OUP user menu

The role of NT-proBNP in the diagnostics of isolated diastolic dysfunction: correlation with echocardiographic and invasive measurements

Carsten Tschöpe, Mario Kašner, Dirk Westermann, Regina Gaub, Wolfgang C. Poller, Heinz-Peter Schultheiss
DOI: http://dx.doi.org/10.1093/eurheartj/ehi406 2277-2284 First published online: 13 July 2005


Aims Diastolic heart failure is a frequent entity but difficult to diagnose. N-terminal pro-B type natriuretic peptide (NT-proBNP) was therefore investigated as a possible non-invasive parameter to diagnose isolated diastolic dysfunction.

Methods and results Sixty-eight symptomatic patients with isolated diastolic dysfunction and preserved left ventricular ejection fraction (LVEF) (≥50%) and 50 patients with regular left ventricular (LV) function were examined by conventional echocardiography, tissue Doppler imaging (TDI), and left and right heart catheterization. Plasma NT-proBNP levels were determined simultaneously. Median NT-proBNP plasma levels were elevated [189.54 pg/mL (86.16–308.27) vs. 51.89 pg/mL (29.94–69.71); P<0.001] and increased with greater severity of the diastolic dysfunction (R=0.67, P<0.001). According to the recevier operating characteristic analysis, LV end-diastolic pressure [area under the curve (AUC) 0.84] was the most specific parameter, which had a low sensitivity (61%), however. The reliability of NT-proBNP was similar to TDI indices (AUC 0.83 vs. 0.81) and improved when compared with conventional echocardiography (AUC 0.59–0.70). NT-proBNP levels had the best negative predictive value of all methods (94%) and correlated strongly with indices of LV filling pressure, as determined by invasive measurements. Multivariable linear regression analysis confirmed NT-proBNP as an independent predictor of diastolic dysfunction with an Odds ratio of 1.2 (1.1–1.4, CI 95%) for every unit increase of NT-proBNP.

Conclusion NT-proBNP can reliably detect the presence of isolated diastolic dysfunction in symptomatic patients and is an useful tool to rule out patients with reduced exercise tolerance of non-cardiac origin.

  • Left ventricular diastolic dysfunction
  • N-terminal pro-B type natriuretic peptide
  • Heart failure
  • Biomarker


Heart failure with normal or minimally impaired systolic function is attributed to diastolic dysfunction. More than one-third of patients presenting with symptoms and signs of congestive heart failure (CHF) have isolated diastolic dysfunction, which is associated with a poor prognosis.1,2 Clinical examination cannot distinguish between systolic and diastolic heart failure.3 The diagnosis of diastolic heart failure is rather based on exclusion. Doppler echocardiography is used for a quick bedside assessment of diastolic function, but its sensitivity and specificity in detecting diastolic abnormalities are unsatisfactory due to many factors including heart rate, age, left ventricular (LV) loading conditions, and operators skills. Similar problems may be expected for tissue Doppler imaging (TDI), AV-plane displacement, and magnetic resonance techniques. Invasive investigations via left and right heart catheterization to measure LV filling pressures are more reliable, but also load-dependent and not useful for wide spread clinical implementation. For these reasons, simple and reliable diagnostic criteria for diastolic heart failure are lacking and a rapid non-invasive diagnostic test would be of high clinical value.

Abnormal diastolic filling pressure, the key functional abnormality in diastolic heart failure, leads to a release of cardiac neurohormones including natriuretic peptides.4 Previous studies have reported that B type natriuretic peptide (BNP) and its biologically inactive fragment N-terminal proBNP (NT-proBNP), both of which are released predominantly by the ventricles in response to stretch, may be used for the diagnosis of systolic heart failure.5 However, the role of these peptides in patients with diastolic heart failure is still under investigation.611 Although it has been found that BNP and NT-proBNP correlate with diastolic abnormalities in patients with reduced systolic function9 and in patients with advanced forms of isolated diastolic heart failure,6,12,13 several Doppler echocardiographic studies found them not to be useful for the detection of mild diastolic failure.1117 The latter is associated with increased filling pressures at exertion only and will be missed by conventional Doppler echocardiography, which is performed at rest. Because NT-proBNP circulates at higher plasma concentrations and has a longer half-life when compared with BNP,18 NT-proBNP could be useful for the detection of all degrees of diastolic dysfunction. To study this hypothesis, we have investigated the potential of NT-proBNP to detect isolated diastolic dysfunction, according to the guidelines of the European Study Group On Diastolic Heart Failure,19 and its accuracy in comparison to established invasive and non-invasive methods including left and right heart catheterization, transmitral Doppler echocardiography, pulmonary venous Doppler, and TDI in a series of patients with clinically suspected CHF despite preserved LV systolic contractility and dimensions.


Patient population

We investigated prospectively 118 patients admitted to our unit who had preserved LV function and normal LV dimensions as determined by echocardiography and ventriculography. Sixty-eight patients with exertional dyspnoea (NYHA class I–III) who met the criteria for isolated diastolic dysfunction were compared with 50 patients with normal diastolic function. Patients with atrial fibrillation, lung diseases, renal dysfunction, significant heart valve disease, or other severe concomitant diseases were excluded. Medications that can influence haemodynamics (diuretics, beta-blockers, calcium-blockers, and ACE- and AT1-receptor inhibitors) were all paused for 48 h before examinations were performed.

All investigated patients gave informed written consent for invasive diagnostic procedures, including echocardiography and right and left cardiac catheterization, which were performed by using standard techniques. Our study complies with the Declaration of Helsinki. The locally appointed Ethics Committee has approved the research protocol.

Definition and assessment of diastolic dysfunction

Following the guidelines from the European Study Group on Diastolic Heart Failure,19 the diagnosis of diastolic dysfunction was defined after the evidence of abnormal LV relaxation, filling, and/or diastolic distensibility in the presence of clinical signs of CHF, with demonstrable normal or only mildly impaired systolic function (EF≥50%). During LV angiography, slow isovolumic LV relaxation was indicated by an increase in dP/dtmin (≥−1100 mmHg/s) and/or a prolongation of the time constant of LV pressure decay (Tau ≥48 ms) as derived and calculated from digitalized LV pressure recordings using a modified commercially available software program (MacLab, Wisstech, Germany), and/or during echocardiography (VINGMED System Five operating at 2.5–3.5 MHz) by a prolongation of isovolumic relaxation time (IVRT ≥94–100 ms) as derived from LV outflow tract Doppler signals. Slow early LV filling was indicated by a reduction of the ratio of E-wave (early filling) to A-wave (atrial filling) peak velocities (E/A≤1) and/or of the E-wave deceleration time (DT≥220–260 ms) as derived from LV Doppler signals. Because the specificity and sensitivity of any one of these three echocardiographic parameters per se are low, diastolic dysfunction was considered when at least two of these parameters were abnormal after age and heart rate relaxation.20 Confirmation included also pulmonary vein flow analysis in the right upper paraseptal pulmonary vein where systolic (S) and diastolic (D) velocities were measured, as well as the velocity (Ar) and duration of the atrial reversal wave. Using TDI, the ratio of early-to-late annular velocity of lateral mitral leaflets origin (E′/A′) was determined in the apical four-chamber view. All echocardiography data were copied to VHS videotape for subsequent playback, analysis, and measurement. Reduced LV diastolic distensibility was indicated by an increase in LV end-diastolic pressure (LVEDP≥16 mmHg) as derived from LV pressure recordings and/or by pulmonary capillary wedge pressure (PCWP) measured by right heart catheterization (Swan–Ganz catheter) at rest (≥12 mmHg) or during exercise (≥20 mmHg) (bicycle ergometry). Exercise was performed by bicycle ergometry in a recumbent position with the use of a continuous protocol, which began with a workload of 25 W that was increased by 25 W every 2 min. Patients were asked to cycle at a rate of 50 r.p.m. at each stage, and exercise was terminated when the patient reached 75% of the calculated maximum heart rate or because of dyspnoea or of the occurrence of leg/general fatigue. NT-proBNP plasma levels were determined at rest using the commercially available Elecsys® proBNP sandwich immunoassay on an Elecsys 2010 (Roche Diagnostics, Germany). All measurements were obtained without knowledge of NT-proBNP data.

Overall diastolic stage, determined from the pattern of mitral and pulmonary venous flow, was defined as impaired relaxation (stage I: E/A ratio <1, DT>220 ms; S/D ratio >1, Ar<35 cm/s), pseudonormal (stage II: E/A ratio 1–2, DT>150–220 ms; S/D ratio <1, Ar>35 cm/s), or restrictive (stage III: E/A ratio >2, DT<150 ms; S/D ratio <1, Ar>35 cm/s).21,22 Reduction of the E′/A′ ratio (<1) in TDI confirmed diastolic dysfunction, including stages I and II. E′ and A′ amplitude <8 cm/s indicated a restrictive flow pattern.

LV mass index was calculated according to Devereuxs formula23 divided by body surface area. We calculate left atrial volume index (LAVI) according to the already described formula: LAV=π/6 (diameter in parasternal long-axis view)×(short-axis in the apical four-chamber)×(long-axis in the apical four-chamber)24 at ventricular end-systole divided by body surface area.

Statistical analysis

SPSS Inc. for Windows Standard version 11.0.1 was used for statistical analysis. Continuous variables are expressed as mean±SD (range). Quantitative normally distributed data were compared using the Student'apos; t-test. Non-normally distributed variables were analysed with non-parametric Mann–Whitney U test. To compare qualitative data, χ2 test was performed. Receiver operating characteristic (ROC) curves were constructed to determine the ability of NT-proBNP throughout the range of concentrations (cut-off points) to identify diastolic dysfunction. The diagnostic utility of NT-proBNP alone was compared with the echocardiographic probability of LV dysfunction through the estimation of the area under the curve (AUC) for each parameter separately. Linear regression analysis was used to determine correlations between continuous variables and the log-transformed NT-proBNP levels were used, as the values were not normally distributed. Spearman correlation was used for correlation assessment of NT-proBNP with the groups of LVDD. Multivariable logistic regression (backward likelihood ratio model) was performed to evaluate the ability of NT-proBNP to identify diastolic dysfunction over and above the information provided by other indicators from Table 1 [age, sex, hypertension, diabetes mellitus, CAD, LV-hypertrophy, and body mass index (BMI)]. A value of P<0.05 was considered statistically significant and α-level adjustment (Bonferroni correction) with factor 3 (P<0.017) is used to avoid the experiment-wise Type I error due to multiple testing between the LV diastolic dysfunction groups.

View this table:
Table 1

Patients characteristics [variable expressed as mean±SD (range)]

Population (n=118)Diastolic dysfunction (n=68)Regular diastolic function (n=50)P-valuea
Demographic variables
 Women/men [n (%)]53 (45)/59 (55)31 (46)/37 (54)22 (44)/28 (56)0.708
 Age (years)49±13 (70–26)51±9 (69–26)49±10 (70–28)0.091
 BMI (kg/m2)26.1±4.1 (32.7–16.3)26.8±4.0 (31.9–16.5)25.3± 4.0 (32.7–17.9)0.120
Vital signs
 MAP (mmHg)91.9±12 (123–61)93.6±11 (117–73)90.1±12 (123–61)0.172
 NYHA II–III [n (%)]66 (56)58 (85)8 (16)0.001
 Heart rate (min−1)71±11 (58–90)70±12 (60–82)71±11 (58–90)0.865
Left heart dimensions
 LVEDD (mm)50±6 (38–59)51±6 (40–59)50±4 (38–58)0.473
 LVMI (g/m2)107±31 (65–198)114±28 (65–198)103±26 (72–154)0.371
 LAVI (mL/m2)18.8±5.9 (11.3–34.2)21.0±6.8 (10.4–34.2)18.1±4.5 (11.3–27.9)0.303
 LVEF (%)67±10 (84–51)68±9 (78–51)65±10 (84–52)0.093
Concomitant diseases [n (%)]
 Hyperlipoproteinemia49 (41)30 (44)19 (37)0.687
 Smoker33 (28)21 (31)12 (24)0.433
 Arterial Hypertension44 (37)29 (43)15 (30)0.060
 Diabetes mellitus11 (9) 9 (13) 2 (0.4)0.011
 Coronary artery disease43 (34)28 (41)15 (30)0.086
 Viral/inflammatory myocardiopathy28 (24)22 (32) 6 (12)0.023

aDiastolic dysfunction vs. regular diastolic function.


Study participants

From 180 patients initially planned to enter the study, 11 of them were excluded because of the low quality of data recorded from echocardiography or heart catheterization measurements. Four patients refused to be examined. In addition, 47 patients were excluded by which the right heart catheterization could not perform and, therefore, no PCWP at rest or exercise was determined. Finally, the study group consisted of 118 patients. From these 118 patients, 68 patients with diastolic dysfunction (mean age 51±9; 31 women and 37 men) were studied prospectively and compared with 50 control subjects with normal diastolic function (mean age 49±10; 22 women and 28 men) (Table 1). All 118 patients had normal LV systolic function (EF>55%) and normal LVEDD in the parasternal long axis (<60 mm). There were no significant differences between both groups with respect to LVEF (68±9 vs. 65±10%, P=0.093), LVEDD [51±6 (40–59) vs. 50±4 (38–58) mm, P=0.473], sex, BMI, LVMI, or LAVI. As expected, the prevalence of committed disease was increased in patients with diastolic dysfunction,1,25 indicated by a trend for a higher prevalence of arterial hypertension and coronary heart disease (CAD) and a higher frequency of diabetes mellitus and suspected myocarditis with abnormal diastolic function (Table 1).

In 68 patients, diastolic dysfunction was confirmed by abnormal values of LVEDP, Tau, IVRT, DT, and/or by the E/A ratio. Pseudonormal and restrictive flow pattern were diagnosed using pulmonary vein flow and TDI. Among 118 patients who underwent right heart catheterization, 61 had an increased PCWP at rest and/or at exercise (Table 2). dP/dtmin, which probably reflects systolic relaxation rather than early diastolic filling, did not differ between both groups. Diastolic dysfunction was diagnosed with a high rate of objective measured abnormal diastolic indices. Fifty-one per cent of patients with diastolic dysfunction had three or four abnormal diastolic echocardiographic and/or invasive parameters, 22% had more than five and 15% had more than seven positive indices.

View this table:
Table 2

Echocardiographic and invasive parameters in patients with diastolic dysfunction vs. regular diastolic function

AllDiastolic dysfunctionRegular diastolic functionP-valuea
LV/RV catheter
 LVEDP (mmHg)12.3±5.7 (27–4)15.5±5.3 (27–5)9.1±4.5 (15–4)<0.001
 Tau (ms)41.5±9.1 (72–23)43.7±9.9 (72–28)38.9±7.3 (46–23)0.009
 dP/dtmin (mmHg/s)−1613±378 (−2738–(−864))−1574±376 (−2487–(−864))−1659±380 (−2738–(−1115))0.271
 PCWP rest (mmHg)8.6±5.1 (29–2)11.0±6.0 (29–3)6.4±3.2 (11–2)<0.001
 PCWP exercise (mmHg)19.1±8.8 (47–5)21.7±10.3 (47–7)13.8±5.5 (20–5)0.002
 IVRT (ms)105±27 (160–65)113±30 (160–65)97±19 (119–65)0.002
 DT (ms)193±44 (265–131)198±48 (265–134)187±36 (251–131)0.250
E/A ratio1.2±0.42 (0.42–3.56)0.9±0.39 (0.42–3.56)1.3±0.29 (0.54–1.84)<0.001
S/D (PV)1.1±0.34 (0.49–2.0)1.3±0.38 (0.49–2.00)0.9±0.24 (0.64–1.64)0.008
E′/A′ ratio (TDI)1.20±0.54 (2.14–0.29)0.87±0.26 (1.36–0.49)1.51±0.56 (2.14–0.63)<0.001

Variable expressed as mean±SD (range).

aDiastolic dysfunction vs. regular diastolic function.

NT-proBNP levels in patients with isolated diastolic dysfunction

NT-proBNP levels were four-fold elevated in patients with diastolic abnormalities when compared with control patients [189.54 pg/mL (86.16–308.27) vs. 51.89 pg/mL (29.9–69.7); P<0.001). There was no significant difference between men and women with diastolic abnormalities [164.3 pg/mL (72.6–260.5) vs. 204.0 pg/mL (108.9–318.9); P=0.112]. NT-proBNP levels increased significantly according to the severity of overall diastolic dysfunction, ranging from impaired relaxation [151.6 pg/mL (90.6–278.1) vs. controls 51.89 pg/mL (29.9–69.7), P<0.001] to pseudonormal filling [308.1 pg/mL (261.7–568.2) vs. 151.6 pg/mL (90.6–278.1), P=0.003] and restrictive filling [2307.1 pg/mL (1592.0–6440.1) vs. 308.1 pg/mL (261.7–568.2), P<0.001] and correlated with the Spearman coefficient R=0.67, P<0.001 (Figure 1). Even after α-level adjustment, the differences between LVDD groups remained significant (P<0.017). The proportion of patients with abnormal sex- and age-adjusted NT-proBNP levels among the control, impaired relaxation, pseudonormal, and restrictive subgroup amounted to 8, 74, 94, and 100%, respectively (Figure 2).

Figure 1 NT-proBNP levels in the patients with LV diastolic dysfunction are significantly elevated and correlate with the severity of disease: controls 51.89pg/mL (29.9–69.7) vs. impaired relaxation 151.6pg/mL (90.6–278.1) (P<0.001) vs. pseudonormalization 308.1pg/mL (261.7–568.2) (P=0.003) vs. restriction 2307.1 pg/mL (1592.0–6440.1) (P<0.001).

Figure 2 Proportion of patients with abnormal NT-proBNP levels. Controls, impaired relaxation, pseudonormal, and restrictive filling pattern amount to 8, 74, 94, and 100%, respectively.

Similarly, NT-proBNP levels increased significantly according to the NYHA-classification, ranging from NYHA class I [97.5 pg/mL (77.2–120.6) vs. controls 55.7 pg/mL (32.7–86.3), P<0.040] to NYHA class II [177.3 pg/mL (74.1–293.3) vs. 97.5 pg/mL (77.2–120.6), P=0.008] and NYHA class III [334.7 pg/mL (180.2–976.8) vs. 177.3 pg/mL (74.1–293.3), P<0.015], and correlated with the Spearman coefficient R=0.48, P>0.001.

Diagnostic accuracy of NT-proBNP to diagnose isolated diastolic dysfunction

ROC curve analyses revealed an AUC for NT-proBNP of 0.83, between the AUCs of LVEDP (0.84) and TDI (0.81) (Table 3). All others were less reliable with the following AUC value: PCWP at exercise 0.76, PCWP at rest 0.74, E/A ratio 0.70, Tau 0.64, IVRT 0.63, DT 0.59, and dP/dtmin 0.58. Thus, NT-proBNP was almost as good as the LVEDP and TDI and better than the E/A ratio of Doppler echocardiography.

View this table:
Table 3

Sensitivity/specificity/positive and negative predictive values by different cut-off values

NT-proBNP cut-off (pg/mL)Sensitivity (%)Specificity (%)PPV (%)NPV (%)

At a cut-off value of 110 pg/mL (Table 3), NT-proBNP showed a high sensitivity of 72%, a specificity of 97%, a positive predictive value of 84%, and the highest negative predictive value of all investigated techniques (94%) (Table 4 and Figure 3). After performing the multivariable logistic regression analysis, NT-proBNP was an independent predictor of isolated diastolic dysfunction with an Odds ratio of 1.2 (1.1–1.4, CI 95%), for every unit increase in the NT-proBNP level (Table 5).

Figure 3 ROC analysis revealed the reliability of NT-proBNP to diagnose an isolated diastolic dysfunction with 82.7%.

View this table:
Table 4

AUC, sensitivity, specificity, and positive and negative predictive values by NT-proBNP at a cut off-value of 120 pg/mL

ParameterAUC (95% CI)Sensitivity (%)Specificity (%)PPV (%)NPV (%)
LVEDP0.84 (0.73–0.91)61946992
NT-proBNP0.83 (0.78–0.89)69916393
E′/A′ (TDI)0.81 (0.75–0.90)71875593
PCWP exercise0.76 (0.65–0.84)39945988
PCWP rest0.74 (0.59–0.85)42874187
E/A0.70 (0.62–0.77)53793688
Tau0.64 (0.52–0.74)33914586
IVRT0.63 (0.58–0.77)69602790
DT0.59 (0.46–0.68)33792684
dP/dtmin0.58 (0.47–0.67)11953383
View this table:
Table 5

Multivariable logistic regression for evaluating the ability of NT-proBNP to identify diastolic dysfunction when compared with other indicators

IndicatorORP-value95% CI
Diabetes mellitus1.0420.9140.340–4.710
LV mass index1.0280.1430.991–1.067

After backward likelihood ratio, analysis reminds NT-proBNP as an independent predictor with OR 1.2 (1.1–1.4). OR, Odds ratio; CI, confidence interval; CAD, coronary artery disease; BMI, body mass index.

Correlation of the NT-proBNP plasma levels with Doppler echocardiographic indices, TDI, and invasive measured diastolic indices

Linear regression analysis revealed significant correlations between NT-proBNP and various echocardiographic parameters (Table 6). Thereby, a moderate relation to early diastolic velocity (E′) (R=−0.41, P<0.001), early-to-late diastolic velocity ratio (E′/A′) (R=−0.36, P<0.001) of mitral lateral annular movement, and a weak relation to A–Ar duration (R=−0.31, P=0.037) were found. NT-proBNP and E′/A′ analysis by TDI correlated significantly with all invasive LV diastolic parameters as measured directly by left and right heart catheterization. The best correlation was found with LVEDP, PCWP at rest, and at maximal exercise with R-values of 0.45, 0.42, 0.49 (P<0.001), respectively. NT-proBNP did not correlate significantly with LVEDD, LVMI, DT, early and late mitral flow, mitral E/A ratio, PV systolic and diastolic, PV systolic to diastolic ratio (S/D), and reversal atrial flow.

View this table:
Table 6

Correlation of NT-proBNP levels with invasive and non-invasive diastolic indices according to univariate linear regression analysis

Systolic RR0.280.014
Diastolic RR0.060.593
Heart rate0.270.011
Mitral flow
PV flow
 A–Ar duration−0.310.037
Tissue Doppler
 PCWP, rest0.42<0.001
 PCWP, exercise0.49<0.001


To investigate the role of NT-proBNP in the detection of isolated diastolic dysfunction, we formed a comparative study of NT-proBNP, invasive haemodynamics, and comprehensive echocardiography. Our study showed that NT-proBNP plasma levels were increased in patients with diastolic dysfunction and preserved systolic function and correlated with the severity of the disease. NT-proBNP had the best negative predictive value of all methods investigated. The best correlation was found between the NT-proBNP levels and the invasive parameters of LV filling pressure. NT-proBNP levels had similar diagnostic accuracy for diastolic heart failure as TDI and were superior to conventional echocardiography.

The role of NT-proBNP in the detection of diastolic dysfunction

Instant measurement of plasma natriuretic peptides that are normally increased as part of the neurohormonal activation in both diastolic and systolic heart failure is a powerful biochemical test. There are several reports on the efficacy of BNP 26,27 or NT-proBNP25,2832 in diagnosing chronic heart failure due to LV systolic dysfunction. A particular advantage of that test is its high negative predictive value, allowing to rule out systolic heart failure. The usefulness of BNPs in detecting diastolic heart failure is still under investigation. Few studies have examined the value of BNPs in diagnosing diastolic dysfunction in comparison with conventional echocardiography. Significantly higher BNP and NT-proBNP levels were found in patients with advanced diastolic dysfunction (restrictive and pseudonormal)6,12,13 and our study supports these findings. Although these peptides were found to correlate with severe diastolic dysfunction, however, their role in the detection of mild diastolic heart failure is uncertain.14 Lubien et al.6 found that BNP levels were increased in patients with exertional dyspnoea and impaired LV relaxation (202 pg/mL). Although Mottram et al.12 also found BNP levels increased in patients with hypertension-caused impaired LV relaxation compared with controls, however, in up to 75%, BNP levels were in the normal range (89 pg/mL) in patients with impaired relaxation in this study. Wei et al.15 also reported low but significantly increased BNP-levels in hypertension-caused diastolic dysfunction, which showed a moderate sensitivity but an excellent specificity in detecting ventricular diastolic dysfunction in hypertensive patients. Similarly, Dahlström17 found only a trend for both BNP and NT-proBNP to be increased into the upper normal reference range of patients with little changes of Doppler echocardiography indices. These conflicting reports may reflect the heterogeneity of the diastolic heart failure populations investigated, the impact of a variety of clinical features, and differences in study design and emphasize the need for a particularly careful characterization of study patients in this field of research. Most previous studies used conventional echocardiographic parameters as the sole diagnostic criteria to identify isolated diastolic dysfunction. Obviously, the accuracy of transmitral flow analysis is limited as also observed in our study. To overcome this key problem, we have defined abnormal diastolic performance by a combination of invasive and comprehensive echocardiographic measurements. We found that increased NT-proBNP levels correlate with the severity of diastolic disease. This even includes the subgroup of clinically stable patients with near normal LVEDP at rest, impaired relaxation, and exertional dyspnoea. It appears that the prolonged half-life of NT-proBNP when compared BNP is advantageous under these circumstances and enables NT-proBNP to identify mild isolated diastolic dysfunction already. Hammerer-Lercher et al.10 recently reported that the performance of NT-proBNP and BNP for the diagnosis of diastolic dysfunction was assay-dependent with a clear benefit for the Elecsys NT-proBNP and Triage® BNP assays, but further comparative evaluations are warranted.

Impaired LV relaxation was detected with a specificity of 90% and a negative predictive value of 94% by a NT-proBNP of 120 pg/mL in 75% of symptomatic patients. This value is similar to the recently recommended official NT-proBNP cut-off value of 120 pg/mL. However, this cut-off value is an average value from several community studies focusing more on LV systolic function and not based on the complicated diagnostic, which is necessary to analyse also sufficient isolated diastolic dysfunction. Our study shows that even lower NT-proBNP levels (90–110 pg/mL) reached the negative predictive value of 94% accompanied with only a slight reduction in specifity (Table 3). Therefore, NT-proBNP is an useful tool to rule out patients with reduced exercise tolerance of non-cardiac origin. From the clinical point of view, the negative predictive value of ≥94% appears most useful, because in patients with exercise-induced dyspnoea at NT-proBNP <90 pg/mL, a cardiac origin of symptoms is highly unlikely, whereas at NT-proBNP >110 pg/mL, diastolic dysfunction is the prominent differential diagnosis. But as advanced forms of diastolic dysfunction may reach NT-proBNP levels similar to those in patients with severe systolic heart failure, NT-proBNP cannot differentiate between diastolic and systolic heart failure and is not a surrogate for echocardiography.

Although the sensitivity of NT-proBNP to detect isolated diastolic dysfunction was moderate (69%) and therefore not accurate for a screening test as already also shown by Redfield et al.,16 NT-proBNP had one of the best sensitivity of the tested methods in our study.

In conclusion, NT-proBNP reliably detects diastolic dysfunction in patients with filling abnormalities and preserved LV systolic function, however, in line with the recent European guidelines incorporating BNPs in to the algorithm for the diagnosis of heart failure.33

NT-proBNP vs. invasive and echocardiographic measurements

This study is the first to compare NT-proBNP for the diagnosis of isolated diastolic dysfunction in symptomatic patients with a broad panel of established methods including left and right heart catheterization, transmitral, PV Doppler, and TDI. Gold standard for the assessment of diastolic function is invasive measurement with a conductance catheter providing pressure–volume relationships. In the clinical practice, the more easily available measurement of LV pressures during routine left and right heart catheterization also enables direct analysis of the early (dP/dtmin, Tau) and late period (LVEDP and PCWP) of diastole. However, these values are volume- and heart rate-dependent, which is reflected by our finding of a high specificity (89–95%) but low sensitivity (<61%) of LVEDP and PCWP, respectively, to detect diastolic dysfunction (Table 4). The differences between PCWP at rest and exercise were not as high as expected, because several patients with diastolic dysfunction did not really reach high exercise levels (mean maximal outcome: 83 W; mean exercise duration time 7.3 min) due to dyspnoea, fatigue, or leg pain, which shows the limitation of this method in the clinical routine. Nevertheless, LVEDP and PCWP, both reflecting LV filling pressure, correlated stronger with NT-proBNP levels. PCWP at exercise, used to investigate symptomatic patients with near normal filling pressures at rest, correlated best with NT-proBNP, as we had also shown recently,34 and is consistent with the fact that the release of BNPs being triggered by ventricular volume expansion and pressure overload.35 In addition, dP/dtmin and tau, reflecting systolic relaxation or early diastolic relaxation rather than myocardial filling pressure, correlated weaker with NT-proBNP levels.

As the aforementioned reliable invasive measurements are not routinely available in the diagnostic workup of suspected diastolic heart failure, appropriate non-invasive tests are needed for the pre-invasive screening. Doppler echocardiography is currently the non-invasive method of choice for the assessment of diastolic function, but its implementation in clinical routine is suboptimal due to operator dependency and limited availability in primary health care. Furthermore, E/A, IVRT, and DT do not directly measure diastolic function but diastolic flow over the mitral valve only and this may explain its rather low accuracy.6,12,36 It is also in agreement with our finding that none of the conventional echocardiographic parameters reached sufficient sensitivity and specificity (33–79%) for reliable detection of isolated diastolic function and may explain why the NT-proBNP levels of our study population did not correlate with DT and most of the indices determined by PV Doppler. Although IVRT correlated significantly the diastolic function in patients with concomitant systolic heart failure, Mottram et al.12 could not verify IVRT as an important diagnostic marker in patients with isolated diastolic dysfunction. Similarly, although a short DT indicates an increase in LA mean pressure in patients with systolic32 or restrictive diastolic heart failure,6 our data do not support a diagnostic relevance of DT, reflecting early diastolic filling, as 24% of our patients with diastolic failure had prolonged DTs. This confirms Dahlström and co-workers11,17 who also found no significant relationship between DT and BNP or NT-proBNP in a study population with mild diastolic abnormalities. In summary, owing to many limitations, Doppler echocardiography cannot provide unequivocal evidence of diastolic dysfunction.37

TDI is a more recent technique that measures myocardial velocities directly. The early diastolic mitral annular velocity (E′/A′) has been shown to be a relatively load-independent measure of myocardial relaxation. It correlated significantly better with the NT-proBNP levels in our study population than the E/A ratio determined by conventional echocardiography, consistent with studies of diastolic function comparing TDI or the ratio of mitral velocity to early diastolic velocity of the mitral annulus (E/E′) derived from TDI and BNP levels in intensive care patients38 or patients with systolic or diastolic heart failure.9,34,39 Thus, NT-proBNP levels have similar diagnostic accuracy for diastolic heart failure as TDI and are superior to conventional echocardiography.

In conclusion, according to its ROC, NT-proBNP stands in the line with invasive parameters of LV filling pressure and is characterized by a high negative predictive value. Therefore, NT-proBNP can reliably detect the presence of isolated diastolic dysfunction in symptomatic patients and is a useful tool to rule out patients with reduced exercise tolerance of non-cardiac origin.


The study was supported by the Deutsche Forschungsgesellschaft (DFG: SFB/TR 19).


  • The first two authors contributed equally.


View Abstract