NCBI Bookshelf. A service of the National Library of Medicine, National Institutes of Health.

Seely D, Kanji S, Yazdi F, et al. Dietary Supplements in Adults Taking Cardiovascular Drugs [Internet]. Rockville (MD): Agency for Healthcare Research and Quality (US); 2012 Apr. (Comparative Effectiveness Reviews, No. 51.)

  • This publication is provided for historical reference only and the information may be out of date.

This publication is provided for historical reference only and the information may be out of date.

Cover of Dietary Supplements in Adults Taking Cardiovascular Drugs

Dietary Supplements in Adults Taking Cardiovascular Drugs [Internet].

Show details


The methods for this Comparative Effectiveness Review follow the methods suggested in the Methods Reference Guide for Effectiveness and Comparative Effectiveness Reviews, Version 1.0 published by the Agency for Healthcare Research and Quality (AHRQ).17 Unless stated otherwise, all methods and analyses were determined a priori and documented in a research protocol that was publicly posted by AHRQ for comments. Screening of literature and inclusion/exclusion of studies were tracked and presented according to the PRISMA methodology.124

Topic Development and Refinement

This evidence report addresses several Key Questions regarding the effects of concomitant use of specific dietary supplements and cardiovascular drugs compared with cardiovascular drugs alone or with different dietary supplements. The original topic nomination was for a comparative effectiveness review (CER) of the risks and benefits for elderly patients taking cardiovascular medication concomitantly with herbal supplements. Discussions were held with Key Informants representing the FDA, the National Center for Complementary and Alternative Medicine (NCCAM), and several other institutions on scope of the topic, research questions to be asked, and methodology of evidence synthesis to be adopted. The panel included clinicians (e.g., cardiologists, naturopathic doctors, a clinical pharmacology specialist, and a nutritionist), a patient (consumer), and systematic review research methodologists.

As preliminary searching indicated scarce data in the elderly subgroup, it was decided with the Key Informants to broaden the review topic to the benefits and risks of dietary supplement use in adults taking drugs for prevention and treatment of cardiovascular disease (CVD). Subgroups of interest were added to a preliminary list of demographic and clinical categories that included those with renal dysfunction and genetic polymorphisms in CYP2D6, 2C9, and 2C19. Drug-functional food and drug-conventional food interactions were considered but it was decided that should be evaluated as a separate research project due to the size of such an undertaking. Five research Key Questions were finalized in the topic refinement process regarding the comparative efficacy, effectiveness, harms and pharmacokinetics of a dietary supplement coadministered with a cardiovascular drug. Subquestions sought to investigate subgroup effects and evidence of specific drug-supplement statistical interaction. An indirect question enquired about evidence based on human studies that some of the most commonly used dietary supplements cause alterations in cytochrome P450 isozyme activity and in cellular drug transport mechanisms.

All dietary supplements, according to the Dietary Supplement Health and Education Act of 1994 (DSHEA) definition, were of investigational interest. The dietary supplement was defined as one of the following substances:

  • a vitamin;
  • a mineral;
  • an herb or other botanical;
  • an amino acid;
  • a dietary substance for use by man to supplement the diet by increasing the total dietary intake (e.g., enzymes or tissues from organs or glands); or
  • a concentrate, metabolite, constituent or extract.

Furthermore, it must also conform to the following criterion:

  • intended for ingestion in pill, capsule, tablet, powder or liquid form not represented for use as a conventional food or as the sole item of a meal or diet.

AHRQ’s Effective Health Care Program posted the proposed Key Questions for public comment on their Web site from August 16 through September 13, 2010. In response, the following five general comments were received: (1) the review scope should be restricted to the most common dietary supplements and cardiovascular drugs; (2) the review should focus on patient-oriented outcomes; (3) the review should examine issues related to quality, dose and purity of the dietary supplements of interest; (4) the review should distinguish between regular and occasional users of dietary supplements; (5) the review should distinguish between nutrient and non-nutrient supplements. The Evidence-based Practice Center (EPC), with input from Key Informants and the Technical Expert Panel (TEP), reviewed and refined the Key Questions to ensure that the questions were specific and explicit about what information was being reviewed.

Protocol and Project Scope Amendment

The review according to the approved topic refinement document was proven to require resources beyond those allocated to the EPC center. The search of included data bases yielded more than 32,000 records; beyond expectations founded on targeted search during topic refinement. A random 10 percent title, abstract and full text screening was conducted to estimate the final size of the review and the projected number of included studies was beyond review resources. Several telephone and e-mail discussions were held with the TEP members and the Task Order Officer. Finally, three modifications were made as an amendment to the original research protocol based on the topic refinement document:

  • Restrict to dietary supplements commonly used in adults and elderly taking cardiovascular medication for which current evidence on possible drug-supplement interaction is lacking. With input from the TEP, and surveys of the general and cardiovascular populations in the Unites States reported in literature,7,1823 the revised list of dietary supplements was arrived at as follows:
    1. dietary supplements commonly perceived to provide cardiovascular benefit, with a high probability of being used simultaneously with cardiovascular drugs: coenzyme Q10; garlic; ginger; Ginkgo biloba; ginseng; multivitamins; vitamin A or beta carotene; vitamin D; vitamin E;omega-3 fatty acids or fish oils; niacin; and magnesium.
    2. other dietary supplements commonly used by the population using cardiovascular drugs, for which there is some reason to believe there may be an interaction with cardiovascular drugs: Echinacea; St. John’s wort; red yeast rice; resveratrol; hawthorn; and vitamin K.
    3. supplements falling into categories A or B may be omitted from the list if there is already well understood interaction and a review would be redundant. On this basis, it was proposed that these be excluded: St. Johns wort; therapeutic doses of niacin (greater than 250 mg/day); and magnesium as infusion or injection.
  • Eliminate the indirect question enquiring about alterations in cytochrome P450 isozyme activity and in cellular drug transport mechanisms.
  • Restrict foreign-language report inclusion to German only.

Language restriction was based on the fact that during the screening of the 10 percent random sample of a total of 32, 000 records, most commonly identified foreign languages were German (24 percent), Russian (16 percent), and Chinese (44 percent), comprising 84 percent of reports in language other than English. Other languages included Norwegian (4 percent), and Italian, Polish, Swedish, Dutch, and French (2 percent each). Because of uncertain applicability of Russian and Chinese language studies to the United States population and settings, the only foreign language considered to be eligible was German.

The list of relevant dietary supplements, rationale for the above-mentioned project scope amendments, and other details, are presented in the Table 3 below.

Table 3. Protocol amendments.

Table 3

Protocol amendments.

Development of the Analytic Framework

The analytic framework depicts the causal pathways forming the basis of the Key Questions (Figure 1). The framework outlines the conceptual basis or expectations for adding a dietary supplement to prescribed cardiovascular drugs. Expectations include improvement in efficacy and reduction in harms related to cardiovascular drugs. Outcomes that represent these changes may be clinical, surrogate (proxy for clinical), or pharmacokinetic. Altered outcomes may be a result of an add-on effect of the dietary supplement and/or supplement-drug interaction. Drug interactions may be seen as biologic (i.e., pharmacodynamic or pharmacokinetic interactions reflected in altered clinical and surrogate outcomes of benefit and harms) or statistical.125 Statistical interactions measure biologic interactions in terms of a product term added to a linear model. The Key Questions posed therefore examine whether outcomes of benefit and harms change with the addition of a dietary supplement. In addition, they examine pharmacokinetic and statistical interactions.

Figure 1. Analytic framework. This is an analytical framework of the key questions. The figure describes the pathway by which administration of dietary supplements to the target population (i.e. adults taking cardiovascular drugs commonly used in outpatient settings) may lead to changes in pharmacokinetic parameters, intermediate outcomes or biological effects (i.e. lipids, blood pressure, electrocardiogram measurements, other serum markers, and other diagnostic tests), clinical outcomes (i.e. mortality, ischemic heart disease, arrhythmias, other heart disease, nonfatal cerebrovascular disease, peripheral vascular (arterial) disease, cardiovascular disease surgery and procedures, quality of life, and others), and harms.

Figure 1

Analytic framework. Abbreviations: CVD = cardiovascular disease; ECG = electrocardiography; KQ = Key Question.

In the context of the population of interest, the Key Questions are indicated along the corresponding arrows connecting treatment and outcomes of interest. The framework includes seven sections. The first section represents the adult population, within which the target population is included. The remaining six sections are: (1) target population (adults taking cardiovascular drugs commonly used in the outpatient setting); (2) treatment (dietary supplements); (3) intermediate outcomes/biological effects (lipid levels, blood pressure, electrocardiogram measurements, other serum markers, and diagnostic tests); (4) clinical outcomes (mortality, ischemic heart disease, arrhythmias, cerebrovascular disease, peripheral arterial disease, cardiovascular disease surgery and procedures, and quality of life); (5) harms (allergic reactions, significant bleeding, and neurological and gastrointestinal adverse events); and (6) pharmacokinetic outcomes (measures of drug absorption, distribution, metabolism, and excretion).

Literature Search Strategy

An experienced medical information specialist developed and tested the electronic search strategies by using a combination of controlled vocabulary and free text, in consultation with the team. An independent information specialist peer-reviewed the search strategies according to the PRESS checklist.126

To identify primary study reports (see Appendix A), we searched the following electronic databases: Ovid MEDLINE® In-Process and other Non-Indexed Citations, and Ovid MEDLINE® (1950 to Sept 1, 2011); Embase (1980 to Sept 1, 2011); Cochrane Library via the Wiley interface (Sept 1 2011) including Cochrane Database of Systematic Reviews, Database of Abstracts of Reviews of Effects, and Health Technology Assessment Database and Cochrane Central Register of Controlled Trials; International Bibliographic Information on Dietary Supplements (IBIDS) on October 26, 2010; and Allied and Complementary Medicine Database (AMED) 1985 to September 1, 2011. We used controlled vocabulary (e.g., “Dietary Supplements,” “Drugs, Chinese Herbal,” “Phytotherapy”) and keywords (e.g., nutritional supplements, garlic, ginger) in combination with controlled vocabulary and keywords related to cardiovascular agents. A broad range of controlled vocabulary was used to address the various synonyms associated with this topic, as well as to cover any evolutionary gaps associated with the introduction of certain vocabulary terms. Results were refined using filters for systematic reviews, RCTs, non-RCTs and observational studies, and safety. A more specific strategy related solely to herb–drug interactions was run in the same databases using only a systematic review filter.

To identify systematic reviews addressing research questions similar to the Key Questions outlined for this review, we used a slightly modified strategy from that indicated above, by including a systematic review filter (for MEDLINE and Embase) and limiting the search of the Cochrane Library to the Cochrane Database of Systematic Reviews, the Database of Abstracts of Reviews of Effects, and the Health Technology Assessment Database (see Appendix Y). The identification of relevant systematic reviews addressing any one of the Key Questions would serve the purpose of replacing a de novo process involving primary synthesis with existing systematic reviews.

We restricted our searches to human studies without imposing any language or date restrictions. We attempted to identify unpublished literature (including abstracts and conference proceedings) through searches of trial registries (e.g.,, Current Controlled Trials, Clinical Study Results, World Health Organization Clinical Trials), the Cambridge Scientific Abstracts (CSA) Conference Papers Index, and Scopus. Additional references were identified by scanning bibliographies of relevant systematic reviews and clinical trials. We also contacted Technical Expert Panel (TEP) members.

Through the Scientific Resource Center, we contacted the industry (see Appendix B for list of drug manufacturers contacted) for scientific information packets (SIPs) on drug-supplement interaction and requested:

  • a current product label;
  • randomized controlled trials, published or unpublished; and
  • observational studies, published or unpublished.

All identified citations were downloaded into a Reference Manager 12127 database for duplicate removal and early analysis. The unique references were then uploaded into DistillerSR,128 a review management Web application.

Study Selection

Process Description

The reviewers involved in initial screening attended a screening training session and pilot tested the screening forms. Titles were screened by one reviewer; all exclusions were independently screened by a second reviewer. A similar process of screening was followed to screen abstracts of studies that passed the title screen. Two reviewers independently screened the full text of all included records. Discrepancies were resolved by consensus.

Eligibility Criteria for Systematic Reviews

When available, topically relevant reviews were to be included to answer one or more of the Key Questions. As per the Cochrane Collaboration definition,129 a systematic review includes: a specific research question; a search strategy (e.g., sources such as electronic databases, period covered by the search); and methods used to assess the risk of bias of studies included in the review. Narrative reviews were excluded. We limited our review to those systematic reviews judged to be of “Good” quality (see below for how quality of a review was assessed). Reasons for exclusion were noted. We planned to replace de novo evidence synthesis with good quality systematic review evidence only when it was deemed current, obviating the need to update.

Eligibility Criteria for Primary Studies

We aimed to include hypothesis-testing as opposed to hypothesis generating studies (case-reports and series). We thus included experimental and observational comparative studies (i.e., those with independent controls) evaluating the benefits or harms of concomitant dietary supplement use in adults taking cardiovascular medications versus no dietary supplement (or other dietary supplement). In a post hoc decision, we expanded the study design criteria to allow inclusion of single arm controlled before-after pharmacokinetic studies in which posttreatment values were compared with baseline values. This was because we considered such evidence as a reasonable comparison between drug-supplement and drug alone for pharmacokinetic outcomes when an adequate washout period was employed. This review was limited to human studies with specific eligibility criteria presented in Table 4.

Table 4. Study inclusion criteria.

Table 4

Study inclusion criteria.

Data Extraction and Data Management

Prior to data abstraction, we iteratively developed and pilot tested a standardized data extraction form. One reviewer extracted relevant data from each study and a second reviewer independently verified data from a 10 percent random sample of studies. In contrast with data pertaining to other Key Questions, where errors were few and sporadic (less than 0.5 percent), we noted some missed harms outcomes; therefore a second reviewer verified that all reported outcomes data of interest were extracted. Before this harms data verification was carried out, we clarified understanding of the reporting of harms. For example, we decided that when authors reported regular laboratory harms surveillance by testing for liver enzymes, glomerular filtration rate, blood urea nitrogen and serum creatinine, and then disclosed that no adverse events were identified, zero patients with events of raised levels of these enzymes should be extracted. Likewise a statement that no thrombotic event was identified was extracted as zero patients with stroke, angina and myocardial infarction. We could not generate extractable harms data from reports stating for example, “no significant adverse events were observed,” “no adverse effects of clinical importance occurred,” or “all remaining adverse events were mild in severity.” We extracted all harms data, not merely those thought to be drug related. Discrepancies in data extractions were resolved through discussions or with a help of third reviewer. Data extractors were not blinded to study information. If a study was reported in multiple publications, we extracted data from the latest and/or most complete publication and supplemented it with data from companion publications, as appropriate. We sought additional information from authors, when necessary. Data were input into DistillerSR.128 During the data extraction process, one reviewer with clinical background rated study populations’ 10-year CHD risk as per Table 5 below, according to the National Cholesterol Education Program Adult Treatment Panel III (NCEP, ATP III) guidelines.24 When all participants were healthy non-smokers, study level 10-year CHD risk was categorized as low.

Table 5. Ten-year coronary heart disease risk strata used to categorize study participants.

Table 5

Ten-year coronary heart disease risk strata used to categorize study participants.

There was some additional attrition after studies were included during screening. Of the included studies, we extracted data when the following additional criteria were met:

  • The dietary supplement was added to one or more cardiovascular drug(s) that were all taken by 100 percent or the majority (at least 80 percent) of participants in randomized controlled trials;
  • Nonrandomized controlled trials, or observational studies reported effect estimates reflective of a supplement plus drug(s) versus drug(s) alone comparison (or with another supplement);
  • Studies reported, or data could be obtained from authors, for at least one relevant outcome;
  • The design of the study was lower in the hierarchy of evidence (i.e., nonrandomized experimental or observational study in presence of higher RCT evidence), and did not meaningfully add (by not being a longer-term or pragmatic study reporting conclusive results) to the already included evidence from a higher study design category;
  • Studies were on a cardiovascular drug not marketed in the U.S.A.;
  • Dosing of the dietary supplement was specified.

Remaining studies were transparently eliminated from data synthesis with reasons.

Extracted Data Elements

The following elements were extracted:

  • Study and report characteristics (first author, study design, study setting, duration of followup, year and language of publication, funding source, treatment sequence generation, treatment allocation concealment, use of blinding)
  • Population characteristics (inclusion/exclusion criteria, number of enrolled and analyzed participants, age, gender, ethnicity/race, health status, comorbidities, baseline nutrient exposures and/or background diet, methods used to assess baseline nutrient exposures, specific cardiovascular drugs and drug classes, drug doses, dose regimens, duration of treatment, potential confounders such as blood pressure, concomitant medication, smoking, lipid levels, number and reasons for withdrawals or drop-outs)
  • Intervention characteristics (name, brand, and country of manufacturer of a dietary supplement or extract, license/registration of the product in the country of manufacture, source from which a dietary supplement or extract was manufactured, method of authentication, dosage regimen and quantitative description, dose form, qualitative testing for authenticity of herbal species, purity assessment for contamination/substitution, standardization, storage conditions and length, methods and instruments for assessing nutrient-intake exposures including validation by using nutrient biomarkers)
  • Control (comparator) intervention characteristics
    • Placebo - description/definition in placebo-controlled trials, duration of treatment, similarity of treatments)
    • Other dietary supplement - same data as above
  • Outcomes
    • Continuous measures (mean baseline and final values, within-treatment arm mean change, between-treatment arm mean difference, standard deviation, if reported means – the corresponding variance, standard error, 95 percent confidence interval)
    • Binary outcomes (number of participants with an event, risk of an event, odds of an event, risk ratio, odds ratio, hazard ratio, variance, standard error, 95 percent confidence interval, crude or adjusted measures of effect)
    • Definitions
    • Measurement method(s)
    • Timing of measurement
    • Data analysis details
    • Statistical test used
    • Regression models (model type, covariates)

For systematic reviews, we planned to extract data on the research question, search strategy, design of individual studies included in a review, risk of bias assessment methods, population characteristics (inclusion and exclusion criteria, type of cardiovascular drug intake), and treatment characteristics (name and type of dietary of supplement).

Assessment of Study Risk of Bias and Quality of Systematic Reviews

Risk of bias of individual studies was assessed according to outcomes. One reviewer assessed risk of bias of all individual primary studies (randomized trials, nonrandomized controlled trials, cohort, and case-control studies). The risk of bias for gradable outcomes was verified by two reviewers, while a random ten percent of risk of bias assessment was verified for other outcomes. Disagreements were resolved by discussions with a third team member.

We used generic criteria to assess study risk of bias (Table 6). These criteria estimate risk of bias across five domains (selection bias, performance bias, attrition bias, detection bias, and other bias). The domains were tailored according to study design. For example, specific sub-domains such as randomization sequence generation were assessed only for randomized trials (parallel arm and crossover). Crossover randomized trials were assessed with the following additional criteria: appropriateness of crossover design; washout period (the length of time had to be at least 3 times the half life of drug elimination); and the report of appropriate data analysis (i.e., based on within-subject differences). The subdomain of blinding (within the Performance Bias domain) of participants, healthcare providers, and outcome assessors to treatment allocation was assessed only for experimental designs (i.e., randomized and non-randomized trials). We assessed the completeness of outcome data (i.e., attrition bias due to loss to followup or withdrawals) by comparing the number of participants who entered the study with the number of participants reported in outcome table(s). We then assessed whether there was complete followup of all participants, small lost to followup (less than 20 percent) or differential lost to followup.

Table 6. Criteria for risk of bias assessment.

Table 6

Criteria for risk of bias assessment.

Other forms of bias assessed included potential financial conflict of interest and selected criteria from the McMaster Quality Assessment Scale of Harms (McHarm).25 All domains had response options of “Yes,” “No,” or “Unclear” with allowance for justifications of such judgments. For each gradable outcome in a study we provided an overall risk of bias rating designated as high, moderate, or low (Table 7). In order to be classified as high risk of bias, a study must have demonstrated some apparent and major flaw (within that study design category) that would invalidate results.

Table 7. Overall risk of bias ratings.

Table 7

Overall risk of bias ratings.

Methodological quality of included systematic reviews was assessed using the AMSTAR tool,131 which rates each systematic review with a “Yes,” “No,” “Cannot answer,” or “Not applicable” across the 11 domains. The overall assessment of quality for each systematic review was based on a reviewer’s overall judgment given their responses to the individual AMSTAR items, and had three overall ratings: “Good,” “Fair,” or “Poor.” In general, “Good” quality systematic reviews were defined as those having few or no methodological/reporting shortcomings (low risk of bias). “Fair” quality systematic reviews were defined as those having some methodological flaws although not sufficient to seriously bias or invalidate the review results. “Poor” quality systematic reviews were defined as those having serious flaws sufficient to seriously bias or invalidate the review results, and were not eligible for inclusion in the evidence synthesis. An independent reviewer helped to resolve any discrepancies regarding the AMSTAR tool assessment between the reviewers.

Grading the Strength of the Body of Evidence

In principle, a body of evidence originating in randomized trials starts with a presumed high strength of evidence, and is downgraded across the domains when there is important overall risk of bias of contributing studies, inconsistency in direction of intervention effect, indirectness of the outcome of interest (e.g., a surrogate outcome, rather than a clinical health outcome) and imprecision in effect estimates of an extent that neither important benefit nor harm can be ruled out. For nonrandomized studies, the body of evidence starts with a presumed low strength of evidence but may be upgraded across certain domains. The strength of a body of evidence was graded based on the following four domains, per previously published guidance: overall risk of bias by outcome, consistency, directness, and precision.26 A methodologist and a content expert graded the strength of the body of evidence as “High,” “Moderate,” “Low,” or “Insufficient” (Table 8). A third methodologist with clinical background adjudicated to resolve disagreements. Given the results we found, optional domains such as dose-response association and existence of confounders were not applicable in this comparative effectiveness review. Given the uncertainties involved in interpreting asymmetry tests for publication bias in most reviews, especially in presence of heterogeneity in effect estimates, we did not plan to investigate publication bias in this review.102,103

Table 8. Strength of evidence grade and definition.

Table 8

Strength of evidence grade and definition.

The strength of evidence was graded insufficient when there was no evidence for an outcome, when direction of estimates were inconsistent between studies without an identifiable cause, or when the body of evidence from the contributing study/studies was underpowered for the outcome of interest (imprecise estimate). That is, when the effect estimate associated with confidence intervals was not only nonsignificant, but wide enough such that the clinical action would differ if the upper versus the lower boundary of the CI represented the truth, we rated the estimate as imprecise, reflecting our uncertainty about clinically important benefits, harms or clinically unimportant differences in effect estimates between the contrasting interventions.

Customarily only a subset of important outcomes are chosen to grade the strength of evidence—outcomes that are most meaningful for decision making given a specific Key Question.26 In consultation with the Technical Expert Panel (TEP), the review team chose the following outcomes for grading the strength of the body of evidence:

Key Question 1.

Mortality (all-cause and vascular death), myocardial ischemic events (fatal myocardial infarction, nonfatal myocardial infarction, unspecified myocardial infarction, and acute coronary syndromes), cerebrovascular events (hemorrhagic/ischemic/unspecified stroke), quality of life, hospitalization, arrhythmia, and clinical outcomes of peripheral arterial disease

Key Question 2.

Blood pressure (systolic and diastolic), lipid profile (low density lipoprotein, high density lipoprotein, and non–high density lipoprotein cholesterol; and triglycerides), international normalised ratio for coumarin derivatives, incidence of metabolic syndrome, and change in 10-year Framingham risk profile.

Key Question 3.

Serious adverse events (composite outcome according to the Food and Drug Administration definition of serious adverse events),27 withdrawal due to adverse events, clinical bleeding (intracranial, gastrointestinal, genitourinary, subretinal, etc.), renal dysfunction (e.g., proteinuria, elevated creatinine, need for transplant, glomerular filtration rate), hepatotoxicity (elevated enzymes or fulminant failure), and QT prolongation

Key Question 4.

Area under the plasma cardiovascular drug concentration-time curve (AUC), maximum drug concentration (Cmax), drug half-life(t1/2), and oral clearance.


We followed previously published guidance, and summarized the population, intervention, comparator, outcome, timing and setting (PICOTS) to assess the applicability of the body of evidence for outcomes or categories of similar outcomes.28 We considered age, race/ethnicity and gender representation; strictness of exclusion criteria; 10-year CHD risk; study setting; whether or not the cardiovascular drug(s) was administered in therapeutic doses; frequency of monitoring of adherence; surrogacy of outcomes; and short versus long-term treatment duration as important aspects determining applicability. Applicability is reported for all conclusive results. When direction of effect was indeterminate either because of lack of evidence or under powered evidence (with imprecise and nonsignificant data), applicability of evidence was not reported. Studies that evaluated representative patient populations in usual or routine care conditions lasting long enough to meaningfully measure health outcomes of both benefit and harm while comparing the intervention with standard of care were considered pragmatic or effectiveness studies as opposed to efficacy studies examining intermediate efficacy outcomes in highly selected patients less likely to experience harms.29

Data Synthesis and Analysis

All analyses compared the combination of dietary supplement plus cardiovascular drug with cardiovascular drug alone (or plus placebo or another dietary supplement). The decision to meta-analyze or qualitatively synthesize outcome specific evidence from primary studies depended upon the presence or absence of homogeneity in clinical and methodological characteristics across studies, and the statistical format of outcome reporting. For pharmacokinetic outcomes in Key Question 4, we followed the FDA guidance for analysis and interpretation of drug interaction studies.30 Because data are usually skewed, the guidance recommends that the pharmacokinetic outcomes in drug interaction studies be reported, after log transformation, as geometric mean ratios (GMR) with their 90 percent confidence intervals (CI) based on a procedure termed the ‘two one-sided test procedure.’ A conservative margin of bioequivalence for most drug interactions is recommended to be between the lower and upper bound 90 percent GMR CI of 0.8 and 1.25. We refer to this as the zone of clinical nonsignificance.

Meta-analysis was considered, or studies were considered suitable for pooling if they were randomized trials that included similar populations in terms of demographics, morbidity, and intake of cardiovascular drug(s) or classes of drugs (e.g., participants aged 65 years or younger, healthy or with diabetes or history of myocardial infarction, use of warfarin or nitrates), compared the same type of dietary supplement (e.g., niacin, oral magnesium, fish oil) versus comparator treatment (e.g., other dietary supplement, no treatment, placebo), and reported the same outcome measures in the same statistical format (e.g., mean difference or geometric mean ratios [GMR]). Where risk of bias differed across studies, this was explored by sensitivity analyses. We did not plan to meta-analyze observational and experimental studies; however, we pooled parallel with crossover randomized studies.32 We did not consider pre-crossover data for synthesis except when it was judged that the treatment given to participants in a given crossover trial was not appropriate for the condition under consideration.31,32 Similarly, we did not pool crossover trials that had not employed sufficient washout period between the two treatment periods because of bias arising from carryover treatment effects. We planned not to meta-analyze observational studies because of the differences in adjustment for confounders and residual confounding.

We used a DerSimonian and Laird random-effects model to generate pooled estimates of relative risk (RR) if an outcome was measured on a dichotomous scale, and weighted between-group mean difference (for end points and within group changes), if an outcome was measured on a continuous scale.33 The measure of variability of the pooled estimates was a 95 percent confidence interval. Statistical heterogeneity was assessed using Cochran’s Q (α=0.10) and the I2 statistic. All analyses were performed using Comprehensive Meta Analysis version 2.2.057 (New Jersey, USA); StatsDirect Ltd. StatsDirect statistical software,132 and R: A Language and Environment for Statistical Computing, Foundation for Statistical Computing.133

When studies failed to report summary statistics (e.g., mean score, standard deviation, standard error), we calculated the needed parameters if individual participant data were provided. If a study reported only a standard error of the mean response, we converted it into a standard deviation. Trials were not meta-analyzed if the mean and standard deviation could not be ascertained. Trials with obvious between-group baseline imbalance in a continuous outcome were not pooled unless the mean change from baseline and corresponding standard deviation for the compared study groups were reported.

We used the Peto odds ratio method when event rates were less than1 percent.32 For studies with zero events in some arms or sparse data overall, we pooled using the fixed effects Mantel-Haenszel method without continuity correction.34 Studies with zero events in both arms were excluded from meta-analysis.32

Outcome results were considered inconclusive when the pooled estimate or the single contributing study estimate had confidence intervals wide enough to incorporate both important benefit and harm (i.e., type II error suggesting underpowered studies unable to precisely conclude benefit, harm or no difference between treatments). When studies could not be pooled, for example when similar outcomes were reported in different statistical formats, or study results pointed in opposite directions, results were also labelled as inconclusive. When inconclusive results were associated with a gradable outcome, strength of evidence was deemed insufficient.

Exploration of heterogeneity was planned to examine clinical and methodological diversity, to answer subquestion (a) of the four Key Questions by carrying out subgroup and sensitivity analyses, and/or meta-regression if there was a sufficient number of studies (4 studies for each categorical subgroup variable, and 6 to 10 for a continuous subgroup variable).32 This was not carried out because of a paucity of studies for each outcome.

Prespecified clinical subgroups included: gender (male, female), ethnicity (including Hispanic, Asian, African American, and Native American), age (those age 65 years and above, and 80 years and above), baseline health status (healthy volunteers, participants according to coronary heart disease risk category, participants at risk for cardiovascular disease, participants with known cardiovascular disease, participants with diabetes, participants with hepatic or renal dysfunction or end-stage renal disease, participants taking a cardiovascular drug for an indication other than cardiovascular disease), and genotypic polymorphisms (e.g., in CYP2D6, 2C9, 2C19).

Sensitivity analyses were to be conducted to explore whether the effect estimates of dietary supplement treatment were influenced by methodological variables such as overall study risk of bias, adequacy of participant selection, important confounding, blinding of outcome assessors, purity/dose/stability of a dietary supplement, or duration of treatment or followup.

Statistical interaction was to be investigated for all four Key Questions. To determine measureable interactions between cardiovascular drugs and dietary supplements, we followed the procedures described below:

  • If a study explicitly examined interaction between cardiovascular drug and dietary supplement, then we extracted the study’s findings along with the method for assessing interaction.
  • When studies did not investigate statistical interaction between the dietary supplement plus cardiovascular drug group and the cardiovascular drug alone group, we carried out test for interactions when data for at least four groups were reported. These groups had to be supplement plus cardiovascular drug group, supplement alone group, cardiovascular drug alone group, and a common comparator placebo or no treatment group. For studies that presented dichotomous data we calculated the synergy index (S).35
Synergy Index (S)=1-RR111-RR01RR10

Where RR11 is the relative risk for subjects exposed to cardiovascular drug and dietary supplement, and RR01 and RR10 are the relative risks for subjects exposed to cardiovascular drug alone and dietary supplement alone respectively. An S-index greater than 1 describes a positive interaction (synergism) and an S-index less than 1 indicates a negative interaction (antagonism). For continuous outcomes, individual patient data were needed to have been modeled in a linear regression with an interaction term.


Pharmacokinetic Measurements: Absorption, Distribution, Metabolism, and Excretion. Adapted from Rowland and Tozer 1995134


The study of the mechanisms of absorption, distribution, metabolism and elimination of an administered drug in the living organism


Absorption can refer to any route of administration except intravenous administration (i.e., enteral, subcutaneous, transdermal).

  1. Fractional bioavailability (F): the fraction of a non-intravenously administered dose of a drug that reaches the systemic circulation (unit: percent)
  2. Time to maximum concentration (tmax): the time at which the maximum plasma concentration of a drug occurs following administration of an extravascular dose (unit: minute or hour)
  3. Maximum concentration (Cmax): the highest drug concentration in plasma after an extravascular dose (unit: μg mL−1)
  4. The area under the plasma drug concentration-time curve (AUC0-t): the measure of an exposure of plasma to a drug over a given time period. It is calculated as the area under the serum concentration-time curve which is based on multiple drug concentration measurements at various time points (unit: μg mL−1hour)


Volume of distribution (Vd) is the apparent volume into which a drug distributes in the body at equilibrium. Alternatively, it is the volume of plasma at the drug concentration required to account for the entire drug in the body (unit: L/kg) Volume is not limited to plasma. For example, volume of distribution will exceed plasma volume or total body water volume for highly lipophylic drugs.


  1. Clearance: the volume of plasma in the vascular compartment cleared of drug per unit time by the processes of metabolism and excretion (unit: mL/hour)
  2. The elimination half-life (t1/2): the time required to reduce the plasma concentration of a drug or the total amount of drug in the body to one half of its initial value (unit: hour)


  • PubReader
  • Print View
  • Cite this Page
  • PDF version of this title (13M)

Recent Activity

Your browsing activity is empty.

Activity recording is turned off.

Turn recording back on

See more...