CHAPTER 19: SEVERE SEPSIS AND SEPTIC SHOCK

(Table 19-1) Animals mount both local and systemic responses to microbes that traverse their epithelial barriers and enter underlying tissues. Fever or hypothermia, leukocytosis or leukopenia, tachypnea, and tachycardia are cardinal signs of the systemic response. To date, attempts to devise precise definitions for the harmful systemic reaction to infection (“sepsis”) have not resulted in a clinically useful level of specificity, in part because the systemic responses to infection, trauma, and other major stresses can be so similar. In general, when an infectious etiology is proven or strongly suspected and the response results in hypofunction of uninfected organs, the term sepsis (or severe sepsis) should be used. Septic shock refers to sepsis accompanied by hypotension that cannot be corrected by the infusion of fluids.

The systemic response to any class of microorganism can be harmful. Microbial invasion of the bloodstream is not essential because local inflammation can also elicit distant organ dysfunction and hypotension. In fact, blood cultures yield bacteria or fungi in only ~20–40% of cases of severe sepsis and 40–70% of cases of septic shock. In a prevalence study of 14,414 patients in intensive care units (ICUs) from 75 countries in 2007, 51% of patients were considered infected. Respiratory infection was most common (64%). Microbiologic results were positive in 70% of individuals considered infected; of the isolates, 62% were gram-negative bacteria (Pseudomonas species and Escherichia coli were most common), 47% were gram-positive bacteria (Staphylococcus aureus was most common), and 19% were fungi (Candida species). This distribution is similar to that reported a decade earlier from eight academic centers in the United States (Table 19-2). In patients whose blood cultures are negative, the etiologic agent is often established by culture or microscopic examination of infected material from a local site; specific identification of microbial DNA or RNA in blood or tissue samples is also used. In some case series, a majority of patients with a clinical picture of severe sepsis or septic shock have had negative microbiologic data.

Severe sepsis is a contributing factor in >200,000 deaths per year in the United States. The incidence of severe sepsis and septic shock has increased over the past 30 years, and the annual number of cases is now >750,000 (~3 per 1000 population). Approximately two-thirds of the cases occur in patients with significant underlying illness. Sepsis-related incidence and mortality rates increase with age and preexisting comorbidity. The rising incidence of severe sepsis in the United States has been attributable to the aging of the population, the increasing longevity of patients with chronic diseases, and the relatively high frequency with which sepsis has occurred in patients with AIDS. The widespread use of immunosuppressive drugs, indwelling catheters, and mechanical devices has also played a role. In the aforementioned international ICU prevalence study, the case–fatality rate among infected patients (33%) greatly exceeded that among uninfected patients (15%).

Invasive bacterial infections are prominent causes of death around the world, particularly among young children. In sub-Saharan Africa, for example, careful screening for positive blood cultures found that community-acquired bacteremia accounted for at least one-fourth of deaths of children >1 year of age. Nontyphoidal Salmonella species, Streptococcus pneumoniae, Haemophilus influenzae, and E. coli were the most commonly isolated bacteria. Bacteremic children often had HIV infection or were severely malnourished.

Sepsis is triggered most often by bacteria or fungi that do not ordinarily cause systemic disease in immunocompetent hosts (Table 19-2). To survive within the human body, these microbes often exploit acquired deficiencies in host defenses, indwelling catheters or other foreign matter, or obstructed fluid drainage conduits. Microbial pathogens, in contrast, can circumvent innate defenses because they (1) lack molecules that can be recognized by host receptors (see below) or (2) elaborate toxins or other virulence factors. In both cases, the body can mount a vigorous inflammatory reaction that results in sepsis or septic shock yet fails to kill the invaders. The septic response may also be induced by microbial exotoxins that act as superantigens (e.g., toxic shock syndrome toxin 1; Chap. 43) as well as by many pathogenic viruses.

Host mechanisms for sensing microbes

Animals have exquisitely sensitive mechanisms for recognizing and responding to certain highly conserved microbial molecules. Recognition of the lipid A moiety of lipopolysaccharide (LPS, also called endotoxin; Chap. 2) is the best-studied example. A host protein (LPS-binding protein) binds lipid A and transfers the LPS to CD14 on the surfaces of monocytes, macrophages, and neutrophils. LPS then is passed to MD-2, a small receptor protein that is bound to Toll-like receptor (TLR) 4 to form a molecular complex that transduces the LPS recognition signal to the interior of the cell. This signal rapidly triggers the production and release of mediators, such as tumor necrosis factor (TNF; see below), that amplify the LPS signal and transmit it to other cells and tissues. Bacterial peptidoglycan and lipopeptides elicit responses in animals that are generally similar to those induced by LPS, although they interact with different TLRs. Having numerous TLR-based receptor complexes (10 different TLRs have been identified in humans) allows animals to recognize many conserved microbial molecules; others include lipopeptides (TLR2/1, TLR2/6), flagellin (TLR5), undermethylated DNA CpG sequences (TLR9), single-stranded RNA (TLR7, 8), and double-stranded RNA (TLR3). The ability of some TLRs to serve as receptors for host ligands (e.g., hyaluronans, heparan sulfate, saturated fatty acids, high-mobility group box 1) raises the possibility that they also play a role in producing noninfectious sepsis-like states. Other host pattern-recognition proteins that are important for sensing microbes include the intracellular NOD1 and NOD2 proteins, which recognize discrete fragments of bacterial peptidoglycan; the inflammasome, which senses some pathogens and produces interleukin (IL) 1β and IL-18; early complement components (principally in the alternative pathway); mannose-binding lectin and C-reactive protein, which activate the classic complement pathway; and Dectin-1 and complement receptor 3, which sense fungal β-glucan.

A host’s ability to recognize certain microbial molecules may influence both the potency of its own defenses and the pathogenesis of severe sepsis. For example, MD-2–TLR4 best senses LPS that has a bisphosphorylated, hexaacyl lipid A moiety (i.e., one with two phosphates and six fatty acyl chains). Most of the commensal aerobic and facultatively anaerobic gram-negative bacteria that trigger severe sepsis and shock (including E. coli, Klebsiella, and Enterobacter) make this lipid A structure. When they invade human hosts, often through breaks in an epithelial barrier, they are typically confined to the subepithelial tissue by a localized inflammatory response. Bacteremia, if it occurs, is intermittent and low grade because these bacteria are efficiently cleared from the bloodstream by TLR4-expressing Kupffer cells and splenic macrophages. These mucosal commensals seem to induce severe sepsis most often by triggering severe local tissue inflammation rather than by circulating within the bloodstream. One exception is Neisseria meningitidis. Its hexaacyl LPS seems to be shielded from host recognition by its polysaccharide capsule. This protection may allow meningococci to transit undetected from the nasopharyngeal mucosa into the bloodstream, where they can infect vascular endothelial cells and release large amounts of endotoxin and DNA. Host recognition of lipid A may nonetheless influence pathogenesis, as meningococci that produce pentaacyl LPS were isolated from the blood of patients with less severe coagulopathy than was found in patients whose isolates produced hexaacyl lipid A; underacylated N. meningitidis LPS has also been found in many isolates from patients with chronic meningococcemia. In contrast, gram-negative bacteria that make lipid A with fewer than six acyl chains (Yersinia pestis, Francisella tularensis, Vibrio vulnificus, Pseudomonas aeruginosa, and Burkholderia pseudomallei, among others) are poorly recognized by MD-2–TLR4. When these bacteria enter the body, they may initially induce relatively little inflammation. When they do trigger severe sepsis, it is often after they have multiplied to high density in tissues and blood. The importance of LPS recognition in disease pathogenesis was demonstrated by engineering of a virulent strain of Y. pestis that makes tetraacyl LPS at 37°C to produce hexaacyl LPS; unlike its virulent parent, the mutant strain stimulated local inflammation and was rapidly cleared from tissues. These findings were subsequently replicated in F. tularensis. For at least one large class of microbes—gram-negative aerobic bacteria—the pathogenesis of sepsis thus depends, at least in part, on whether the bacterium’s major signal molecule, LPS, can be sensed by the host.

Local and systemic host responses to invading microbes

Recognition of microbial molecules by tissue phagocytes triggers the production and/or release of numerous host molecules (cytokines, chemokines, prostanoids, leukotrienes, and others) that increase blood flow to the infected tissue (rubor), enhance the permeability of local blood vessels (tumor), recruit neutrophils and other cells to the site of infection (calor), and elicit pain (dolor). These reactions are familiar elements of local inflammation, the body’s frontline innate immune mechanism for eliminating microbial invaders. Systemic responses are activated by neural and/or humoral communication with the hypothalamus and brainstem; these responses enhance local defenses by increasing blood flow to the infected area, augmenting the number of circulating neutrophils, and elevating blood levels of numerous molecules (such as the microbial recognition proteins discussed above) that have anti-infective functions.

Cytokines and other mediators

Cytokines can exert endocrine, paracrine, and autocrine effects. TNF-α stimulates leukocytes and vascular endothelial cells to release other cytokines (as well as additional TNF-α), to express cell-surface molecules that enhance neutrophil endothelial adhesion at sites of infection, and to increase prostaglandin and leukotriene production. Whereas blood levels of TNF-α are not elevated in individuals with localized infections, they increase in most patients with severe sepsis or septic shock. Moreover, IV infusion of TNF-α can elicit fever, tachycardia, hypotension, and other responses. In animals, larger doses of TNF-α induce shock and death.

Although TNF-α is a central mediator, it is only one of many proinflammatory molecules that contribute to innate host defense. Chemokines, most prominently IL-8 and IL-17, attract circulating neutrophils to the infection site. IL-1β exhibits many of the same activities as TNF-α. TNF-α, IL-1β, interferon γ, IL-12, IL-17, and other proinflammatory cytokines probably interact synergistically with one another and with additional mediators. The nonlinearity and multiplicity of these interactions have made it difficult to interpret the roles played by individual mediators in both tissues and blood.

Coagulation factors

Intravascular thrombosis, a hallmark of the local inflammatory response, may help wall off invading microbes and prevent infection and inflammation from spreading to other tissues. IL-6 and other mediators promote intravascular coagulation initially by inducing blood monocytes and vascular endothelial cells to express tissue factor. When tissue factor is expressed on cell surfaces, it binds to factor VIIa to form an active complex that can convert factors X and IX to their enzymatically active forms. The result is activation of both extrinsic and intrinsic clotting pathways, culminating in the generation of fibrin. Clotting is also favored by impaired function of the protein C–protein S inhibitory pathway and depletion of antithrombin and proteins C and S, whereas fibrinolysis is reduced by increases in plasma levels of plasminogen activator inhibitor 1. Thus, there may be a striking propensity toward intravascular fibrin deposition, thrombosis, and bleeding; this propensity has been most apparent in patients with intravascular endothelial infections such as meningococcemia (Chap. 52). Evidence points to tissue factor–expressing microparticles derived from leukocytes as a potential trigger for intravascular coagulation. The contact system is activated during sepsis but contributes more to the development of hypotension than to that of disseminated intravascular coagulation (DIC).

Neutrophil extracellular traps (NETs) are produced when neutrophils, stimulated by microbial agonists or IL-8, release granule proteins and chromatin to form an extracellular fibrillar matrix. NETs kill bacteria and fungi with antimicrobial granule proteins (e.g., elastase) and histones. It has been reported that NETs can form within hepatic sinusoids in animals injected with large amounts of LPS, and platelets can induce NET formation without killing neutrophils. A role played by NETs in organ hypofunction during sepsis has been proposed but not established.

Control mechanisms

Elaborate control mechanisms operate within both local sites of inflammation and the systemic compartment.

Local control mechanisms

Host recognition of invading microbes within subepithelial tissues typically ignites immune responses that rapidly kill the invaders and then subside to allow tissue recovery. The forces that put out the fire and clean up the battleground include molecules that neutralize or inactivate microbial signals. Among these molecules are intracellular factors (e.g., suppressor of cytokine signaling 3 and IL-1 receptor–associated kinase 3) that diminish the production of proinflammatory mediators by neutrophils and macrophages; anti-inflammatory cytokines (IL-10, IL-4); and molecules derived from essential polyunsaturated fatty acids (lipoxins, resolvins, and protectins) that promote tissue restoration. Enzymatic inactivation of microbial signal molecules (e.g., LPS) may be required to restore homeostasis; a leukocyte enzyme, acyloxyacyl hydrolase, has been shown to prevent prolonged inflammation in mice by inactivating LPS.

Systemic control mechanisms

The signaling apparatus that links microbial recognition to cellular responses in tissues is less active in the blood. For example, whereas LPS-binding protein plays a role in recognizing LPS, in plasma it also prevents LPS signaling by transferring LPS molecules into plasma lipoprotein particles that sequester the lipid A moiety so that it cannot interact with cells. At the high concentrations found in blood, LPS-binding protein also inhibits monocyte responses to LPS, and the soluble (circulating) form of CD14 strips off LPS that has bound to monocyte surfaces.

Systemic responses to infection also diminish cellular responses to microbial molecules. Circulating levels of cortisol and anti-inflammatory cytokines (e.g., IL-6 and IL-10) increase even in patients with minor infections. Glucocorticoids inhibit cytokine synthesis by monocytes in vitro; the increase in blood cortisol levels that occurs early in the systemic response presumably plays a similarly inhibitory role. Epinephrine inhibits the TNF-α response to endotoxin infusion in humans while augmenting and accelerating the release of IL-10; prostaglandin E₂ has a similar “reprogramming” effect on the responses of circulating monocytes to LPS and other bacterial agonists. Cortisol, epinephrine, IL-10, and C-reactive protein reduce the ability of neutrophils to attach to vascular endothelium, favoring their demargination and thus contributing to leukocytosis while preventing neutrophil-endothelial adhesion in uninflamed organs. Studies in rodents have found that macrophage cytokine synthesis is inhibited by acetylcholine that is produced by choline acetyltransferase–secreting CD4+ T cells in response to stimulation by norepinephrine, whereas acetylcholine-producing B cells reduce neutrophil infiltration into tissues. Several lines of evidence thus suggest that the body’s neuroendocrine responses to injury and infection normally prevent inflammation within organs distant from a site of infection. There is also evidence that these responses may be immunosuppressive.

IL-6 plays important roles in the systemic compartment. Released by many different cell types, IL-6 is an important stimulus to the hypothalamic-pituitary-adrenal axis, is the major procoagulant cytokine, and is a principal inducer of the acute-phase response, which increases the blood concentrations of numerous molecules that have anti-infective, procoagulant, or anti-inflammatory actions. Blood levels of IL-1 receptor antagonist often greatly exceed those of circulating IL-1β, for example, and this excess may inhibit the binding of IL-1β to its receptors. High levels of soluble TNF receptors neutralize TNF-α that enters the circulation. Other acute-phase proteins are protease inhibitors or antioxidants; these may neutralize potentially harmful molecules released from neutrophils and other inflammatory cells. Increased hepatic production of hepcidin (stimulated largely by IL-6) promotes the sequestration of iron in hepatocytes, intestinal epithelial cells, and erythrocytes; this effect reduces iron acquisition by invading microbes while contributing to the normocytic, normochromic anemia associated with inflammation.

It may thus be said that both local and systemic responses to infectious agents benefit the host in important ways. Most of these responses and the molecules responsible for them have been highly conserved during animal evolution and therefore may be adaptive. Elucidating how they become maladaptive and contribute to lethality remains a major challenge for sepsis research.

Organ dysfunction and shock

As the body’s responses to infection intensify, the mixture of circulating cytokines and other molecules becomes very complex: elevated blood levels of more than 60 molecules have been found in patients with septic shock. Although high concentrations of both pro- and anti-inflammatory molecules are found, the net mediator balance in the plasma of these extremely sick patients seems to be anti-inflammatory. For example, blood leukocytes from patients with severe sepsis are often hyporesponsive to agonists such as LPS. In patients with severe sepsis, persistence of leukocyte hyporesponsiveness has been associated with an increased risk of dying; at this time, the most predictive biomarker is a decrease in the expression of HLA-DR (class II) molecules on the surfaces of circulating monocytes, a response that seems to be induced by cortisol and/or IL-10. Apoptotic death of B cells, follicular dendritic cells, and CD4+ T lymphocytes also may contribute significantly to the immunosuppressive state.

Endothelial injury

Given the vascular endothelium’s important roles in regulating vascular tone, vascular permeability, and coagulation, many investigators have favored widespread vascular endothelial injury as the major mechanism for multiorgan dysfunction. In keeping with this idea, one study found high numbers of vascular endothelial cells in the peripheral blood of septic patients. Leukocyte-derived mediators and platelet-leukocyte-fibrin thrombi may contribute to vascular injury, but the vascular endothelium also seems to play an active role. Stimuli such as TNF-α induce vascular endothelial cells to produce and release cytokines, procoagulant molecules, platelet-activating factor, nitric oxide, and other mediators. In addition, regulated cell-adhesion molecules promote the adherence of neutrophils to endothelial cells. Although these responses can attract phagocytes to infected sites and activate their antimicrobial arsenals, endothelial cell activation can also promote increased vascular permeability, microvascular thrombosis, DIC, and hypotension.

Tissue oxygenation may decrease as the number of functional capillaries is reduced by luminal obstruction due to swollen endothelial cells, decreased deformability of circulating erythrocytes, leukocyte-platelet-fibrin thrombi, or compression by edema fluid. On the other hand, studies using orthogonal polarization spectral imaging of the microcirculation in the tongue found that sepsis-associated derangements in capillary flow could be reversed by applying acetylcholine to the surface of the tongue or by giving nitroprusside intravenously; these observations suggest a neuroendocrine basis for the loss of capillary filling. Oxygen utilization by tissues may also be impaired by changes (possibly induced by nitric oxide) that decrease oxidative phosphorylation and ATP production while increasing glycolysis. The local accumulation of lactic acid, a consequence of increased glycolysis, may decrease extracellular pH and contribute to the slowdown in cellular metabolism that occurs within affected tissues.

Remarkably, poorly functioning “septic” organs usually appear normal at autopsy. There is typically very little necrosis or thrombosis, and apoptosis is largely confined to lymphoid organs and the gastrointestinal tract. Moreover, organ function usually returns to normal if patients recover. These points suggest that organ dysfunction during severe sepsis has a basis that is principally biochemical, not structural.

Septic shock

The hallmark of septic shock is a decrease in peripheral vascular resistance that occurs despite increased levels of vasopressor catecholamines. Before this vasodilatory phase, many patients experience a period during which oxygen delivery to tissues is compromised by myocardial depression, hypovolemia, and other factors. During this “hypodynamic” period, the blood lactate concentration is elevated and central venous oxygen saturation is low. Fluid administration is usually followed by the hyperdynamic vasodilatory phase, during which cardiac output is normal (or even high) and oxygen consumption declines despite adequate oxygen delivery. The blood lactate level may be normal or increased, and normalization of central venous oxygen saturation may reflect improved oxygen delivery, decreased oxygen uptake by tissues, or left-to-right shunting.

Prominent hypotensive molecules include nitric oxide, β-endorphin, bradykinin, platelet-activating factor, and prostacyclin. Agents that inhibit the synthesis or action of each of these mediators can prevent or reverse endotoxic shock in animals. However, in clinical trials, neither a platelet-activating factor receptor antagonist nor a bradykinin antagonist improved survival rates among patients with septic shock, and a nitric oxide synthase inhibitor, L-N^G-methylarginine HCl, actually increased the mortality rate.

Severe sepsis: a single pathogenesis?

In some cases, circulating bacteria and their products almost certainly elicit multiorgan dysfunction and hypotension by directly stimulating inflammatory responses within the vasculature. In patients with fulminant meningococcemia, for example, mortality rates have correlated directly with blood levels of endotoxin and bacterial DNA and with the occurrence of DIC (Chap. 52). In most patients infected with other gram-negative bacteria, in contrast, circulating bacteria or bacterial molecules may reflect uncontrolled infection at a local tissue site and have little or no direct impact on distant organs; in these patients, inflammatory mediators or neural signals arising from the local site seem to be the key triggers for severe sepsis and septic shock. In a large series of patients with positive blood cultures, the risk of developing severe sepsis was strongly related to the site of primary infection: bacteremia arising from a pulmonary or abdominal source was eightfold more likely to be associated with severe sepsis than was bacteremic urinary tract infection, even after the investigators controlled for age, the kind of bacteria isolated from the blood, and other factors. A third pathogenesis may be represented by severe sepsis due to superantigen-producing S. aureus or Streptococcus pyogenes; the T cell activation induced by these toxins produces a cytokine profile that differs substantially from that elicited by gram-negative bacterial infection. Further evidence for different pathogenetic pathways has come from observations that the pattern of mRNA expression in peripheral-blood leukocytes from children with sepsis is different for gram-positive, gram-negative, and viral pathogens.

The pathogenesis of severe sepsis thus may differ according to the infecting microbe, the ability of the host’s innate defense mechanisms to sense and respond to it, the site of the primary infection, the presence or absence of immune defects, and the prior physiologic status of the host. Genetic factors are probably important as well, yet despite much study very few allelic polymorphisms have been associated with sepsis severity in more than one or two analyses. Further studies in this area are needed.

The manifestations of the septic response are superimposed on the symptoms and signs of the patient’s underlying illness and primary infection. The rate at which severe sepsis develops may differ from patient to patient, and there are striking individual variations in presentation. For example, some patients with sepsis are normo- or hypothermic; the absence of fever is most common in neonates, in elderly patients, and in persons with uremia or alcoholism.

Hyperventilation, producing respiratory alkalosis, is often an early sign of the septic response. Disorientation, confusion, and other manifestations of encephalopathy may also develop early on, particularly in the elderly and in individuals with preexisting neurologic impairment. Focal neurologic signs are uncommon, although preexisting focal deficits may become more prominent.

Hypotension and DIC predispose to acrocyanosis and ischemic necrosis of peripheral tissues, most commonly the digits. Cellulitis, pustules, bullae, or hemorrhagic lesions may develop when hematogenous bacteria or fungi seed the skin or underlying soft tissue. Bacterial toxins may also be distributed hematogenously and elicit diffuse cutaneous reactions. On occasion, skin lesions may suggest specific pathogens. When sepsis is accompanied by cutaneous petechiae or purpura, infection with N. meningitidis (or, less commonly, H. influenzae) should be suspected (see Fig. 14-42); in a patient who has been bitten by a tick while in an endemic area, petechial lesions also suggest Rocky Mountain spotted fever (see Fig. 83-1). A cutaneous lesion seen almost exclusively in neutropenic patients is ecthyma gangrenosum, often caused by P. aeruginosa. This bullous lesion surrounded by edema undergoes central hemorrhage and necrosis (see Fig. 61-1). Histopathologic examination shows bacteria in and around the wall of a small vessel, with little or no neutrophilic response. Hemorrhagic or bullous lesions in a septic patient who has recently eaten raw oysters suggest V. vulnificus bacteremia, whereas such lesions in a patient who has recently sustained a dog bite may indicate bloodstream infection due to Capnocytophaga canimorsus or Capnocytophaga cynodegmi. Generalized erythroderma in a septic patient suggests the toxic shock syndrome due to S. aureus or S. pyogenes.

Gastrointestinal manifestations such as nausea, vomiting, diarrhea, and ileus may suggest acute gastroenteritis. Stress ulceration can lead to upper gastrointestinal bleeding. Cholestatic jaundice, with elevated levels of serum bilirubin (mostly conjugated) and alkaline phosphatase, may precede other signs of sepsis. Hepatocellular or canalicular dysfunction appears to underlie most cases, and the results of hepatic function tests return to normal with resolution of the infection. Prolonged or severe hypotension may induce acute hepatic injury or ischemic bowel necrosis.

Many tissues may be unable to extract oxygen normally from the blood, so that anaerobic metabolism occurs despite near-normal mixed venous oxygen saturation. Blood lactate levels rise early because of increased glycolysis as well as impaired clearance of the resulting lactate and pyruvate by the liver and kidneys. The blood glucose concentration often increases, particularly in patients with diabetes, although impaired gluconeogenesis and excessive insulin release on occasion produce hypoglycemia. The cytokine-driven acute-phase response inhibits the synthesis of transthyretin while enhancing the production of C-reactive protein, fibrinogen, and complement components. Protein catabolism is often markedly accelerated. Serum albumin levels decline as a result of decreased hepatic synthesis and the movement of albumin into interstitial spaces.

Cardiopulmonary complications

Ventilation-perfusion mismatching produces a fall in arterial Po₂ early in the course. Increasing alveolar epithelial injury and capillary permeability result in increased pulmonary water content, which decreases pulmonary compliance and interferes with oxygen exchange. In the absence of pneumonia or heart failure, progressive diffuse pulmonary infiltrates and arterial hypoxemia occurring within 1 week of a known insult indicate the development of mild acute respiratory distress syndrome (ARDS) (200 mmHg < Pao₂/Fio₂ ≤ 300 mmHg), moderate ARDS (100 mmHg < Pao₂/Fio₂ ≤ 200 mmHg), or severe ARDS (Pao₂/Fio₂ ≤100 mmHg). Acute lung injury or ARDS develops in ~50% of patients with severe sepsis or septic shock. Respiratory muscle fatigue can exacerbate hypoxemia and hypercapnia. An elevated pulmonary capillary wedge pressure (>18 mmHg) suggests fluid volume overload or cardiac failure rather than ARDS. Pneumonia caused by viruses or by Pneumocystis may be clinically indistinguishable from ARDS.

Sepsis-induced hypotension (see “Septic Shock,” above) usually results initially from a generalized maldistribution of blood flow and blood volume and from hypovolemia that is due, at least in part, to diffuse capillary leakage of intravascular fluid. Other factors that may decrease effective intravascular volume include dehydration from antecedent disease or insensible fluid losses, vomiting or diarrhea, and polyuria. During early septic shock, systemic vascular resistance is usually elevated and cardiac output may be low. After fluid repletion, in contrast, cardiac output typically increases and systemic vascular resistance falls. Indeed, normal or increased cardiac output and decreased systemic vascular resistance distinguish septic shock from cardiogenic, extracardiac obstructive, and hypovolemic shock; other processes that can produce this combination include anaphylaxis, beriberi, cirrhosis, and overdoses of nitroprusside or narcotics.

Depression of myocardial function, manifested as increased end-diastolic and systolic ventricular volumes with a decreased ejection fraction, develops within 24 h in most patients with severe sepsis. Cardiac output is maintained despite the low ejection fraction because ventricular dilation permits a normal stroke volume. In survivors, myocardial function returns to normal over several days. Although myocardial dysfunction may contribute to hypotension, refractory hypotension is usually due to low systemic vascular resistance, and death most often results from refractory shock or the failure of multiple organs rather than from cardiac dysfunction per se.

Adrenal insufficiency

The diagnosis of adrenal insufficiency may be very difficult in critically ill patients. Whereas a plasma cortisol level of ≤15 μg/mL (≤10 μg/mL if the serum albumin concentration is <2.5 mg/dL) indicates adrenal insufficiency (inadequate production of cortisol), many experts now feel that the adrenocorticotropic hormone (CoSyntropin®) stimulation test is not useful for detecting less profound degrees of corticosteroid deficiency in patients who are critically ill. The concept of critical illness–related corticosteroid insufficiency (CIRCI) was proposed to encompass the different mechanisms that may produce corticosteroid activity that is inadequate for the severity of a patient’s illness. Although CIRCI may result from structural damage to the adrenal gland, it is more commonly due to reversible dysfunction of the hypothalamic-pituitary axis or to tissue corticosteroid resistance resulting from abnormalities of the glucocorticoid receptor or increased conversion of cortisol to cortisone. The major clinical manifestation of CIRCI is hypotension that is refractory to fluid replacement and requires pressor therapy. Some classic features of adrenal insufficiency, such as hyponatremia and hyperkalemia, are usually absent; others, such as eosinophilia and modest hypoglycemia, may sometimes be found. Specific etiologies include fulminant N. meningitidis bacteremia, disseminated tuberculosis, AIDS (with cytomegalovirus, Mycobacterium avium-intracellulare, or Histoplasma capsulatum disease), or the prior use of drugs that diminish glucocorticoid production, such as glucocorticoids, megestrol, etomidate, or ketoconazole.

Renal complications

Oliguria, azotemia, proteinuria, and nonspecific urinary casts are frequently found. Many patients are inappropriately polyuric; hyperglycemia may exacerbate this tendency. Most renal failure is due to acute tubular necrosis induced by hypovolemia, arterial hypotension, or toxic drugs, although some patients also have glomerulonephritis, renal cortical necrosis, or interstitial nephritis. Drug-induced renal damage may greatly complicate therapy, particularly when hypotensive patients are given aminoglycoside antibiotics. Nosocomial sepsis following acute renal injury is associated with a high mortality rate.

Coagulopathy

Although thrombocytopenia occurs in 10–30% of patients, the underlying mechanisms are not understood. Platelet counts are usually very low (<50,000/μL) in patients with DIC; these low counts may reflect diffuse endothelial injury or microvascular thrombosis, yet thrombi have only infrequently been found on biopsy of septic organs.

Neurologic complications

Delirium (acute encephalopathy) is often an early manifestation of sepsis. Depending on the diagnostic criteria used, it occurs in 10–70% of septic patients at some point during the hospital course. When the septic illness lasts for weeks or months, “critical illness” polyneuropathy may prevent weaning from ventilatory support and produce distal motor weakness. Electrophysiologic studies are diagnostic. Guillain-Barré syndrome, metabolic disturbances, and toxin activity must be ruled out. Recent studies have documented long-term cognitive loss in survivors of severe sepsis.

Immunosuppression

Patients with severe sepsis often become profoundly immunosuppressed. Manifestations include loss of delayed-type hypersensitivity reactions to common antigens, failure to control the primary infection, and increased risk for secondary infections (e.g., by opportunists such as Stenotrophomonas maltophilia, Acinetobacter calcoaceticus-baumannii, and Candida albicans). Approximately one-third of patients experience reactivation of herpes simplex virus, varicella-zoster virus, or cytomegalovirus infections; the latter are thought to contribute to adverse outcomes in some instances.

Abnormalities that occur early in the septic response may include leukocytosis with a left shift, thrombocytopenia, hyperbilirubinemia, and proteinuria. Leukopenia may develop. The neutrophils may contain toxic granulations, Döhle bodies, or cytoplasmic vacuoles. As the septic response becomes more severe, thrombocytopenia worsens (often with prolonged thrombin time, decreased fibrinogen, and the presence of d-dimers, suggesting DIC), azotemia and hyperbilirubinemia become more prominent, and levels of aminotransferases rise. Active hemolysis suggests clostridial bacteremia, malaria, a drug reaction, or DIC; in the case of DIC, microangiopathic changes may be seen on a blood smear.

During early sepsis, hyperventilation induces respiratory alkalosis. With respiratory muscle fatigue and the accumulation of lactate, metabolic acidosis (with increased anion gap) typically supervenes. Evaluation of arterial blood gases reveals hypoxemia that is initially correctable with supplemental oxygen but whose later refractoriness to 100% oxygen inhalation indicates right-to-left shunting. The chest radiograph may be normal or may show evidence of underlying pneumonia, volume overload, or the diffuse infiltrates of ARDS. The electrocardiogram may show only sinus tachycardia or nonspecific ST–T wave abnormalities.

Most diabetic patients with sepsis develop hyperglycemia. Severe infection may precipitate diabetic ketoacidosis that may exacerbate hypotension. Hypoglycemia occurs rarely and may indicate adrenal insufficiency. The serum albumin level declines as sepsis continues. Hypocalcemia is rare.

There is no specific diagnostic test for sepsis. Diagnostically sensitive findings in a patient with suspected or proven infection include fever or hypothermia, tachypnea, tachycardia, and leukocytosis or leukopenia (Table 19-1); acutely altered mental status, thrombocytopenia, an elevated blood lactate level, respiratory alkalosis, or hypotension also should suggest the diagnosis. The systemic response can be quite variable, however. In one study, 36% of patients with severe sepsis had a normal temperature, 40% had a normal respiratory rate, 10% had a normal pulse rate, and 33% had normal white blood cell counts. Moreover, the systemic responses of uninfected patients with other conditions may be similar to those characteristic of sepsis. Examples include pancreatitis, burns, trauma, adrenal insufficiency, pulmonary embolism, dissecting or ruptured aortic aneurysm, myocardial infarction, occult hemorrhage, cardiac tamponade, postcardiopulmonary bypass syndrome, anaphylaxis, tumor-associated lactic acidosis, and drug overdose.

Definitive etiologic diagnosis requires identification of the causative microorganism from blood or a local site of infection. At least two blood samples should be obtained (from two different venipuncture sites) for culture; in a patient with an indwelling catheter, one sample should be collected from each lumen of the catheter and another via venipuncture. In many cases, blood cultures are negative; this result can reflect prior antibiotic administration, the presence of slow-growing or fastidious organisms, or the absence of microbial invasion of the bloodstream. In these cases, Gram’s staining and culture of material from the primary site of infection or from infected cutaneous lesions may help establish the microbial etiology. Identification of microbial DNA in peripheral blood or tissue samples by polymerase chain reaction may also be definitive. The skin and mucosae should be examined carefully and repeatedly for lesions that might yield diagnostic information. With overwhelming bacteremia (e.g., pneumococcal sepsis in splenectomized individuals; fulminant meningococcemia; or infection with V. vulnificus, B. pseudomallei, or Y. pestis), microorganisms are sometimes visible on buffy coat smears of peripheral blood.

TREATMENT

Approximately 20–35% of patients with severe sepsis and 40–60% of patients with septic shock die within 30 days. Others die within the ensuing 6 months. Late deaths often result from poorly controlled infection, immunosuppression, complications of intensive care, failure of multiple organs, or the patient’s underlying disease. Case–fatality rates are similar for culture-positive and culture-negative severe sepsis. Prognostic stratification systems such as APACHE II indicate that factoring in the patient’s age, underlying condition, and various physiologic variables can yield useful estimates of the risk of dying of severe sepsis. Age and prior health status are probably the most important risk factors (Fig. 19-1). In patients with no known preexisting morbidity, the case–fatality rate remains <10% until the fourth decade of life, after which it gradually increases to >35% in the very elderly. Death is significantly more likely in severely septic patients with preexisting illness. Septic shock is also a strong predictor of both short- and long-term mortality. Cognitive impairment may be significant in survivors, particularly those who are elderly.

Prevention offers the best opportunity to reduce morbidity and mortality from severe sepsis. In developed countries, most episodes of severe sepsis and septic shock are complications of nosocomial infections. These cases might be prevented by reducing the number of invasive procedures undertaken, by limiting the use (and duration of use) of indwelling vascular and bladder catheters, by reducing the incidence and duration of profound neutropenia (<500 neutrophils/μL), and by more aggressively treating localized nosocomial infections. Indiscriminate use of antimicrobial agents and glucocorticoids should be avoided, and optimal infection-control measures (Chap. 17) should be used. Studies indicate that 50–70% of patients who develop nosocomial severe sepsis or septic shock have experienced a less severe stage of the septic response on at least one previous day in the hospital. Research is needed to identify patients at increased risk and to develop adjunctive agents that can modulate the septic response before organ dysfunction or hypotension occurs.