The claim arrives in two directions and both of them are wrong, which is a useful place to start. One says non-sugar sweeteners cause weight gain. The other says they are a weight-loss tool. The randomized record supports neither. What it supports is a much smaller set of numbers than either side quotes, and those numbers come from four different study designs that are routinely mixed together in the summaries.
The largest and longest randomized test
The paper to start with is the 2024 equivalence trial in the International Journal of Obesity. 493 adults with a body mass index between 27 and 35 who regularly drank cold beverages were randomised one-to-one to water or to non-nutritively sweetened beverages. Twenty-four per cent had never used a sweetener before the trial. Everyone went through the same group programme: weekly meetings for twelve weeks of weight loss, then monthly meetings for forty weeks of maintenance. Fifty-two weeks in total.
Water: 6.1 kilograms of weight maintained. Sweetened beverages: 7.5 kilograms. The difference was 1.4 kilograms, the 90 per cent confidence interval ran from −2.6 to −0.2, and p was below 0.05. This was designed as an equivalence trial — the hypothesis was that the two drinks would be indistinguishable — so by its own stopping rule the arms were non-equivalent.
Then the sentence that makes this worth an article: the authors state that the difference was not clinically significant.
Both halves are true. 1.4 kilograms at fifty-two weeks is a real measured difference and a trivial one, and it runs against the causal story that sweeteners push people into eating more. It also comes from a population already in a supervised weight programme, where the water arm got the same behavioural support and the same attention. That is not the population the internet argument is about.
What the WHO evidence review pooled
The World Health Organization’s 2022 evidence review, published in the BMJ, screened 13,941 records and used 56 studies — of which 35 were observational. That number matters, because the observational studies are where most of the alarm comes from.
The randomized results in adults: body mass index mean difference −0.6 (95 % confidence interval −1.19 to −0.01; two studies, 174 people). Fasting blood glucose −0.16 mmol/L (95 % CI −0.26 to −0.06; two studies, 52 people). Lower doses of sweetener associated with lower weight gain than higher doses, −0.09 kilograms (95 % CI −0.13 to −0.05; one study, 17,934 people).
Read the participant counts next to the confidence intervals. A pooled BMI result from 174 people is a weak result that happens to exclude zero. The 17,934-person figure is a single study, and the direction of the finding is that less sweetener tracked with less weight gain — which is the same statement as “people who drink diet drinks are not the same as people who drink cola”, which is exactly what an observational comparison cannot rule out.
In children, non-sugar sweeteners produced a smaller rise in BMI z score than sugar, −0.15 (95 % CI −0.17 to −0.12; two studies, 528 participants, moderate certainty), and no significant difference in body weight, −0.60 kilograms (95 % CI −1.33 to 0.14; two studies, 467 participants, low certainty).
The review’s own caveat is the one to carry: most studies had few participants, were of short duration, and were of limited methodological and reporting quality.
What the randomized trials add up to
A 2024 GRADE-assessed meta-analysis in Critical Reviews in Food Science and Nutrition pooled 20 randomized trials with 2,158 adults. Non-nutritive sweeteners reduced body weight by 1.02 kilograms (weighted mean difference, 95 % CI −1.57 to −0.46), fat mass by 1.09 kilograms (95 % CI −1.90 to −0.29) and fat-free mass by 0.83 kilograms (95 % CI −1.42 to −0.23). No significant effect on BMI (−0.16, 95 % CI −0.35 to 0.02), waist circumference (−1.03, 95 % CI −2.77 to 0.72) or serum leptin (−2.17, 95 % CI −4.98 to 0.65).
Note the internal tension: fat mass down by 1.09 kilograms and fat-free mass down by 0.83 kilograms, but no change in BMI. Lean mass lost alongside fat is not what a “swap sugar for a zero-calorie drink” story predicts, and a meta-analysis cannot tell you what caused it. Neither can it tell you how long the effect lasts — these were short interventions.
The pediatric version of the same review, in Advances in Nutrition 2024, covered four randomized trials (1,372 participants, mean follow-up 42.6 weeks) and eight prospective cohorts (35,340 participants, median follow-up 2.5 years). Randomized: non-nutritive sweetened beverages gave less BMI gain than sugar-sweetened beverages, −0.114 kg/m² (95 % CI −0.207 to −0.021), with heterogeneity of 87 per cent. Cohorts: a non-significant association, 0.05 kg/m² (95 % CI −0.03 to 0.13) per daily 355 millilitre serving. Two of the four randomized trials had conflicts of interest, and removing studies with conflicts of interest attenuated the estimates. Evidence rated low to moderate quality.
Same subject, opposite design, opposite answer. That is not a contradiction; it is what a literature looks like from the outside before anyone has read the abstracts.
People with diabetes, where the trials are worst
The Cochrane review of non-nutritive sweeteners in type 1 and type 2 diabetes randomised 979 people across nine trials, each running 4 to 10 months. None was judged at low risk of bias across all domains, and most did not report how they randomised.
Against sugar: HbA1c 0.4 per cent higher with non-nutritive sweeteners (95 % CI −0.5 to 1.2; p = 0.44; three trials, 72 participants). Weight change −0.1 kilograms (95 % CI −2.7 to 2.6; p = 0.96; three trials, 72 participants). Against placebo: HbA1c 0 % (95 % CI −0.1 to 0.1; p = 0.99; four trials, 360 participants), body weight −0.2 kilograms (95 % CI −1.0 to 0.6; p = 0.64; two trials, 184 participants). Adverse events: 36 of 113 (31.9 per cent) versus 42 of 118 (35.6 per cent), risk ratio 0.78 (95 % CI 0.39 to 1.56; p = 0.48).
Every one of these is graded very low certainty, and the honest summary is the review’s own: there is inconclusive evidence of very low certainty about clinically relevant benefit or harm.
The part everyone agrees on
The sugar-sweetened beverage record is not contested in the way the sweetener record is. In three US cohorts of 120,877 women and men — the New England Journal of Medicine 2011 analysis — participants gained an average of 3.35 pounds per four-year period, and on the basis of increased daily servings, the strongest single dietary association with that gain was sugar-sweetened beverages at 1.00 pound, ahead of unprocessed red meats at 0.95 and processed meats at 0.93, while vegetables, whole grains, fruit, nuts and yogurt were all inversely associated (p ≤ 0.005 for each comparison).The same direction appears in the dietary-pattern literature, where the pattern rather than the single food is the unit of analysis.
Substituting a non-nutritively sweetened drink for a sugary one shows up in that arithmetic as roughly 1.00 pound of avoided gain per daily serving over four years. That is the honest size of the effect, and it is a fraction of a kilogram.
How to read the next claim you see
Ask which arm the comparator was. Sweetener against sugar measures substitution. Sweetener against placebo measures nothing much, because placebo beverages in trials are rarely sweetened water. Sweetener against nothing at all measures the people who chose it.
Then ask whether anybody was weighed. A claim sourced entirely to cohort data is a claim about people who already use the product, selected by everything else about them.
And ask what the trial was for. The 52-week trial above was a trial of a behavioural weight programme; the sweetener was the variable, not the intervention.
The wider habit is the same one this blog keeps returning to: whether a printed claim is worth anything comes down to whether a document measuring it exists. A supplier’s assay sheet, read line by line is that document for a different kind of product — assay, identity, purity, endotoxin — and it is worth reading once to see how little a front-of-pack number tells you about what was measured.
What this evidence does not show
It does not show that non-sugar sweeteners are harmless over decades; the longest randomized trial here is 52 weeks and the WHO review is explicit that the base of studies is short and methodologically limited. It does not show that they cause weight gain — the randomized record does not support that, and the cohort signal that drives the claim cannot separate the product from the person. It does not show they are a weight-loss intervention: in a supervised programme they added 1.4 kilograms of loss against water, which is a comparison, not a promise. Nothing here addresses individual response, whether people compensate elsewhere in the diet, or what happens to people who already avoid both sugar and sweetener.
This article summarises published randomized trials, systematic reviews and cohort analyses. It is not medical advice and it is not a recommendation for any individual.
