Evidenced Health

Foods and Dietary Components Hub

8 min read

Navigator for the specific-food and component clusters — whether a named food, additive, or compound class moves a patient-important outcome enough to change what someone eats or buys, and whether the food category is doing the work. Nucleus of the label side: Organic vs Conventional Food.

Cancer / neoplasia prevention

The diet/activity/adiposity -> cancer domain (WCRF/AICR CUP). Cross-cutting — a dedicated Cancer Prevention Hub is owed once the domain grows beyond these; catalogued here for now.

  • Diet Physical Activity and Cancer Prevention — the cancer-prevention nucleus: WCRF’s graded causal matrix (convincing/probable/limited), which exposures move cancer risk and how strong the evidence is; the big rocks are body fatness, alcohol, physical activity
  • Body Fatness and Cancer Risk — the largest diet-adjacent cancer lever: body fatness a convincing/ probable cause of ~12 of 17 cancers, its hallmark mechanisms, and the young-adulthood breast paradox
  • Dietary Acrylamide and Cancer Risk — the dietary-acrylamide nucleus: the hyped cooking-formed “toxin” (fried/baked starches, coffee) that is IARC 2A on animal/mechanism grounds but null across ~20 non-gynecological cancer sites in humans (Filippini dose-response MA, no threshold); the symmetric-standards / attention-anti-signal test — near-zero Layer-1 lever
  • Alcohol and Cancer Risk — the per-drinker site-specific dose-response (Bagnardi 2014 gold MA, 572 studies / 23 sites): heavy-drinking RRs up to 5.13 (oral/pharynx), significant risk from light drinking for aerodigestive sites + breast (no safe threshold there), acetaldehyde/ALDH2 mechanism; the effect leg paired with Rumgay’s burden on Alcohol and Mortality and Vascular Disease (cluster alcohol)

Foods and beverages

  • Fish and Seafood Consumption — the fish-seafood nucleus: EFSA’s benefit-vs-risk weighing with numbers (EPA/DHA -> CHD mortality + fetal neurodevelopment vs methylmercury), why the balance flips by stratum and species (apex predators carry the mercury; oily low-trophic fish carry the benefit), and why “fish” is the wrong exposure
  • Coffee Consumption and Health — what habitual coffee does to patient-important outcomes, for whom, at what dose, and how much is causal
  • Tea Consumption and Cardiovascular Risk — the beverage-cells tea opener: per-cup RR within 1-4% of 1.0 for CVD/stroke/all-cause mortality (Chung 2020 dose-response MA, cohorts only), but the benefit goes NULL in the best-exposure-measured studies (RoB gradient) and the source is Unilever-funded; not a big rock, food-vs-component question wide open (no MR, no decaf-analog)
  • Flavonoid Intake and Mortality — the beverage-cells flavonoid cell: dietary flavonoid intake -> lower total (RR 0.87) + CVD (0.85) mortality (Mazidi 2020 MA, 16 cohorts), cancer NULL. But the exposure is FFQ-ESTIMATED from the foods (fruit/veg/tea), so a “flavonoid” MA does NOT isolate the component — it re-expresses the same dietary-pattern signal, NOT independent backing for tea; the food-vs-component gap stays open. Heterogeneity claimed without I2; borderline (all-cause CI to 0.99); confidence low
  • Flavonoid Intake and Cognitive Function — the cognition sibling of the flavonoid-mortality cell: flavonoid intake -> adverse cognitive events OR 0.90 (0.83-0.98), driven by cognitive DECLINE with dementia + AD both NULL (Peng 2025 SR+MA, 26 studies / 269,574). Same FFQ-estimated exposure that does NOT isolate the component; nested inside the Zhou F&V signal so type-F/G, NOT independent-E; fitted- linear dose-response of unstated range; confidence low
  • Soy Isoflavones and Cognitive Function — the RCT counterweight to the flavonoid-cognition null: supplemental soy isoflavones -> pooled test scores SMD 0.19 (0.07-0.32), memory-carried (0.15), every non-memory domain NULL (Cui 2020 RCT MA, 16 trials). A surrogate (test score, <=2.5 y), not a decline event; the tofu-harm scare does NOT survive confounding; RCT-vs-observational split is a type-F/B distinction not a tension; equol-producer route-(b) unestablished; confidence low
  • Nut Consumption and Mortality — the plant-foods opener: nuts -> CHD/CVD/cancer/all-cause & cause-specific mortality (Aune dose-response MA, per 28 g/day). Sizeable inverse RRs, an observed plateau at 15-20 g/day, but observational-only (healthy-user ceiling, no MR); “20 g optimal” is a study-edge, not a derived optimum
  • Fruit and Vegetable Intake and Health — plant-foods nucleus: F&V -> CHD/stroke/CVD/cancer/ all-cause (Aune 2017 dose-response MA, per 200 g/day; ~8-10% lower CVD/death per serving, cancer weak). Lowest risk at the 800 g/day study-edge (double 5-a-day), not an optimum; the benefit does NOT distribute evenly across types (leafy greens/citrus carry it, grapes/berries too thin); measurement error attenuates toward null; observational, same-lab F-not-E vs the nut/whole-grain MAs
  • Red and Processed Meat and Cancer — whether red or processed meat causes cancer, by how much, and whether the evidence warrants reducing intake
  • Lean Red Meat and Atherogenic Lipoproteins — the two lean-red-meat feeding RCTs (Roussell BOLD 2012, Bergeron APPROACH 2019): within a low-SFA diet, LDL-C/apoB are set by background SFA and by meat-vs-plant, NOT by beef amount or red-vs-white color; red = white on lipids; the rise is large-LDL (weakly atherogenic), so the surrogate over-reads. Surrogate-only, no events
  • Poultry and White Meat Consumption — the white-meat opener: poultry+rabbit (EXCL fish) -> all-cause mortality (Lupoli 2021 MA, OR 0.94 highest-vs-lowest, a 6% reduction) but NEUTRAL on CV mortality and non-fatal CV events. No dose-response (heterogeneous highest-vs-lowest contrast); very high heterogeneity; the all-cause signal is inseparable from red-meat displacement (a comparator problem). Observational, single gold MA; stroke (Kim) + CVD/T2D (Ramel) endpoints await
  • Vegetarian Dietary Patterns and Mortality — the AHS-2 cohort (Orlich 2013) leg of the dietary-patterns cluster: vegetarian/vegan/pesco/lacto-ovo patterns -> all-cause mortality ~0.88 (pesco/vegan-in-men steepest, semi null; strictness does not reward), plus a diabetes/renal channel and a male-concentrated CV signal (diet x sex P=.01 for CVD). Single confounded cohort — the healthy-adherent bundle is the ceiling; the EPIC-Oxford divergence shows “vegetarian” is not one exposure
  • Diet Quality Scores and Cardiovascular Risk — the diet-quality-scores nucleus (PURE, Mente 2023, 80 countries): an unweighted 6-food protective-pattern score -> CVD/mortality (score >=5 vs <=1 mortality 0.70), with a LOCATED dose-response knee ~score 4/6 (the lever is raising low consumers to moderate, not optimizing an adequate diet), the under-nutrition-not-over-nutrition global reframe (regional heterogeneity = curve-position), PURE ~= Med/HEI/DASH but Planetary/EAT-Lancet null. Weak discrimination (AUC 0.52-0.61), ~1/3 attenuation under adjustment, dairy-industry-funded — observational, no RCT
  • Low-Fat Dietary Pattern and Cardiovascular Disease — the WHI DM Trial (n=48,835, mean 8.1 yr): the largest RCT of a total-fat-reduction pattern, NULL on CHD/stroke/CVD (HR ~0.97/1.02/0.98). A diluted-contrast (achieved ~70% of design, arm rebounded 24->29%E) + wrong-exposure (fat->carbohydrate, not protective-foods) + ~40%-power null — a no-effect-on-the-tested-contrast result, NOT a refutation of the recommended pattern; the fat-quality lever is elsewhere. Symmetric-standards counterweight to observational diet-pattern signals
  • Dietary Patterns and COPD Risk — the fabric’s first respiratory outcome: a healthy overall pattern -> lower COPD (cross-sectional OR 0.88, I2=0%; cohort RR 0.56 but “essentially two cohorts”), the unhealthy arm null with I2=91% (Parvizian 2020, observational-only SR+MA). The precision inversion (tightest number is the least causal), smoking mostly adjusted, no QoL/mortality studied. A secondary lever behind the smoking big rock and confirmatory of the CV diet-pattern case — confidence low
  • Dietary Fibre and COPD Risk — the fibre-COPD bridge (fibre-cluster orbiter): fibre isolated out of the pattern-level signal, dose-response on COPD incidence (Valisoltani 2023, 5 cohorts, n≈213,912). Total RR 0.72 (0.64-0.80), cereal/fruit significant, vegetable null; per +10 g/day total/cereal/ fruit -26%/-21%/-37%. NutriGrade LOW, ROBINS-E moderate. A type-F component-refinement of Dietary Patterns and COPD Risk (broadens the incidence base ~two -> five cohorts) — confidence low
  • Fermented Foods and Health — the fermented-foods nucleus: do live-fermented foods move a patient-important outcome, and is the effect the live cultures, the food matrix, or fermentation’s biochemistry? Held evidence is thin (one surrogate-outcome RCT + two observational fermented-dairy CVD MAs, Guo refining Zhang downward); the live-vs-pasteurized natural experiment is unrun; trendy ferments are a named GAP. No halo
  • Dairy and Cardiometabolic Health — the dairy nucleus: what dairy does to CVD, mortality and T2D once you stop treating “dairy” as one food. Broadly neutral (CVD/mortality) to modestly inverse (T2D, women); the milk-mortality scare is a single-cohort confounding artifact; the matrix mechanism AWAITS Thorning. Non-fermented axes only (fermented-dairy CVD lives on the fermented-foods nucleus)

Sugars, sweeteners, and seed oils

Is the compound-class scare real?

  • Organic vs Conventional Food — whether the “organic” label delivers a health benefit large enough to change buying, and whether it is the certificate or an underlying exposure doing the work
  • Dietary AGEs and Cooking Method — whether preferring low-heat cooking to reduce advanced glycation end products improves outcomes
  • Antinutrients in Plant Foods — whether “antinutrients” are a reason to avoid plant foods, or ordinary preparation defuses the concern
  • Soy Products and Health — the soy nucleus: the three fears tested under symmetric standards — hormonal (Reed 2021: male-hormone NULL; isoflavones are SERMs), breast cancer (Chen 2014: neutral-to-protective, menopausal-status/region/design-conditional), LDL (Anderson 1995: modest lowering via isolated soy protein, baseline-dependent, DATED). Forms differ (whole / fermented / isolated / condiment); thyroid + natto-K2 are named GAPs

Graph View