Place · Level 3 · Macro
Ultra-processed Foods (UPF)
NOVA 分级 · Hall 2019 NIH RCT 同卡路里 +500 kcal/d · 超 60% 美国人热量 · 营养学最重要的范式转移
Last updated
Story path
Chapter 1
NOVA 4-tier classification
NOVA 4-tier classification
The NOVA food classification is a framework proposed by Monteiro's team at the University of São Paulo (Brazil) in 2009, now adopted by WHO, FAO, the EU, and Canadian dietary guidelines. It classifies food by processing degree, not by macro or calorie:
Tier 1 · unprocessed or minimally processed:
Whole fruit, vegetables, whole grains, legumes, nuts, meat, eggs, fish, dairy (no additions)Processing only for preservation or convenience (frozen, dried, vacuumed, cut)Examples: apple, spinach, brown rice, chicken breast, plain yogurt, raw eggs
Tier 2 · culinary ingredients:
Substances extracted from Tier 1 — sugar, oil, salt, starch, butterNot meant to be eaten alone, but cooking inputsExamples: olive oil, sugar, sea salt, flour, butter, vinegar
Tier 3 · processed foods:
Simple combinations of Tier 1 + Tier 2, usually 2–3 ingredientsTraditional preservation / cooking methods (fermentation, smoking, curing, baking)Examples: canned tomatoes, traditional bread (flour + water + yeast + salt), cheese, salted fish, smoked meat, homemade jam
Tier 4 · ultra-processed foods (UPF):
5+ ingredients, usually containing things you don't have in a home kitchen:Industrial sugars (high-fructose corn syrup HFCS / invert sugar)Industrial fats (hydrogenated oil / refined palm oil / trans fats)Industrial proteins (soy isolate / hydrolyzed wheat protein)Flavorings (MSG + 5'-inosinate + various 'natural flavors')Emulsifiers (polysorbate 80 / carrageenan / xanthan gum / mono- and diglycerides)Thickeners, preservatives, artificial colors, artificial sweetenersProduced by industrial extrusion / high-temperature pressure / chemical modificationDesign goal: hyper-palatable + long shelf life + ready-to-eatExamples: soft drinks, instant noodles, breakfast cereals, packaged pastries, sausages, chicken nuggets, ice cream, candy, frozen meals, most bag snacks, chain bakery breads (with 10+ additives)
Quick identification rules:
If the ingredient list contains 'emulsifier / thickener / flavor enhancer / artificial sweetener / maltodextrin / high-fructose syrup' → UPFIf a home kitchen can't make it from these ingredients, it's UPFIf the ingredient list has 5+ items, most are UPF (with rare exceptions: complex homemade sauces can have 5+ items too)
Why not classify by macros:
The same 200 kcal of chicken breast vs a 'protein bar' produces completely different metabolic effects in the body (next scene's Hall RCT proves this)Calorie accounting cannot explain UPF's health harm → a new framework is needed
Tier 1 · unprocessed or minimally processed:
Whole fruit, vegetables, whole grains, legumes, nuts, meat, eggs, fish, dairy (no additions)Processing only for preservation or convenience (frozen, dried, vacuumed, cut)Examples: apple, spinach, brown rice, chicken breast, plain yogurt, raw eggs
Tier 2 · culinary ingredients:
Substances extracted from Tier 1 — sugar, oil, salt, starch, butterNot meant to be eaten alone, but cooking inputsExamples: olive oil, sugar, sea salt, flour, butter, vinegar
Tier 3 · processed foods:
Simple combinations of Tier 1 + Tier 2, usually 2–3 ingredientsTraditional preservation / cooking methods (fermentation, smoking, curing, baking)Examples: canned tomatoes, traditional bread (flour + water + yeast + salt), cheese, salted fish, smoked meat, homemade jam
Tier 4 · ultra-processed foods (UPF):
5+ ingredients, usually containing things you don't have in a home kitchen:Industrial sugars (high-fructose corn syrup HFCS / invert sugar)Industrial fats (hydrogenated oil / refined palm oil / trans fats)Industrial proteins (soy isolate / hydrolyzed wheat protein)Flavorings (MSG + 5'-inosinate + various 'natural flavors')Emulsifiers (polysorbate 80 / carrageenan / xanthan gum / mono- and diglycerides)Thickeners, preservatives, artificial colors, artificial sweetenersProduced by industrial extrusion / high-temperature pressure / chemical modificationDesign goal: hyper-palatable + long shelf life + ready-to-eatExamples: soft drinks, instant noodles, breakfast cereals, packaged pastries, sausages, chicken nuggets, ice cream, candy, frozen meals, most bag snacks, chain bakery breads (with 10+ additives)
Quick identification rules:
If the ingredient list contains 'emulsifier / thickener / flavor enhancer / artificial sweetener / maltodextrin / high-fructose syrup' → UPFIf a home kitchen can't make it from these ingredients, it's UPFIf the ingredient list has 5+ items, most are UPF (with rare exceptions: complex homemade sauces can have 5+ items too)
Why not classify by macros:
The same 200 kcal of chicken breast vs a 'protein bar' produces completely different metabolic effects in the body (next scene's Hall RCT proves this)Calorie accounting cannot explain UPF's health harm → a new framework is needed
Global UPF intake
Global UPF share of total energy intake (Monteiro 2018 + GBD 2024):US: 57–60% of total energy from UPF (Martínez Steele 2016) — world's highestUK: 50–55%Canada / Australia: 40–50%Brazil: 20–30% (lower than developed countries but rising fast)France: 30–35% (Mediterranean tradition resisted some incursion)Italy: 18–20% (Latin culture + home cooking tradition)Japan / Korea: 30–40% (traditional diet + convenience food parallel)China: 18–25% (CHNS 2018 data), rising rapidly — from ~5% in 2000 to ~25% in 2020, a 4× rise in 20 years (Du 2022)
China's UPF growth is one of the fastest globally:
Urbanization + dual-income families + delivery food spread + packaged food industrializationChildren and adolescents eat more of it than adults: in the same national survey (CHNS) under the same NOVA definition, mean daily UPF among ages 6-18 rose from 9.7 g in 1997 to 60.0 g in 2011, against 12.0 g → 41.5 g for adults 20+ (Li & Shi). ⚠️ The data stop at 2011, before the delivery-app and milk-tea waveFrozen noodles + packaged snacks + sugar-sweetened beverages + instant noodles + chain bakery breads are the main vehicles'Health claim' foods are largely UPF too: 'low-fat' yogurt (with thickeners + sweeteners), 'protein' bars (10+ additives), 'whole grain' breakfast cereals (sugar + HFCS + flavorings)
Why this number matters:
UPF intake correlates linearly positively with obesity, T2D, cardiovascular disease, depression, multiple cancers, and all-cause mortality (Lane 2023 BMJ meta of 32 outcomes)There is no 'safe UPF intake share' threshold — the dose-response curve is linearEach +10% UPF share → all-cause mortality ↑ 14%, cardiovascular events ↑ 12%, depression ↑ 22% (Lane 2023)
'Eat less UPF but don't become extreme':
Don't demonize UPF to the point of 'absolutely never touch it' (this triggers health anxiety and eating disorders)The goal is shifting UPF from 60% of total energy down to < 20%, not 'zero'Method: one home-cooked meal per week → several meals → most meals; replace graduallyFocus on replacing sugar-sweetened beverages + processed meat + packaged snacks — these three give the biggest health gains
wang-2021-jama-us-youth-upf
Chapter 2
Hall 2019 NIH RCT
Hall 2019 NIH RCT
This is one of the most important RCTs in nutrition over the past 10 years — it upgraded 'UPF is a problem' from observational hypothesis to mechanistic proof:
Hall et al. 2019 Cell Metabolism (NIH NIDDK metabolic ward, N = 20):
Design:
Inpatient metabolic ward RCT (controlled environment, not self-reported food questionnaires)Crossover design: each subject did two 2-week periods — one eating completely ultra-processed (UPF) food, one eating completely unprocessed (matched whole foods)Key control: diets were precisely matched for calories, protein, carbs, fat, sugar, sodium, and fiberSubjects could eat ad libitum (unrestricted) — no caloric limitMeasured: body weight + spontaneous intake + metabolic rate + multiple hormones (ghrelin + leptin + glucagon-like peptide-1: A gut hormone released after eating that makes you feel full and helps lower blood sugar.)
Results (paradigm-shifting):
UPF group ate ~500 kcal more per day (508 ± 106 kcal/day, p < 0.0001)UPF group gained +0.9 kg in 2 weeks, unprocessed group lost 0.9 kg — net difference 1.8 kg (4 lb)UPF group ate faster (~17 kcal/min vs 14, p = 0.01) — meaning the brain's 'satiety signal' arrived too lateUPF group PYY (satiety hormone) rose less, ghrelin (hunger hormone) was suppressed lessMetabolic rate showed no difference — so weight gain came purely from intake difference
Why this is revolutionary:
Same calories + macros + sugar + salt + fiber, and the UPF group still ate more'A calorie equals a calorie' was falsified by this single studyThis points to UPF harm coming not just from 'what it contains' (sugar + trans fat + sodium) but from 'its structure':Food matrix is destroyed → fiber + nutrient release rate changes → glycemic response differsHyper-palatability → bypasses natural satiety mechanismsHigh energy density (kcal/g) → more calories consumed for the same volumeSoft + easy to chew → eating rate is fast → satiety signal lagsLacks chewing + saliva-mixing-phase gut hormone signals (cephalic phase impaired)
This is the paradigm shift: not 'eat less sugar / less fat', but 'eat less industrially restructured food'.
Follow-up: Hamano 2024 (Japan, randomised crossover, n = 9, one week per arm) reproduced the direction with an even larger gap, not a smaller one — though at that sample size it is a direction, not a magnitude.
Hall et al. 2019 Cell Metabolism (NIH NIDDK metabolic ward, N = 20):
Design:
Inpatient metabolic ward RCT (controlled environment, not self-reported food questionnaires)Crossover design: each subject did two 2-week periods — one eating completely ultra-processed (UPF) food, one eating completely unprocessed (matched whole foods)Key control: diets were precisely matched for calories, protein, carbs, fat, sugar, sodium, and fiberSubjects could eat ad libitum (unrestricted) — no caloric limitMeasured: body weight + spontaneous intake + metabolic rate + multiple hormones (ghrelin + leptin + glucagon-like peptide-1: A gut hormone released after eating that makes you feel full and helps lower blood sugar.)
Results (paradigm-shifting):
UPF group ate ~500 kcal more per day (508 ± 106 kcal/day, p < 0.0001)UPF group gained +0.9 kg in 2 weeks, unprocessed group lost 0.9 kg — net difference 1.8 kg (4 lb)UPF group ate faster (~17 kcal/min vs 14, p = 0.01) — meaning the brain's 'satiety signal' arrived too lateUPF group PYY (satiety hormone) rose less, ghrelin (hunger hormone) was suppressed lessMetabolic rate showed no difference — so weight gain came purely from intake difference
Why this is revolutionary:
Same calories + macros + sugar + salt + fiber, and the UPF group still ate more'A calorie equals a calorie' was falsified by this single studyThis points to UPF harm coming not just from 'what it contains' (sugar + trans fat + sodium) but from 'its structure':Food matrix is destroyed → fiber + nutrient release rate changes → glycemic response differsHyper-palatability → bypasses natural satiety mechanismsHigh energy density (kcal/g) → more calories consumed for the same volumeSoft + easy to chew → eating rate is fast → satiety signal lagsLacks chewing + saliva-mixing-phase gut hormone signals (cephalic phase impaired)
This is the paradigm shift: not 'eat less sugar / less fat', but 'eat less industrially restructured food'.
Follow-up: Hamano 2024 (Japan, randomised crossover, n = 9, one week per arm) reproduced the direction with an even larger gap, not a smaller one — though at that sample size it is a direction, not a magnitude.
Why same kcal, +500 still
Hall 2019 broke down why UPF makes people eat more at the molecular level — several non-exclusive mechanisms operate in parallel:1. Energy density (kcal/g):
UPF group meal averaged 2.06 kcal/g, unprocessed group 1.20 kcal/gSame volume / weight intake → UPF group consumes 70% more calories'Stomach size' in physiology is calculated by volume, not calories — this is a satiety mechanism honed over hundreds of millions of years of evolution
2. Eating rate:
UPF group ~17 kcal/min, unprocessed group ~14 kcal/minSatiety signal takes ~20 minutes to reach the brain — eating faster means stopping at higher intakeUPF industrial extrusion + high moisture + soft texture → less chewing + faster swallowing
3. Food matrix destruction:
In whole foods, fiber + protein + fat + carbs are structurally bound — slow gastric emptying + flat glycemic responseUPF deconstructs and reassembles them: 'flour + sugar' vs 'whole wheat bread' — same carbs, but completely different gastric emptying and glycemic responseThe classic comparison: whole apple vs apple juice vs apple-flavored gummies — similar calories, vastly different in-body behavior
4. Flavoring and central effects of additives:
UPF's hyper-palatability comes from precisely engineered sugar + fat + salt + flavoring combinationsDopamine reward pathways activated repeatedly → behavioral-addiction-like neural adaptation (Schulte 2015)Bliss point (Howard Moskowitz 1970s experiments) — food scientists specifically hunt for the 'can't stop' combination
5. Microbiome + emulsifiers:
Industrial emulsifiers disrupt gut microbiota + inflammation + metabolic syndrome in mice (Chassaing 2015 Nature). ⚠️ That study tested carboxymethylcellulose and polysorbate-80 — it never tested carrageenan, which often gets named alongside themHuman evidence is still accumulating
6. Sugar-sweetened beverages' special position:
Liquid calories don't trigger satiety (Pan 2011, DiMeglio 2000)A 350 ml soft drink's ~150 kcal does not correspondingly reduce dinner intake by 150 kcal — it's almost pure additional intakeIf you can only intervene on one thing, cutting sugar-sweetened beverages has the highest ROI
hamano-2024-upf-crossover
Chapter 3
Engineered, not accidental
Engineered, not accidental
UPF didn't accidentally become harmful — it was engineered to be:
Bliss point:
Howard Moskowitz, a food scientist in the 1970s–80s, did research for Coca-Cola / General Foods / VlasicThrough systematic tuning of hundreds of versions, he found the sugar + fat + salt + flavor combination that makes people eat / drink the mostThis ratio is not 'most delicious', it's 'hardest to stop' — they aren't the sameVlasic 'Zesty' pickles / Prego marinara / Dr Pepper / Oreo are all bliss-point design examples
Vanishing caloric density:
Cheetos / chips / corn puffs → melt in the mouth — the brain registers 'didn't eat much' while density is sky-highFood scientist Steven Witherly's book *Why Humans Like Junk Food* describes thisThe chemistry: rapid melt → triggers swallowing → no chewing needed → brain registers 'haven't finished'
Key tools in formulation:
Flavor systems: engineered combinations across thousands of 'natural / artificial' flavor compoundsTexture modifiers: emulsifiers + thickeners + shortening → recreate the mouthfeel of cream / butter / full-fat milkSweetness systems: HFCS + polysaccharides + artificial sweeteners + enhancers combined → more tolerable than pure sucroseSalt modulation: MSG + 5'-inosinate + sodium glutamate + potassium salt partly replacing sodium → 'low sodium' label + same saltiness
Economic incentives:
UPF gross margins are typically 30–60% vs fresh foods 5–15%Long shelf life reduces logistics and inventory lossStandardized production reduces costBrand + marketing value: most of a bag of chips' price is brand, not ingredientsResult: the food industry concentrates marketing budgets on UPF, while fresh food vendors have no marketing budget
Big Food's tobacco-industry playbook:
Marion Nestle (NYU) + Michael Pollan + Anthony Warner have shown: the food industry adopted the full 1960s–90s tobacco industry playbookFund favorable research + suppress unfavorable (Kearns 2016 JAMA Intern Med exposed sugar industry funding of Harvard in the 1960s)Shift attention ('it's not sugar, it's fat' / 'it's not UPF, it's lack of exercise')Regulatory revolving door (former FDA / USDA officials moving to the food industry and back)'Personal responsibility' rhetoric (responsibility lies with the consumer, not the product)Direct result: UPF improves too slowly and regulation is too weak
This section isn't atlas's routine debunking — it's a statement of structural problem:
Individual choice matters, but 'eat less UPF' is a privilege of the few when the food environment is severely unbalancedTime, cooking skills, price, accessibility, and marketing barrage all favor UPFWhat this island can do is lay out the mechanisms so you can see them, and let you make choices within your capacity.
Bliss point:
Howard Moskowitz, a food scientist in the 1970s–80s, did research for Coca-Cola / General Foods / VlasicThrough systematic tuning of hundreds of versions, he found the sugar + fat + salt + flavor combination that makes people eat / drink the mostThis ratio is not 'most delicious', it's 'hardest to stop' — they aren't the sameVlasic 'Zesty' pickles / Prego marinara / Dr Pepper / Oreo are all bliss-point design examples
Vanishing caloric density:
Cheetos / chips / corn puffs → melt in the mouth — the brain registers 'didn't eat much' while density is sky-highFood scientist Steven Witherly's book *Why Humans Like Junk Food* describes thisThe chemistry: rapid melt → triggers swallowing → no chewing needed → brain registers 'haven't finished'
Key tools in formulation:
Flavor systems: engineered combinations across thousands of 'natural / artificial' flavor compoundsTexture modifiers: emulsifiers + thickeners + shortening → recreate the mouthfeel of cream / butter / full-fat milkSweetness systems: HFCS + polysaccharides + artificial sweeteners + enhancers combined → more tolerable than pure sucroseSalt modulation: MSG + 5'-inosinate + sodium glutamate + potassium salt partly replacing sodium → 'low sodium' label + same saltiness
Economic incentives:
UPF gross margins are typically 30–60% vs fresh foods 5–15%Long shelf life reduces logistics and inventory lossStandardized production reduces costBrand + marketing value: most of a bag of chips' price is brand, not ingredientsResult: the food industry concentrates marketing budgets on UPF, while fresh food vendors have no marketing budget
Big Food's tobacco-industry playbook:
Marion Nestle (NYU) + Michael Pollan + Anthony Warner have shown: the food industry adopted the full 1960s–90s tobacco industry playbookFund favorable research + suppress unfavorable (Kearns 2016 JAMA Intern Med exposed sugar industry funding of Harvard in the 1960s)Shift attention ('it's not sugar, it's fat' / 'it's not UPF, it's lack of exercise')Regulatory revolving door (former FDA / USDA officials moving to the food industry and back)'Personal responsibility' rhetoric (responsibility lies with the consumer, not the product)Direct result: UPF improves too slowly and regulation is too weak
This section isn't atlas's routine debunking — it's a statement of structural problem:
Individual choice matters, but 'eat less UPF' is a privilege of the few when the food environment is severely unbalancedTime, cooking skills, price, accessibility, and marketing barrage all favor UPFWhat this island can do is lay out the mechanisms so you can see them, and let you make choices within your capacity.
China-specific industrial path
China's UPF industrial localization characteristics (2000s–2020s):Foreign entry → domestic rise:
1990s: McDonald's (1990) / KFC (1987) / Coca-Cola (1979) entered2000s: domestic giants rose — Master Kong / Uni-President (instant noodles) / Nongfu Spring (beverages) / Three Squirrels (snacks) / HeyTea / Mixue2010s onward: e-commerce + delivery + short-video marketing pushed UPF to tier-3 and tier-4 cities
'National food' UPF-ification:
Instant noodles: invented by Japan's Nissin in 1958, exploded in China from the 1980s — now a ¥220 billion/year market, per-capita 30–40 packs/year (highest globally)Packaged snacks: duck neck, latiao (konjac + heavy flavorings + seasonings), went from small workshops to industrial production in the 2010sTea drinks + sugary beverages: China's per-capita 16–25 kg/year sugar intake comes mainly from tea drinks and soft drinks (Wang 2023 BMC Public Health)Packaged breakfast: whole-grain bread / purple sweet potato buns / oat bran / high-fiber cookies — reading the ingredient list closely, most are UPF with sugar + HFCS + additives
'Health claim' packaging is the most deceptive:
'0 sucrose' → usually contains HFCS + maltodextrin + artificial sweeteners (still UPF, still unhealthy)'Low fat' → usually compensated with sugar + thickeners (deeper UPF)'Whole grain' → 5% whole wheat flour + 95% refined flour + caramel color + malt extract'Protein' bar → 10+ additives + hidden sugar'Children's / pregnancy / elderly' nutrition: marketing labels, ingredient list usually identical to the adult version'Imported / influencer / anti-inflammatory / detox': marketing words, usually premium-priced UPF
Ingredient list reading 30-second rule:
1. Count items: 5+ is usually UPF (home recipes rarely have 5+ ingredients)
2. What's the first item: the first item is the largest by weight. If it's 'water' / 'refined flour' / 'sugar' / 'syrup' / 'HFCS' → warning
3. Look for 'not in a home kitchen' words: emulsifiers / thickeners / flavorings / flavor essences / maltodextrin / carrageenan / anti-caking agents / preservatives / 'E-numbers'
4. Check sugar per serving: > 5 g per serving + a second serving the next day → cumulative danger
5. Check trans fat: '0 g trans fat' ≠ 0; regulations allow < 0.5 g per serving to be labeled 0; look for 'partially hydrogenated oil / shortening / non-dairy creamer' wording
How to break the pattern (practical for Chinese families):
Cutting the three biggest categories yields the biggest gains: sugar-sweetened beverages → milk tea → instant noodles + frozen foodsReplacements: home-cooked noodles + egg + greens (5 minutes, complete nutrition, < ¥5/meal)Breakfast: simple boiled egg + milk + a slice of ordinary bread + an apple → better than any 'nutritious breakfast' packaged foodSnacks: nuts + fruit + plain yogurt (with honey added by you) → replaces puffed snacks + packaged pastriesTakeout: hard to fully avoid, but choosing simple cuisines (stir-fried / steamed / boiled) over heavily seasoned dishes (spicy / oily / heavy seasoning) reduces UPF
Kids are the critical population:
Ages 6–18 are the taste preference formation period; UPF exposure shapes lifelong food preferencesThe cultural expectation of 'buying snacks for kids is how I express love' is the deepest hook of UPF marketingSolution: replace with 'labor of love' — cook together, cut fruit together, fold dumplings together
du-2022-china-upfli-shi-2021-china-adult-upfli-shi-2025-china-child-upfchina-added-sugar-adults-2023china-added-sugar-children-2022chassaing-2015-emulsifiers-microbiotaforde-2020-energy-intake-raterobinson-2014-eating-rate-metavolkow-2008-addiction-obesity-circuitshernandez-1988-food-cocaine-dopamine
Chapter 4
Disease links + meta
Disease links + meta
UPF intake dose-response with hard endpoints (Lane 2023 BMJ umbrella meta + GBD 2024):
Lane 2023 BMJ — synthesis of 14 meta-analyses covering 9.8 million people across 32 health outcomes:
Direct evidence (strong / convincing):All-cause mortality: every +10% UPF share of total energy → +14% (HR 1.14, 95% CI 1.10–1.18)Cardiovascular disease: +12% (HR 1.12)Type 2 diabetes: +12–15%Depression: +22% — but this one is pooled observational (Lane 2022, HR 1.22), so it belongs a tier lower than the rows above itTotal cancer: +6% (Fiolet 2018 BMJ)
Highly suggestive:Obesity / overweight: +3–4 kg/m² BMI in top vs bottom UPF quintile (Mendonça 2017)Coronary heart disease + stroke: +8–15%Breast cancer: +11% (Fiolet 2018)Colorectal cancer: +12%NAFLD / MASLD biomarkers (Konieczna 2022, PREDIMED-Plus, n = 5,867): each +10% UPF intake over one year → fatty liver index +1.6 points, hepatic steatosis index +0.43. Note this is a continuous index shift, not an incidence risk ratio
Moderate evidence:Inflammatory bowel disease (Crohn + UC): +70–80% (Narula 2021 BMJ). ⚠️ The French NutriNet-Santé cohort (Vasseur 2021) found no association — one positive cohort and one null, not two positivesChildhood allergy / asthma: +20–40%Male sperm quality: associated with female conception difficulty
Key confounders:
UPF intake strongly correlates with obesity, income, education, and exercise → must be controlledThe Lane 2023 + Fiolet 2018 + Bonaccio 2022 + Srour 2019 series of studies show the association persists after full adjustmentBut the gold-standard 'complete causality' RCT count remains low — only Hall 2019 directly proved causality
Dose-response:
UPF intake and hard endpoints have a near-linear relationship — no clear 'safety threshold' (like smoking or alcohol)'Zero dose' has no 'lowest risk' plateau either — but reductions at any level are beneficial
Compared with traditional dietary patterns:
Mediterranean / DASH / Nordic diets = NOVA 1–3 dominant + UPF < 10–20%Typical American / modern urban Chinese diets = UPF 30–60%The essence of a 'good dietary pattern' is partly just 'less UPF', not some secret nutrient
Clinical intervention evidence:
MedDiet replacement RCT (Estruch 2018 NEJM PREDIMED): replacing UPF with MedDiet → cardiovascular events -30%
Bottom line:
UPF intake and hard endpoint associations come from multiple sources + consistent direction + dose-response + reversibility — satisfying the causal evidence chainCompared with any single 'superfood' or nutrient intervention, lowering UPF share is the broadest, highest-ROI dietary intervention
Lane 2023 BMJ — synthesis of 14 meta-analyses covering 9.8 million people across 32 health outcomes:
Direct evidence (strong / convincing):All-cause mortality: every +10% UPF share of total energy → +14% (HR 1.14, 95% CI 1.10–1.18)Cardiovascular disease: +12% (HR 1.12)Type 2 diabetes: +12–15%Depression: +22% — but this one is pooled observational (Lane 2022, HR 1.22), so it belongs a tier lower than the rows above itTotal cancer: +6% (Fiolet 2018 BMJ)
Highly suggestive:Obesity / overweight: +3–4 kg/m² BMI in top vs bottom UPF quintile (Mendonça 2017)Coronary heart disease + stroke: +8–15%Breast cancer: +11% (Fiolet 2018)Colorectal cancer: +12%NAFLD / MASLD biomarkers (Konieczna 2022, PREDIMED-Plus, n = 5,867): each +10% UPF intake over one year → fatty liver index +1.6 points, hepatic steatosis index +0.43. Note this is a continuous index shift, not an incidence risk ratio
Moderate evidence:Inflammatory bowel disease (Crohn + UC): +70–80% (Narula 2021 BMJ). ⚠️ The French NutriNet-Santé cohort (Vasseur 2021) found no association — one positive cohort and one null, not two positivesChildhood allergy / asthma: +20–40%Male sperm quality: associated with female conception difficulty
Key confounders:
UPF intake strongly correlates with obesity, income, education, and exercise → must be controlledThe Lane 2023 + Fiolet 2018 + Bonaccio 2022 + Srour 2019 series of studies show the association persists after full adjustmentBut the gold-standard 'complete causality' RCT count remains low — only Hall 2019 directly proved causality
Dose-response:
UPF intake and hard endpoints have a near-linear relationship — no clear 'safety threshold' (like smoking or alcohol)'Zero dose' has no 'lowest risk' plateau either — but reductions at any level are beneficial
Compared with traditional dietary patterns:
Mediterranean / DASH / Nordic diets = NOVA 1–3 dominant + UPF < 10–20%Typical American / modern urban Chinese diets = UPF 30–60%The essence of a 'good dietary pattern' is partly just 'less UPF', not some secret nutrient
Clinical intervention evidence:
MedDiet replacement RCT (Estruch 2018 NEJM PREDIMED): replacing UPF with MedDiet → cardiovascular events -30%
Bottom line:
UPF intake and hard endpoint associations come from multiple sources + consistent direction + dose-response + reversibility — satisfying the causal evidence chainCompared with any single 'superfood' or nutrient intervention, lowering UPF share is the broadest, highest-ROI dietary intervention
Depression + UPF
The UPF-depression association is one of the most overlooked health findings of the past 5 years:Key studies:
Lane 2022 *Nutrients* — meta-analysis of 17 observational studies (n = 385,541; 15 cross-sectional, 2 prospective). Pooled prospective HR 1.22 (1.16-1.28) for higher vs lower UPF intake. This is where the widely quoted +22% comes from — not from any RCTAdjibade 2019 BMC Med (NutriNet-Santé N = 26,000): UPF share rising is significantly associated with depressive symptoms risingLai 2014 AJCN is observational, not an RCT: 13 pooled observational studies, healthy dietary pattern associated with lower depression risk (OR 0.84), with the authors explicitly unable to establish temporalityJacka 2017 SMILES Trial: Mediterranean diet (essentially low-UPF) RCT significantly outperformed social support in depression treatment — the first nutrition-psychiatry RCT proving causality
Possible mechanisms:
Gut-brain axis: UPF emulsifiers + additives → gut microbiota dysbiosis + endotoxin → systemic inflammation → blood–brain barrier: The 'security gate' on brain vessels that blocks most substances in blood from entering the brain. permeability → central inflammation → changes in 5-HT / dopamine / BDNFMicronutrient deficits: UPF replaces whole foods → reduced intake of B vitamins + magnesium + zinc + omega-3 + polyphenols → impaired neurotransmitter synthesis and neuroplasticityGlycemic fluctuations: high-GI UPF → sharp postprandial glucose → reactive hypoglycemia → anxiety / fatigueReward pathway neural adaptation: UPF dopamine repeats → addiction-like neural circuit → natural rewards (exercise / sunlight / social) numbed → 'nothing makes me happy' core depression experience
Clinical translation:
Mediterranean diet is now recommended adjunct treatment for depression (Australian + UK + Canadian guidelines)Replacing SSD (sugar-sweetened drinks) with water is the single highest-ROI interventionIntegrated psychiatry: increasingly, clinicians discuss dietary patterns alongside SSRI prescriptions
Connections to other atlas content:
vitamin-d/immune L4 (D and mood / seasonal affective disorder)fats-omega-3 (EPA depression meta Mocking 2016)digestive/microbiome + endocrine/hypothalamic–pituitary–adrenal axis: The body's stress-response chain (hypothalamus → pituitary → adrenal) that releases cortisol. (gut-brain axis + stress response)insomnia story (sleep + mood)These stories + UPF together form a complete atlas of mental health nutrition intervention
Important warning:
UPF intervention ≠ substitute for antidepressants — moderate-to-severe depression still needs SSRI / psychotherapyUPF improvement is adjunct / upstream interventionClinical depression (PHQ-9 ≥ 10) should see psychiatry
adjibade-2019-upf-depressionjacka-2017-smiles-triallane-2022-upf-mental-health-metalai-2014-dietary-patterns-depression-meta
Chapter 5
Practical decision tree
Practical decision tree
'How to reduce UPF' — realistically executable version:
Step 1 · identify:
Log everything you eat for one week, tag each item N1 (unprocessed) / N2 (culinary ingredient) / N3 (processed) / N4 (UPF)Don't aim for perfection — estimating in the 60% / 30% / 5% / 5% magnitude is enoughYou'll likely find more N4 than you imagined — that's an honest starting point
Step 2 · cut the three highest-leverage categories (highest ROI):
First choice · sugar-sweetened beverages (SSB), single highest ROI
Includes soft drinks, milk tea, juice, flavored coffee, sports drinks.Replace with: water, unsweetened tea, black coffee, soda water + lemon.One milk tea is ~300–500 kcal and 30–60 g sugar; cutting it for a week saves 2000–3500 kcal + 200–400 g sugar.
Second choice · packaged snacks
Includes chips, cookies, candy, industrial cake, chocolate (high-sugar industrial type), puffed snacks.Replace with: nuts, fruit, plain yogurt, a square of ≥ 70% cocoa dark chocolate.
Third choice · processed meats
Includes sausage, bacon, ham, lunch meat, fish balls, meat floss.Replace with: fresh meat / fish / eggs / legumes.WHO IARC classifies processed meat as a Group 1 carcinogen; 50 g/day → ~18% increase in colorectal cancer risk.
Step 3 · three low-cost swaps:
Breakfast: packaged cereal + flavored yogurt → boiled egg + milk + oats + a fruitLunch takeout: spicy hot pot / Shaxian → simple stir-fry / steamed rice + 1 vegetable + 1 proteinDinner frozen dumplings / instant noodles: replace with fresh noodles + egg + greens + a few meat slices (5 minutes, < ¥5)
Step 4 · no perfectionism:
Target: UPF from 50–60% → 20–30%Going to zero is unnecessary (birthday cake / occasional milk tea / holiday snacks preserve quality of life)80/20 principle: 80% clean eating, 20% flexible — sustainable long-term
Step 5 · family + social environment:
Don't stock UPF at home: you won't run downstairs at midnight to buy chips, but you will open the pantry and grab a bagAlign with family / roommates: consistency beats going aloneSupermarket shopping list by NOVA 1–3: one weekly tripRestaurant choice: home-style Chinese > Western fast food > industrial milk tea shops
Step 6 · when to seek professionals:
Binge eating / loss of appetite control: see a psychologist — this isn't a 'self-discipline issue'Clear obesity + multiple metabolic diseases: dietitian + endocrine clinicDepression + high UPF intake: psychiatry + nutrition counseling (doesn't replace medication)
Wrapping up: UPF isn't 'poison' — it's 'structurally restructured food'. Occasional consumption is fine; but when 60% of calories come from UPF for decades, 30 years later your health trajectory is noticeably changed. This is a collective environmental problem, not entirely attributable to personal choice — but personal choices at the margin do matter, and getting them right at the margin is enough.
What we can do is explain the mechanisms clearly and lay out the evidence — neither selling panic nor selling perfectionism — and the decision is still yours.
Step 1 · identify:
Log everything you eat for one week, tag each item N1 (unprocessed) / N2 (culinary ingredient) / N3 (processed) / N4 (UPF)Don't aim for perfection — estimating in the 60% / 30% / 5% / 5% magnitude is enoughYou'll likely find more N4 than you imagined — that's an honest starting point
Step 2 · cut the three highest-leverage categories (highest ROI):
First choice · sugar-sweetened beverages (SSB), single highest ROI
Includes soft drinks, milk tea, juice, flavored coffee, sports drinks.Replace with: water, unsweetened tea, black coffee, soda water + lemon.One milk tea is ~300–500 kcal and 30–60 g sugar; cutting it for a week saves 2000–3500 kcal + 200–400 g sugar.
Second choice · packaged snacks
Includes chips, cookies, candy, industrial cake, chocolate (high-sugar industrial type), puffed snacks.Replace with: nuts, fruit, plain yogurt, a square of ≥ 70% cocoa dark chocolate.
Third choice · processed meats
Includes sausage, bacon, ham, lunch meat, fish balls, meat floss.Replace with: fresh meat / fish / eggs / legumes.WHO IARC classifies processed meat as a Group 1 carcinogen; 50 g/day → ~18% increase in colorectal cancer risk.
Step 3 · three low-cost swaps:
Breakfast: packaged cereal + flavored yogurt → boiled egg + milk + oats + a fruitLunch takeout: spicy hot pot / Shaxian → simple stir-fry / steamed rice + 1 vegetable + 1 proteinDinner frozen dumplings / instant noodles: replace with fresh noodles + egg + greens + a few meat slices (5 minutes, < ¥5)
Step 4 · no perfectionism:
Target: UPF from 50–60% → 20–30%Going to zero is unnecessary (birthday cake / occasional milk tea / holiday snacks preserve quality of life)80/20 principle: 80% clean eating, 20% flexible — sustainable long-term
Step 5 · family + social environment:
Don't stock UPF at home: you won't run downstairs at midnight to buy chips, but you will open the pantry and grab a bagAlign with family / roommates: consistency beats going aloneSupermarket shopping list by NOVA 1–3: one weekly tripRestaurant choice: home-style Chinese > Western fast food > industrial milk tea shops
Step 6 · when to seek professionals:
Binge eating / loss of appetite control: see a psychologist — this isn't a 'self-discipline issue'Clear obesity + multiple metabolic diseases: dietitian + endocrine clinicDepression + high UPF intake: psychiatry + nutrition counseling (doesn't replace medication)
Wrapping up: UPF isn't 'poison' — it's 'structurally restructured food'. Occasional consumption is fine; but when 60% of calories come from UPF for decades, 30 years later your health trajectory is noticeably changed. This is a collective environmental problem, not entirely attributable to personal choice — but personal choices at the margin do matter, and getting them right at the margin is enough.
What we can do is explain the mechanisms clearly and lay out the evidence — neither selling panic nor selling perfectionism — and the decision is still yours.
六步实操 · 从记录到替换
怎么减少 UPF — 现实可执行版:Step 1 · 识别:
一周记录所有吃的, 标 N1 (未加工) / N2 (烹饪原料) / N3 (加工) / N4 (UPF)不需要完美——估计 60% / 30% / 5% / 5% 这个量级即可你大概率会发现 N4 比想象多——这是诚实的起点
Step 2 · 砍三类最大杠杆 (性价比最高):
首选 · 含糖饮料 (SSB), 单一最高性价比
包含软饮、奶茶、果汁、调味咖啡、运动饮料。替代: 水、无糖茶、黑咖啡、苏打水 + 柠檬。一杯奶茶大约 300-500 kcal、30-60 g 糖, 一周砍掉等于减 2000-3500 kcal + 200-400 g 糖。
次选 · 包装零食
包含薯片、饼干、糖果、工业蛋糕、巧克力 (高糖工业型)、膨化食品。替代: 坚果、水果、原味酸奶、一块 ≥ 70% 可可的黑巧。
第三选 · 加工肉
包含香肠、培根、火腿、午餐肉、鱼丸、肉松。替代: 新鲜肉、鱼 / 蛋、豆。WHO IARC 把加工肉列为 1 类致癌物, 每天 50 g → 结直肠癌风险大约升 18%。
Step 3 · 三个低成本替换:
早餐: 包装麦片 + 调味酸奶 → 煮蛋 + 牛奶 + 燕麦 + 一个水果午餐外卖: 麻辣烫、沙县 → 简单炒菜、蒸饭 + 一份菜 + 一份肉晚餐速冻饺子、即食面: 替换为新鲜挂面 + 鸡蛋 + 青菜 + 几片肉 (5 分钟, < 5 元)
Step 4 · 不要完美主义:
目标: UPF 从 50-60% → 20-30%完全清零是不必要的 (生日蛋糕、偶尔奶茶、节日零食保留生活质量)80/20 原则: 80% 正常吃干净, 20% 灵活, 长期可持续
Step 5 · 家庭 + 社会环境:
不在家里囤 UPF: 你不会半夜冲下楼买薯片, 但会打开柜子取一袋和家人、室友达成共识: 一致比单打独斗强超市购物清单按 NOVA 1-3 类列: 周末一次性采购饭店选择: 中餐家常 > 西式快餐 > 工业化奶茶店
Step 6 · 何时找专业人士:
暴食、食欲失控: 看心理医生, 这不是自律问题明确肥胖 + 多种代谢病: 营养师 + 内分泌科抑郁 + UPF 高摄入: 精神科 + 营养辅导 (不替代药物)
Packaging traps
'Health' words on packaging are mostly UPF red flags — quick identification checklist:Definite warnings (these words = likely UPF):
'0 sucrose' → check the ingredients: usually contains HFCS / maltodextrin / artificial sweeteners / concentrated fruit juice'Low fat / fat-free' → check the ingredients: usually compensated with sugar + thickeners (deeper UPF)'Whole grain' → if the first ingredient isn't 'whole wheat flour / oats' but 'wheat flour + caramel color' → fake whole grain'Dietary fiber' → often added through processed fiber (inulin + maltodextrin + resistant dextrin), not natural fiber'Protein +6g / 10g' → check the protein source: whey isolate / soy isolate → UPF; milk / eggs / meat → not'Children / pregnancy / elderly' nutrition: marketing labels — ingredient lists usually identical to the adult version'Imported / influencer / anti-inflammatory / detox': marketing words, usually premium-priced UPF
Ingredient list 30-second rule:
1. Count items: 5+ is usually UPF (home recipes rarely have 5+ items)
2. What's the first item: first ingredient is the largest by weight. If it's 'water' / 'refined flour' / 'sugar' / 'syrup' / 'HFCS' → warning
3. Look for 'not in a home kitchen' words: emulsifiers / thickeners / flavorings / flavor essences / maltodextrin / carrageenan / anti-caking agents / preservatives / 'E-numbers'
4. Check sugar per serving: > 5 g per serving + a second serving the next day → cumulative danger
5. Check trans fat: '0 g trans fat' ≠ 0; regulations allow < 0.5 g/serving to be labeled 0; look for 'partially hydrogenated oil / shortening / non-dairy creamer' wording
Common UPF disguises:
| Looks healthy | Actually is |
|---|---|
| Flavored yogurt | Sugar + thickeners + flavorings |
| Instant oatmeal cups | Sugar + syrup + flavorings |
| Whole-wheat cookies | 5% whole wheat + 95% refined sugar and oil |
| Vegetable chips | 0.5% spinach powder + 99% industrial chips |
| Protein bar | 10+ additives + hidden sugar |
| 'Functional' drinks | Sugar + caffeine + vitamins (get vitamins from food) |
| Breakfast cereal | Sugar + malt syrup + caramel color + flavorings |
| 'High-end' infant rice powder | Simple ground rice + marketing premium (most versions OK) |
Best practice: 'foods a home kitchen can make' = default safe; 'only a factory can make' = default scrutinize.
References · 10
- Monteiro, C. A., Cannon, G., Moubarac, J.-C., Levy, R. B., Louzada, M. L. C., & Jaime, P. C. (2018). The UN Decade of Nutrition, the NOVA food classification and the trouble with ultra-processing. Public Health Nutrition, 21(1), 5-17. 10.1017/S1368980017000234
- Monteiro, C. A., Cannon, G., Levy, R. B., Moubarac, J. C., Louzada, M. L. C., Rauber, F., et al. (2019). Ultra-processed foods: what they are and how to identify them. Public Health Nutrition, 22(5), 936-941. 10.1017/S1368980018003762
- Hall, K. D., Ayuketah, A., Brychta, R., Cai, H., Cassimatis, T., Chen, K. Y., et al. (2019). Ultra-processed diets cause excess calorie intake and weight gain: an inpatient randomized controlled trial of ad libitum food intake. Cell Metabolism, 30(1), 67-77.e3. 10.1016/j.cmet.2019.05.008
- Nestle, M. (2013). Food Politics: How the Food Industry Influences Nutrition and Health (Revised and Expanded 10th Anniversary Edition). University of California Press. www.ucpress.edu/book/9780520275966/food-politics
- Kearns, C. E., Schmidt, L. A., & Glantz, S. A. (2016). Sugar industry and coronary heart disease research: a historical analysis of internal industry documents. JAMA Internal Medicine, 176(11), 1680-1685. 10.1001/jamainternmed.2016.5394
- Lane, M. M., Gamage, E., Du, S., Ashtree, D. N., McGuinness, A. J., Gauci, S., et al. (2024). Ultra-processed food exposure and adverse health outcomes: umbrella review of epidemiological meta-analyses. BMJ, 384, e077310. 10.1136/bmj-2023-077310
- Fiolet, T., Srour, B., Sellem, L., Kesse-Guyot, E., Allès, B., Méjean, C., et al. (2018). Consumption of ultra-processed foods and cancer risk: results from NutriNet-Santé prospective cohort. BMJ, 360, k322. 10.1136/bmj.k322
- Konieczna, J., Fiol, M., Colom, A., Martínez-González, M. Á., Salas-Salvadó, J., Corella, D., et al. (2022). Does Consumption of Ultra-Processed Foods Matter for Liver Health? Prospective Analysis among Older Adults with Metabolic Syndrome. Nutrients, 14(19), 4142. PREDIMED-Plus, continuous liver-index outcomes, not incident NAFLD risk. No "+30-40%" figure exists in it. 10.3390/nu14194142
- Vasseur, P., Dugelay, E., Benamouzig, R., Savoye, G., Lan, A., Srour, B., et al. (2020). Dietary Patterns, Ultra-processed Food, and the Risk of Inflammatory Bowel Diseases in the NutriNet-Santé Cohort. Inflammatory Bowel Diseases, 27(1), 65-73. A NULL result: neither dietary patterns nor UPF proportion were significantly associated with incident IBD (75 cases in 105,832 participants). It was cited as joint support for a +70-80% risk increase. 10.1093/ibd/izaa018
- Bouvard, V., Loomis, D., Guyton, K. Z., Grosse, Y., Ghissassi, F. E., Benbrahim-Tallaa, L., et al. (2015). Carcinogenicity of consumption of red and processed meat. Lancet Oncology, 16(16), 1599-1600. IARC classifies processed meat as Group 1 carcinogen, red meat as Group 2A. 10.1016/S1470-2045(15)00444-1