Place · Level 3 · Macro
Artificial & Non-nutritive Sweeteners
阿斯巴甜 IARC 2B (2023) · 赤藓糖醇 Hazen 2023 Nat Med MACE · 三氯蔗糖肠菌 · 天然 ≠ 安全 · 0 卡 ≠ 中性
Last updated
Story path
Chapter 1
Eight families · landscape
Eight families · landscape
'Artificial sweetener' is a colloquial umbrella covering eight chemically distinct families with widely different metabolic and safety profiles. Lumping them together is where most misunderstandings start.
1. Aspartame — discovered 1965, FDA-approved 1981
Chemistry: L-aspartate + L-phenylalanine methyl ester; ~200× the sweetness of sucroseMetabolism: hydrolyzed in the gut to aspartate + phenylalanine + trace methanolContraindicated in phenylketonuria (PKU) — these patients cannot metabolize phenylalanine normallyBrands: NutraSweet, Equal; used in virtually all 'Zero' diet sodasIn 2023 WHO IARC classified it Group 2B 'possibly carcinogenic' on limited human + limited animal evidence
2. Sucralose (Splenda) — discovered 1976
Chemistry: three -OH groups on sucrose replaced with Cl; not metabolized by humans (~85% excreted)Sweetness ~600× sucroseStrength: heat-stable, doesn't decompose in bakingNew concern: Schiffman 2023 *J Toxicol Environ Health* reported that sucralose-6-acetate — partly a manufacturing residue, partly produced when gut bacteria transform sucralose — is genotoxic in vitro. The concern is that molecule, not baking. FDA + EFSA still consider sucralose safe, but regulators are re-evaluating
3. Saccharin (Sweet'N Low) — discovered 1879, the oldest artificial sweetener
Not metabolized, excreted directly1970s rat bladder cancer → 1980s warning label → later shown to be rat-specific → 2001 warning removedSweetness ~300-400× sucrose
4. Acesulfame-K (Ace-K): not metabolized, ~200× sucrose — usually blended with other sweeteners (the blend tastes closer to sugar).
5. Stevia (RebA) — marketed as 'natural'
Extract from the South American *Stevia rebaudiana* plantMain sweet molecules are stevioside and rebaudioside ANot metabolized in humans; sweetness ~200-300× sucrose'Natural = safe' is mostly marketing rhetoric: FDA only approves high-purity rebaudioside A; whole-leaf stevia remains pending GRAS review
6. Monk fruit (Luo Han Guo): traditional Chinese sweetener; key molecule mogroside V; ~200× sucrose; FDA GRAS 2010.
Those six families above are all non-nutritive — not metabolized, near-zero calories. But the 'sweetener' umbrella also covers two more families that do carry calories and have real metabolic effects — lumping those in with the first six is another common misunderstanding. Turn the page for sugar alcohols + allulose.
1. Aspartame — discovered 1965, FDA-approved 1981
Chemistry: L-aspartate + L-phenylalanine methyl ester; ~200× the sweetness of sucroseMetabolism: hydrolyzed in the gut to aspartate + phenylalanine + trace methanolContraindicated in phenylketonuria (PKU) — these patients cannot metabolize phenylalanine normallyBrands: NutraSweet, Equal; used in virtually all 'Zero' diet sodasIn 2023 WHO IARC classified it Group 2B 'possibly carcinogenic' on limited human + limited animal evidence
2. Sucralose (Splenda) — discovered 1976
Chemistry: three -OH groups on sucrose replaced with Cl; not metabolized by humans (~85% excreted)Sweetness ~600× sucroseStrength: heat-stable, doesn't decompose in bakingNew concern: Schiffman 2023 *J Toxicol Environ Health* reported that sucralose-6-acetate — partly a manufacturing residue, partly produced when gut bacteria transform sucralose — is genotoxic in vitro. The concern is that molecule, not baking. FDA + EFSA still consider sucralose safe, but regulators are re-evaluating
3. Saccharin (Sweet'N Low) — discovered 1879, the oldest artificial sweetener
Not metabolized, excreted directly1970s rat bladder cancer → 1980s warning label → later shown to be rat-specific → 2001 warning removedSweetness ~300-400× sucrose
4. Acesulfame-K (Ace-K): not metabolized, ~200× sucrose — usually blended with other sweeteners (the blend tastes closer to sugar).
5. Stevia (RebA) — marketed as 'natural'
Extract from the South American *Stevia rebaudiana* plantMain sweet molecules are stevioside and rebaudioside ANot metabolized in humans; sweetness ~200-300× sucrose'Natural = safe' is mostly marketing rhetoric: FDA only approves high-purity rebaudioside A; whole-leaf stevia remains pending GRAS review
6. Monk fruit (Luo Han Guo): traditional Chinese sweetener; key molecule mogroside V; ~200× sucrose; FDA GRAS 2010.
Those six families above are all non-nutritive — not metabolized, near-zero calories. But the 'sweetener' umbrella also covers two more families that do carry calories and have real metabolic effects — lumping those in with the first six is another common misunderstanding. Turn the page for sugar alcohols + allulose.
六类非营养型 · 一类一条
人工甜味剂 这个口语词其实涵盖 8 个完全不同的化学家族, 它们的安全证据和代谢效应差异很大。把它们当一类讨论, 是大多数误解的起点。先看前 6 类非营养型 (几乎不被人体代谢、近零热量), 一行一个:
阿斯巴甜 (Aspartame): 甜度约蔗糖 200 倍, 在肠道被拆解; 苯丙酮尿症 (PKU) 患者禁用; 几乎所有零度软饮用的都是它。2023 年 WHO IARC 把它列为 2B 类可能致癌。三氯蔗糖 (Sucralose, Splenda): 甜度蔗糖 600×, 基本不被人体代谢, 耐热能烘焙; 新担忧不在烘焙, 而在一个叫三氯蔗糖-6-乙酸酯的分子 —— 它一部分是生产过程留下的残留, 一部分是肠菌把三氯蔗糖改造出来的, 体外试验显示它有基因毒性 (Schiffman 2023), 监管层在重新评估。糖精 (Saccharin): 最老的人工甜味剂, 甜度蔗糖 300-400×, 不代谢直接排出; 当年的大鼠膀胱癌警告后来被证明是大鼠特异性, 2001 年撤销。安赛蜜 (Acesulfame-K): 不代谢, 甜度约 200×, 常和别的甜味剂混用 (合起来更接近蔗糖味)。甜菊糖 (Stevia): 从南美甜叶菊提取, 不被人体代谢, 甜度约蔗糖 200-300 倍; 主打天然营销, 但天然不等于安全 (FDA 只认高纯度的那种提取成分)。罗汉果 (Monk Fruit): 中国传统甜味剂, 甜度约 200×, 2010 年获 FDA 认可。
为什么甜度倍数本身就是一条机制
甜味受体不管你放了多少克, 它只管有多少个分子卡进了它的口袋。上面这几类分子和受体贴得比蔗糖紧得多, 所以极少的量就能把受体按满 —— 你尝到的甜是满格的, 而真正进到身体里的物质只有针尖那么点。近零热量并不是因为它们身上没有能量, 是因为你根本用不着放多少。
阿斯巴甜是这六类里唯一真被拆开的那个
它一进小肠就被剪成两个氨基酸 (天冬氨酸 + 苯丙氨酸) 加一点甲醇。那两个氨基酸不是什么新鲜东西, 你吃一口鸡蛋拿到的比这多得多。它之所以还算零卡, 靠的不是不被代谢, 而是上面那一行写的甜度倍数 —— 用量太小, 拆出来那点氨基酸的热量可以忽略。
真正要盯的因此不是热量, 是被剪下来的苯丙氨酸落到谁身上。对绝大多数人它就是一枚普通氨基酸, 走的是和蛋白质里那些一模一样的路; 对少数天生处理不了它的人, 它会堵在血里 —— 这也是标签上写着含苯丙氨酸而不是含阿斯巴甜的原因 (细节在阿斯巴甜那一幕)。
但甜味剂 这把伞底下还有 2 类是有部分热量 + 真代谢效应的, 把它们和上面 6 类混为一谈是另一类误解 —— 翻一页看糖醇 + 阿洛酮糖.
Sugar alcohols + allulose · different from the 6 above
7. Sugar alcohols — distinct from the six above; they carry some calories and metabolic effects:Erythritol: 4-carbon, 0.24 kcal/g, non-fermenting (low FODMAP), ~70% sucrose sweetnessXylitol: 5-carbon, 2.4 kcal/g, anti-cariogenic (oral bacteria can't metabolize it) → toothpaste + gum; acutely lethal to dogsSorbitol / maltitol / isomalt / mannitol: partially fermented → bloating + osmotic diarrhea (overdo it and you'll have GI issues)
8. Allulose — emerging 'rare sugar':
Fructose isomer, ~70% sucrose sweetness, 0.4 kcal/g, doesn't enter the tricarboxylic acid (Krebs) cycle: The mitochondrial hub cycle that fully oxidizes fuel and harvests electrons for energy. cycleFDA excludes it from the 'total sugars' label (since 2019)Doesn't raise blood glucose — genuinely 'glucose-metabolism neutral' — but expensive
Why this matters:
'I don't eat sugar' while drinking lots of Diet Coke (aspartame) is not the same intervention as 'I use xylitol as a sugar substitute' — the first is metabolically neutral, the second carries 2.4 kcal/g and ferments'Sugar-free chocolate' (maltitol) causing diarrhea at higher intake is a direct consequence of that physical chemistryMetabolic effects and safety evidence differ family by family; the next scenes take each major health question one family at a time
China common-brand cheat sheet
Cheat sheet — ingredients in common '0 sugar' / 'sugar-free' drinks and foods on the Chinese market:Soft drinks / teas:
GenkiForest sparkling water: erythritol + sucraloseCoke Zero: aspartame + Ace-KPepsi Max: aspartame + Ace-KOriental Leaves (sugar-free): no sweetener (real plain tea)Nongfu Spring NFC: real juice, no added sugar but contains natural fructose (not 'zero-sugar')
Milk tea + coffee ('light' / 'sugar-free' labels):
Typically = 50% less sugar + aspartame / Ace-K / sucralose + still contains starch pearls / coconut jelly / jam (all high-sugar)'Sugar-free' milk tea ≠ low-calorie — can still hit 300-400 kcal
Packaged snacks ('0 sugar' label):
Sugar-free cookies / cakes: maltitol + sucralose + still contain flour + fat (carb + fat calories unchanged)Sugar-free chocolate: maltitol (causes diarrhea in excess) + cocoa
Children + infants:
Infant formula: artificial sweeteners are not permitted (Chinese + international regulation)Children's vitamin gummies: most contain sucrose / glucose syrup (not actually 'sugar-free')School-age '0 sugar' drinks: same as adult versions; safety data sparse. WHO's 2023 'not recommended for weight control' guidance applies to everyone except people with pre-existing diabetes — it is not a children-only rule, and WHO itself grades it a conditional recommendation
Diabetic / weight-loss populations:
'Diabetic cookies': typically use maltitol / sorbitol — low glycemic but not actually lower in calories + cause diarrhea at higher intakeSweeteners vs equivalent sucrose for short-term weight: Toews 2019 *BMJ* meta-analysis reports a BMI difference of about −0.6, resting on two small studies (174 people), rated low-to-very-low certainty. The issue isn't that it's 'non-significant' — it's that we have very little confidence in the number
The core reminder: '0 sugar' doesn't equal '0 calories', let alone 'healthy'. Judge a packaged food by reading the ingredient list + total calories + total carbs + protein + fat together — don't trust a single 'sugar-free' label.
Chapter 2
Aspartame IARC 2B
Aspartame IARC 2B
In July 2023 the WHO International Agency for Research on Cancer (IARC) classified aspartame as Group 2B 'possibly carcinogenic' — a headline that was widely misread; the truth needs to be unpacked.
The IARC classification system:
Group 1 (carcinogenic): tobacco, alcohol, processed meat, solar UV, asbestos, formaldehyde (sufficient human evidence)Group 2A (probably carcinogenic): red meat, night-shift work, acrylamide, very hot beverages above 65°C (limited human + sufficient animal evidence)Group 2B (possibly carcinogenic): aspartame, gasoline, aloe extract (limited human + limited animal evidence)Group 3 (not classifiable): most chemicals — including coffee itselfThe former Group 4 (probably not carcinogenic): abolished in the 2019 revision of the Monographs preamble; its single entry (caprolactam) was reassigned to Group 3
Those two lines deserve a second look, because they are routinely stated backwards: in 2016 the same IARC evaluation moved coffee down from 2B to Group 3 (no adequate evidence that coffee itself is carcinogenic) and classified drinking beverages above 65°C as Group 2A. The problem is the temperature, not the coffee — and that hazard sits one rung above aspartame. Using 'coffee causes cancer' as the worked example of Group 2B teaches this lesson backwards.
What IARC evaluates / doesn't evaluate:
IARC assesses **whether there is *any* carcinogenic potential, not the actual risk at typical doses**This is commonly confused: 'Group 2B' ≠ 'this dose is dangerous'Risk assessment is done by JECFA (FAO/WHO Joint Expert Committee on Food Additives)
The parallel JECFA 2023 evaluation:
Reaffirmed aspartame ADI (acceptable daily intake) at 40 mg/kg body weightA can of diet soft drink contains 200 or 300 mg of aspartame (WHO's own figure)A 70 kg adult's ADI = 2,800 mg — so you would need more than 9-14 cans per day to cross itThe vast majority of consumers are far below the ADI
The basis for the IARC 2B classification:
Primarily three observational epidemiology studies (NutriNet-Santé, Ramazzini rat experiments, partial liver-cancer signal in a cohort)Evidence rated 'limited' — insufficient to rule out but not strongControversial: Ramazzini methodology has been criticized; the population studies have confounding (diabetics + obese populations have elevated baseline cancer risk and also higher sweetener intake)
Real-world meaning:
1-2 Diet Cokes/day: far below ADI, weak epidemiology, very small risk — but not zero5-10/day: on WHO's per-can figure this already reaches or crosses the low end of the ADI — worth reconsideringChildren / pregnant women: limited data, caution is reasonableSwitching to water / tea: safer at any level
How to read the 'Diet Coke and cancer' story:
Clickbait: 'Aspartame causes cancer' — technically wrong; IARC didn't say 'causes', it said 'possible'Truth: weak evidence + small actual risk, but not 'completely safe'Better than regular sugar? For someone who already drinks soda regularly, the Diet version is probably slightly better (no glucose + weight); but the best choice is not drinking soda at all
This section is also a key Atlas meta-lesson: 'risk' ≠ 'magnitude of harm' ≠ 'actual exposure' ≠ 'whether you should change behavior' — four concepts routinely confused in science communication.
The IARC classification system:
Group 1 (carcinogenic): tobacco, alcohol, processed meat, solar UV, asbestos, formaldehyde (sufficient human evidence)Group 2A (probably carcinogenic): red meat, night-shift work, acrylamide, very hot beverages above 65°C (limited human + sufficient animal evidence)Group 2B (possibly carcinogenic): aspartame, gasoline, aloe extract (limited human + limited animal evidence)Group 3 (not classifiable): most chemicals — including coffee itselfThe former Group 4 (probably not carcinogenic): abolished in the 2019 revision of the Monographs preamble; its single entry (caprolactam) was reassigned to Group 3
Those two lines deserve a second look, because they are routinely stated backwards: in 2016 the same IARC evaluation moved coffee down from 2B to Group 3 (no adequate evidence that coffee itself is carcinogenic) and classified drinking beverages above 65°C as Group 2A. The problem is the temperature, not the coffee — and that hazard sits one rung above aspartame. Using 'coffee causes cancer' as the worked example of Group 2B teaches this lesson backwards.
What IARC evaluates / doesn't evaluate:
IARC assesses **whether there is *any* carcinogenic potential, not the actual risk at typical doses**This is commonly confused: 'Group 2B' ≠ 'this dose is dangerous'Risk assessment is done by JECFA (FAO/WHO Joint Expert Committee on Food Additives)
The parallel JECFA 2023 evaluation:
Reaffirmed aspartame ADI (acceptable daily intake) at 40 mg/kg body weightA can of diet soft drink contains 200 or 300 mg of aspartame (WHO's own figure)A 70 kg adult's ADI = 2,800 mg — so you would need more than 9-14 cans per day to cross itThe vast majority of consumers are far below the ADI
The basis for the IARC 2B classification:
Primarily three observational epidemiology studies (NutriNet-Santé, Ramazzini rat experiments, partial liver-cancer signal in a cohort)Evidence rated 'limited' — insufficient to rule out but not strongControversial: Ramazzini methodology has been criticized; the population studies have confounding (diabetics + obese populations have elevated baseline cancer risk and also higher sweetener intake)
Real-world meaning:
1-2 Diet Cokes/day: far below ADI, weak epidemiology, very small risk — but not zero5-10/day: on WHO's per-can figure this already reaches or crosses the low end of the ADI — worth reconsideringChildren / pregnant women: limited data, caution is reasonableSwitching to water / tea: safer at any level
How to read the 'Diet Coke and cancer' story:
Clickbait: 'Aspartame causes cancer' — technically wrong; IARC didn't say 'causes', it said 'possible'Truth: weak evidence + small actual risk, but not 'completely safe'Better than regular sugar? For someone who already drinks soda regularly, the Diet version is probably slightly better (no glucose + weight); but the best choice is not drinking soda at all
This section is also a key Atlas meta-lesson: 'risk' ≠ 'magnitude of harm' ≠ 'actual exposure' ≠ 'whether you should change behavior' — four concepts routinely confused in science communication.
IARC 在给什么排序 · 谁管剂量
2023 年 7 月, WHO 国际癌症研究机构 (IARC) 把阿斯巴甜列为 2B 类可能致癌物——这条新闻被广泛误读, 真相需要拆开看:IARC 分级系统:
1 类 (致癌): 烟草、酒精、加工肉、太阳辐射、石棉、福尔马林 (人类证据充分)2A 类 (很可能致癌): 红肉、夜班、丙烯酰胺、> 65°C 的极热饮品 (有限人证据 + 充分动物证据)2B 类 (可能致癌): 阿斯巴甜、汽油、芦荟提取物 (有限人证据 + 有限动物证据)3 类 (无法分类): 大多数化学品, 也包括咖啡本身原 4 类 (可能不致癌): 2019 年修订评估手册时已经取消这一档, 当年唯一那一项 (己内酰胺) 改列 3 类
上面那两行值得单独看一眼, 因为它常被说反: 2016 年 IARC 同一次评估干了两件事 —— 把咖啡从 2B 降到 3 类 (找不到咖啡本身致癌的充分证据), 同时把喝 > 65°C 的极热饮品评为 2A。也就是说, 出问题的是温度, 不是咖啡; 而且极热饮品这一档比阿斯巴甜还高一级。用咖啡致癌当 2B 的例子, 会把这一整堂课教反。
IARC 评估什么、不评估什么:
IARC 评估是否有致癌可能, 不评估实际剂量下风险大小这点常被混淆: 2B 类 ≠ 这剂量危险风险评估由 JECFA (FAO/WHO 联合食品添加剂专家委员会) 做
JECFA 2023 同步评估:
重申阿斯巴甜 ADI (每日可接受摄入量) 40 mg/kg 体重一罐无糖汽水含 200 或 300 mg 阿斯巴甜 (WHO 自己给的口径)70 kg 成人 ADI = 2800 mg, 换算下来要每天 9-14 罐才越过这条线绝大多数人摄入远低于 ADI
这张表排的是我们有多确定, 不是有多危险
同一档里躺着加工肉和石棉, 并不意味着吃火腿等于吸石棉。分级排的是证据的确定性: 越往上, 说明这东西能致癌这句话越难被推翻; 它完全没有说在你实际吃到的量上会发生什么。
把这两件事分给两个机构去做是刻意的分工: 一个负责问有没有信号, 另一个负责问在真实摄入量上信号有多大。读新闻时把它们混起来, 就会得到一个既恐慌又无从行动的结论 —— 你既不知道该改什么, 也不知道改到哪算够。
iarc-2016-coffee
证据为什么只算有限 · 混杂藏在哪里
IARC 2B 的依据:
主要基于 3 项观察性流行病学研究 (NutriNet-Santé / Ramazzini 大鼠实验、部分人群肝癌信号)证据等级被评为有限 —— 不充分但不能排除争议很大: Ramazzini 实验有方法学批评, 人群研究有混杂 (糖尿病 + 肥胖人群本来癌症风险高, 同时也是甜味剂高摄入人群)
现实意义:
每天 1-2 罐 Diet Coke: 远低于 ADI, 流行病学证据弱, 风险非常小——但不是 0每天 5-10 罐: 按 WHO 的每罐口径, 这已经贴上甚至越过 ADI 下界了, 应该重新考虑儿童、孕妇: 数据少, 谨慎是合理的替换为水、茶: 任何水平上都是更安全选择
Diet Coke 与癌症 解读:
标题党: 阿斯巴甜致癌——技术上不对, IARC 没说致癌, 说可能真相: 证据弱 + 实际风险小, 但不是完全安全比正常糖更好? —— 对已经习惯喝软饮的人, Diet 版本可能略好 (避免血糖 + 体重); 但最佳选择是不喝软饮
这一段是 atlas 上一个重要的元教学: 风险 ≠ 危害大小 ≠ 实际暴露 ≠ 该不该改变行为 —— 这 4 个概念在科学传播中常被混为一谈。
混杂的方向, 值得慢慢看一遍
这里的因果箭头很容易被读反。不是甜味剂把人变成了糖尿病或者肥胖, 而是已经血糖高、已经体重高的人才会改喝无糖饮料 —— 而这批人本来的癌症风险就高于一般人群。于是喝无糖饮料的人癌症多一点这条关联里, 有一部分其实是先有病、才有这瓶饮料。
要把这一部分拆掉, 需要的是随机分组: 把人随机分进喝与不喝两组, 让两组在其它方面平均起来一样。但终点是几十年后的癌症, 这种试验几乎没人做得起、也等不到。所以这一档证据大概率会长期停在有限上 —— 不是因为科学家偷懒, 而是这个问题的形状不允许更硬的证据。
知道这一点, 你就能自己判断下一条同类新闻: 先问它是关联还是随机分组, 再问它的暴露量和你的日常差多远。
PKU + pregnancy
Aspartame's clear contraindications and concerns.1. Phenylketonuria (PKU):
PKU: autosomal-recessive deficiency of phenylalanine hydroxylase (PAH); newborn-screening positivity ~1/10,000 globallyPatients cannot metabolize phenylalanine → neurotoxicity → intellectual development is impaired (severe but preventable with early dietary management)Aspartame is 50% phenylalanine → PKU patients must avoid itAll aspartame-containing products are legally required to display 'contains phenylalanine'PKU is not 'rare bad luck' — it's part of standard neonatal screening in China and globally; people who have it usually know
2. Pregnancy:
FDA + EFSA + WHO consensus: safe within ADIBut newer studies raise concerns:Azad 2016 *JAMA Pediatr* (CHILD cohort, N = 3,033): maternal sweetener use during pregnancy → ↑ infant BMI at age 1Plows 2022 + 2024 epidemiology: signals of higher offspring metabolic-syndrome riskConsensus position: 'don't ban, but reduce'; favor water + unsweetened drinks
3. Depression / mood disorders:
Some observational studies (Strawbridge 2017) link aspartame with depressive symptoms (possible central phenylalanine → tyrosine → dopamine pathway disturbance)Evidence is weak, insufficient to ban, but patients with severe depression can try a discontinuation trial and watch for effect
4. Migraine — this one was tested double-blind and failed to reproduce:
Schiffman 1987 *NEJM* recruited exactly the people who reported repeated headaches after consuming aspartame (n = 40) and challenged them double-blind, crossover: headache incidence after aspartame was 35% versus 45% after placebo — not significantly different. The authors concluded that in this population aspartame is no more likely to produce headache than placeboThat doesn't mean the headaches aren't real; it means this molecule is probably not what triggers them. The usual triggers — sleep, menstrual cycle, dehydration, a skipped meal, a sudden drop in stress — routinely coincide with reaching for a diet drink, which is how the drink gets the blameA headache diary can still log aspartame, but log it to read your own data, not to confirm a claim that has already been tested and failed
A special note on erythritol:
The '0 cal + 0 GI' star sweetener, long considered safe**Hazen 2023 *Nat Med* (Cleveland Clinic, N > 4,000 + mouse experiments): highest plasma-erythritol quartile vs lowest → MACE risk ~ 2× elevated** (HR 1.80 and HR 2.21 in the two validation cohorts), linked to enhanced platelet aggregationThis is an observational study; causality isn't established, but the signal is strong enough that FDA is reassessingPractical: erythritol is still legal, but 'safe + 0 cal + 0 GI = best choice' as a default judgment has been shaken
Sugar-alcohol diarrhea (sorbitol / maltitol / isomalt):
Non-fermenting erythritol is fine, but other sugar alcohols are fermented by gut bacteria → osmotic diarrhea + bloatingIndividual variation is large: 10-50 g can already trigger symptoms'Sugar-free' chocolates / candies often use maltitol → 100 g can cause diarrhea where the same dose of sucrose wouldn't
azad-2016-jama-peds-sweetenersschiffman-1987-nejm-aspartame-headache
Chapter 3
Gut microbiome + glucose
Gut microbiome + glucose
Artificial sweeteners + gut microbiome + blood glucose — the most disruptive research line of the past decade.
**Suez 2014 *Nature*** (Weizmann Institute, Israel):
Mouse experiments: feeding sucralose + aspartame + saccharin → significant gut-microbiota changes → impaired glucose toleranceHuman pilot (N = 7): 7 days of saccharin → 4/7 subjects had worsening glucose tolerance + parallel microbiome shiftsCritical evidence: transplanting the microbiota of the 'responder' subjects into germ-free mice → the mice also developed glucose intolerance → the microbiome was the mediator
**Suez 2022 *Cell*** follow-up (N = 120 healthy adults, RCT):
6 arms of 20 people each: four sweetener arms (saccharin / sucralose / aspartame / stevia) + two controls (glucose vehicle, no supplement), 2 weeks, all doses below the ADISaccharin + sucralose: significantly perturbed the microbiome + altered glucose response in some subjectsAspartame + stevia: smaller microbiome changesHuge individual variation: same dose, completely different metabolic response between people
Mechanistic hypothesis:
Non-metabolized sweeteners come into direct contact with gut bacteriaSome taxa (Clostridiales / Bacteroidales) utilize or are affected → dysbiosis → changes in short-chain fatty acid + bile-acid metabolism → altered insulin sensitivity / glucose responseSucralose contains chlorine → may disrupt the microbiome more strongly
'0 GI' ≠ 'glucose-neutral':
Classical nutrition: 'no sugar → no blood-glucose effect'The Suez series shows 'indirect impact via the microbiome''Glucose-metabolism-neutral' sweeteners may not be as neutral as assumed
Clinical translation:
WHO 2023 guideline: 'Non-sugar sweeteners are NOT recommended for weight control or for reducing non-communicable disease risk' (based on a systematic review of 283 studies)This is a major policy shift — the 'Diet soda = healthy substitute' framing from 20 years ago is no longer WHO's positionThree qualifiers travel with that sentence and are usually dropped: WHO grades it a conditional recommendation, not a strong one; it applies to everyone except people with pre-existing diabetes; and it explicitly excludes sugar alcohols (polyols), so erythritol and xylitol are outside its scopeAnd WHO didn't ban anything — just 'not recommended'; actual regulation remains national
How do they compare to sugar?
Sugar (sucrose / HFCS) at high intake: clearly causes weight gain, T2D, fatty liver, CVDArtificial sweeteners at high intake: may disrupt microbiome / glucose metabolism / appetite regulationNeither is good, but the evidence for sugar is stronger and the harm largerBest practice: the replacement target is water / tea / black coffee, not 'Diet replacing Regular'
Real-food contrast:
Whole fruit contains natural sugar, but also fiber + polyphenols + vitamins — net healthyHoney + maple syrup + other 'natural' sugars are still sugar — no significant advantage'Natural vs artificial' is the wrong axis; 'whole food vs industrial product' is the real axis
**Suez 2014 *Nature*** (Weizmann Institute, Israel):
Mouse experiments: feeding sucralose + aspartame + saccharin → significant gut-microbiota changes → impaired glucose toleranceHuman pilot (N = 7): 7 days of saccharin → 4/7 subjects had worsening glucose tolerance + parallel microbiome shiftsCritical evidence: transplanting the microbiota of the 'responder' subjects into germ-free mice → the mice also developed glucose intolerance → the microbiome was the mediator
**Suez 2022 *Cell*** follow-up (N = 120 healthy adults, RCT):
6 arms of 20 people each: four sweetener arms (saccharin / sucralose / aspartame / stevia) + two controls (glucose vehicle, no supplement), 2 weeks, all doses below the ADISaccharin + sucralose: significantly perturbed the microbiome + altered glucose response in some subjectsAspartame + stevia: smaller microbiome changesHuge individual variation: same dose, completely different metabolic response between people
Mechanistic hypothesis:
Non-metabolized sweeteners come into direct contact with gut bacteriaSome taxa (Clostridiales / Bacteroidales) utilize or are affected → dysbiosis → changes in short-chain fatty acid + bile-acid metabolism → altered insulin sensitivity / glucose responseSucralose contains chlorine → may disrupt the microbiome more strongly
'0 GI' ≠ 'glucose-neutral':
Classical nutrition: 'no sugar → no blood-glucose effect'The Suez series shows 'indirect impact via the microbiome''Glucose-metabolism-neutral' sweeteners may not be as neutral as assumed
Clinical translation:
WHO 2023 guideline: 'Non-sugar sweeteners are NOT recommended for weight control or for reducing non-communicable disease risk' (based on a systematic review of 283 studies)This is a major policy shift — the 'Diet soda = healthy substitute' framing from 20 years ago is no longer WHO's positionThree qualifiers travel with that sentence and are usually dropped: WHO grades it a conditional recommendation, not a strong one; it applies to everyone except people with pre-existing diabetes; and it explicitly excludes sugar alcohols (polyols), so erythritol and xylitol are outside its scopeAnd WHO didn't ban anything — just 'not recommended'; actual regulation remains national
How do they compare to sugar?
Sugar (sucrose / HFCS) at high intake: clearly causes weight gain, T2D, fatty liver, CVDArtificial sweeteners at high intake: may disrupt microbiome / glucose metabolism / appetite regulationNeither is good, but the evidence for sugar is stronger and the harm largerBest practice: the replacement target is water / tea / black coffee, not 'Diet replacing Regular'
Real-food contrast:
Whole fruit contains natural sugar, but also fiber + polyphenols + vitamins — net healthyHoney + maple syrup + other 'natural' sugars are still sugar — no significant advantage'Natural vs artificial' is the wrong axis; 'whole food vs industrial product' is the real axis
两组研究怎么做的 · 无菌鼠那一步为什么关键
人工甜味剂 + 肠菌 + 血糖 —— 过去 10 年最颠覆的研究路线:Suez 2014 Nature (Weizmann Institute, 以色列):
小鼠实验: 喂三氯蔗糖 + 阿斯巴甜 + 糖精 → 肠道菌群组成显著改变 → 葡萄糖耐量受损人体小试验 (N = 7): 7 天糖精 → 4/7 受试者葡萄糖耐量恶化, 同时菌群组成变化关键证据: 把变坏 受试者的菌群移植到无菌鼠 → 鼠也变葡萄糖耐量差 → 直接证明菌群是中介
Suez 2022 Cell 续作 (N = 120 健康成人 RCT):
一共 6 个臂, 每臂 20 人: 4 个甜味剂臂 (糖精、三氯蔗糖、阿斯巴甜、甜菊糖) + 2 个对照臂 (葡萄糖载体对照 / 不补充对照), 补 2 周, 剂量都在 ADI 以内糖精 + 三氯蔗糖: 显著扰乱菌群 + 部分人血糖反应改变阿斯巴甜 + 甜菊糖: 菌群变化较小个体差异巨大: 同样剂量, 不同人代谢响应完全不同
机制猜想:
不被人体代谢的甜味剂直接接触肠菌某些菌 (Clostridiales / Bacteroidales) 利用、受影响 → 菌群失调 → 短链脂肪酸 + 胆汁酸代谢改变 → 胰岛素敏感性、血糖响应变化三氯蔗糖含氯 → 可能干扰菌群更显著
0 GI ≠ 对血糖中性:
经典营养学认为不含糖 → 不影响血糖Suez 系列研究表明通过菌群间接影响血糖葡萄糖代谢中立 的甜味剂可能没有想象的中立
临床转化:
WHO 2023 指南: 不推荐非糖甜味剂作为体重控制或减少非传染性疾病风险的工具 (基于 283 项研究系统综述)这是重大政策转变——20 年前的Diet 软饮 = 健康替代 已不是 WHO 立场但要把这条建议读准, 有三个限定词不能丢: ① 它在 WHO 自己的分级里是有条件建议 (conditional), 不是强推荐; ② 适用范围是所有人, 已确诊糖尿病的人除外; ③ 它明确不覆盖糖醇 (polyols) —— 赤藓糖醇、木糖醇不在这份指南的射程里, 拿它去评价它们是越界而且 WHO没禁, 只是不推荐, 实际监管由各国
为什么把菌群搬进无菌鼠那一步最关键
只观察到血糖变差的人菌群也怪, 有两种同样说得通的解释: 菌群把血糖弄差了, 或者血糖差反过来把菌群弄怪了。光看人身上的数据, 这两种解释长得一模一样。
无菌鼠把它们分开了。这种鼠从出生起就没接触过细菌, 血糖本来正常。研究者只搬走一样东西 —— 菌群 —— 别的什么都没给它。如果症状跟着菌群一起搬了家, 箭头就只能是从菌群指向血糖。这是关联和因果之间少见的一座真桥, 也是这条研究线值得认真对待的原因。
至于为什么三氯蔗糖的动静格外大
它的骨架本来就是蔗糖, 只是有几个位置被换成了氯原子。骨架还在, 所以肠菌多少还认得它、还想去处理它; 而那几个氯又让人的酶下不了刀, 于是它整整齐齐地被送到大肠、交给细菌。一个既像食物、又不属于你的分子天天出现在同一批住客面前 —— 这就是上面那条含氯可能干扰更显著的猜想的直观版本。
和糖比谁更糟 · 真正的那根轴
和糖比谁更糟糕?
糖 (蔗糖 / HFCS) 在大量摄入时: 明确导致 体重增 / T2D / 脂肪肝、心血管病人工甜味剂 在大量摄入时: 可能干扰 菌群、葡萄糖代谢、食欲调节两个都不是好的, 但糖的证据更强 + 危害更大最佳实操: 替换的目标是水、茶 / 黑咖啡, 不是用 Diet 替换 Regular
真食物的对比:
整水果含天然糖, 但有纤维 + 多酚 + 维生素——净健康蜂蜜 + 枫糖等天然糖仍是糖, 没有显著优势天然 vs 人工 是错误的轴, 整食物 vs 工业制品 才是真正的轴
整食物对工业制品, 为什么才是真正的那根轴
一个苹果里的糖和一罐饮料里的糖是同一种分子, 但它们到达小肠的速度完全不同。苹果的糖被锁在细胞壁和纤维搭起来的架子里, 你得先咬碎, 再让消化酶一点一点把它放出来; 饮料里的糖已经溶在水里, 一进胃就随时可以过闸。同样多的糖, 一个是细水长流, 一个是开闸放水, 你的胰岛素要应付的是两条完全不同的曲线。
天然和人工这根轴分不出这两者的差别 —— 蜂蜜是天然的, 但它站在开闸的那一边。真正分得开的那根轴, 是这口东西还带不带着它原来的结构。
'Sweetness in the brain'
A second non-metabolic pathway for artificial sweeteners: the brain's reward system.'Sweet anticipation → metabolic response' (the cephalic phase):
Before sugar even arrives, seeing / smelling / thinking of it → the brain triggers insulin + GI-hormone releaseThis is 'pre-digestion priming', established for ~50 yearsDrinking a Diet beverage → triggers the same anticipation, but no sugar actually arrives → 'inverse anticipation' neuroadaptation
Long-term hypothesis (Yang 2010, Mattes 2009 review):
Repeated sweetness without sugar → the reward circuit blunts → 'real sugar no longer satisfies'Simultaneously triggers sugar craving: Mattes 2009 shows Diet-soda drinkers don't end up with lower total intake — they compensate at other mealsCounter-evidence: pooling the randomised trials, sugar-for-sweetener swaps do produce a short-term reduction — but a much smaller one than usually quoted. Miller 2014 *AJCN* pools to −0.80 kg (95% CI −1.17 to −0.43); Toews 2019 *BMJ* reports a BMI difference of −0.6 (95% CI −1.19 to −0.01 — that is BMI, not kilograms) resting on just 2 small studies in 174 people, with certainty rated low to very lowSo the counter-evidence is itself soft: it says the direction may be right, not how many kilosReconciliation: short-term effective (calorie cut), long-term possibly compensated (neuroadaptation) → weight rebounds
Special concern for children:
The 6-18 taste-preference window strongly shapes lifelong food preferencesRepeated sweetness → natural sweetness 'doesn't taste sweet enough' → rejection of fruit / vegetables'Kids' no-sugar drinks' look healthy but reinforce the mental model that 'all drinks should be sweet'WHO + AAP recommend: water + milk for children — don't anchor habits to sweet beverages
The 'sugar addiction' analogy:
Sugar genuinely activates dopamine reward pathways; some animal experiments show opioid-receptor-like addictive featuresBut human 'sugar addiction' is behavioral + neuroadaptive, not physical dependenceDiet replacement doesn't solve sugar addiction — the core problem is 'expecting sweetness', which Diet also fulfillsReal solution = lowering the sweetness threshold — for 2-4 weeks reduce all sweet input (sugar + sweeteners), let taste buds reset → fruit will taste sweeter afterward
Operational psychology:
'Diet replacing Regular' is an effective short-term calorie-cutting tool, not a long-term solutionLong-term goal: lower the overall sweetness expectation of beverages + snacksThis is a gradual process, not overnight cold-turkey — body + brain both need time to adapt
yang-2010-sweetener-brainmattes-2009-sweeteners-energymiller-2014-ajcn-lcs-weighttoews-2019-bmj-nss-review
Chapter 4
Erythritol · Hazen 2023
Erythritol · Hazen 2023
Erythritol was long considered the 'perfect sweetener' — until a 2023 Cleveland Clinic study changed that assessment.
Where the 'perfect sweetener' reputation came from:
0.24 kcal/g (essentially calorie-free)Zero blood-glucose response (doesn't enter glucose metabolism)Non-fermenting (low FODMAP, no diarrhea except at massive doses)Anti-cariogenic (oral bacteria can't metabolize it)Naturally present (in fruit + fermentation)~70% sucrose sweetness — the closest taste profile to sugar
**Hazen 2023 *Nat Med*** (Cleveland Clinic, Stan Hazen's group):
N = 4,000+ cardiovascular-high-risk patients, prospective + mouse + in vitro validation:
1. Highest blood-erythritol quartile vs lowest: 3-year MACE (MI + stroke + death) risk ~ 2× elevated — measured separately in two independent validation cohorts: US (n = 2,149) HR 1.80 (95% CI 1.18-2.77) and European (n = 833) HR 2.21 (95% CI 1.20-4.07)
2. Mechanism: in vitro, rising erythritol concentration → enhanced platelet aggregation → ↑ thrombotic risk
3. Mouse experiments: doses equivalent to a single human erythritol drink → shortened thrombus formation time
4. Healthy adults given 30 g erythritol (a typical single-beverage dose): plasma levels rose > 1000×, stayed elevated ≥ 2 days
Why this is significant:
The '0 cal + 0 GI + natural = safe' core rationale has been seriously shakenFDA + EU + Chinese food regulators are re-examiningMany 'keto' / 'diabetic-friendly' / 'weight-loss' products + GenkiForest + LMNT + most '0 sugar' sodas use erythritol
Limitations of the Hazen study (stated honestly):
Observational + correlational: cardiovascular-high-risk patients have baseline elevated MACE and may also consume more '0-sugar' products (confounding)No RCT: there's no controlled 'give healthy people erythritol vs placebo and watch hard endpoints' trialMechanism mouse studies + the 1000× plasma rise in healthy volunteers weaken the 'just correlation' defense — dose-response + biological plausibility are credibleReplication studies are underway, with more data expected 2024-2025
Clinical operating stance:
History of cardiovascular disease / high thrombotic risk (atrial fibrillation, prior stroke, anticoagulation): reducing erythritol exposure is reasonableHealthy adults: moderate intake (once or twice daily) probably carries small risk, but you can no longer claim it is completely safeAlternatives: allulose / monk fruit / stevia / very small amounts of real sucrose with whole foods'No sweeteners at all' remains the optimum — water + tea + black coffee
Similar concern for xylitol:
Witkowski 2024 *Eur Heart J* (Hazen group sequel): a xylitol-MACE association is also presentXylitol is used in diabetic foods + anti-cariogenic gum + toothpasteXylitol is acutely lethal to dogs (massive insulin release in dogs → severe hypoglycemia + acute liver failure) — dog owners should absolutely never let their dog access any xylitol-containing product
Bottom line:
The 30-year consensus that 'sugar alcohols = safe' is being revised by the scienceNo need to panic — the absolute risk is still small, the signal is realBut for anyone who treated '0-cal erythritol as a harmless substitute', that view no longer holds
Where the 'perfect sweetener' reputation came from:
0.24 kcal/g (essentially calorie-free)Zero blood-glucose response (doesn't enter glucose metabolism)Non-fermenting (low FODMAP, no diarrhea except at massive doses)Anti-cariogenic (oral bacteria can't metabolize it)Naturally present (in fruit + fermentation)~70% sucrose sweetness — the closest taste profile to sugar
**Hazen 2023 *Nat Med*** (Cleveland Clinic, Stan Hazen's group):
N = 4,000+ cardiovascular-high-risk patients, prospective + mouse + in vitro validation:
1. Highest blood-erythritol quartile vs lowest: 3-year MACE (MI + stroke + death) risk ~ 2× elevated — measured separately in two independent validation cohorts: US (n = 2,149) HR 1.80 (95% CI 1.18-2.77) and European (n = 833) HR 2.21 (95% CI 1.20-4.07)
2. Mechanism: in vitro, rising erythritol concentration → enhanced platelet aggregation → ↑ thrombotic risk
3. Mouse experiments: doses equivalent to a single human erythritol drink → shortened thrombus formation time
4. Healthy adults given 30 g erythritol (a typical single-beverage dose): plasma levels rose > 1000×, stayed elevated ≥ 2 days
Why this is significant:
The '0 cal + 0 GI + natural = safe' core rationale has been seriously shakenFDA + EU + Chinese food regulators are re-examiningMany 'keto' / 'diabetic-friendly' / 'weight-loss' products + GenkiForest + LMNT + most '0 sugar' sodas use erythritol
Limitations of the Hazen study (stated honestly):
Observational + correlational: cardiovascular-high-risk patients have baseline elevated MACE and may also consume more '0-sugar' products (confounding)No RCT: there's no controlled 'give healthy people erythritol vs placebo and watch hard endpoints' trialMechanism mouse studies + the 1000× plasma rise in healthy volunteers weaken the 'just correlation' defense — dose-response + biological plausibility are credibleReplication studies are underway, with more data expected 2024-2025
Clinical operating stance:
History of cardiovascular disease / high thrombotic risk (atrial fibrillation, prior stroke, anticoagulation): reducing erythritol exposure is reasonableHealthy adults: moderate intake (once or twice daily) probably carries small risk, but you can no longer claim it is completely safeAlternatives: allulose / monk fruit / stevia / very small amounts of real sucrose with whole foods'No sweeteners at all' remains the optimum — water + tea + black coffee
Similar concern for xylitol:
Witkowski 2024 *Eur Heart J* (Hazen group sequel): a xylitol-MACE association is also presentXylitol is used in diabetic foods + anti-cariogenic gum + toothpasteXylitol is acutely lethal to dogs (massive insulin release in dogs → severe hypoglycemia + acute liver failure) — dog owners should absolutely never let their dog access any xylitol-containing product
Bottom line:
The 30-year consensus that 'sugar alcohols = safe' is being revised by the scienceNo need to panic — the absolute risk is still small, the signal is realBut for anyone who treated '0-cal erythritol as a harmless substitute', that view no longer holds
它凭什么被叫做完美甜味剂 · 那组数字
赤藓糖醇 (Erythritol) 长期被认为是完美甜味剂——但 2023 年 Cleveland Clinic 的研究改变了这个判断:完美甜味剂 的赞誉来源:
0.24 kcal/g (几乎无热量)零血糖反应 (不进葡萄糖代谢)不发酵 (低 FODMAP, 不导致腹泻——除非极大量)抗龋齿 (不被口腔菌代谢)天然来源 (水果 + 发酵生成)甜度 ~ 70% 蔗糖, 口感最接近糖
Hazen 2023 Nat Med (Cleveland Clinic, Stan Hazen 团队):
N = 4,000+ 心血管病高风险患者, 前瞻性 + 鼠 + 体外验证:
1. 血浆赤藓糖醇水平最高四分位 vs 最低: 3 年内 MACE (心梗 + 中风 + 死亡) 风险 ↑ ~ 2 倍 —— 两个独立验证队列各算一次: 美国队列 (n = 2,149) HR 1.80 (95% CI 1.18-2.77), 欧洲队列 (n = 833) HR 2.21 (95% CI 1.20-4.07)
2. 机制: 体外实验显示赤藓糖醇浓度增 → 血小板凝集增强 → 血栓风险增
3. 鼠实验: 喂赤藓糖醇剂量等同人喝一杯赤藓糖醇饮料 → 血栓形成时间缩短
4. 健康人口服 30 g 赤藓糖醇 (典型一杯饮料剂量): 血浆水平上升 > 1000 倍, 持续 ≥ 2 天
为什么这是大事:
0 卡 + 0 GI + 天然 = 安全这条核心理由被严重动摇FDA + 欧盟 + 中国食品监管已重新审视大量生酮 / 糖尿病专用 / 减重 食品 + 元气森林 + LMNT + 大部分0 糖 软饮都用赤藓糖醇
为什么血浆浓度那一条比风险比值更有说服力
单看关联, 总能反问一句: 血里赤藓糖醇高的人, 会不会本来就病得更重、吃的无糖加工食品也更多? 这个反问很难被证伪。
但健康人喝下普通一杯的剂量后血浓度直接冲上去、而且好几天下不来, 说的是另一件事: 你日常吃到的那个量, 确实足以把血里的浓度推进研究里出事的那个区间。剂量和暴露对上了 —— 这一步把只是关联的辩护变难了很多, 因为它把血浓度高的人是特殊人群换成了普通人喝一杯就会到那个浓度。
再加上体外和动物那两层, 三段是接得上的: 浓度升高, 血小板更容易聚集, 血栓形成得更快。缺的那一环是把健康人随机分进喝与不喝两组、追踪硬终点 —— 这一环目前没人做。
证据强在哪弱在哪 · 谁该现在减量
Hazen 研究的局限性 (诚实陈述):
观察性 + 关联: 心血管病高危人群本来 MACE 高, 这些人也可能更多吃0 糖 食品 (混杂)没 RCT: 没有给健康人喝赤藓糖醇 vs 安慰剂 → 看终点 的对照试验机制鼠实验 + 健康人血浆 1000× 上升这两点削弱了只是关联 的辩护——剂量-效应 + 机制可信复制研究进行中, 2024-2025 应有更多数据
临床实操立场:
心血管病史、高血栓风险 (房颤、既往中风、抗凝治疗): 减少赤藓糖醇暴露是合理的健康人: 适量 (一天一两次) 风险可能小, 但不能再宣称完全安全替代: 阿洛酮糖、罗汉果、甜菊糖、极少量蔗糖 + 整食物不用任何甜味剂 仍是最优选——水 + 茶 + 黑咖啡
木糖醇 (Xylitol) 类似担忧:
Witkowski 2024 Eur Heart J (Hazen 同团队后续): 木糖醇与 MACE 关联也存在木糖醇用于糖尿病食品 + 抗龋齿口香糖 + 牙膏狗摄入木糖醇是急性致命的 (狗胰岛素强烈分泌 → 严重低血糖 + 急性肝衰竭) —— 家有狗者绝对不要让狗碰任何含木糖醇的食品
底线:
糖醇 = 安全 这个 30 年的共识正在被科学修正不需要恐慌——风险信号是真的, 但绝对风险仍小但对认为0 卡赤藓糖醇是无害替代的人, 这一观点已经不成立
木糖醇为什么对狗是急诊, 对人不是
这不是狗比人脆弱这么笼统的事, 而是一个具体的识别错误。木糖醇进了狗的血以后, 狗的胰腺会把它当成需要立刻处理的糖信号, 于是猛放胰岛素; 人的胰腺不会对它有这个反应。胰岛素一多, 血里的葡萄糖被大批赶进细胞, 血糖在很短时间里被拖到危险的低位, 再叠加急性肝衰竭。
所以一小块无糖口香糖掉在地上被狗捡走, 是真正的急诊, 不是吃一点没事。家里有狗, 木糖醇牙膏和口香糖要收进狗够不到的地方。
'Natural' ≠ 'safe' meta-lesson
The erythritol case + the whole sweetener discussion → a core Atlas meta-lesson:'Natural' ≠ 'safe', 'artificial' ≠ 'dangerous'.
Common manifestations of the 'natural' fallacy:
'Honey is natural sugar, better than white sugar' — honey and white sugar have nearly identical calories and metabolic responses'Brown sugar has minerals so it's better' — the mineral amount is trivial; not a nutritional benefit'Maple syrup / coconut sugar / monk-fruit sugar are natural' — all are sugar; harmful at high intake'Stevia / monk fruit are natural' — they're sweeteners, not sugars; but 'natural' doesn't prove safe
Real dangers from the 'natural' camp:
Cyanogenic glycosides: bitter almonds + raw cassava + ginkgoPlant toxins: mushrooms + certain leaves + certain seedsNatural hormones: soy isoflavones + red yeast rice (statin-like) + some herbsNatural carcinogens: safrole (sassafras root beer) + aflatoxins (natural mold contamination of grains) + aristolochic acid
Real advantages of 'artificial':
Precise dose control: pharmaceuticals + nutrient fortification (iodized salt, folic-acid-fortified flour)Reproducible + regulatableContaminant removal: modern food safety greatly reduced exposure to mold, parasites, and heavy metals in 'natural foods'
The real axes:
'Whole food vs industrial product' — regardless of natural or artificial'Level of evidence' — which class has more evidence / clinical endpoints / long-term follow-up'Dose + accumulation' — every 'poison' is a matter of dose
Applied back to sweeteners:
Stevia (natural) vs sucralose (artificial): don't pick stevia just because it's 'natural' — look at evidence + individual response + priceHoney (natural) vs erythritol (natural but processed): both are 'natural', yet metabolic response + clinical evidence differ completelyWhole fruit vs any sweetener: whole fruit is the default best, because it's not just a sweetness vehicle
This is a concrete embodiment of Atlas's teaching philosophy: don't decide by label — check what's inside + what it does in the body + what the evidence shows.
Chapter 5
Decision tree
Decision tree
Practical decision tree — who to use and who to skip:
Q1: Can you avoid sweetened drinks entirely?
Yes (best): water, unsweetened tea, black coffee, sparkling water, homemade lemon waterNo (realistic): use sweeteners to reduce the harm of full-sugar sodas (Diet > Regular)This is the primary decision — not 'which sweetener' but 'do I need anything sweet at all'
Q2: If using one, which?
Best options (balancing evidence + individual response):
Whole fruit + a small amount of real sugar in home cooking: don't add sugar to drinksMonk fruit (mogroside V): traditional Chinese; few side-effect signals; OK for home dessertsStevia (high-purity rebaudioside A): short-term safety evidence strong, long-term data limited; some report a 'bitter aftertaste'Allulose: genuinely metabolically neutral, but expensive + uncommon
Conditional use (OK short-term, taper long-term):
Aspartame (Diet sodas): within ADI, far below typical intake; absolutely contraindicated in PKUSucralose: same caveats; the open question is sucralose-6-acetate (manufacturing residue + gut-bacterial conversion, genotoxic in vitro, Schiffman 2023), not bakingAcesulfame-K: usually used in blends; no clear standalone problem
Use with caution (new evidence):
Erythritol: the Hazen 2023 signal isn't ruled out — cardiovascular patients / high-thrombotic-risk individuals should reduce intakeXylitol: Witkowski 2024 points the same direction — dog owners must absolutely keep it away from dogsOther sugar alcohols (sorbitol / maltitol): diarrhea risk
Avoid:
'0 sugar' formulations blending 'HFCS + maltodextrin + sweetener': these are almost all UPF, not healthy substitutesChildren regularly drinking any '0 sugar' / 'Diet' beverage: taste-preference shaping + sparse dataHigh intake during pregnancy: within ADI is safe, but the Azad 2016 infant-BMI signal → caution
Q3: What's your 'cut sugar' goal?
Goal A: Reduce calories / lose weight: short-term sweeteners may help; long-term needs full dietary structureGoal B: Control blood glucose (T2D / prediabetes): sweeteners + whole foods + exercise + medication togetherGoal C: Reduce dental caries: xylitol gum + toothpaste (keep away from dogs)Goal D: Reduce UPF: don't replace 'sugar-sweetened UPF' with 'sweetener-sweetened UPF' — the point is replacing with whole foods
Q4: A 'reset the sweetness threshold' plan:
If you want to fundamentally change your reliance on sweetness4-week plan: drastically reduce all sweet input (sugar + sweeteners both)Weeks 1-2: hard, strong cravingsWeeks 3-4: the threshold begins resetting; fruit starts to taste genuinely sweetOutcome: afterwards, small amounts of sweetness will satisfy — no need to rely on large quantities
To close the topic:
Sweeteners are neither evil nor a perfect substitute. Swapping full-sugar soda for Diet has a marginal benefit at the 'less direct sugar harm' level, but the root problem — 'needing sweetness' — is untouched. What actually solves the root is swapping sweetened drinks for water / tea.
Direction matters more than intensity: head toward 'less sweetness dependency' — no need to force overnight zero.
Q1: Can you avoid sweetened drinks entirely?
Yes (best): water, unsweetened tea, black coffee, sparkling water, homemade lemon waterNo (realistic): use sweeteners to reduce the harm of full-sugar sodas (Diet > Regular)This is the primary decision — not 'which sweetener' but 'do I need anything sweet at all'
Q2: If using one, which?
Best options (balancing evidence + individual response):
Whole fruit + a small amount of real sugar in home cooking: don't add sugar to drinksMonk fruit (mogroside V): traditional Chinese; few side-effect signals; OK for home dessertsStevia (high-purity rebaudioside A): short-term safety evidence strong, long-term data limited; some report a 'bitter aftertaste'Allulose: genuinely metabolically neutral, but expensive + uncommon
Conditional use (OK short-term, taper long-term):
Aspartame (Diet sodas): within ADI, far below typical intake; absolutely contraindicated in PKUSucralose: same caveats; the open question is sucralose-6-acetate (manufacturing residue + gut-bacterial conversion, genotoxic in vitro, Schiffman 2023), not bakingAcesulfame-K: usually used in blends; no clear standalone problem
Use with caution (new evidence):
Erythritol: the Hazen 2023 signal isn't ruled out — cardiovascular patients / high-thrombotic-risk individuals should reduce intakeXylitol: Witkowski 2024 points the same direction — dog owners must absolutely keep it away from dogsOther sugar alcohols (sorbitol / maltitol): diarrhea risk
Avoid:
'0 sugar' formulations blending 'HFCS + maltodextrin + sweetener': these are almost all UPF, not healthy substitutesChildren regularly drinking any '0 sugar' / 'Diet' beverage: taste-preference shaping + sparse dataHigh intake during pregnancy: within ADI is safe, but the Azad 2016 infant-BMI signal → caution
Q3: What's your 'cut sugar' goal?
Goal A: Reduce calories / lose weight: short-term sweeteners may help; long-term needs full dietary structureGoal B: Control blood glucose (T2D / prediabetes): sweeteners + whole foods + exercise + medication togetherGoal C: Reduce dental caries: xylitol gum + toothpaste (keep away from dogs)Goal D: Reduce UPF: don't replace 'sugar-sweetened UPF' with 'sweetener-sweetened UPF' — the point is replacing with whole foods
Q4: A 'reset the sweetness threshold' plan:
If you want to fundamentally change your reliance on sweetness4-week plan: drastically reduce all sweet input (sugar + sweeteners both)Weeks 1-2: hard, strong cravingsWeeks 3-4: the threshold begins resetting; fruit starts to taste genuinely sweetOutcome: afterwards, small amounts of sweetness will satisfy — no need to rely on large quantities
To close the topic:
Sweeteners are neither evil nor a perfect substitute. Swapping full-sugar soda for Diet has a marginal benefit at the 'less direct sugar harm' level, but the root problem — 'needing sweetness' — is untouched. What actually solves the root is swapping sweetened drinks for water / tea.
Direction matters more than intensity: head toward 'less sweetness dependency' — no need to force overnight zero.
要不要甜 · 用谁不用谁
到底用谁不用谁 - 实操决策:Q1: 你能不能不喝甜味饮料?
能 (最优解): 水 + 茶 (无糖) + 黑咖啡 + 苏打水 + 自制柠檬水不能 (现实): 用甜味剂减少全糖软饮造成的危害 (Diet > Regular)这是首要决策, 不是用哪个甜味剂 而是到底要不要甜的
Q2: 如果用, 用哪个?
最佳选项 (证据 + 个体响应权衡):
整水果 + 极少量真糖在自做食物: 不在饮料里加糖罗汉果 (mogroside V): 中国传统, 副作用证据少, 适合家用甜点甜菊糖 (高纯 rebaudioside A): 短期安全证据强, 长期数据少, 部分人有苦后味阿洛酮糖: 真的代谢中立, 但贵 + 罕见
条件用 (短期 OK, 长期减用):
阿斯巴甜 (Diet 软饮): ADI 内, 远低于一般人摄入; PKU 绝对禁忌三氯蔗糖: 同上; 新出现的问号是三氯蔗糖-6-乙酸酯 (生产残留 + 肠菌转化, 体外基因毒性, Schiffman 2023), 不是烘焙安赛蜜: 通常和其它甜味剂混用, 单独无明显问题
谨慎用 (新证据出现):
赤藓糖醇: Hazen 2023 信号未排除——心血管病人、高血栓风险者减少木糖醇: Witkowski 2024 同方向——家有狗的人绝对不要让狗碰其它糖醇 (山梨醇、麦芽糖醇): 腹泻风险
避免:
0 糖 配方含HFCS + 麦芽糊精 + 甜味剂 混合: 几乎都是 UPF, 不是健康替代儿童经常喝任何 0 糖 / Diet 饮料: 味觉偏好塑造 + 数据少怀孕大量摄入: ADI 内安全, 但 Azad 2016 婴儿 BMI 信号 → 谨慎
你的目标是哪一个 · 把甜度阈值降回来
Q3: 戒糖 的目标是什么?
目标 A: 减卡路里、减重: 短期甜味剂可能有效, 长期需配合饮食结构目标 B: 控血糖 (糖尿病、糖尿病前期): 甜味剂 + 整食物 + 运动 + 药物综合目标 C: 减龋齿: 木糖醇口香糖 + 牙膏 (不要让狗碰)目标 D: 减 UPF: 不要把用甜味剂的 UPF 替换用糖的 UPF——核心是替换为整食物
Q4: 重置甜度阈值 计划:
想从根本改变对甜的依赖4 周计划: 大幅减少所有甜味输入 (糖 + 甜味剂都减)第 1-2 周: 困难期, 渴望强第 3-4 周: 阈值开始重置, 水果开始变得真的甜结果: 之后即使少量甜味也满足, 不需要再依赖大量摄入
阈值为什么真的能降回来
舌头和它后面那条奖赏通路是按相对量工作的, 不是按绝对量。长期泡在高甜度里, 同一颗草莓落在这条曲线上就显得平淡; 把输入整体降下来一段时间, 同一颗草莓重新变得够甜。变的不是草莓, 是你拿来比较的那把尺子。
所以戒糖靠意志力这个说法把问题看小了。真正在动的是校准: 你不是在忍住不吃甜, 你是在把尺子调回来。这也解释了为什么最难的是开头那几天, 而不是后面几周 —— 尺子还没动, 落差却已经在了。
'Sugar-free people' psychology
The psychology of 'I drink Diet, so I'm healthy':Moral licensing (Khan & Dhar 2006):
A series of experiments (Khan & Dhar 2006): after making a virtuous choice, people lean toward an indulgent one next — 'I already drank Diet, so I can have a slice of cake'Clinically: Diet-soda drinkers don't end up with lower total energy intake
Health-halo effect:
'0 sugar' label → automatic 'healthy' association → consumption risesResult: the '0 sugar + weight loss' intent is offset by higher intake
'Willpower outsourcing' mindset:
'I can drink Diet because it's a safe sugar substitute' → psychologically the 'problem is solved' → underlying diet structure is no longer examinedBut the root problem (UPF + processed-food overconsumption) hasn't moved
Reverse 'withdrawal' risk:
Suddenly stopping all sugar + sweeteners → headache / irritability / cravings (short-term, 3-7 days)This isn't a dramatic 'sugar addiction breaking out', but it is a temporary imbalance in the reward circuitHandling: gradual, not abrupt; substitute sweet inputs (whole fruit / dark chocolate / spices like cinnamon)
The 'perfectionism' trap:
'I must cut all sugar' → one slip → 'I failed' → total collapseThe best health behavior isn't perfect + intermittent; it's 80% sustained
A sustainable cognitive model:
Sweetness = a taste, not 'love' / 'reward' / 'emotion management'If you use sweet foods to self-soothe / celebrate / cope with stress, that's a behavioral problem, not a nutritional one, and nutritional intervention can't touch the coreReal solution = find reward + emotional-regulation pathways outside sweetness (exercise, social, creative work, nature, reading)
The intent of this island isn't to teach you the 'perfect' way to pick a sweetener — it's to help you see clearly what role sweetness plays in your life, and then adjust within your means.
References · 9
- World Health Organization. (2023). Use of non-sugar sweeteners: WHO guideline. Geneva: WHO. www.who.int/publications/i/item/9789240073616
- Schiffman, S. S., Scholl, E. H., Furey, T. S., & Nagle, H. T. (2023). Toxicological and pharmacokinetic properties of sucralose-6-acetate and its parent sucralose: in vitro screening assays. Journal of Toxicology and Environmental Health, Part B, 26(6), 307-341. In-vitro screening reported genotoxicity for sucralose-6-acetate, a compound that arises as a manufacturing impurity of sucralose and can also be formed when gut bacteria acetylate sucralose. The work is cell-based screening, not human exposure data. 10.1080/10937404.2023.2213903
- International Agency for Research on Cancer. (2023). IARC Monographs evaluation of the carcinogenicity of aspartame, methyleugenol, and isoeugenol. Lyon: IARC. www.iarc.who.int/news-events/aspartame-hazard-and-risk-assessment-results-released
- Joint FAO/WHO Expert Committee on Food Additives. (2023). Aspartame: summary of safety evaluation. 97th JECFA meeting. www.who.int/news/item/14-07-2023-aspartame-hazard-and-risk-assessment-results-released
- Suez, J., Korem, T., Zeevi, D., Zilberman-Schapira, G., Thaiss, C. A., Maza, O., et al. (2014). Artificial sweeteners induce glucose intolerance by altering the gut microbiota. Nature, 514(7521), 181-186. 10.1038/nature13793
- Suez, J., Cohen, Y., Valdés-Mas, R., Mor, U., Dori-Bachash, M., Federici, S., et al. (2022). Personalized microbiome-driven effects of non-nutritive sweeteners on human glucose tolerance. Cell, 185(18), 3307-3328.e19. 10.1016/j.cell.2022.07.016
- Witkowski, M., Nemet, I., Alamri, H., Wilcox, J., Gupta, N., Nimer, N., et al. (2023). The artificial sweetener erythritol and cardiovascular event risk. Nature Medicine, 29(3), 710-718. 10.1038/s41591-023-02223-9
- Witkowski, M., Nemet, I., Li, X. S., Wilcox, J., Ferrell, M., Alamri, H., et al. (2024). Xylitol is prothrombotic and associated with cardiovascular risk. European Heart Journal, 45(27), 2439-2452. 10.1093/eurheartj/ehae244
- Khan, U., & Dhar, R. (2006). Licensing effect in consumer choice. Journal of Marketing Research, 43(2), 259-266. Across a series of randomised experiments, making a virtuous choice (or merely intending one) increased subsequent preference for an indulgent or luxury option. The authors attribute the effect to a boost in self-concept that licenses the later indulgence rather than to changes in guilt. 10.1509/jmkr.43.2.259