Historical perspectives

Our understanding of the cardiovascular manifestations of septic shock has evolved over the years, as new techniques to assess cardiovascular performance have become available. Before the introduction of the pulmonary artery catheter (PAC), two distinct cardiovascular clinical presentations of septic shock were described: a high cardiac output (CO) state, associated with warm, dry skin and a bounding pulse despite hypotension (warm shock); and a low CO state, associated with hypotension, cold, clammy skin and a thready pulse (cold shock) [5]. Clowes et al. [6], in a 1966 study, described these two clinical pictures as different stages of septic shock: patients were believed to initially experience a hyperdynamic phase (warm shock), and then to either recover or progress to hypodynamic shock (cold shock) and death. This view was reinforced by other clinical studies [5,7] that correlated survival with high cardiac index (CI). Only a few studies hinted at the relationship between volume status, the CI and outcome [8,9]. All these studies that supported the concept of terminal cold shock suffered from the fact that they used central venous pressure (CVP) as the best available estimate of left ventricular end-diastolic volume and adequacy of resuscitation. Evidence accumulated over the past 40 years shows that CVP, as a reflection of right ventricular preload, is a poor estimate of left ventricular preload in critically ill patients, and particularly in sepsis [10].

The introduction of the PAC (which could measure pulmonary artery wedge pressure as a more accurate estimate of left ventricular preload) has allowed for better definition of the cardiovascular dysfunction in septic shock and has improved volume resuscitation. Several studies have shown that adequately resuscitated septic shock patients consistently manifest a hyperdynamic circulatory state with high CO and low systemic vascular resistance (SVR) [11,12]. In contrast to previous belief, this hyperdynamic state usually persists until death in nonsurvivors (Fig. 1) [13,14]. Despite the strong evidence characterizing sepsis as a hyperdynamic state, studies that examined myocardial performance still showed left ventricular dysfunction (illustrated by decreased left ventricular stroke work index) in properly resuscitated septic patients [15]. The depression in the Frank–Starling curve demonstrated in these studies, however, could be explained by either a change in myocardial contractility or compliance.

The development of portable radionuclide cineangiography and its application to critically ill patients has further improved our understanding of cardiovascular dysfunction in septic shock, by allowing differentiation between impaired contractility and impaired compliance.

Left ventricular function

Parker et al. [16] showed, using radionuclide cineangiography, that survivors of septic shock demonstrated a decreased left ventricular ejection fraction (LVEF) and an acutely dilated left ventricle, as evidenced by an increased left ventricular end-diastolic volume index (LVEDVI) (Fig. 2). These parameters returned to normal over 7–10 days in survivors. Nonsurvivors maintained normal LVEF and LVEDVI throughout the course of their illness until death. All patients in this study [16] had normal or elevated CI and low SVR, as measured by the PAC.

In 1988, Ognibene et al. compared left ventricular performance curves (plotting left ventricular stroke work index versus LVEDVI) of septic and nonseptic critically ill patients (Fig. 3). They showed a flattening of the curve in septic shock patients, with significantly smaller left ventricular stroke work index increments in response to similar LVEDVI increments when compared with nonseptic critically ill controls [17]. Subsequent studies have confirmed the presence of significant left ventricular systolic dysfunction in septic patients [18,19].

Left ventricular diastolic function in septic shock is not as clearly defined. The dilatation of the left ventricle [16] and the lack of discordance between pulmonary artery wedge pressure (PAWP) and left ventricular end-diastolic volume [17] both argue against significant diastolic dysfunction in sepsis. More recent studies using echocardiography, however, have demonstrated slower left ventricular filling [20] and aberrant left ventricular relaxation [21,22] in septic patients, suggesting that impaired compliance may significantly contribute to myocardial depression in sepsis.

Right ventricular function

Low peripheral vascular resistance in sepsis leads to decreased left ventricular afterload. However, the right ventricular afterload is frequently elevated due to increased pulmonary vascular resistance from acute lung injury [23]. These different physiologic conditions mean that the right ventricle cannot be expected to behave like the left ventricle in septic patients. Multiple studies have therefore specifically examined right ventricular function in sepsis.

A number of studies have documented right ventricular systolic dysfunction in volume-resuscitated septic patients, as evidenced by decreased right ventricular ejection fraction (RVEF) and right ventricular dilation [24,25,26,27]. Kimchi et al. [24] and Parker et al. [26] also showed that right ventricular dysfunction occurred independently of pulmonary vascular resistance and pulmonary artery pressure, suggesting that increased right ventricular afterload could not be the dominant cause of right ventricular depression in septic shock. Parker et al. [26] further demonstrated a close temporal parallel between right ventricular and left ventricular dysfunction in sepsis. In their study, survivors experienced significant right ventricular dilation and decreased RVEF and right ventricular stroke work index, all of which returned to normal within 7–14 days (Fig. 4). Nonsurvivors had moderate right ventricular dilation and a marginally decreased RVEF, neither of which improved throughout their illness.

There is also evidence of right ventricular diastolic dysfunction in septic patients. Kimchi et al. [24] noticed a lack of correlation between right atrial pressure and right ventricular end-diastolic volume, suggesting altered right ventricular compliance. Schneider et al. [25] similarly identified a subgroup of patients who failed to exhibit an increased right ventricular end-diastolic volume index in response to volume loading, despite a rise in CVP. However, the relative contribution of systolic and diastolic dysfunction to right ventricular depression in sepsis remains largely unknown.

Cardiovascular prognostic factors in septic shock

Early studies of cardiovascular dysfunction in septic shock suggested that a low or decreasing CI invariably carried a poor prognosis [5,6,7,8]. As previously discussed, these studies relied on CVP measurements to assess volume status. It is now known that CVP is a poor reflection of left ventricular preload in critical illness and cannot accurately determine adequacy of resuscitation. Introduction of the PAC showed that adequately volume-resuscitated septic shock patients (as measured by PAWP) predictably exhibited a high CI and low SVR, including nonsurvivors [11,12]. CI is therefore not a reliable predictor of mortality in sepsis.

Recognition of the significant post-resuscitation peripheral vasodilation (low SVR) in septic shock led to the theory that peripheral vascular failure could be a major determinant of mortality in septic shock. Baumgartner et al. [28] noticed that patients with an extremely high CI (>7.0 l/min/m²) and low SVR had a uniformly poor outcome. Groeneveld et al. [29] retrospectively examined data from septic shock patients. They found that, for equivalent CI, nonsurvivors had a lower SVR than survivors. They concluded that peripheral vascular resistance was closely linked to outcome in septic shock.

Parker et al. [13] reviewed hemodynamic data from septic shock patients on presentation and at 24 hours to identify prognostic value. They found that, on presentation, only a heart rate <106 beats/min suggested a favorable outcome. At 24 hours, a heart rate <95 beats/min, a SVR index >1529 dynes s cm⁵/m², a decrease in heart rate >18 beats/min and a decrease in CI >0.5 l/min/m² all predicted survival. In a subsequent study [14], the same authors confirmed previous findings of decreased LVEF and increased LVEDVI in survivors of septic shock but not in non-survivors. However, although nonsurvivors as a whole did not exhibit left ventricular dilation, they could be divided into two groups: the first with decreasing LVEDVI and decreasing CI on serial measurements, and the other group with progressively increasing LVEDVI and maintained CI.

There are three hemodynamic patterns of death in septic shock [13]. Early deaths are due either to distributive shock (low SVR and refractory hypotension despite preserved CI) or to a cardiogenic form of septic shock (decreased CI). Late deaths are due to multisystem organ failure. Correlating this with Parker et al.'s [14] findings, it is possible that nonsurvivors who are unable to dilate their left ventricle (decreasing LVEDVI and CI) succumb to the cardiogenic form of septic shock. Those who have increasing LVEDVI and preserved CI die of the classic distributive shock.

The value of right ventricular performance parameters in predicting outcome is less clear. Multiple studies [24,25,26,27] have shown that, in sepsis, the right ventricle behaves similar to the left ventricle, exhibiting acute dilation and decreased RVEF. Persistence of right ventricular systolic dysfunction is associated with poor outcome [26,27]. Opinions diverge, however, on whether initial right ventricular dilation and decreased RVEF portend poor prognosis. Vincent et al. [27] suggested that survival correlates with higher initial RVEF, whereas Parker et al. [26] observed lower RVEF in survivors of septic shock than in nonsurvivors. The question requires further study.

Patients with sepsis and septic shock can show resistance to the vasopressor and inotropic effects of catecholamines. Nonsurvivors of septic shock have been shown to have an attenuated inotropic response to a dobutamine stress test compared with survivors [30]. Conversely, increased SVI, increased mixed venous oxygen saturation, ventricular dilation and drop in diastolic blood pressure in response to a dobutamine stress test all predict survival in septic patients [31].

Myocardial hypoperfusion

The possibility of myocardial dysfunction in sepsis was originally proposed and described in the 1960s. Its etiology, however, remained a mystery. For many years, the leading theory was that sepsis was associated with a globally decreased myocardial perfusion, leading to ischemic injury and myocardial depression. Two studies disproved that view. Cunnion et al. [32] performed serial measurements of coronary blood flow and metabolism using thermodilution coronary sinus catheters in septic patients (Fig. 5). They found normal or elevated coronary blood flow in septic patients compared with normal controls with comparable heart rates, and found no difference in blood flow between patients who developed myocardial dysfunction and those who did not. There was no net myocardial lactate production. Dhainaut et al. [33], using the same technique, confirmed these findings.

Furthermore, studies on animal models of sepsis [34] demonstrated that myocardial oxygen metabolism and high-energy phosphates were preserved in septic shock, neither of which is compatible with myocardial ischemia. However, Turner et al. [35] recently measured increased troponin I levels in patients with septic shock, demonstrating some degree of myocardial cell injury in the course of septic shock. It remains unclear from their study whether direct cardiac injury plays a role in sepsis-induced myocardial dysfunction or is the result of other factors, including a myocardial depressant substance (MDS) or exogenous catecholamine administration.

Circulatory depressant substances

Wiggers' landmark report [36] in 1947 postulating the presence of a circulating myocardial depressant factor in hemorrhagic shock provided the basis for the accepted current theory of myocardial dysfunction in septic shock. The presence of a myocardial depressant factor in sepsis was later confirmed experimentally by Lefer [37] in the late 60 s. Clinical studies performed at the same time associated death from septic shock with a hypodynamic circulatory state marked by a decreased CO [6,7,8]. Although these earlier studies depended on the measurement of CVP to assess preload, a factor now recognized as unreliable in the critically ill, Lefer's early contributions furthered research into MDS in sepsis [37].

The first study to actually show a link between septic shock-associated myocardial depression in humans with the myocyte depressant effects of a patient's own septic serum was carried out by Parrillo et al. in 1985 [38]. Serum from patients with septic shock generated a significant concentration-dependent depression of in vitro myocyte contractility (Fig. 6). The investigators were also able to demonstrate a strong correlation between the timing and degree of septic shock-associated decrease in LVEF in vivo and cardiac myocyte depression in vitro induced by exposure to serum from the same patients.

Subsequent work focused on identifying the presence of a MDS. In a study of 34 patients with septic shock [39], researchers noted high levels of myocardial depressant activity that correlated with higher peak serum lactate, with increased ventricular filling pressures and with increased LVEDVI. In addition, a trend towards higher mortality was present in patients with high levels of myocardial depressant activity compared with patients with lower or absent activity (36% versus 10%) [39]. These studies would argue in favor of a circulating substance rather than hypoperfusion as the causative factor in septic shock-associated myocardial depression.

Myocardial depression in sepsis: cytokines

Although the existence of MDS was demonstrated by the previous studies [36,37,38,39], the identity of the molecules remained in question. Potential circulating inflammatory mediators that could cause septic myocardial depression include the prostaglandin group, leukotrienes, platelet activating factor, histamine and endorphins. However, the substance was found to be heat labile, soluble in water, and its activity in filtration studies was present in the >10 kDa fraction [39]. Although full molecular characterization was not possible with the available data, the characteristics were most consistent with either a polypeptide or protein. The list of potential cytokine mediators of myocardial depression is exhaustive; however, TNF-α and IL-1β play a central role and deserve further consideration.

TNF-α shares a similar biochemical profile with MDS, making it a plausible mediator of the myocardial effects of sepsis and septic shock [38,40,41]. Experimentally, increased levels of TNF-α produce fever, lactic acidosis, disseminated intravascular coagulation, acute lung injury and death. The cardiovascular effects are similar to clinical septic shock; namely, hypotension, increased CO and low SVR [42,43]. Human volunteers given TNF-α infusions demonstrate similar responses [44,45].

A number of studies have shown that, when TNF-α is administered to human and animal myocardial tissue in vitro or ex vivo, the result is a concentration-dependent depression of contractility [46,47,48,49]. Furthermore, removal of TNF-α from the serum of patients with septic shock partially eliminates its myocardial depressant effect [47]. Although larger phase III clinical trials have shown no overall survival benefit when anti-TNF-α monoclonal antibody has been administered to patients with septic shock, left ventricular function did improve in this patient group [50].

IL-1β induces similar pathophysiologic responses to TNF-α. IL-1β is increased in both human and animal models of sepsis and septic shock [51]. When exposed to IL-1β, in vitro as well as ex vivo myocardial contractility is depressed [49,52,53]. Immunoabsorption of IL-1β partially neutralizes cardiac myocyte depressant activity of human septic serum [47]. Administration of IL-1β antagonist attenuates the hemodynamic and metabolic manifestations of septic shock [54,55]. As with TNF-α, overall mortality has not been improved in randomized trials with IL-1β [54,55]

A number of studies have postulated that cytokine synergy plays a key role in septic myocardial depression. When considering IL-1β and TNF-α independently, substantially higher concentrations of each cytokine are required to depress contractility in rat cardiac myocytes than when these two cytokines are combined [47]. These findings have also been validated in ex vivo studies of isolated human atrial trabeculae [56]. The combination of TNF-α and IL-1β may cause myocyte depression at concentrations 50–100 times lower than would be required if applied individually. Such concentrations are well within those found in blood during human septic shock. Available data supports a causative role for TNF-α and IL-1β acting synergistically in septic myocardial depression.

Cellular mechanisms of septic myocardial depression

Depression of myocardial contractility by TNF-α, IL-1β and septic serum in vitro occurs in two distinct time frames. The early phase of cardiac myocyte depression occurs within minutes of exposure either to TNF-α, to IL-1β, to TNF-α and IL-1β given together or to septic serum [47,48]. In vivo canine studies also demonstrate the ability of TNF-α to induce rapid myocardial depression [43,57]. Furthermore, there is a convincing relationship between the degree of in vitro early cardiac myocyte depression produced by human septic serum and the decrease in LVEF from those same patients during acute septic shock [38]. Other studies also document a delayed depressant effect of TNF-α, IL-1β and supernatants of activated macrophages on in vitro myocardial tissue [52,53,57,58]. This effect begins hours after exposure and may persist for days. This late phase of myocardial depression appears to occur by a distinctly different biochemical pathway than the early depression, and may involve de novo protein synthesis.

The generation of NO may be central to both early and late depression of myocytes in in vitro and in vivo models of septic shock. The role of NO in cardiac contractility is still under study. NO is produced by the conversion of L-arginine to L-citrulline by the nitric oxide synthase, which exists in inducible nitric oxide synthase (iNOS) and constitutive nitric oxide synthase (cNOS) forms. Studies support the role of NO generated by cNOS in the physiologic regulation of cardiac contractility [59,60,61]. In vitro studies of cardiac myocytes superperfused with either NO, nitroprusside (NO donor) or the NO donor SIN-1 demonstrate reductions in myocardial contractility [62]. In human studies, nitroprusside infusion into coronary arteries depresses intraventricular pressure while improving diastolic relaxation and distensibility [63].

In sepsis and septic shock, the pathophysiologic production of NO may contribute to cardiovascular dysfunction. The initial process is thought to occur from sequential NO and cGMP generation via cNOS activation rather than by de novo synthesis of iNOS. This is consistent with the short time frame to the onset of myocardial depression [48,56,64]. These studies cannot, however, account for the late-onset of cytokine-driven myocardial depression discussed earlier. Further studies have shown, in in vitro myocardial preparations, that the generation of iNOS, NO and cGMP may be responsible for late onset myocardial depression [48,53,65,66].

Studies thus suggest that NO-mediated depression of myocardial contractility appears in two separate time frames by two distinct pathways. Prior to Kinugawa et al.'s study [67], it might have been thought these mechanisms were mutually exclusive. Kinugawa et al. demonstrated, using an avian cardiac myocyte model, a biphasic response to IL-6 administration. IL-6 given in high concentration produced both early (<30 min) and late (24 hours) cardiac myocyte depression, associated with a decrease in intracellular calcium. Early depression could be blocked by pretreatment with a calcium chelator, implicating that the calcium/calmod-ulin-dependent cNOS is involved. However, late depression could not be blocked by the chelator, supporting the role for the calcium/calmodulin-independent iNOS.

The study of Kinugawa et al. clearly suggests that cytokines (or sepsis) can stimulate the heart to sequentially produce NO by both cNOS and iNOS. The data would suggest that TNF-α, IL-1β and possibly other cytokines involved in septic shock could mediate their cardiodepressant effects by at least two distinct mechanisms. Early cardiodepressant activity may involve both a NO-dependent but β-adrenoreceptor-independent mechanism and a NO-independent defect of β-adrenoreceptor signal transduction [68]. Late depression involves an iNOS-dependent defect of β1-adrenergic signal transduction with further addition of an undefined defect of myocardial contractility. While NO generation may contribute to the cardiodepressant effects seen in sepsis, its precise role and the role of other mechanisms underlying septic myocardial dysfunction continue to be defined.