Effects of Nutritional Supplements and Dietary Interventions on Cardiovascular Outcomes

Abstract

Current U.S. dietary guidelines recommend several healthy eating patterns, including U.S., Mediterranean, and vegetarian diets (1). Although the guidelines recognize the occasional need for nutritional supplementation or food fortification for specific nutrients that may be consumed in inadequate amounts, they do not recommend routine use of any dietary supplement to reduce risk for cardiovascular disease (CVD) or other chronic diseases. Despite these recommendations, most U.S. adults use supplements to enhance their diets, with uncertain health benefits (2, 3). From 1999 to 2012, the NHANES (National Health and Nutrition Examination Survey) reported that 52% of participants used at least 1 and 10% used at least 4 dietary supplements (4). From 2011 to 2014, the NHANES reported that among participants aged 60 years or older, 70% used at least 1 and 29% used at least 4 supplements, and 41% of supplement takers reported that they did so to improve their overall health (5). In 2013, the U.S. Preventive Services Task Force conducted a systematic review of the utility of vitamin and mineral supplements for CVD prevention and found little evidence to support use (6). More recently, Jenkins and colleagues published a meta-analysis of randomized controlled trials (RCTs) of dietary supplements published through October 2017 (7). They found some stroke benefit conferred by folate; no CVD benefit for multivitamins, vitamin C, vitamin D, or calcium; and evidence for mortality harm for niacin and antioxidants. Since then, several landmark RCTs evaluating the efficacy of fish oils (810) and vitamin D (11, 12) for CVD prevention have been published, which add to the evidence level. In addition, the quality of the evidence base of these various nutritional supplements and dietary interventions still needs to be evaluated to ascertain the confidence in their efficacy. Thus, we performed a systematic review of existing meta-analyses of RCTs and generated an evidence map for efficacy of nutritional supplements and dietary interventions for CVD prevention. Methods Search Strategy We used PubMed, CINAHL, and the Cochrane Library from inception to March 2019 to find meta-analyses published in the English language about vitamins, minerals, dietary supplements or products, and dietary interventions using the following search terms: (*minerals OR *vitamins OR *diet AND *cardiovascular outcomes) and (meta-analy* OR metaanaly* OR systematic review*). After selecting systematic reviews on the basis of a priori criteria, the search timelines of the systematic reviews were reviewed for recency and an updated search for RCTs published in English was performed starting from the end date of searches from selected systematic reviews until March 2019 (Supplement Table 1). Additional sources included Web sites of major cardiovascular and medicine journals (www.onlinejacc.org; https://academic.oup.com/eurheartj; www.ahajournals.org/journal/circ; www.nejm.org; https://jamanetwork.com; and http://annals.org/aim) and bibliographies of relevant studies. We also searched ClinicalTrials.gov (10 March 2019) to check for publication bias and to identify any new or ongoing trials (Supplement Table 2). Supplement. Supplemental Material. Study Selection The prespecified inclusion criteria were meta-analyses of RCTs assessing efficacy of nutritional supplements (vitamins, minerals, dietary supplements) or dietary interventions in adult participants (18 years) that report effect estimates for all-cause mortality and cardiovascular outcomes of interest and were written in English. Because the nutritional and dietary recommendations are universal, there were no restrictions on baseline health status, race, or sex of the participants. Meta-analyses of observational studies or those reporting efficacy of interventions on surrogate or other outcomes, such as blood pressure, lipid values, inflammatory markers, electrolytes, renal values, or quality-of-life indicators, were excluded. Systematic reviews reporting meta-analyses of both clinical trials and observational studies were reviewed for data related to RCTs only. In case of multiple meta-analyses of the same intervention and outcome, we preferred the most recent, largest, and updated meta-analysis. However, the competing meta-analyses were screened for any additional trials not included in the selected meta-analysis. After removing duplicates and following the selection criteria, we screened the retrieved articles at the title and abstract level and then at the methods level. The search, selection, and abstraction processes were performed independently by 2 authors (M.U.K. and S.V.). Any discrepancies were resolved by discussion and mutual consensus, referring to the original study or third-party review (S.U.K.). Data Extraction, Outcomes, and Quality Assessment We first extracted information from eligible meta-analyses on first author, journal, year of publication, interventions, outcomes of interest, number of trials, whether an appropriate study search and selection criteria was reported, method of pooling estimates (fixed or random effects), methods of detecting publication bias, measure of heterogeneity, and risk-of-bias assessment. Second, we generated the pool of clinical trials by identifying trials contained in the selected meta-analyses and screening competing meta-analyses for additional trials and trials published after the selected meta-analyses (Supplement Table 3). Among new clinical trials for omega-3 long-chain polyunsaturated fatty acid (LC-PUFA) (810), we excluded REDUCE-IT (Reduction of Cardiovascular Events With EPA-Intervention Trial) (9) because icosapent ethyl, a highly purified form of eicosapentaenoic acid (EPA), does not qualify as a dietary supplement according to the Dietary Supplement Health and Education Act of 1994 (13). Third, after removing duplicates, we abstracted data on trial name, first author, year, intervention, outcomes, raw events, and sample sizes for each group. The main outcome of interest was all-cause mortality. The secondary outcomes were cardiovascular mortality, myocardial infarction (MI), stroke, and coronary heart disease (CHD). Two independent reviewers (V.O. and M.S.K.) assessed the methodological quality of meta-analyses on specific potential factors that may affect the validity of summary estimatesthat is, appropriate search and selection criteria, number of trials and participants included, risk-of-bias assessment of included trials, method of pooling the estimates, assessment of publication bias, and degree of heterogeneity (Supplement Table 4). Data Synthesis and Analysis We created an evidence map that displays the plausible benefits of each intervention and the certainty of the evidence (14). The certainty of the evidence was evaluated using the GRADE (Grading of Recommendations Assessment, Development and Evaluation) approach (GRADEpro GDT) (https://gdt.gradepro.org/app/) (14) and was classified as high, moderate, low, or very low (Supplement Table 5). Two reviewers (V.O. and M.S.K.) performed these assessments under the supervision of a third reviewer (S.U.K.). Estimates were pooled according to MantelHaenszel random-effects model. The PauleMandel method was used for reestimating outcomes. HartungKnapp small-sample adjustments were applied when the number of studies was less than 10 (15). Effect sizes were reported as risk ratios (RRs) with 95% CIs. We used I 2 statistics to estimate the extent of unexplained heterogeneity; I 2 greater than 50% was considered a high degree of between-study heterogeneity. We calculated the Egger regression test as an estimate of publication bias for any reanalysis that included at least 10 studies (16). Statistical analyses were conducted using meta, version 4.9-4 (R Project for Statistical Computing). Statistical significance was set at 0.05 for all analyses except for the Egger regression test, which had a threshold less than 0.10 because of the test s limited statistical power. Role of the Funding Source The study received no funding. Results Search Results Of 942 citations, after removing duplicates and screening at the title and abstract level we reviewed 140 full-text articles for eligibility. We excluded 131 articles because they focused on nonrandomized studies, were not relevant, or were outdated, as well as 5 systematic reviews that assessed intake of nuts (17), fruits and vegetables (18), fiber (19), and green or black tea (20) and those focusing on low-carbohydrate and low-fat diets (21) that did not report cardiovascular outcomes of interest. Ultimately, we included 9 systematic reviews and 4 new RCTs for a total of 105 meta-analyses of 24 interventions (277 RCTs, 992129 participants) (7, 2229) (Figure 1). The interventions evaluated in the meta-analyses included 16 types of supplements (antioxidants, -carotene, vitamin B complex, multivitamins, selenium, vitamin A, vitamin B3 or niacin, vitamin B6, vitamin C, vitamin E, vitamin D, calcium plus vitamin D, calcium, folic acid, iron, and omega-3 LC-PUFA) and 8 types of dietary interventions (Mediterranean diet and intake of reduced saturated fat, modified dietary fat, reduced dietary fat, reduced salt among hypertensive and normotensive participations, increased omega-3 -linolenic acid [ALA], and increased omega-6 PUFA) (Supplement Table 6). Figure 1. Evidence search and selection. RCT = randomized controlled trial. Quality Assessment All included studies were trial-level meta-analyses (7, 2228), except the study by Mente and colleagues, which was a patient-level analysis of 4 studies (29) (Supplement Table 4). All trial-level systematic reviews reported comprehensive search and selection criteria as well as quality assessment of studies by using the Cochrane Risk of Bias Tool (30). Six systematic reviews primarily used random-effects models for meta-analyses, of which 4 used fixed-effects models for sensitivity analyses. Two studies primarily used fixed