WHO’s 2023 NSS guideline. The whole verdict turns on two moves: the comparator decides the effect, and the discordance between short-term trials and long-term cohorts is left genuinely open — not resolved either way. NSS are defined as «all synthetic and naturally occurring or modified non-nutritive sweeteners that are not classified as sugars» (acesulfame K, aspartame, sucralose, saccharin, stevia, neotame, cyclamates, advantame); sugar alcohols/polyols are excluded. [EXTRACTED (WHO - Non-Sugar Sweeteners 2023) chunk 01]
The recommendation
«WHO suggests that non-sugar sweeteners not be used as a means of achieving weight control or reducing the risk of noncommunicable diseases (conditional recommendation).» [EXTRACTED (WHO - Non-Sugar Sweeteners 2023) chunk 01]
- Conditional recommendation, LOW certainty overall — both stated plainly: «The recommendation is based on evidence of low certainty overall». Read them together — this is a suggestion against on weak evidence, not a strong recommendation against. It says the expected value of using NSS for weight control is not positive on current evidence, not that NSS are demonstrated to harm. [EXTRACTED (WHO - Non-Sugar Sweeteners 2023) chunk 01]
- Scope guard — does NOT apply to people with existing diabetes (excluded from the reviewed evidence), and does not apply to NSS in medicines/hygiene products. Weight control means both weight loss in overweight/obesity and preventing unhealthy gain. [EXTRACTED (WHO - Non-Sugar Sweeteners 2023) chunk 01]
- Class-level, not per-sweetener. The recommendation covers NSS «as a class of compounds», because
«the evidence is currently insufficient to make recommendations for individual NSS» — even though each
has a distinct structure and «may have different physiological effects in humans». So a finding on
saccharin or aspartame is not automatically a finding on stevia.
type-B(one label, chemically distinct objects). [EXTRACTED (WHO - Non-Sugar Sweeteners 2023) chunk 01]
The comparator IS the effect (adults, RCTs)
The single most important reading of the trial evidence: NSS only lowers weight when it displaces sugar, and the effect shrinks toward null the closer the trial gets to the real-world question of replacing sugar with NSS.
effect_measure: body weight MD -0.71 kg (95% CI -1.13 to -0.28), 29 RCTs, n=2433; BMI -0.14 kg/m2 (-0.30 to 0.02), ns
population_and_comparator: adults; higher vs lower/no NSS in any manner (pooled across comparators)
outcome: body weight / BMI (surrogate for sustained-weight / NCD outcomes — see below)
uncertainty: Low certainty (RCTs); most trials <=3 months
certainty: Low
confidence: low[EXTRACTED (WHO - Non-Sugar Sweeteners 2023) chunk 01]
The pooled -0.71 kg dissolves once the comparator is pinned:
- vs free sugars → largest effect. «with the largest effects when NSS were compared with sugars».
- vs water / nothing / placebo → no effect. «NSS compared with water showed no effect on body weight and a non-significant increase in BMI»; «When NSS were compared with nothing/placebo or water … no effects on body weight or BMI were observed.»
- As an actual sugar replacement → attenuates to null. In the RCTs that instructed habitual sugar consumers to switch to NSS versions, «the effect on body weight was significantly weakened … and an effect on BMI was no longer observed» — quantified as body weight MD -0.61 kg (95% CI -1.28 to 0.06), non-significant, BMI -0.01 (-0.38 to 0.35). [EXTRACTED (WHO - Non-Sugar Sweeteners 2023) chunk 01]
Mechanism is displacement, not the molecule. The weight effect «is mediated at least in part by lower energy intake as a result of decreased free sugars intake, rather than primarily by an inherent property of NSS» — a significant weight/BMI difference appeared «only … in trials that reported a reduction in energy intake, and energy intake was only significantly different in the arms of trials that compared NSS with free sugars». Energy intake fell -569 kJ/day (95% CI -859 to -278) and sugars intake -38.4 g/day (-57.8 to -19.1) — but only vs sugar, not vs water/placebo. [EXTRACTED (WHO - Non-Sugar Sweeteners 2023) chunk 01, chunk 02]
The RCT-vs-cohort discordance (adults)
The signature tension in the whole file: short-term trials show a small benefit; long-term observational cohorts show the opposite-signed associations with hard disease.
| Outcome | RCT (short-term) | Cohort (long-term, ~up to 13 yr follow-up) | Cohort certainty |
|---|---|---|---|
| Body weight | MD -0.71 kg (-1.13 to -0.28) | MD -0.12 / -0.01 kg, ns | Very low |
| BMI | -0.14 kg/m2, ns | +0.14 (0.03 to 0.25) | Very low |
| Incident obesity | — | HR 1.76 (1.25 to 2.49) — a 76% higher risk | Low |
| Type 2 diabetes (beverages) | glucose/insulin/HbA1c all null | HR 1.23 (1.14 to 1.32) | Low |
| Type 2 diabetes (tabletop) | — | HR 1.34 (1.21 to 1.48) | Low |
| CVDs | BP / LDL null in RCTs | HR 1.32 (1.17 to 1.50) | Low |
| Stroke | — | HR 1.19 (1.09 to 1.29) | Low |
| CVD mortality | — | HR 1.19 (1.07 to 1.32) | Low |
| All-cause mortality | — | HR 1.12 (1.05 to 1.19) | Very low |
| Cancer (any) | — | HR 1.02 (0.95 to 1.09) — null | Very low |
| Cancer mortality | — | HR 1.02 (0.92 to 1.13) — null | Very low |
| Bladder cancer (saccharin, case-control) | — | OR 1.31 (1.06 to 1.62) | Very low |
[EXTRACTED (WHO - Non-Sugar Sweeteners 2023) chunk 01, Table 1]
Two things the table says on its own:
- The RCT biomarkers are flat. Fasting glucose MD -0.01 mmol/L (-0.05 to 0.04, Moderate), fasting insulin -0.49 pmol/L (-4.99 to 4.02), HbA1c 0.02%, LDL 0.03 mmol/L, systolic BP -1.33 mmHg (ns) — no short-term cardiometabolic signal in either direction. The insulin/glucose route is quiet, which is why a CGM sees nothing (-> Continuous Glucose Monitoring as a Health Intervention).
- Cancer is null in cohorts — the overall-cancer association is flat (HR ~1.02). The saccharin-bladder case-control signal is the exception, at very low certainty. The aspartame/cancer hazard question is not in this guideline (see gap below).
How WHO handles the discordance — the load-bearing appraisal point
This is where symmetric standards bite: WHO does not wave the cohorts away as reverse causation, and it does not accept them as proven harm either. It ran the check and returned a partial verdict.
-
Why short-term weight loss is not banked as benefit: «weight loss and maintenance of a healthy weight must be sustained over the long term to have a meaningful impact on health», so «evidence of minor weight loss or reduced BMI over several months or less … does not represent a health benefit». The trial benefit is on a surrogate (short-term weight), and WHO declines to let it stand for the sustained outcome (-> Surrogate Outcomes). «The discordant results between the RCTs and prospective cohort studies suggest that the small amount of weight loss … in short-term experimental settings may not be relevant to … long-term NSS use in the general population.» [EXTRACTED (WHO - Non-Sugar Sweeteners 2023) chunk 01]
-
The reverse-causation check was RUN — and the association survived it (weakened). Reverse causation here = «those already at elevated risk of disease initiated or increased use of NSS because of their risk status» (people already heavy/dysglycaemic switch to sweeteners because of their condition). Study authors attacked it: adjusting for BMI, stratifying by weight, excluding those who lost weight before baseline, and dropping early follow-up years. The result was mixed — «some results were attenuated, some were strengthened» — but «in the majority of studies, particularly for type 2 diabetes, associations persisted … in fully adjusted models». WHO’s conclusion:
«It was concluded that, although reverse causation and residual confounding may be contributing factors, the available evidence suggests that the associations observed between NSS use and health outcomes in observational studies cannot be dismissed as being solely a result of reverse causation or residual confounding.» [EXTRACTED (WHO - Non-Sugar Sweeteners 2023) chunk 02]
-
The conservative move. «Because of lack of certainty about the overall balance … including the possibility that reverse causation may have contributed to one or more of the associations …, a conservative approach was taken, leading to a conditional recommendation.» The uncertainty about which direction the cohort arrow points is exactly what makes the recommendation conditional rather than strong. [EXTRACTED (WHO - Non-Sugar Sweeteners 2023) chunk 01]
The honest read (symmetric standards). Do not report this as the cohorts are just reverse causation — WHO explicitly refused that framing. Nor as NSS cause diabetes and CVD — the certainty is Low/Very-low and the RCT biomarkers are flat. The defensible statement is: the long-term association is unexplained, cannot be fully attributed to reverse causation, and is not established as causal. This is the The U-Shaped Association Artifact machinery applied and returning not-dissolved, which is a different outcome from alcohol (dissolved) — file it there.
Decision relevance — substitution frame
- vs sugar: a small, short-lived weight benefit that fades to null in true-replacement trials. Better than sugar on weight, marginally, in the short run.
- vs water / unsweetened: no benefit. WHO’s preferred alternative is explicit: «Replacing free sugars in the diet with sources of naturally occurring sweetness, such as fruits, as well as minimally processed unsweetened foods and beverages … should be the preferred alternatives». Swapping sugar for NSS leaves «the overall quality of the diet … largely unaffected». [EXTRACTED (WHO - Non-Sugar Sweeteners 2023) chunk 01]
- NSS are «not essential dietary factors and have no nutritional value» — so the null-benefit finding carries no offsetting cost of not using them. [EXTRACTED (WHO - Non-Sugar Sweeteners 2023) chunk 01]
- Layer-3 caveat: for a person whose realistic alternative is sugar-sweetened soda and who will not switch to water, NSS is still the better of those two — the guideline optimizes a population question (don’t build policy on NSS), not this individual’s next drink.
Other strata
- Children: more limited; one good RCT showed body-fatness reductions, but pooled BMI z-score showed no effect (Moderate certainty); overall certainty across adult+child data judged Low. Two stevia RCTs showed lower dental-caries indicators.
- Pregnancy: higher NSS use associated with preterm birth, OR 1.25 (1.07 to 1.46), Low certainty (three cohorts, dose-response in two); offspring birth-weight/adiposity associations inconsistent. Extrapolated from adult evidence «without downgrading for indirectness». [EXTRACTED (WHO - Non-Sugar Sweeteners 2023) chunk 01]
Named gaps
- Aspartame / cancer HAZARD (IARC 2023) and the ADI (JECFA 2023) are NOT in this guideline. WHO NSS
2023 covers the outcome side (cohort cancer null; saccharin-bladder case-control) and explicitly
defers the hazard identification and risk-assessment to IARC and JECFA, whose 2023 evaluations post-date
it. The hazard-vs-risk distinction for the Are Artificial Sweeteners Bad deliverable needs those
sources — a
G-gap here, not held. AWAITS IARC Aspartame 2023 / JECFA Aspartame 2023 — closes the cancer sub-question (hazard label + unchanged ADI). - Appetite / cravings / gut-microbiome routes are named as putative mechanisms (sweet-taste receptors outside the mouth; microbiota) but «a detailed discussion … is beyond the scope of this guideline» — so they stay insufficient evidence, not findings. AWAITS a sweetener-microbiome RCT — closes the glucose-tolerance/microbiome sub-question.
Limits
- Single-source page (WHO NSS 2023). The reverse-causation adjudication rests on the systematic-review authors’ own sensitivity analyses as WHO reports them, not on an independent re-analysis.
- Open loop (R1): this grades WHO’s reasoning for coherence and source-fidelity; whether NSS actually move a hard outcome is a validity fact no operation here can see.
- NON-independent of the WHO free-sugars family — see Free Sugars Intake. This is the same body’s companion guideline, so agreement between them is not type-E independent corroboration.