This review followed the Cochrane Handbook for Systematic Reviews of Interventions to formulate the research question, define eligibility criteria, conduct the comprehensive literature searches, assess risk of bias, and synthesize evidence [25]. The PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) statement informed the reporting structure of this manuscript [26], ensuring all the key components of a high-quality systematic review were included, including that search strategies and data extraction methods were clearly outlined. The protocol of this review was registered in PROSPERO (CRD42023452276) to enhance transparency, minimize bias, and prevent duplication of effort.

are shown in Textbox 1. Inclusion criteria were guided by the populations, interventions, comparisons, outcomes, and study designs (PICOS) framework. Feasibility results were not used as eligibility criteria, as this study aims to investigate the effectiveness of the intervention. However, we extracted this information to inform future research.

A comprehensive literature search was conducted across 5 electronic databases, covering the period from database inception to December 2024: Embase (from 1910), Medline (from 1946), PubMed (from 1997), CINAHL (from 1981), and the Cochrane Library (from 1996). The selection of these databases aligned with the research question by ensuring broad coverage of medical, psychological, and health-related literature, maximizing the likelihood of identifying relevant studies. The search strategy used a combination of Medical Subject Headings (MeSH) terms and free-text keywords to capture variations in terminology. Boolean operators were applied to refine the search, combining terms such as “cardiovascular disease” AND (“psychological intervention” OR “eHealth”) AND (“depression”). The search strategies for all databases are provided in Multimedia Appendix 1. Additionally, a hand search of bibliographies from relevant review articles was performed to identify any further studies [30].

Endnote (version X9; Clarivate) was used to identify and remove duplicate records; 2 reviewers (JJS and RL) independently screened studies by evaluating titles and abstracts against the predefined eligibility criteria, focusing on the relevance to the research question and the predefined PICOS criteria. Studies that met the inclusion criteria or had unclear eligibility progressed to a full-text review. Disagreements between the 2 reviewers throughout the review process were resolved through discussion and, when necessary, adjudicated by a third researcher (LB). The third researcher’s role ensured objectivity in the decision-making process and the transparency of study selection.

Methodological quality was assessed independently by 2 researchers (JJS and RL) using the Cochrane Risk of Bias tool (version 2), evaluating each study across 5 domains: randomization, intervention deviations, missing data, outcome measurement, and selective reporting. Bias risk was categorized as “low,” “some concerns,” or “high.” Studies with “some concerns” in 3 or more domains were deemed high risk. Due to the limited number of studies, funnel plots were not generated. Publication bias was evaluated by comparing outcome variables presented in randomized controlled trials with their trial registry and published protocols.

Study data were extracted independently by 2 authors (JJS and RL). The authors developed a table for data extraction, which included the following: (1) origin of the articles, including authors, year, and country; (2) sample characteristics, such as gender, sample size, age, setting, and diagnosis; (3) trial design, consisting of a brief description of the intervention and control groups and study duration; (4) data collection time points; and (5) feasibility data regarding recruitment, attrition, and adherence.

The overall rates of recruitment, attrition, and intervention adherence were documented using median, minimum, and maximum values. Predefined cutoff values for the feasibility criteria, specifically a recruitment rate of at least 20% and an attrition of less than 25%, were determined based on clinical relevance as established in a previous study [31].

Review Manager (version 5.3; Cochrane) was used for data pooling due to its capacity for combining outcomes from multiple studies with varying sample sizes. Data pooling was conducted when ≥3 studies were reporting the same outcome; otherwise, synthesis without meta-analysis (SWiM) was conducted [32]. The SD, if unavailable, was calculated from the CI. Psychological well-being was considered the primary outcome. Where the same study reported outcomes at multiple end points, those from immediately after the intervention were used to investigate the intervention effect. For continuous outcomes, intervention effects were expressed as standardized mean differences (SMDs) using Cohen d, which standardizes differences between groups, enabling comparison across studies with different measurement scales. Cohen d values were interpreted as representing small (0.2-0.5), moderate (0.5-0.8), or large (>0.8) effect sizes, which provided a framework to understand the magnitude of an intervention’s impact on psychological outcomes. For dichotomous outcomes, intervention effects were summarized as risk ratios using the Mantel-Haenszel method, which accounts for study weights in pooled analyses and is appropriate for binary data.

Statistical heterogeneity was assessed using the I² statistic, which quantifies the proportion of variability in effect estimates due to between-study differences rather than chance. I² values were interpreted as follows: 0% to 25% indicated low heterogeneity, 26% to 50% indicated moderate heterogeneity, and >50% indicated substantial heterogeneity. Random-effects models were used when substantial heterogeneity was detected to provide more conservative estimates where heterogeneity was present. Leave-one-out sensitivity analysis was conducted where significant heterogeneity was observed to identify the impact of individual studies on the overall findings by systematically excluding one study at one time to detect outliers that disproportionally contributed to heterogeneity. Due to the limited number of studies, meta-regression and subgroup analysis were not performed [25].

Figure 1 presents the PRISMA flow diagram illustrating the search results. Initially, 6701 articles were identified across the 5 databases. Following the removal of 31 duplicates, 6670 studies remained for relevance assessment. After screening titles and abstracts, the full texts of 212 studies were examined for eligibility. Ultimately, 12 trials were included in this review.

Multimedia Appendix 1 provides an overview of the included articles. Trials were conducted across 10 countries in Asia, Europe, and North America, including a total of 2319 participants, with study sample sizes ranging from 50 to 562. Participants had a mean age of 58.68 (SD 5.85) years, with a majority being male (62.8%). Participants were diagnosed with heart failure (111/2319, 4.79%) and [24,33] coronary heart disease (628/2319, 27.08%) [34-36], and there was a mixture of patients with CVDs and at high CVD risk (562/2319, 24.2%) [37], controlled arrhythmia (491/2319, 21.17%) [38-40], and general CVDs (527/2319, 22.72%) [31,41,42]. Eight trials focused on individuals with mild to moderate levels of anxiety or depression (1561/2319, 67.31%) [31,33-38,41]. Outcomes were assessed immediately after the interventions. Long-term follow-up data, ranging from 1 to 5 years after the interventions, was only collected in 3 studies [34,38,39].

Figure 2 shows the results of the risk of bias assessment. Most trials were at low risk regarding random sequence generation, deviations from intended interventions, and missing outcome data. “Some concerns” were related to inadequate reporting. Specific issues around detection bias were noted in trials by He et al [40], Humphries et al [34], and Clays et al [24], who explicitly stated that blinding was not implemented due to the behavioral nature of interventions. This lack of blinding raises concerns about potential bias in outcome measurement, as participants and data assessors may have been influenced by knowledge of intervention assignments. Selection bias was consistently rated as low across all studies. In summary, although some studies were categorized as low risk, the presence of high or unclear risks in certain domains necessitates a cautious interpretation of the compiled results. In addition, we recommend adherence to standardized reporting practices using the CONSORT (Consolidated Standards of Reporting Trials) guidelines, which provide a structured framework for transparent reporting of study design, procedures, and outcomes, reducing the potential for bias and improving replicability.

Four types of psychological interventions were examined across the 12 trials, most commonly internet-based cognitive behavioral therapy (ICBT) [24,33-38,41], followed by mindfulness-based interventions [40,42], problem-solving therapy [39], and stress management [31].

For ICBT interventions, most incorporated core CBT techniques like behavioral activation, worry management, and problem-solving. Specific areas such as psychoeducation [33,35] or the psychosocial aspects of implantable cardioverter defibrillators [38] were the focus of some interventions. Additionally, certain programs integrated additional therapeutic methods, including interpersonal therapy [37] or mindfulness [24]. The interventions used a variety of platforms, from multimedia supports to telemonitoring devices and expert systems for automated guidance. Some featured interactive web-based group settings with discussion forums and secure email systems for queries. For instance, one study provided customization options that allowed participants to tailor their treatment paths after completing compulsory modules [34].

Mindfulness interventions also differed in their approaches. One used a brain-computer interface app with real-time electroencephalography monitoring to facilitate dynamic monitoring over a 35-minute mindfulness exercise, including body scanning and relaxation techniques [40]. The other mindfulness intervention offered a more traditional 12-week online course, encompassing various meditation types, like self-reflection and yoga, supported by biweekly reminders [42]. The 12-week problem-solving therapy, focusing on psychoeducation, used biweekly emails to foster adherence [39]. For stress management, a 2-stage strategy was adopted: participants initially engaged in either a self-guided or lay-coach–supported web program, with nonresponders escalated to a more intensive program that included motivational interviewing [31].

The interventions were delivered by a diverse group of professionals and nonprofessionals. In 4 trials, the intervention was delivered by mental health professionals, including a mental health nurse [33] and psychologists [35,38,39]; in other studies, the intervention was delivered by a multidisciplinary team [24,36], a therapist [34], nurses [40,41], a trained lay coach [31], and, finally, only a self-directed approach in 2 studies [37,42].

Recruitment and attrition rates were considered acceptable [31]. The included trials showed an overall recruitment rate of 42% (ranging from 10% to 85%). Regarding attrition, the median attrition rate was 19% (ranging from 5% to 24%) in intervention groups and 8% (ranging from 3% to 43%) in control groups. In the intervention groups, there were 231 dropouts out of 1213 participants, with 6 (4%) dropouts due to health-related reasons and 225 (96%) due to unspecified non–health-related reasons. In the control groups, there were 108 dropouts out of 1106 participants, with 3 (6%) due to unspecified health-related reasons and 105 (94%) due to non–health-related reasons. It is noteworthy that unspecified non-health reasons were the most common cause of dropouts in both the intervention and control groups (Multimedia Appendix 1).

Adherence to the eHealth interventions varied across the interventions, ranging from 20% to 73%; 2 studies did not report the adherence rate [36,42]. Key factors contributing to low adherence included technical issues (eg, insufficient computer literacy) [24,34], time constraints, and emotional discomfort evoked by the intervention. In addition, one study found that coach calls, originally designed to last 10 to 15 minutes, often exceeded this time frame due to technical challenges [31].

Three studies that delivered ICBT recorded participants at risk of deliberate self-harm, as detected by a validated depression scale, and reported similar numbers (n=21 in the intervention group and n=23 in the control group [37] vs n=2 in the intervention group and n=3 in the control group [35]) or fewer numbers (n=1 in the intervention group and n=2 in the control group) [41].

Four trials comparing ICBT to usual care reported cardiovascular outcomes. One reported significant reductions in 1-year and 3-year mortality risk within the group but not between groups [24]; another lower hospitalization-related costs, although this was nonsignificant (–$1010, 95% CI –$3294 to $1274; P=.39) [36]; yet another similar frequencies of hospitalization [38], and finally, another a nonsignificantly higher number of cardiovascular events (intervention group: n=36 cases; control group: n=25 cases) [34]. One study providing mindfulness training reported 5 and 10 participants with fluctuations of >50 mmHg in systolic blood pressure in the intervention and control groups, respectively [40].

The aim of this review was to assess the effects of eHealth psychological interventions in improving psychological outcomes among individuals with CVDs. This is the first review of such interventions in CVD demonstrating they effectively reduce depression. The significant improvement identified could be attributed to the use of structured, evidence-based interventions, chiefly ICBT compared to usual care, website-based discussion forums, or education. These findings are consistent with other reviews in more general populations or those with other chronic conditions on the benefits of ICBT [43]. While the effects of ICBT on anxiety were not significant, more data on the effects of other eHealth psychological interventions, as well as their effects on stress, HRQoL, and cardiovascular events, are needed prior to drawing conclusions.

The effectiveness of eHealth psychological interventions in reducing depression among CVD patients is noteworthy. This finding is consistent with previous literature that reported the effectiveness of traditional (ie, non–technology-based) nonpharmacological psychological interventions in alleviating depression among CVD patients [44]. Another similarity is that CBT was identified as the most commonly used psychological therapy for CVD patients, followed by mindfulness and other stress management interventions [44,45]. However, previous CVD meta-analyses of traditional psychological intervention showed inconsistent clinical outcomes, including cardiac events and mortality [45,46]. One review found a small but nonsignificant effect of nonpharmacological psychological interventions in reducing cardiac mortality (risk ratio 0.80, 95% CI 0.64 to 1.00; I²=0%; P=.56) for patients with coronary heart disease [47]. Conversely, one meta-analysis suggested significant improvement in cardiac mortality (risk ratio 0.81, 95% CI 0.68 to 0.96) and the occurrence of cardiac events among participants who received psychological interventions compared to control group participants [45]. In this review, the clinical outcomes were not assessed or reported by most of the included studies, which hindered evaluation of the effect of eHealth psychological intervention for this parameter.

The impact of eHealth interventions on anxiety and HRQoL were inconclusive. This may be because the applied interventions did not fully address the inherent complexities of anxiety or the multifaceted nature of HRQoL faced by patients with CVDs [48]. For anxiety, specific therapeutic strategies—such as exposure therapy or specialized CBT approaches tailored and proven effective for anxiety management [49]—were not the primary focus of these trials. Additionally, improving HRQoL in CVD goes beyond psychological well-being to include improvement in physical endurance, social engagement, and daily functioning [48]. Effective interventions would need to integrate elements such as physical rehabilitation exercises, social skills training, and occupational therapy to holistically address HRQoL. This highlights a gap in the current therapeutic approaches and underscores the necessity for more integrative eHealth intervention designs tailored to such needs of CVD patients. The varied effectiveness likely also stems from the differing technology used and guidance levels in the interventions. For instance, mindfulness has been shown to be effective in face-to-face sessions [50], but its key strengths—deep interaction and personalized adjustment—may not translate as effectively to eHealth formats.

This review supports recommendation of ICBT as a standard treatment for depression in CVD patients [51]. However, considering a previous review focusing on face-to-face CBT that showed comprehensive and significant improvements in depression as well as anxiety and HRQoL among patients with CVDs [52], patient concerns or diagnoses and preferences must be considered. Two factors might explain the differences: (1) the prior review focused exclusively on participants diagnosed with mental disorders, whereas this review included a broader demographic of CVD patients; and (2) the inherent interpersonal interactivity and potential personalization of face-to-face CBT, which can help patients recognize and modify unhelpful thinking patterns, might be superior to online portals and discussion forums. Despite this, the resource-intensive and costly aspects of traditional CBT [53], coupled with the dynamic nature of mental health conditions among participants, highlight the potential of alternative delivery methods such as eHealth.

Moreover, a deeper examination of the eHealth modalities used by the studies included in this review reveals several unique strengths compared to traditional methods. First, the temporal attributes of some ICBT interventions, which allowed dynamic telemonitoring and daily tracking of heart rate, skin response, temperature, and physical activity levels, as well as support systems to provide tailored advice based on participants’ telemonitoring results and progress, are promising [24]. Second, in some trials, intervention adherence was tracked using built-in website counters of online course participation, with professionals providing individualized feedback on participants’ ICBT assignments for cognitive restructuring and behavioral activation [24,33-38,41]. Cognitive restructuring requires that participants continuously complete homework, identifying and modifying detrimental thought patterns and engaging in positive and meaningful activities. Thus, supporting participants in practicing ICBT in their daily lives, coupled with online support from professionals through time, could be one unique strength of this mode, leading to sustainable effects [54]. Lastly, the eHealth modality allows for a confidential platform in which participants can reveal negative thinking patterns and professionals can monitor progress and provide timely feedback or treatment plan adjustments. Thus, eHealth offers unique strengths in accessibility, scalability, and personalization [55].

When harnessing the capabilities of eHealth, it is important to note that technology serves as a tool and cannot replace the therapeutic encounter and rapport. Mental health professionals play a crucial role as gatekeepers in the creation and implementation of eHealth psychological interventions [5,56]. Their analysis of patients’ self-monitoring data and provision of real-time feedback are crucial for quality care and outcome optimization. Additionally, these professionals are instrumental in tailoring treatment to the patient’s context, such as by managing workplace stress or other inciting factors in their physical and psychosocial environments, thereby enhancing intervention applicability and impact [56].

Some practice implications are important to consider. First, many CVD patients continue to have unrecognized comorbid psychological conditions. With more patients adopting technology and with technological advances, the adoption, acceptance, quality, and efficacy of psychological eHealth interventions may improve. Second, it may be important to be more selective in allocating patients to psychological eHealth interventions, such that patients who do not prefer technology, or who have no (or who have severe) psychological conditions should be provided with alternative interventions. Finally, in addition to considering patient preference, stepped care, such as was implemented in one included trial, should be pursued so nonresponding patients can receive care through to remission and so that their excess cardiovascular risk is reduced [36].

Several implications for future studies are noteworthy. Studies may consider combining ICBT with antidepressants, as guidelines from the American Heart Association and American Academy of Family Physicians recommend such combination for the treatment of post–acute coronary syndrome depression [57,58], addressing both psychological and physiological causes as well as outcomes of depression in this population. Strategies such as enhancing eHealth platform interactivity for real-time feedback, using adaptive algorithms for personalized therapy, and integrating virtual or augmented reality for immersive experiences could enable achievement of comparable benefits to traditional therapeutic sessions [59].

The review is limited by the limited available trials in this field. Moreover, while publication bias was evaluated by comparing findings with registered trials and published protocols, meta-regression and funnel plot asymmetry assessment for publication bias were not possible. The participants were predominantly middle-aged or young-older adults from developed countries, which may limit the generalizability of the findings to older age groups or cultures unfamiliar with Western psychological treatment approaches. Caution is needed in interpreting the results and replicating the interventions, as most included studies used mixed eHealth approaches, combining multiple intervention modalities. Future studies should also ensure best practices are applied in terms of randomization, allocation, and blinding methods to enhance evidence quality.

This first review of eHealth-based psychological interventions for CVD patients found that these interventions, particularly ICBT, significantly reduced depressive symptoms but not anxiety, with unclear effects on stress, HRQoL, and cardiovascular events. Research is urgently needed to assess the long-term impact, scalability, and accessibility of these interventions, which could play a transformative role in improving psychological well-being in this population. Collaboration among researchers, policy makers, and multidisciplinary cardiovascular health professionals is critical to advancing this field. Incorporating persuasive design elements, such as goal setting and motivational feedback, harnessing the temporal (ie, time-sensitive) intervention attributes of eHealth, and involving mental health professionals will be key to maximizing the effectiveness of eHealth approaches for CVD care.