Left ventricular (LV) diastolic properties are a major determinant of LV filling and subsequent stroke volume. On clinical grounds, LV diastolic function is commonly assessed using echocardiography Doppler 1. Unfortunately, pulsed wave Doppler indices used for the evaluation of LV diastolic properties have long been known to be altered by numerous factors, including loading conditions and heart rate 234. Accordingly, their use to identify LV diastolic dysfunction in clinical settings characterized by abrupt variations in preload or afterload, such as severe sepsis or septic shock 567, may be of limited value. Recently, animal and clinical studies have suggested that Doppler tissue imaging (DTI) of the mitral annulus and early diastolic blood flow propagation velocity measured using colour M-mode may constitute parameters of LV diastolic function that are relatively preload independent 8910111213.

Haemodialysis provides a unique opportunity to evaluate the effect of a preload reduction on Doppler parameters currently used to assess LV diastolic properties. In this specific setting, recent clinical studies conducted in patients with chronic renal failure have yielded discrepant results 141516171819. The volume of ultrafiltration withdrawn from patients is presumably a major determinant of the sensitivity of Doppler parameters to preload reduction. Accordingly, we sought to evaluate whether the new indices of LV diastolic function provided by DTI of the mitral annulus and colour M-mode were altered by intermittent haemodialysis when performed in two distinct clinical settings: high-volume ultrafiltration in patients with chronic renal failure and lower volume ultrafiltration in ventilated critically ill patients with acute renal failure receiving vasopressor therapy. We also evaluated the potential influence of LV systolic function on variability in Doppler indices induced by haemodialysis.

The protocol was approved by the Ethics Committee of the Société de Réanimation de Langue Française, which waived the need of signed informed consent. All patients were informed about the study either at the time of echocardiography (haemodialysis patients) or as early as possible after recovery (ventilated patients).

No perdialytic hypotension was observed and haemodynamics remained fairly stable throughout the study period in the two patient groups (Table 1). In haemodialysis patients, the greater volume of ultrafiltration tended to further reduce LV end-diastolic volume index compared with ICU patients (16 ± 8 ml/m² versus 11 ± 4 ml/m²; P = 0.08). As a result, cardiac index significantly decreased after ultrafiltration in haemodialysis patients but not in ICU patients (Table 1). In ambulatory patients, blood pressure decreased after haemodialysis whereas it remained stable in ICU patients receiving vasopressors, because of a compensatory increase in systemic vascular resistance (Table 1).

Despite the absence of relevant tachycardia secondary to ultrafiltration-related intravascular volume reduction, haemodialysis induced substantial alterations in both mitral and pulmonary vein Doppler patterns in the two study groups (Table 1). In haemodialysis patients, isovolumic relaxation time and E-wave deceleration time increased after ultrafiltration. Because the E-wave maximal velocity decreased whereas the A-wave maximal velocity remained unchanged after ultrafiltration, the E/A ratio was significantly reduced by haemodialysis (Figure 1). Opposite variations were observed with the S/D ratio, because the D-wave maximal velocity decreased in parrallel to the E-wave velocity, whereas the S-wave maximal velocity remained unchanged after ultrafiltration. Overall, variations in pulsed wave Doppler indices induced by haemodialysis were less pronounced in ICU patients (Table 1).

DTI and colour M-mode Doppler findings obtained before and after haemodialysis in the two study groups are shown in Table 2. In six patients, Vp was not measured because of inadequate imaging quality, whereas TDI was successfully performed in all studied patients. At baseline, 18 haemodialysis patients (49%) had low E' maximal velocities, consistent with abnormal LV relaxation, whereas Vp was decreased in 14 patients (38%). Among them, eight patients had concentric LV hypertrophy and 10 had LV systolic dysfunction. TDI E' velocity remained stable after ultrafiltration in ICU patients, whereas it decreased significantly only when recorded at the level of the septal mitral annulus in haemodialysis patients (Figure 1). Vp uniformly decreased after ultrafiltration in the two study groups, the difference being significant only in haemodialysis patients (Table 2). Both the E/E' and E/Vp ratios decreased after haemodialysis in ambulatory patients, whereas they were not significantly altered by ultrafiltration in ICU patients (Table 2).

In haemodialysis patients, baseline LV systolic function apparently failed to alter the effect of ultrafiltration on Doppler parameters of LV diastolic function (Table 3). The volume of ultrafiltration was similar in the two subsets of patients (3.0 ± 1.2 l versus 2.8 ± 0.9 l; P = 0.43) and induced a comparable preload reduction, reflected by the close decrease in LV end-diastolic volume index (18 ± 12 ml/m² versus 20 ± 15 ml/m²; P = 0.9). DTI E' velocity tended to be less affected by ultrafiltration when measured at the lateral than at the septal aspect of the mitral ring, and variations were less pronounced in the presence of LV systolic dysfunction. In contrast, Vp tended to decrease with preload reduction, regardless of LV systolic function (Table 3).

Inter-observer variability in measurement of Doppler parameters ranged from 1% to 13%, whereas intra-observer variability ranged from 2% to 7%. Notably, measurement of DTI E' velocity at the level of the mitral ring was more reproducible than that of Vp (Table 4).

Haemodialysis represents an unparalleled model of abrupt alteration in LV loading conditions. Ultrafiltration allows achievement of marked preload reduction, as reflected in our patients by a significant decrease in LV end-diastolic volume index, a commonly used surrogate of LV preload (Table 1). Importantly, careful reduction of intravascular volume induced by ultrafiltration failed to trigger reflex tachycardia (Table 1), thus avoiding the confounding effect of heart rate on pulsed wave Doppler velocity profiles 1. In our ICU patients, the absence of blood volume expansion or change in vasopressor infusion rate throughout the study period limited additional variations in LV loading conditions. This allowed us to assess more specifically the effect of haemodialysis-related preload reduction on Doppler indices that are routinely used in the assessment of LV diastolic function 122. Echocardiographic studies were performed immediately before and at least one hour after haemodialysis to reach a relatively stable volaemic state, the process of plasma refilling being completed by that time 16. Finally, measurements of Doppler parameters were reproducible, as reported earlier in our ICU for two-dimensional measurements 26. In keeping with the results reported by Bouhemad and coworkers 27, variability in Vp measurement was greater than that in DTI E' velocity in our patients (Table 4).

The high prevalence of LV diastolic dysfunction found at baseline in our patients with end-stage chronic renal failure using previously proposed diagnostic criteria based on DTI and colour M-mode 22 is related to the frequency of LV hypertrophy and congestive heart failure in this population 2829. Previous studies reported similar results with use of conventional pulsed wave Doppler 303132. Not surprisingly, the present study confirms the load dependence of the pulsed wave Doppler parameters used to evaluated LV diastolic properties. As previously reported 151632333435, the abrupt preload reduction induced by ultrafiltration significantly decreased E-wave maximal velocities and the E/A ratio because A-wave velocities remained unaffected (Table 1). Both the isovolumic relaxation time and E-wave deceleration time were significantly prolonged by volume reduction, as previously shown 1415323335. Finally, the evolution of pulmonary vein Doppler D wave after intravascular volume withdrawal paralleled that of mitral Doppler E wave, and the S/D ratio increased after haemodialysis (Table 1), as was previously reported 1417. In light of these findings, the conclusions of clinical studies using conventional pulsed wave Doppler parameters to identify transient LV diastolic dysfunction in acute settings (for instance, septic shock) should be interpreted with caution 567.

In the present study, certain recently proposed Doppler parameters of LV diastolic function appeared also to be sensitive to acute and marked variations in LV preload related to ultrafiltration. In haemodialysis patients, TDI E' velocity decreased with preload reduction when recorded at the septal portion of the mitral annulus, as previously reported 16, whereas TDI E' velocities recorded at the lateral aspect of the mitral ring remained unaffected by haemodialysis (Table 2). Similar discrepancies in haemodialysis-related variations in E' maximal velocities according to site of measurement were previously reported 1417. As previously observed 16, Vp globally decreased after haemodialysis in ambulatory patients (Table 2).

In our ventilated ICU patients, ultrafiltration-induced variations in pulsed wave Doppler parameters were similar to those observed in haemodialysis patients but were of lesser magnitude (Table 1). Of note, Vp tended to decrease after haemodialysis whereas DTI E' velocity remained fairly stable, regardless of the site of measurement (Table 2). This is presumably related to the lower volume of ultrafiltration withdrawn during haemodialysis from these critically ill patients under vasopressors to limit perdialytic hypotension 36, when compared with our haemodialysis patients. Overall, we found no correlation between the magnitude of variation in DTI velocity or Vp and the volume of ultrafiltration withdrawn from our patients (data not shown). Nevertheless, the amount and rapidity of preload reduction may play a major role in the decreases in E' septal maximal velocity and Vp observed in certain patients, as has previously been demonstrated in other clinical settings 37.

Although of lesser magnitude, a substantial decrease in septal E' velocity and Vp induced by ultrafiltration was observed in the subset of haemodialysis patients with altered LV systolic function (Table 3). In a similar clinical setting, Liu and coworkers 1819 found that Vp was preload dependent in patients with preserved LV systolic function, whereas it was not significantly altered by haemodialysis in patients with LV systolic dysfunction. This apparently discrepant findings may be attributable to the lower volume of ultrafiltration used in these studies 1819. Indeed, small preload variations induced by Valsalva manoeuvre, passive leg lifting and sublingual nitroglycerin fail to alter Vp, regardless of LV systolic function 12. Animal studies suggested that TDI velocities are partly influenced by preload, but their preload dependence appears to decrease with worsening relaxation and associated LV diastolic dysfunction 383940. Although most of our haemodialysis patients had severe LV diastolic dysfunction associated with high filling pressure, as reflected by elevated E/Vp and E/E' ratios 10212741, most of the Doppler indices of LV diastolic function were significantly altered by the marked preload reduction related to the high volumes of ultrafiltration used in the present study.

The study has several limitations. Our results cannot be extrapolated to clinical settings in which variations in LV loading conditions are less abrupt than those obtained by intermittent haemodialysis. The group of ICU patients was fairly small, and observed results in this specific subset of patients remain to be confirmed. The hypothesis raised by the present data is that smaller changes in Doppler indices observed in ICU patients are presumably related to the lower volume of ultrafiltration withdrawn from these patients compared with haemodialysis patients, rather than a potential lusitropic effect of catecholamines. Finally, we did not assess the potential role of afterload variations and changes in cardiac contractility induced by haemodialysis on observed variations in Doppler indices 42.

This study confirms the preload dependence of conventional pulsed wave Doppler parameters routinely used to evaluate LV diastolic function. Among the recently proposed Doppler indices for the evaluation of LV diastolic properties, TDI E' maximal velocity measured at the lateral portion of the mitral annulus appeared to be relatively preload independent and reproducible. In contrast, E' septal velocity and Vp appeared sensitive to a marked and abrupt preload reduction. The preload dependence of Doppler parameters was not significantly influenced by LV systolic function. Accordingly, Doppler parameters should be used cautiously to evaluate LV diastolic properties and to diagnose transient diastolic dysfunction in clinical settings characterized by abrupt and marked changes in cardiac loading conditions.

• Abrupt variations in cardiac preload markedly alters conventional pulsed wave Doppler parameters routinely used to assess LV diastolic properties.

• Marked preload reduction may also alter both DTI E' velocity when recorded at the septal mitral ring and Vp measured by colour M-mode.

• DTI E' velocity recorded at the lateral aspect of the mitral ring appears to be relatively insensitive to variations in preload.

• These variations were apparently unaffected by LV systolic function.

DTI = Doppler tissue imaging; E/A = mitral early diastole/atrial contraction maximal velocity ratio; E' wave = DTI early diastolic velocity of the mitral annulus; ICU = intensive care unit; LV = left ventricular; S/D = pulmonary vein systolic/diastolic maximal velocity ratio; TEE = transoesophageal echocardiography; TTE = transthoracic echocardiography; Vp = propagation velocity of LV inflow at early diastole measured using color M-mode.

The authors declare that they have no competing interests.

PV, JCA and HG contributed to the design of the clinical study, data analysis and preparation of the manuscript. VA, JL, JFM and BF contributed to the recruitment and clinical evaluation of patients throughout the study period.