Food: What Human Trials Have Shown
Foods and ingredients, what trials found when people cut back on some of them, whole eating patterns, and the way food is cooked. Every letter is copied from the record that owns it and rests on randomized trials in people; foods no trial has tested are named as untested.
- Every letter here is copied from the record that owns it and rests on randomized trials in people who changed what they ate or how it was cooked.
- Several letters rest on a measured risk factor such as LDL cholesterol or blood pressure rather than on disease, and each entry's endpoint line says which.
- Trials that failed keep their D, and foods no trial has tested are named as untested rather than left out.
Every record here rests on randomized trials in people who changed what they ate or how it was cooked. Each letter is copied at build time from the record that owns it, including the D grades and the records EBL does not grade.
Foods and ingredients
Single foods and food ingredients, each graded on the outcome its trials measured. Several letters rest on a measured risk factor such as LDL Low-density lipoprotein cholesterol, the cholesterol fraction most closely tied to plaque in the arteries. cholesterol or blood pressure rather than on disease; the endpoint The outcome a trial measures to decide whether the treatment worked, such as heart attacks, blood pressure or a symptom score. line says which.
- AHard clinical endpointProtocol · Three cluster- or kitchen-randomized trials with hard cardiovascular endpointsEndpoint: Stroke, major cardiovascular events and death
- BSurrogate biomarkerProtocol · Meta-analyses A study that pools the numbers from several trials into one combined estimate. of randomized trials, with a molecular-weight trialEndpoint: LDL cholesterol
- BSurrogate biomarkerProtocol · Meta-analyses of randomized trials, psyllium-specific and genus-levelEndpoint: Total and LDL cholesterol
- BSurrogate biomarkerProtocol · Meta-analysis of randomized trialsEndpoint: LDL cholesterol
- BSurrogate biomarkerProtocol · Meta-analysis of 14 randomized trials and a GRADE systematic review A review that searches for every study on a question using a method set in advance, then assesses them together.Endpoint: LDL cholesterol
- CSelf-reportProtocol · One randomized crossover A trial design in which each participant receives both the treatment and the comparison, one after the other, and serves as their own control. trial and one exploratory partially randomized trialEndpoint: Number of complete spontaneous bowel movements per week
- CSelf-reportProtocol · Two small randomized trials in different populationsEndpoint: Number of complete spontaneous bowel movements per week
- CSurrogate biomarkerProtocol · One large double-blind Neither the participants nor the people assessing them know who received the real treatment until the trial ends. trial and two meta-analysesEndpoint: Blood pressure
- CSurrogate biomarkerProtocol · Cochrane meta-analysis of 35 short randomized trialsEndpoint: Blood pressure
- CHard clinical endpointProtocol · Two randomized trials of dietary advice, one positive and one negativeEndpoint: All-cause and cardiac death
- CSurrogate biomarkerProtocol · Dose-response meta-analysis of 38 randomized trialsEndpoint: LDL cholesterol
- CSurrogate biomarkerProtocol · Meta-analysis of 61 randomized trials and a 2-year walnut trialEndpoint: LDL cholesterol
- CSurrogate biomarkerProtocol · Meta-analysis of 41 randomized trialsEndpoint: LDL cholesterol
- CSurrogate biomarkerProtocol · Meta-analysis of randomized trialsEndpoint: LDL cholesterol
- CSurrogate biomarkerProtocol · Meta-analysis of 14 randomized trialsEndpoint: LDL cholesterol
- CSurrogate biomarkerProtocol · One 6-month double-blind randomized trial and supporting short trialsEndpoint: Flow-mediated dilation
- CHard clinical endpointProtocol · One large cluster randomized trial with fracture outcomesEndpoint: Fractures
- CHard clinical endpointProtocol · One large pragmatic randomized trial with hard outcomesEndpoint: Adverse clinical outcome and death at 30 days
- DSurrogate biomarker · Tested, no benefitProtocol · One large multicenter randomized trial, primary endpoint The one outcome a trial names in advance as its main test. Results on other outcomes carry less weight. not metEndpoint: Visceral fat by MRI
- DHard clinical endpoint · Not yet testedProtocol · No randomized human trial of the claim; observational Research that watches what people already do rather than assigning a treatment. It can show that two things travel together, not that one causes the other. and marker studies onlyEndpoint: Cancer incidence
- Not gradedProtocol · One large randomized trialEndpoint: Cardiovascular events in older adults: primary endpoint missed in one large trial; not graded.
- Not gradedProtocol · One small randomized study (pilot size)
Cutting back
What randomized trials found when people ate or drank less of something, or swapped it for something else.
- BSurrogate biomarkerProtocol · Controlled feeding trial and a meta-analysis of 60 feeding trialsEndpoint: LDL and HDL cholesterol
- BSurrogate biomarkerProtocol · Meta-analysis of randomized trials (Cochrane)Endpoint: Systolic and diastolic blood pressure (Cochrane meta-analysis)
- BSurrogate biomarkerProtocol · Observational synthesis for the headline; randomized trials and pooled analyses on blood pressure and atrial fibrillationGrade for: Blood pressure in people drinking more than two drinks a day
- CHard clinical endpointProtocol · Meta-analyses of randomized diet trials, recovered data from two 1960s and 1970s randomized trials, and randomized feeding trials of blood lipids and inflammatory markers; cohort studies An observational study that follows a group of people over time and compares what happens to them. and a non-randomized hospital trial are listed as excluded evidence and carry no letterGrade for: Coronary and cardiovascular events when linoleic-acid oils replace saturated fat
- CSurrogate biomarkerProtocol · Meta-analysis of randomized trials within a broader systematic reviewEndpoint: Body weight and energy intake
- CSurrogate biomarkerProtocol · Randomized controlled trial People are assigned to the treatment or a comparison by chance, so the groups differ only in what they received. It is the design that can show cause and effect. (inpatient metabolic ward)Endpoint: Short-term energy intake and body weight
- CSurrogate biomarkerProtocol · Meta-analysis of 36 randomized trials, by comparison dietEndpoint: LDL cholesterol
- CSurrogate biomarkerProtocol · Meta-analysis of 14 randomized trialsEndpoint: LDL cholesterol
- Not gradedProtocol · One small randomized trialEndpoint: One 6-month trial found a liter a day of sugared cola raised liver and visceral fat compared with milk, diet cola or water; not graded.
Eating patterns
Whole diets and meal timing, tested as a pattern rather than as single foods.
- AHard clinical endpointProtocol · Randomized controlled trial (PREDIMED)Endpoint: Major cardiovascular events (PREDIMED)
- BSurrogate biomarkerProtocol · Randomized controlled feeding trials (DASH, DASH-Sodium) + meta-analyses of randomized trialsEndpoint: Systolic and diastolic blood pressure (DASH, DASH-Sodium)
- BSurrogate biomarkerProtocol · Meta-analyses of randomized trials (Morton, Tagawa, Nunes) and randomized trialsGrade for: Lean (fat-free) mass during resistance training, healthy adults
- BSurrogate biomarkerProtocol · Randomized controlled trials + meta-analysisGrade for: Blood pressure against an unrestricted diet, adults without diabetes
- BSurrogate biomarkerProtocol · Randomized controlled trials (DiRECT, DIADEM-I) with a 5-year follow-upEndpoint: Type 2 diabetes remission at 12 months: HbA1c A blood test that reflects average blood sugar over the previous two to three months. below 6.5% with all glucose-lowering drugs stopped
- CSurrogate biomarkerProtocol · Umbrella review and meta-analyses of randomized trials + epilepsy RCTs Randomized controlled trial: people are assigned to the treatment or a comparison by chance, so the groups differ only in what they received. It is the design that can show cause and effect.Grade for: Longevity and cardiometabolic markers in adults
- CSurrogate biomarkerProtocol · Meta-analyses of randomized trials + one-year RCTsEndpoint: Weight and cardiometabolic markers, compared with daily calorie restriction
- CSurrogate biomarkerProtocol · Randomized controlled trialsEndpoint: Weight and metabolic markers with calories held equal
- CSurrogate biomarkerProtocol · Randomized controlled trial (CALERIE-2)Endpoint: Metabolic markers and pace of aging in healthy non-obese adults
- CHard clinical endpointProtocol · Small randomized controlled trials of the whole program, each run by the program's developers and none independently repeatedGrade for: Coronary narrowing and cardiac events in coronary heart disease
- DHard clinical endpoint · Not yet testedProtocol · Randomized controlled trials and meta-analyses of randomized trials (LDL cholesterol, HbA1c); cohort evidence only for events and mortalityGrade for: Heart attack, stroke, mortality and longevity
- DSurrogate biomarker · Tested, harm foundProtocol · Meta-analysis of 13 randomized trialsEndpoint: Body weight and energy intake
- DMeasured function · Tested, no benefitProtocol · One 3-year randomized trial, primary endpoint not metEndpoint: Global cognition
How food is cooked
Cooking method changes what ends up in the food. The char and processed meat evidence, which no trial can test against cancer, is on the environmental exposures page.
- CSurrogate biomarkerProtocol · One randomized trialEndpoint: Insulin resistance (HOMA-IR)
- Char from grilling and frying meat: what studies and trials found →
- Processed and red meat: what studies and trials found →
- Acrylamide from frying and baking: what studies found →
- Organic food and pesticide residues: what studies and trials found →
- Nutrition foundations →
- Nutrient gaps in America →
- Myths, tested against trials →
- Healthy aging: the evidence by area →
Common questions
Which food entry rests on outcomes like stroke and death?
Potassium-enriched salt substitute used in place of regular salt, which rests on three cluster- or kitchen-randomized trials whose endpoints were stroke, major cardiovascular events and death (PMIDs 34459569, 37055566, 16762939).
Why are some foods graded on cholesterol rather than on heart disease?
Because LDL cholesterol is what their trials measured. Oat beta-glucan, psyllium and plant sterols are graded on that endpoint, not on heart attacks or strokes.
Did eating an avocado a day reduce belly fat?
No. One large multicenter randomized trial missed its primary endpoint, visceral fat measured by MRI (PMID 35861827).