Heart doi:10.1136/heartjnl-2012-302489
  • Heart failure
  • Original Article

Exercise-induced torsional dyssynchrony relates to impaired functional capacity in patients with heart failure and normal ejection fraction

  1. Francisco Leyva1
  1. 1Department of Cardiovascular Medicine, University of Birmingham, Birmingham, UK
  2. 2University Hospital of North Staffordshire and Institute for Science and Technology in Medicine, Keele University, Stoke-on-Trent, UK
  3. 3Department of Medicine & Therapeutics, The Chinese University of Hong Kong, Shatin, China
  1. Correspondence to Dr Yu Ting Tan, Department of Cardiovascular Medicine, The Medical School, University of Birmingham, Edgbaston, Birmingham B15 2TT, UK; ytingtan99{at}
  • Received 30 May 2012
  • Revised 6 October 2012
  • Accepted 9 October 2012
  • Published Online First 6 December 2012


Background Left ventricular (LV) systole and diastole are intimately dependent on myocardial torsion, which involves coupling between myocardial rotation (twisting in systole and untwisting in diastole) and longitudinal motion. Heart failure with normal ejection fraction (HFNEF) is known to involve exercise-induced wall motion abnormalities, but torsion on exercise has not been explored. We hypothesised that torsional dyssynchrony may also be involved and be exaggerated by exercise.

Methods and Results 67 patients (age 73±7 years, 45 female) with HFNEF and 38 controls underwent cardiopulmonary exercise testing and echocardiography at rest and on supine exercise. Analysis of three plane motions was performed using speckle tracking and tissue Doppler imaging. Torsional dyssynchrony was quantified as the SD of the time to peak systolic motion (SDSM) (basal and apical rotation, longitudinal and radial displacement); the time difference between peak twist and peak longitudinal displacement (twist-longitudinal motion delay, TLMD) and the ratio of untwist to longitudinal extension (UT:LE). At rest, HFNEF patients had similar SDSM, TLMD and UT:LE compared with controls. Exercise was associated with significantly more dyssynchrony in the HFNEF patients (SDSM 38.8±27.6 ms vs 25.9±15.5 ms, p=0.02; TLMD 28.4±46.2 ms vs 2.9±31.2 ms, p=0.005 and UT:LE 10.4±15.3 vs 3.3±3.8, p=0.022). The SDSM correlated positively with LV wall thickness (r=0.31, p=0.015) and negatively with peak oxygen consumption (r=−0.299, p=0.01) and changes in stroke volume on exercise (r=−0.371, p=0.001).

Conclusions HFNEF involves exercise-induced torsional dyssynchrony in systole and diastole, which relates to LV hypertrophy as well as exercise capacity.


Heart failure with normal ejection fraction (HFNEF) accounts for approximately half of patients with signs and symptoms of heart failure.1 It is increasingly recognised that the effect of HFNEF on survival, functional capacity and quality of life are comparable with those with heart failure and a reduced ejection fraction (EF) or systolic heart failure.2 ,3 Whereas there are proven effective therapies for systolic heart failure, such as drugs and devices, there is none presently for HFNEF.

Pathophysiologically, HFNEF involves many mechanisms including a rise in left ventricular (LV) end-diastolic pressure, consequent upon disturbances of ventricular contraction and relaxation, diastolic distensibility, and end-diastolic stiffness.4 Speculatively, LV dyssynchrony could impact upon these factors and therefore, could contribute to the haemodynamic disturbance observed in HFNEF. Previously, several studies have shown a high prevalence of dyssynchrony in HFNEF.5 ,6 These studies, however, assessed dyssynchrony in single myocardial planes, assessing time delays between different LV segments with techniques such as tissue Doppler imaging (TDI). Importantly, however, myocardial motion occurs in various directions, namely longitudinal, radial, circumferential and oblique directions, reflecting the orientations of myocardial fibres throughout the LV wall.7 The combined effect of such motion is torsion, or wringing and unwringing of the ventricle in systole and diastole, respectively.

The notion of torsional dyssynchrony arises from this recognition that cardiac motion occurs in three dimensions. Changes in LV wall properties such as LV hypertrophy and fibrosis may lead to temporal dispersion of wall motion in systole and diastole. Thus, torsional dyssynchrony could account for the impairment of functional capacity observed in patients with HFNEF. As an extension of our findings of exercise-induced systolic dysfunction in patients with HFNEF,8 the present study explores whether torsional dyssynchrony in systole or diastole, at rest or during exercise, relates to cardiac output and functional capacity in patients with HFNEF. We have used speckle tracking echocardiography and TDI for their ability to provide measurements of the individual components of myocardial motion.


We assessed LV systolic and diastolic function non-invasively at rest and on exercise in patients with the clinical diagnosis of HFNEF and in healthy controls. Ninety patients with adequate echocardiographic images from 130 patients seen in heart failure clinics were preselected and recruited into the study. All patients had symptoms of heart failure with New York Heart Association class II or class III and were on medical treatment for symptoms at the time of study (table 1). Exclusion criteria were pulmonary disease, congenital or valvular heart disease, electrical pacemakers or implantable cardiac defibrillators and established history of ischaemic heart disease.

Table 1

Clinical characteristics and basic echocardiographic data

Healthy, age-matched controls were recruited from local primary care clinics. Only those with no previous past medical history and on no regular medications were recruited into this study.

All subjects underwent cardiopulmonary exercise testing as previously described.8 All subjects gave written informed consent prior to their participation and the study was approved by the Institutions Research Ethical Committees.

Two-dimensional and tissue Doppler echocardiography

All subjects underwent full echocardiography examination using a GE Vivid Seven scanner (Horton, Norway) at rest and on exercise. Symptom-limited (fatigue or dyspnoea) exercise testing was done on a semirecumbent and tilting bicycle ergometer (Lode BV, Netherlands) in patients or to a maximum heart rate of 100 bpm in healthy controls (ie, submaximal exercise to maximise frame rates). The heart rate, symptom status, brachial blood pressure and heart rhythm were monitored continuously during exercise. Image acquisitions of all subjects were made by one operator and all echocardiographic measurements and off-line data analysis were done by two experienced independent observers blinded to each other's results.

Two-dimensional (2D) images and colour-coded tissue Doppler images (TDI) from parasternal (long axis and short axis at basal, mid-ventricular and apical levels) and apical views (two and four chamber views) were optimised, obtained and stored digitally as previously described.8 At least three sets of images with loops consisting of at least three consecutive cardiac cycles each were stored for offline analysis using a customised software package (EchoPac, GE). LV dimensions and wall thickness were measured according to the recommendation of the American Society of Echocardiography.9 LV volume and EF were measured using the modified biplane Simpson's method from the apical four and two chamber views.9 LV mass was calculated according to Devereux formula.10 Left atrial (LA) volume was calculated using the biplane area-length method from the apical four and two chamber views and indexed to body surface area to derive LA volume index.11

Peak mitral annular myocardial velocities of the left ventricle septal and lateral walls were recorded with real time pulse wave TDI method, as previously described.12 The early diastolic mitral annular velocity (e′) was measured and E/e′, an index of LV filling pressure, was calculated.13 Colour coded TDI images were also acquired over three consecutive cardiac cycles for septal and lateral LV walls and data was analysed as previously described.8 Longitudinal displacement was measured directly from longitudinal displacement curves derived from colour TDI images.

Speckle tracking on short-axis images at basal and apical levels was used to assess LV rotation, as previously described.14–17 Twist was calculated as the sum of apical and basal rotation. Torsion was derived from dividing LV twist by the diastolic length. The raw data from speckle tracking of apical and basal rotation, radial displacement from the short axis mid-ventricular level, the longitudinal displacement derived from TDI of the septal and lateral mitral annulus, were transferred and analysed using a custom-written Microsoft Excel algorithm, which interpolated all co-ordinates and time intervals, to enable comparison of events of different durations within the cardiac cycle. The raw data were used to plot and construct the twist-displacement loops to illustrate the co-ordination of these two motions at rest and on exercise. The algorithm was applied from methods used by Borg et al.18

Torsional dyssynchrony


Dyssynchrony in three planes namely longitudinal, radial and rotation, was calculated using the interpolated timing information. The SD of the time to peak systolic motions (SDSM), timed from R-wave to the peak of the individual motions (basal rotation, apical rotation, longitudinal displacement and radial displacement), was taken as a measure of systolic torsional dyssynchrony. The maximum time difference between the duration to reach peak twist and peak longitudinal displacement was calculated and expressed as twist-longitudinal motion delay (TLMD). All measurements were undertaken at rest and on exercise.


In diastole there is no obvious peak motion due to the split in an early (E) and a late (A) component of the motion. Therefore, the ratio of the percentage change in untwist to longitudinal extension (UT:LE) over a set time duration (isovolumic relaxation time, IVRT) was taken as a measure of torsional dyssynchrony in diastole. Accordingly, UT:LE was calculated as follows:Formula

This ratio is consistent with the early slope of the diastolic limb (red arrow) of the twist-displacement loops in figure 2.


Statistical analysis was performed using SPSS V.18.0 (Chicago, Illinois, USA) and Medcalc 12.0.1 (Mariakerke, Belgium). Continuous variables were expressed as mean±SD. Fisher's exact test was conducted for nominal variables. Comparisons between HFNEF patients and controls were performed using unpaired t-test for normally distributed data. Non-normally distributed data was analysed using Mann-Whitney U Test. Comparisons between rest and exercise results were done by paired student's t test for normally distributed data whereas Wilcoxon test was used for non-normally distributed data. Logarithmic transformation was used when variables were not normally distributed. Pearson's correlation coefficient was used to examine associations between variables.

Inter-observer and intra-observer variabilities were assessed using Bland-Altman analyses, based on readings from at least 10 randomly selected subjects for measurements that were applied for torsional dyssynchrony (rotational by speckle tracking, radial strain by speckle tracking, longitudinal displacement by colour TDI).


A total of 148 subjects (90 patients with HFNEF and 58 asymptomatic healthy subjects) were identified as potential participants. With respect to patients with HFNEF, 23/90 were excluded (seven had evidence of respiratory restriction, one had significant coronary artery disease, two were unable to exercise, three failed to increase their heart rate on exercise, five had atrial fibrillation and five did not have adequate images for analysis). The remaining 67 symptomatic patients fulfilled the Framingham HFNEF criteria.19

With respect to the initial 58 asymptomatic subjects considered, 20/58 were excluded after further investigation (16 had previously undiagnosed hypertension, one was tachycardic at rest due to anxiety, and three did not have adequate images for analysis). The 38 remaining healthy controls had no relevant past medical history and were not taking any medications.

The mean age of the patients was 73±7 years and 67% were female. Control subjects were of comparable age (71±7 years, p=NS) and 76% female. The past medical history and drug history of patients are summarised in table 1. All patients had symptoms of heart failure with New York Heart Association class II or class III despite being on medications. Patients had a significantly higher body mass index (BMI) compared with controls but peak oxygen consumption (VO2) which was indexed to BMI was significantly lower in patients compared with controls. All subjects achieved respiratory exchange ratio of >1 on cardiopulmonary exercise testing. (table 1)

2D echocardiography

The left ventricular ejection fraction, fractional shortening, end-systolic and end-diastolic dimensions, E/A ratio, deceleration time, and IVRT were all comparable between the two groups (table 1). Patients had a significantly increased LV wall thickness (posterior and interventricular septal walls), LV mass index (LVMI), left atrial volume index, mitral inflow E and A waves and E/e′ (table 1).


Heart rate and blood pressure were comparable between the groups at rest and were maintained at comparable level on exercise (table 2). Stroke volume index was not different at rest but was significantly lower in HFNEF patients on exercise (p=0.045).

Table 2

Haemodynamic, tissue Doppler, speckle tracking and torsional dyssynchrony data

Tissue Doppler imaging and speckle tracking

LV twist and torsion which were reduced at rest, became even more obvious on exercise in HFNEF patients. In fact, the magnitude of these measurements on exercise in patients remained lower than the resting results of control subjects (table 2). Longitudinal and radial displacement were comparable at rest but became significantly different on exercise (table 2).

There was a good correlation between the magnitude of LV twist and longitudinal displacement on exercise (r=−0.511, p<0.001) confirming the important mechanical link between these two motions which contribute to the overall mitral annular motion. Any time delay of these motions would contribute to the uncoupling of LV mechanics (as illustrated below).

Torsional dyssynchrony


At rest, the time to reach peak systolic motion (peak longitudinal displacement, radial displacement, basal and apical rotation), expressed as SDSM, was comparable between HFNEF and controls (47.7±32.5 ms vs 43.1±25.3 ms, p=0.469). On exercise, controls were able to reduce SDSM significantly (p=0.002) whereas patients failed to show a significant decrease (p=0.655) (table 2) (figure 1). The SDSM correlated inversely with peak oxygen consumption (r=−0.299, p=0.01) as well as with the increase in stroke volume on exercise (r=−0.371, p=0.001). In multivariable regression analyses, SDSM on exercise (coefficient −0.09, t-value=−2.86, p=0.006) was independent of BMI (coefficient=−0.79, t-value=−4.57, p<0.0001) (F-value=14.0, p<0.0001 for analysis) as a predictor of peak oxygen consumption. On the other hand, SDSM on exercise (coefficient −0.78, t-value=−1.96, p=0.055) was not independent of LVMI (coefficient=−0.06, t-value=−2.02, p=0.049) (F-value=5.34, p=0.008 for analysis) as a predictor of peak oxygen consumption. Importantly, the lower SDSM in control subjects on exercise was mainly due to a reduction in the maximum TLMD in contraction (25.5±54.3 ms at rest and 2.9±31.2 ms on exercise in controls, p=0.025). However, this did not occur to the same extent in HFNEF patients (56.1±67.0 ms at rest and 28.4±46.2 ms on exercise, p=0.205) (table 2). In addition, SDSM on exercise correlated with interventricular septal wall thickness (r=0.31, p=0.015) and LV posterior wall (PW) thickness (r=0.35, p=0.012).

Figure 1

Torsional dyssynchrony on exercise. Long.Disp., Longitudinal displacement; Rad.Disp., Radial displacement; AVC, aortic valve closure; red line (time of aortic time closure); green lines (time delay between twist and longitudinal motion).


The ratio of UT:LE was 23.1±54.9 in HFNEF and 7.1±10.7 in control subjects at rest (p=0.064). On exercise, the UT:LE was 10.4±15.3 in HFNEF patients and 3.3±3.8 in control subjects (p=0.022) (table 2). The logarithmic transformation of UT:LE (UT:LE(log)) was calculated in view of non-normally distributed UT:LE data. UT:LE(log) was 1.79±1.58 in HFNEF and 1.28±1.18 in control subjects at rest (p=0.130). On exercise, the UT:LE(log) was 1.63±1.15 in HFNEF patients and 0.67±1.3 in control subjects (p=0.003). As shown in figure 2, the relation between twist and longitudinal displacement was very different in HFNEF patients, with a steeper slope in HFNEF patients indicating that more untwist occurring before a given longitudinal displacement is achieved. In addition, HFNEF involved a delay in untwisting and longitudinal displacement, particularly on exercise. In contrast, diastolic UT and LE appeared to occur simultaneously in healthy controls. This uncoupling of diastolic LV motion on exercise correlated with interventricular septal wall thickness (r=0.428, p=0.001) and PW thickness (r=0.441 p=0.001).

Figure 2

Twist-Longitudinal displacement loops. (A) Twist-Longitudinal displacement loop of healthy control at rest. (B) Twist-Longitudinal displacement loop of heart failure with normal ejection fraction (HFNEF) patient at rest. (C) Twist-Longitudinal displacement loop of healthy control on exercise. (D) Twist-Longitudinal displacement loop of HFNEF patient on exercise. Twist and longitudinal displacement relationship during systole and diastole. Green curve, systolic motion; Red curve diastolic motion; Orange arrow, untwist (UT) without longitudinal extension (LE) in diastole; Blue arrow, UT and LE in diastole; Purple arrow, LE without untwist in diastole. In addition to the synchronous untwist and longitudinal motion in diastole (blue arrow) seen in control subject, there are two distinct separations of motions (orange and purple arrows) in patient, particularly on exercise (2b and 2d). AVO, aortic valve opening; AVC, aortic valve closure; MVO, mitral valve opening; MVC, mitral valve closure; Peak-E, peak mitral inflow E wave; End-E, end of mitral inflow E wave; Peak EJ, peak ejection; Peak T, peak twist; Peak TR, peak twist rate; Peak UTR, peak untwist rate.

There was significant correlation between TLMD and UT:LE (log) at rest (r=0.499, p<0.001) which became stronger on exercise (r=0.807, p<0.001) indicating the close relation between systole and diastole particularly during exercise.

Inter-observer and intra-observer variability

In Bland-Altman analyses, the mean bias (95% confidence limit of agreement) for interobserver at rest varied from 3% (−28.6 to 24.6%, colour TDI) to 4% (−25.3% to 30.7%, speckle tracking) and on exercise from 3.9% (−32.1 to 25.9%, colour TDI) to 6% (−32.6% to 16%, speckle tracking). For intraobserver speckle tracking and tissue Doppler analysis variabilities at rest were 2% (−18% to 13.7%) and 1.5% (−19 to 15%) respectively. On exercise, intraobserver variabilities were 6% (−24.8 to 28%) and 4.3% (−23.4 to 23.5%) respectively.


In previous studies,8 ,20 we found that in HFNEF patients, exercise leads to widespread abnormalities of longitudinal systolic and diastolic functional reserve, rotation, twist and untwisting, suction and atrial function. In the present study, we have employed novel measures of torsional dyssynchrony, which are derived from LV myocardial rotation, radial and longitudinal displacement as well as the coupling between twist and longitudinal displacement. We have used the temporal dispersion of these parameters, quantified in terms of SD, as a measure of torsional dyssynchrony. Using this approach, we have found that in controls, exercise leads to marked reductions in the SDSM and the TLMD, reflecting a reduction in the temporal dispersion of motion in the three planes as well as coupling of LV twist with longitudinal motion. In contrast, HFNEF was associated with lesser reductions in the SDSM and TLMD, reflecting a proportionally greater temporal dispersion and uncoupling of systolic LV twist and longitudinal motion on exercise, when patients became breathless. In addition, HFNEF was also associated with uncoupling of untwist and longitudinal motion (extension) in diastole, measured using the ratio UT:LE. These observations indicate that HFNEF involves exercise-induced torsional dyssynchrony, in systole and diastole.

The coupling between rotational and longitudinal wall motion is crucial for LV ejection and suction. During LV systole, contraction of the longitudinal fibres pulls the mitral annulus towards the apex at the same time as the oblique fibres contract to further bring the mitral annulus towards the apex and the wringing motion maximises LV ejection. During diastole, the release of potential energy enables the oblique fibres to spring back and hence untwist the left ventricle, creating an apex-base pressure gradient to enhance diastolic filling while the simultaneous longitudinal fibres relax to facilitate mitral annulus movement towards the base, enveloping the column of blood into the ventricle. The efficiency and coupling of these motions are crucial to maintain LV systolic and diastolic function, and particularly on exercise as illustrated in the strong correlation between TLMD and UT:LE.

In this study, we have shown that there is no difference in torsional dyssynchrony between HFNEF patients and age-matched controls at rest. On exercise, however, dyssynchrony was observed in patients with HFNEF, evidenced by a relatively higher SDSM compared with control subjects. Importantly, this measure of systolic torsional dyssynchrony correlated negatively with changes in stroke volume and with peak oxygen consumption on cardiopulmonary exercise testing. This suggests that in HFNEF, torsional dyssynchrony contributes to the inability to increase stroke volume and that this is related to dyspnoea and reduced exercise tolerance. Accordingly, a mechanical effect may ensue even in the absence of changes in traditional echocardiographic parameters, such as LV volumes. Secondly, the mechanical response may need to be assessed at rest and on exercise. However, it is possible that cardiac resynchronisation therapy (CRT) could correct exercise-induced torsional dyssynchrony by the same mechanism that it corrects dyssynchrony in systolic heart failure.

We have observed that in HFNEF, untwisting occurs before LE, unlike the coupled and nearly simultaneous longitudinal and untwisting motion observed in control subjects. This temporal dispersion or uncoupling of diastolic motions amounts to diastolic torsional dyssynchrony. Speculatively, early diastolic torsional dyssynchrony could lead to reduced suction and therefore, an increased end-diastolic LV and LA pressure and hence the symptom of breathlessness, the hallmark of HFNEF. Correction of systolic torsional dyssynchrony could lead to correction of early diastolic torsional dyssynchrony simply by correcting dyssynchrony before it continues into the diastolic phase.

The primary abnormality most probably resides in the extracellular matrix and the nature of the collagen related to LV hypertrophy, and this may be another promising therapeutic target.21 It is well established that most of the known conditions associated with HFNEF such as diabetes, hypertension, ischaemia and ageing all affect the extracellular matrix.22 In addition, the subendocardial situated fibres that sustain the long axis movement by the nature of their position are more prone to injury.23 ,24 LV hypertrophy by itself is known to be associated with abnormal coronary flow reserve that may result in subendocardial ischaemia particularly on exercise.25 But it is unlikely there is one pathophysiology mechanism common to all HFNEF subjects and a variety of abnormalities of LV mechanics will be present depending on the exact underlying pathology.

In contrast to previous studies of dyssynchrony in HFNEF5 ,6 we found a close relationship between the degree of systolic and diastolic dyssynchrony. This makes more sense as physiologically, systole and diastole are closely intertwined. For example, there is a close relationship between annular systolic and diastolic velocities across a wide range of LV EFs.26 In addition, LV wall thickness was found to correlate to the degree of dyssynchrony suggesting that there is a conduction delay particularly between the longitudinal fibres found mainly in the subendocardium and the oblique fibres located mainly in the subepicardium.

Limitations of the study

A recognised limitation of speckle tracking echocardiography is a relatively low frame rate, which could preclude detection of peak velocities. However, exercise during echocardiography was deliberately submaximal to keep the heart rate less than 100 bpm. Despite this, all patients became symptomatic on exercise. The SD of twist and torsion measurement was relatively large but these were comparable with other studies using the same method.27 This may reflect the limitation of speckle tracking itself which requires adequate image quality and frame rate, not present in approximately 35% of the screened cohort. Plane motion measurements and timings were not taken from the same frame but these were all corrected to cycle length and interpolated. Given that this is the first study to demonstrate dyssynchrony in different planes, there was no standard value of plane motion dyssynchrony to compare. This study has provided proof of concept that exercise-induced measures of torsional dyssynchrony occur in HFNEF, but it does not provide the clinician with defined cut-offs that can be applied in a clinical practice. While TDI28 and speckle-tracking echocardiography29 have been validated against the gold-standard of sonomicrometry, our derived measures of torsional dyssynchrony have not. Arguably, underlying myocardial ischaemia could contribute to our findings and systematic screening for myocardial ischaemia with myocardial perfusion studies and coronary angiography may have been useful. Importantly, however, 30% of patients had coronary angiography on the basis of previous clinical indications, and no evidence of obstructive coronary heart disease was observed. The remaining patients, who did not have a history of chest pain, did not exhibit ischaemic ECG changes on cardiopulmonary exercise testing.


Using novel measures of rotational dyssynchrony, we have demonstrated that in HFNEF, exercise leads to systolic and diastolic torsional dyssynchrony. This amounts to an increase in the temporal dispersion of motion in three planes as well as an uncoupling of LV twist and longitudinal motion. Further studies are needed to determine whether or not these disturbances are amenable to correction by CRT.


The authors are grateful for the customised algorithm software provided by Dr Alexander Borg and Dr Simon Ray (Manchester University), and also thank Eveline Lee, Rebekah Weaver and Stuart Wragg for their assistance.


  • Contributors YTT: British Heart Foundation funded research fellow involved in the design, planning, subject recruitment, data acquisition and analyses, and writing of manuscript. FWGW: British Heart Foundation funded research assistant involved in data acquisition and analyses. JES: supervision of research study, fund application, quality control and writing of manuscript. FL: data interpretation and writing of manuscript.

  • Funding The project was funded by a project grant from the British Heart Foundation (PG/06/006/106/21472) and an equipment grant from the North Staffordshire Heart Committee.

  • Competing interests None.

  • Ethics approval Solihull Local Research Ethics Committee 06/Q2706/69.

  • Provenance and peer review Not commissioned; externally peer reviewed.


Free sample
This recent issue is free to all users to allow everyone the opportunity to see the full scope and typical content of Heart.
View free sample issue >>

Don't forget to sign up for content alerts so you keep up to date with all the articles as they are published.