ASSOCIATION OF LEFT VENTRICULAR DIASTOLIC DYSFUNCTION WITH ELEVATED NT-pro-BNP IN GENERAL INTENSIVE CARE UNIT PATIENTS WITH PRESERVED EJECTION FRACTION

A COMPLEMENTARY ROLE OF TISSUE DOPPLER IMAGING PARAMETERS AND NT-pro-BNP LEVELS FOR ADVERSE OUTCOME

Abstract

The mechanisms of the N-terminal-pro-brain natriuretic peptide (NT-pro-BNP) release in intensive care unit (ICU) patients with preserved ejection fraction (EF) are unclear. We investigated whether left ventricular (LV) dysfunction, as assessed by tissue Doppler imaging (TDI), is related to NT-pro-BNP levels in ICU patients with preserved EF and has a complementary value to NT-pro-BNP in the determination of in-hospital mortality. We examined 58 mechanically ventilated patients with no history of heart failure (age, 60 ± 18 years; EF, 63% ± 7%). The systolic (S) and early diastolic (E′) velocity of the mitral annulus by TDI and the E/E′ as well as NT-pro-BNP, troponin, lactate acid, blood oxygen (PO₂/FiO₂), sepsis, and ICU mortality were assessed. Systolic, E′, and E/E′ correlated with age, PO₂/FiO₂, lactate acid, NT-pro-BNP, troponin, history of arterial hypertension, and diabetes (P < 0.05). By multivariate analysis, the determinants of NT-pro-BNP were S (P = 0.024), E/E′ (P = 0.017), and sepsis (P = 0.015). An NT-pro-BNP greater than 941 pg/mL was a reliable predictor of LV diastolic dysfunction defined as a composite of E′ less than or equal to 8 cm/s and/or mean E/E greater than or equal to 13 (area under the curve, 75%; P = 0.03). Patients with combined NT-pro-BNP greater than 941 pg/mL and abnormal TDI markers had increased creatinine levels and a lower MAP, PO₂/FiO₂, and survival rate than those with abnormal TDI or NT-pro-BNP alone or patients with normal TDI markers and NT-pro-BNP (25%, 60%, 70%, and 84%, respectively; P < 0.05). The addition of abnormal TDI in a model including NT-pro-BNP and sepsis increased the model's value for in-hospital mortality (P for change = 0.01). In ICU patients with preserved EF, LV diastolic dysfunction and sepsis determine NT-pro-BNP levels. Tissue Doppler imaging markers and NT-pro-BNP have a complementary value for in-hospital mortality.

INTRODUCTION

Several studies have investigated the use of various clinical and biochemical markers to assess outcome in critically ill patients (^{1, 2}).

Recently, the use of natriuretic peptides as prognostic indicators in critical illness has aroused great interest (^1-8). Brain natriuretic peptide (BNP) is a polypeptide neurohormone, which is mainly produced and secreted by cardiomyocytes. The main stimuli for its synthesis and release is myocyte stretch because of elevated cardiac pressure and/or volume (²). Thus, increased myocardial wall stress because of hypertension, diabetes and myocardial ischemia, structural myocardial disease such as left ventricular (LV) hypertrophy, and excessive intravascular volume causing myocardial stretch and cardiac dilatation may increase BNP (^{2, 9}). However, inflammation (¹⁰), neurohumoral activation, angiotensine II, stress hormones (^{2, 11}), and hypoxia (¹²), as well as treatment with inotropic agents and fluids (¹¹), are also considered as triggering factors for BNP production in critically ill patients. The pharmacologic effect of BNP at the cellular level is mediated by increases in cyclic guanosine monophosphate, which lead to relaxation of vascular smooth muscle (²). Brain natriuretic peptide induces vasodilation; inhibits renin, aldosterone, and angiotensin production; increases diuresis; and is thus an important regulator of fluid homeostasis (²). Studies have postulated that the vasodilatory effects of BNP may cause hypotension and impairment of renal function in septic patients (¹³) or patients with systolic heart failure (^14-16). Brain natriuretic peptide is secreted into the blood as a prohormone, where it is cleaved into active BNP and inactive metabolite N-terminal-pro-BNP (NT-pro-BNP) (²). Brain natriuretic peptide and NT-pro-BNP are produced in equimolar amounts but are removed from the circulation by different mechanisms at different time rates, making the plasma concentration unequal (²). N-Terminal- pro-BNP is mainly excreted by the kidneys, has a longer half-life time, and a better in vitro stability than BNP, which is cleared by specific clearance cell receptors and enzyme neutral endopeptidase (²). Brain natriuretic peptide and NT-pro-BNP have been widely used as excellent markers in the diagnosis of LV and prognosis of heart failure patients (^5-7). In critically ill patients, NT-pro-BNP production may be triggered by 1) LV dysfunction either preexisting, because of cardiac disease (⁹), or acquired, because of hypoxia, hypoperfusion, and toxic effects of inflammation (^{11, 12}); 2) renal dysfunction (³); 3) inflammatory mediators released during infection and/or sepsis (^{10, 11}); 4) elevation of stress hormones and neurohumoral activation (^{2, 11}) and thus may be a predictor of adverse outcome.

Traditionally, natriuretic peptides have been used in the diagnosis and prognosis of systolic and/or diastolic heart failure (⁹). Elevated levels of these biomarkers have also been recognized in noncardiac diseases such as pulmonary embolism, sepsis, chronic obstructive pulmonary disease, and renal insufficiency (¹⁷). The previously discussed noncardiac causes of elevated NT-pro-BNP are often present in critically ill patients admitted in a general intensive care unit (ICU). Tissue Doppler imaging (TDI) indices are able to detect LV diastolic and/or systolic dysfunction (^18-23) and have been associated with elevated natriuretic peptide levels during exercise (²⁴) in patients with suspected diastolic heart failure despite the presence of a normal ejection fraction (EF). Furthermore, studies have shown that TDI markers of LV function bear a prognostic impact on patients with cardiovascular disease (²⁵).

N-Terminal-pro-BNP levels are related with LV diastolic dysfunction in patients with cardiovascular risk factors such as hypertension and diabetes or overt cardiac disease (^{9, 10}). However, the association of LV diastolic dysfunction and NT-pro-BNP levels in critically ill patients with preserved EF, no history of chronic heart failure, and a high incidence of noncardiac causes of elevated NT-pro-BNP during hospitalization in a general ICU has not been clearly defined.

In agreement with other investigators, we have previously shown (³) that nonsurvivors have higher aminoterminal part of BNP (NT-pro-BNP) levels on admission in ICU than survivors, and that NT-pro-BNP is an independent predictor of ICU mortality (^3-8). Furthermore, we have previously shown that the use of inotropes, renal impairment, sepsis, and age accounted for nearly 50% of the NT-pro-BNP variation in ICU patients. However, in our previous study, we did not investigate whether LV diastolic dysfunction, as assessed by tissue Doppler echocardiography, is an additional determinant of elevated NT-pro-BNP levels and whether TDI markers of LV diastolic dysfunction have an independent and incremental value to NT-pro-BNP levels in the assessment of the in-hospital mortality. Thus, in the present study, we hypothesized that LV diastolic dysfunction is an additional determinant of increased NT-pro-BNP levels in critically ill patients with preserved ejection function and no history of heart failure admitted in a general ICU, and thus, NT-pro-BNP may be a reliable marker of LV diastolic dysfunction in these patients. We also hypothesized that elevated NT-pro-BNP and abnormal indices of LV function as assessed by TDI may have a complementary value in the determination of in-hospital mortality.

The aim of our study was 1) to investigate whether NT-pro-BNP levels are independently related to LV diastolic dysfunction as assessed by TDI in critically ill patients with preserved ejection function and no history of heart failure admitted in a general ICU and 2) to examine whether TDI markers of LV diastolic function and NT-pro-BNP levels have complementary value in the determination of the in-hospital mortality in these patients. We have chosen to study patients with preserved ejection as an association between systolic cardiac dysfunction, BNP levels, and in-hospital mortality has been demonstrated in critically ill patients (²⁶).

PATIENTS AND METHODS

Patients

This prospective single-center study included serial recruitment of critically ill patients, aged older than 18 years, requiring support with mechanical ventilation, admitted in our general ICU during a 12-month period. The study was approved by the hospital's ethics committee, and informed consent was obtained from patients' relatives. Exclusion criteria included chronic heart failure, defined by known history of hospitalization due to heart failure decompensation and/or an LV ejection fraction (LVEF) of less than 45% on admission echocardiography, atrial fibrillation, and preexistent renal insufficiency (history of serum creatinine >1.8 mg/dL before ICU entry). Patients' clinical information included age, sex, reason for admission, disease severity according to the Acute Physiology and Chronic Health Evaluation II score, degree of organ dysfunction quantified by the Sequential Organ Failure Assessment (SOFA) score, the presence of sepsis by using widely accepted guidelines (²⁷), requirement for inotropic agents (norepinephrine, dobutamine, or dopamine), MAP, and central venous pressure (CVP). Blood gas analysis, including partial pressure of oxygen (PO₂), carbon dioxide (PcO₂), and lactate acid, was also available. The ratio of partial pressure of oxygen in blood to the oxygen concentration given during mechanical ventilation (PO₂/FiO₂) was used as a marker of oxygenation. In-hospital outcome was defined as mortality during hospitalization.

Echocardiography

Transthoracic echocardiography was performed on admission using a Vivid 4 (GE Medical Systems, Horten, Norway) phased array system. Studies were digitally stored and analyzed by two observers (I.I., M.N.) blinded to clinical and laboratory data using a computerized station (Echopac GE). All patients had adequate images for analysis. For the analysis of segmental wall motion abnormalities, a 16-segment protocol was used (²⁸). Each segment was scored as follows: 1 indicates normal; 2, hypokinetic; 3, akinetic; and 4, dyskinetic. A total wall motion score was calculated as the sum of all 16 segments. The wall motion score index was defined as the ratio of the number of segments with wall motion abnormalities divided by the total of all 16 segments (normal value, 1). Patients with wall motion score index greater than or equal to 1.125 (corresponding to wall motion abnormalities in at least two segments) were considered to present segmental wall motion abnormalities (²⁸). However, none of our patients showed evidence of segmental wall motion abnormalities. The following parameters were measured from cross-sectional echocardiographic images of the LV: 1) end-diastolic (LVEDD) and end-systolic diameter (in millimeters); 2) interventricular septal (IVS), posterior wall (PW) thickness (in millimeters); 3) relative wall thickness (RWT) as the sum of the IVS and PW thickness divided by the LVEDD; 4) fractional shortening and EF (percentage); and 5) left atrium dimension (in millimeters). Transmitral pulsed wave Doppler velocities were recorded in the four-chamber view with the sample volume at the tip of mitral valve leaflets. E and A wave velocities and deceleration time of early transmitral flow velocity were measured. Myocardial velocities were recorded using color TDI to record low-velocity, high-intensity myocardial signals at a high frame rate (120 MHz). A 5-mm sample volume was placed in septal and lateral corner of the mitral annulus in the apical four-chamber views to record the systolic velocity (S′), early diastolic velocity (E′), and late diastolic velocity (A′). The mean value of the S′, E′, and A′ in the septal and lateral corner was used for analysis. The ratio of E wave of the mitral inflow measured by pulsed wave Doppler to the E′ was calculated as an index of LV diastolic filling pressures. All Doppler markers were measured at the end-expiration (²⁹). Interobserver and intraobserver variability of these measurements was 3% and 1.7%, respectively. Patients were stratified into those with or without a composite of abnormal TDI indices of LV diastolic dysfunction. The composite was defined as mean E′ less than or equal to 8 cm/s and/or septal E′ younger than 40 years less than 9 cm/s, septal E′ 40 to 60 years less than 7 cm/s, E′ older than 60 years less than 6 cm/s, E/E′ greater than 12 for the lateral corner, E/E′ greater than 15 for the septal corner of the mitral annulus, mean E/E′ greater than or equal to 13 using previously published cutoff values (^{9, 18-23}) Furthermore, the patients were also stratified into those with an S less than or equal to 8 cm/s and those with higher S values, suggesting a normal systolic longitudinal function of the LV (¹⁹).

Laboratory measurements

Blood samples were drawn on admission to determine NT-pro-BNP. The blood was centrifuged and stored at −70°C until assayed. N-Terminal-pro-BNP was measured by the same study-assigned laboratory using commercially available kits (NT-pro-BNP; Elecsys 2010, Roche Diagnostics, Mannheim, Germany). The interassay coefficient of variance is less than 3.0%. The analytic range is from 20 to 35,000 pg/mL. Patients' hematological and biochemical profile included hemoglobin, white blood cells, platelets, INR creatinine, electrolytes, albumin, bilirubin, and troponin T.

Statistical analysis

All data were tested for normal distribution by the Kolmogorov-Smirnov test. Results are presented as medians and ranges for parameters with non-normal distribution or mean values and SDs. Group comparisons were performed by Mann-Whitney and chi-square tests. Pearson and Spearman correlation coefficient assessed the associations between variables. The logarithm of the variables with a skewed distribution underwent logarithmic transformation for analysis. By receiver operator characteristic analysis, we assessed the predictive value of NT-pro-BNP for the presence of LV diastolic dysfunction. The area under the receiver operator characteristic curve and corresponding confidence intervals (CIs), as well as the corresponding sensitivity and specificity, were calculated. Patients' cohort was also stratified as patients with a composite of normal versus abnormal diastolic TDI indices (as previously defined), and the corresponding survival curves were compared by the log-rank test. The combination of elevated NT-pro-BNP and abnormal diastolic TDI markers for ICU mortality was also examined by survival curves. Univariate Cox regression analysis was used to calculate the relative risk and CIs of the examined variables for in-hospital mortality. Multivariate Cox regression analysis was used to assess the incremental value of the composite of abnormal TDI indices over other prognostic factors. The change of -2LogLikelihood after addition of the composite of abnormal TDI in the multivariable model was used to assess statistical significance. A P < 0.05 was considered statistically significant in all analyses.

RESULTS

Of 75 patients fulfilling the inclusion criteria, 5 had inadequate echo image quality due to severe chronic obstructive pulmonary disease, 5 died on ICU admission and thus did not have an echo study, and 7 had extensive chest trauma excluding the feasibility of a transthoracic echo study. Thus, 58 patients (39 men) with a mean age of 60 ± 18 years (range, 18-85 years) under mechanical ventilation were finally included in the study. Admitting diagnosis was related to medical (n = 35), surgical (n = 20), or multiple trauma (n = 3) critical states. Additionally, 39 of 58 patients were diagnosed to have sepsis, and 5 more had evidence of infection. Of 58 patients, 22 died during hospitalization, yielding an in-hospital mortality rate of 38%. All patients had preserved LVEF, with a mean value of 63% ± 7%. Echocardiographic findings are shown in Table 1.

Table 1: Patients' baseline echocardiographic evaluation

TDI parameters

S, E′, and E/E′ correlated with age, NT-pro-BNP, myocardial damage as assessed by troponin levels, PO₂/FiO₂, lactate acid, history of hypertension, and RWT (P < 0.05; for all correlations). Additionally, E′ and E/E′ were related with history of diabetes (P < 0.05) and E′ with presence of sepsis (P = 0.04). As assessed by TDI markers, 27 patients (46%) had a composite of abnormal TDI indices of LV diastolic function. Of those, 18 patients also had an S less than 8 cm/s. Patients with a composite of abnormal indices of LV diastolic function had higher NT-pro-BNP, troponin, lactate acid levels, incidence of history of hypertension and diabetes, RWT than patients with normal TDI (P < 0.05; Table 2).

Table 2: Differences between patients with or without LV diastolic dysfunction as assessed by TDI

Interestingly, patients with abnormal TDI markers a lower PO₂/FiO₂ than patients with normal TDI, suggesting that elevated LV filling pressures may have caused pulmonary congestion contributing to hypoxia (P < 0.05; Table 2). MAP and CVP were similar between patients with abnormal and normal TDI markers (P = not significant).

N-Terminal-pro-brain natriuretic neptide

In addition to TDI markers, NT-pro-BNP was also associated with age, myocardial damage as assessed by troponin levels, RWT, presence of sepsis, inotrope infusion, SOFA score, and tissue oxygenation as assessed by PO₂/FiO₂ (P < 0.05 for all correlations) and renal function as assessed by creatinine levels (regression coefficient b, 0.44; P = 0.02). In a multivariate linear regression model, including TDI indices, and univariate determinants of NT-pro-BNP levels with P less than 0.05 (troponin, age, creatinine, RWT, presence of sepsis, inotrope infusion, SOFA score, PO₂/FiO₂) the most important determinants of NT-pro-BNP levels were S (regression coefficient b, -0.33; P = 0.024,), E/E′ ratio (regression coefficient b, 0.36; P = 0.017), and presence of sepsis (regression coefficient b, 0.35; P = 0.012). By receiver operating analysis, an NT-pro-BNP greater than 941 pg/mL predicted a composite of abnormal TDI markers of LV diastolic function with 73% sensitivity and 70% specificity (area under the curve,75%; 95% CI, 68-94; P = 0.03).

Interestingly, patients with an NT-pro-BNP greater than 941 pg/mL had lower MAP and similar CVP than patients with lower levels (75 ± 11 vs. 84 ± 14 mmHg [P = 0.03] and 10.8 ± 3.5 vs. 10.9 ± 4.0 mmHg [P = 0.83], respectively), suggesting the presence of a lower BP in the presence of similar preload or at least similar right atrial "stretch" in patients with increased NT-pro-BNP levels compared with those with lower levels. Additionally, patients with an NT-pro-BNP greater than 941 pg/mL had higher creatinine levels than those with lower levels (1.75 ± 1.1 vs. 1.04 ± 0.5 mg/L; P = 0.03).

Outcome

Of 58 patients, 22 died during hospitalization (59.0 ± 35 days), yielding a mortality rate of 38%. Survivors had lower Acute Physiology and Chronic Health Evaluation II score, troponin, and NT-pro-BNP levels, as well as higher PO₂/FiO₂, incidence of sepsis, and higher TDI velocities (P < 0.05 for all comparisons). By Kaplan-Meier curve analysis, patients with a composite of abnormal TDI indices for LV diastolic function had an increased mortality (survival rate, 37%) compared with patients with normal TDI markers (survival rate, 83%; log-rank, 11,1; P = 0.001; Fig. 1, A). The patients with a combination of abnormal TDI markers of LV diastolic function and NT-pro-BNP greater than or equal to 941 pg/mL had higher hospital mortality (survival rate, 25%) compared with patients with only abnormal TDI (survival rate, 60%), only BNP greater than 941pg/mL (survival rate, 70%), or a combination of normal TDI markers and NT-pro-BNP less than 941 pg/mL (survival rate, 84%; log-rank, 11,9; P = 0.007; Fig. 1, B). By Cox regression analysis, presence of sepsis, NT-pro-BNP levels, and the composite of abnormal TDI indices of diastolic LV function were found significant univariate predictors of in-hospital mortality among all the previously examined variables. The corresponding relative risk of mortality was 4.6 (95% CI, 1.6-12; P = 0.003) for the composite of abnormal TDI markers, 3.1 (95% CI, 1.01-10.7; P = 0.035) for sepsis, and 2.9 (95% CI, 1.07-7.9; P = 0.03) for NT-pro-BNP, respectively.

Fig. 1: A, Kaplan-Meier curves for the prediction of in-hospital mortality in patients with a composite of abnormal TDI indices of LV function versus those with normal TDI indices. B, Kaplan-Meier plots for the prediction of in-hospital mortality in patients with 1) normal TDI and NT-pro-BNP less than 941 pg/mL, 2) normal TDI and NT-pro-BNP greater than 941 pg/mL, 3) abnormal TDI and NT-pro-BNP less than 941 pg/mL, and 4) abnormal TDI and NT-pro-BNP greater than 941 pg/mL.

By multivariable Cox regression analysis, the addition of the composite of abnormal TDI markers of LV diastolic function, in a model including NT-pro-BNP and sepsis, significantly increased the prognostic value of the model for ICU mortality (−2LogLikelihood from −143.8 to −133.6; P for change = 0.01, respectively).

Interestingly, patients with a combination of abnormal TDI markers of LV diastolic function and NT-pro-BNP greater than or equal to 941 pg/mL had the lowest MAP compared with patients with only abnormal TDI, only BNP greater than 941 pg/mL, or a combination of normal TDI markers and NT-pro-BNP less than 941 pg/mL (72 ± 8 vs. 79 ± 7 vs. 75 ± 10 vs. 86 ± 15 mmHg; F = 3.8; P = 0.014) in the presence of similar CVP (11.0 ± 3.8 vs. 12.7 ± 3.2 vs. 11.7 ± 4.3 vs. 10.1 ± 3.5 mmHg; F = 1.07; P = 0.37) as well as the lowest PO₂/FiO₂ (127 ± 98 vs. 198 ± 86 vs. 248 ± 100 vs. 298 ± 122 mmHg/%; F = 3.8; P = 0.014) and elevated creatinine levels (1.7 ± 0.5 vs. 1.1 ± 0.6 vs. 1.9 ± 1.2 vs. 1.1 ± 0.7. mg/L; F = 3.9; P = 0.01). Thus, patients with a combination of abnormal TDI markers and NT-pro-BNP greater than or equal to 941 pg/mL had the lowest PO₂/FiO₂ and MAP compared with the remaining patients, suggesting a greater degree of reduced blood oxygenation and tissue hypoperfusion.

DISCUSSION

In the present study, we have shown that approximately half of critically ill patients with "normal" EF and no history of heart failure had evidence of LV diastolic dysfunction, as assessed by TDI, and one third of them showed evidence of impaired longitudinal systolic function. Moreover, this is the first study that demonstrates the independent association of elevated NT-pro-BNP levels with LV diastolic dysfunction in mechanically ventilated, critically ill patients with preserved EF. Furthermore, we have shown that TDI markers of LV function and NT-pro-BNP levels are complementary predictors of in-hospital mortality in patients with preserved EF on admission to a general ICU.

LV diastolic dysfunction in ICU patients

Tissue Doppler imaging of the systolic mitral annulus motion permits an early detection of the longitudinal systolic dysfunction even if EF remains "normal" (^{20, 22, 23}). Furthermore, the advantage of the E′ over the mitral inflow Doppler markers of diastolic function is that it behaves as a relatively preload independent index of LV relaxation (^{18, 22}). E/E′ is a widely used marker to assess simply and accurately LV filling pressures (^{9, 23}). Recently, transthoracic estimation of the E/E′ ratio at the end-expiration has been shown to correlate closely with pulmonary artery occlusion pressure in mechanically ventilated ICU patients (²⁹). The reduced TDI velocities of the mitral annulus (S, E′, E/E′) have been shown to predict mortality or cardiovascular events in patients with heart failure, myocardial infarction, or hypertension (²⁵).

In the present study, we have shown that approximately half of mechanically ventilated, critically ill patients with "normal" EF and no history of heart failure have evidence of LV diastolic dysfunction, as assessed by TDI. Additionally, 31% showed evidence of impaired longitudinal systolic function as assessed by an S′less than 8 cm/s in addition to LV diastolic dysfunction despite the presence of a preserved EF. A reduced early systolic velocity (S) of the mitral annulus by TDI has been described in patients with preserved EF and LV diastolic dysfunction such as patients with hypertension, coronary artery disease, diabetes, or cardiomyopathies (²⁰). This finding is thought to reflect an impairment of the LV longitudinal systolic performance caused by dysfunction of subendocardial myocardial layers on the ground of hypertrophy, fibrosis, elevated filling pressures, and subendocardial ischemia (²⁰). In our study, there was a close association between TDI indices of LV diastolic function and age, history of hypertension or diabetes, and relative wall thickness. It is known that aging, arterial hypertension, diabetes mellitus, and increased wall thickness may impair the diastolic properties in the left ventricle (^{20-22, 30}). Additionally, in our study, impaired TDI markers were also related with presence of sepsis and increased levels of lactate acid. This finding suggests that acidosis and/or inflammation may play an additional role in myocardial depression in ICU patients causing LV diastolic dysfunction. Studies have shown that lactate acidosis (³¹) and inflammation (^{32, 33}) are associated with cardiac systolic or diastolic dysfunction. Therefore, we may hypothesize that each one of the previously discussed factors or their combination may worsen the myocardial diastolic properties and longitudinal systolic function, as assessed by TDI in critically ill patients with preserved EF. In support of this hypothesis, TDI markers of LV function were related with troponin levels, confirming that LV diastolic dysfunction is related with a new-onset myocardial damage in critically ill patients with preserved EF on admission to ICU.

In conclusion, LV diastolic dysfunction and impaired longitudinal systolic function in our ICU patients with preserved EF may originate from a preexisting increased wall stress on the grounds of hypertension, diabetes, and increased wall thickness (⁹), or may be caused from direct effects of hypoxia, acidosis, and the release of toxic inflammatory mediators during infection and/or sepsis on myocardium (^{11, 22, 31-33}).

Determinants of NT-pro-BNP levels in ICU

Most studies in ICU patients have focused on LV systolic dysfunction as assessed by means of an impaired EF and its impact on natriuretic peptides and mortality (^6-8). The independent association between NT-pro-BNP and TDI markers of LV diastolic function found in the present study suggests that LV diastolic dysfunction is a major determinant of elevated NT-pro-BNP levels in ICU patients with preserved EF. Studies have shown that myocyte stretch because of elevated LV filling pressures may cause BNP production (²). A further stimulus of NT-pro-BNP release is tissue hypoxia (¹²). Thus, reduced blood oxygenation because of pulmonary congestion on the grounds of LV diastolic dysfunction may contribute to BNP production. In our study, an NT-pro-BNP greater than 941 pg/mL was a reliable predictor of LV diastolic dysfunction with 73% sensitivity and 70% specificity. In ICU patients, the myocyte stretch leading to elevated NT-pro-BNP may originate from the LV filling pressures on the basis of hypertension, diabetes, and increased wall thickness (^{9, 24, 34}). Supporting this hypothesis, studies have shown that LV stroke work, as well as pulmonary artery occlusion pressure assessed after cardiac catheterization, correlated with NT-pro-BNP in ICU patients (^{29, 34}). However, increased NT-pro-BNP production may be triggered from the direct effect of inflammatory mediators during sepsis (^{10, 11, 13}), elevated levels of stress hormones (^{2, 11}), renal dysfunction (²), or from overtreatment with fluids, vasoactive medications (¹¹), and positive pressure ventilation in critically ill patients. Sepsis is associated with high cardiac output, and therapy includes administration of fluids. It is possible that cardiac distension occurs, with preserved systolic function causing elevation of BNP secretion (^{11, 35}). In agreement with other investigators (^{4-8, 10}), we have shown in the present study that NT-pro-BNP also correlated with age, creatinine levels, inotrope infusion, presence of sepsis, and hypoxia. Furthermore, the independent association, by multivariate analysis, of NT-pro-BNP levels with S, E/E′, and presence of sepsis supports the hypothesis that sepsis may cause NT-pro-BNP release in critically ill patients in addition to diastolic LV dysfunction. The interrelation between sepsis, TDI markers of LV function, and NT-pro-BNP in our study suggests that the inflammatory processes and release of inflammatory mediators in septic conditions may cause LV diastolic dysfunction (^10-22), leading to elevated NT-pro-BNP levels in ICU patients. However, sepsis may cause elevated NT-pro-BNP levels through mechanisms other than myocardial damage such as renal impairment, increased stress hormones, and elevated levels of inflammatory cytokines (¹¹) or coexistent noncardiac causes of elevated NT-pro-BNP (¹⁰). In summary, elevated NT-pro-BNP may be a marker of preexisting LV diastolic dysfunction on the grounds of hypertension diabetes and LV hypertrophy; of acquired LV diastolic dysfunction because of sepsis, inflammation, hypoxia, and acidosis; or of the systemic inflammatory response during infection or sepsis in our critically ill patients.

Determinants of ICU mortality

In our previous study (³), we have shown that NT-pro-BNP levels are independent predictors of outcome in a large mixed ICU population. In the present study, we have demonstrated that LV diastolic dysfunction may determine NT-pro-BNP levels in addition to other comorbidities. Furthermore, we have shown that LV diastolic dysfunction as assessed by TDI is a prognostic marker for ICU mortality in patients with preserved EF in addition to elevated NT-pro-BNP levels and presence of sepsis. In our study, patients with a composite of abnormal TDI markers of diastolic LV function had a 4-fold risk of death during hospitalization. Furthermore, the addition of the composite of abnormal TDI markers of LV diastolic function, in a model including NT-pro-BNP and sepsis, significantly increased the prognostic value of the model for ICU mortality. This finding suggests that in the presence of low NT-pro-BNP levels or sepsis, TDI markers become particularly useful in distinguishing patients at a substantially lower or higher risk of an adverse event.

Additionally, the combination of normal TDI markers and low NT-pro-BNP was associated with a survival rate of 84%, whereas the combination of abnormal TDI markers and BNP has a survival rate of 25%. Thus, the combination of NT-pro-BNP levels with the TDI parameters of LV function could discriminate those with an excellent compared with those with a poor prognosis during ICU hospitalization.

Patients with the combination abnormal TDI markers and high NT-pro-BNP levels had a lower survival rate compared with patients with abnormal TDI or NT-pro-BNP alone or to patients with normal TDI markers and low BNP levels.

Thus, abnormal TDI markers provided additional prognostic information in patients with high NT-pro-BNP levels and vice versa.

Our results suggest that TDI markers and NT-pro-BNP levels may be complementary predictors of ICU mortality in patients in critically ill patients with preserved EF and are in line with studies demonstrating the independent and incremental prognostic value of TDI markers in cardiovascular disease (²⁵).

Considering the causal relation of LV diastolic dysfunction with outcome, LV diastolic dysfunction raises the left atrial and pulmonary vein pressure (^{9, 22-25}) and thus causes pulmonary congestion contributing to reduced blood oxygenation and, consequently, to tissue hypoxia, acidosis, and multiple organ failure. Indeed, in our study, patients with LV diastolic dysfunction had a reduced PO₂/FiO₂ and increased lactic acid. Additionally, LV diastolic dysfunction may not permit an adequate increase of stroke volume in the presence of excessive peripheral vasodilation during sepsis or under conditions of low output, leading to tissue hypoperfusion (¹¹). In critically ill patients, elevated NT-pro-BNP may be a cumulative marker of 1) LV diastolic dysfunction either preexisting, because of cardiac disease, or acquired, because of hypoxia, hypoperfusion, and toxic effects of inflammation; 2) renal dysfunction, 3) inflammatory process and tissue hypoxia during infection and/or sepsis; 4) elevation of stress hormones and neurohumoral activation and thus may be a valid predictor of adverse outcome (^{2, 10-17}).

However, BNP is an active hormone that increases the cellular production of cyclic guanosine monophosphate, leading to relaxation of vascular smooth muscle cells and renin inhibition (²). Thus, elevated NT-pro-BNP may cause inappropriate vasodilation, especially in septic conditions, leading to reduced MAP and thus may aggravate tissue hypoperfusion, renal dysfunction, and multiple organ failure. Brain natriuretic peptide antagonizes angiotensin II-induced efferent arteriolar vasoconstriction and thus reduces glomerular filtration rate, leading to increases in creatinine levels (¹⁵).

In agreement with this hypothesis, treatment with nesiritide, a synthetic form of BNP, has been associated with adverse outcome in patients with heart failure because of hypotension and deterioration of renal function (^14-16). In the present study, we have shown that patients with NT-pro-BNP greater than 941 pg/mL had lower MAP and serum creatinine than patients with lower NT-pro-BNP. Thus, through these mechanisms, NT-pro-BNP levels may have a causal association with outcome in our critically ill patients. Furthermore, patients with combination of elevated NT-pro-BNP and impaired TDI indices had the lowest MAP and PO₂/FiO₂ in addition to elevated serum creatinine compared with the remaining patients. Thus, pulmonary congestion and/or inadequate increase of stroke volume under conditions of peripheral vasodilation, because of LV diastolic dysfunction, combined with a low MAP, because of the vasodilatory effects of increased BNP, may contribute to reduced blood oxygenation, tissue hypoperfusion, hypoxia, and further deterioration of renal function. The previously discussed mechanisms may at least partially explain the complementary role of LV diastolic dysfunction and increased NT-pro-BNP levels in the prediction of adverse outcome in our cohort of critically ill patients.

Study limitations

One of the limitations of our study was the absence of hemodynamic evaluation of LV filling pressures with a Swan-Ganz catheter even if the safety and use of these catheters have been questioned. Furthermore, several studies have shown a close correlation between elevated E/E′ and invasively assessed LV filling pressures (^{9, 18}). The relatively small sample of our patient cohort should also be acknowledged. However, in our study, we have shown that TDI indices of LV function are a useful prognostic tool, which are easily performed and readily available by the bedside of critically ill patients and may discriminate subtle diastolic abnormalities of the cardiac wall, leading to increased NT-pro-BNP levels and adverse outcome. Thus, the assessment of TDI velocities of the mitral annulus is a useful tool to rule out the origin of elevated NT-pro-BNP levels (LV dysfunction or noncardiac factors) in critically patients with preserved EF and to stratify more accurately the risk of adverse outcome on patients' admission to ICU.

Because patients are admitted into the ICU at varying time points in the progression of their illness, serial assessment of TDI markers and NT-pro-BNP over time may be more useful in the determination of prognosis and the effects of treatment.

In our study, nonsurvivors had reduced oxygenation, a higher level of acidosis, a higher incidence of sepsis, and impaired indices of LV diastolic function than survivors. The close link between adverse outcome and LV diastolic dysfunction, reduced oxygenation, acidosis, and sepsis in our study suggests that preexisting diastolic dysfunction may generate a more "vulnerable" myocardium to toxic effects of hypoxia, acidosis, and sepsis. It should be further investigated whether prompt treatment of these factors may reverse at least partly LV diastolic function and, consequently, patients' prognosis of during ICU hospitalization. In patients with impaired TDI indices and elevated NT-pro-BNP implying elevated LV filling pressures, initiation of diuresis, ultrafiltration, and/or inotropes should be considered individually according to a patient's underlying disease, clinical condition, blood pressure, loading conditions as assessed by CVP, renal function, and hourly urine volume. However, LV diastolic dysfunction may also be a marker of increased severity of illness and thus may not respond to these interventions.

CONCLUSION

In the present study, we have shown that LV diastolic dysfunction, as assessed by TDI and sepsis, determine NT-pro-BNP levels in critically ill patients with preserved EF. Moreover, abnormal TDI indices of LV function and NT-pro-BNP have a complementary value for in-hospital mortality.