Multiple Risk Factor Interventions for Primary Prevention of CVD in LMIC A Cochrane Review

This study sought to determine the effectiveness of multiple risk factor interventions aimed at modifying major cardiovascular risk factors for the primary prevention of cardiovascular disease in low- and middle-income countries (LMIC). We searched electronic databases for randomized controlled trials of health promotion interventions to achieve behavior change. The pooled effect indicated a reduction in systolic blood pressure ( (cid:2) 6.72 mm Hg; 95% con ﬁ dence interval [CI]: (cid:2) 9.82 to (cid:2) 3.61; I 2 ¼ 91%), diastolic blood pressure ( (cid:2) 4.40 mm Hg; 95% CI: (cid:2) 6.47 to (cid:2) 2.34; I 2 ¼ 92%), body mass index ( (cid:2) 0.76 kg/m 2 ; 95% CI: (cid:2) 1.29 to (cid:2) 0.22; I 2 ¼ 80%), and waist circumference ( (cid:2) 3.31 cm; 95% CI: (cid:2) 4.77 to (cid:2) 1.86; I 2 ¼ 55%) in favor of multiple risk factor interventions. There is some evidence that multiple risk factor interventions may lower blood pressure levels and anthropometrics in populations in LMIC settings at high risk of hypertension and diabetes. Structured summary Provide a structured summary including, as applicable: background; objectives; data sources; study eligibility criteria, participants, and interventions; study appraisal and synthesis methods; results; limitations; conclusions and implications of key ﬁ ndings; systematic review number.

Many low-and middle-income countries (LMIC) are now experiencing epidemiological transition, the change from a burden of infectious diseases to chronic diseases [1], due to dramatic changes in diet and lifestyle. The epidemiological transition in LMIC is happening in a shorter time frame than that experienced historically by highincome countries [2]. Urbanization and consumption of unhealthy diets are the main causes of this epidemic in LMIC [2e4]. In addition, LMIC are dealing not only with the emerging burden of noncommunicable diseases, but also with the current burden of infectious diseases [5e8]. Therapeutic lifestyle modification, including increasing physical activity, changing eating habits, and eliminating addictions, has been seen as a cornerstone of therapy for managing people with metabolic syndrome [9]. Lifestyle modifications have been shown to decrease the incidence of type 2 diabetes mellitus by 58% among people with impaired glucose intolerance [10,11] and significantly lowered systolic blood pressure between À5.4 and À11.4 mm Hg [12]. Therapeutic lifestyle interventions have been found to be at least as effective as pharmacotherapies [13] at little cost and with minimum risk [14]. In contrast to most pharmacotherapies, lifestyle modifications can also prevent or control other chronic conditions [10,15]. However, it has been suggested that in order for therapeutic lifestyle modification to be effective, it is important to pay attention not just to a single cardiovascular risk factor but to several factors simultaneously [16]. It is therefore generally recommended that lifestyle modifications should be implemented as a group [17].
A comprehensive Cochrane review has examined the effectiveness of multiple risk factor interventions in all settings, predominantly high-income countries [18] and found that "counselling and education interventions designed to change health behaviors do not reduce total or coronary heart disease mortality or clinical events in general populations, but they may be effective in reducing mortality in high-risk hypertensive and diabetic populations." This Cochrane review [18], in which most studies were based in high-income countries, concluded that health promotion interventions have limited use in general populations. Caution is needed in generalizing evidence from high-income countries to the current LMIC context because of the differences in settings and the nature of the communities, as well as the targeted populations. The objective of this review was to determine the effectiveness of multiple risk factor interventions (with or without pharmacological treatment) aimed at modifying major cardiovascular risk factors for the primary prevention of cardiovascular disease (CVD) in LMIC [19].

Protocol and registration
This systematic review's rationale and methods were specified in advance and documented in a protocol that was published in the PROSPERO (International Prospective Register of Systematic Reviews) (http://www.crd.york. ac.uk/PROSPERO/CRD42015019312) [20].
increasing activity levels in adult populations (!18 years of age); conducted in LMIC; and reported at least of 1 the following outcomes: 1) combined fatal and nonfatal CVD events (including myocardial infarction, unstable angina, need for coronary bypass grafting or percutaneous coronary intervention, stroke, peripheral artery disease); 2) adverse events; 3) all-cause mortality; 4) changes in CVD risk factors (blood pressure, lipid levels, diabetes, and obesity); and 5) changes in health knowledge, attitudes, and intention.

Information sources and search strategy
We identified trials through systematic searches of the following bibliographic databases (from January 1, 1950 to June 27, 2014): Cochrane Central Register of Controlled Trials; MEDLINE; EMBASE Classic þ EMBASE; Science Citation Index Expanded (SCI-EXPANDED); Conference Proceedings Citation Index-Science (CPCI-S) on Web of Science; Database of Abstracts of Reviews of Effects; LILACS (Bireme); Global Health; and ELDIS. We adapted the preliminary search strategy for MEDLINE (Ovid) for use in the other databases (Online Appendix 1). We checked the reference lists of all primary studies and review articles for additional references.

Study selection
Two authors (O.A.U. and L.H.) independently screened the titles and abstracts of all the potential studies we identified as a result of the search and coded them as "retrieve" (eligible or potentially eligible/unclear) or "do not retrieve." In case of any disagreements, we asked a third author (K.R.) to arbitrate.

Data abstraction
We used a data collection form for study characteristics and outcome data, which had been piloted on at least 1 study in the review. One author (O.A.U.) extracted study characteristics from the included studies. Two authors (O.A.U. and L.H.) independently extracted outcome data from the included studies. We resolved disagreements by consensus or by involving a third author (K.R.). One author (O.A.U.) transferred data into the Review Manager 5 software (Cochrane Informatics and Knowledge Management Department, Copenhagen, Denmark). We doublechecked that data had been entered correctly by comparing the data presented in the systematic review with the study reports. A second author (L.H.) spot-checked study characteristics for accuracy against the trial report.

Assessment of risk of bias in included studies
Two authors (O.A.U. and L.H.) independently assessed risk of bias for each study. We resolved any disagreements by discussion or by involving another author (K.R.). We assessed the risk of bias according to the following domains: random sequence generation; allocation concealment; blinding of outcome assessment; incomplete outcome data; selective outcome reporting; and other bias. We graded each potential source of bias as high, low, or unclear.

Measures of treatment effect
We used Review Manager 5 to manage the data and to conduct the analyses. We reported dichotomous outcomes as risk ratios (RR) with 95% confidence intervals (CI). For continuous outcomes, we calculated mean differences (MD) with 95% CI when the studies use the same scale. We included cluster-randomized trials in the meta-analysis along with individually randomized trials. Clusterrandomized trials are labelled with a (C). For clusterrandomized trials to be included in the meta-analyses, we adjusted for design effect using an "approximation method" [21]. The "approximation method" entailed calculation of an "effective sample size" for the comparison groups by dividing the original sample size by the "design effect," which is 1 þ (MÀ1) ICC, where M is the average cluster size and ICC is the intracluster correlation coefficient. For dichotomous data, we divided both the number of participants and the number who experienced the event by the same design effect, whereas for continuous data, only the sample size was reduced (means AE SD were left unchanged). We used the following reported ICCs for calculating the "design effects" (DE) [22]:

Data synthesis
We summarized and analyzed all eligible studies in Review Manager 5. Two authors (O.A.U. and L.H.) extracted the data; the first author entered all data and the second author checked all entries. We resolved disagreements by discussion. We undertook meta-analyses only where this was meaningful, that is, if the treatments, participants, and the underlying clinical question were similar enough for pooling to make sense. We combined the data using a random-effects model, due to anticipated heterogeneity that may result from the differences in methodology and study settings. We used the I 2 statistic to measure heterogeneity among the trials in each analysis [23]. When we identified substantial heterogeneity (I 2 value >50%), that is, more than 50% of the variation is due to heterogeneity rather than chance [24], we reported it and explored possible causes by pre-specified subgroup analysis. We used funnel plots and Egger tests [25] to assess potential small-study biases and publication bias for those outcomes with more than 10 trials (i.e., systolic and diastolic blood pressure).

Study selection and characteristics
The literature searches yielded 13,468 potentially relevant articles after duplicates were removed (Figure 1, Online Appendix 2). After scanning titles and abstracts, we identified 413 potentially relevant articles and assessed full-text copies against the inclusion criteria. Of these, 13 randomized controlled trials met the inclusion criteria [9,22,26e36]. Online Table 1 presents details and reasons for exclusion for the studies that most nearly missed the inclusion criteria. The characteristics of the included studies are summarized in Table 1. Where this was reported, the trials were conducted between 2001 and 2010 and were published between 2004 and 2012. Three trials were conducted in Turkey [27,29,33]. Two trials each were conducted in China [35,36] and Mexico [9,26]. One trial recruited participants from both China and Nigeria [22]. The other trials were conducted in Brazil [28], India [30], Pakistan [31], Romania [32], and Jordan [34]. The randomization unit for most trials was individual participants [9,26e30,32e36]. Two trials used cluster randomization (primary care facilities [22] and households [31]). Only 2 trials [27,36] recruited participants from healthy or general population. Most trials (n ¼ 11) recruited high-risk groups: known hypertensive people [22,26,29,31,33]; pre-hypertensive people [9]; metabolic syndrome [32,34]; obese participants [28]; and people with impaired glucose regulation [30,35]. The content of the interventions varied across the trials. Most of the trials included dietary advice and advice on physical activity. The follow-up period ranged from 6 months to 30 months (mean 13.3 months).

Risk of bias in included studies
The risk of bias of included studies is shown in Figure 2. The generation of allocation sequence was adequate in 4 trials [29,31,34,36], unclear in 7 trials [9,22,26e28,30,35], and inadequate in 2 trials [32,33]. Avram et al. [32] and Hacihasanoglu et al. [33] used the calendar date for generating allocation sequence. Allocation concealment was adequate in 1 trial [29], inadequate in 2 trials [32,33], and unclear in the remaining 10 trials. Four trials [27,29,31,32] masked outcome assessors to treatment allocation and 1 trial [33] did not. It is not clear whether the remaining trials masked outcome assessors to treatment allocation. The potential risk of bias likely to be introduced by incomplete data was high in only 1 trial [28], unclear in 3 trials [22,30,32], and low in the remaining 9 trials. The risk of selective reporting bias was unclear in Avram et al. 2011 [32], and low in the remaining 12 trials. The risk of bias likely to be introduced by other potential sources of bias was low in 2 trials [22,31] and unclear in the remaining 11 trials.

Effects of interventions
Combined cardiovascular events. One trial [30] reported cardiovascular events as an outcome. There was no significant difference between intervention and control groups in the rates of cardiovascular events (RR: 0.57; 95% CI: 0.11 to 3.07; 232 participants). This result is imprecise (wide confidence interval and small sample size) and makes it difficult to draw a reliable conclusion.
Blood pressure. Systolic blood pressure and diastolic blood pressure were reported in 11 trials (5,106 participants randomized) [9,22,26,28e31,33e36]. The pooled effect showed a statistically significant reduction in systolic blood pressure (MD: À6.72 mm Hg; 95% CI: À9.82 to À3.61; 4,868 participants) ( Figure 3) in favor of multiple risk factors interventions, but with evidence of statistically significant substantial between-trial heterogeneity (I 2 ¼ 91%; p ¼ 0.0001). There was no evidence of funnel plot asymmetry for systolic blood pressure (Figure 4), suggesting no evidence of small-study bias (p ¼ 0.270 for the Egger regression asymmetry test). In a pre-specified subgroup analysis, the pooled intervention effect estimate tended to be more pronounced   Similarly, the pooled effect showed a statistically significant reduction in diastolic blood pressure (MD: À4.40 mm Hg; 95% CI: À6.47 to À2.34; 4,701 participants) ( Figure 3) in favor of multiple risk factors interventions, but with evidence of statistically significant substantial between-trial heterogeneity (I 2 ¼ 92%; p ¼ 0.0001). There was no evidence of funnel plot asymmetry for diastolic blood pressure (Figure 4), suggesting no evidence of small-study bias (p ¼ 0.446 for the Egger regression asymmetry test). In a pre-specified subgroup analysis, the pooled intervention effect estimate tended to be more pronounced among high-risk groups (MD: À4.55; 95% CI: À7.26 to À1.85; 10 trials, 2,739 participants) than in the general population (MD: À3.18; 95% CI: À3.90 to À2.46; 1 trial, 1,962 participants); however, this difference did not reach a statistically significant level (p ¼ 0.34 for interaction). Kisioglu et al. [27] found no statistically significant difference between intervention and control groups in the rate of high blood pressure (RR: 0.87; 95% CI: 0.54 to 1.40; 400 participants).

Glycosylated hemoglobin (hemoglobin A 1c )
One trial [35] reported glycosylated hemoglobin as an outcome. There was no statistically significant difference between the intervention and control groups in mean change from baseline percentage hemoglobin A 1c (MD: À0.08%; 95% CI: À0.38 to 0.22; 181 participants).

Fruits and vegetables consumption
One trial [22] (2,166 participants randomized) reported increased fruit and vegetable consumption as an outcome.

Summary of main results
This review of multiple risk factor interventions for primary prevention of CVD in LMICs has brought together  jg REVIEW evidence from 13 randomized controlled trials primarily from the last 10 years, incorporating 7,310 participants. We found that evidence for effects on CVD events was scarce, with only 1 trial reporting these. We found that multiple risk factor interventions have an effect on some risk factors, especially on systolic blood pressure, diastolic blood pressure, body mass index, and waist circumference. However, the risk factor changes associated with interventions should be interpreted with caution. The meta-analyses of risk factor changes were highly heterogeneous, making pooled estimates of effect questionable. The observed risk factor changes associated with multiple risk factor interventions were modest, but are probably spurious as attributions of effect are inherently difficult to demonstrate in these interventions. These apparent reductions in risk factors may well be due to several factors, including failure to carry out intention-to-treat analysis owing to losses to follow-up, regression to the mean, nonblinded assessment of outcomes, etc. [18]. Furthermore, there are many problems in relating trial outcome to a risk measure that is itself dependent on the outcome in meta-analysis [37]; it is not possible to separate the benefits of the use of antihypertensive drugs in this set of trials because trials that included participants at high risk of developing CVD are more likely to include participants with high rates of use of antihypertensive drugs [18].

Study limitations and strengths
Overall, the studies included in this review were at some risk of bias, and the results should be treated with caution. We found statistically significant heterogeneity in all the meta-analyses of changes in CVD risk factors, thus suggesting that the percentage of the variability in effect estimates that is due to heterogeneity rather than to sampling error (chance) is important. The heterogeneity may be due to differences in study follow-up, geographical location, baseline differences in blood pressure values, and content of the multiple risk factor interventions. We conducted a comprehensive search across major databases for multiple risk factor interventions. We also screened systematic review reference lists, and we contacted trial authors when necessary. Two authors independently carried out all screening, inclusion and exclusion, data abstraction, and data entry and analysis. It is unlikely that the methods used in the review could have introduced bias.

Comparison with similar studies
Ebrahim et al. [18] conducted a Cochrane review to assess the effects of multiple risk factor interventions for reducing total mortality, fatal and nonfatal coronary heart disease events, and cardiovascular risk from factoring, among adults assumed to be without clinical evidence of previous coronary heart disease. The review included 55 trials that enrolled 163,471 participants and found that "interventions using counselling and education aimed at behavior change do not reduce total or [coronary heart disease] mortality or clinical events in general populations but may be effective in reducing mortality in high-risk hypertensive and diabetic populations" [18]. Another recent systematic review [12] examined the effects of lifestyle-related interventions on blood pressure in LMIC. The review included 8 multiple-intervention trials (defined as more than 1 lifestyle-related intervention delivered at the same time)  Heterogeneity: Tau 2 = 1.10; Chi 2 = 6.72, df = 3 (P = 0.08); I 2 = 55% Test for overall effect: Z = 4.47 (P < 0.00001) and found that the studies combining physical activity and diet or behavioral counselling interventions significantly reduced both systolic blood pressure (pooled MD: À6.1 mm Hg; 95% CI: À8.9 to À3.3) and diastolic blood pressure (pooled MD: À2.4 mm Hg; 95% CI: À3.7 to À1.1) [12]. Joshi et al.
[38] conducted a cluster randomized trial in rural Andhra Pradesh to develop, implement, and evaluate 2 CVD prevention strategies (clinical and health promotion interventions). The health promotion intervention included posters, street theater, rallies, and community presentations designed to increase the knowledge of the adult population about stopping tobacco use, heart-healthy eating, and physical activity [38]. The trial found no detectable effect of the health promotion interventions on the primary outcome of knowledge about 6 lifestyle factors affecting CVD risk and on both systolic and diastolic blood pressures [38]. The trial was excluded from this review because they reported no usable outcomes for the meta-analyses.  jg REVIEW CONCLUSIONS Due to the limited evidence available, currently we can draw no conclusions as to the effectiveness of multiple risk factor interventions on combined CVD events and mortality. Risk factor modification programs may be effective in altering risk factors in people living in LMIC. However, the evidence comes from studies at some risk of bias, and there was statistical variation between the results of the studies. There is a paucity of randomized controlled trials looking at the effects of multiple risk factor interventions for the primary prevention of CVD events and mortality over the long term. Therefore, there is a need for well-designed randomized controlled trials to fill this research gap. Further research is also needed to identify which components of multiple risk factor interventions, which modes of delivery, and which settings are key for an effective multiple risk factor program. State the process for selecting studies (e.g., screening, eligibility, included in systematic review, and, if applicable, included in the meta-analysis).

5
Data collection process 10 Describe method of data extraction from reports (e.g., piloted forms, independently, in duplicate) and any processes for obtaining and confirming data from investigators.

Data items 11
List and define all variables for which data were sought (e.g., PICOS, funding sources) and any assumptions and simplifications made.

5
Risk of bias in individual studies 12 Describe methods used for assessing risk of bias of individual studies (including specification of whether this was done at the study or outcome level), and how this information is to be used in any data synthesis.