Community-Based Interventions for Cardiovascular Disease Prevention in Low-and Middle-Income Countries: A Systematic Review

Objectives: To synthesize evidence on the effectiveness of community-based interventions for cardiovascular disease (CVD) prevention in low- and middle-income countries (LMICs) to inform design of effective strategies for CVD prevention. Methods: We searched MEDLINE, EMBASE, CINAHL, Cochrane register of controlled studies and PSYCINFO databases for studies published between January 2000 and June 2019. Other studies were identified from gray literature sources and review of reference lists of included studies. The primary outcomes for the review were those aimed at primary prevention of CVD targeting physical activity, diet, smoking and alcohol consumption. Results: Database searches yielded 15,885 articles and 94 articles were identified through snowball searching. After screening, the articles from LMICs were 32 emanating from 27 studies: 9 cluster randomized trials, eight randomized controlled trials and 10 controlled before and after studies. Community-based interventions successfully improved population knowledge on CVD and risk factors and influenced physical activity and dietary practices. Evidence of interventions on smoking cessation and reduced alcohol consumption was inconsistent. Conclusion: This evidence should inform policy makers in decision-making and prioritizing evidence-based interventions.


INTRODUCTION
Cardiovascular disease (CVD) continues to disproportionately cause morbidity and mortality in low-and middle-income countries (LMICs). Of the 17.9 million CVD related deaths reported worldwide in 2016, 75% occurred in LMICs [1]. In many LMICs, epidemiological transition, industrialization, infectious diseases burden and globalization have influenced changes in lifestyle observed through changes in physical activity, diet, alcohol and smoking behavior among others [2]. These lifestyle changes have contributed to the upsurge in CVD metabolic risk factors such as obesity, hypertension and diabetes [2]. It is estimated that over half of the 671 million obese population in the world live in 10 countries, eight of which are LMICs [3]. Moreover, the number of people living with diabetes in LMICs is estimated to rise to 228 million by 2030 from 84 million in 1995 [4]. A 2015 systematic review reported a pooled prevalence of hypertension in LMICs of 32.3% (95% CI: 29.4-35.3) [5] while a prevalence of 57.0% (95% CI 52-61%) was reported in another review among the African population aged 50 years and above [6].
Targeting lifestyle factors such as physical inactivity, poor diets, smoking and alcohol intake, and metabolic risk factors including dyslipidemia, hypertension and diabetes can reduce the overall burden of CVD [7]. Community-based interventions target change among individuals, groups, and organizations to avoid development of CVD risk factors or control them and often incorporate strategies to create policy and environmental changes [8,9]. Through community-based interventions, reduction in CVD burden and risk can be achieved in entire communities impacting population level knowledge and perceptions and risk reduction practices [10,11]. Population level public health measures are also likely to be more cost effective [12] than treatment oriented programmes for which most LMICs lack capacity to implement on a large scale [13].
Although community-based interventions aimed at CVD prevention have been implemented in LMICs, gaps remain regarding their effectiveness in these settings. Previous reviews on community based interventions have not been specifically directed to LMICs [14,15], evaluated only a few of the interventions or outcomes [16][17][18], or do not include recent evidence [19]. This systematic review was aimed at providing upto-date and comprehensive evidence on the effectiveness of community-based interventions for CVD prevention to support the design of effective strategies for CVD prevention. This review therefore answers two key research questions considering LMICs: 1) What community-based interventions and strategies have been implemented for CVD prevention in LMICs? 2) What is the effectiveness of community-based interventions for CVD prevention in LMICs?

METHODS
This study was conducted and reported in accordance with the Preferred Reporting Items for Systematic reviews and Meta-Analysis Protocols (PRISMA-P) 2015 statement (S1 Checklist) [20]. The study protocol was registered in the Prospero International prospective register of systematic reviews (Registration Number: CRD42019119885).

Eligibility Criteria
This review included studies conducted between January 2000 and June 2019 to obtain recent relevant evidence on communitybased interventions for cardiovascular disease prevention to inform policy and practice applicable in the current social dynamics. Studies were included if they met the criteria below: • Study population: Studies conducted among adults aged 18 years and above. • Intervention: Studies that reported interventions carried out within the community for either primordial or primary prevention of CVD aimed to improve cardiovascular risk knowledge and healthy lifestyle such as physical activity, healthy dietary habit, cessation of smoking and alcohol consumption. These interventions include health education and promotion, community mobilization, lifestyle counseling and coaching, screening, and treatment. The studies ought to have been implemented in a community setting including households, workplaces, schools, sport centers, pharmacies, primary health care units, community health worker posts among others but not secondary care health facilities. Interventions that started at health facilities and later linked to the community were included. • Comparator: studies where intervention was compared with another intervention, usual care or nothing. • Outcomes: The primary outcomes were changes in knowledge regarding CVD, physical activity, diet, smoking, and alcohol consumption. Among studies that had at least one primary outcome, secondary outcomes of body weight, systolic and diastolic blood pressure, blood glucose and lipid levels were reported. Studies among patients with CVD conditions and those whose aim was not to prevent CVD or their risk factors were excluded. • Study designs: Individual level or cluster randomized controlled trials, controlled before and after and controlled interrupted time series studies. • Context: This was a broad review that was not restricted to any geographical location. However, this article includes only filtered studies conducted in LMICs as defined by the World Bank Gross National Income per capita, calculated using the World Bank Atlas method as of June 2019. • Language: This review was restricted to articles published in the English language. • Other considerations: We included only studies that had a sample size of at least 150 participants, a follow-up period of at least nine months and a participant attrition rate of less than 40% to minimize bias from included studies. We also excluded duplicate publications, systematic or narrative reviews, reviews, abstracts, letters to the editor, comments, case reports, conference presentations and study protocols.

Information Sources and Search Strategy
We searched MEDLINE, EMBASE, CINAHL, Cochrane register of controlled studies and PSYCINFO. Other sources of publications including thesis online, OpenGrey, ProQuest, Google Scholar and the World Health Organization (WHO) International Clinical Trials registry platform were searched. In addition, we searched reference lists of included studies and similar systematic reviews for potential eligible studies to include in this review. A comprehensive search strategy relating to the population, intervention, and outcomes was developed in MEDLINE (S2 Search strategy) and adjusted to suit other databases. Appropriate limits were applied and only studies conducted between January 2000 to June 2019 were retrieved. The search was however repeated before submission of the systematic review article to include any newer articles. Articles from all comprehensive searches of databases and gray literature and those obtained from reference lists of other articles were exported as EndNote files (including titles and abstracts) and then imported into EndNote as a single library. Duplicate articles from the searches were verified and removed. The remaining articles were imported into rayyan.QCRI.org [21], a web-based tool that facilitates screening and collaboration among researchers, for screening.

Screening and Data Extraction
Screening was conducted at two levels, title and abstract and full text independently in rayyan. QCRI by two reviewers (RN and HH) using defined criteria and in case of any disagreements, a third reviewer (HB or GM) made the final decision. We contacted authors by email in case of any key missing information in the articles. All reasons for exclusion of articles were noted and the review process was summarized within the PRISMA flow chart ( Figure 1) highlighting the process of screening articles [22]. Data extraction was also independently conducted by two of the authors (RN and HH), thereby extracting all relevant information from included full text articles into a standardized Excel spreadsheet and later comparing and resolving any discrepancies. Data were extracted on study and participant characteristics, context, study design, methodology, intervention characteristics, comparator group(s) and outcome measures. For the outcomes, any effect estimates and observed changes in knowledge about CVDs, uptake of physical activity and diet, or reduction in smoking and alcohol use were recorded and data on secondary outcomes also extracted. Results of studies presented in multiple papers for the same population were only included in the review once.

Risk of Bias Assessment
The risk of bias of included studies was assessed using the revised Cochrane tool for Risk of Bias (RoB2) for randomized studies [23] and the Risk of Bias In Non-randomized Studies-of Interventions (ROBINS-I) for non-randomized studies [24]. The risk of bias assessment was conducted independently by two reviewers (RN and HH) who resolved any differences through consensus and if necessary, after consultation with a third reviewer (HB or GM).

Strategy for Data Synthesis
Data for this review were synthesized narratively while answering the aforementioned research questions. Findings have been descriptively presented and discussed while elaborating about the interventions and primary and secondary outcomes. Data have been presented in tabular form for comparison highlighting country, year of study, study objective, intervention, context, population and outcomes among others.

Search Results
Databases searched for this review yielded 15,885 articles. On top of that, 94 articles were identified through gray literature and snowball searching. At title and abstract screening, we retained 805 articles that underwent full text screening yielding 124 articles that met the inclusion criteria. Of the 124 articles, 32 were from LMICs representing 27 studies and were included in this review. A flow chart including details of the article screening process is shown in Figure 1.

Risk of Bias
Among the included studies, the risk of bias categorization was: low (7), moderate (7), some concerns (9) and high/serious (4). The major sources of bias within included studies were: outcome measurement bias due to limited use of objective measures or validated tools, missing data bias due to omission of missing data or not applying appropriate analytical techniques, and the lack of control for some confounders for non randomized studies. Figures 2-4 show the risk of bias graphs and their summary drawn using the visualization tools by McGuinness and Higgins [57]. Overall, the randomized controlled and cluster randomized controlled trials had a lower risk of bias compared to the controlled before and after studies.

Knowledge on CVDs and Risk Factors
Six studies half of which had moderate/some concerns RoB [40,46,47] and the rest had a high/serious RoB [38,45,49] examined changes in CVD knowledge following implementation of community-based interventions. In five of the studies, knowledge significantly improved in the intervention groups related to dietary and lifestyle factors [38,47,49], hypertension [38,46,49] and diabetes [40] compared to the control groups. In another study in China, although tobacco related knowledge increased, diet and physical activity knowledge decreased in the intervention compared to the control group [47]. In a study with a high RoB carried out in India, there was no statistically significant effect of the intervention on knowledge about the six lifestyle factors affecting CVD risk [45]. Interventions that were effective in enhancing CVD knowledge majorly involved training, community mobilization, health education and consultation delivered through campaigns, group meetings, workshops, use of mobile technologies and of health workers, community health workers or peers.

Physical Activity
A total of 23 studies recorded changes in physical activity across study populations. Among these, 16 studies compared improvements in physical activity between the intervention FIGURE 2 | Risk of bias graph (above) and summary (below); authors' judgements about each risk of bias item presented as percentages across all included randomized controlled trials, Community-based interventions for cardiovascular disease prevention in low-and middle-income countries: a systematic review, low-and middle-income countries, 2000-2019.         and control groups and eight with low (1), moderate/some concerns (7) RoB found changes in favor of the intervention group in engaging in physical activity [26,41,44], leisure-time activity [44], proportion of participants physically active [30,32,51], adherence to physical activity [50] and energy expenditure for physical activities [26,39]. In one study with a serious RoB, significant improvements were higher in the control group [49] while there was no significant difference between groups in the remaining studies all of which were of low, moderate or some concerns RoB [34,37,42,43,[54][55][56]. In seven studies, differences in physical activity were tested within groups and significant improvements were noted in the intervention group [26,27,47,48], both intervention and control groups [46] or none of them [36,38,40]. Some studies compared physical activity levels between males and females and found more activity among males than females [41] and higher activity levels among females compared to males [26]. Majority of effective studies utilized health education through mass media, public places, and information, education and communication materials, training of health workers and community volunteers, community mobilization through campaigns and structural changes in the environment.

Diet
Twenty-one studies determined changes in diet following implementation of interventions. Fruits and vegetable consumption significantly improved among intervention compared to control group in five studies of low [34], some concerns [40], moderate [26,55], or high RoB [38] unlike others [32,33,44,45,56]. In one study with some concerns RoB, the changes were only observed in fruit intake [42]. Several studies also showed significant differences in favor of the intervention in lower salt consumption [41,49,55], fat intake [26,45,49] and snacks consumption [42] when compared with control groups. Other studies measured overall diet quality and found significantly greater improvements in diet score or adherence to recommended diets in intervention populations relative to control group [37,48,50]. In six studies, there were no significant differences between the intervention and control groups in dietary components examined [30,36,46,47,51,56]. Effective dietary interventions mostly focused on providing advice on healthy diets, community health education during meetings and public places, community mobilization, tailoring interventions based on readiness to change and follow up of persons such as through phone calls.

Alcohol Use
Eight studies explored outcomes related to alcohol consumption as a risk factor for CVD [32,34,41,42,46,49,50,56]. In two studies of low and serious RoB, intervention group participants exhibited a greater reduction in alcohol consumption compared to control participants [34,49] while in another study of moderate RoB, significant differences were observed in both intervention and control communities [41]. Effective interventions for alcohol reduction were mainly communitybased peer support with group sessions, community-wide activities such in public places, health education through pamphlets, newspapers, classes and training of health workers. In the remaining studies with some concerns RoB, there was no significant reduction in proportion of persons consuming alcohol [42,46,56] or adhering to its moderate use [50]. On the other hand, significant reduction in alcohol use at population level and among patients with hypertension was observed in the control group of a study with moderate RoB conducted in Kenya [32].

Smoking
Among the 19 studies that examined changes in smoking levels, nine with low (2), moderate/some concerns (4), high/serious (3) RoB found a statistically significant difference in smoking prevalence in favor of the intervention compared to the control group [26,35,38,39,43,47,52,53,55]. Indeed, these differences were in smoking cessation [26,35,39,47,52,53,55] and reduced use of smokeless tobacco [38,43]. In two controlled before and after studies with moderate RoB, stratified analysis found significantly greater reductions in smoking cessation among men but not women [26,55]. Effective smoking interventions involved health education through social gatherings, pamphlets, brochures and booklets; community mobilization through training, involvement of local leaders and use of media campaigns; and group or individual counseling. In ten studies with low (3), moderate (2), some concerns (3), high (1) and serious (1)

Body Weight
Fourteen studies assessed changes in body weight among study participants with six of low (2), moderate (1) and some concerns (3) RoB reporting significant reduction in body mass index [42,44,46], waist hip ratio [48], body weight [30,36,42,44] and abdominal obesity [30] in intervention compared to control groups. Eight studies of low (3), some concerns/moderate (4) and high (1) RoB did not find any significant difference between groups [34,37,39,41,45,50,51,55]. Health education groups, health management programmes including training and counseling, nutrition education classes, training and engaging community volunteers were effective interventions for reducing body weight indicators.

Blood Pressure
Following implementation of interventions among the population groups, compared with the control groups, in seven of the 16 studies with low (1) and some concerns/ moderate (6) RoB, the intervention groups registered significant reductions in systolic [39,48,56] and diastolic blood pressure [30,51] or both [41,50]. In addition, two studies with high/serious RoB demonstrated a positive effect of the intervention on awareness, treatment and control of hypertension [49] and adherence to antihypertensive drugs [43]. Studies that showed effectiveness in blood pressure outcomes involved several strategies including health education awareness raising through mass media and public places, trainings, lifestyle promotion campaigns, health management programmes providing information, skills and tools for self-management, counseling sessions and improving access to blood pressure screening. There was no significant difference observed between intervention and control groups in systolic [32,43,51] and diastolic [39,48] blood pressure or both [34,37,40,42,45], and hypertension prevalence [30,32,40] and control [39,40,46] even though within group differences were observed in some of the studies.

Blood Glucose
The ten studies that evaluated blood glucose related outcomes with low (4) and some concerns/moderate (6) RoB found significant reductions in fasting blood sugar [30,48,50,51,54], and prevalence of diabetes and intermediate hyperglycaemia [40], risk of developing diabetes [34] and incidence of diabetes [26,36,37] in the intervention compared to control group. In one study with a low RoB, there was no significant difference in the incidence of diabetes between the intervention and control groups [34]. In another study with some concerns RoB, although differences were observed in the community mobilization arm, the m-health intervention arm did not influence the combined prevalence of intermediate hyperglycaemia and diabetes or the incidence of diabetes [40].
Interventions among effective studies included health education and advice, trainings, nutrition education through classes and print media such as brochures and pamphlets, provision of information and tools for self-management, secondary prevention activities in clinics and counseling.

Lipids
Seven studies evaluated changes in blood lipids and in four studies with low (1) and some concerns/moderate (3) RoB, significantly lower measures were registered for low density lipoprotein (LDL) [30,37], triglycerides [30,46] and total cholesterol [30,46,50] and a high prevalence of high density lipoprotein (HDL) [51] in intervention compared to control groups. Conversely, within the intervention group of a study in Iran with moderate RoB, prevalence of low HDL cholesterol increased at follow-up [30]. Health education sometimes personalized for individuals to support lifestyle changes or conducted in groups through classes and print media or mobile messages, health management programmes, and secondary prevention activities in clinics formed effective interventions. There was no statistically significant difference between intervention and control groups in total cholesterol [34,46], LDL cholesterol [34,46], serum cholesterol [37] and triglycerides [34,37,48] in other studies.

DISCUSSION
This systematic review examined community-based interventions for CVD prevention in LMICs and their effectiveness. Among the 27 studies that were included in the review, most employed health education and awareness creation, trainings, lifestyle consultation and counseling and community mobilization. Community-based interventions successfully improved population knowledge on CVD and risk factors and influenced physical activity and dietary practices unlike reduction in smoking and alcohol consumption. The interventions also led to significant improvements in blood pressure and blood sugar measurements. Overall, the effective interventions mostly involved community mobilization and social activities; health education and communication through use of information, education and communication materials including mass media; individual or group counseling; and trainings of providers including community health workers, peers or health workers.
Health education and health communication includes verbal and written measures to influence and empower individuals, populations, and communities to make healthier choices [58]. Many studies included more than one intervention strategy similar to other reviews [59] which is desirable as multistrategy interventions have been demonstrated to be more effective in influencing uptake of behaviors compared to individual strategies [58]. These strategies aim at changing people's knowledge, attitudes and/or behaviors and are considered a crucial first step in addition to community mobilization activities that help create a favourable enabling environment [58]. In addition, with the huge knowledge and capacity gaps in CVD services delivery in many LMICs [60][61][62], the high number of training interventions is not surprising.
The review also examined the effectiveness of communitybased interventions in increasing knowledge and uptake of CVD prevention practices. We found significant improvements in knowledge on behavioral and metabolic risk factors for CVD prevention in favor of the community-based interventions similar to a previous review [14]. These improvements were achieved mostly through training, community mobilization, health education and consultation [38,40,45,46,49]. Worthy of note is that knowledge on CVD is still very low in many LMICs [63,64] and thus interventions are more likely to lead to significant changes. Regarding, changes in behavioral practices following implementation of intervention, greatest improvements were around physical activity with increase in proportion physically active and leisure-time activity. Also, improvements in dietary practices were observed with changes reported in fruit and vegetable consumption, salt intake, and fats and snacks intake. An earlier review also observed improvements in physical activity and diet as significant lifestyle changes reported by a sizable proportion of studies [19]. Improvements in other risk factors were inconsistent with only few studies reporting changes in smoking and alcohol consumption behaviors. In high-income contexts, changes in smoking have been reported following community programmes [59]. Changes in alcohol and smoking practices usually takes a longer time yet many studies in our review had shorter follow-up periods of one to two years. Moreover, sustaining changes in behavior over long periods has been shown to be challenging [65,66]. In some studies, the intensity of the intervention was low with broad interventions and self-reported outcome measures which gaps should be bridged in future studies. Moreover, there is need for further studies of longer durations to examine changes in CVD prevention practices and inform the growing evidence base for CVD prevention intervention effectiveness in LMICs. This is especially important because lifestyle interventions have been reported to be more cost effective than pharmacological interventions [67] which most LMICs cannot afford amidst the growing epidemic of NCDs [13]. When planning such intervention studies, multiple intervention strategies should be considered, adapted to context and informed by appropriate theories, frameworks and models and evaluated using robust study designs.
This review also examined changes in metabolic risk factors as secondary outcomes and found more evidence in favor of reduction in systolic blood pressure, diastolic blood pressure and blood glucose following community-based interventions which was from relatively lower ROB studies. Effectiveness of community-based intervention in reducing metabolic risk factors has been reported previously for several risk factors including systolic blood pressure [14,19,68,69], diastolic blood pressure [69], incidence of diabetes [70] and HbA1C [70] including in high-income countries. Positive changes in behavioral risk factors should lead to improvements in metabolic risk factors for example changes in dietary behaviors and/or physical activity should impact blood pressure indicators. Although changes in metabolic risk factors were examined for only included studies as secondary outcomes and thus not comprehensive, this review provides insights into how community-based interventions impact both primary and secondary level CVD risk factors.
Overall, this systematic review notes that there have been interventions for CVD prevention in LMICs but these are mostly limited to middle-income Asian countries which could have been influenced by their higher CVD burden. The lack of studies from low-middle income sub-Saharan African countries is concerning and calls for systematic approaches to address the gaps. Targeted research funding and establishing dedicated research centers to support evidence generation and translation is recommended. Although majority of the studies were randomized controlled (17), only seven had a low risk of bias which affects the strength of currently available evidence. The design of future studies should consider measures to minimize identified sources of bias where possible including using objective measures of outcomes and/or using validated tools, properly designing interventions to avoid foreseeable deviations and applying appropriate techniques to deal with missing data. Moreover, randomized studies should ensure that randomization is effectively done to minimize baseline imbalances in study characteristics while non-randomized studies should control for most sources of confounding. We note that although the review intended to provide a comprehensive overview of community-based interventions and literature search widely done, the strict inclusion criteria designed to provide robust evidence on intervention effectiveness could have limited this. This review only considered studies published in the English language due to inadequate resources which could have led to publication bias hence future reviews should consider including other languages. We also excluded conference abstracts which           could have provided recent studies. However, such abstracts sometimes do not contain adequate information and their results may be inconclusive. The heterogeneity of the outcome measures also meant that a meta-analysis was not possible. Nevertheless, this systematic review provides evidence on the effectiveness of community-based interventions for CVD prevention in LMICs and their effectiveness in improving knowledge and uptake of healthy lifestyles in addition to changing metabolic risk factors essential for CVD prevention.

Conclusion
This review found several community-based interventions implemented for CVD prevention in LMICs which significantly influenced knowledge about CVD and risk factors, and changes in physical activity and dietary behaviors for CVD prevention. However, evidence on reducing smoking and alcohol consumption were inconsistent necessitating further research. Regarding the CVD metabolic risk factors, communitybased interventions significantly impacted blood pressure and blood sugar measurements. The most effective interventions utilized community mobilization, health education and information sharing, individual or group counseling, and trainings of providers. Evidence from this review can inform policy makers in decision-making and prioritizing evidencebased interventions for CVD prevention in LMICs.

FUNDING
This work was supported by the SPICES project in Uganda which received funding from the European Commission through the Horizon 2020 research and innovation action grant agreement No 733356 to implement and evaluate a comprehensive CVD prevention program in five settings: a rural and semi-urban community in a low-income country (Uganda), middle income (South Africa) and vulnerable groups in three highincome countries (Belgium, France and United Kingdom). The funder had no role in the design, decision to publish, or preparation of the manuscript.