Supplements · Sweeteners
stevia is safe to consume
In plain terms: Is stevia safe to consume?
Leans support(preclinical) Supplements 🐭 Non-human evidence💰 Industry COI noted
Part of: • Stevia
RefutedContestedStrong support
consensus score 1.00
⚖️ Thin evidence — read the needle loosely. The score shows which way the studies lean, but there are too few independent, high-quality ones to place it firmly. Expect this to move as better evidence arrives.
🔬 Capped to leans support — preclinical evidence only. The raw signal looked like strong support, but a claim about human health can't be graded that high on animal or lab evidence alone. It stays capped until human trials weigh in.
By the weight of the evidence, yes - it's not genotoxic or cancer-causing at normal amounts, and regulators (FDA, EFSA) set a safe daily intake of about 4 mg per kg of body weight. The old cancer scare doesn't hold up. Honest caveat: a lot of the foundational safety testing was industry-funded, but independent reviews reach the same reassuring conclusion.
📅 Last reviewed: 2026-07-15 ⓘ
Evidence ladder
How far up the ladder this claim has climbed. A high consensus on a low rung means "consistent so far," not "proven in people."
Top evidence so far: Animal studies (Animal)
MechanismIn-vitroAnimalObservationalRCTMeta-analysis
How the studies fall
3 support 0 contradict 0 tested null 0 mixed · 3 sources, 3 independent groups
What the evidence shows
By the weight of the evidence, stevia is safe at normal intakes: it is not genotoxic or carcinogenic across large batteries of tests, and regulators (FDA GRAS; EFSA/JECFA) set an acceptable daily intake of 4 mg/kg body-weight per day. The old 'stevia causes mutations/cancer' scare does not hold up.
The evidence (9)
| Source | Grade | Stance | Quality | Finding |
|---|---|---|---|---|
| Mendoza-Pérez S et al 2024 · PLoS One | animal | supports | low | Wistar rats, n=8/group/sex, 120 days from weaning, 0.94% crude S. rebaudiana LEAF water infusion vs 10% glucose water vs plain water. Authors conclude the crude extract 'did not show negative effects on the normal blood serum levels of glucose, triglycerides, cholesterol, and the hormones insulin, glucagon, leptin, ghrelin, and glucose dependent insulinotropic peptide, GIP'. Two stevia-vs-blank differences in FEMALES only: lower cumulative body-mass gain (158-171 g vs blank 178-192 g, p=0.0058) and lower GIP (p<0.0001); males null throughout (glucose p=0.2547, insulin p=0.406, body mass p=0.218). Narrow safety window - fasting serum panel only, no histopathology, organ weights or genotoxicity ('Internal organs were collected for future research'). Crude leaf infusion, not the >=95% purified steviol glycosides that hold GRAS status. |
| Lea IA, Chappell GA, Wikoff DS 2021 · Mutat Res Genet Toxicol Environ Mutagen | observational # was meta-analysis — narrative weight-of-evidence review, no pooling; own extract says "not a meta-analysis" (reread #295, precedent k) | supports | low | Narrative weight-of-evidence review (not a meta-analysis; no pooled estimate, no stated search protocol, no registration) surveying genotoxicity of five low- and no-calorie sweeteners including steviol glycosides: 'The weight-of-evidence demonstrates overall negative findings across assay types for each sweetener when considering the totality of study design, reliability and reporting quality, as well as the lack of carcinogenic responses (or lack of responses relevant to humans) in animal cancer bioassays as well as observational studies in humans.' Explicitly downstream of prior regulatory appraisal - 'Data from peer-reviewed literature and the ToxCast/Tox21 database were evaluated and integrated with the most recent weight-of-evidence evaluations from authoritative sources' and 'This conclusion is consistent with the opinions of authoritative sources'. All three authors affiliated 'ToxStrategies, Inc.' (a risk-assessment consultancy, NOT Exponent as previously recorded); no funding or competing-interests statement in the available text. Abstract-grade only - no OA full text exists. |
| Williams LD, Burdock GA 2009 · Food Chem Toxicol | in-vitro | supports | moderate | OECD-guideline battery: RebA non-mutagenic (Ames, chromosomal aberration, mouse lymphoma) and non-genotoxic (micronucleus, UDS) at high doses. [industry consultancy] |
| Marchitti SA et al. 2025 · Adv Nutr | animal | supports | low | Narrative weight-of-evidence review (no new data): the authors 'conducted a high-level evaluation of available carcinogenicity and mechanistic (e.g., genotoxicity) evidence from animal and in vitro studies' for 8 nonsugar sweeteners. On stevia: 'Genotoxicity testing of steviol glycosides has largely been negative, showing no consistent evidence of genotoxicity, including in tests performed to detect mutagenicity, chromosomal aberrations, micronuclei formation, and general DNA damage'; 'steviol glycosides were shown to be inactive in all 17 genotoxicity assays of the NTP ToxCast/Tox21 program'; 'Three GLP-compliant standard rodent cancer bioassays have been conducted on steviol glycosides and have demonstrated no treatment-related increases in the formation of tumors or non-neoplastic lesions'; conclusion: 'Taken together, the available experimental evidence supports previous conclusions that steviol glycosides are not genotoxic or carcinogenic.' Direction is confirmatory, but the paper pools nothing and restates dossiers the vault already counts, so CONVENTIONS §5 zero-weights it. Quality low: 'ABA provided funding for this paper' and 'All authors are employed by Gradient or the American Beverage Association (ABA)', with no protocol, no risk-of-bias instrument, and the conclusion asserted in the title. |
| Usami M, Sakemi K, Kawashima K, Tsuda M, Ohno Y 1995 · Eisei Shikenjo Hokoku | animal | supports | moderate | Rat teratogenicity study: no increased fetal malformation up to 1000 mg/kg/day; NOAEL >1000 mg/kg/day. |
| Smirnova MG 2001 · Vopr Pitan. 2001;70(4):41-4 | mechanism | supports | low | Narrative literature review (Russian; English abstract only): 'The review consist of modern data of physiologic and toxic effects on organism of sweetening stevioside with low energy value... His sweetening is considered by the most of investigators as not toxic, not mutadenic and not cardinogenic one.' No primary data, no doses, no study count, no citations visible - the safety statement is an attribution to other investigators, not a finding of this paper. |
| Magnuson BA, Carakostas MC, Moore NH, Poulos SP, Renwick AG 2016 · Nutr Rev | mechanism | supports | low | Narrative review of the chemistry, regulatory status and biological fate of five low-calorie sweeteners, stevia leaf extract among them ('saccharin, stevia leaf extract (steviol glycoside), and sucralose.'). It reports no data of its own; on safety it restates the regulators - 'safety of LNCS use has been affirmed by regulatory agencies worldwide.' Review-role echo, not independent evidence. Industry-adjacent authorship: 'S.P. Poulos is with the Calorie Control Council, Atlanta, Georgia, USA.', 'Published by Oxford University Press on behalf of the' ILSI. |
| EFSA Panel on Food Additives and Flavourings (FAF) 2024 · EFSA Journal 2024;22(11):e9045 | meta-analysis | mixed | high | EFSA FAF Panel opinion on steviol glycosides (E 960a–d), on an International Stevia Council dossier. No new safety data: 'The Panel noted that no new biological and toxicological data were provided in support of this request.' Two conclusions, both qualifying rather than affirming: (1) it REFUSED the applicant's request to raise the ADI from 4 to 6 or 16 mg/kg bw/day — 'there is insufficient justification to increase the current ADI… of 4 mg/kg bw per day', the ADME data 'not sufficiently robust' to support a chemical-specific adjustment factor; (2) at regulatory maximum permitted levels the ADI is already exceeded at the 95th percentile — '4.1 and 4.8 mg/kg bw per day for infants and toddlers', rising to 6.9 for toddlers under the proposed extension of use. The Panel calls the exposure model conservative ('an overestimation', exceedances from one country per age group), and the 4 mg/kg ADI itself is the 2010 value from a 100-fold uncertainty factor on a 2-year rat carcinogenicity NOAEL — restated here, not re-derived. [independent regulator; industry dossier; review layer, no new data] |
| Urban JD, Carakostas MC, Brusick DJ 2013 · Food Chem Toxicol | meta-analysis | supports | low | Narrative review, not a meta-analysis: it pools nothing and states its own method as 'this review evaluates the specific genotoxicity studies that are the sources of concern, and evaluates the adequacy of the database' — no search strategy, no pooled estimate, no registration; PubMed types it Review. Conclusion is supportive but scoped to one axis: 'The current database of in vitro and in vivo studies for steviol glycosides is robust and does not indicate that either stevioside or rebaudioside A are genotoxic' and 'a lack of evidence for neoplasm development in rat bioassays, establish the safety of all steviol glycosides with respect to their genotoxic/carcinogenic potential' — genotoxicity/carcinogenicity only, silent on metabolic, renal, reproductive and microbiome safety. Written to rebut published mutagenicity concerns and leans on 'approximately 20 expert panels' rather than on new analysis. Quality low: first/corresponding author affiliation is 'ToxStrategies, Inc., 9390 Research Blvd., Suite 250, Austin, TX 78717, United States' (product-defense consultancy), co-author Carakostas is the Cargill-side stevia scientist (group_id urban-cargill), no author is academic or public-agency, and no funding statement was reachable in the abstract-grade text. |
Disagree, or know a study we missed?
We grade by evidence, not opinions. The way to weigh in is to point us to a study we haven't cited (check the evidence table above first), or to flag a problem with one we have. Every submission is reviewed; if it holds up, the grade updates and shows in Science Changes Its Mind.
Opens a short form. You'll sign in with Google so submissions are tied to a real account — we don't display your identity, and we only accept a link we can verify (PubMed, DOI, ClinicalTrials.gov).
Educational only, not medical advice. Grades and scores reflect published evidence weighted by study design and quality; see the methodology.