Sweeteners and blood pressure: the comparator decides

A sweetening system is bought to take sugar out of a formula. Everything that follows from that removal — fewer calories, a different mouthfeel, a different label — belongs to the sugar that left, not to the molecule that arrived. A meta-analysis published in 2026 makes that distinction unusually easy to see, because it reports a blood pressure effect and then, in its own conclusion, hands the credit to the comparator.

The finding matters commercially for one reason only: it is the kind of number that ends up in a sales deck with the comparator stripped off. So it is worth reading slowly, next to a second 2026 synthesis of the same question that reached the opposite answer, and next to the counts of how many trials sit behind each estimate.

What the 18 trials actually compared

Moghtaderi and colleagues searched PubMed, Scopus, ISI Web of Science and Google Scholar to July 2025 for randomised controlled trials of non-nutritive sweeteners in adults, and pooled the effect on blood pressure, heart rate and pulse rate. The comparator was not one thing: trials compared sweeteners against water, against sugar-sweetened beverages, or against placebo, and all three were pooled together.

That is the first thing to hold on to. A trial in which a participant drinks a sweetened beverage instead of a sugared one tests sugar removal. A trial in which a participant drinks a sweetened beverage instead of water tests sweetener addition. These are different experiments with different expected results, and a single pooled number describes neither.

The numbers, and how thin some of them are

The pooled estimates, as reported, are these. Resting systolic pressure fell by 3.25 mmHg (95% CI −5.74 to −0.76) across 18 trials and 1,898 participants. Resting diastolic pressure fell by 2.34 mmHg (−4.28 to −0.40) across 17 trials and 1,675 participants. Twenty-four-hour diastolic pressure fell by 2.30 mmHg (−4.14 to −0.47), but that estimate pools two trials and 176 people. Twenty-four-hour systolic pressure was not significant. Heart rate and pulse rate were unaffected.

The two largest effects in the whole analysis are the supine ones: 6.10 mmHg systolic (−11.51 to −0.69) and 5.70 mmHg diastolic (−10.59 to −0.81). Each is pooled from one trial of 37 participants. They are reported honestly by the authors, with the counts attached. They are also the numbers most likely to be lifted out of context, because they are the biggest.

All values as reported by Moghtaderi et al. 2026. Negative values mean lower pressure.

Why the comparator is the intervention

The review’s own summary sentence says the reduction holds “especially when compared with sugar-sweetened beverages”. A second 2026 synthesis lets that sentence be tested. The consensus statement of the Korean Diabetes Association and the Korean Nutrition Society pooled only trials that substituted an artificially sweetened beverage for a sugar-sweetened one — six trials, 8 to 52 weeks — and found systolic pressure unchanged (−1.21 mmHg, 95% CI −3.20 to 0.78) and diastolic borderline but not significant (−1.45 mmHg, −2.98 to 0.09). Both were graded low certainty.

So the two syntheses do not line up, and the public text does not let anyone reconcile them: the comparator-stratified figures behind the first review are not in its abstract, and the full article sits behind the publisher’s paywall with no open repository copy. What can be said is that the same pattern shows up in body weight, where the numbers are available. Pooling 19 trials, replacing sugar with a sweetener changed body weight by −0.79 kg (95% CI −1.55 to −0.04), and the authors report that the effect was present only when the comparator was caloric sugar (−0.94 kg), and only in trials shorter than 18 weeks.

Read together, the honest reading is that these are trials of taking sugar out, not trials of putting sweetener in. That is not a weaker claim for anyone selling a sweetening system. It is simply the accurate one, and it points the specification at the right question.

What it means for a formulator

Specify the sugar that leaves, not the sweetener that arrives. If the measured effect belongs to sugar removal, then the design target is how much sugar comes out of the formula and what the replacement does to the rest of the matrix — bulk, freezing point, browning, water activity. The sweetener is dosed to a sweetness target and then checked against the maximum for that food category in Annex II of Regulation (EC) No 1333/2008, which is expressed in steviol equivalents, not in grams of material. Our notes on steviol glycoside systems start from that conversion rather than from a potency multiple.

Sweetener identity is a document, not a word. The trials in these reviews describe their interventions as aspartame, sucralose, saccharin or stevia. “Stevia” is not a specification. In the EU, steviol glycosides are four separate additives with separate specifications and separate label identities: E 960a from the leaf, E 960b rebaudioside M from fermentation, E 960c produced enzymatically and E 960d glucosylated. A certificate of analysis that names a glycoside profile, with an acceptance criterion per component, is the only thing that makes two offers comparable — the same argument that applies to a fraction in a multi-source protein system.

Decide whether the specification is at release or at expiry. Blood pressure moves with the measurement protocol; sweetness moves with storage. A steviol glycoside profile can rearrange under heat and low pH, so a total-glycosides assay can pass material that has already shifted. Fix the method, fix the acceptance criterion per glycoside, and state explicitly whether the number is taken at packing or at end of shelf life. Both numbers exist; only one of them supports the label.

Limits, stated plainly

The main review is available to us as its indexed abstract only. Critical Reviews in Food Science and Nutrition does not deposit in PubMed Central and the publisher’s site refused access, so we can report the pooled estimates and their confidence intervals but not the heterogeneity statistics, the risk-of-bias assessment, the certainty grading or the stratified results. Every figure quoted here comes from the abstract as indexed; none is inferred.

Within what is reported, three weaknesses are visible without the full text. Pooled estimates built from one trial of 37 people sit in the same table as estimates built from 18 trials. Several chemically unrelated sweeteners are pooled as one exposure. And three comparators that test different hypotheses are pooled as one comparison. The weight data carry the same caveat: the subgroup figures from the 19-trial review are abstract-level, and two of them are reported without confidence intervals, so they are quoted here in prose and left out of the chart.

One further point belongs here for balance. The same Korean consensus statement that found no blood pressure effect also reviewed 20 systematic reviews of prospective cohorts and reported that habitual intake of artificially sweetened beverages was associated with a higher risk of hypertension (relative risks 1.08 to 1.14), similar in direction to sugar-sweetened beverages, at low certainty and with reverse causation explicitly flagged. Short-term trials and long-term cohorts answer different questions. Anyone quoting the trial result should be willing to quote the cohort one.

Specifying a system against numbers like these, rather than against a brochure, is the conversation our request line is for.