Advertisement

The effects of foods on LDL cholesterol levels: A systematic review of the accumulated evidence from systematic reviews and meta-analyses of randomized controlled trials

  • Malin Schoeneck
    Affiliations
    Dalarna County Council, Norslund-Svärdsjö Academic Primary Health Care Center, Björkvägen 2, SE-790 23, Svärdsjö, Sweden
    Search for articles by this author
  • David Iggman
    Correspondence
    Corresponding author. Dalarna County Council, Norslund-Svärdsjö Academic Primary Health Care Center, Björkvägen 2, SE-790 23, Svärdsjö, Sweden.
    Affiliations
    Dalarna County Council, Norslund-Svärdsjö Academic Primary Health Care Center, Björkvägen 2, SE-790 23, Svärdsjö, Sweden

    Center for Clinical Research Dalarna, Falun, Sweden

    Unit for Clinical Nutrition and Metabolism, Department of Public Health and Caring Sciences, Uppsala University, Uppsala, Sweden
    Search for articles by this author
Open AccessPublished:January 15, 2021DOI:https://doi.org/10.1016/j.numecd.2020.12.032

      Highlights

      • Several novel foods that can help reduce LDL cholesterol were discovered.
      • These include flaxseeds, almonds, avocados, tomatoes, turmeric, and green tea.
      • LDL cholesterol can however be increased by unfiltered coffee and possibly by sugar.

      Abstract

      Aims

      To systematically evaluate the evidence regarding the effects of foods on LDL cholesterol levels and to compare the findings with current guidelines.

      Data synthesis

      From inception through June 2019, we searched PubMed, Cochrane Database of Systematic Reviews, and Cochrane Central Register of Controlled Trials for guidelines, systematic reviews, and RCTs (for coffee intake only) of at least 13 days duration. Additionally, we searched Trip database for guidelines from 2009 through Oct 2019. Language was restricted to English. The strength of evidence was evaluated using The Grading of Recommendations Assessment, Development, and Evaluation (GRADE). A total of 37 guidelines, 108 systematic reviews, and 20 RCTs were included. With high evidence, foods high in unsaturated and low in saturated and trans fatty acids (e.g. rapeseed/canola oil), with added plant sterols/stanols, and high in soluble fiber (e.g. oats, barley, and psyllium) caused at least moderate (i.e. 0.20–0.40 mmol/L) reductions in LDL cholesterol. Unfiltered coffee caused a moderate to large increase. Soy protein, tomatoes, flaxseeds, and almonds caused small reductions. With moderate evidence, avocados and turmeric caused moderate to large reductions. Pulses, hazelnuts, walnuts, high-fiber/wholegrain foods, and green tea caused small to moderate reductions, whereas sugar caused a small increase. Other identified foods were either neutral or had low or very low evidence regarding their effects.

      Conclusions

      Several foods distinctly modify LDL cholesterol levels. The results may aid future guidelines and dietary advice for hypercholesterolemia.

      Graphical abstract

      Keywords

      Introduction

      Lifestyle management remains the foundation for the prevention of cardiovascular disease (CVD). Dietary guidelines targeting healthy individuals [
      U.S. Department of Health and Human Services and U.S. Department of Agriculture
      2015 – 2020 dietary guidelines for Americans.
      ], dyslipidemia [
      • Mach F.
      • Baigent C.
      • Catapano A.L.
      • Koskinas K.C.
      • Casula M.
      • Badimon L.
      • et al.
      2019 ESC/EAS Guidelines for the management of dyslipidaemias: lipid modification to reduce cardiovascular risk.
      ], or CVD prevention [
      • Arnett D.K.
      • Blumenthal R.S.
      • Albert M.A.
      • Buroker A.B.
      • Goldberger Z.D.
      • Hahn E.J.
      • et al.
      2019 ACC/AHA guideline on the primary prevention of cardiovascular disease: a report of the American college of cardiology/American heart association task force on clinical practice guidelines.
      ] typically advocate healthy dietary patterns such as the Mediterranean diet. However, there has been a call for more food-based dietary advice [
      • Mozaffarian D.
      • Ludwig D.S.
      Dietary guidelines in the 21st century-a time for food.
      ]. The effects of individual foods on CVD outcomes are challenging to study in randomized controlled trials (RCTs), whereas effects on risk factors can be determined from strictly controlled interventions of shorter duration. Moreover, the dietary choices that influence the level of low-density lipoprotein (LDL) cholesterol are distinctly different from those concerning hypertriglyceridemia, which can be improved by omega-3 supplements and reduced intakes of alcohol or carbohydrates [
      • Mach F.
      • Baigent C.
      • Catapano A.L.
      • Koskinas K.C.
      • Casula M.
      • Badimon L.
      • et al.
      2019 ESC/EAS Guidelines for the management of dyslipidaemias: lipid modification to reduce cardiovascular risk.
      ]. This distinction may require clarification, as only LDL cholesterol is considered as a critical [
      World Health Organization & Mensink RP
      Effects of saturated fatty acids on serum lipids and lipoproteins: a systematic review and regression analysis. Geneva.
      ], causal, and cumulative risk factor [
      • Mach F.
      • Baigent C.
      • Catapano A.L.
      • Koskinas K.C.
      • Casula M.
      • Badimon L.
      • et al.
      2019 ESC/EAS Guidelines for the management of dyslipidaemias: lipid modification to reduce cardiovascular risk.
      ]; and represents the primary target for CVD risk reduction [
      • Mach F.
      • Baigent C.
      • Catapano A.L.
      • Koskinas K.C.
      • Casula M.
      • Badimon L.
      • et al.
      2019 ESC/EAS Guidelines for the management of dyslipidaemias: lipid modification to reduce cardiovascular risk.
      ]. The potential for improvement in LDL cholesterol through diet is also substantial, as demonstrated by studies combining several foods, e.g. the Portfolio diet, which emphasizes nuts, plant protein, plant sterols, and soluble fiber [
      • Kendall C.W.
      • Jenkins D.J.
      A dietary portfolio: maximal reduction of low-density lipoprotein cholesterol with diet.
      ]. In recent years, several additional foods have emerged as potentially effective. Thus, a detailed update would be valuable, both to caregivers and to individuals seeking to improve their cholesterol levels. This study aimed to systematically evaluate the evidence from RCTs reporting effects of foods on LDL cholesterol levels and to compare the findings with current guidelines.

      Methods

      The study contains both an umbrella review of guidelines and systematic reviews, and a systematic review and meta-analysis of RCTs.

      Data sources and searches

      Based on preliminary searches, seven foods were selected, for which evidence for effects was indicated but not considered unequivocal; garlic, coffee, tea, cocoa/chocolate, probiotics, nuts, and soy protein. Searches were then performed in two predetermined steps. First, PubMed and Cochrane Database of Systematic Reviews were searched for guidelines (considered separately) and systematic reviews concerning the effects of foods on LDL cholesterol levels. Second, for the selected foods above, we set out to perform our own systematic reviews (and meta-analyses) of RCTs, if they had not been included in any published systematic review with low risk of bias. This second search for RCTs was performed in PubMed and Cochrane Central Register of Controlled Trials. The search strategies were developed with assistance from librarians at Falu hospital and are described in Supplemental Table 1. The first search (for guidelines and systematic reviews) was performed on 13 March 2018 and the second search (for RCTs) on 2 May 2018. Both were updated on 4 June 2019. A complementary search for guidelines was performed in Trip database on 31 Oct 2019, as several guidelines had been recently updated. Abstracts were screened by both authors; in cases of disagreement, the paper was generally included for further review. The bibliographies of included RCTs were screened for other potentially relevant studies.

      Study selection

      For both searches, the target population was adults who were not treated with lipid-lowering medications. The interventions of interest were foods or nutrients related to specific foods; but not supplements, weight loss diets, or dietary patterns. The outcome was LDL cholesterol in mmol/L (conversion factor from mg/dL 0.02586). Comparison foods and no treatment were acceptable as controls. Language was restricted to English. Guideline documents should be publically available, related to dyslipidemia or CVD prevention, and be no older than 10 years. In the second search (for RCTs), at least a 13-day isocaloric intervention was required. No other restrictions were applied.

      Data extraction and risk of bias assessment

      Available data were extracted from articles read in full text. Authors were contacted if required. Risk of bias in systematic reviews was assessed (by both authors) using a modified version of the A MeaSurement Tool to Assess systematic Reviews (AMSTAR) tool. For low risk of bias grading, a reproducible search strategy was mandatory. Additionally, at most one of the following criteria was allowed omitted: having at least two independent reviewers; searching at least two databases; providing a complete list of excluded studies read in full text, with reasons; providing a table describing relevant details of included studies; and providing a risk of bias assessment of included studies. Narrative reviews and systematic reviews not classified as having low risk of bias were excluded from the qualitative summary. Excluded studies near the threshold were however mentioned in text (but not in main tables and figures) if their results could affect the overall conclusions. The RCTs found during the second search were assessed (by both authors) using the Cochrane Risk of Bias Tool 2.0. We did not rely on previously performed evaluations of the evidence. Instead, the Grading of Recommendations Assessment, Development and Evaluation (GRADE) strength of evidence was determined for each systematic review, by predetermined criteria for risk of bias, inconsistency, indirectness, imprecision, publication bias, large effects, dose–response relationship, and opposing bias. In marginal cases, discussions were extended until a consensus was reached, applying overall judgments. For each food, the effect and strength of evidence were then evaluated, taking into account all included systematic reviews, bearing in mind that some RCTs may be included in several different systematic reviews.

      Data synthesis and analysis

      Meta-analyses were conducted using Review Manager 5.3. In studies with several study arms, the weighted averages of means and standard deviations were calculated. The mean change from baseline was used whenever standard deviations were given or possible to calculate. Otherwise, the end-of-study means and standard deviations were used, at the latest available time point. Crossover studies were combined with parallel RCTs in the meta-analyses, with adjusted weights; in cases where the variances of mean differences were unavailable, a correlation R = 0.82 was imputed (the calculated mean from studies with available data), and sensitivity analyses were performed with a more conservative weighting of R = 0.5. If heterogeneity was considered low or moderate (I2<50%), fixed effects models were preferred. As sensitivity analyses, only the end-point data (with imputed data on standard deviations if missing) were used, parallel and crossover studies were analysed separately, and only one arm per study (the most relevant) was included. Estimated food doses were calculated as weighted means from included studies with reported data, using the weights from the corresponding meta-analyses.
      The study was conducted according to the Swedish Agency for Health Technology Assessment and Assessment of Social Services Method Handbook [
      Statens beredning för medicinsk och social utvärdering (SBU)
      Utvärdering av metoder i hälso- och sjukvården och insatser i socialtjänsten - en handbok. Stockholm.
      ] and the Cochrane Handbook for Systematic Reviews of Interventions [
      The Cochrane Collaboration
      Cochrane Handbook for systematic reviews of interventions version 5.1.0.
      ]. The review protocol is available at www.crd.york.ac.uk/PROSPERO, identifier: CRD42018089661. No ethics approval was required.

      Results

      The study selections in the two searches are presented as Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) flowcharts (Fig. 1A and B). Two hundred potential systematic reviews were read in full text, and 108 were included in the qualitative summary (Supplemental Table 2).
      Figure 1
      Figure 1A. Flow chart for the first search (for systematic reviews). B. Flow chart for the second search (for randomized controlled trials on coffee).
      Thirty-seven guideline documents were identified (Supplemental Table 3). Their recommendations were partly consistent but also included some variations. Several recommended a Mediterranean (advice included in 11 guidelines), Dietary Approaches to Stop Hypertension (DASH, 6 guidelines), or a similar dietary pattern; high in fruits and vegetables, nuts, non-tropical oils, legumes, whole-grain/high-fiber foods, and fish. Single guidelines recommended other dietary patterns, e.g. vegetarian/vegan, plant-based wholefood, Portfolio, low-fat, or Terapeutic Lifestyle Changes diet. Seven guidelines recommended foods high in or enriched with plant sterols and/or stanols. Seven guidelines recommended soy products. Restrictions in sugar and soft drinks (16 guidelines), trans fatty acids (17 guidelines) or saturated fatty acids (24 guidelines) were common. Some included restrictions in red or processed meats (14 guidelines), salt (9 guidelines), alcohol (8 guidelines), refined carbohydrates (6 guidelines) or dietary cholesterol (11 guidelines). Only one guideline advised against unfiltered coffee intake. Brief summaries of each guideline's dietary recommendations and evidence gradings (included in 15 guidelines) are listed in Supplemental Table 3. In addition, two WHO reports were found, which described the effects of saturated and trans fatty acids on LDL cholesterol. These were not published in peer-reviewed journals and did not fulfil the AMSTAR criteria for systematic reviews, but were still considered relevant for evaluation (see Fatty foods, below).
      All of the pre-specified foods except coffee were included in published systematic reviews with low risk of bias. Thus, coffee was the only food included in the second search for RCTs, from which 20 RCTs (1518 participants) fulfilled our inclusion criteria and were included in the qualitative summary (Table 1). Nineteen RCTs (1335 participants) were included in the meta-analyses (Fig. 2). The overall estimates of the effect and strength of and evidence for each food are summarized in Table 2. The effects of foods with moderate or high evidence are further visualized in Fig. 3. The excluded systematic reviews and RCTs are listed, with reasons, in Supplemental Tables 4 and 5.
      Table 1Characteristics of the 20 included RCTs that evaluated the effect of coffee on LDL cholesterol.
      Author, yearDesignInterventions/comparisonDaily dose (mean ± SD and/or range)DurationParticipants analysed, n overall (per group, % male)PopulationAge in years (mean ± SD and/or range)Baseline LDL-C (mean ± SD mmol/L)Risk of biasCountryFunding
      Agudelo-Ochoa [
      • Agudelo-Ochoa G.M.
      • Pulgarín-Zapata I.C.
      • Velásquez-Rodriguez C.M.
      • Duque-Ramírez M.
      • Naranjo-Cano M.
      • Quintero-Ortiz M.M.
      • et al.
      Coffee consumption increases the antioxidant capacity of plasma and has No effect on the lipid profile or vascular function in healthy adults in a randomized controlled trial.
      ], 2016
      ParallelThree groups: high vs medium chlorogenic acid coffee vs no coffee
      Filtered compared with no coffee.
      400 mL (780 mg chlorogenic acids in high and 420 mg in medium group)8 w74 (24 + 25 + 25, 51%)Healthy20–602.78 ± 0.59Some concerns (randomization)ColombiaN/A
      Ahola [
      • Ahola I.
      • Jauhiainen M.
      • Aro A.
      The hypercholesterolaemic factor in boiled coffee is retained by a paper filter.
      ], 1991
      CrossoverBoiled coffee vs boiled and filtered coffee
      Filtered compared with unfiltered coffee.
      6–10 dL4 w20 (15%)Healthy45 ± 83.8 ± 0.9Some concerns (randomization)FinlandFood Research Foundation
      Aro [
      • Aro A.
      • Kostiainen E.
      • Huttunen J.
      Effects of coffee and tea on lipoproteins and prostanoids.
      ], 1985
      CrossoverThree interventions: instant coffee vs instant tea vs rosehip ‘tea’a,b8 cups3 w12 (50%)Healthy33–453.02 ± 0.19Some concerns (randomization)FinlandN/A, Academy of Finland
      Aro [
      • Aro A.
      • Tuomilehto J.
      • Kostiainen E.
      • Uusitalo U.
      • Pietinen P.
      Boiled coffee increases serum low density lipoprotein concentration.
      ], 1987
      CrossoverThree interventions: boiled coffee vs filtered coffee vs teab,c8 cups4 w42 (50%)Hyper-cholesterolemic49 (31–60)6.04 ± 0.16Some concerns (randomization)FinlandFinnish Food Research Foundation
      Aro [
      • Aro A.
      • Teirilä J.
      • Gref C.
      Dose-dependent effect on serum cholesterol and apoprotein B concentrations by consumption of boiled, non-filtered coffee.
      ], 1990
      CrossoverBoiled coffee vs filtered coffeec5.7 (2–14) cups4 w41 (32%)Healthy45 (23–61)~3.5Some concerns (randomization)FinlandFood Research Foundation
      Bak [
      • Bak A.
      • Grobbee D.
      The effect on serum cholesterol levels of coffee brewed by filtering or boiling.
      ], 1989
      ParallelThree groups: boiled coffee vs filtered coffee vs no coffeea,c4–6 cups12 w101 (33 + 34 + 34, 48%)Healthy26 ± 43.2 ± 1.1Some concerns (randomization)NetherlandsCommittee on Physiological Effects of Coffee, Netherlands Prevention Fund and Netherlands Heart Foundation
      Corrêa [
      • Corrêa T.
      • Rogero M.
      • Mioto B.
      • Tarasoutchi D.
      • Tuda V.L.
      • César L.A.
      • et al.
      Paper-filtered coffee increases cholesterol and inflammation biomarkers independent of roasting degree: a clinical trial.
      ], 2013
      CrossoverMedium-light vs medium roast coffee
      Darker compared with lighter roast.
      3 or 4 cups4 w20 (30%)Healthy50 ± 93.1 ± 0.5Some concerns (randomization)BrazilFAPESP, National Council for Scientific and Technological Development
      D'Amicis [
      • D’Amicis A.
      • Scaccini C.
      • Tomassi G.
      • Anaclerio M.
      • Stornelli R.
      • Bernini A.
      Italian style brewed coffee: effect on serum cholesterol in young men.
      ], 1996
      ParallelThree groups: espresso vs mocha vs tea
      Coffee compared with tea.
      3.1 ± 1.2 vs 2.8 ± 1.1 cups (espresso, 25–35 mL/cup and mocha, 40–50 mL/cup)6 w84 (28 + 28 +28, 100%)Healthy soldiers27 ± 13.0 ± 0.6Some concerns (randomization)ItalyInstitute for Scientific Information on Coffee
      Dusseldorp [
      • van Dusseldorp M.
      • Katan M.B.
      • Demacker P.N.
      Effect of decaffeinated versus regular coffee on serum lipoproteins. A 12-week double-blind trial.
      ], 1990
      CrossoverRegular coffee vs decaffeinated coffee
      Regular compared with decaffeinated coffee.
      4–6 cups6 w45 (49%)Healthy38 ± 7 (25–45)N/ALowNetherlandsNetherlands Heart Foundation
      Dusseldorp [
      • van Dusseldorp M.
      • Katan M.B.
      • van Vliet T.
      • Demacker P.N.
      • Stalenhoef A.F.
      Cholesterol-raising factor from boiled coffee does not pass a paper filter.
      ], 1991
      ParallelThree groups: boiled coffee vs boiled and filtered coffee vs no coffee
      Filtered compared with no coffee.
      ,
      Filtered compared with unfiltered coffee.
      6 cups (0.9 L)79 d64 (22 + 21 +21, 52%)Healthy39 ± 83.4 ± 0.8Some concerns (randomization)NetherlandsNetherlands Heart Foundation
      Fried [
      • Fried R.
      • Levine D.
      • Kwiterovich P.
      • Diamond E.L.
      • Wilder L.B.
      • Moy T.F.
      • et al.
      The effect of filtered-coffee consumption on plasma lipid levels. Results of a randomized clinical trial.
      ], 1992
      ParallelFour groups: regular coffee in high vs low dose vs decaffeinated coffee vs no coffee
      Filtered compared with no coffee.
      ,
      Regular compared with decaffeinated coffee.
      720 mL in high dose and de-caffeinated groups, 360 mL in low dose group8 w100 (25 + 25 + 25 + 25, 100%)Healthy44 ± 103.3 ± 0.8Some concerns (randomization)USANational Coffee Association and Outpatient Clinical Research Center from the National Institutes of Health, Bethesda
      Kempf [
      • Kempf K.
      • Kolb H.
      • Gärtner B.
      • Bytof G.
      • Stiebitz H.
      • Lantz I.
      • et al.
      Cardiometabolic effects of two coffee blends differing in content for major constituents in overweight adults: a randomized controlled trial.
      ], 2015
      ParallelDark roast vs medium roast coffee
      Darker compared with lighter roast.
      4–5 cups3 mo114 (56 + 58, 34%)Overweight49 ± 123.5 ± 0.9Some concerns (randomization)GermanyTchibo GmbH
      Rosmarin [
      • Rosmarin P.
      • Applegate W.
      • Somes G.
      Coffee consumption and serum lipids: a randomized, crossover clinical trial.
      ], 1990
      CrossoverCoffee vs no coffee
      Filtered compared with no coffee.
      3.6 (2.1–6.7) cups2 mo21 (100%)Healthy, white35 ± 6 (22–45)3.2 ± 0.8LowUSAUniversity Physicians Foundation of the University of Tennessee
      Sanguigni [
      • Sanguigni V.
      • Gallu M.
      • Ruffini M.
      • Strano A.
      Effects of coffee on serum cholesterol and lipoproteins: the Italian brewing method. Italian group for the study of atherosclerosis and dismetabolic diseases, Rome II center.
      ], 1995
      CrossoverRegular moka vs decaffeinated moka
      Regular compared with decaffeinated coffee.
      3 cups5 w49 (51%)Healthy23 (21–28)2.7 ± 0.6Some concerns (randomization)ItalyN/A
      Shaposhnikov [
      • Shaposhnikov S.
      • Hatzold T.
      • Yamani N.E.
      • Stavro P.M.
      • Lorenzo Y.
      • Dusinska M.
      • et al.
      Coffee and oxidative stress: a human intervention study.
      ], 2018
      ParallelThree groups: high vs low dose coffee vs water
      Filtered compared with no coffee.
      5 (high dose, water) or 3 (low dose) cups8 w160 (53 + 53 + 54, N/A)Healthy (56% overweight)51 ± 12N/ASome concerns (randomization)NetherlandsKraft Foods and University of Olso
      Superko [
      • Superko H.
      • Bortz W.
      • Williams P.
      • Albers J.
      • Wood P.
      Caffeinated and decaffeinated coffee effects on plasma lipoprotein cholesterol, apolipoproteins, and lipase activity: a controlled, randomized trial.
      ], 1991
      ParallelThree groups: regular coffee vs decaffeinated coffee vs no coffee
      Filtered compared with no coffee.
      ,
      Regular compared with decaffeinated coffee.
      4.5 ± 1.1 (3–6) cups8 w181 (62 + 61 + 58, 100%)Healthy46 ± 103.7 ± 0.9Some concerns (randomization)USANational Institutes of Health
      Urgert [
      • Urgert R.
      • Meyboom S.
      • Kuilman M.
      • Rexwinkel H.
      • Vissers M.N.
      • Klerk M.
      • et al.
      Comparison of effect of cafetière and filtered coffee on serum concentrations of liver aminotransferases and lipids: six month randomised controlled trial.
      ], 1996
      ParallelCafetiere vs filtered coffee
      Filtered compared with unfiltered coffee.
      5–6 cups (0.9 L)24 w46 (22 + 24, 50%)Healthy29 ± 103.0Some concerns (randomization)NetherlandsNetherlands Heart Foundation through the Netherlands Organisation of Scientific Research
      Wahrburg [
      • Wahrburg U.
      • Martin H.
      • Schulte H.
      • Walek T.
      • Assmann G.
      Effects of two kinds of decaffeinated coffee on serum lipid profiles in healthy young adults.
      ], 1994
      ParallelThree groups: regular arabica coffee vs arabica decaffeinated coffee vs arabica/robusta decaffeinated coffee
      Regular compared with decaffeinated coffee.
      750–1000 mL6 w116 (39 + 39 + 38, 51%)Healthy students25 ± 33.4 ± 0.6Some concerns (randomization)GermanyInstitute for Scientific Information on Coffee
      Wedick [
      • Wedick N.M.
      • Brennan A.M.
      • Sun Q.
      • Hu F.B.
      • Mantzoros C.S.
      • van Dam R.M.
      Effects of caffeinated and decaffeinated coffee on biological risk factors for type 2 diabetes: a randomized controlled trial.
      ], 2011
      ParallelThree groups: instant regular coffee vs instant decaffeinated coffee vs water
      Filtered compared with no coffee.
      ,
      Regular compared with decaffeinated coffee.
      5 cups (885 mL)8 w45 (16 + 14 + 15, 36%)Healthy, overweight41 ± 132.5 ± 0.8LowUSABoston Obesity Nutrition Research Center and National Center for Research Resources
      Yamaguchi [
      • Yamaguchi T.
      • Chikama A.
      • Mori K.
      • Watanabe T.
      • Shioya Y.
      • Katsuragi Y.
      • et al.
      Hydroxyhydroquinone-free coffee: a double-blind, randomized controlled dose-response study of blood pressure.
      ], 2007
      ParallelFive groups: Regular coffee vs zero-dose vs low-dose vs middle-dose vs high-dose hydroxyhydro-quinone coffee1 cup (183 mL)4 w183 (37 + 37 + 37 + 35 + 37, 50%)Mildly hyper-tensive49 ± 103.3 ± 0.8Some concerns (randomization)JapanKao Corporation
      Abbreviations: mo – months, LDL – low-density lipoprotein, N/A – not available, RCT – randomized controlled trial, w – weeks.
      Included in corresponding meta-analyses (Fig. 2A–E):
      a Filtered compared with no coffee.
      b Coffee compared with tea.
      c Filtered compared with unfiltered coffee.
      d Regular compared with decaffeinated coffee.
      e Darker compared with lighter roast.
      Figure 2
      Figure 2Forest plots of the effect of coffee intake on LDL cholesterol. A. Filtered coffee compared with no coffee intake. B. Coffee compared with tea intake. C. Filtered compared with unfiltered coffee intake. D. Regular compared with decaffeinated coffee intake. E. Darker compared with lighter roast coffee intake. Units are in mmol/L. Ahola 1991, Aro 1985, Aro 1987, Aro 1990, Corrêa 2013, Dusseldorp 1990, Rosmarin 1990, and Sanguigni 1995 are crossover studies. Data are expressed as end-of-study means±standard deviations, means±standard deviations in absolute change from baseline, or mean differences±standard errors.
      Table 2GRADE table for the effects of foods on LDL cholesterol.
      FoodEffect on LDL cholesterol
      Small <0.20, moderate 0.20–0.40, large reduction >0.40 mmol/L.
      GRADE evidence
      Foods high in n-6 PUFA and/or MUFA and low in SFA; e.g. canola oilModerate to large reduction
      For comparison between foods high in MUFA and/or PUFA vs. SFA and/or trans fatty acids.
      High
      Foods high in soluble fiber; e.g. psyllium, oats, and barleyModerate reductionHigh
      Foods with added plant sterols or stanolsModerate reductionHigh
      Flaxseeds (whole)Small to moderate reductionHigh
      Soy proteinSmall to moderate reductionHigh
      TomatoesSmall to moderate reductionHigh
      AlmondsSmall reductionHigh
      FishNo clear effectHigh
      Decaffeinated coffee (in place of regular coffee)No effectHigh
      Filtered coffeeNo effectHigh
      Foods high in SFA or trans fatty acids (i.e. solid and tropical fats)Moderate to large increase
      For comparison between foods high in MUFA and/or PUFA vs. SFA and/or trans fatty acids.
      High
      Unfiltered coffee (in place of filtered coffee)Moderate to large increaseHigh
      AvocadosModerate to large reductionModerate
      TurmericModerate to large reductionModerate
      HazelnutsSmall to moderate reductionModerate
      PulsesSmall to moderate reductionModerate
      Green teaAt least small reductionModerate
      Fiber, whole grainsSmall reductionModerate
      WalnutsSmall reductionModerate
      Darker roast coffeeNo clear effectModerate
      Fructose (in place of sucrose/glucose)No clear effectModerate
      Marine oils (high in long-chain n-3 PUFA)Very small increaseModerate
      Free sugarsSmall increaseModerate
      Coffee (in place of tea)Small to moderate increaseModerate
      Garlic powderSmall to moderate reductionLow
      Probiotics and prebioticsSmall to moderate reductionLow
      CuminSmall to moderate reductionLow
      GingerSmall reductionLow
      EggsSmall increase
      Possibly larger effect in some individuals and little effect in other individuals
      Low
      Foods high in resistant starchSmall reductionLow
      High-polyphenol olive oil (in place of low-polyphenol)Small reductionLow
      Foods high in α-linolenic acid, e.g. flaxseed oilNo clear effectLow
      Foods high in medium-chain (in place on of long-chain) SFANo clear effectLow
      GrapefruitsNo effectLow
      BerriesSmall to moderate reductionVery low
      GarlicSmall to moderate reductionVery low
      Black teaAt least small reductionVery low
      Dark chocolate/cocoa productsAt least small reductionVery low
      Alcoholic drinksSmall reductionVery low
      Dairy products (all, high-fat, low-fat)No clear effectVery low
      Grape polyphenolsNo clear effectVery low
      SynbioticsNo clear effectVery low
      Whey proteinNo clear effectVery low
      Fruit juiceNo effectVery low
      Red meatNo effectVery low
      SweetenersNo effectVery low
      Abbreviations: GRADE – The Grading of Recommendations Assessment, Development, and Evaluation, LDL – low-density lipoprotein, MUFA – monounsaturated fatty acids, PUFA – polyunsaturated fatty acids, SFA – saturated fatty acids.
      a Small <0.20, moderate 0.20–0.40, large reduction >0.40 mmol/L.
      b For comparison between foods high in MUFA and/or PUFA vs. SFA and/or trans fatty acids.
      c Possibly larger effect in some individuals and little effect in other individuals
      Figure 3
      Figure 3Foods that modify LDL cholesterol by effect and strength of evidence. Larger circles indicate high GRADE evidence. Smaller circles indicate moderate GRADE evidence.

      Alcohol

      One study conducted in 1999 was excluded, while one systematic review conducted in 2011 was included. There was a tendency towards a small reduction in LDL cholesterol but the evidence was considered very low.

      Chocolate and cocoa

      Two studies conducted in 2006–2016 were excluded, while five systematic reviews conducted in 2010–2016 were included. Overall, the evidence was considered very low for a small reduction in LDL cholesterol by chocolate and cocoa.

      Dairy

      Two studies conducted in 2015–2018 were excluded. One systematic review on dairy products conducted in 2013 was rated as having unclear/low risk of bias, but was included. Increased dairy intake had no clear effect on LDL cholesterol, but the evidence was considered very low. For high-fat dairy products (which are high in SFA), see also Fatty foods. For yogurt, see Probiotics.

      Eggs and other cholesterol-rich foods

      One systematic review on cholesterol conducted in 2015 and two on eggs conducted in 2017–2018 were included. As discussed in the most recent systematic review, there may be large individual differences in response to dietary cholesterol, which may partly explain the inconsistency of the results. However, with regard to the overall claim that dietary cholesterol and eggs can increase LDL cholesterol, the evidence was considered low for a small effect.

      Fatty foods

      The two WHO documents [
      World Health Organization & Mensink RP
      Effects of saturated fatty acids on serum lipids and lipoproteins: a systematic review and regression analysis. Geneva.
      ,
      • Brouwer I.
      World Health Organization
      Effect of trans-fatty acid intake on blood lipids and lipoproteins: a systematic review and meta-regression analysis. Geneva.
      ] from 2016 presented meta-regressions from well-controlled dietary RCTs for each percentage energy replacement of SFA and trans fatty acid intakes by other types of fats or carbohydrate. The reductions were largest when SFA or trans fatty acids were replaced by MUFA and PUFA, with high GRADE evidence according to their authors.
      Twenty-four studies conducted in 1992–2018 were excluded, while 15 systematic reviews conducted in 2009–2018 were included. Among individual foods, the results were most clear for rapeseed/canola oil; in a systematic review conducted in 2019, the evidence was considered high for a moderate (0.20–0.40 mmol/L) to large reduction in LDL cholesterol, when approximately 17 percent of energy intake (%E) (weighted mean in 5 of 9 RCTs which reported units as %E) of rapeseed/canola oil (representing about half of total fat intake in most Western populations) replaced foods high in SFA. There was also low evidence for a potential beneficial effect of polyphenols on LDL cholesterol, as demonstrated in studies on olive oil.
      For α-linolenic acid (ALA), in two systematic reviews conducted in 2016–2018, the evidence was considered low and moderate for no clear effect on LDL cholesterol. Note that ALA and flaxseed oil are presented together in Table 2 (for flaxseeds, see Nuts and seeds, below). For marine omega-3 sources, the evidence was considered moderate for a very small increase.

      Fiber and whole grains

      Twelve studies conducted in 1997–2018 were excluded, while 13 systematic reviews conducted in 2010–2018 were included. The evidence was most clear for soluble (or viscous) fibers, e.g. psyllium (weighted mean dose 11 g per day compared with foods without or low in fiber or containing insoluble fiber, in 29 of 29 comparisons with units reported as g per day) or β-glucans from barley (weighted mean dose 5.6 g per day of β-glucan compared with diets low in β-glucan, in 14 of 14 RCTs which reported units as g per day) or oats (weighted mean dose 4.8 g per day of β-glucan compared with food products without soluble fiber or high in insoluble fiber, in 33 of 33 comparisons with units reported as g per day), with high evidence for moderate reductions in LDL cholesterol. The effects of other types of fiber (approximate weighted mean dose 12 g per day compared with lower-fiber diets, in 27 of 34 RCTs which reported units as g per day) and wholegrain (approximate weighted mean dose 90 g per day compared with lower-wholegrain diets, mostly in weight stable conditions, in 25 of 28 comparisons with units reported as g per day) on LDL cholesterol were slightly less clear (moderate evidence, small reductions). For pulses (weighted mean dose 114 g wet weight per day compared with usual diets, in 2 of 4 RCTs which reported units in this format), there was moderate evidence for a small to moderate reduction in LDL cholesterol (see also under Protein-rich foods).
      In a series of reviews [
      • Reynolds A.
      • Mann J.
      • Cummings J.
      • Winter N.
      • Mete E.
      • Te Morenga L.
      Carbohydrate quality and human health: a series of systematic reviews and meta-analyses.
      ] conducted in 2019, results on blood lipids were also presented for starch digestability and glycemic index of carbohydrate. These results and other systematic reviews on dietary patterns low in glycemic index or glycemic load were considered outside the scope of this review.

      Fruits, berries, and vegetables

      Five studies conducted in 2015–2018 were excluded, while eight systematic reviews conducted in 2011–2018 were included. For tomatoes (approximate weighted mean dose of tomatoes or tomato products 300 g per day compared with cucumber or a usual diet without tomatoes, in 2 of 6 RCTs which reported units as g per day), the evidence was considered high for a small to moderate reduction in LDL cholesterol. As non-randomized trials and many different foods were included in the review on lycopene [
      • Ried K.
      • Fakler P.
      Protective effect of lycopene on serum cholesterol and blood pressure: meta-analyses of intervention trials.
      ] conducted in 2011, which contributed to the reduction in the level of evidence, we relied mostly on the evidence of the systematic review on tomatoes [
      • Cheng H.M.
      • Koutsidis G.
      • Lodge J.K.
      • Ashor A.
      • Siervo M.
      • Lara J.
      Tomato and lycopene supplementation and cardiovascular risk factors: a systematic review and meta-analysis.
      ] conducted in 2017.
      Despite their moderate to large effect on LDL cholesterol, the evidence for avocados (substituted for other fat sources in 10 of 12 RCTs and added to the habitual diet in 2 of 10 RCTs, approximate weighted mean dose 219 g per day in 5 of 14 comparisons with units reported as g per day, or 22.5 %E in 3 of 14 comparison with units reported in this format, or 1 avocado per day in 2 of 14 comparisons with units reported in this format) was only considered moderate, due to risk of bias and inconsistency. Berries and other fruits and vegetables caused small to moderate reductions, but the evidence was considered very low.

      Garlic

      Two studies conducted in 2016–2019 were excluded, while three systematic reviews conducted in 2013–2017 were included. In the two studies not restricted to individuals with diabetes, the evidence was low (specifically for garlic powder) or very low for a small to moderate effect on LDL cholesterol.

      Grapes and wine

      One systematic review of grape polyphenols conducted in 2017 was included. Only two of its 29 RCTs demonstrated an effect. Although no meta-analysis was performed, the evidence was considered very low for no clear effect on LDL cholesterol.

      Herbs

      Three studies conducted in 2007–2015 were excluded, but no systematic reviews were included.

      Nuts and seeds

      Seven studies conducted in 2010–2018 were excluded, while nine systematic reviews conducted in 2005–2018 were included. High evidence was demonstrated for almonds (weighted mean dose 60 g per day compared with no almonds or a variety of provided control foods, from 18 of 18 RCTs which reported units as g per day). Moderate evidence was demonstrated also for hazelnuts (reported mean dose 39 g per day compared with control diets without hazelnuts, in 3 RCTs which were included in a Bayesian meta-analysis) and walnuts (approximate weighted mean dose 46 g, in 22 of 24 RCTs which reported units as g per day, or 16 %E, in 14 of 24 RCT which reported units as %E; compared with a variety of control diets low in walnuts). Their effects on LDL cholesterol were small or possibly moderate (for hazelnuts). Notably, in the systematic review [
      • Pan A.
      • Yu D.
      • Demark-Wahnefried W.
      • Franco O.H.
      • Lin X.
      Meta-analysis of the effects of flaxseed interventions on blood lipids.
      ] conducted in 2009, the 10 RCTs evaluating the effects of consuming whole flaxseeds included two RCTs on ground flaxseeds and one RCT on defatted flaxseeds. Still, (whole) flaxseeds obtained high evidence for a small reduction in LDL cholesterol (weighted mean dose 38 g per day compared with a variety of control foods or a diet without flax seeds, in 11 of 11 comparisons with units reported as g per day). For flaxseed oil, see also Fatty foods, above.
      One systematic review on tree nuts [
      • Del Gobbo L.C.
      • Falk M.C.
      • Feldman R.
      • Lewis K.
      • Mozaffarian D.
      Effects of tree nuts on blood lipids, apolipoproteins, and blood pressure: systematic review, meta-analysis, and dose-response of 61 controlled intervention trials.
      ] conducted in 2015 was considered just outside the set AMSTAR threshold. It was nonetheless the largest systematic review, with 42 RCTs and 2102 participants included, and thus merits mentioning. The LDL cholesterol reduction was small but highly significant (mean and 95% CI −0.12 (−0.14 to −0.11) mmol/L) and the evidence for tree nuts overall was considered high. However, evidence gradings and effects are presented separately for individual nuts and seeds in Table 2.

      Plant sterols and stanols

      Seven studies conducted in 2007–2014 were excluded, while five systematic reviews from conducted in 2005–2016 were included. There was high evidence for a moderate reduction in LDL cholesterol by foods enriched with plant sterols and stanols (approximate weighted mean dose 2.2 g per day compared with diets without foods enriched with plant sterols and stanols, in 58 of 59 RCTs which reported units as g per day).

      Probiotics and prebiotics

      Five studies conducted in 2016–2018 were excluded, while 19 systematic reviews conducted in 2011–2019 were included. Most of the papers considered probiotics of various bacterial strains. Some studies were restricted to specific populations, e.g. individuals with diabetes. One systematic review [
      • Guo Z.
      • Liu X.M.
      • Zhang Q.X.
      • Shen Z.
      • Tian F.W.
      • Zhang H.
      • et al.
      Influence of consumption of probiotics on the plasma lipid profile: a meta-analysis of randomised controlled trials.
      ] conducted in 2011 provided moderate evidence for a small reduction in LDL cholesterol by probiotics. However, three similar and larger systematic reviews [
      • Sun J.
      • Buys N.
      Effects of probiotics consumption on lowering lipids and CVD risk factors: a systematic review and meta-analysis of randomized controlled trials.
      ,
      • Shimizu M.
      • Hashiguchi M.
      • Shiga T.
      • Tamura H.O.
      • Mochizuki M.
      Meta-analysis: effects of probiotic supplementation on lipid profiles in normal to mildly hypercholesterolemic individuals.
      ,
      • Cho Y.A.
      • Kim J.
      Effect of probiotics on blood lipid concentrations: a meta-analysis of randomized controlled trials.
      ] conducted in 2015 only provided low or very low evidence for reductions in LDL cholesterol by probiotics. Thus, we considered the overall evidence for probiotics and prebiotics as low for a small to moderate effect on LDL cholesterol. For synbiotics, there was no clear effect on LDL cholesterol and very low evidence.

      Protein-rich foods

      Eleven papers conducted in 2010–2019 were excluded, while seven systematic reviews conducted in 2008–2019 were included. Soy protein (weighted mean dose 26 g per day compared with non-soy protein, in 42 of 42 comparisons with units reported as g per day) obtained high evidence for a small to moderate reduction in LDL cholesterol, albeit based on a dated systematic review conducted in 2008. For plant protein (soy, pulses, and nuts combined), the overall evidence was considered moderate for a small reduction in LDL cholesterol. Fish intake (approximate weighted mean 86 g per day compared with a diet without or low in fish, in 11 of 14 RCTs which reported units as g per day) had high evidence for no clear effect on LDL cholesterol, but foods from animal sources other than eggs had no clear evidence for effects on LDL cholesterol (see Eggs and cholesterol above).
      One systematic review on red meat [
      • Guasch-Ferré M.
      • Satija A.
      • Blondin S.A.
      • Janiszewski M.
      • Emlen E.
      • O’Connor L.E.
      • et al.
      Meta-analysis of randomized controlled trials of red meat consumption in comparison with various comparison diets on cardiovascular risk factors.
      ] conducted in 2019 was excluded as it fell just outside the set risk of bias threshold. If included, its results would have indicated (with high evidence) no overall effect of red meat on LDL cholesterol (−0.03 (−0.08 to 0.02) mmol/L), which was consistent with the included systematic review conducted in 2017. However, compared with plant protein sources, red meat caused a small to moderate increase in LDL cholesterol (0.20 (0.07–0.33) mmol/L). Compared with fish intake, red meat instead caused a small to moderate reduction in LDL cholesterol (−0.17 (−0.26 to −0.09) mmol/L). There were no clear effects on LDL cholesterol for other comparisons.

      Rice

      One study conducted in 2015 was excluded, but no systematic reviews were included.

      Salt

      One study conducted in 2017 was excluded, but no systematic reviews were included.

      Spices

      One study conducted in 2014 was excluded, while three systematic reviews study conducted in 2017–2018 were included. For turmeric, the results on LDL cholesterol were only expressed as standardised mean difference (−0.34 (−0.53 to −0.15), I2 = 42%) in the systematic review [
      • Qin S.
      • Huang L.
      • Gong J.
      • Shen S.
      • Huang J.
      • Ren H.
      • et al.
      Efficacy and safety of turmeric and curcumin in lowering blood lipid levels in patients with cardiovascular risk factors: a meta-analysis of randomized controlled trials.
      ] conducted in 2017, because one RCT did not clearly report units. After contacting its authors, we were able to reproduce the results of the meta-analysis as weighted mean difference. The effect of turmeric (approximate weighted mean dose compared with placebo or no turmeric intake was 2.3 g per day, in 2 of 6 RCTs which reported units in this format, or 600 mg of curcuminoids, in 4 of 6 RCTs which reported units in this format) on LDL cholesterol was moderate to large, without substantial heterogeneity (−0.35 (−0.48 to −0.22) mmol/L, I2 = 25%, Supplemental Fig. 1). The evidence was considered moderate, downgraded for indirectness. For cumin [
      • Hadi A.
      • Mohammadi H.
      • Hadi Z.
      • Roshanravan N.
      • Kafeshani M.
      Cumin (Cuminum cyminum L.) is a safe approach for management of lipid parameters: a systematic review and meta-analysis of randomized controlled trials.
      ] and ginger [
      • Pourmasoumi M.
      • Hadi A.
      • Rafie N.
      • Najafgholizadeh A.
      • Mohammadi H.
      • Rouhani M.H.
      The effect of ginger supplementation on lipid profile: a systematic review and meta-analysis of clinical trials.
      ], the evidence was considered low for a small (to moderate) effect on LDL cholesterol.

      Sugar

      Three studies conducted in 2013–2018 were excluded, while three systematic reviews from 2014 to 2017 were included. The largest systematic review, on free sugars [
      • Te Morenga L.A.
      • Howatson A.J.
      • Jones R.M.
      • Mann J.
      Dietary sugars and cardiometabolic risk: systematic review and meta-analyses of randomized controlled trials of the effects on blood pressure and lipids.
      ] conducted in 2014, demonstrated a small increase in LDL cholesterol with moderate evidence, when approximately 17 %E of sugar (weighted mean in 12 of 22 RCTs which reported units as %E) or 80 g per day (weighted mean in 12 of 22 RCTs which reported units as g per day) was compared with lower sugar intakes isocalorically or ad libitum. For fructose compared with sucrose/glucose, there was no clear effect on LDL cholesterol but high imprecision and moderate evidence.

      Sweeteners

      One systematic review conducted in 2011 was included. No meta-analysis was performed but none of the five included RCTs demonstrated significant effects on LDL cholesterol. The number of participants was low and the evidence was very low for a lack of effect on LDL cholesterol.

      Tea

      Two studies conducted in 2008–2012 were excluded, while 11 systematic reviews on the effects on LDL cholesterol of green tea, black tea, or both conducted in 2011–2018 were included. The systematic review on both green and black tea [
      • Hartley L.
      • Flowers N.
      • Holmes J.
      • Clarke A.
      • Stranges S.
      • Hooper L.
      • et al.
      Green and black tea for the primary prevention of cardiovascular disease.
      ] conducted in 2013 stood out, with large reductions in LDL cholesterol compared with the later and more comprehensive systematic reviews, but was restricted to RCTs of at least three months duration. Taken together, we judged the evidence somewhat conservatively as moderate for green tea (weighted mean dose 263 mg of catechins which corresponds to about 2–5 cups per day, in 17 of 17 RCTs which reported units as g per day in the systematic review from 2016) and very low for black tea, for an at least small reduction in LDL cholesterol.

      Other foods

      In addition to the abovementioned individual foods and food groups, one systematic review and network meta-analysis from 2018 was included. In 66 RCTs of at least four weeks duration with a total of 3595 participants, the authors indirectly compared 10 different food groups: refined grains, whole grains, fruits and vegetables, nuts, legumes, eggs, dairy, red meat, fish, and sugar-sweetened beverages. According to its authors, the GRADE evidence was very low or low for all effects of foods on LDL cholesterol.

      Coffee

      A series of meta-analyses on five different comparisons was conducted. Nine RCTs compared filtered coffee vs. no coffee (Fig. 2A), three compared coffee vs. tea (Fig. 2B), six compared filtered vs. unfiltered coffee (Fig. 2C), six compared regular vs. decaffeinated coffee (Fig. 2D), and two compared different coffee roasts (Fig. 2E). Most studies had some concerns of risk of bias related to incomplete reporting of randomization/allocation concealment, but none were considered of high risk. We found no indications of publication bias or small study effects, although reliable investigations by tests or funnel plots were prevented by the limited number of studies in each analysis. The sensitivity analyses only marginally affected the results (data not shown).
      The most notable effect on LDL cholesterol was observed when filtered coffee was replaced with unfiltered coffee, e.g. Scandinavian style boiled coffee at a mean daily dose of 6.1 cups (high evidence, further upgraded for large effects). By contrast, 4.6 daily cups of filtered coffee did not increase the LDL cholesterol level compared with water or no coffee intake (high evidence).
      Compared with tea, 6.8 daily cups of coffee (mocha, espresso, instant, or filtered coffee) tended to increase LDL cholesterol in our meta-analysis of three RCTs. The evidence was considered moderate (downgraded for imprecision). However, as noted above (see Tea), tea intake may decrease LDL cholesterol.
      No clear effects on LDL cholesterol were observed when comparing regular coffee with decaffeinated coffee (high evidence) or when comparing coffee based on the roasting process (moderate evidence). One study [
      • Wahrburg U.
      • Martin H.
      • Schulte H.
      • Walek T.
      • Assmann G.
      Effects of two kinds of decaffeinated coffee on serum lipid profiles in healthy young adults.
      ] conducted in 1994 also compared (decaffeinated) coffee from different beans (arabica compared with arabica/robusta blend), also with no effect on LDL cholesterol. Two RCTs compared special coffees (different levels of chlorogenic acids and hydroxyhydroquinone) for which no meta-analysis was performed, but there was no significant effect on LDL cholesterol in either study [
      • Agudelo-Ochoa G.M.
      • Pulgarín-Zapata I.C.
      • Velásquez-Rodriguez C.M.
      • Duque-Ramírez M.
      • Naranjo-Cano M.
      • Quintero-Ortiz M.M.
      • et al.
      Coffee consumption increases the antioxidant capacity of plasma and has No effect on the lipid profile or vascular function in healthy adults in a randomized controlled trial.
      ,
      • Yamaguchi T.
      • Chikama A.
      • Mori K.
      • Watanabe T.
      • Shioya Y.
      • Katsuragi Y.
      • et al.
      Hydroxyhydroquinone-free coffee: a double-blind, randomized controlled dose-response study of blood pressure.
      ].

      Discussion

      In this umbrella review of guidelines and systematic reviews, as well as systematic review and meta-analysis of RCTs on coffee, previously established evidence was consolidated for the beneficial effects of foods high in unsaturated and low in saturated or trans fatty acids, e.g. non-tropical oils in place of solid animal and industrially processed fats; for soluble/viscous fiber, especially from oats, barley, and psyllium; for functional foods with added plant sterols and stanols; and for foods rich in plant proteins, especially soybeans. It was also established with high evidence that tomatoes, whole flaxseeds, and almonds can reduce LDL cholesterol, whereas boiled/unfiltered coffee can increase it. There was moderate evidence for beneficial effects by avocados, hazelnuts, and walnuts (high in unsaturated fatty acids), and pulses (high in soluble fiber and plant protein); by turmeric and green tea; and for a small detrimental effect by free sugars.
      Current guidelines for dyslipidemia and CVD prevention were fairly consistent and similar to recommendations for healthy populations, but did not always evaluate the strength of evidence for their included advice. Notably, some foods were not explicitly included in any of these guidelines despite moderate or strong evidence in their favor, e.g. flaxseeds, tomatoes, turmeric, avocados, and green tea. Unfiltered coffee was mentioned only in one guideline, despite its clear detrimental effect on blood lipids. Conversely, much emphasis was put on other foods (e.g. sugar) with less clear evidence regarding effects on LDL cholesterol. However, some of the included guidelines included other aspects than hypercholesterolemia, e.g. effects on triglycerides or hypertension, which may partly explain this discrepancy. Also, potential adverse effects on body weight by energy-dense foods should be taken into account, as most studies have been performed in isocaloric or even hypocaloric conditions, and weight gain may increase LDL cholesterol levels.
      We demonstrated clear effects for the comparison between unfiltered and filtered coffee, in line with results from subgroup analyses in a previous (excluded) meta-analysis and with recently published results from pooled Norwegian cohorts, in which total cholesterol, CVD mortality, and mortality was higher in persons consuming unfiltered compared with filtered coffee [
      • Tverdal A.
      • Selmer R.
      • Cohen J.M.
      • Thelle D.S.
      Coffee consumption and mortality from cardiovascular diseases and total mortality: does the brewing method matter?.
      ]. Unfiltered coffee contains high levels of the diterpenes kahweol and cafestol, providing a mechanistic link. For several other foods, the cholesterol-lowering mechanisms are at least partly unknown. Various suggested mechanisms have been discussed in previous systematic reviews but are not always clearly established. For instance, the effects of turmeric may not be related to the polyphenol curcumin, as there was no clear effect in the excluded review on curcumin supplementation [
      • Sahebkar A.
      A systematic review and meta-analysis of randomised controlled trials investigating the effects of curcumin on blood lipid levels.
      ]. Better biological understanding (including potential adverse effects or drug interactions) could increase the incentive to include novel foods (e.g. tomatoes, turmeric, and green tea) in future dietary recommendations and help identify other relevant foods. The presence or absence of a known plausible mechanism could also warrant additional modifications of the strength of evidence, as this feature is not fully captured by the GRADE criteria. For complex foods such as probiotics and prebiotics, improved biological understanding may be required before further attempts to systematically evaluate the effects. Narrative reviews by field experts may even be preferred until then.
      The present study has limitations. The searches were only performed in two databases. Only English literature was included, and no grey literature. Despite the large scope of the study, only two reviewers were included, which increases vulnerability to bias and errors. Also, the majority of the results relies on the work of others and potential conflicts of interest in the included systematic reviews were not considered. Although they were still evaluated for important information, many systematic reviews were excluded based on the strict criteria of our modified version of the AMSTAR tool, which in other cases may have disregarded importance biases, due to its abbreviated form. In addition, systematic reviews may not have been published for all relevant foods. Thus, all effective foods may not have been captured by our searches. Also, considerable uncertainties exist for the effect estimates for most included foods. Beneficial effects may be exaggerated, particularly when the quality of evidence is less than high. Moreover, it is implausible that all included foods would jointly produce additive effects, although such have been demonstrated for certain food combinations [
      • Kendall C.W.
      • Jenkins D.J.
      A dietary portfolio: maximal reduction of low-density lipoprotein cholesterol with diet.
      ]. Furthermore, the illustration (Fig. 3/Graphical abstract) does not fully take into account potential replacement foods. For some individuals, reduced intakes of foods high in SFA or sugar may be more achievable and thus effective than attempts to increase consumption of foods high in unsaturated fatty acids or soluble fiber, and vice versa. Both strategies may provide beneficial effects on LDL cholesterol depending upon individual background intakes and preferences. Moreover, we cannot with this approach precisely establish which food doses are required for clinically meaningful effects. The weighted mean doses should for some foods be considered as rough approximations, as all RCTs did not always report doses in the same units. However, the highlighted foods may principally modify LDL cholesterol levels, at least at intake levels that have been acceptable in short-term interventions.
      Future studies should further investigate foods whose effects showed moderate (e.g. turmeric and green tea) or low evidence (e.g. eggs, garlic, cumin, ginger, and probiotics). Moreover, the effects of red meat may require further clarification. By contrast, the evidence seems unequivocal for plant sterols, soluble fiber, unfiltered coffee, and possibly soy, as these foods were upgraded even beyond the threshold for high evidence. The short- and longer-term effects of several of the highlighted foods in combination may also warrant further research in RCTs, as well as combining the evidence on LDL cholesterol for individual foods with long-term observational studies on CVD morbidity and mortality. Although LDL cholesterol is causally related to atherosclerotic CVD, foods are complex by nature and may have divergent effects on both known and unknown risk factors. In such instances, priority should always be given to patient-oriented health outcomes.
      In conclusion, several foods can distinctly modify LDL cholesterol levels. This updated summary of the accumulated evidence may help inform clinicians and future guidelines for dyslipidemia and CVD prevention.

      Declaration of competing interest

      There are no conflicts to report.

      Acknowledgments

      We thank Katarina Eriksson for creating the illustration (with permission).

      Appendix A. Supplementary data

      The following are the Supplementary data to this article:

      Funding

      This research did not receive any specific grant from funding agencies in the public, commercial, or not-for-profit sectors.

      Deviations from protocol

      Total cholesterol was stated as a secondary outcome of interest. However, in order to keep the report concise, we chose to focus only on the primary outcome of interest, LDL cholesterol. Also, results as percentage change was omitted as this was seldom given. A complementary search for guidelines was performed in the Trip database.

      References

        • U.S. Department of Health and Human Services and U.S. Department of Agriculture
        2015 – 2020 dietary guidelines for Americans.
        8th ed. December 2015
        • Mach F.
        • Baigent C.
        • Catapano A.L.
        • Koskinas K.C.
        • Casula M.
        • Badimon L.
        • et al.
        2019 ESC/EAS Guidelines for the management of dyslipidaemias: lipid modification to reduce cardiovascular risk.
        Eur Heart J. 2019; (00): 1-78
        • Arnett D.K.
        • Blumenthal R.S.
        • Albert M.A.
        • Buroker A.B.
        • Goldberger Z.D.
        • Hahn E.J.
        • et al.
        2019 ACC/AHA guideline on the primary prevention of cardiovascular disease: a report of the American college of cardiology/American heart association task force on clinical practice guidelines.
        Circulation. 2019; 140: e596-e646
        • Mozaffarian D.
        • Ludwig D.S.
        Dietary guidelines in the 21st century-a time for food.
        J Am Med Assoc. 2010; 304: 681-682
        • World Health Organization & Mensink RP
        Effects of saturated fatty acids on serum lipids and lipoproteins: a systematic review and regression analysis. Geneva.
        2016
        • Kendall C.W.
        • Jenkins D.J.
        A dietary portfolio: maximal reduction of low-density lipoprotein cholesterol with diet.
        Curr Atherosclerosis Rep. 2004; 6: 492-498
        • Statens beredning för medicinsk och social utvärdering (SBU)
        Utvärdering av metoder i hälso- och sjukvården och insatser i socialtjänsten - en handbok. Stockholm.
        2017
        https://www.sbu.se/sv/var-metod/
        Date accessed: March 17, 2020
        • The Cochrane Collaboration
        Cochrane Handbook for systematic reviews of interventions version 5.1.0.
        ([updated March 2011])
        https://handbook-5-1.cochrane.org/
        Date: 2011
        Date accessed: March 17, 2020
        • Brouwer I.
        • World Health Organization
        Effect of trans-fatty acid intake on blood lipids and lipoproteins: a systematic review and meta-regression analysis. Geneva.
        2016
        • Reynolds A.
        • Mann J.
        • Cummings J.
        • Winter N.
        • Mete E.
        • Te Morenga L.
        Carbohydrate quality and human health: a series of systematic reviews and meta-analyses.
        Lancet. 2019; 393: 434-445
        • Ried K.
        • Fakler P.
        Protective effect of lycopene on serum cholesterol and blood pressure: meta-analyses of intervention trials.
        Maturitas. 2011; 68: 299-310
        • Cheng H.M.
        • Koutsidis G.
        • Lodge J.K.
        • Ashor A.
        • Siervo M.
        • Lara J.
        Tomato and lycopene supplementation and cardiovascular risk factors: a systematic review and meta-analysis.
        Atherosclerosis. 2017; 257: 100-108
        • Pan A.
        • Yu D.
        • Demark-Wahnefried W.
        • Franco O.H.
        • Lin X.
        Meta-analysis of the effects of flaxseed interventions on blood lipids.
        Am J Clin Nutr. 2009; 90: 288-297
        • Del Gobbo L.C.
        • Falk M.C.
        • Feldman R.
        • Lewis K.
        • Mozaffarian D.
        Effects of tree nuts on blood lipids, apolipoproteins, and blood pressure: systematic review, meta-analysis, and dose-response of 61 controlled intervention trials.
        Am J Clin Nutr. 2015; 102: 1347-1356
        • Guo Z.
        • Liu X.M.
        • Zhang Q.X.
        • Shen Z.
        • Tian F.W.
        • Zhang H.
        • et al.
        Influence of consumption of probiotics on the plasma lipid profile: a meta-analysis of randomised controlled trials.
        Nutr Metabol Cardiovasc Dis. 2011; 21: 844-850
        • Sun J.
        • Buys N.
        Effects of probiotics consumption on lowering lipids and CVD risk factors: a systematic review and meta-analysis of randomized controlled trials.
        Ann Med. 2015; 47: 430-440
        • Shimizu M.
        • Hashiguchi M.
        • Shiga T.
        • Tamura H.O.
        • Mochizuki M.
        Meta-analysis: effects of probiotic supplementation on lipid profiles in normal to mildly hypercholesterolemic individuals.
        PloS One. 2015; 10e0139795
        • Cho Y.A.
        • Kim J.
        Effect of probiotics on blood lipid concentrations: a meta-analysis of randomized controlled trials.
        Medicine (Baltim). 2015; 94e1714
        • Guasch-Ferré M.
        • Satija A.
        • Blondin S.A.
        • Janiszewski M.
        • Emlen E.
        • O’Connor L.E.
        • et al.
        Meta-analysis of randomized controlled trials of red meat consumption in comparison with various comparison diets on cardiovascular risk factors.
        Circulation. 2019; 139: 1828-1845
        • Qin S.
        • Huang L.
        • Gong J.
        • Shen S.
        • Huang J.
        • Ren H.
        • et al.
        Efficacy and safety of turmeric and curcumin in lowering blood lipid levels in patients with cardiovascular risk factors: a meta-analysis of randomized controlled trials.
        Nutr J. 2017; 16: 68
        • Hadi A.
        • Mohammadi H.
        • Hadi Z.
        • Roshanravan N.
        • Kafeshani M.
        Cumin (Cuminum cyminum L.) is a safe approach for management of lipid parameters: a systematic review and meta-analysis of randomized controlled trials.
        Phytother Res. 2018; 32: 2146-2154
        • Pourmasoumi M.
        • Hadi A.
        • Rafie N.
        • Najafgholizadeh A.
        • Mohammadi H.
        • Rouhani M.H.
        The effect of ginger supplementation on lipid profile: a systematic review and meta-analysis of clinical trials.
        Phytomedicine. 2018; 43: 28-36
        • Te Morenga L.A.
        • Howatson A.J.
        • Jones R.M.
        • Mann J.
        Dietary sugars and cardiometabolic risk: systematic review and meta-analyses of randomized controlled trials of the effects on blood pressure and lipids.
        Am J Clin Nutr. 2014; 100: 65-79
        • Hartley L.
        • Flowers N.
        • Holmes J.
        • Clarke A.
        • Stranges S.
        • Hooper L.
        • et al.
        Green and black tea for the primary prevention of cardiovascular disease.
        Cochrane Database Syst Rev. 2013; 6
        • Wahrburg U.
        • Martin H.
        • Schulte H.
        • Walek T.
        • Assmann G.
        Effects of two kinds of decaffeinated coffee on serum lipid profiles in healthy young adults.
        Eur J Clin Nutr. 1994; 48: 172-179
        • Agudelo-Ochoa G.M.
        • Pulgarín-Zapata I.C.
        • Velásquez-Rodriguez C.M.
        • Duque-Ramírez M.
        • Naranjo-Cano M.
        • Quintero-Ortiz M.M.
        • et al.
        Coffee consumption increases the antioxidant capacity of plasma and has No effect on the lipid profile or vascular function in healthy adults in a randomized controlled trial.
        J Nutr. 2016; 146: 524-531
        • Yamaguchi T.
        • Chikama A.
        • Mori K.
        • Watanabe T.
        • Shioya Y.
        • Katsuragi Y.
        • et al.
        Hydroxyhydroquinone-free coffee: a double-blind, randomized controlled dose-response study of blood pressure.
        Nutr Metabol Cardiovasc Dis. 2008; 18: 408-414
        • Tverdal A.
        • Selmer R.
        • Cohen J.M.
        • Thelle D.S.
        Coffee consumption and mortality from cardiovascular diseases and total mortality: does the brewing method matter?.
        Eur J Prev Cardiol. 2020; 27: 1986-1993
        • Sahebkar A.
        A systematic review and meta-analysis of randomised controlled trials investigating the effects of curcumin on blood lipid levels.
        Clin Nutr. 2014; 33: 406-414
        • Ahola I.
        • Jauhiainen M.
        • Aro A.
        The hypercholesterolaemic factor in boiled coffee is retained by a paper filter.
        J Intern Med. 1991; 230: 293-297
        • Aro A.
        • Kostiainen E.
        • Huttunen J.
        Effects of coffee and tea on lipoproteins and prostanoids.
        Atherosclerosis. 1985; 57: 123-128
        • Aro A.
        • Tuomilehto J.
        • Kostiainen E.
        • Uusitalo U.
        • Pietinen P.
        Boiled coffee increases serum low density lipoprotein concentration.
        Metabolism. 1987; 36: 1027-1030
        • Aro A.
        • Teirilä J.
        • Gref C.
        Dose-dependent effect on serum cholesterol and apoprotein B concentrations by consumption of boiled, non-filtered coffee.
        Atherosclerosis. 1990; 83: 257-261
        • Bak A.
        • Grobbee D.
        The effect on serum cholesterol levels of coffee brewed by filtering or boiling.
        N Engl J Med. 1989; 321: 1432-1437
        • Corrêa T.
        • Rogero M.
        • Mioto B.
        • Tarasoutchi D.
        • Tuda V.L.
        • César L.A.
        • et al.
        Paper-filtered coffee increases cholesterol and inflammation biomarkers independent of roasting degree: a clinical trial.
        Nutrition. 2013; 29: 977-981
        • D’Amicis A.
        • Scaccini C.
        • Tomassi G.
        • Anaclerio M.
        • Stornelli R.
        • Bernini A.
        Italian style brewed coffee: effect on serum cholesterol in young men.
        Int J Epidemiol. 1996; 25: 513-520
        • van Dusseldorp M.
        • Katan M.B.
        • Demacker P.N.
        Effect of decaffeinated versus regular coffee on serum lipoproteins. A 12-week double-blind trial.
        Am J Epidemiol. 1990; 132: 33-34
        • van Dusseldorp M.
        • Katan M.B.
        • van Vliet T.
        • Demacker P.N.
        • Stalenhoef A.F.
        Cholesterol-raising factor from boiled coffee does not pass a paper filter.
        Arterioscler Thromb. 1991; 11: 586-593
        • Fried R.
        • Levine D.
        • Kwiterovich P.
        • Diamond E.L.
        • Wilder L.B.
        • Moy T.F.
        • et al.
        The effect of filtered-coffee consumption on plasma lipid levels. Results of a randomized clinical trial.
        J Am Med Assoc. 1992; 267: 811-815
        • Kempf K.
        • Kolb H.
        • Gärtner B.
        • Bytof G.
        • Stiebitz H.
        • Lantz I.
        • et al.
        Cardiometabolic effects of two coffee blends differing in content for major constituents in overweight adults: a randomized controlled trial.
        Eur J Clin Nutr. 2015; 54: 845-854
        • Rosmarin P.
        • Applegate W.
        • Somes G.
        Coffee consumption and serum lipids: a randomized, crossover clinical trial.
        Am J Med. 1990; 88: 349-356
        • Sanguigni V.
        • Gallu M.
        • Ruffini M.
        • Strano A.
        Effects of coffee on serum cholesterol and lipoproteins: the Italian brewing method. Italian group for the study of atherosclerosis and dismetabolic diseases, Rome II center.
        Eur J Epidemiol. 1995; 11: 75-78
        • Shaposhnikov S.
        • Hatzold T.
        • Yamani N.E.
        • Stavro P.M.
        • Lorenzo Y.
        • Dusinska M.
        • et al.
        Coffee and oxidative stress: a human intervention study.
        Eur J Nutr. 2018; 57: 533-544
        • Superko H.
        • Bortz W.
        • Williams P.
        • Albers J.
        • Wood P.
        Caffeinated and decaffeinated coffee effects on plasma lipoprotein cholesterol, apolipoproteins, and lipase activity: a controlled, randomized trial.
        Am J Clin Nutr. 1991; 54: 599-605
        • Urgert R.
        • Meyboom S.
        • Kuilman M.
        • Rexwinkel H.
        • Vissers M.N.
        • Klerk M.
        • et al.
        Comparison of effect of cafetière and filtered coffee on serum concentrations of liver aminotransferases and lipids: six month randomised controlled trial.
        BMJ. 1996; 313: 1362-1366
        • Wedick N.M.
        • Brennan A.M.
        • Sun Q.
        • Hu F.B.
        • Mantzoros C.S.
        • van Dam R.M.
        Effects of caffeinated and decaffeinated coffee on biological risk factors for type 2 diabetes: a randomized controlled trial.
        Nutr J. 2011; 10: 93