Introduction

Sepsis: historical perspective

The term sepsis is derived from the Greek word sepsin, which means ‘to make putrid’. Early descriptions of disease mediated by “small invisible creatures” were made in the second century B.C., and the concepts of contagion and isolation of diseased individuals followed. Despite attempts at prevention pan-epidemic infections have caused the deaths of millions of persons throughout history. The first documented observations of living bacteria were made by van Leeuwenhoek in 1674 and classification of bacterial morphology in the early nineteenth century. However, the relationship between infectious disease, its aetiology, and its pathogenesis remained elusive.

The principles of disinfection and anti-septic practices pioneered by Semmelweis and later by Lister were adopted only several decades later. The importance of the host response to infection was first described in the 1880s and classified separately in terms of cell-mediated and humoral immunity. The subsequent use of drug anti-metabolites to ameliorate the effects of syphilis at the turn of the century, and the discovery of antibiotic sulphonamides, moulds and vaccination, led to a revolution in the treatment and prevention of infection. It was believed that these developments had the potential of eradicating sepsis from the modern age, but the problems of changing disease patterns and antibiotic resistance dampened early ambitions and sepsis has remained a formidable problem in many areas of medical practice.

Sepsis and multiple organ dysfunction: epidemiology

Sepsis, the host response to an infectious process, is termed severe when complicated by predefined organ system dysfunction [1]. Together, the systemic inflammatory response syndrome (SIRS), sepsis and septic shock have been termed the ‘sepsis syndromes’ [1, 2, 3]. The nature of infectious organisms associated with sepsis is changing. Thus, whilst Gram-negative bacteria were traditionally responsible for the majority of hospital-acquired infections, Gram-positive organisms (30–50% of cases) and multidrug-resistant bacteria or fungi (25%) are now more common [4, 5]. Moreover, the burden of sepsis-related disease is also rising; from 82.7 to 240.4 cases per 100,000 population in the United States and to 51 cases per 100,000 population (1997 figures) in the United Kingdom, where 27.1% of adult ICU admissions had severe sepsis in the first 24 h [6].

Severe sepsis and shock are characterised by tissue hypoperfusion, cellular hypoxia and metabolic dysfunction. Consequently the majority of patients with SIRS and its sequelae who fail to survive succumb to multiple organ dysfunction syndrome (MODS). Multiple organ failure (i.e. demonstrable failure of two or more organs) within the ICU was first documented in 1977. Bacterial sepsis was aetiologically significant in 69% of the cases described [7]. Indeed, the onset of MODS, synonymous with multiple organ system failure, was thought originally to follow a temporal sequence (lung, liver, gastric mucosa and kidney) [8]. Moreover, whilst strongly linked to uncontrolled infection (in particular intra-abdominal), it is now recognised that MODS can occur independently of sepsis. The commonest manifestation of MODS is acute lung injury, defined by refractory hypoxaemia attributable to high permeability pulmonary oedema [9]. Its extreme manifestation, the acute respiratory distress syndrome (ARDS), occurs in more than 40% of patients with sepsis and severe sepsis [6, 10].

There has been an evolution in the appreciation of mechanisms that result in sepsis and subsequent MODS. Thus, initially, a link between infections, which were recognised and treated and their inflammatory consequences was not appreciated. Indeed, progression to MODS, in spite of evidence of clearing of infection, nurtured the hypothesis of the body's response to infection associated systemic inflammation (by now autonomous from the initial infection) as being crucial to outcome. The process of increasing understanding of sepsis-associated MODS has required a number of key components, namely: (a) defining the biophysiological pathways arising from a systemic inflammatory insult, (b) clear epidemiological definitions of the spectrum of sepsis syndromes (often misused terms), (c) understanding the pathophysiological processes of the clinically apparent systemic disturbances during early and later stages and (d) testing different therapeutic approaches, directed at specific implicated inflammatory markers or at abnormal physiological parameters. Many therapeutic ‘bedside’ approaches have been proven wrong, yet providing insights into further ‘bench’ studies.

In summary, the sepsis syndromes and their sequelae, specifically MODS, represent the leading cause of death in adult general ICUs, with an associated mortality of 30–45%, consumption of 45% of ICU and 33% of hospital bed days and an estimated cost of $16.7 billion [6, 11].

Pathophysiology of MODS in sepsis

It is unknown why sepsis progresses to MODS in only certain individuals, or what the exact pathway is that leads to this. If the inflammatory process that characterises the systemic response to infectious pathogens becomes self-sustaining and progressive, organ dysfunction ensues. An extraordinarily complex and intricate cascade of inflammatory mediators, extra- and intracellular cell signalling pathways is activated. Prevailing wisdom suggests that these result in either microvascular dysregulation and/or mitochondrial dysfunction (so-called cytopathic hypoxia). These processes result in tissue hypoperfusion, and a further cascade of biochemico-physical alterations culminating in MODS [12].

Microvascular dysfunction

Early in the course of sepsis cardiac output (CO) rises to maintain blood pressure and organ perfusion in the face of reduced peripheral vascular resistance (hyperdynamic sepsis). As sepsis progresses, cardiac output is frequently reduced (so-called hypodynamic sepsis), which has a poor prognosis. Cardiac dysfunction per se is apparent in up to 44% of critically ill septic patients, with the aetiological agents suspected to be circulating depressant factors. Myocardial function tends to recover in survivors, and the prognostic significance of dysfunction in sepsis remains debatable [13]. Redistribution of capillary blood flow has been demonstrated in both animal models and in clinical sepsis [14, 15]. The use of investigatory tools such as intravital videomicroscopy, now applicable in the clinical setting, has provided evidence of simultaneous structural and functional abnormalities in sepsis, strengthening the association between tissue hypoperfusion and organ dysfunction. However, contradictory evidence from animal studies suggests that such hypoperfusion does not invariably lead to organ dysfunction and death.

Cytopathic hypoxia

Elevated tissue oxygen levels have been demonstrated in animals during experimental sepsis and in human skeletal muscle, suggesting that cellular inefficiency of oxygen utilisation rather than a failure of oxygen delivery (DO2) to tissues occurs in sepsis. By contrast, in cardiogenic shock tissue oxygen is reduced [16, 17]. Tissue oxygen consumption occurs normally principally through ATP production by oxidative phosphorylation in mitochondria. Reduced ATP concentrations in skeletal muscle during sepsis are associated with increasing severity of, and poor outcome from, septic shock [18]. The pathophysiological consequences of both regional flow alterations and mitochondrial dysfunction undoubtedly co-exist in the septic state, but do not appear to lead to significant histopathological correlates detectable at post-mortem examination.

Inflammatory cytokines in sepsis

The development of sequential organ failure in critically ill patients with sepsis is strongly predictive of mortality. However, the mechanisms involved in the dynamic interaction between different organ systems are dictated by the intricate interplay of haemodynamics, oxygen transport and metabolic disturbances. Genetic predisposition is almost certainly relevant in upregulating the expression of inflammatory mediators [e.g. tumour necrosis factor (TNF), interleukin (IL) 1, IL-8, triggering receptor on myeloid cells 1, high mobility group box 1), thereby influencing adversely the anti-/pro-inflammatory balance. Genetic predisposition seems more important for some infectious diseases such as meningococcaemia, but polymorphisms such as for TNF-α gene promoter can play a more general role in susceptibility to septic shock associated mortality [19]. Neuroendocrine systems and prothrombotic pathways (e.g. tissue factor) are activated with downregulation of fibrinolytic systems (i.e. anti-thrombin III, activated protein C and tissue factor pathway inhibitor) [20]. Inflammatory mediators TNF, IL-1, nitric oxide and reactive oxygen species are believed to disrupt communication pathways between organs which precedes organ failure [21]. Indeed, epithelial dysfunction has been proposed as a final common pathway for organ dysfunction in sepsis [22]. The tight junctions between these cells are affected in experimental models of sepsis. This may be particularly relevant in the gastrointestinal tract, which has been variously proposed as the ‘seat of sepsis’ and the ‘motor of multiple organ failure’ [23, 24]. Bacterial translocation (i.e. direct transcellular transport of microbes from the enterocytes to the submucosal layer) across a permeable intestinal luminal mucosa into the splanchnic circulation has been proposed as the initiator and propagator of sepsis following a remote insult. Mechanisms for this mucosal injury are multifactorial, including reduced intestinal blood flow and tissue hypoxia. Impaired hepatic clearance of toxins may also be relevant [25, 26, 27].

The prevailing theories of sepsis as an uncontrolled inflammatory response, which have been based on extensive animal studies, do not necessarily reflect the human clinical pattern. They used relatively large doses of bacteria or endotoxin and mortality was therefore the result of a ‘cytokine storm’, that if blocked improved survival. Meningococcaemia is perhaps the only human form of sepsis in which circulating levels of TNF-α are high and correlated with mortality [28]. Furthermore, there is much evidence of immune suppression during sepsis. Anergy (a state of non-responsiveness to antigen) through lymphocyte apoptosis has been demonstrable in vivo, and from autopsy studies of patients dying from sepsis [29]. Cellular hibernation or ‘stunning’ as occurs during myocardial ischaemia has been postulated as a mechanism for sepsis-associated MODS based on the notable findings of discordance between histological findings and the degree of organ dysfunction from patients who died of sepsis [30].

An emerging concept is the variable immune response during sepsis; from hyperimmune to hypoimmune, depending on factors that include virulence of the organism, size of the inoculum, pre-existing co-morbidity, genetic polymorphisms in candidate genes and the inflammatory insults during the course of sepsis. Therefore it is perhaps too simplistic to consider an overactive immune system as the reason for sepsis and associated MODS but rather a dynamic state where a severely compromised immune system might prevent adequate eradication of pathogens [29].

Clinical relevance of organ dysfunction: severity of illness scoring systems

Scoring systems as risk prediction tools rely on acute derangements in acute physiological parameters which are numerically assigned by degree and aggregated. Such generic (as distinct from disease-specific) scoring systems are best exemplified by the Acute Physiology and Chronic Health Evaluation (APACHE) system [31] which has led to the development of a number of other organ-based failure scores [32, 33, 34, 35].

Perhaps the most widely applied in current practice is the Sequential Organ Failure Assessment Score (SOFA, previously called the Sepsis-Related Organ Failure Assessment). Daily SOFA scores provide an important physiological tracking system for the dynamic course of critically ill patients with sepsis. Whilst not designed for mortality prediction, worse scores are strongly associated with mortality [36]; the mean and highest SOFA scores are predictors of poor prognosis, whilst a worsening of SOFA within the first 48 h predicts the likelihood of mortality 50% or higher [37]. However, whether organ-based scoring systems direct the timing, degree and duration of appropriate interventions to prevent MODS in sepsis is uncertain.

Detecting organ dysfunction in sepsis

Continuous monitoring of clinical and physiological variables, recognition of the significance of any changes in monitored parameters, and an appropriate response, are the cornerstones and defining characteristic of modern-day intensive care medicine. Electrocardiographic, peripheral temperature (as an indicator of shock or its response) [38], non-invasive oxygen saturations [39], arterial blood gas, end tidal CO2, metabolism (i.e. lactate), central venous, and cardiac output monitoring have become routine in practice. Specific organ system monitoring can guide management in certain circumstances such as intracranial pressure monitoring in traumatic head injury [40], whilst other more novel techniques such as gastric tonometry, and hepatic blood flow devices are under evaluation in the setting of sepsis [41].

Metabolic monitoring

Hyperlactataemia is multifactorial in origin. Nevertheless, there is a good relationship in sepsis between lactic acidosis, organ failure and poor outcome [42]. Indeed, blood lactate sampling is established and now recommended as an important parameter for monitoring in international guidelines on the management of severe sepsis [43].

Cardiac output monitoring

The history of the development of flow-directed, balloon-tipped, pulmonary artery catheters (PAC) saw them adopt a pivotal role in continuous bedside cardiopulmonary monitoring, and coincidently propagated the value of central venous catheters [44, 45, 46]. However, the SUPPORT [47] investigators identified an increased odds ratio for mortality and resource utilisation with the use of the PAC, even after adjustment for treatment selection bias. The ‘attributable’ morbidity associated with PAC use was thought more likely due to misinterpretation of the values thereby derived than to physical complications on insertion [48]. However, such work has led to the development of a number of other monitoring devices utilising arterial waveform analysis (i.e. pulse contour cardiac output, lithium dilution cardiac output), oesophageal Doppler and bioimpedance. Whilst all are relatively less invasive than the PAC, none provides the additional information about the pulmonary circulation. By contrast, the use of echocardiography is becoming more widespread in assessing cardiac function in sepsis [49, 50, 51, 52].

Mixed venous oxygen saturation

The value of reduced mixed venous oxygen tensions/saturations sampled from indwelling PACs as an accurate reflection of inadequate DO2 due to reduced CO in cardiorespiratory failure was first demonstrated in patients undergoing cardiac surgery in whom a close correlation between venous oxygen saturation, CO and outcome was demonstrated [53]. Central venous oxygen saturation is now regarded as a crucial physiological surrogate for identifying and directing the correction of ‘hidden’ oxygen debt [54, 55, 56].

Management of organ dysfunction in sepsis

The principles of management of severe sepsis and associated organ dysfunction have evolved concomitantly with an increasing evidence base. Some critical concepts and studies that have helped this development are discussed below.

Diagnosis, source control and anti-microbial therapy

Early diagnosis of infection, ‘source control’ and appropriate anti-microbial treatment have been reported as crucial to outcome in sepsis for many years [57]. By contrast, up to eight-fold higher mortality is observed in prospective cohort studies of antibiotic misuse [58, 59], while inadequate surgical source control predicts MODS and increases mortality [7, 60].

Resuscitation-fluid management

Prompt and adequate haemodynamic resuscitation in patients with severe sepsis is pivotal in preventing progression to MODS and death. International recommendations suggest achieving a central venous pressure of 8–12 mmHg (or 12–15 mmHg in mechanically ventilated patients [56]. Which type of fluid replacement (i.e. crystalloid vs. colloid or albumin) to administer is more contentious [61, 62, 63], although a recent position statement by the American Thoracic Society is helpful in this regards [64].

Haemodynamic goals in sepsis

Fluid resuscitation in septic shock is directed at achieving adequate tissue perfusion and oxygenation, thereby overcoming tissue oxygen ‘debt’ which relates in part to inadequate DO2. However, an early demonstration that dobutamine and adequate volume resuscitation improve DO2 (and oxygen consumption, VO2) as well as haemodynamic parameters post-operatively [65, 66, 67] was not reproduced in patients with sepsis-induced organ failures. Indeed, a strategy of goal directed supranormal oxygen delivery (cardiac index 4.5 l min−1 m−2, DO2 > 60 ml min−1 m−2, VO2 > 170 ml min−1 m−2) using dobutamine in volume resuscitated critically ill patients increased mortality (54%) compared to controls (34%) [68]. In fact, the dobutamine-‘driven’ patients did not increase their VO2 beyond those of adequately volume resuscitated controls. A second study with similar outcomes [69] helped to establish a number of facts. First, patients with sepsis and septic shock who can improve their haemodynamic indices through adequate fluid resuscitation are likely to do better than those who do not. Second, supranormal targets for DO2/VO2 are at best unnecessary, and at worst increase mortality. Third, a beneficial response to fluid resuscitation is more likely in the acute phase, before established critical illness. Thus patients with severe sepsis and septic shock resuscitated to standard haemodynamic goals, who additionally achieve central venous oxygen saturation of 70% or higher within the first 6 h by fluid resuscitation, red cell transfusion to a haematocrit of 30%, and/or dobutamine (up to 20 μg kg−1 min−1) display significantly lower 30- and 60-day mortality rates [56].

Ventilatory strategies

In those patients with sepsis who develop acute lung injury and require mechanical ventilatory support low tidal volumes (approx. 6 ml/kg) and inspiratory plateau pressures below 30 cmH2O should be used where possible. Such recommendations have emerged from animal studies [70, 71] and a retrospective analysis of patients with ARDS, which demonstrated that pressure-limited ventilation with so-called permissive hypercapnia reduced hospital mortality compared with APACHE II predictions (18.6% vs. 37.8%) [72]. It was, however, the pivotal ARDSnet study that demonstrated a 9% absolute mortality reduction (31% vs. 39.8% for controls) in patients with ARDS randomised to receive a tidal volume of 6 ml/kg with plateau pressure limited to less than 30 cmH2O [73]. By contrast, higher positive end expiratory pressures, prone positioning and the use of inhaled nitric oxide and surfactant have demonstrated only short-term improvements in oxygenation. The results of a large randomised controlled trial of steroid therapy in late stage ARDS based upon an encouraging single-centre study are awaited [74].

Management of renal dysfunction

The importance of maintaining regional perfusion in sepsis is increasingly recognised, not least the hepatosplanchnic circulation. Since the first experiences of arteriovenous haemofiltration in anuric intensive care patients with fluid overload resistant to diuretics in the 1970s [75], acute renal failure in the critically ill has been recognised to be of multifactorial aetiology. Hypotension, nephrotoxic drug insults, sepsis and preceding renal dysfunction may all be relevant [76]. Acute renal failure is an independent risk factor for mortality in the critically ill, which varies from 45% to 70% when associated with sepsis [77, 78]. Factors predicting a poor outcome are advanced age, altered previous health status, later onset of acute renal failure, sepsis, oliguria and severity of illness [79]. The use of low-dose dopamine has been shown to be ineffective in halting the progression to acute renal failure in the critically ill [80, 81]. Daily intermittent haemodialysis is better than alternate-day haemofiltration in critically ill patients who require renal replacement therapy, improving the time to resolution and survival at 14 days [82]. Continuous renal replacement therapy has equivalent outcomes to intermittent renal replacement therapy for acute renal failure in critical illness, although the former may offer easier management of fluid balance in the haemodynamically unstable septic patient. Whether higher doses (i.e. ultrafiltration rates 35–45 vs. 20 ml kg−1 h−1) of continuous renal replacement therapy confer a survival advantage in acute renal failure awaits corroboration [83].

Metabolic management

Impaired adrenoceptor responsiveness has long been recognised in endotoxic shock, partially reversible by corticosteroids [84, 85]. However, high doses of steroids (methylprednisolone 30 mg/kg or dexamethasone), administered on day 1 of septic shock failed to show an outcome benefit in two multicentre randomised controlled trial in the 1980s, with the abandonment of empirical steroid treatment, except for those with demonstrable adrenocortical insufficiency [86, 87, 88]. However, later work employing the prospective characterisation of the adrenal status of patients in septic shock, through the use of a 250 μg ACTH stimulation test, into so-called responders (proposed unimpaired adrenocortical axis) and non-responders (proposed relative adrenocortical insufficiency) proved more encouraging. Thus non-responders randomised to 50 mg hydrocortisone every 6 h plus 50 μg oral fludrocortisone for 7 days displayed a significantly better 28-day vasopressor-withdrawal effect and survival advantage than those receiving placebo [89]. Overall survival between the hydrocortisone and placebo groups was not statistically different [90]. An ongoing trial (EUROCORTICUS) aims to address previous findings and investigate the risk-benefit ratio of low-dose steroids in non-refractory septic shock.

Glycaemic control, whilst avoiding potentially deleterious episodes of hypoglycaemia, plays an important role in outcomes of sepsis-associated organ failures and mortality. Tight glucose control (4.4–6.1 mmol/l) compared with standard care confers significant survival advantage in post-operative cardiac surgery patients. Multiple-organ failure with a proven focus of sepsis was also decreased [91]. Recent studies further support tight but less stringent control of blood glucose in critically ill patients (8.0 mmol/l or less) but suggest that glucose control, rather than insulin dose per se, is more important in determining outcome [92].

Anti-thrombotic strategies

The inflammatory response in severe sepsis is integrally related to procoagulant activity and endothelial activation. Protein C is activated by complexing with thrombin and endothelial cell thrombomodulin. Activated protein C (APC) then modulates inflammation, coagulation and endothelial cell function. A deficiency of APC and lower levels of protein C activity in sepsis are correlated with higher mortality rates [93, 94]. The PROWESS trial of drotrecogin alfa (activated) (recombinant human APC, rhAPC) showed that patients with severe sepsis who were randomised to 96-h infusions of rhAPC (24 μg kg−1 h−1) within 24 h of inclusion had significantly lower 28-day all-cause mortality vs. placebo (24.7% vs. 30.8% respectively). The incidence of serious bleeding was higher in the rhAPC group (3.5% vs. 2.0%, p = 0.06) [95], and it seems that sicker patients (APACHE II>25) benefit most from this therapy [96]. The effect was not reproduced in a large scale trial of anti-thrombin III in severe sepsis (mortality 38.9%, anti-thrombin group vs. 38.7% for placebo group) in spite of favourable indications from preclinical and phase II trials [97], in this sense mirroring experience with many other putative therapeutic interventions (i.e. anti-endotoxin, anti-TNF and nitric oxide synthase inhibition) trialled in patients with sepsis over many years [98, 99, 100, 101, 102, 103, 104, 105, 106, 107]. This failure (PROWESS notwithstanding) of new pharmacological therapies and immunotherapies in patients with sepsis may in part reflect the complexity of mechanisms leading to organ dysfunction and the consequent heterogeneity of the patient population. Whether new definitions are needed that may identify critically ill patients more likely to respond to novel therapies remains unclear [106].

Other strategies

Blood transfusion requirements in the critically ill have evolved from reports of its beneficial use dating back to 1935 and the appreciation of its value in improving tissue DO2 in early resuscitation [108]. However, the Transfusion in Critical Care Trial demonstrated that a conservative strategy employing a hemoglobin threshold of 7.0 g/dl (to maintain hemoglobin between 7 and 9 g/dl) is not associated with higher mortality than with a liberal transfusion protocol (i.e. threshold 10 g/dl), previously accepted as standard practice. However, only 6% of patients enrolled had sepsis, and in patients with ischaemic cardiac disease a higher threshold was recommended [109]. The optimal haemoglobin levels of specific groups of critically ill patients are therefore as yet unstudied, and the value of recombinant erythropoietin remains unclear.

Stress-ulcer prophylaxis to prevent clinically important bleeding from the gastrointestinal tract in critically ill patients is well established, and the predisposing factors (i.e. coagulopathy, hypotension and mechanical ventilation) are frequently present in patients with sepsis [110]. However, relatively small percentages of patients develop clinically important bleeding from recent observational studies. Moreover, the pursuit of early enteral nutrition where possible, together with a trend to an increased incidence of ventilator associated pneumonia by H2 antagonists/proton pump inhibitors, means that identifying subgroups of patients who may benefit most from stress ulcer prophylaxis remains difficult.

Table 1 Diagnostic criteria for sepsis and associated organ dysfunction in adults. Adapted from [2]: infection (documented or suspected—a pathological process induced by a micro-organism) and some of the following variables
Full size table
Table 2 The Sequential Organ failure Assessment score (MAP mean arterial blood pressure, Nor norepinephrine, Dop dopamine, Dob dobutamine, Epi epinephrine; FIO2 fraction of inspired oxygen, GCS Glasgow Coma Scale score) (adapted from [31])
Full size table
Table 3 Management guidelines for ‘early’ (the initial few hours following suspected sepsis) and ‘late’ (the period beyond the first few hours of severe sepsis) severe sepsis and septic shock (ALI acute lung injury, ARDS acute respiratory distress syndrome, APACHE Acute Physiology and Chronic Health Evaluation, MODS multiple organ dysfunction syndrome) (adapted from [38])
Full size table

Conclusions

Multiple organ dysfunction complicating sepsis remains the commonest cause of mortality in the ICU. However, its mechanisms remain unknown, and the results of pathological autopsy studies show no correlation with degree of organ dysfunction or with specific causes of death. Nevertheless, these mechanisms continue to be unravelled, alongside emerging genetic predisposing targets. Moreover, the concept of a variable immune status, which can be tracked during sepsis and modulated, provides an increasing number of potential new therapeutic targets. A body of evidence accrued over decades reemphasises the fundamental importance of early recognition of physiological surrogates of tissue dysoxia in reducing associated organ dysfunction. Local and International clinical strategies, through a phased approach of the development of evidenced-based guidelines (incorporating proven strategies in sepsis), their implementation and evaluation, have undertaken the challenge of effecting improved survival in this patient population.