Stress Management Interventions for Adults with Heart Failure: Systematic Review and Meta-Analysis

Abstract

Objective:

Stress management interventions (SMIs) targeting psychological stress and other psychosocial factors associated with heart failure (HF) morbidity and mortality are increasingly recommended for adults with HF. SMI content and delivery varies widely and meta-analyses are needed to synthesize current findings to identify gaps in the literature. The purpose of this meta-analysis is to examine the efficacy of SMIs for improving anxiety, depressive symptoms, exercise capacity, and disease-specific quality of life in adults with HF.

Method:

Comprehensive searches of ten electronic bibliographic databases identified peer-reviewed, published, randomized controlled trials (RCTs) of SMIs for adults with HF.

Results:

Twenty-three RCTs were included (N = 2,294; mean age = 63.09 ± 7.27 years; 40% women, 56% white). Pooled effects indicated greater improvements in anxiety (d+ = 0.49, 95% CI = 0.09–0.89, k = 10), depressive symptoms (d+ = 0.39, 95% CI = 0.03–0.75, k = 13), disease-specific quality of life (d+ = 0.82, 95% CI = 0.40–1.24, k = 16), and exercise capacity (d+ = 0.57, 95% CI = 0.20–0.95, k = 14) among SMI recipients relative to controls at the first post-intervention assessment. The benefits were not maintained at follow-up. Participant characteristics (e.g., proportion women, HF severity), but not intervention type, moderated the findings.

Conclusions:

SMIs for adults with HF demonstrated short-term improvements in anxiety, depressive symptoms, quality of life, and exercise capacity. Future research sampling patients who are psychologically distressed with more thorough assessment of stress and longer follow-ups can elucidate the benefits of SMIs among adults with HF.

Heart failure (HF) affects more than 26 million adults worldwide (Ponikowski et al., 2014) and carries limiting physical symptoms characterized by dyspnea, fatigue, fluid retention (Bekelman et al., 2007), and impaired exercise capacity (Pollentier et al., 2010). Long-term management of HF and recurrent exacerbations of HF symptoms can be physically and psychologically stressful. Many adults with HF report elevated anxiety and depressive symptoms (Yohannes et al., 2010) and reduced quality of life (QoL)(Erceg et al., 2013). Consistent with a psychoneuroimmunology perspective of stress (Scott-Sheldon et al., 2008), chronic psychological distress (depression, anxiety) undermines autonomic functioning and neurohormonal responses and elevates risk of further morbidity and mortality (Endrighi et al., 2016; Sokoreli et al., 2016). The strongest evidence exists for the impact of depression in HF. Alleviating distress among adults with HF is critical to reduce psychological symptoms and improve outcomes.

Stress management interventions (SMIs), broadly defined as approaches to strengthen an individual’s skills to identify, understand, and cope with psychological and physical stress, may offer a promising non-pharmacological approach of supporting emotional responses to disease-related stress and also improving disease management and progression (cf. Scott-Sheldon et al., 2008). By improving coping and reducing stress (hypothesized mediator), SMIs may engender a normalization of physiological stress responses. In addition, some SMIs may also encourage increased health behavior engagement which also supports improved prognosis. Many SMIs have been studied (Varvogli & Darviri, 2011); some interventions are based on a theory-driven psychotherapeutic framework (e.g., cognitive behavioral therapy [CBT]) whereas others emphasize physical movement (e.g., balance, flexibility, or strength exercises) and mental strategies, relaxation techniques, or meditative practices (e.g., yoga). Numerous differences exist in the theoretical rationale, content, and delivery across SMIs. However, SMIs are aligned in their efforts to improve coping and alleviate psychological and/or physical markers of stress.

Although meta-analyses have begun to organize the evidence for the efficacy of SMIs for adults with HF, many focused on specific interventions or outcomes with mixed findings (Gomes-Neto et al., 2014; Gu et al., 2017; Jiang et al., 2018; Pan et al., 2013; Ren et al., 2017). A recent meta-analysis of psychological HF self-care interventions (e.g., CBT, relaxation) reported favorable effects on disease-specific QoL but not on anxiety, depressive symptoms, or exercise capacity (Jiang et al., 2018). This meta-analysis, however, was primarily focused on interventions targeting HF self-care as opposed to psychological stress and did not focus exclusively on interventions with a stress component (i.e., other SMIs such as tai chi and yoga were omitted). It remains unclear whether SMIs including a physical, non-aerobic exercise component may enhance physical function and exercise capacity, which often contribute to poor QoL, anxiety, and depressive symptoms. Additionally, prior meta-analyses have largely focused on interventions outcomes without characterizing intervention dimensions which may be most responsible for the effects (i.e., moderators). For example, the impact of intervention characteristics such as the type and dose on outcomes has not been reported. Similarly, sample characteristics such as age, sex, and HF disease severity may moderate effects. Intervention differences such as whether the intervention included a physical movement component may also be important. Given the breath of available non-pharmacological SMIs, simultaneous evaluation of SMIs is needed to understand which SMI offers the most benefit or whether moderating factors impact the efficacy of SMIs.

In light of the growing number of non-pharmacological interventions that have been tested, many of which individually had positive effects, a comprehensive review and meta-analysis of the literature pertaining to all types of SMIs that seek to improve a psychological outcome for patients with HF is needed to (a) evaluate whether broadly defined SMIs improve psychological and physical function, (b) clarify whether specific types or delivery modalities offer greater benefits relative to others, (c) to understand whether certain individuals may be more or less likely to benefit, and (d) to identify remaining ambiguities in the literature. Therefore, the purpose of this meta-analysis was to evaluate the efficacy of SMIs in alleviating psychological distress and improving physical function among HF patients. We hypothesized that SMIs would lead to improvements in psychological and physical function outcomes: (1) Perceived stress, depressive symptoms, anxiety, and QoL were selected as the psychological outcomes given their associations with physiological markers of stress and poor health outcomes; and (2) exercise capacity was selected as a physical function outcome given its importance as a functional status indicator for adults with HF. We also examined potential moderators of treatment effects to clarify who might benefit from SMIs. Finally, we followed the Preferred Reported Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines (Moher et al., 2009).

Method

Eligibility Criteria

Eligible studies evaluated the effects of a SMI in HF patients with preserved or reduced ejection fraction [EF]. For this review, SMIs included psychological, physical, or a combination of psychological and physical components to improve coping skills and reduce stress such as CBT, cognitive behavioral stress management, coping skills interventions, meditative movement (e.g., yoga, tai chi), mindfulness meditation, and relaxation techniques (e.g., biofeedback, progressive muscle relaxation). Interventions that primarily included disease self-management targets but also integrated a significant and clearly definable stress component were also eligible. Aerobic exercise-only interventions and studies for which the only comparison group was aerobic exercise-based were excluded because these interventions do not explicitly address stress. Studies evaluating cardiac rehabilitation or interventions added to cardiac rehabilitation were also excluded because rehabilitation has a substantial aerobic exercise focus. Additional criteria included publication in a peer-reviewed journal and use of a RCT design to assess the SMI’s impact on psychological outcomes (anxiety, depressive symptoms, stress), functional status (exercise capacity), or disease-related QoL.

Information Sources and Search Strategy

The primary search strategy involved searching ten electronic bibliographic databases (e.g., PubMed, PsycINFO) using a Boolean search strategy (terms included keywords associated with “complementary and alternative medicine practices” and “heart failure”) through October 2020 (last search conducted on November 9, 2020). Specific search strings were created to match the parameters of each database with no language or geographical restrictions (see Supplemental Material 1). Secondary sources included a review of (a) the reference sections of reviews, meta-analyses, and included studies, (b) electronic tables of contents of related journals, and (c) online databases of funded research (e.g., NIH RePORTer) and clinical trials (ClinicalTrials.gov).

Study Selection

Database searches were imported into reference management software; duplicate records were removed. The lead author (ECG) evaluated the title and/or abstract of each record based on pre-determined eligibility criteria. Record selection was independently and thoroughly checked by the PI (LAJSS) in consultation with the second author (ESB), a cardiologist, as needed. Full text articles were reviewed by the first author and independently verified by the PI (LAJSS). Discrepancies were resolved through discussion until consensus was reached. Secondary sources were cross-checked with the results from database searches and relevant new studies were added. If two or more manuscripts reported information from the same study, the manuscripts were linked and coded as a single study. Published manuscripts with the most complete information were considered the primary study record; additional documents (including any non-published reports such as theses) were considered supplemental material. Authors were contacted for additional information or data as needed. Fourteen authors were contacted; 64% responded and provided clarification about their study (4 out of 9) or data (5 out of 9).

Data Collection Process, Data Items, and Reliability

Coders were trained by the PI (LAJSS) to extract study data. Two coders reviewed each study and independently extracted study (e.g., publication year, location), sample (e.g., age, race, sex), design (e.g., RCT), and intervention details (e.g., type of SMI, facilitators, dose, materials, format, theoretical rationale). If a study provided insufficient data to compute effect sizes (ES), authors were contacted for additional information. Coding discrepancies were resolved through discussion; the lead author (ECG) and PI (LAJSS) jointly resolved any remaining discrepancies. Inter-rater reliability for categorical variables was 93% (mean Cohen’s κ = 0.85) and the average intra-class correlation coefficient for the continuous variables was 0.84 (median = 1.00).

Risk of Bias in Individual Studies

Methodological quality (MQ) of the studies were assessed using a 17-item rating form (total possible score = 25) adapted from published measures (Downs & Black, 1998; Fowkes & Fulton, 1991; Jadad et al., 1996; Miller et al., 1995). Inter-rater reliability for the methodological quality items was 88% (mean Cohen’s κ = 0.74). The proportion of MQ items satisfied by the study was calculated and used in the analyses; larger proportions indicate stronger MQ.

Risk of Bias Across Studies

Visual inspection of funnel plots and statistical tests assessed possible publication bias when ≥10 studies reported the same outcome (Begg & Mazumdar, 1994; Egger et al., 1997; Lau et al., 2006; Sterne & Egger, 2001). Trim-and-fill methods estimated the number of potentially missing studies and adjusted for the overall effect estimate when the possibility of publication bias was detected from the visual and statistical tests (Duval & Tweedie, 2000).

Study Outcomes

Outcomes included psychological (perceived stress, anxiety, depressive symptoms), physical function (exercise capacity), and disease specific QoL measured using self-report (e.g., Beck Depression Inventory—II (Beck et al., 1996), Minnesota Living with Heart Failure Questionnaire (Rector, 1987)) or clinical (e.g., walk tests, step test) assessments.

Summary Measures

Between-group ES measured differences between patients who received a SMI and those in a comparison group. Standardized mean difference ES (d) were calculated by dividing the post- vs. pre-test mean by the pooled standard deviation, and correcting for sample size bias (Hedges, 1981) and baseline differences (Becker, 1988; Morris & DeShon, 2002). Positive ES indicated improvements in SMIs relative to the controls.

Synthesis of Results

Random-effects procedures using maximum likelihood assumptions was used to calculate the pooled ES (d+) and 95% confidence intervals (Lipsey & Wilson, 2001). Heterogeneity was assessed with the Q statistic (Cochran, 1954), I² index, and τ² (Higgins & Thompson, 2002). Meta-regression was used to assess potential moderators (Lipsey & Wilson, 2001). Moderators of interest were (a) mean age of sample, (b) proportion of women in the sample, (c) HF severity via New York Heart Association (NHYA) class, (d) mean EF at baseline, (e) intervention type (psychological vs. blended physical/psychological) and (f) delivery modality (individual vs group format, presence of home practice recommendations). Analyses were conducted in Stata/SE 15.1 (StataCorpLP, 2017) using published macros (Lipsey & Wilson, 2001; Wilson, 2001); forest plots were generated using Comprehensive Meta-Analysis (Borenstein et al., 2013).

Results

Twenty-three 23 RCTs (k = 26 interventions) were included in the meta-analysis (Figure 1). Studies were conducted in the USA (n = 14), China (n = 2), Brazil (n = 1), Hong Kong (n = 1), India (n = 1), Philippines (n = 1), Sweden (n = 1), Taiwan (n = 1), and the United Kingdom (n = 1). Fifty percent of the studies were published after 2010 (range: 2004 – 2019).

Figure 1.

Screening and Selection Procedures

Sample Characteristics

Data were available from 2,294 [out of 2,581] participants (1,139 SMI and 1,155 controls).¹ On average, samples included 100 adults with HF (range = 15–870). Samples were 63±7 years of age, 40% women (n = 22), and 56% white (n = 14). All studies recruited participants from clinical settings. Three studies recruited patients experiencing elevated depression (Dekker et al., 2012; Freedland et al., 2015; Gary et al., 2010); no studies specifically recruited individuals based on distress level. Five studies explicitly sampled participants with an EF of ≤40% or ≤45%, whereas others sampled a wider EF range (Krishna et al., 2014; Swanson et al., 2009), or included patients with reduced or preserved systolic function (Dekker et al., 2012; Gary et al., 2010; Powell et al., 2010; Pullen et al., 2010). The average EF was 36±11 in the 14 studies reporting mean EF. In the five studies reporting the time since diagnosis, the average length of time was 43 months (SD = 11). Nineteen studies provided some indication of pharmacotherapy: beta-blockers (n = 16), angiotensin-converting enzyme inhibitors (n = 14), diuretics (n =12), statins (n = 7), and aspirin (n = 3). Of the 18 studies reporting NYHA class, the average proportion of participants classified as NYHA III/IV was 38% (SD = 0.20).

Intervention and Control Characteristics

Interventions (k = 26) included biofeedback (k = 6 interventions), cognitive therapy or CBT (k = 4), tai chi (k = 5), yoga (k = 3), coping skills training (k = 2), meditation (k = 2), progressive muscle relaxation (k = 1), diaphragmatic breathing (k = 1), self-management counseling that included cognitive restructuring and deep breathing training (k = 1), and a relaxation response intervention (k = 1). The median number of sessions was 12. Sessions lasted 75 minutes on average (SD = 110).

Characteristics of the interventions were not fully described across the studies. Only seven studies reported using theory to guide the intervention content; these interventions were guided by cognitive behavioral (Beck & Greenberg, 1984), social cognitive (Bandura, 1986), and stress and coping (Lazarus & Folkman, 1984) theory. Most interventions (14 out of 26) indicated some tailoring of the intervention content. Interventions were most often delivered face-to-face; a single study was conducted via telephone only (Sherwood et al., 2017). Four interventions offered telephone follow-up or boosters after the in-person session(s) (Dekker et al., 2012; Freedland et al., 2015; Seo et al., 2016; Yu et al., 2007). Facilitators included nurses (k = 10), certified instructors (k = 6), and psychologists or masters- and doctoral-level therapists and/or students (k = 5); five studies did not specify the facilitator. Fourteen interventions recommended home practice; participants were asked to complete a median of 42 sessions (range: 6–365) of 20 minutes (range: 5–70). Participant compliance was described in 16 of the 22 RCTs, but only 6 RCTs reported compliance rates (range: 72–100%). Materials were provided to participants in the form of handouts, audio recordings, or other relevant materials (k = 15 out of 26).

The most common control condition used was standard care (n = 13). Details regarding the standard care control groups received varied across studies but, when provided, generally included pharmacotherapy and outpatient medical supervision from HF nurses, cardiologists, primary care providers, or other healthcare providers, and/or risk factor modification recommendations. Other control conditions included sleep hygiene education (Wang et al., 2014), weekly discussions about stress (Curiati et al., 2005), and other active controls that did not include SM (e.g., health education; telephone check-ins, n = 7), three of which were matched to intervention conditions for time and contact. One study included used an attention control (Swanson et al., 2009). See Supplemental Material 2 for study, intervention, and control condition details.

Synthesis of Results

Psychological Outcomes²

Depressive symptoms were typically assessed using self-report measures. One study used three measures to assess depressive symptoms; ES were averaged to represent a composite depression ES (Freedland et al., 2015). Intervention participants reported greater reductions in depressive symptoms compared to controls at the first post-intervention assessment (d+ = 0.39, 95% CI = 0.03–0.75, k = 13; M weeks = 1.62 ± 3.66, range = 0–12 weeks). Examination of variability in the observed effects revealed that the effects were heterogeneous (Q [12] = 36.43, p <.001; I² = 67%, 95% CI = 41–82%, τ² = 0.13, 95% CI: 0.06–1.09, SE = .09). Heterogeneity exceeded the 50% threshold and uncertainty limits were wide. Depressive symptoms did not differ significantly between groups at the last assessment (d+ = 0.25, 95% CI = −0.27–0.78, k = 7; M weeks = 21.86 ± 14.65, range = 12–52 weeks; see Table 1).

Table 1.

The Overall Weighted Mean Effects Comparing Stress Management Interventions versus Controls on Depressive Symptoms at First Post-Intervention Assessment.

Note. The overall weighted mean effect sizes are calculated using random-effects models with methods of moments (MM) and full information maximum likelihood (ML) methods to estimate the between-study variance. Weighted mean effect sizes (d₊) are positive for differences that favor the stress management intervention group relative to controls. d₊, weighted mean effect size; SE, standard error; CI, confidence interval.

Anxiety was assessed via self-report. One study used two measures of anxiety; ES were averaged (Freedland et al., 2015). SMI participants reported greater reductions in anxiety relative to controls at the first post-intervention assessment (d+ = 0.49, 95% CI = 0.09–0.89, k = 10; M weeks = 0.9 ± 2.18, range = 0–7 weeks);³ however, heterogeneity exceeded the 75% threshold (Q[9] = 39.50, p < .001; I²=77, 95% CI = 58–88%, τ² = .17, 95% CI = 0.08–1.04, SE = 0.11). Between-group differences were not maintained at the last assessment (d+ = 0.14, 95% CI = −0.59–0. 88, k = 3; M weeks = 30.33 ± 19.86, range = 13–52 weeks; see Supplemental Material 3).

Exercise Capacity

Exercise capacity was assessed using the six-minute walk test (6MWT) in all studies except one study where the Incremental Shuttle Walking Test was used (Barrow et al., 2007). SMIs increased exercise capacity relative to control participants (d+ = 0.57, 95% CI = 0.20 – 0.95, k = 14; M weeks = 0.93 ± 3.20, range = 0 – 12 weeks) at the first post-intervention assessment. High heterogeneity was noted and exceeded 75% (Q [12] = 148.88, p<.001; I²=91%, 95% CI = 87–94%, τ² = 0.52, 95% CI = 0.56 – 3.30, SE = 0.37).⁴ The improvements in exercise capacity between the SMI and control groups were not maintained at the last assessment (d+ = 0.29, 95% CI = −0.38 – 0.95, k = 5; M weeks = 15.13 ± 5.56, range = 12 – 25 weeks; see Table 2).

Table 2.

The Overall Weighted Mean Effects Comparing Stress Management Interventions versus Controls on Exercise Capacity at First Post-Intervention Assessment.

Note. The overall weighted mean effect sizes are calculated using random-effects models with methods of moments (MM) and full information maximum likelihood (ML) methods to estimate the between-study variance. Weighted mean effect sizes (d₊) are positive for differences that favor the stress management intervention group relative to controls. d₊, weighted mean effect size; SE, standard error; CI, confidence interval.

Quality of Life, Disease-Specific

Thirteen studies measured QoL using HF-specific QoL questionnaires. Participants in the SMIs reported improved disease-specific QoL relative to controls (d+ = 0.82, 95% CI = 0.40–1.24, k = 16; M weeks = 1.41 ± 3.32, range = 0–12). The effects for disease-specific QoL were heterogeneous (Q [15] = 152.07, p < .001, I² = 90%, 95% CI = 86–93%, τ² = .95, 95% CI = 0.4–3.82, SE = 0.47); heterogeneity was high and exceeded the 75% threshold. No significant differences were observed between SMIs and controls at the last assessment (d+ = 0.06, 95% CI = −0.65–0.76, k = 5; M weeks = 15.00 ± 5.61, range = 12–25 weeks; see Supplemental Material 3).

Moderator Analyses

Moderator tests evaluated whether demographic (age, proportion of women), disease (proportion NYHA class III/IV,⁵ proportion using beta-blockers, mean EF), intervention type (psychological vs. physical/psychological) or delivery modality (group vs individual; home practice recommended) explained the heterogeneity observed in the effects for each outcome at the first post-intervention assessment. SMIs were more successful in reducing anxiety when samples were composed of more women (β = 2.99, SE = 0.84, p = .007; adjusted R² = 76%). Effects on QoL were smaller among samples with a higher proportion of participants with a NYHA class III/IV HF (β = −3.13, SE = 1.34, p = .042; adjusted R² = 34%). Age, EF, proportion using beta-blockers, intervention type or delivery modality did not moderate the effects on anxiety or QoL. There were no significant moderators of depressive symptoms or exercise capacity.

Methodological Quality

All studies randomized participants to groups; twelve studies (52%) completely randomized participants to the groups; and ten studies (44%) randomized participants via a block-randomization scheme or by matching participants on specified characteristics (e.g., age, gender) prior to randomization. Controls were matched (22%) or assessed for equivalency (78%). One study blinded participants to the assigned group but 65% of the studies reported some attempt to blind members of the research team. Ten studies (44%) reported high retention rates (i.e., ≥85% of the initial sample). Most (87%) studies performed appropriate statistical analyses; 52% of studies used statistical approaches to account for missing data (e.g., intent-to-treat). On average, 73% percent of MQ items were satisfied. MQ was not a significant predictor of the pooled ES for any outcome (ps ≥ .05). See Supplemental Material 4 for full MQ details.

Risk of Bias Across Studies

Visual inspection of the funnel plots for all outcomes at the first assessment revealed possible asymmetries (see Supplemental Material 5). Publication bias was assessed only at the first assessment because fewer than 10 studies were available at the last measurement occasion. Begg (Begg & Mazumdar, 1994) and Egger’s (Egger et al., 1997) statistical tests did not indicate the presence of small study effects for any outcome: depressive symptoms (Begg: Δx-y = −24, z = −1.46, p = .14; Egger: bias coefficient = −1.09, SE = 1.26, p = .41), anxiety (Begg: Δx-y = 3, z = 0.27, p = .79; Egger: bias coefficient = 1.38, SE = 2.59, p = .61), exercise capacity (Begg: Δx-y = 27, z = 1.48, p = .14; Egger: bias coefficient = 2.31, SE = 1.39, p = .12), or QoL (Begg: Δx-y = 36, z = 1.62, p = .11; Egger: bias coefficient = 2.63, SE = 2.14, p = .24).

Discussion

This review represents the first comprehensive meta-analysis of SMIs to target both psychological and functional status in adults with HF and to evaluate whether sample and intervention characteristics moderated observed effects. While we found that SMIs improved psychological outcomes and exercise capacity, most failed to directly measure the most proximal variable of interest, psychological stress. The overall magnitudes of the effect sizes for depressive symptoms, anxiety, disease specific QoL, and exercise capacity are medium to large (d+ = 0.39–0.82) at the first post-intervention assessment and suggests that 64–80% of adults with HF who received the SMI showed improvements on psychological or physical outcomes relative to the average for those in the control or comparison. The benefits, however, were not maintained at the follow-up assessment and the moderator tests were limited by the currently available literature. In addition, critical stress processes such as perceived stress and coping were reported in an insufficient number of studies to examine via meta-analysis.

Short- Term Effects and Sustainability of SMIs

The primary purpose was to test the hypotheses that SMIs improve psychological stress, exercise capacity, and disease-specific QoL among adults with HF. Testing the hypothesis that SMIs would alleviate psychological stress could not be completed due to infrequent assessment. Other hypotheses related to depressive symptoms, anxiety, disease specific QoL, and exercise capacity were confirmed at the initial post-intervention assessment. It is encouraging that SMIs demonstrated this benefit because even mild symptoms of depression can have a negative impact on HF (Bhatt et al., 2016) and because anxiety undermines engagement in health recommendations (Celano et al., 2016). Our findings corroborate prior reviews of psychological interventions (Jeyanantham et al., 2017; Jiang et al., 2018) and mind-body interventions (Woltz et al., 2012; Zou et al., 2019) among adults with HF. Clinically meaningful improvements were also observed in disease-specific QoL following SMIs. Importantly, some adults with HF prefer a better QoL to a longer lifespan (Kraai et al., 2013; MacIver et al., 2008).

Despite the benefits observed to depressive symptoms, anxiety, and disease-specific QoL, stress processes (the hypothesized mediator) could not be assessed due to insufficient reporting. The few studies measuring these outcomes also reported inconsistent findings. Jayadevappa et al. (2007) observed no significant between-group differences in perceived stress at three- or six-months post-intervention. Bose et al. (2016) did not show any statistically significant differences between the SMI and control groups on coping processes. Future rigorous RCTs that measure, and report, stress processes will be necessary to determine whether SMIs successfully reduce perceived stress and improve coping processes (cf. Scott-Sheldon et al., 2008).

Although behavioral or biological measures of stress were not routinely assessed, the broader literature suggests that psychological stress may contribute to or exacerbate depressive symptoms and anxiety (Tafet and Nemeroff, 2016). Specific SMIs may reduce psychological distress (depressive symptoms, anxiety) by improving stress appraisal, interrupting maladaptive thought-mood relationships, and encouraging adaptive coping behaviors and emotion regulation strategies. Relaxation, meditative practice, and mind-body movement techniques increase awareness of and promote reduction in stress-related muscle tension, reduce rumination through encouragement of mental focus on breath or meditative thought, and lessen stress appraisals and emotional reactivity to stress. These techniques also induce the relaxation response (Benson, Beary, & Carol, 1974). Physiological disruptions commonly present with depression (e.g., inflammation, SNS and HPA dysregulation) or anxiety (e.g., elevated heart rate) may also be improved (Creswell and Lindsay, 2014; Kong et al. 2019; Riley & Park, 2015). Importantly, much research is needed to clarify how different SMIs may operate through distinct pathways.

SMIs were successful in improving exercise capacity which is especially noteworthy because SMIs did not include an aerobic exercise component. For comparison, cardiac rehabilitation, a primarily exercise-based intervention, consistently demonstrates improvements in exercise capacity. Mean 6MWT changes of 60 meters are seen in cardiovascular outpatients completing rehabilitation (Bellet et al., 2012), whereas we found a mean 6MWT increase of 20 meters in HF patients who received a SMI versus a decline of 25 meters in controls. SMIs with movement components may improve exercise capacity through increased muscle strength, flexibility, and endurance (Tran et al., 2001). The mechanisms through which SMIs without a movement component improve exercise capacity are less clear but associated reductions in depressive symptoms and anxiety may be associated with increases in other exercise engagement or mitigation of stress-related sympathetic activation. Importantly, higher exercise capacity is associated with a lower risk of HF-related hospitalizations (Palau et al., 2018) and lower cardiovascular mortality (Rostagno et al., 2003). These findings, however, should be interpreted within the context of the scientific literature reviewed such that removal of a single outlier reduced the effect of SMIs on exercise capacity. This outlier represented the only yoga study assessing exercise capacity and the only study published in India (Krishna et al., 2014). The larger effect size may reflect the cultural importance, and perceived benefits, of yoga in India (Newcombe, 2017). Additional research is needed to clarify the benefits of yoga and other SMIs on functional status.

Another important purpose of this meta-analysis was to investigate the sustainability of any observed benefits of SMIs. The benefits observed in the short-term were not maintained at follow-up. Few of the reviewed studies (10 out of 23) included follow-ups, and most follow-ups occurred within six months of intervention completion. Strikingly, approximately half of studies that did include assessments of depressive symptoms, anxiety, QoL, or exercise capacity beyond end-of-treatment did not demonstrate significant improvements relative to controls at follow-up. Thus, one possibility is that improvements observed at post-intervention were not maintained due to the limited number of studies assessing outcomes at a longer follow-up (i.e., low power). Alternatively, physical decline or lack of continued engagement in stress management skills or other health behaviors may have led to dissipating effects over time. More studies with longer follow-ups, and exploring the added value of a booster session to promote the long-term efficacy of these treatments, are needed to establish the potential benefits of SMIs for adults with HF.

Factors that Moderate the Short-Term Efficacy of SMIs

We also examined dimensions of the studies as potential moderators of SMIs. Only two significant moderators were identified: (1) Improvements in anxiety were more pronounced in studies sampling more women. Baseline anxiety may have been higher in women relative to men indicating more room for change (McLean et al., 2011) but we could not test this hypothesis empirically due to the limited sex-specific reporting in the studies. In addition (or alternatively), women are more likely than men to engage in meditative practices and perceive greater benefits accruing from SMIs (Upchurch & Johnson, 2019). Thus, samples with more women may have observed larger effects due to greater women engagement and practice. (2) Studies sampling more HF patients identified as NYHA class III/IV were less likely to report improvements on disease-related QoL. The greater physical symptom burden and elevated depressive symptoms characteristic of a higher NYHA class may have attenuated the observed effect for QoL.

Limitations and Future Research

Limitations of the current literature limited the scope of our meta-analysis. First, the absence of consistent stress measurement significantly limits the current review. Behavioral or biological markers of stress were not typically assessed, and inclusion of such measures are critical for determining if improved coping skills or reduced stress mediates the association between SMIs and improvements in psychological outcomes or physical functioning. Second, few studies specifically targeted participants with clinically elevated psychological stress, depression, or anxiety at study entry and therefore the potential impact of SMIs on clinical levels of distress is limited. Third, evaluation of mediating mechanisms of intervention efficacy were largely absent, which limited our ability to identify mechanisms of action. Future evaluation of coping skills, stress reactivity, and self-management will deepen understanding of how SMIs may benefit psychological and physical outcomes. Fourth, intervention content varied widely which prevented additional moderator tests to determine which SMI components may be most beneficial for adults with HF. Also, due to the limited reporting of study, sample, and intervention details in the primary studies, we could not test other important moderators of intervention efficacy which may help to better understand the observed heterogeneity (e.g., intervention adherence, ischemic versus non-ischemic HF etiology, preserved versus non-preserved ejection fraction, psychiatric and medical comorbidity). Fifth, follow-up assessments were often not included which limited power to determine the longer-term benefits of SMIs. Sixth, our MQ assessment instrument was adapted from prior measures to assess the quality of the studies more completely but has not been formally validated. Finally, underreported details related to standard care or participant’s ongoing physical activity outside of the research program may impact study findings in relevant ways given that standard care may differ across regions or countries and patient physical activity or participation in cardiac rehabilitation may be an unmeasured confound.

The ambiguities in the literature precludes clear clinical recommendations regarding which SMIs should be considered as part of HF management at this time. We offer several recommendations for future research based on the findings of this review in order to identify efficacious interventions that can increase implementation effectiveness:

1. Interventions to treat psychological distress, rather than prevent the development or onset of psychological distress, should be purposefully targeted to individuals in need of support. Researchers should consider recruiting individuals with elevated stress at study entry or reporting analyses of those with elevated distress in addition to the overall sample effects.

2. Future SMI trials should incorporate self-reported and biomarkers of stress to evaluate the impact of SMIs on theoretically relevant outcomes.

3. Trials are encouraged to explicitly measure and report theoretical mechanisms of action (e.g., coping skills, stress reactivity) and elements recommended in intervention reporting guidelines (Hoffmann et al., 2014).

4. Follow-up and prognostic outcome assessments should be incorporated, whenever possible.

To our knowledge, this is the first meta-analysis examining both the psychological and functional benefits of a broad range of SMIs among adults with HF. Overall, this meta-analysis suggests that SMIs offer short-term benefits on depressive symptoms, anxiety, exercise capacity, and disease-related QoL and supports a Class IIB, Level of Evidence A recommendation that SMIs may be considered to alleviate some psychological symptoms and improve exercise capacity in the short-term. Significant gaps in the literature currently limit identification of specific SMIs that may offer the most benefit. Additional methodologically rigorous RCTs that adhere to intervention reporting standards are needed to increase power to enable a thorough examination of intervention components and the sustainability of intervention benefits. Thereafter, future meta-analyses using a taxonomic meta-analytic approach may better clarify whether distinct intervention components confer greater benefits relative to others (Hedges et al., 2020). Such clarification may help to identity whether non-pharmacological approaches that target psychological distress may be useful to incorporate into HF management.