Effects of exercise training on depression in patients with heart failure: a systematic review and meta-analysis of randomized controlled trials

Authors


Abstract

Aims

The aim of this review is to assess the effects of exercise training on the symptoms of depression in heart failure (HF) patients.

Methods and results

Randomized controlled trials of exercise training in patients with HF and symptoms of depression were identified. The depression data were pooled using meta-analysis, and 19 studies were identified, with a total of 3447 patients, of which 16 (3226 patients) provided data for the meta-analysis. Exercise training demonstrated significant reductions in the symptoms of depression [standardized mean difference (SMD) –0.38, 95% confidence interval (CI) –0.55 to −0.21], and its antidepressive effect was consistent in a number of HF groups, such as in ages under and over 65 years (SMD −0.14, 95% CI −0.22 to −0.07 vs. SMD −0.44, 95% CI −0.61 to −0.27) and EFs of <50% (SMD −0.38, 95% CI −0.56 to −0.20), as well as in a range of interventional strategies, including the aerobic mode (SMD −0.40, 95% CI −0.61 to −0.19), centre, home, or combined setting (SMD −0.61, 95% CI −0.95 to −0.27 vs. SMD −0.25, 95% CI −0.44 to −0.07 vs. SMD −0.13, 95% CI −0.21 to −0.05), and short or longer training programmes (≤12 weeks, SMD −0.50, 95% CI −0.73 to −0.27; 12–26 weeks, SMD −0.47, 95% CI −0.82 to −0.11; >26 weeks, SMD −0.12, 95% CI −0.20 to −0.04). The beneficial effects were preserved when blind design trials were considered (SMD −0.14, 95% CI −0.22 to −0.07).

Conclusion

Exercise training significantly decreased the symptoms of depression in patients with HF. This benefit remained unclear in cases of HF with a normal EF and combined aerobic and strength training. Random controlled trials are needed to verify the benefit of exercise in these populations, and in very old, asymptomatic, and severe HF patients.

Introduction

Symptoms of depression are a common co-morbidity, affecting as many as 42% of patients with heart failure (HF), and are associated with a poor quality of life and adverse prognosis.[1-3] Symptoms of depression have negative impacts not only on daily social and domestic activities, but also on hospitalizations and mortality rates in HF patients.[4] Thus, a reduction in the patient's depressive symptoms has become of prime importance for therapeutic intervention. Despite recent advances in pharmacological management, however, treating the symptoms of depression is especially challenging in HF patients because they are often on complex medications and have multiple co-morbidities.[1] Non-pharmacological interventions, such as exercise training, have been demonstrated as potential treatments for depression and may be comparable with antidepressant therapy,[5-7] but these effects are not well identified in patients with HF.

Theoretically, exercise may provide important advantages over pharmacotherapy, including fewer drug interactions and more involvement of patients in their self-care. Some studies,[8-17] including the largest and recently completed Heart Failure—A Controlled Trial Investigating Outcomes of Exercise Training (HF ACTION),[16] indicated that exercise resulted in a modest reduction in depressive symptoms. Several other studies reported no benefits with exercise training on depressive symptoms in HF patients.[18-28] Thus, uncertainty remains with regard to the effects of exercise training on depressive symptoms in patients with HF.

To our knowledge, randomized controlled trials (RCTs) of the effects of exercise training on depressive symptoms among patients with HF have not been systematically reviewed. The aim of this meta-analysis is to determine the effect of exercise training on symptoms of depression in HF patients using various aspects, such as age, LVEF, type of exercise, length of the intervention, and exercise setting.

Methods

Search strategy

Electronic searches of MEDLINE, EMBASE, PUBMED, and the Cochrane Central Register of Controlled Trials (CENTRAL) in The Cochrane Library, from 1966 to August 2013, using the search terms ‘heart failure, heart failure with preserved ejection fraction, cardiac failure, left ventricular failure, ischemic cardiomyopathy, exercise, training, and cardiac rehabilitation’, were performed to identify RCTs. Hand searches of bibliographies from published meta-analyses, review articles, and proceedings booklets from conferences were also conducted to ensure the inclusion of all pertinent studies for the preliminary review. The searches were limited to studies published in English.

Eligibility criteria

Studies were considered to be eligible if they were (i) RCTs; (ii) included patients (>18 years old) with either systolic HF or HF with a preserved EF of any aetiology in the control and in the intervention group (diagnosis based on LVEF or clinical findings); (iii) if patients received exercise training either alone or as part of a comprehensive cardiac rehabilitation programme (i.e. included components such as psychological intervention or health education); (iv) compared with a standard medical treatment or education placebo control group; and (v) reported the effect of exercise training on depression or depressive symptoms. In addition, if several articles reported one study but with different follow-up times or outcome results, we adopted the one that provided more complete and updated data; however, in these cases the previous study was used as an important and necessary information supplement.

Data extraction and quality assessment

Two investigators (R.-H.T. and Z.-D.B.) independently extracted data from each study. We used a standardized extraction form to evaluate the study quality, including the study design, population characteristics (age, NYHA class, LVEF, and aetiology of HF), intervention (mode, frequency, intensity, duration per session, and length of the programme) and control, and depression data (mean, standard difference between intervention and control group). The risk of bias was assessed by consideration of: assessments of randomization methods, blinding of allocation to treatment, baseline balance of groups, loss to follow-up, intention-to-treat analysis, and complete outcome reported. The data were checked for consistency between the two investigators, and disagreements were discussed with all seven investigators. If the seven investigators could not resolve the discrepancies, consensus with Z.-Y.Z. was required. When necessary, the original investigators were contacted to clarify data or provide additional data.

Statistical analysis

The statistical analyses were performed using RevMan, version 5.0,[29] and STATA, version 11.0.[30] Data were analysed using the change from baseline data for both the intervention and control groups. When the standard deviation for change was not provided, it was calculated by using the within-patient correlation coefficient.[31] If depression measures used the same scale, the results were combined as a weighted mean difference (WMD); otherwise, a standardized mean difference (SMD) was presented. In studies reporting more than one depression outcome, to prevent double counting in the meta-analysis, we chose one at random. A fixed effects model was used to pool data from each trial, and 95% confidence intervals (CIs) were used for safety analysis.

Statistical heterogeneity was quantified using the χ2 test, and, when heterogeneity was present, a random effects model was calculated and a sensitivity analysis was performed.[32] Subgroup analyses were undertaken to investigate the role of the various study characteristics on the observed effect. Publication bias was assessed visually using the funnel plot (i.e. scatter plots of the mean intervention effect vs. the inverse of variance of the intervention effect for each study) and quantitatively using the Begg adjusted-rank correlation test.[33] We defined publication bias as the tendency for positive trials to be published and the tendency for negative trials not to be published.

Results

Study selection

A total of 2496 relevant references from all databases were found. After title and abstract screenings, 55 full papers were retrieved for detailed evaluation, and, of these, 34 were excluded. The reasons for exclusion were: outcomes not relevant to this review (i.e. quality of life, n = 20), non-randomized design (n = 3), review article (n = 4), pre–post comparison study (n = 2), articles used the same or some of the same patients (n = 1), inappropriate study populations (n = 3), and study protocol (n = 1). Therefore, 21 papers were included reporting the results of 19 distinct studies. The study identification and selection process is summarized in Figure 1.

Figure 1.

Study flow. RCT, randomized controlled trial.

Design and patient characteristics

The 19 trials included a total of 3447 HF patients.[8-28] The baseline patient characteristics and design of the trials are described in Table 1. The majority of the subjects were male and in NYHA classes II and III, and most trials were restricted to patients with LVEFs of <40%, while three trials included patients with LVEFs of >45%.[10-12, 25] The median age of the populations ranged between 54 and 81 years. Of the trials, 5 examined the effects of combined aerobic and strength training,[13, 17, 19, 20, 28] 1 examined strength training (respiratory muscle training) alone,[9] and the other 13 examined aerobic training.[8, 10-12, 14-16, 18, 21-27] Exercise training programmes ranged widely across the studies: exercise mode, walking, bicycle, treadmill, games, jogging, calisthenics, Tai Chi Chuan, and strength training; frequency, 2–7 days/week; duration, 20–60 min per session; intensity, 40–80% of maximum heart rate, 60–70% of maximal heart rate reserve, or maximum oxygen uptake; and overall length of the programme, 6 weeks to 1.5 years. The exercises were centre-based in nine trials,[8, 12-15, 18, 23-25] entirely home-based in five trials,[10, 11, 17, 19-21] and initially within-centre and then at home in the remainder.[9, 16, 22, 26-28] The controls received the usual care, including medication and education advice, but no formal exercise training.

Table 1. Description of included studies
StudyDetails of populationTraining programmeDepression assessmentJadad scale (1–7)
 Group ET/C (n)Age (years)Males (%)LVEF (%)NYHA classModeFrequency (days/week)IntensityDuration (min)Length (weeks)Exercise setting  
  1. BDI, Beck Depression Inventory; C, control; CBA-H, Cognitive Behavioural Assessment Hospital form; CESD, Center for Epidemiologic Studies Depression Scale; ET, exercise training; GDS, Geriatric Depression Scale; HADS, the Hospital Anxiety and Depression Scale; HAM-D, Hamilton Rating Scale for Depression; HDCDS, Hare-Davis Cardiac Depression Scale; HRR, heart rate reserve; IMT, inspiratory muscle training; MAAC, Multiple Affect Adjective Checklist; MADRS, Montgomery Asberg Depression Rating Scale; MHR, maximal heart rate; MIP, maximal inspiratory pressure; NR, not reported; PGWB, Psychological General Well-being Index; PMS, the Profile of Mood States; RHR, resting heart rate; 1RM, one repetition maximum; RPE, rating of perceived exertion; SCL, Symptom Checklist 90; THR, target heart rate; VO2 peak, peak oxygen uptake.
Barrow et al.[8]32/3368.281.5NRII/IIITai Chi Chuan25516CentreSCL4
Berg-Emons et al.[18]18/1658.673.525.6II/IIICycle, walk, games2RHR + 60% of HRR6012CentreHADS2
Blumenthal et al.[16]1158/11645971.525II–IVTreadmill, walk, cycle3–560–70% of HRR3052Centre + homeBDI-II6
Bosnak-Guclu et al.[9]18/1867.68034.8II/IIIIMT740% of MIP306Centre + homeMADRS6
Dracup et al.[19]86/875471.726.4II–IVWalk, resistance training760% of MHR, 80% of 1RM4552HomeMAAC2
Gary et al.[10, 11]16/1668055.5II/IIIWalk360% of THR3012HomeGDS2
Gary et al.[20]12/12605025II/IIIWalk, resistance training5–6RHR +70% of HRR, RPE (12–15)60–9026HomeBDI2
Gary et al.[21] (A)20/1765.841.9≥15II/IIIWalk3RPE (<15)6012HomeHAM-D4
Gary et al.[21] (B)18/1965.841.9≥15II/IIIWalk3RPE (<15)6012HomeHAM-D4
Giannuzzi et al.[22]40/4053.59534I/IICycle, walk3–780% of MHR6026Centre + homeCBA-H2
Gottlieb et al.[26]17/1665.566.723.5II/IIICycle, treadmill3Borg Scale (12–13)4526CentreCESD2
Jolly et al.[17]84/856874.6≤40I–IIIWalk, muscle strengthening570% of VO2 peak, Borg Scale (12–13)20–3024HomeHADS4
Kitzman et al.[12]26/2769.513≥50II/IIIWalk360–70% of HRR6016CentreCESD4
Klocek et al.[24] (A)14/145410033.4II/IIICycle360–75% of MHR2524CentrePGWB4
Klocek et al.[24] (B)14/1455.910033.7II/IIICycle360–75% of MHR2524CentrePGWB4
Koukouvou et al.[13]18/1152.6100<40II/IIICycle, jog, walk, resistance training3–470% of VO2 peak, RPE (12–14)6026CentreBDI2
Kulcu et al.[14]27/2659.472.6NRII/IIITreadmill360–70% of MHR608CentreBDI4
Smart et al.[25]16/1464.55257.8I–IICycle360–70% of VO2 peak3016CentreHDCDS6
Witham et al.[26, 27]41/4180.555NRII/IIISeated exercise2Borg Scale (11–13)2076Centre + homeHADS7
Witham et al.[28]53/548067.3NRII/IIIElasticized bands, walk2Borg Scale (11–13)6024Centre + homeHADS7
Yeh et al.[15]50/5067.46429I–IIITai Chi Chuan26012CentrePMS7

Risk of bias assessment

Half of the included trials performed intention-to-treat analyses and blinding of the outcome assessments, and three trials described the concealment of random allocation (see Table 2 and Supplementary material online, Table S1). Losses to follow-up varied across the study assessments, from 0 to 20.7%. The trials tended to be assessed as being of moderate quality.

Table 2. Risk of bias assessment (I)
StudyAdequate sequence generationAllocation concealmentOutcome blindingLoss to follow-upIntention-to-treat analysisComplete outcome reportedGroups balanced at baseline
  1. NR, not reported.
  2. aNot stated by authors but implicit from the study report. er for Epidemiologic Studie
Barrow et al.[8]YesNRNR13/65 (20%)NRYesYes
Berg-Emons et al.[18]NRNRNR0/34YesaYesNo
Blumenthal et al.[16]YesNRYes584/2322 (15%)YesYesYes
Bosnak-Guclu et al.[9]YesNRYes6/36 (7%)NRYesYes
Dracup et al.[19]NRNRNR0/173YesaYesNo
Gary et al.[10, 11]NRNRNR5/32 (15%)NRYesYes
Gary et al.[20]NRNRNR0/24YesaYesNo
Gary et al.[21] (A)NRNRYes6/37 (16%)NoYesYes
Gary et al.[21] (B)NRNRYes6/37 (16%)NoYesYes
Giannuzzi et al.[22]NRNRNR3/40 (7.5%)NoaYesYes
Gottlieb et al.[23]NRNRNR8/33 (7.5%)NRNoYes
Jolly et al.[17]YesNRNo12/169 (7.2%)YesYesYes
Kitzman et al.[12]NRNRYes7/53 (13.2%)YesaYesYes
Klocek et al.[24]NRNRYesNRNRNoNo
Koukouvou et al.[13]NRNRNR3/29 (10.3%)NoaYesYes
Kulcu et al.[14]YesNRNR11/53 (20.7%)NoaYesYes
Smart et al.[25]YesNRYes5/30 (16.7%)YesYesYes
Witham et al.[26, 27]YesYesYes14/82 (17%)NRYesNo
Witham et al.[28]YesYesYes20/107 (18.6%)YesNoNo
Yeh et al.[15]YesYesYes4/100 (4%)YesYesYes

Outcomes

The outcomes for symptoms of depression are summarized in Table 3.

Table 3. Meta-analysis results
Depression outcome, total or subgroupStudies (n)Patients (n)Statistical methodEffect estimate, mean (95% CI) and P-valueStatistical heterogeneity, I2 and P-value
  1. CI, confidence interval; SMD, standardized mean difference; WMD, weighted mean difference.
Total163266SMD (random effects)−0.38 (−0.55 to −0.21)<0.0000154%, 0.004
Total (remove Barrow et al.)153214SMD (fixed effects)−0.17 (−0.24 to −0.10)<0.0000127%, 0.15
Ejection fraction      
<50%133155SMD (random effects)−0.38 (−0.56 to −0.20)<0.000158%, 0.003
<50% (remove Barrow et al.)123103SMD (fixed effects)−0.17 (−0.24 to −0.10)<0.0000129%, 0.15
>50%3111SMD (fixed effects)−0.39 (−0.82 to 0.03)0.0719%, 0.29
Age      
<65 years82730SMD (fixed effects)−0.14 (−0.22 to −0.07)0.000241%, 0.11
>65 years8536SMD (fixed effects)−0.44 (−0.61 to −0.27)<0.0000140%, 0.1
Length of exercise intervention      
≤12 weeks6298SMD (fixed effects)−0.50 (−0.73 to −0.27)<0.00010, 0.79
12–26 weeks7431SMD (random effects)−0.47 (−0.82 to −0.11)0.0164%, 0.01
12–26 weeks (remove Barrow et al)6279SMD (fixed effects)−0.30 (−0.50 to −0.09)0.00425%, 0.25
>26 weeks32537SMD (fixed effects)−0.12 (−0.20 to −0.04)0.0030, 0.94
Exercise mode      
Aerobic training112844SMD (random effects)−0.40 (−0.61 to −0.19)0.000255%, 0.01
Aerobic training (remove Barrow et al.)102792SMD (fixed effects)−0.16 (−0.23 to −0.08)<0.00012%, 0.42
Aerobic + strength4392SMD (random effects)−0.31 (−0.67 to 0.06)0.158%, 0.07
Aerobic + strength (remove Koukouvou et al.)3366SMD (fixed effects)−0.19 (−0.40 to 0.01)0.070%, 0.59
Strength training130WMD (fixed effects)−4.50 (−8.48 to −0.52)0.03Not applicable
Exercise setting      
Centre-based7339SMD (random effects)−0.61 (−0.95 to −0.27)0.000454%, 0.04
Centre-based (remove Barrow et al.)6287SMD (fixed effects)−0.44 (−0.68 to −0.21)0.000213%, 0.33
Home-based5456SMD (fixed effects)−0.25 (−0.44 to −0.07)0.0080, 0.48
Centre + home42471SMD (fixed effects)−0.13 (−0.21 to −0.05)0.00116%, 0.31
Blind      
Yes72639SMD (fixed effects)−0.14 (−0.22 to −0.07)0.00030%, 0.61
No9627SMD (random effects)−0.52 (−0.81 to −0.23)0.000462%, 0.007
No (remove Barrow et al.)8575SMD (fixed effects)−0.33 (−0.49 to −0.16)0.000138%, 0.13

Exercise training and depression

Because of the diversity of the dimensions, we pooled the symptoms of depression data by SMD (see Table 1). In addition, there were three studies which could not be incorporated into the meta-analysis because of the incomplete data presented.[23, 24, 28] The study by Gottlieb et al. included 33 elderly patients (mean age 65.5 ± 8.9), with moderate to severe HF, undertaking 6-month supervised and graded aerobic training (n = 17) or usual care (n = 16).[23] The depression values were only provided at the 6-month follow-up in the exercise group (9 ± 9 to 8 ± 8) but not the controls, and there was no significant change based on the CES-D (Center for Epidemiological Studies-Depression Scale) questionnaire.

In the study of Klocek et al., 42 men (mean age 55.9 ± 8.1) with ischaemic HF (NYHA classes II and III) were randomized into three groups: group A with a constant workload (n = 14), group B with a progressive workload (n = 14), and group C who were not trained (n = 14).[24] At 6 months, an improvement in depressed mood was observed in groups A (10.1 to 14.4 vs. 10.1 to 9.6, P < 0.01) and B (9.7 to 15.7 vs. 10.1 to 9.6, P < 0.01); however, no standard difference was reported. The study of Witham et al. was a parallel-group, single-blind, RCT,[28] where 107 HF patients (mean age 80 ± 5.4) were randomized to either 24 weeks of aerobic exercise plus strength training (n = 53) or usual care (n = 54). The symptoms of depression were not provided separately between the exercise group and the control group, and did not decrease when compared with those of the control group at 8 weeks (−0.2, 95% CI −1.3 to 0.8, P = 0.64) or at 24 weeks (−0.9, 95% CI −2.1 to 0.2, P = 0.11) based on the HADS (Hospital Anxiety and Depression Scale).

Pooling across 16 trials (3266 patients), regardless of the depression measure used, provided strong evidence of a decrease in the symptoms of depression with exercise (random effects SMD −0.38, 95% CI −0.55 to −0.21, P < 0.00001; Figure 2), and this analysis demonstrated substantial heterogeneity (I2 = 54%, P = 0.004). The sensitivity analysis showed that the study by Barrow et al. was separate from the others;[8] therefore, we removed this study and re-evaluated the meta-analysis, and the statistical heterogeneity disappeared (I2 = 27%, P = 0.17; Table 3). This change did not modify the results from the meta-analysis, which continued to favour the exercise training group.

Figure 2.

Meta-analysis: depression in heart failure patients.

Subgroup analysis (left ventricular ejection fraction)

Three trials did not report mean LVEF data, and we calculated that they were restricted to patients with LVEFs of <50%, based on their inclusion criteria.[8, 14, 26] Thirteen trials testing LVEFs of <50% demonstrated relatively consistent benefits in depressive symptoms (3155 patients, random effects SMD −0.38, 95% CI −0.56 to −0.20, P < 0.0001; heterogeneity I2 = 58%, P = 0.003). In this meta-analysis, there was statistical heterogeneity; however, three trials testing LVEFs of >50% showed a trend for significance (111 patients, fixed effects SMD −0.39, 95% CI: −0.82 to 0.03, P = 0.07; heterogeneity I2 = 19%, P = 0.29).

Subgroup analysis (age)

The positive effect of exercise training on the symptoms of depression was observed in both relatively young patients (mean age <65 years) experiencing HF (eight trials, 2730 patients, fixed effects SMD −0.14, 95% CI −0.22 to −0.07, P = 0.0002; heterogeneity I2 = 41%, P = 0.11) and in the elderly populations (mean age >65 years, eight trials, 536 patients, fixed effects SMD −0.44, 95% CI −0.61 to −0.27, P < 0.00001; heterogeneity I2 = 40%, P = 0.1).

Subgroup analysis (duration of exercise intervention)

When limited to a 12-week intervention, the exercise training demonstrated significant reductions in depressive symptoms (six trials, 298 patients, fixed effects SMD −0.50, 95% CI −0.73 to −0.26, P < 0.0001; heterogeneity I2 = 0%, P = 0.79). The positive effect lasted from 12 to 26 weeks after the programme (seven trials, 431 patients, random effects SMD −0.47, 95% CI −0.82 to −0.11, P = 0.01; heterogeneity I2 = 64%, P = 0.01), and 26 weeks after the programme (three trials, 2537 patients, fixed effects SMD −0.12, 95% CI −0.20 to −0.04, P = 0.003; heterogeneity I2 = 0, P = 0.94).

Subgroup analysis (exercise mode)

Eleven of 16 trials examined the effects of aerobic training on HF, and, when pooling the data together, a significant reduction in the symptoms of depression was observed (2844 patients, random effects SMD −0.40, 95% CI −0.61 to −0.19, P = 0.0002; heterogeneity I2 = 55%, P = 0.01). However, four trials examining the effects of combined aerobic and strength training were inconclusive (392 patients, random effects SMD −0.31, 95% CI −0.67 to 0.06, P = 0.1; heterogeneity I2 = 58%, P = 0.07). The only trial examining strength training alone was found to be conclusive (WMD −4.50, 95% CI −8.48 to −0.52, P = 0.03).

Subgroup analysis (exercise setting)

There was a significant reduction in the symptoms of depression in the HF patients, no matter which exercise training was undertaken, in the centre setting (seven trials, 339 patients, random effects SMD −0.61, 95% CI −0.95 to −0.27, P = 0.0004; heterogeneity I2 = 54%, P = 0.04), home setting (five trials, 456 patients, fixed effects SMD −0.25, 95% CI −0.44 to −0.07, P = 0.008; heterogeneity I2 = 0, P = 0.48), or initially within the centre and then at home (four trials, 2471 patients, fixed effects SMD −0.13, 95% CI −0.21 to −0.05, P = 0.001; heterogeneity I2 = 16%, P = 0.31).

Subgroup analysis (blind or not)

Blinding is an important design feature, given that exercise and the outcomes of symptoms of depression have a significant degree of subjectivity and susceptibility to observer bias. For the blinded design, seven trials were included in the meta-analysis, showing that exercise training had a marked effect on the reduction in symptoms of depression (2639 patients, fixed effects SMD −0.14, 95% CI −0.22 to −0.17, P = 0.0003; heterogeneity I2 = 0, P = 0.61). A similar effect was observed with exercise training in the non-blinded trials (nine trials, 627 patients, random effects SMD −0.52, 95% CI −0.81 to −0.23, P = 0.0004; heterogeneity I2 = 62%, P = 0.007).

Publication bias

There was evidence of publication bias on the visual assessment of the funnel plot for the symptoms of depression, as it demonstrated asymmetry (see Supplementary material online, Figure S1). However, this finding was not supported on the Begg adjusted-rank correlation test (P = 0.202), based on comparisons between the exercise and control groups, using P ≤ 0.1 as the level of significance.

Discussion

Over 3000 HF patients with depression or depressive symptoms were evaluated in this meta-analysis. The results showed that exercise training was associated with a significant reduction in depressive symptoms, and its antidepressive effect was not influenced by age, duration of the exercise intervention, or exercise setting, but rather by LVEF and the exercise mode; HF patients with LVEFs of <50%, as well as aerobic exercise intervention, demonstrated consistent benefits on depressive symptoms. The beneficial effects were preserved when blind design trials were considered.

To our knowledge, this is the first meta-analysis to evaluate the effects of exercise training on symptoms of depression in HF patients. Most previous systematic reviews of exercise training for HF have focused on the impact of exercise-based interventions on the mortality, hospitalization, LV remodelling, health-related quality of life, and peak oxygen consumption.[34, 35] Several recent systematic reviews have demonstrated the benefits of exercise in the treatment of symptoms of depression in the general population, and in patients with a chronic illness.[5, 7]

The mechanism whereby exercise attenuates symptoms of depression is probably multifactorial. Social contact may be an important mechanism. The depressed patient who exercises may, as a result, obtain positive feedback from another person, which boosts their sense of self-worth and self-esteem. Physiological effects may be another important mechanism, in that depression has been linked to heightened sympathetic nervous system activity, decreased heart rate variability, increased inflammation, hypercoagulable blood, and endothelial dysfunction, each of which has been linked to adverse clinical outcomes in patients with HF.[36] Moreover, all of the above factors improved significantly after exercise.[37] Most importantly, exercise offers a mechanism of improving survival in patients with HF, primarily by decreasing the adverse consequences of depression.[38]

Our evidence of the benefits of exercise training in HF with symptoms of depression was from patients with LVEFs of <40%, in NYHA classes II or III. While symptoms of depression in HF with preserved LV function occur frequently, especially in elderly patients, only three trials were available regarding the effect of an exercise programme in this population, and the benefit was less conclusive. Indeed, the characteristics of patients with reduced LVEF were different from those with normal LVEF in our analysis, with the latter likely to be older and predominantly female. These differences might partially explain the limited benefits experienced by HF patients with normal LVEFs.[25] Therefore, multicentre, prospective RCTs are needed to verify the positive effects of exercise training in this population.

Although the efficiency of exercise-based cardiac rehabilitation has been confirmed predominantly in young patients (<65 years), exercise training has also been proven to be beneficial in elderly patients (>65 years) in this meta-analysis. Even if exercise capacity in elderly patients was more frequently limited by a non-cardiovascular cause, such as lung diseases, neurological disorders, musculoskeletal abnormalities, or orthopaedic problems, the benefit was similar to that found in younger HF patients.[39]

Our analysis did not focus on the exercise training effects among patients with a clinical diagnosis of depression. Nonetheless, 5 of 19 effects (26.3%) came from symptom scores high enough to indicate a clinical elevation. Although depression was not diagnosed, it seemed that patients with depressive disorders were included.

In the present study, the efficacy of exercise in the treatment of symptoms of depression was influenced by the exercise mode. An unexpected finding was that the favourable antidepressive effect of aerobic exercise was not verified when this mode of exercise was combined with strength training. Indeed, the type of exercise training might determine the level of benefit. Blumenthal et al. confirmed that aerobic training was as effective as antidepressant medication in reducing depression.[16] Current guidelines for exercise training in patients with HF acknowledge the benefits of aerobic training;[6] however, the beneficial effects of strength training in the treatment of the symptoms of depression are still being challenged.[40]

In this study, the funnel plot for the symptoms of depression outcomes did demonstrate asymmetry, which suggests the existence of publication bias, although other alternative explanations should also be explored. A funnel plot is subjective, and the same graph may be interpreted differently by different observers. Therefore, we formally evaluated publication bias by using Begg's statistical method, which quantifies the potential presence of publication bias. However, it was inconsistent with the funnel plot, and, since the impact of publication bias cannot be ruled out, the results should be interpreted with caution.

Limitations

A number of limitations in the present analysis should be noted. First, our analysis of the symptoms of depression data was restricted to an SMD as numerous depression scales were used in the included trials; therefore, the results should be interpreted with this in mind.

Secondly, our conclusions are constrained by the restricted nature of the trial participants: the majority of HF patients in this meta-analysis were in NYHA classes II or III, while two trials included patients with NYHA class I[15, 25] and IV;[16, 19] unfortunately, they did not provide the symptoms of depression data separately. Therefore, the benefits of exercise training in asymptomatic HF patients and in those with more severe HF have not been evaluated. Although we showed that exercise training was equally effective in patients under and over 65 years of age, it is not clear whether exercise training has the same effect in very old patients (mean age of 80 years). In the present meta-analysis, only two studies focused on very old patients with HF.[26-28] In one study,[28] the symptoms of depression data were not consistently provided in a poolable format, because the change from baseline was not reported separately between the intervention and the control groups. Therefore, we could not access the effect of exercise training on depression in very old HF patients.

Another limitation is the presence of statistical heterogeneity between the trials in this meta-analysis. In order to provide the studies with more uniform weight, we used the random effects model. In addition, a sensitivity analysis and subgroup analysis were also performed to lower the heterogeneity. However, the consistency of the pooled estimates is good in the different models when we remove one trial to lower the heterogeneity (see Table 3).

An additional limitation is the lack of trials examining the effect of strength training alone on symptoms of depression.

Finally, our conclusions were constrained by the quality of the trials reviewed. The main shortcomings were the lack of blinding procedures, and the absence of concealment of the randomization and of intention-to-treat analyses. However, we did not find a consistent difference in the effects demonstrated in the blinded design or non-blinded trials. Large, high quality RCTs, especially in very old patients and in those with more severe HF and normal LVEF, are required to determine the benefits of exercise training in the treatment of the symptoms of depression.

Conclusion

Exercise training was associated with a demonstrable benefit in the symptoms of depression in clinically stable HF patients. The beneficial effects appeared to be consistent regardless of age, length of the intervention, exercise setting, and trial quality. However, the efficacy of exercise in the treatment of the symptoms of depression was not confirmed in normal LVEF or for combined aerobic and strength training. Large-scale high quality RCTs are needed to verify the positive effects of exercise training in those populations. The evidence verified here should encourage physicians to recommend exercise as a clinically effective way to reduce the symptoms of depression in patients with HF.

Conflict of interest: none declared.

Authors contributions: Conception, Z.-Y.Z., R.-H.T., and G.-Q.Z.; design, Z.-Y.Z., R.-H.T., and Z.-D.B.; acquisition of data, R.-H.T., W.-F.W., Y.-J.L., and Y.H.; analysis, R.-H.T., Z.-D.B., and Z.-Y.Z.; interpretation of data, R.-H.T., W.-Q.H., and L.-M.Y.; and composition of the first draft of the review, R.-H.T. and Z.-D.B. All authors contributed to the final manuscript, and Z.-Y.Z. is the guarantor.

Ancillary