Comparison of outcomes from sepsis between patients with and without pre-existing left ventricular dysfunction: a case-control analysis

Severe sepsis and septic shock represent serious medical conditions with high morbidity and mortality rates. Sepsis is a leading cause of death [1, 2], and frequently causes end-organ damage [3], as well as cognitive and physical disability in survivors [4]. In the United States, over $14 billion were spent on hospitalizations for sepsis in 2008 alone [5]. The hospitalization rate for sepsis has more than doubled between 2000 and 2008, with a rate of 24 per 10,000 persons in 2008 [2]. Patients hospitalized with sepsis have a length of stay nearly twice that of other patients and are eight times more likely to die than other patients [2, 6]. Early aggressive treatment for sepsis increases the chances for survival [7]. Tenets of aggressive treatment include early antibiotic therapy [8] and aggressive resuscitative strategies to achieve monitored clinical goals [9].

Patients with sepsis are a heterogeneous group with various predisposing features and sources of infection. Patients who are critically ill from all causes have a higher mortality when they have cardiac dysfunction [10]. Sepsis may be associated with concurrent reversible left ventricular dysfunction (LVD) [11] that some authors have attributed to circulating factors [2, 12]. Patients with LVD that occurs during sepsis may have a worse prognosis than patients who do not have this condition [12, 13]. Cardiac disease has a high prevalence in hospitalized patients [14], which may complicate the management of sepsis. However, the outcomes for septic patients who have pre-existing LVD have not been carefully studied.

Early sepsis management requires aggressive fluid resuscitation to defined clinical endpoints, whereas it may be necessary to be more circumspect in the administration of fluids in later phases of sepsis [15]. The presence of LVD in critically ill septic patients complicates this strategy and leads to concerns about the development of pulmonary edema and respiratory failure from excessive fluid resuscitation. Alterations in the cardiac compliance in patients with LVD may diminish the utility of the central venous pressure as a preload assessment tool [16]. Both the mixed venous oxygen saturation (ScVO₂) and the serum lactate level may be affected by the presence of LVD as well as sepsis, potentially confounding their utility in patients with both disorders [17‐19]. More data are needed to guide clinicians in the management of these difficult patients.

Our institution has maintained an extensive database since 2005 for all hospitalized patients treated for sepsis. We selected a population of patients with both acute sepsis and prior evidence of LVD for this study. In order to determine if differences exist between septic patients with pre-existing LVD and those without, we selected case-matched control patients with normal pre-sepsis LV function for comparison purposes. We chose to analyze clinical parameters of both case and control patients in order to identify risk factors for in-hospital mortality. The identification of differences between septic patients with and without pre-existing LVD, and the ascertainment of sepsis mortality risk factors in each group, may lead to the refinement of management algorithms in the subpopulation of patients with pre-existing LVD.

Clinical data were collected in a quality improvement database at our institution for 1,717 hospitalized patient episodes with a diagnosis of sepsis between 1 January 2005 and 30 June 2010. Of these patients, 510 died during their hospital stay, while 1,207 survived until discharge, for an in-hospital all-cause mortality rate of 29.7%. We identified 197 study subjects who met inclusion criteria and had both sepsis and pre-existing LVD. We chose 197 control subjects with normal left ventricular function from the quality improvement database matched to cases according to age and gender. Comparisons between the mean values for various clinical variables for cases and controls are presented in Table 1. Comparisons of clinical outcomes for cases and controls are presented in Table 2. When controlled for ScVO₂, hematocrit, lactate levels and intubation status by multivariable logistic regression analysis using mortality as the dependent variable, the mortality risk between cases and controls was not different (P = 0.183, OR = 1.41). The other risk factors studied, ScVO₂ (P = 0.005, OR = 0.998), intubation (P = 0.000, OR = 2.56), hematocrit (P = 0.012, OR = 0.996) and initial lactate level (P = 0.037, OR = 1.078) were all significantly associated with mortality.

In order to study the effect of the administration of intravenous fluid volume on respiratory gas exchange, we examined the relationship between the total intravenous fluid volume administered during the first 24 hours of patient admission with the ratio of the partial pressure of oxygen to the fraction of inspired oxygen (PO₂/FiO₂) obtained at 24 hours by linear regression analysis in both cases and controls. Data for the volume of fluid administered were available for most patients (n = 190 for cases; n = 192 for controls), but the data availability for the PO₂/FiO₂ at 24 hours was more limited (n = 108 for cases; n = 115 for controls). We found no relationship between intravenous fluid volume and the PO₂/FiO₂ at 24 hours in either cases or controls (r² = 0.019 for cases, r² = 0.001 for controls; Figures 1 and 2).

We examined the population of patients with pre-existing LVD to determine if there were differences between those subjects with more severe LVD compared to those subjects with less severe dysfunction other than the reduction in LVEF. There were 81 patients with an LVEF of less than 35%, compared to 116 patients with an LVEF greater than or equal to 35% but less than 50%. Patients with an LVEF less than 35% received less intravenous fluid during their resuscitation than did patients with a higher LVEF (3.40+/-2.07 L compared with 4.42+/-3.09 L, P = 0.012). However, mortality rate (27% versus 35%, P >0.05), intubation rate (47% versus 43%, P >0.05), use of vasopressors (46% versus 44%, P >0.05), Acute Physiology and Chronic Health Evaluation II (APACHE II) score (20+/-7 versus 19+/-6, P = 0.285), ScVO2 (67.9+/-16.0% versus 66.0+/-17.1%, P = 0.472), and initial lactate levels (3.8+/-3.7 mmol/L versus 3.1+/-2.5 mmol/L, P = 0.121) were not statistically different between patients with an LVEF of less than 35% and those with an LVEF between 35% and 50%.

We studied both the case and the control populations to determine the risk factors for mortality in each group. We applied univariable analysis of 25 potential risk factors with mortality as the dependent variable to both the cases and control groups. Results are presented in Table 3. Significant correlations by Pearson’s correlation were demonstrated in both populations between the APACHE II score and the lactate level (P = 0.001). We chose to exclude APACHE II from the multivariable model. We also found a significant correlation between the hematocrit and respiratory rate in the case but not the control subjects (P = 0.003 for the cases, P = 0.210 for the controls). We chose to exclude respiratory rate from the model as well. The model used for the multivariable logistic regression analysis, with mortality as the dependent variable, and which was applied to the case and control population in turn, included the following covariates: intubation status, ScVO₂, initial hematocrit, initial platelet count, compliance with a sepsis bundle, lung infection, serum lactate level and the use of vasoactive agents to augment blood pressure. The results of this analysis are presented in Table 4. Risk factors for mortality were different between cases and controls. Significant risk factors for mortality among patients with pre-existing left ventricular dysfunction included the ScVO₂, compliance with a sepsis algorithm, intubation status and a source of infection other than the lung. Significant risk factors for mortality among patients with pre-existing normal left ventricular function included the initial platelet count and the initial lactate level. While the ScVO₂ was similar between survivors and non-survivors among control patients, the ScVO₂ was significantly lower among non-surviving patients with pre-existing LVD compared with surviving patients (Table 5).

To our knowledge, this is the first study to compare patients with sepsis and pre-existing LVD to patients with sepsis with no significant pre-existing cardiac abnormalities. We found that septic patients with pre-existing LVD differed only modestly from patients with preserved left ventricular function at our institution. Both populations had similar degrees of severity of illness, as assessed by the APACHE II score. Although patients with pre-existing LVD had significantly lower ScVO₂ values, hematocrit levels and initial lactate levels than did patients with normal pre-sepsis ventricular function, other demographic and laboratory determinants were similar. Important clinical outcomes, such as in-hospital mortality and intubation rates, were not significantly different between the two groups of patients, though there was a trend towards increased mortality in the patients with pre-existing LVD.

Fluid management in critically ill patients with sepsis is complex, in part because the clinical endpoints for fluid administration change during the course of the septic episode. Initially, there is a general consensus that fluid administration should be liberal and aggressive [5, 9]. Clinical management during the initial 24 hours of therapy emphasize the use of resuscitation goals that implicate sufficient tissue delivery of oxygen and include not only traditional hemodynamic measurements and urine output but also variables related to oxygen delivery and consumption, such as hemoglobin and mixed venous oxygen saturation. Subsequently, however, patient outcomes are improved in critically ill septic patients by observing conservative fluid management strategies [20]. The supporting rationale for the application of conservative fluid management strategies during the late stages of sepsis stem from observations of improved patient outcomes from the use of such strategies for the treatment of acute lung injury [21, 22]. Complex patients with both pre-existing LVD and sepsis might be perceived as being more susceptible than other septic patients to developing respiratory complications from aggressive initial fluid resuscitation because pulmonary edema is a component of the congestive heart failure syndrome. Although patients with an LVEF below 35% did not have a different severity of illness nor different outcomes than patients with LVD but with higher LVEF, they did receive statistically less intravenous fluid during this resuscitation. We speculate that this may reflect practice patterns directed by the assumption just noted. We observed that patients with both sepsis and pre-existing LVD were no more likely to be intubated and require mechanical ventilation than case-matched controls, although the need for intubation was a mortality risk factor in these patients. Additionally, we found no correlation between the amount of resuscitative fluid administered over the first 24 hours and oxygenation in that group of cases and controls for whom data were available. We suggest that patients with sepsis and pre-existing LVD may be no more likely than other patients to develop respiratory compromise from goal-directed volume resuscitation in early sepsis.

Studies involving animal models over 40 years ago demonstrated that serum lactate levels were elevated when tissue oxygenation was decreased by experimental manipulation of the determinants of oxygen delivery [23, 24]. Increased lactate levels have been demonstrated to be associated with poor outcomes in critically ill patients [25], patients with circulatory shock [26], and patients with severe sepsis [27]. Production of lactate appears to be increased both in patients with circulatory shock and sepsis, though the putative mechanisms may be different: in circulatory shock mechanisms implicating oxygen delivery play an important role, whereas in sepsis constitutive cellular mechanisms of lactate production are prominent [28]. While there are some data that suggest that lactate clearance may be preserved in circulatory shock [29], lactate clearance seems to be diminished in some patients with sepsis [30]. We observed that the mean initial lactate level for patients who had sepsis and pre-existing LVD was less than that of the control group in our study, though the mean lactate levels for both groups were elevated compared to normal values. Initial lactate levels were closely associated with mortality in our septic control patients, but not in our septic patients with pre-existing LVD. The reasons for these findings are not clear, but may be related to the complex interaction of multiple factors effecting lactate metabolism in patients with multi-system disease. We suggest that initial serum lactate levels may be less important a predictor of outcome in septic patients with pre-existing LVD than they are in other septic patients.

Kasnitz and co-workers studied the relationship between ScVO₂ and lactate levels in patients with cardiopulmonary disease, noting that ScVO₂ levels were decreased in such patients and associated with increased lactate levels and mortality [17]. The presumed mechanism for this phenomenon is that diminished cardiac output leads to decreased oxygen delivery to tissues, increased oxygen extraction and, therefore, diminished oxygen saturation in blood returning to the heart [31]. In a large cohort of patients undergoing coronary artery bypass graft surgery, a threshold value for ScVO₂ less than 60.1% was identified below which mortality was increased [32]. In contrast to the findings in patients with cardiogenic shock, early work suggested that ScVO₂ was normal or increased in patients with septic shock [33]. Mechanistically, it has been suggested that this phenomenon occurs because of decreased oxygen extraction and utilization. Krafft and co-workers provided evidence supporting this hypothesis in general, but suggested that short-term decrements in ScVO₂ might be associated with adverse clinical outcomes [34]. More recently, Rivers and associates demonstrated that a strategy that, in part, targets augmenting ScVO₂ in septic patients leads to improved clinical outcomes [9]. However, the control group in the latter study had a mean ScVO₂ of 49%, which is different from that expected from prior work. Concerns have been raised about the ability to generalize the findings of Rivers et al. given that most septic populations have ScVO₂ levels in the normal or supranormal range [35].

We found that patients with pre-existing LVD and sepsis had a lower mean initial ScVO₂ than a control group of septic patients, a finding which is supported by previous work [36]. Patients with pre-existing LVD in our study who died had a significantly lower ScVO₂ than survivors, and low ScVO₂ was an independent risk factor for mortality. These findings were not observed in the control cohort. The additional observation that failure to comply with a protocol targeted to improve oxygen delivery variables was an independent risk factor for mortality in cases, but not controls, suggests to us that patients with pre-existing LVD may have a different physiological response profile during sepsis than other septic patients. We postulate that sepsis is a heterogeneous disorder, and management strategies that target oxygen delivery variables may be particularly appropriate for patients with pre-existing LVD.

Our work suffers from a number of important limitations. This study is from a single center, which may limit the applicability of the results. We retrospectively reviewed a quality improvement database; so that there is a risk that unappreciated confounding variables may be present. We did not assess left ventricular function systematically during the sepsis event in any patients, leading to the possibility that controls or cases may have had unrecognized synchronous LVD. The number of patients included in our study was relatively small, limiting the strength of our conclusions. Larger, prospective studies will need to be performed to confirm our findings.

This single center retrospective study demonstrates that patients with sepsis and pre-existing LVD may have similar outcomes when compared to patients with sepsis without pre-existing LVD. No correlation was demonstrated between the amount of intravenous fluid administered over the first 24 hours and oxygenation at 24 hours in septic patients either with or without pre-existing LVD. Mortality risk factors for patients with pre-existing LVD and sepsis included a low ScVO₂, failure to comply with a sepsis bundle of therapies, intubation status, and a source of infection other than the lung, whereas mortality risk factors for septic patients with no previous evidence of cardiac disease included increased lactate levels and a low platelet count. Further study will be needed to determine if improvement of oxygen delivery during sepsis is more important in ameliorating outcome in patients with pre-existing LVD compared to those without.

DRO is a Senior Staff Physician in the Pulmonary and Critical Care Medicine Division at Henry Ford Hospital and is an Associate Professor of Medicine at Wayne State University School of Medicine. SS is a Fellow in Pulmonary and Critical Care Medicine at Henry Ford Hospital.