfusion is acceptable, pulmonary edema ensues, or fluid fails to augment perfusion.

Careful monitoring is essential in this approach because patients with sepsis typically develop noncardiogenic pulmonary edema (ie, ALI, ARDS). In patients with ALI or ARDS who are hemodynamically resuscitated, a liberal approach to intravenous fluid administration prolongs the duration of mechanical ventilation, compared to a more restrictive approach that typically requires large doses of furosemide [24]. Thus, while the early, aggressive fluid therapy is appropriate in severe sepsis and septic shock, fluids may be unhelpful or harmful when the circulation is no longer fluid-responsive. (See "Supportive care and oxygenation in acute respiratory distress syndrome", section on 'Fluid management'.)

• Crystalloid versus colloid — Clinical trials have failed to consistently demonstrate a difference between colloid and crystalloid in the treatment of septic shock [25,26].

In the saline versus albumin fluid evaluation (SAFE) trial, 6997 critically ill patients were randomly assigned to receive 4 percent albumin or normal saline for up to 28 days [27]. There were no differences between groups for any endpoint, including the primary endpoint, mortality. Among the patients with severe sepsis (18 percent of the total group), there were also no differences in outcome.

Another randomized trial compared pentastarch (a colloid) to modified Ringer's lactate (a crystalloid) in patients with severe sepsis — the Efficacy of Volume Substitution and Insulin Therapy in Severe Sepsis (VISEP) trial [28]. There was no difference in 28-day mortality, but the trial was stopped early because there was a trend toward increased 90-day mortality among patients who received pentastarch.

In our clinical practice, we generally use crystalloid because of the higher cost of colloid. We believe that giving a sufficient quantity of intravenous fluids rapidly and targeting appropriate goals is more important than the type of fluid chosen.

Vasopressors — Vasopressors are second line agents in the treatment of severe sepsis and septic shock; we prefer intravenous fluids as long as they increase perfusion without seriously impairing gas exchange [29]. However, intravenous vasopressors are useful in patients who remain hypotensive despite adequate fluid resuscitation or who develop cardiogenic pulmonary edema.

There is no definitive evidence of the superiority of one vasopressor over another (table 4). We prefer norepinephrine, although dopamine is also a reasonable first-choice among vasopressors [7]. Phenylephrine, a pure alpha-adrenergic agonist, may be particularly useful when tachycardia or arrhythmias preclude the use of agents with beta-adrenergic activity. Limited experience with vasopressin (antidiuretic hormone) suggests that this agent may be useful in vasodilatory septic shock. (See "Use of vasopressors and inotropes".)

Additional therapies — When the ScvO2 remains <70 percent after optimization of intravenous fluid and vasopressor therapy, it is reasonable to consider additional therapies, such as inotropic therapy or red blood cell transfusion.

• Inotropic therapy — For patients who have myocardial dysfunction, a trial of inotropic therapy is warranted if ScvO2 remains <70 percent after all of the interventions discussed above [7,8,13,30,31]. Inotropic therapy should not be used to increase the cardiac index to supranormal levels [7].

Dobutamine is the usual inotropic agent. At low doses, dobutamine may cause the blood pressure to decrease because it can dilate the systemic arteries. However, as the dose is increased, blood pressure usually rises because cardiac output increases out of proportion to the fall in vascular resistance.

• Red blood cell transfusions — Early goal-directed therapy aggressively utilizes red blood cell transfusions to raise the ScvO2. In the trial discussed above, nearly 70 percent of patients in the early goal-directed therapy group received transfusions, compared to 45 percent in the conventional therapy group [13]. However, other data support a more cautious approach to transfusion in critically ill patients [32]. (See "Use of blood products in the critically ill", section on 'Red blood cells'.)

There are several possible explanations for the conflicting data.

• - Outcome may be related to when a red blood cell transfusion is given. Transfusions administered as part of early goal-directed therapy were given early in the course of illness, whereas studies that support a more cautious approach typically gave transfusions later in the course of illness.
• - The apparent benefit of red blood cell transfusions may be due to other interventions. In other words, red blood cell transfusion was just one of several interventions during early goal-directed therapy and it is possible that the benefit was due to one or more of the other interventions, not the red blood cell transfusion per se.

Ongoing management — There are two possible outcomes following the interventions described above:

• Despite aggressive therapy, the patient may have persistent hypoperfusion and progressive organ failure. This should prompt reassessment of the adequacy of the above therapies, antimicrobial regimen, and control of the septic focus, as well as the accuracy of the diagnosis and the possibility that unexpected complications or coexisting problems have intervened (eg, pneumothorax following CVC insertion).
• The patient may have responded to the above interventions with restored perfusion and a ScvO2 greater than 70 percent. Such patients should continue to have their clinical and laboratory parameters followed closely. These include blood pressure, arterial lactate, urine output, creatinine, platelet count, Glasgow coma score, serum bilirubin, liver enzymes, oxygenation (ie, arterial oxygen tension or oxyhemoglobin saturation), and gut function (table 5). Gastric tonometry may also be helpful, if available. Reevaluation is indicated if any of these parameters worsen or fail to improve.

In early sepsis, most lactate is probably a byproduct of anaerobic metabolism due to organ hypoperfusion. Supporting this view, early goal-directed therapy decreases lactate levels faster than conventional therapy [13]. After the restoration of perfusion, however, lactate is probably due to causes other than anaerobic metabolism and further increasing oxygen delivery to the peripheral tissues is unlikely to decrease its levels [33]. As a result, lactate values are generally unhelpful following restoration of perfusion, with one exception — a rising lactate level should prompt reevaluation of perfusion (see "Arterial and mixed venous blood gases in lactic acidosis".

It would be ideal if hypoxia could be detected for individual organs, because tests that combine output from many organs (eg, arterial lactate) may obscure the presence of significant ischemia in an individual organ [34]. Gastric tonometry indirectly measures perfusion to the gut by estimating the gastric mucosal PCO2. It can be used to detect gut hypoxia by calculating the gastric to arterial PCO2 gap [35,36]. But, gastric tonometry is not widely available and it is uncertain whether it can successfully guide therapy. Additional studies and clinical experience are needed.

The American Thoracic Society (ATS) statement on the detection, correction, and prevention of tissue hypoxia, as well as other ATS guidelines, can be accessed through the ATS web site at www.thoracic.org/sections/publications/statements/index.html.

CONTROL OF THE SEPTIC FOCUS — Prompt identification and treatment of the primary site or sites of infection are essential [37-39]. This is the primary therapeutic intervention, with most other interventions being purely supportive.

Identification of the septic focus — A careful history and physical examination may yield clues to the source of sepsis and help guide microbiologic evaluation (table 6). As an example, sepsis arising after trauma or surgery is often due to infection at the site of injury or surgery. The presence of a urinary or vascular catheter increases the chances that these are the source of sepsis.

Gram stain of material from sites of possible infection may give early clues to the etiology of infection while cultures are incubating. As examples, urine should be routinely Gram stained and cultured, sputum should be examined in a patient with a productive cough, and an intra-abdominal collection in a postoperative patient should be percutaneously sampled under radiologic guidance.

Blood should be taken from two distinct venipuncture sites and inoculated into standard blood culture media. (See "Blood cultures for the detection of bacteremia".)

There is no single test that immediately confirms the diagnosis of severe sepsis or septic shock. However, several laboratory tests, all of which are still investigational, have been studied as diagnostic markers of active bacterial infection [6]:

• The plasma concentration of soluble TREM-1 (triggering receptor expressed on myeloid cells), a member of the immunoglobulin superfamily that is specifically upregulated in the presence of bacterial products, is increased in patients with sepsis [40-42]. In a small trial, increased TREM-1 levels were both sensitive and specific for the diagnosis of bacterial sepsis (96 and 89 percent, respectively) [40]. Serial monitoring of TREM-1 may also provide prognostic information in patients with established sepsis [41,42].
• Elevated serum procalcitonin levels are associated with bacterial infection and sepsis [43-45]. But, a meta-analysis of 18 studies found that procalcitonin distinguished sepsis from nonseptic systemic inflammation poorly (sensitivity of 71 percent and specificity of 71 percent) [44].

Evaluation of the clinical usefulness of both TREM-1 and procalcitonin is still in its earliest stages and should be considered preliminary. Until additional clinical investigations have been performed, we do not suggest the routine use of either.

Eradication of infection — Effective treatment of the active infection is essential to the successful treatment of severe sepsis and septic shock. Source control (physical measures undertaken to eradicate a focus of infection and eliminate or treat ongoing microbial proliferation and infection) should be undertaken since undrained foci of infection may not respond to antibiotics alone (table 3). As examples, potentially infected foreign bodies (eg, vascular access devices) should be removed when possible, and abscesses should undergo percutaneous or surgical drainage. Some patients require extensive soft tissue debridement or amputation; in rare cases, fulminant Clostridium difficile-associated colitis may necessitate colectomy [46].

Antimicrobial regimen — Intravenous antibiotic therapy should be initiated immediately after obtaining appropriate cultures. The choice of antibiotics can be complex and should consider the patient's history (eg, recent antibiotics received), comorbidities, clinical context (eg, community- or hospital-acquired), Gram stain data, and local resistance patterns [7,47,48].

Poor outcomes are associated with inadequate or inappropriate antimicrobial therapy (ie, treatment with antibiotics to which the pathogen was later shown to be resistant in vitro) [49-54]. They are also associated with delays in initiating antimicrobial therapy, even short delays (eg, an hour).

• A prospective cohort study of 2124 patients demonstrated that inappropriate antibiotic selection was surprisingly common (32 percent) [52]. Mortality was markedly increased in these patients compared to those who had received appropriate antibiotics (34 versus 18 percent).
• A retrospective analysis of 2731 patients with septic shock demonstrated that the time to initiation of appropriate antimicrobial therapy was the strongest predictor of mortality [53].

When the potential pathogen or infection source is not immediately obvious, we favor broad-spectrum antibiotic coverage directed against both gram-positive and gram-negative bacteria. Few guidelines exist for the initial selection of empiric antibiotics in severe sepsis or septic shock. In our practice, if Pseudomonas is an unlikely pathogen, we favor combining vancomycin with one of the following:

• Cephalosporin, 3rd or 4th generation (eg, ceftriaxone or cefotaxime), or
• Beta-lactam/beta-lactamase inhibitor (eg, piperacillin-tazobactam, ticarcillin-clavulanate), or
• Carbapenem (eg, imipenem or meropenem).

Alternatively, if Pseudomonas is a possible pathogen, we combine vancomycin with two of the following (see "Treatment of Pseudomonas aeruginosa infections":

• Antipseudomonal cephalosporin (eg, ceftazidime, cefepime), or
• Antipseudomonal carbapenem (eg, imipenem, meropenem), or
• Antipseudomonal beta-lactam/beta-lactamase inhibitor (eg, piperacillin-tazobactam,ticarcillin-clavulanate), or
• Fluoroquinolone with good anti-pseudomonal activity (eg, ciprofloxacin), or
• Aminoglycoside (eg, gentamicin, amikacin), or
• Monobactam (eg, aztreonam)

Selection of two agents from the same class, for example, two beta-lactams, should be avoided. We emphasize the importance of considering local susceptibility patterns when choosing an empiric antibiotic regimen.

Staphylococcus aureus is associated with significant morbidity if not treated early in the course of infection [55]. There is growing recognition that methicillin-resistant S. aureus (MRSA) is a cause of sepsis not only in hospitalized patients, but also in community dwelling individuals without recent hospitalization [56,57]. Many of these staphylococci have the Panton-Valentine leukocidin virulence factor, which causes severe, necrotizing infections [58]. For these reasons, we recommend that severely ill patients presenting with sepsis of unclear etiology be treated with intravenous vancomycin (adjusted for renal function) until the possibility of MRSA sepsis has been excluded. Linezolid is a reasonable alternative if there are contraindications to vancomycin.

After culture results and antimicrobial susceptibility data return, we recommend that therapy be pathogen- and susceptibility-directed, even if there has been clinical improvement while on the initial antimicrobial regimen. Gram-negative pathogens have historically been covered with two agents from different antibiotic classes. However, several clinical trials and two meta-analyses have failed to demonstrate superior overall efficacy of combination therapy compared to monotherapy with a third generation cephalosporin or a carbapenem [52,59-63]. Furthermore, one meta-analysis found double coverage was associated with an increased incidence of adverse events [62,63]. For this reason, we recommend use of a single agent with proven efficacy and the least possible toxicity, except in patients who are either neutropenic or whose severe sepsis is due to a known or suspected Pseudomonas infection [7,61]. (See "Pseudomonas aeruginosa bacteremia and endocarditis" and "Treatment of Pseudomonas aeruginosa infections".)

Regardless of the antibiotic regimen selected, patients should be observed closely for toxicity, evidence of response, and the development of nosocomial superinfection [64]. The duration of therapy is typically 7 to 10 days, although longer courses may be appropriate in patients who have a slow clinical response, an undrainable focus of infection, or immunologic deficiencies [7]. In patients who are neutropenic, antibiotic treatment should continue until the neutropenia has resolved. In non-neutropenic patients in whom infection is thoroughly excluded, antibiotics should be discontinued to minimize colonization or infection with drug-resistant microorganisms and superinfection with other pathogens.

ADDITIONAL THERAPIES

Recombinant human activated protein C — A number of coagulation abnormalities have been noted in severe sepsis and septic shock (table 1), and several reports suggested that protein C supplementation may produce clinical benefit, particularly in the setting of purpura fulminans [65-68].

These observations provided the rationale for the multicenter PROWESS trial that supported the efficacy of recombinant human activated protein C (drotrecogin alfa; Xigris) in patients with severe sepsis or septic shock [69]. This trial randomly assigned 1690 patients to receive a 96 hour infusion of drotrecogin alfa or placebo, beginning within 24 hours of presentation. Patients with acute renal failure, but not those with chronic renal failure were included. The full inclusion and exclusion criteria are shown in the table (table 7A-B).

The following findings were noted:

• The 28-day mortality rate was significantly lower in the drotrecogin-treated group (24.7 versus 30.8 percent).
• There was an increased incidence of serious bleeding, including fatal intracranial hemorrhage, in patients receiving drotrecogin alfa that did not achieve statistical significance.
• Based upon post-hoc analysis of the study data, drotrecogin alfa was of greater benefit in the most acutely ill patients, as identified by an APACHE II score ≥25 (calculator 1). An analysis of secondary endpoints suggested that the incidence of multiple organ dysfunction was lower in patients treated with drotrecogin alfa, and that therapy was associated with more rapid recovery of cardiac and pulmonary function [70]. (See "Predictive scoring systems for the severity of illness in the intensive care unit".)

Interpretation of these results is complicated by the fact that the study protocol was modified after the enrollment of 720 patients, excluding those with metastatic cancer, pancreatitis, and most organ transplant recipients [71]. In addition, the master cell line used by the manufacturer to produce drotrecogin alfa was changed during the trial. These modifications appeared to divide the trial into two phases during which the effect of drotrecogin alfa on 28-day mortality was different; there was no effect with original protocol (RR 0.94, 95% CI 0.75-1.17), but improvement with the modified protocol (RR 0.71, 95% CI 0.57-0.87) [72]. A new randomized, placebo-controlled trial is underway in an attempt to address these limitations.

A second open-label, single arm study of patients with severe sepsis (the ENHANCE trial) noted that 28-day all cause mortality for patients treated with drotrecogin alfa (24 mcg/kg per hour for 96 hours) was similar to that observed in PROWESS, although the incidence of intracranial hemorrhage was higher [73]. In addition, ENHANCE found that patients treated within 0 to 24 hours from their first sepsis-induced organ dysfunction had significantly lower mortality than those treated after 24 hours (22.9 compared to 27.4 percent).

The suggested dosing regimen of drotrecogin alfa is 24 mcg/kg per hour for 96 hours. Extending the duration of the infusion in patients who are still vasopressor dependent at the end of the infusion does not improve outcomes [74]. There is no suggested dose adjustment for patients with renal failure [75,76].

Patients with certain risk factors for bleeding were excluded from the PROWESS trial because the major adverse effect of drotrecogin alfa is bleeding. These risk factors are listed in the table (table 7A-B). However, many of these risk factors are not listed as contraindications on the product label. Whether patients with such risk factors are at increased risk for bleeding was evaluated in a retrospective cohort study of 73 patients who received drotrecogin alfa [77]. The incidence of serious bleeding was significantly higher among patients with one or more risk factors for bleeding, compared to patients without any risk factors (35 versus 4 percent). Serious bleeding was defined as a hemoglobin fall of ≥2 g/dL over <48 hours, a transfusion requirement of ≥4 units over 48 hours, objective evidence of bleeding, and clinician documentation of clinical consequences of the bleeding. This study suggests that extra caution is warranted when deciding whether to administer drotrecogin alfa to patients who have one or more of the risk factors for bleeding that were used as exclusion criteria during the PROWESS trial.

The economic impact and cost effectiveness of drotrecogin alfa therapy have been assessed using two separate decision-analysis models [78,79]. These studies found that, for patients with APACHE II scores ≥25, the cost per year of life saved with drotrecogin alfa is a cost-effective therapy in patients with severe sepsis and septic shock (calculator 1).

Less severe disease — The role of drotrecogin alfa in the treatment of adults with less severe disease was the subject of several trials [80-82].

The largest of these trials, the ADDRESS trial, evaluated patients with severe sepsis and a low risk of death (defined as an APACHE II score <25 or single organ dysfunction). It was designed to randomly assign 11,000 adult patients to treatment with drotrecogin alfa (24 mcg/kg per hour for 96 hours) or placebo within 48 hours of presentation [82]. This trial was stopped after 2600 patients completed the protocol, based on interim analysis suggesting that, despite its massive size, the trial was unlikely to demonstrate a significant difference in mortality among these patients with sepsis and a low overall risk of mortality. The 28-day and in-hospital mortality rates were similar in both groups; however, there was a significantly increased risk of serious bleeding among patients treated with drotrecogin alfa (3.9 versus 2.2 percent in the placebo group). At one year, the mortality rate remained similar in both groups [83]. There were no additional serious adverse events during the period between hospital discharge and one year.

In addition to the possible harm caused by drotrecogin alfa in patients with severe sepsis and a low risk of death, two studies found that drotrecogin alfa therapy was not cost-effective for patients with APACHE II scores <25 (calculator 1) [78,79].

These findings suggest drotrecogin alfa has little role in the management of patients with sepsis and a low risk of death, based on APACHE II score <25 or single organ failure (calculator 1).

Children — A large clinical trial of drotrecogin alfa in pediatric patients with sepsis (RESOLVE trial) was stopped early when interim analysis suggested that this intervention did not improve mortality, and was associated with an increased risk of intracranial hemorrhage, particularly in infants ≤60 days [84]. (See "Septic shock: Initial evaluation and management in children".)

SIRS — The value of drotrecogin alfa in patients with severe systemic inflammatory response syndrome (SIRS) is unclear. As an example, there are case reports of patients with severe acute pancreatitis, without evidence of infection, who have been treated with drotrecogin alfa [85]. However, larger dedicated trials to demonstrate therapeutic efficacy in SIRS patients will be needed before this approach can be recommended.

VTE prophylaxis during infusion — In theory, venous thromboembolism (VTE) prophylaxis with heparin during infusion of drotrecogin alfa may be unnecessary because drotrecogin alfa has antithrombotic and profibrinolytic properties [86]. It may also be harmful because heparin increases the clearance of drotrecogin alfa in vitro [87].

The Xigris and Prophylactic HeparRin Evaluation in Severe Sepsis (XPRESS) trial randomly assigned patients with severe sepsis or septic shock to receive subcutaneous unfractionated heparin (n = 511), subcutaneous low molecular weight heparin (n = 493), or placebo (n = 990) during their 96-hour drotrecogin alfa infusion [86]. Compared to patients who received placebo, those who received unfractionated or low molecular weight heparin had significantly more bleeding complications (10.8 versus 8.1) and fewer ischemic strokes (0.3 versus 1.3), as well as a nonstatistically significant reduction in 28-day mortality (28 versus 32 percent). Although total bleeding complications were increased among those who received heparin, the rate of serious bleeding events (intracranial hemorrhage, retinal hemorrhage, hemarthrosis, spinal hemorrhage, or other life threatening bleeding) was not [88].

Recognizing that many patients would receive heparin VTE prophylaxis prior to the drotrecogin alfa infusion, the investigators prospectively defined four subgroups [86]:

• Patients who received heparin both before and during the infusion
• Patients who received heparin before the infusion, but placebo during the infusion
• Patients who received no prophylaxis before the infusion, but heparin during the infusion
• Patients who received no prophylaxis before the infusion and placebo during the infusion

The 28-day mortality increased among patients whose heparin was replaced by placebo (36 percent), but was similar among the other groups (27 to 29 percent). This trial suggests that heparin VTE prophylaxis should not be discontinued during infusion of drotrecogin alfa, unless the potential risks of heparin outweigh the potential benefits.

Glucocorticoids — Glucocorticoids have long been investigated as therapeutic agents in sepsis because the pathogenesis of sepsis involves an intense and potentially deleterious host inflammatory response. This topic is discussed in detail separately. (See "Corticosteroid therapy in septic shock".)

Nutrition — There is consensus that nutritional support improves nutritional outcomes in critically ill patients, such as body weight and mid-arm muscle mass. However, it is uncertain whether nutritional support improves important clinical outcomes (eg, duration of mechanical ventilation, length of stay, mortality), or when nutritional support should be initiated. This topic is reviewed in detail elsewhere. (See "Overview of nutritional support in critically ill patients".)

Intensive insulin therapy — Hyperglycemia and insulin resistance are common in critically ill patients, independent of a history of diabetes mellitus [89]. As a result, intensive insulin therapy has been studied and a body of evidence has accumulated. This topic is discussed separately. (See "Glycemic control and intensive insulin therapy in critical illness".)

Protocols — Sepsis treatment protocols that incorporate early empiric antibiotics, restoration of tissue perfusion, glucocorticoids, glucose control, and recombinant human activated protein C may improve outcome [90-92]. This was illustrated by an observational cohort study of 120 patients with septic shock [92]. Implementation of a standardized hospital order set was associated with greater likelihood that the initial antibiotic regimen targeted the culprit microorganism (87 versus 72 percent), shorter hospital stay (9 versus 12 days), and lower 28-day mortality (30 versus 48 percent), compared to historical controls. It is impossible to determine which component or components of the protocol conferred the benefit.

SUMMARY AND RECOMMENDATIONS

• Sepsis is a clinical syndrome characterized by systemic inflammation and widespread tissue injury due to infection. There is a continuum of illness severity ranging from sepsis to severe sepsis and septic shock. When infection is absent, the clinical syndrome is termed systemic inflammatory response syndrome (SIRS). (See 'Introduction' above.)

• Initial management is aimed at securing the airway and correcting hypoxemia. Intubation and mechanical ventilation may be required. (See 'Stabilize respiration' above.)

• Once the patient's respiratory status has been stabilized, the adequacy of perfusion should be assessed. Hypotension is the most common indicator that perfusion is inadequate. However, critical hypoperfusion can also occur in the absence of hypotension, especially during early sepsis. Common signs of hypoperfusion include cool, vasoconstricted skin due to redirection of blood flow to core organs (although warm, flushed skin may be present in the early phases of sepsis), obtundation or restlessness, oliguria or anuria, and lactic acidosis. (See 'Assess perfusion' above.)

• Once it has been established that hypoperfusion exists, early restoration of perfusion is necessary to prevent or limit multiple organ dysfunction, as well as reduce mortality. Tissue perfusion should be promptly restored using intravenous fluids, vasopressors, red blood cell transfusions, and inotropes (algorithm 1). We recommend patients be managed with therapy aimed at achieving a central (or mixed) venous oxygen saturation ≥70 percent within six hours of presentation (Grade 1B). It is reasonable to simultaneously aim for a central venous pressure 8 to 12 mmHg, mean arterial pressure (MAP) ≥65 mmHg, and urine output ≥0.5 mL per kg per hour. (See 'Restore perfusion' above.)

• We recommend boluses of intravenous fluids as first-line therapy in patients who demonstrate impaired perfusion (Grade 1B). Fluid boluses are repeated until blood pressure and tissue perfusion are acceptable, pulmonary edema ensues, or there is no further response. These parameters should be assessed before and after each fluid bolus. There are no data to support preferential administration of crystalloid or colloid. (See 'Intravenous fluids' above,)
• We recommend vasopressors for patients who remain hypotensive following intravascular volume repletion (Grade 1B). Although there is no definitive evidence of the superiority of one vasopressor over another, we suggest beginning with norepinephrine (Grade 2C). (See 'Vasopressors' above.)

• For patients whose ScvO2 remains <70 percent after intravenous fluid and vasopressor therapy, it is reasonable to administer additional therapies, including blood transfusions or inotropic therapy. (See 'Additional therapies' above.)

• Prompt identification and treatment of the site of infection are essential. Sputum and urine should be collected for gram stain and culture. Intra-abdominal fluid collections should be percutaneously sampled. Blood should be taken from two distinct venipuncture sites and from indwelling vascular access devices and cultured aerobically and anaerobically. (See 'Identification of the septic focus' above.)

• Antibiotics should be administered immediately after appropriate cultures have been obtained. We recommend empiric broad spectrum antibiotics when a definite source of infection can not be identified (Grade 1B). (See 'Antimicrobial regimen' above.)

• Potentially infected vascular access devices should be removed (if possible), abscesses should be drained, and extensive soft tissue infections should be debrided or amputated (table 3). (See 'Eradication of infection' above.)

• In patients with septic shock or severe sepsis with a high risk of death, defined as an APACHE II score >25, multiple organ dysfunction, or sepsis-induced acute respiratory distress syndrome, we suggest that recombinant human activated protein C be administered if contraindications do not exist (Grade 2B). Effort should be made to initiate the infusion within 24 hours from the first-sepsis induced organ dysfunction. (See 'Recombinant human activated protein C' above.)
• Recombinant human activated protein C appears to confer no benefit in children or in patients with severe sepsis and a low risk of death (defined as an APACHE II score <25 or single organ dysfunction) and is associated with increasing bleeding. We suggest that recombinant human activated protein C not be administered to these patient populations (Grade 2B). Its impact in SIRS is unclear. (See 'Recombinant human activated protein C' above.)

• Glucocorticoid therapy, nutritional support, and glucose control are additional issues that are important in the management of patients with severe sepsis or septic shock. Each is discussed in detail in separate topic reviews. (See "Corticosteroid therapy in septic shock" and "Overview of nutritional support in critically ill patients" and "Glycemic control and intensive insulin therapy in critical illness".)