Heart failure has an important association with type 2 diabetes mellitus (T2DM),1 although coronary artery disease is responsible for much of the cardiovascular burden of this disease.2 One potential mediator of this phenomenon is subclinical myocardial dysfunction, which is present in the absence of left ventricular (LV) hypertrophy at rest3 and during exercise.4 5

This subclinical disease corresponds to stage B heart failure, as classified by the American Heart Association/American College of Cardiology guidelines.6 The appropriate management of this entity is undefined—small studies have suggested that myocardial dysfunction in these patients may be reversed by pharmaceutical interventions aimed at improving glycaemic control.7 However, although exercise training has been shown to improve myocardial function in an animal model of T2DM,8 to date this has not been assessed in humans. Therefore, the aims of this study were to investigate the effects of a 1-year exercise intervention on myocardial function in T2DM and to identify the determinants of improvements in myocardial function in order to better understand how to manage this condition. We hypothesised that exercise training would improve myocardial function through improved glycaemic control, and that the patients who would derive the most benefit would be those with the greatest metabolic disturbance at baseline, especially those with impaired baseline myocardial function.

MATERIALS AND METHODS

Patient selection

Patients with T2DM from hospital clinics and the community were eligible for inclusion if they were between 18 and 75 years of age, excluding only those with a serious co-morbidity (life expectancy <6 months), pregnancy or known cardiovascular disease. Recruited patients (n = 248) underwent screening for occult coronary artery disease using exercise echocardiography, which excluded 25 patients. Based on increases of myocardial diastolic velocity in previous pharmaceutical intervention studies (an increment of 1.1 cm/s in patients with a baseline diastolic tissue velocity of 5.5 (SD 1.7) cm/s)7 we anticipated that 56 patients per group would provide a 90% power. Therefore, 112 patients per group were recruited in the expectation that 50% would drop out or fail to train adequately.

After baseline testing, patients were randomly allocated to either usual care (n = 112) or exercise intervention (n = 111) by an independent statistician using random number generation software. Thirteen groups of 15–20 patients were randomised every 3 months between 2003 and 2006, and randomisation was stratified within each group.

After randomisation, 12 declined the intervention, and in the course of 12 months of follow-up, 15 patients withdrew from the study and 20 did not attend for follow-up. Thus, 176 of 223 patients (79%) underwent follow-up, of whom 88 received exercise training. The 12 patients who declined the intervention cited lack of time and personal reasons. Of the 15 patients who withdrew from the study four moved from the area, two patients could not be contacted and the remaining nine patients withdrew for unstated or personal reasons. The trial was approved by research ethics committees of the Princess Alexandra Hospital and the University of Queensland, and was registered with the Australian Clinical trial registry (ACTRN12607000060448). Informed consent was obtained from all participants.

Usual care

All participants received standard risk factor intervention for T2DM, including support to maintain a target blood pressure of <130/80, smoking cessation, and attainment of lipid targets (low-density lipoprotein (LDL) cholesterol <2.6 mmol/l, triglycerides <1.7 mmol/l and high-density lipoprotein (HDL) cholesterol >1.0 mmol/l). Usual care was coordinated through the Diabetes Clinic of Princess Alexandra Hospital, a teaching hospital in Brisbane, Australia.

Exercise intervention

In addition to usual care, patients randomised to the lifestyle intervention group received a program of individualised exercise training with the addition of some general dietary management designed to be feasible for routine implementation and based on the current guidelines9 (achieving a balanced diet—50% carbohydrates, <35% fat and a polyunsaturated: monounsaturated: saturated fat (PMS) ratio of 1:1:1). Dietary changes were derived from seven day food records and analysed using a nutrition software program (Foodworks Professional 2006, Xyris Software Pty Ltd, Brisbane, Australia).

An accredited exercise physiologist (AEP) supervised the exercise intervention, which was provided in two stages—an initial, 4-week supervised, gym-based training programme (stage 1), followed by home-based training with regular telephone exercise counselling (stage 2). Stage 1 of the training programme has been described elsewhere.10 Briefly, it aimed to achieve a minimum of 150 minutes of at least moderate exercise (12–13 on Borg 20-point scale) each week from a individualised combination of aerobic and resistance training. Patients attended two 1-hour supervised exercise sessions and were asked to perform an additional 30 minutes of exercise at home each week.

During the telephone-led home stage of the intervention, telephone contact was made weekly for 3 months, fortnightly for 3 months and monthly thereafter for the remainder of the study. The telephone counselling aimed to monitor and record levels of exercise and physical activity, as well as to ensure each patient achieved, maintained and, where possible, increased the amount of exercise through motivation, changing the exercise prescription and implementing strategies to account for disruption to the exercise routine and overcoming barriers.

Myocardial function

Tissue Doppler imaging was used to assess myocardial function and has been previously described.3 11 Briefly, images were acquired (Vivid 7, GE Medical Systems, Horten, Norway) in three apical views (apical four-chamber, two-chamber and long-axis views). Peak myocardial systolic and diastolic tissue velocities were obtained by placing a sample volume in specific locations and tracking the wall motion through ventricular contraction (Echopac, GE Medical Systems). Myocardial deformation (strain and strain rate) curves were extracted from an average of three cycles of tissue Doppler imaging data (Echopac, GE Medical Systems), by placing a 12-mm sample length in each segment in the three apical views,12 13 manually tracked to wall motion. All measurements were taken from each of the six LV basal wall segments (septal, lateral, anteroseptal, posterior, inferior and anterior) and averaged to give global measurements. Baseline and post-intervention values were reviewed in a blinded fashion by an independent sonographer, using a side-by-side display to ensure consistency in the placement of the sample volume. Myocardial dysfunction in our patient cohort has been classified according to population-based norms for E′, strain and strain rate.14 Patients were classified as having abnormal function if either their E′, strain or strain rate were below these population norms. Previously reported kappa values between observers have been in the range of 0.8.14

Conventional echocardiography

Conventional echocardiography measures (LV systolic and diastolic dimensions, ejection fraction and LV mass index) were measured as described previously.15 Briefly, LV mass was determined by Devereux’s formula.16 Resting LV end-diastolic and end-systolic volumes and ejection fraction were computed using a modified Simpson’s biplane method.

Clinical and laboratory measurements

Height, body mass, waist circumference, resting supine blood pressure (BP) (Baumanometer, W.A. Baum Co, New York, NY, USA) were measured in all patients before and after the 1-year intervention. Fasting blood lipid profile, glucose, insulin and HbA_1c were measured using standard procedures by the Queensland Health Pathology Service (Brisbane, Australia) after an overnight fast and at least 24 hours since the last exercise session. Simplified indices of insulin sensitivity (HOMA and QUICKI) were determined using previously derived formulas.17 18

Cardiorespiratory fitness

This was assessed by indirect calorimetry (Vmax29c, SensorMedics, CA, USA), measuring maximal oxygen consumption (VO_2max) during a graded exercise test to exhaustion. Blood pressure and cardiac status (using a 12-lead electrocardiogram) were monitored during the exercise test (CASE, GE Medical Systems, Milwaukee, WI, USA). The percentage of normal VO_2max was assessed from nomograms derived from subclinical populations in men and women.19 20 The 6-minute walk test was used at baseline and post-intervention to assess functional capacity.

Exercise adherence

At baseline patients were classified as sufficiently or insufficiently active using a cut-off of 150 minutes, reflecting physical activity recommendations for patients with type 2 diabetes.21 Exercise and PA levels were assessed at baseline, 1 month and 1 year in the lifestyle intervention and control groups using a self-report, written questionnaire with items from the Active Australia Survey.22 Time spent in walking, moderate and vigorous activity was determined.

Statistical analysis

All data were tested for normality with the Kolmogorov-Smirnov test and the Normal Q-Q plot where appropriate. The following data required log transformation; body mass index (BMI), glucose, HbA_1c, HOMA, QUICKI, triglycerides, HDL, systolic and diastolic BP, VO_2max and the time spent in activity categories. General linear modelling with repeated measures analyses of variance (ANOVA) were used to assess group and time effects. Group × time interaction was used to establish differences between groups. Paired t tests were used to assess within-group changes. Mean absolute change was determined from the individual change. A post-hoc analysis to account for differences in physical activity performed was conducted by analyses of covariance (ANCOVA). The effect of group and the interactions between groups, change in moderate and change in vigorous activity were assessed to predict outcomes in myocardial function (systolic, diastolic, myocardial strain and strain rate) and metabolic function (BMI, waist circumference, percent body fat, HbA_1c, insulin sensitivity (QUICKI), systolic and diastolic BP and VO_2max). Pearson correlations were used to assess associations and multiple regression analysis was used to determine significant independent correlates, with p<0.10 in bivariate testing used as a requirement to enter the model. Significance was assumed if p<0.05. Receiver-operator characteristic (ROC) curves were used to determine cut-off values associated with improvements in myocardial function (one baseline standard deviation of diastolic tissue velocity (1.47 cm/s)). Data were analysed using standard statistical software (SPSS version 15.0 for Windows)

RESULTS

Patient characteristics

Table 1 provides a comparison of baseline clinical and laboratory data in the exercise intervention and control groups, and their change over 1 year for myocardial, echocardiographic, metabolic, cardiovascular, functional capacity and physical activity variables. In general, the intervention and control groups were similar at baseline, although the intervention group had a significantly higher baseline systolic BP (p<0.01). The majority of patients (89% and 87% for the intervention and control groups, respectively) were classified as having myocardial dysfunction. Similar proportions of the intervention (40%) and control patients (36%) were classified as inactive at baseline.

Adherence to intervention

Table 1 shows the time spent in different activity types at baseline and the change at 1 year for the exercise intervention and control groups. Compared to controls, the intervention group spent significantly more time in vigorous activity during the intervention (p<0.01). This was primarily attributable to a significant (p<0.01) increase in vigorous activity at 1 month (30.6 (SD 155.1) minutes). However, the intervention was effective in increasing moderate activity within the intervention group (p<0.05). There were no significant differences in the time spent walking between the intervention and control groups.

Dietary changes

Although the dietary targets for fat intake were reached in the exercise training group, with fat accounting for 29% of total energy consumed, the carbohydrate, polyunsaturated: monounsaturated ratio, fibre and sodium intake all improved from baseline in the intervention group, but failed to reach dietary targets. There were no dietary changes observed in the control group.

Efficacy of the intervention on myocardial function and echocardiographic measures

The intervention failed to reverse myocardial dysfunction in patients with T2DM, with no between-group differences observed in systolic or diastolic function parameters (systolic and diastolic tissue velocity, strain, strain rate) (table 1). Both groups showed significant increases in systolic (18%) and diastolic tissue (17%) velocities, and myocardial strain (5%) over follow-up, although no significant change in myocardial strain rate was identified. Similarly, there were no significant between-group differences for ejection fraction, LV mass, LV mass index or end-systolic and diastolic dimensions.

Efficacy of the intervention on metabolic function

Table 1 summarises significant improvements in body habitus (BMI, waist, fat mass), systolic and diastolic BP, glycaemic control (HbA_1c, glucose), insulin resistance (QUICKI, HOMA) and exercise capacity (VO_2max, 6-minute walk). The intervention group showed significant improvement (p<0.05) in waist circumference, fat mass, glucose, HbA_1c, QUICKI, VO_2max, percentage of normal VO_2max and 6-minute walk compared to controls. The improvement in HbA_1c was a result of preventing the deterioration seen in controls.

Interaction between group and activity levels

Table 2 summarises the significant ANCOVA models for prediction of changes in myocardial and metabolic function accounting for changes in moderate and vigorous activity performed in post-hoc analysis. Change in diastolic tissue velocity was significantly predicted by the interaction between groups, change in moderate activity and change in vigorous activity for the intervention group (p<0.01). Change in myocardial strain rate was significantly predicted by the interaction between groups and change in moderate activity for the control group (p = 0.03). The change in physical activity levels and their interaction with group assignment were not significant predictors of change in S′ or strain.

In the control group increases in HbA_1c and decreases in VO_2max were also significantly predicted by the interaction between groups, change in moderate and change in vigorous activity. Similarly, increases in HbA_1c were significantly predicted by the interaction between groups and change in moderate activity in controls. Improvements in diastolic blood pressure were significantly (p = 0.04) predicted by the interaction between groups and change in vigorous activity. Interaction models were also developed for the remaining variables but these were not significant.

Correlates of change in myocardial function

The correlates of change in myocardial parameters (table 3 and fig 1), showed different associations in systolic and diastolic tissue velocities, myocardial strain and strain rate. The independent determinants of change in systolic tissue velocity were baseline systolic tissue velocity and exercise capacity (percentage of normal VO_2max), with these two variables accounting for 61% of the change in systolic tissue velocity (p<0.001). The independent correlates of change in diastolic tissue velocity were baseline diastolic tissue velocity, change in HbA_1c and change in waist circumference, with these variables accounting for 51% of the change in diastolic tissue velocity (p<0.001) (fig 1). The interaction between group assignment and change in HbA_1c (β = −0.21, p<0.01) was also an independent predictor of change in diastolic tissue velocity (having a stronger association than change in HbA_1c alone). The change in myocardial strain was independently associated with baseline myocardial strain and HOMA, with these two variables accounting for 37% of the change in myocardial strain (p<0.001). Finally, baseline myocardial strain rate, insulin and HbA_1c were independent predictors of change in myocardial strain rate, accounting for 58% of this change.

• Download figure
• Open in new tab
• Download powerpoint

Figure 1

Association between baseline diastolic tissue velocity (A), baseline waist circumference (B), change in HbA_1c (C) and change in waist circumference (D) with change in diastolic tissue velocity (E). Change in diastolic tissue velocity for patients above and below the cut-off value for baseline diastolic tissue velocity (E). The vertical lines in (A) and (B) represent proposed cut-off values and the horizontal lines indicate a clinically significant improvement in diastolic tissue velocity.

Determining thresholds of benefit for lifestyle intervention

ROC curves for predicting improvements in diastolic tissue velocity were plotted for baseline diastolic tissue velocity and waist circumference. The determined cut-off values for baseline diastolic tissue velocity and waist circumference were 5.0 cm/s and 107.8 cm, respectively. Baseline diastolic tissue velocity had a larger area under the curve (AUC = 0.80), higher specificity (0.78) and sensitivity (0.71) compared to baseline waist circumference (0.64, 0.65, 0.59, respectively). Fewer than half of the patients recruited for this study were within the determined cut-off values for diastolic tissue velocity (42%) and waist circumference (47%). Patients who were within the cut-off values significantly increased their diastolic tissue velocity (fig 1E).

DISCUSSION

In this study, a 1-year exercise intervention (gym and home-based training) failed to improve myocardial function in all patients. However, post-hoc analysis revealed that subgroups of patients who had the greatest increases in both moderate and vigorous activity significantly improved diastolic function (tissue velocity), HbA_1c and cardiorespiratory fitness. Further, there was increased systolic deformation (myocardial strain rate) in intervention patients who had the greatest increases in moderate activity compared to controls. The intervention also produced significant improvements in waist circumference, body fat, glycaemic control, insulin sensitivity and functional capacity. Intervention patients who had greater increases in moderate or vigorous activity showed significant improvements in HbA_1c and diastolic blood pressure, respectively. Patients with more impaired myocardial function, greater metabolic upset (BMI, waist circumference, glycaemic control and insulin resistance) and lower functional capacity were more likely to improve their myocardial function over 1 year. The degree of improvement in glycaemic control and the improvement in body fat were associated with improvements in myocardial function.

Myocardial effects

Despite no observed differences in myocardial function between the overall groups in our primary analysis, the interaction between group assignment and changes in activity levels in predicting myocardial change has important implications for exercise prescription recommendations for patients with T2DM. While the metabolic effects of exercise were as effective as other studies, the intervention may have been insufficient to provide cardiovascular changes in all patients. Although previous intervention studies showed improvements in vascular function from three supervised sessions each week,23 24 other exercise interventions have involved 5–7 sessions per week,25 26 and myocardial improvements in animal models have also required greater amounts of exercise (typically five 60-minute sessions per week) than in guidelines.8 Therefore, although based on current recommendations, the prescription used in this study may have had insufficient training dose (volume or intensity) to yield improvements in myocardial function in all subjects. In fact the patients with the greatest increases in moderate and vigorous activity also showed myocardial benefit. Therefore a greater exercise dose, achieved through higher amounts of exercise and/or higher intensities than the current guidelines, may be required to reverse myocardial dysfunction in patients with T2DM. Future research is needed to confirm these findings and identify the dose required for these desired adaptations to occur.

We studied an unselected patient group, the majority of whom had myocardial dysfunction despite being generally physically active, overweight and having fair glycaemic control. The activity levels of the patients at baseline may have contributed to the lack of effect of the intervention. Interestingly, age was not associated with or predictive of changes in myocardial function. Patients who had worse myocardial function, more severe metabolic disturbance or worse functional capacity were significantly more likely to improve their myocardial function over follow-up. These findings are concordant with our own previous work.10

Metabolic effects

There were significant metabolic improvements—in particular, a 1.1 mmol/l reduction in blood glucose and a 0.7% reduction in HbA_1c compared to controls, which is analogous to the respective changes of 1 mmol/l and 0.9% with intensive medical therapy in the UKPDS study,27 as well as previous exercise interventions in similar populations.28

Previously reported changes in body composition with exercise training of patients with T2DM are mixed, with some studies reporting improvements29 and others no improvement30 in body composition. Our results were consistent with this pattern, with small reductions in waist circumference, percentage fat mass and BMI. The varying modes of exercise (resistance vs cardiorespiratory) and assessment techniques may be responsible for such findings.

There were significant (p<0.05) increases in VO_2max (2.4 ml/kg/min), percentage of normal VO_2max (8.1%) and 6-minute walking distance (22.5 m) observed in the intervention group compared to controls. A recent meta-analysis concluded that exercise training resulted in a mean increase of 2.1 ml/kg/min in VO_2max in patients with T2DM.31 Thus, although home-based training using telephone exercise counselling has failed to maintain benefits in other studies,32 this strategy appeared to be as effective as most, if not all,31 more structured strategies.

Limitations

Although dietary changes were minimal, separation of the effects of dietary changes (if any) from exercise training cannot be assessed. Both the control and intervention groups improved their myocardial function over the intervention. It is possible that the activity levels of patients at baseline and a Hawthorne effect (also evidenced by an increase in 6-minute walking distance in controls) may have biased the results, and the improvement of impaired function in both groups may reflect regression to the mean. Further, changes in dose of medication were not recorded. However, the medication profile (number of patients taking different types of medication) did not significantly change in either group (data not shown). Measurement of physical activity was made by self report. This approach, while being subjective, is considered to be reliable.33

Conclusion

Despite the current exercise recommendations failing to improve myocardial function in patients with T2DM, only intervention patients who achieved greater increases in moderate and vigorous activity showed an improvement in myocardial function, glycaemic control and cardiorespiratory fitness. Further, poor baseline myocardial function, body composition, glycaemic control and cardiorespiratory fitness were all associated with greater improvements in myocardial function, and exercise-induced decreases in HbA_1c and body composition independently predicted improvements in myocardial function. These results suggest that the current exercise recommendations for patients with T2DM may be insufficient to reverse myocardial dysfunction and highlight the importance of including both moderate and vigorous activity in exercise prescription for these patients.