Waltz Dancing in Patients With Chronic Heart Failure: New Form of Exercise Training

The protocol was approved by the local Ethical Committee. Patients were randomized after a run-in period of 1 week, during which they signed an informed written consent form, and were visited by a cardiologist. The protocol was designed according to CONSORT statement suggestions.¹⁴ Patients were randomized to supervised exercise training (cycling, treadmill) at 70% of peak V̇o₂ 3 times a week for 8 weeks (group E, n=44) or to a dance protocol (group D, n=44). A group was not exercised and served as control (group C, n=42). They remained in the same allocation throughout the study period. A computer-generated randomization list was drawn up by an independent statistician. Dance sessions were performed at the hospital’s gym 3 times a week for 8 weeks. Heart rate was monitored during exercise sessions and dancing in all patients. On study entry and at 8 weeks, all patients underwent cardiopulmonary exercise testing, 2D-echo with Doppler, endothelium-dependent dilation of the brachial artery, and blood chemistry.

Dance sessions were performed at the hospital’s gym. Under heart rate monitoring, patients started the dance protocol after a warm-up phase of calisthenics, which lasted 10 minutes. The dance chosen was waltz, because of its wide popularity in many countries and the potential generalizability of the results. The sequence of waltz was 5 minutes of slow waltz, 3 minutes of fast waltz, 5 minutes slow, 3 minutes fast, and 5 minutes slow (total 21 minutes). The intensity of dance training was presumed to be similar to that of traditional training, because it was calculated as the heart rate corresponding to 70% of peak V̇o₂. Energy expenditure (EE) during dancing was calculated as recommended by the American College of Sports Medicine (EE=metabolic equivalents of task (METs) � 3.5 � body weight/200). A cool down period after the last waltz was allowed for all patients. Patients danced during all sessions in couple with a partner experienced in dancing. If the habitual partner was unavailable, 2 patients completed the dancing session together.

A supervised exercise training program was performed at the hospital’s gym, 3 times a week for 8 weeks, as previously described.¹³ Exercise intensity was chosen at 70% of peak V̇o₂. Each session lasted about 1 hour, which began with calisthenics for 10 minutes, followed by 30 minutes of aerobic exercise on an electronically braked cycle ergometer (Ergometrics 800 S) or treadmill (Tunturi, Finland) or both. Heart rate was maintained at target intensity for 30 minutes of cycling and was adjusted during training. Blood pressure was measured at rest, at the end of exercise, and after 5 minutes of recovery. Patients were asked to avoid regular exercise at home during the study period.

Plasma lipids and glucose were measured on a Hitachi 917 biochemical analyzer (Hitachi, Tokyo, Japan) by using conventional enzymatic methods (Boehringer Mannheim, Mannheim, Germany). Malondialdehyde levels in plasma samples were measured by using high-performance liquid chromatography with fluorescence detection. Total plasma lipid hydroperoxide concentrations were measured by using the ferrous oxidation-xylenol orange assay in conjunction with triphenyl phosphine as previously described.¹⁵

After a familiarization test, a symptom-limited cardiopulmonary exercise test was performed on an electronically braked cycle ergometer by using a ramp-pattern increase in work rate. All the exercise tests were performed in the morning at least 3 hours after a light breakfast. Expired gases and volumes were analyzed, breath-by-breath, with a metabolic cart (QUARK PFT, COSMED, Rome). Heart rate and blood pressure were measured every minute during increasing work rate exercise and recovery. A 12-lead ECG was recorded every minute. The exercise test was stopped when one or more of the following criteria were present: fatigue, dyspnea, excessive systemic blood pressure increase (≥230/130 mm Hg), ≥2-mm ST depression in at least 2 adjacent leads, and/or angina. The anaerobic threshold was measured by the V-slope method.¹⁶ Peak oxygen uptake was the average oxygen uptake during the last 15 seconds of exercise.

Echocardiographic studies were performed with the patients supine in the left lateral position. Two-dimensional echo images were continuously acquired from the parasternal long axis, short axis, and apical 4- and 2-chamber views by using an electronic scanner (3.5 MHz, Megas, ESAOTE, Italy). End-diastolic and end-systolic volumes were obtained from the apical view by using a modified single-plane Simpson’s rule from which ejection fraction was calculated. Left ventricular flow velocity waveforms from 5 consecutive cardiac cycles were obtained and averaged by positioning the sample size on the mitral leaflet tips during diastole.

Peak velocity of early filling wave (E), peak velocity of late filling (A), and their ratio were then calculated. The echocardiographic images were evaluated in a blinded manner by 2 independent, experienced observers who adopted the same assessment criteria. Disagreement between the 2 observers occurred in 8% of studies. Differences in interpretation were resolved by a third independent cardiologist.

Patients were evaluated in the morning in a fasting state as previously described.¹⁷ After 5 minutes of relaxation in the supine position, a 7.5-MHz ultrasound probe was positioned over the dominant arm to detect good quality brachial artery images (ESAOTE Challenge, Florence, Italy). Acquisition started after fixation of the probe in a stereotaxic arm to avoid artifacts due to operator movements. Images were taken at baseline for 30 seconds and 90 seconds after cuff release (flow-mediated response) according to recently published recommendations.¹⁷ Flow-mediated dilatation was evaluated after release of a small (pediatric size) sphygmomanometer inflated at 240 mm Hg for 4.5 minutes at the wrist. We considered a ≥7% change in diameter from resting values as a normal response (2 SD of the difference between repeated measurements at our laboratory and other laboratories).^17,18 Images were evaluated by 2 independent experienced operators unaware of the clinical picture and blinded as to each other’s interpretation. Intraobserver and interobserver variabilities were assessed in 250 consecutive subjects with a variety of conditions, and results were in agreement with those of other laboratories (1.2�0.8% and 1.9�0.9%, respectively).

The Minnesota Living With Heart Failure Questionnaire (MHFLQ) was used.¹⁹ This method contains 21 self-administered items with 3 domains: the effects of heart failure on (1) physical, (2) emotional, and (3) socioeconomic factors.

Data were analyzed by SPSS 13.0 (for Windows) statistical package. The distribution was normal for all variables considered. We used one-way ANOVA to compare variables at baseline and changes from baseline to 8 weeks and post hoc analysis with the Dunnett’s test to compare each group to the control group.

Regression analysis was performed to determine the relationship between changes in peak V̇o₂ and changes in E/A ratio, endothelium-dependent relaxation, and quality of life. Univariate analysis was also performed to assess the independent predictors of change in functional capacity after dancing. From previous studies from our laboratory, we calculated a number of patients of 42 per group for a power of 80% and α=0.05 to determine a 15% difference in peak V̇o₂ between study entry and the end of exercise training. Data are expressed as mean�SD, with a significance level P<0.05.

EE was not directly measured. We monitored gas exchange and ventilation with a portable analyzer (K4b2, Cosmed, Rome) in a group of patients participating in the study. In the 15 patients studied (10 men, 5 women; age, 60�9 years; weight, 77�17 kg), mean V̇o₂ was 14.3 mL/(kg�min), corresponding to 4.1 METs. In this group, EE was calculated as 5.4 kcal/min or 113 kcal for 21 minutes of dancing exercise. A similar result is obtained by multiplying oxygen uptake by 5 kcal.

After 8 weeks, peak V̇o₂, anaerobic threshold, V̇e/V̇co₂ slope, and V̇o₂/W slope were all similarly improved in both groups E and D (16% and 19%, 20% and 21%, 14% and 15%, 18% and 19%, respectively; P not significant for all comparisons, P<0.001 versus controls). Oxygen pulse, a good approximation of stroke volume, was also similarly improved in the 2 exercise groups (Table 2). We analyzed the relationship between baseline values for the measured variables and the percent change in peak V̇o₂ after dancing. The strongest independent predictors of peak V̇o₂ change were the initial peak V̇o₂ (r=−0.61), E/A ratio (r=−0.54), and age (r=−0.32) (P<0.001 for all). Gender did not enter in the model.

As shown in Figure 2, the endothelium-dependent relaxation was improved in the 2 groups (group E, from 2.6�1.3% to 5.2�1.5%, P<0.001 versus control; group D, from 2.2�1.4% to 5.0�1.5%, P<0.001 versus control). However, the improvement was not significantly different between dancing and exercise groups (P=0.51).

Ejection fraction was not significantly changed in both E and D groups (Table 3). However, mitral inflow diastolic filling pattern was changed as a result of changes in the E/A ratio in both E and D groups (P<0.05 versus controls; P=0.44, E versus D).

Total cholesterol was not significantly reduced in both dancing and exercise training groups (group E: from 198�25 mg/dL to 195�24 mg/dL; group D: from 191�24 to 186�24; P=0.43). Low-density lipoprotein cholesterol had no significant changes from baseline in the 3 groups. However, high-density lipoprotein cholesterol and triglycerides were both improved at the end of the protocol in both E and D groups (high-density lipoprotein cholesterol, group E: from 33�10 mg/dL to 37�11 mg/dL, group D: from 35�9 to 39.5�8; P<0.02 versus controls; triglycerides, group E: from 182�21 mg/dL to 165�22 mg/dL, group D: from 175�23 to 144�22; P<0.01 versus controls). Fasting blood glucose was also reduced in the 2 exercise groups (group E: from 101�13 g/dL to 91�12 g/dL; group D: from 98�12 to 88�11; P<0.03 versus control).

In dancing patients, there was a significant decrease in both plasma malondialdehyde (from 3.75�0.8 μmol/L to 2.36�0.8 μmol/L) and lipid hydroperoxide levels (from 3.88�0.9 μmol/L to 2.21�0.9 μmol/L) as compared with controls (P<0.001 for both). Similar changes were observed in exercise training patients (from 3.81�0.8 μmol/L to 2.44�0.8 μmol/L and from 3.96�1.0 μmol/L to 2.45�0.9 μmol/L, respectively, P<0.001 versus controls).

As shown in Figure 3, the MHFLQ sum score improved significantly after both dancing and exercise training (P<0.001 versus controls). However, the improvement after dancing was not different from the improvement after aerobic exercise (P=0.40). The 3 domains changed are as follows: physical (group D: from 35�8 to 27�9; group E: from 33�7 to 25�8; controls from 35�9 to 33�9), emotional (group D: from 10�5 to 4�4; group E: from 11�4 to 8�4; controls from 10�4 to 9�4), and socioeconomic (group D: from 10�6 to 7�6; group E: from 9�6 to 7�5; controls from 10�6 to 10�6). The adherence to the protocol, calculated as the percentage of training sessions attended, was higher in group D than in group E (91�8% versus 77�8%, P<0.02).

Changes in peak V̇o₂ were correlated with changes in the E/A ratio (r=−0.58, P<0.001), changes in the endothelium-dependent relaxation (r=0.64, P<0.001) (Figure 4), and changes in quality of life sum score (r=0.68, P<0.001). Correlation coefficients were not different between dancing and aerobic exercise.

During both exercise training and dancing, there were minor untoward events. There were no significant differences in systolic blood pressure at rest and recovery in both groups. Sporadic ventricular premature contractions were observed in 6 patients during the exercise training sessions and 8 patients during dancing. Other mild events were dizziness (1E, 2D patients) and hypotension during recovery (2D patients).

This is the first demonstration that a dance protocol improves functional capacity in humans with CHF. Peak V̇o₂ increased by 19% in the dancing group and 16% in the exercise training group (P=0.69). The improvement in peak V̇o₂ after dancing was predicted by the initial peak V̇o₂, as well as E/A ratio and age. Gender did not enter in the model. The amount of improvement in peak V̇o₂ was similar to that of a recent meta-analysis in 800 patients with similar characteristics (17%; range, 12% to 31%).²⁰ In the present study, in both groups, the increased functional capacity was accompanied by significant improvements in ventilation and cardiocirculatory efficiency, as demonstrated by changes in the V̇e/V̇co₂ slope, V̇o₂/W slope, and O₂ pulse, respectively (Table 2). The average heart rate during dance was similar to that recorded during traditional aerobic exercise, but the duration of the waltz protocol was 43% shorter (21 minutes versus 30 minutes).

The profile of heart rate responses during dancing resembles that observed during interval training exercise. On-off exercise may correspond to the sequence slow-fast waltz dance we used (Figure 1).

As recently shown, intermittent exercise determined more marked improvements in functional capacity and endothelium-dependent brachial artery relaxation than did continuous aerobic exercise in a group of CHF patients clinically similar to those studied in the present investigation. As shown by Wisloff et al,²¹ interval training rather than continuous exercise was associated with a more marked increase in total antioxidant status, and this improvement was correlated with improved endothelium-dependent vasorelaxation (r=0.67). Although pre-post changes in peak V̇o₂ and endothelium-dependent dilation observed in this study were in agreement with the results described by Wisloff et al, we found similar decreases in plasma malondialdehyde and lipid hydroperoxide, as compared with initial values, in dancing and exercise training groups. This parallel reduction in free radicals was observed, despite different exercise protocols.

Dancing is a form of exercise that induces favorable effects similar to aerobic exercise training. Both central and peripheral adaptations have been demonstrated in humans with CHF after moderate-intensity aerobic exercise. Exercise training enhances cardiac output at peak workload, improves myocardial perfusion and function, increases the size and volume density of mitochondria and skeletal muscle oxidative enzymes as well, reduces endothelial dysfunction, and improves autonomic balance.^7–12 The mechanisms through which dancing improves functional capacity and endothelial dysfunction have not been demonstrated in the past. From the result of the present investigation, we may hypothesize similar adaptations as aerobic exercise training. In fact, correlation coefficients of the relationship between changes in peak V̇o₂ versus E/A ratio and endothelium-dependent dilation were not different between dancing and aerobic exercise training groups. On the basis of the results of the present investigation, 2 factors seem to play a role in improving functional capacity. One is left ventricle diastolic filling, and the other is the endothelium-dependent vasorelaxation. Previously, our group demonstrated that aerobic exercise improves left ventricle diastolic filling in patients with CHF, which was correlated with peak V̇o₂ increase.²² The results of the present study are in agreement with previous evidence showing that the increase in peak V̇o₂ is related to improvements in the endothelium-dependent vasorelaxation.²³ However, we did not observe any difference between dance and aerobic exercise. At 8 weeks, an increase in high-density lipoprotein cholesterol was observed in both groups and may explain, at least in part, the enhanced endothelium-dependent vasomotor response. The reduction in plasma triglycerides may contribute to this effect, together with a decrease in malondialdehyde and lipid hydroperoxide.

Quality of life was improved in both dance and exercise groups (Figure 3). However, patients involved in the dance protocol, rather than patients randomized to traditional aerobic exercise training, had a more marked improvement in emotional dimension measures of MHFLQ (P=0.04). Moreover, the adherence to the protocol, calculated as the percentage of training sessions attended, was higher in group D than in group E (P<0.02). Evidently, people involved in courses of dance are generally well motivated to continue the program until the end, with a low rate of withdrawal. Dance is a form of exercise in which movement, social interaction, and fun are mixed together. This combination may explain a better compliance with dancing than with traditional exercise training as we observed in the present study.²⁴

Dance therapy is offered as a health promotion service for healthy people and as a complementary tool to reduce distress in patients with cancer.²⁵ Dancing improves mobility and muscle coordination and also reduces muscle tension. More important, dance is reported to improve self-awareness and self-confidence and to facilitate interpersonal sharing of feelings. Not only dance, but also aerobic exercise, increases personal well-being and decreases anxiety and depression.²⁶

The study was not blinded, neither to the investigators nor to the patients. Dance may not be suitable for all individuals. We tested only 1 dance protocol combining fast and slow waltz. We do not know whether other forms of dance may have the same therapeutic effect. We chose waltz because it represents the most popular dance in the western countries and its choice could guarantee the generalizability of the results obtained. However, the new generation of patients is less experienced with classical dances. This could negatively influence the willingness to participate. The E/A ratio appears to be less accurate than tissue Doppler annular velocity (Ea) to measure ventricular relaxation, because it is more sensitive to alterations in preload. However, changes in the E/A ratio was correlated with pre-post training change in peak V̇o₂ in both dance and exercise groups, confirming previous observations in patients with ischemic cardiomyopathy.²² We did not monitor heart rate during calisthenics. However, because heart rate was monitored just after the end of them, the initial heart rate monitored should presumably reflect the heart rate during calisthenics. If this is true, heart rate during calisthenics should be between 85 and 95 bpm, corresponding to 50% to 55% of peak V̇o₂ or 60% to 65% of peak heart rate. Adherence in the dancing group may be biased by the presence of an experienced partner. Conversely, this adherence is limited to the 8 weeks of training. Finally, we did not measure EE. However, we may derive it from the basis of V̇o₂ measured with a portable analyzer in a small group of participants in the study (see Methods).

The results of the present investigation demonstrate that waltz dancing is safe and improves functional capacity and endothelial dysfunction in patients with CHF. Both improvements were similar to those observed after a supervised moderate exercise training program, despite a 43% shorter protocol. The enhanced functional capacity was correlated with quality of life in both exercise groups. However, patients involved in the dance protocol had a more marked improvement in emotional dimension measures of MHFLQ than did patients randomized to traditional aerobic exercise training (P=0.04). Moreover, the adherence to the protocol was higher than that in traditional exercise training. On the basis of the results of the present investigation, waltz dancing may be used in combination with traditional aerobic exercise or as an alternative to it in patients whose adherence to traditional programs is low and/or in those who prefer dancing to other forms of exercise. More studies are needed to confirm these preliminary results, and other forms of dance should also be tested in cardiac rehabilitation programs.