EUTECNIQAtlas

Every page of EUTECNIQ on one mapArrow keys move through results · Enter opens · Esc closes

Academy · Reading the evidence Agency and peer-reviewed sources Educational explorers

Read safety evidence in context, six questions before any conclusion

A cell result, a rat tumor, an old study or one bad sample can be real and still say little about a product people take. Six principles, each tied to documented public cases, show what to check first.

  1. SpeciesA cell or animal result is a signal to test.
  2. DoseHazard becomes risk only with exposure.
  3. MaterialPart, extract and chemotype change the result.
  4. PurityA contaminant or substitute may be the cause.
  5. Data qualityReliable studies carry more weight.
  6. SupplyOne good sample does not make a reliable lot.

Status. Education, not medical advice. This page gives no dosing advice and makes no claim that any botanical treats or prevents disease. The three explorers are educational calculations with invented inputs, not data about any substance or product. EUTECNIQ Sciences is a company in formation with no laboratory or study data.

Level
Introductory, with technical depth in the expandable blocks
Reading time
About 8 minutes; 30 with every block opened
Builds on
Toxicology and dose
Tools
3 interactive explorers, 7 figures
Principle 1 · Species

A positive cell or animal test is a signal to test further

The evidence ladder: what cells, animals and people can and cannot showIARC evaluations
Evidence ladder: cells, animals and people, with what each level can show, what it cannot show by itself, and a public case for each; schematicCells and tissues (in vitro)Can showWhether a compound can trigger a change, suchas DNA damage, at a set concentration.Cannot show by itselfWhether people ever reach that concentration.IARC notes that large in vitro screens were built toprioritize chemicals for more testing.Public caseSodium saccharin gave positive genotoxicityresults in mammalian cells; IARC said highosmolality may explain some, not all, of them.Animal studies (in vivo)Can showEffects in a whole organism, after absorption andmetabolism, across a range of doses.Cannot show by itselfWhether the mechanism works in humans. Aspecies-specific mechanism can make a realrodent tumor irrelevant to people.Public cased-Limonene caused kidney tumors in male ratsthrough α2u-globulin, a male-rat protein; IARCrated it Group 3, not classifiable.People (human data)Can showWhat happens at real exposures in realpopulations.Cannot show by itselfA study without an effect covers only theexposures, cancer types and populations itstudied (IARC Preamble).Public caseAflatoxins caused liver tumors in animals and arelinked to liver cancer in people; IARC Group 1.Most human carcinogens that have been studied adequately in experimental animals gave positive results in one or more species. Without strong evidence of aspecies-specific mechanism, agents with sufficient animal evidence are considered a potential human hazard (IARC Preamble, 2019).Evidence ladder: cells, animals and people, with what each level can show, what it cannot show by itself, and a public case for each; schematicCells and tissues (in vitro)Can showWhether a compound can trigger a change, such as DNAdamage, at a set concentration.Cannot show by itselfWhether people ever reach that concentration. IARC notesthat large in vitro screens were built to prioritize chemicalsfor more testing.Public caseSodium saccharin gave positive genotoxicity results inmammalian cells; IARC said high osmolality may explainsome, not all, of them.Animal studies (in vivo)Can showEffects in a whole organism, after absorption andmetabolism, across a range of doses.Cannot show by itselfWhether the mechanism works in humans. Aspecies-specific mechanism can make a real rodent tumorirrelevant to people.Public cased-Limonene caused kidney tumors in male rats throughα2u-globulin, a male-rat protein; IARC rated it Group 3, notclassifiable.People (human data)Can showWhat happens at real exposures in real populations.Cannot show by itselfA study without an effect covers only the exposures,cancer types and populations it studied (IARC Preamble).Public caseAflatoxins caused liver tumors in animals and are linked toliver cancer in people; IARC Group 1.Most human carcinogens that have been studiedadequately in experimental animals gave positive results inone or more species. Without strong evidence of aspecies-specific mechanism, agents with sufficient animalevidence are considered a potential human hazard (IARCPreamble, 2019).
Schematic. Cases: IARC Monographs vol. 73 (1999), saccharin and its salts and d-limonene; vol. 82 (2002), aflatoxins. Statements on screening and on animal evidence: IARC Monographs Preamble (2019), Part B, sections 2, 3 and 4.

IARC found sufficient evidence of cancer in experimental animals for sodium saccharin and for d-limonene, then rated both Group 3 because it judged the rat mechanism not relevant to humans. Aflatoxins show the opposite: liver tumors in animals, and a Group 1 human carcinogen.

Go deeper: three public cases and the IARC rules, with sources

Saccharin and the rat bladder Animal and in vitro findings

Sodium saccharin produced urinary bladder tumors in male rats in four two-generation studies. IARC (1999) concluded that it acts in rats "by a non-DNA-reactive mechanism that involves the formation of a urinary calcium phosphate-containing precipitate, cytotoxicity and enhanced cell proliferation", and that "this mechanism is not relevant to humans because of critical interspecies differences in urine composition."

IARC had rated saccharin Group 2B in 1987 and moved it to Group 3 in 1999. NTP removed saccharin from its Report on Carcinogens in the ninth edition (2000), and California delisted it under Proposition 65 effective April 6, 2001.

The cell data needed context too. Sodium saccharin gave positive genotoxicity results in mammalian cells in vitro; IARC wrote that increased osmolality "would appear to explain some but not all of the findings". A negative study can be weak as well: a monkey study of up to 23 years found no bladder tumors, but IARC noted its relatively low dose (25 mg/kg bw) and few animals.

d-Limonene and the male rat kidney Animal finding

d-Limonene increased kidney tubular tumors in male rats, which make the protein α2u-globulin in the liver. Female rats, mice and rat strains that do not make the protein did not respond. IARC concluded that the mechanism "is not relevant to humans" and rated d-limonene Group 3.

Animal findings that do hold in people IARC

Aflatoxins induced mainly liver tumors in experimental animals, and IARC classifies naturally occurring aflatoxins as carcinogenic to humans (Group 1), a classification based on studies evaluated in vol. 56 and reviewed again in vol. 82 (2002). The IARC Preamble states that "most human carcinogens that have been studied adequately for carcinogenicity in experimental animals have produced positive results in one or more animal species", and that, without strong evidence of a species-specific mechanism that does not operate in humans, agents with sufficient evidence in animals are considered to pose a potential carcinogenic hazard to humans.

IARC also notes that large in vitro screening programs "were designed to evaluate large chemical libraries in order to prioritize chemicals for additional toxicity testing rather than to identify the hazard of a specific chemical".

Related: NTP two-year studies of botanicals on the toxicology and dose course; in silico and in vitro screens on toxicity and exposure models; carcinogenicity study design on toxicology and exposure assessment.

IARC Monographs vol. 73 (1999): saccharin and its salts, and d-limonene, summaries and evaluations (INCHEM); IARC Supplement 7 (1987), saccharin; IARC Monographs vol. 82 (2002), aflatoxins; IARC Monographs Preamble (2019); California OEHHA, saccharin delisting notice (effective April 6, 2001).

Principle 2 · Dose

The dose makes the poison, even for water

Interactive calculator Change the inputs and the result updates. An educational simulation with stated inputs.

Hazard: a fixed dose–response curve (Hill model)response = dn ⁄ (dn + ED50n)
MOE = reference point ⁄ exposure
Response on the toy curve
Margin of exposure (100 ÷ exposure)

Same hazard, different exposure: move the dose along a fixed curveEducational calculation
Show data
Toy curve values at selected doses
Dose, units per dayResponse, %
10.0011
100.11
100 (reference point)10
300 (ED50)50
1,00091.7
Calculated from the Hill model, one of the candidate dose–response models in EFSA's benchmark-dose guidance (2022), with invented parameters ED50 = 300 and n = 2; the reference point is the dose giving a 10% response (100). An educational calculation, not a measurement or a prediction for any product.
Method, assumptions and limits: dose and margin explorer
What it calculates
The response a fixed, invented dose–response curve gives at the chosen exposure, and the margin of exposure (MOE), the reference point divided by the exposure.
Equations
response(d) = dn ⁄ (dn + ED50n) with n = 2 and ED50 = 300; reference point = ED50 · (0.1 ⁄ 0.9)1/n = 100; MOE = 100 ⁄ d.
Variables and units
d: exposure in arbitrary dose units per day, from 0.01 to about 2,500 (slider in log10 steps of 0.05). Response: share of a test population responding, in percent.
Assumptions
No background response; one endpoint; the curve is known exactly. Real assessments fit models to data, report a confidence bound (BMDL) and add uncertainty factors.
Valid range
The slider is clamped to its range; outside it nothing is shown.
What it does not establish
Anything about a real substance, product or person. EFSA states that the magnitude of an MOE "only indicates a level of concern and does not quantify risk".
References
EFSA Scientific Committee (2022), benchmark-dose guidance, Table 3 (candidate models for quantal endpoints); EFSA Scientific Committee (2012), statement on the margin-of-exposure approach.
Verification
Values in the data table were checked against an independent calculation.
Three familiar substances: published reference amounts, upper limits and reported harmAgency and peer-reviewed data
Published reference points: caffeine 400 mg/day cited by FDA and about 1,200 mg rapidly linked to seizures; vitamin A RDA 700–900 and preformed vitamin A UL 3,000 µg RAE/day; 13% hyponatremia and 0.6% critical among 488 Boston Marathon runnersCaffeine, most adultsmg per day, FDA (2024)101001,00010,000400 mg/day: cited by FDA as not generally associated with negative effects for most adultsAbout 1,200 mg consumed rapidly: toxic effects such as seizures can be observed (FDA)400: not generally associated with negative effects≈1,200 taken rapidly: seizures canbe observedVitamin A, adults 19+µg RAE per day; the UL is forpreformed vitamin A (NIH ODS,2025)1001,00010,000RDA, women 19+: 700 µg RAE/dayRDA, men 19+: 900 µg RAE/dayTolerable Upper Intake Level, adults 19+: 3,000 µg RAE/day of preformed vitamin ARDA 700 (women), 900 (men)UL 3,000: maximum dailyintake unlikely to causeadverse effectsWater, 2002 BostonMarathon% of 488 runners with afinish-line blood sampleHyponatremia, Na ≤ 135 mmol/LHyponatremia, Na ≤ 135 mmol/L: 13%13%Critical, Na ≤ 120 mmol/LCritical, Na ≤ 120 mmol/L: 0.6%0.6%agency reference amountupper limit or reported harmPublished reference points: caffeine 400 mg/day cited by FDA and about 1,200 mg rapidly linked to seizures; vitamin A RDA 700–900 and preformed vitamin A UL 3,000 µg RAE/day; 13% hyponatremia and 0.6% critical among 488 Boston Marathon runnersCaffeine, most adultsmg per day, FDA (2024)101001,00010,000400 mg/day: cited by FDA as not generally associated with negative effects for most adultsAbout 1,200 mg consumed rapidly: toxic effects such as seizures can be observed (FDA)400≈1,200Vitamin A, adults 19+µg RAE per day; the UL is for preformed vitamin A (NIH ODS,2025)1001,00010,000RDA, women 19+: 700 µg RAE/dayRDA, men 19+: 900 µg RAE/dayTolerable Upper Intake Level, adults 19+: 3,000 µg RAE/day of preformed vitamin ARDA 700–900UL 3,000Water, 2002 Boston Marathon% of 488 runners with a finish-line blood sampleHyponatremia, Na ≤ 135 mmol/LHyponatremia, Na ≤ 135 mmol/L: 13%13%Critical, Na ≤ 120 mmol/LCritical, Na ≤ 120 mmol/L: 0.6%0.6%agency reference amountupper limit or reported harm
FDA, "Spilling the beans: how much caffeine is too much?" (content current as of August 28, 2024); NIH Office of Dietary Supplements, Vitamin A fact sheet for health professionals (updated March 10, 2025); Almond CS et al., N Engl J Med 2005;352:1550–1556. Log scales for caffeine and vitamin A.
Show data: reference points
Published reference points drawn in the figure
SubstanceValueWhat the source says
Caffeine400 mg/dayFDA: cited for most adults as an amount "not generally associated with negative effects"
Caffeine≈1,200 mgFDA: "toxic effects, like seizures, can be observed with rapid consumption"
Vitamin A700 / 900 µg RAE/dayRDA, women / men aged 19+ (total vitamin A, as retinol activity equivalents)
Preformed vitamin A3,000 µg RAE/dayTolerable Upper Intake Level, adults 19+: "maximum daily intake unlikely to cause adverse health effects"
Water, 2002 Boston Marathon13%of 488 runners with a usable finish-line sample had hyponatremia (sodium ≤ 135 mmol/L)
Water, 2002 Boston Marathon0.6%had critical hyponatremia (≤ 120 mmol/L)

A hazard is what a substance can do; risk depends on how much of it reaches the body. In the 2002 Boston Marathon, 13% of 488 runners tested at the finish had low blood sodium, which was associated with weight gain during the race.

Go deeper: water, vitamin A and the margin of exposure, with sources

Water Human observational study

Of 766 runners enrolled in the 2002 Boston Marathon study, 488 gave a usable blood sample. On multivariate analysis, hyponatremia was associated with weight gain (odds ratio 4.2; 95% CI 2.2 to 8.2), a racing time over 4:00 hours (odds ratio 7.4 against under 3:30) and body-mass-index extremes. Drinking more than 3 liters of fluid during the race was associated with hyponatremia on univariate analysis.

Vitamin A NIH ODS

The adult upper limit of 3,000 µg RAE applies only to preformed vitamin A (retinol and its esters); no upper limit is set for beta-carotene. NIH ODS lists acute toxicity within days to weeks (severe headache, blurred vision, nausea, dizziness, aching muscles, coordination problems) and states that preformed vitamin A intakes above the UL can cause congenital birth defects.

From a reference point to a margin

A margin of exposure divides a reference point from a study, such as a benchmark dose, by the estimated human exposure. For substances that are both genotoxic and carcinogenic, EFSA considers an MOE of 10,000 or higher, based on an animal BMDL10, to be of low concern, and adds that an MOE "only indicates a level of concern and does not quantify risk". Uncertainty factors, tolerable intakes and an MOE calculator with EFSA data are on the toxicology and dose course.

Almond CS, Shin AY, Fortescue EB, et al. N Engl J Med 2005;352(15):1550–1556; NIH ODS Vitamin A fact sheet (updated March 10, 2025); FDA consumer update on caffeine (August 28, 2024); EFSA Scientific Committee, EFSA J 2012;10(3):2578; EFSA Scientific Committee, EFSA J 2022;20(10):e07584.

Principle 3 · Material

What was tested decides what a result means

One plant, different material: plant part and cytotype change the toxic constituent several-foldPeer-reviewed data
Comfrey leaves contained 0.11–0.24 and roots 0.60–3.28 mg/g of four main pyrrolizidine alkaloids; diploid calamus rhizome contained 0.2% and triploid 4.4% β-asaroneComfrey (Symphytum officinale): four main pyrrolizidine alkaloids, mg/g dry weightPolish samples, LC-MS/MS; bars span the lowest to highest sample (Kimel et al., 2023)LeavesLeaves: 0.11–0.24 mg/g dry weight0.11–0.24RootsRoots: 0.6–3.28 mg/g dry weight0.60–3.280.115Calamus (Acorus calamus) rhizome: β-asarone, % w/wTwo cytotypes, one sample each, micellar electrokinetic chromatography (Hanson et al., 2005)Diploid, USADiploid, USA: 0.2% w/w β-asarone0.2%Triploid, IndiaTriploid, India: 4.4% w/w β-asarone4.4%012345Comfrey leaves contained 0.11–0.24 and roots 0.60–3.28 mg/g of four main pyrrolizidine alkaloids; diploid calamus rhizome contained 0.2% and triploid 4.4% β-asaroneComfrey (Symphytum officinale): four mainpyrrolizidine alkaloids, mg/g dry weightPolish samples, LC-MS/MS; bars span the lowest to highestsample (Kimel et al., 2023)LeavesLeaves: 0.11–0.24 mg/g dry weight0.11–0.24RootsRoots: 0.6–3.28 mg/g dry weight0.60–3.280.115Calamus (Acorus calamus) rhizome: β-asarone, % w/wTwo cytotypes, one sample each, micellar electrokineticchromatography (Hanson et al., 2005)Diploid, USADiploid, USA: 0.2% w/w β-asarone0.2%Triploid, IndiaTriploid, India: 4.4% w/w β-asarone4.4%012345
Kimel K et al. Molecules 2023;28(16):6171 (sum of intermedine, lycopsamine and their N-oxides, Table 4); Hanson KM et al. Electrophoresis 2005;26(4–5):943–946. Comfrey on a log scale.
Show data: constituent content by material
Constituent content by plant part or cytotype
MaterialConstituentContentSource
Comfrey leaves, Polish samplesSum of four main pyrrolizidine alkaloids0.11–0.24 mg/g dry weightKimel 2023
Comfrey roots, Polish samplesSame0.60–3.28 mg/g dry weightKimel 2023
Calamus rhizome, diploid, USA wetlandsβ-Asarone0.2% w/wHanson 2005
Calamus rhizome, triploid, India (commercial)β-Asarone4.4% w/wHanson 2005
Calamus, alcoholic extract, diploid / triploidβ-Asaroneabsent / 11%Bertea 2005

Comfrey leaves held about 6 to 14 times less pyrrolizidine alkaloid than roots in a 2023 survey. LiverTox reports no link between drinking green tea and liver injury, while green tea extracts have been implicated in more than 100 published cases.

Go deeper: green tea, comfrey and calamus, with sources

Green tea: beverage against extract Human case data and trials

LiverTox states that "drinking green tea has not been associated with liver injury or serum aminotransferase elevations", that green tea extract and, more rarely, large amounts of green tea have been implicated in clinically apparent liver injury, and that more than 100 such cases attributed to green tea extract have been reported. It describes the injury as idiosyncratic, citing an HLA association, the lack of clear dose dependency and recurrence on re-exposure.

EFSA (2018) estimated that green tea infusions supply a mean of 90 to 300 mg EGCG a day to EU adults, and food supplements 5 to 1,000 mg a day. Intervention trials showed a statistically significant rise in serum transaminases at 800 mg EGCG a day or more taken as a supplement. The dose chart is on the toxicology and dose course.

Comfrey: root against leaf Analytical survey

Kimel et al. (2023) measured intermedine, lycopsamine and their N-oxides by LC-MS/MS. Their sum was 0.11 to 0.24 mg/g dry weight in leaves and 0.60 to 3.28 mg/g in roots, and root profiles varied more between sources than leaf profiles. Two root samples from herbal stores had the highest contents. How pyrrolizidine alkaloids are activated and how EFSA sets their margin of exposure: toxicology and dose course.

Calamus: chemotype and cytotype Analytical studies

Bertea et al. (2005) found 11% β-asarone in an alcoholic extract of triploid calamus and none in the diploid; they describe β-asarone as limiting calamus use because of its carcinogenic properties. Hanson et al. (2005) measured 4.4% w/w in a commercial rhizome from India that they described as triploid, and 0.2% w/w in a diploid from U.S. wetlands; the Indian value falls in the range EMA gives for tetraploid material (4.4–8.3%), so the cytotype was not confirmed. A safety result for one cytotype therefore says little about another until both are analyzed.

IARC records this for every evaluation: "if the material that has been tested in experimental animals or in vitro systems is different from that to which humans are exposed, these differences are noted." Naming the plant part on a label: botanical identity course; chemotypes and genomes: genomes and species.

LiverTox: Green Tea (NCBI Bookshelf, last update November 20, 2020); EFSA ANS Panel, EFSA J 2018;16(4):5239; Kimel K et al. Molecules 2023;28(16):6171; Bertea CM et al. Phytochemistry 2005;66(5):507–514; Hanson KM et al. Electrophoresis 2005;26(4–5):943–946; IARC Monographs Preamble (2019).

Principle 4 · Purity

A contaminant or a substitute can be the real cause

Plant or contaminant: four documented cases where the hazard came from a substitute or a contaminantPublished cases and agency data
Four documented cases where harm or risk came from a substitution or a contaminant: Aristolochia in place of Stephania; heavy metals in 20% of 70 Ayurvedic products; Salmonella in a moringa leaf powder lot; aflatoxins from moldsProduct and settingActual causeDocumented consequenceSlimming pills, Belgium, 1990–1992Labeled herb: Stephania tetrandraAristolochia fangchi, put in by a manufacturingerror18 of 39 patients with end-stage kidney disease who hadsurgery had urothelial carcinomaAyurvedic products, Boston-areastores, 200370 products bought and testedLead, mercury and/or arsenic in 14 of 70 (20%)Each of the 14, taken as the maker directed, could givemetal intakes above published regulatory standardsGreens powder with moringa leaf,2025Supplement powderSalmonella; FDA traceback to a single lot fromone importer of moringa leaf powder11 people ill in 7 states; 3 hospitalizedMoldy crops: peanuts, corn, somenuts and grainsFood commoditiesAflatoxins made by Aspergillus molds before,during or after harvestIARC Group 1; FDA may consider food above 20 ppb totalaflatoxins as adulteratedFour documented cases where harm or risk came from a substitution or a contaminant: Aristolochia in place of Stephania; heavy metals in 20% of 70 Ayurvedic products; Salmonella in a moringa leaf powder lot; aflatoxins from moldsSlimming pills, Belgium, 1990–1992Labeled herb: Stephania tetrandraAristolochia fangchi, put in by a manufacturing error18 of 39 patients with end-stage kidney disease whohad surgery had urothelial carcinomaAyurvedic products, Boston-area stores, 200370 products bought and testedLead, mercury and/or arsenic in 14 of 70 (20%)Each of the 14, taken as the maker directed, couldgive metal intakes above published regulatorystandardsGreens powder with moringa leaf, 2025Supplement powderSalmonella; FDA traceback to a single lot from oneimporter of moringa leaf powder11 people ill in 7 states; 3 hospitalizedMoldy crops: peanuts, corn, some nuts and grainsFood commoditiesAflatoxins made by Aspergillus molds before, duringor after harvestIARC Group 1; FDA may consider food above 20 ppbtotal aflatoxins as adulterated
Vanherweghem JL et al. Lancet 1993;341:387–391; Nortier JL et al. N Engl J Med 2000;342:1686–1692; Saper RB et al. JAMA 2004;292:2868–2873; CDC media release, October 31, 2025; IARC Monographs vol. 82 (2002); FDA CPG Sec. 555.400 (June 2021).
Show data: the four cases
Cases in the figure with their numbers
CaseCauseNumbers reported
Belgian slimming regimenAristolochia fangchi in place of Stephania tetrandra18 urothelial carcinomas among 39 patients (46%; 95% CI 29–62%)
Ayurvedic products, Boston areaLead, mercury and/or arsenic14 of 70 products (20%; 95% CI 11–31%)
Greens powder with moringa leafSalmonella11 ill, 7 states, 3 hospitalized
Aflatoxins in foodAspergillus moldsFDA CPG: food with total aflatoxins > 20 ppb may be considered adulterated

In the Belgian slimming-clinic cases, kidney failure and later urothelial cancer were traced to Aristolochia fangchi supplied in place of Stephania tetrandra. U.S. cGMP rules require supplement makers to set contamination limits for every component and finished batch, and to determine whether those specifications are met.

Go deeper: the clinic, metals, microbes, molds and cGMP, with sources

Substitution: the Belgian clinic Human case series

In 1993, Vanherweghem et al. described nine women with rapidly progressive kidney fibrosis who had followed the same slimming regimen; the clinic had added two Chinese herbs in May 1990 after 15 years without problems. The capsules "had different alkaloid profiles from those expected". Nortier et al. (2000) reported that, "because of a manufacturing error", Stephania tetrandra had been replaced by Aristolochia fangchi. All tissue samples analyzed contained aristolochic acid-related DNA adducts, and cumulative doses over 200 g were linked to higher cancer risk.

Metals, microbes and molds Surveys and agency data

In 2003, Saper et al. bought every unique Ayurvedic product made in South Asia at the stores within 20 miles of Boston City Hall that their systematic search identified; 14 of 70 contained lead, mercury and/or arsenic. In 2025, CDC reported a Salmonella outbreak in which interviews and leftover-product testing identified a supplement powder containing moringa leaf, and "FDA traceback identified a single lot from a direct importer of organic moringa leaf powder".

Aflatoxins are made by Aspergillus molds; FDA lists temperature, humidity and rainfall around harvest as factors and states that high aflatoxin intake is associated with liver cancer in humans. FDA may consider food with more than 20 ppb total aflatoxins adulterated. Limits for botanicals and spices and the other COA contaminant panels: certificate-of-analysis science; microbial hazards: microbiology of botanicals.

What cGMP requires 21 CFR Part 111

A dietary supplement manufacturer must establish an identity specification and "limits on those types of contamination that may adulterate" for each component (§ 111.70(b)), set the same kinds of limits for the finished batch (§ 111.70(e)), and determine whether specifications are met (§ 111.75). The ABC–AHP–NCNPR Botanical Adulterants Prevention Program has documented how ingredients are substituted, diluted or "fortified" with undisclosed lower-cost ingredients, most often for financial gain, and which test methods detect such fraud (Gafner et al., 2023).

Vanherweghem JL et al. Lancet 1993;341(8842):387–391; Nortier JL et al. N Engl J Med 2000;342(23):1686–1692; Saper RB et al. JAMA 2004;292(23):2868–2873; CDC, Salmonella outbreak linked to moringa leaf (October 31, 2025); FDA CPG Sec. 555.400, Aflatoxins in Human Food (June 2021); IARC Monographs vol. 82 (2002); 21 CFR 111.70 and 111.75 (eCFR, current as of October 28, 2026); Gafner S et al. J Nat Prod 2023;86(2):460–472.

Principle 5 · Data quality

Not every study deserves equal weight

Laboratory and reliability standards arrived between 1978 and 2018, so a study's era tells you which ones it could followRegulations and published methods
Timeline: FDA GLP rule 1978, OECD GLP Principles 1981, Klimisch categories and OECD GLP revision 1997, ToxRTool 2009, SciRAP 2014 and its refinement 20181978: FDA GLP rule21 CFR Part 58, Dec 22, 19781978 · FDA GLP rule1981: OECD GLP Principles1981 · OECD GLP PrinciplesPart of the Mutual Acceptanceof Data decision1997: Klimisch categories; OECD GLP revisedFour reliability classes (Feb);revision adopted Nov 261997 · Klimisch categories;OECD GLP revised2009: ToxRTool2009 · ToxRToolCriteria that assign Klimisch 1,2 or 32014: SciRAPCriteria for in vivo reliabilityand relevance; refined in 20182014 · SciRAPTimeline of laboratory-quality and study-reliability standards, 1978 to 20181978: FDA GLP rule1978 · FDA GLP rule21 CFR Part 58, Dec 22, 19781981: OECD GLP Principles1981 · OECD GLP PrinciplesPart of the Mutual Acceptance of Data decision1997: Klimisch categories; OECD GLP revised1997 · Klimisch categories; OECD GLP revisedFour reliability classes (Feb); revision adopted Nov 262009: ToxRTool2009 · ToxRToolCriteria that assign Klimisch 1, 2 or 32014: SciRAP2014 · SciRAPCriteria for in vivo reliability and relevance; refined in 2018
21 CFR Part 58, source note 43 FR 60013 (December 22, 1978); OECD ENV/MC/CHEM(98)17 (1998), Principles of GLP and Council Acts; Klimisch HJ et al. Regul Toxicol Pharmacol 1997;25:1–5; Schneider K et al. Toxicol Lett 2009;189:138–144; Beronius A et al. J Appl Toxicol 2014;34:607–617 and 2018;38:1460–1470.
Show data: the timeline
Dates of quality and reliability standards
YearEvent
1978FDA Good Laboratory Practice regulations, 21 CFR Part 58 (43 FR 60013, December 22, 1978)
1981OECD Council recommends the Principles of GLP as Annex II of the Mutual Acceptance of Data decision, C(81)30(Final)
1997Klimisch et al. define four reliability categories (February); OECD Council adopts revised GLP Principles (November 26)
2009ToxRTool published; it assigns Klimisch categories 1, 2 or 3
2014SciRAP method for evaluating in vivo toxicity studies first published; refined in 2018

Interactive calculator Change the inputs and the result updates. An educational simulation with stated inputs.

Rate each invented study (Klimisch category)

Pooled, all equal (geometric mean)
Quality-weighted (geometric mean)
Studies carrying weight

Pooling every study as equal against weighting by reliabilityEducational illustration, invented inputs
Show data
Invented studies, their default categories and weights
StudyValue, invented unitsDefault categoryDefault weight
A2040
B1230
C4520.5
D15011
E11011
Result at defaultsPooled 44.7; quality-weighted 104.1
Invented studies and values, describing no real substance. Category names: Klimisch et al. (1997) as listed by Ingre-Khans et al. (2019). The weights (1, 0.5, 0, 0) are a teaching choice, not a regulatory rule.
Method, assumptions and limits: quality-weighting demo
What it calculates
Two summaries of five invented study values: the geometric mean with equal weights, and a weighted geometric mean in which each study's weight depends on the reliability category chosen for it.
Equations
pooled = exp(Σ ln xi ⁄ 5); weighted = exp(Σ wi ln xi ⁄ Σ wi), with w = 1, 0.5, 0, 0 for categories 1, 2, 3, 4.
Variables and units
x: an invented study result in arbitrary units; w: weight, dimensionless.
Assumptions
All five studies measure the same endpoint, so their values can be combined; the weights are invented for teaching.
Valid range
If every study is rated 3 or 4, the weighted mean is undefined and the explorer says so.
What it does not establish
How any agency combines studies. Assessors usually select a critical study and endpoint or model the dose–response; they do not average reference points. The demo shows only that weighting changes the answer.
References
Klimisch et al. 1997; Schneider et al. 2009; Beronius et al. 2014 and 2018; Ingre-Khans et al. 2019.
Verification
Default results were checked against an independent calculation.

Good Laboratory Practice (GLP), a quality system for how safety studies are planned, run, recorded and reported, became an FDA rule in 1978 and was recommended by the OECD Council in 1981. Scoring each study's reliability, then weighting by it, keeps an unreported 1979 abstract from counting as much as a recent GLP study.

Go deeper: GLP, Klimisch, ToxRTool, SciRAP and the limits of scoring, with sources

GLP Regulation and OECD Principles

OECD defines GLP as "a quality system concerned with the organisational process and the conditions under which non-clinical health and environmental safety studies are planned, performed, monitored, recorded, archived and reported." Under the OECD Mutual Acceptance of Data decision, data generated in accordance with OECD Test Guidelines and GLP "shall be accepted in other Member countries". FDA's Part 58 covers nonclinical laboratory studies that support applications for products FDA regulates, including food and color additives and human drugs.

Scoring reliability Published methods

Klimisch et al. (1997) separated reliability, relevance and adequacy, and split reliability into four categories: reliable without restriction, reliable with restriction, not reliable, and not assignable. ToxRTool (2009), developed in a European Commission-funded project, gives explicit criteria for in vivo and in vitro data and leads to Klimisch categories 1, 2 or 3. SciRAP offers criteria and a color-coding tool for reliability and relevance; its in vivo method was first published in 2014 and refined in 2018 after testing by toxicologists and risk assessors.

IARC's checklist for an animal study asks whether the agent was characterized, including impurities; whether doses, duration and route were appropriate; whether group sizes were adequate and animals randomized; whether histopathology was adequate; and whether data were reported and analyzed adequately.

Limits of scoring

An analysis of REACH registrations found that Klimisch-based evaluation "is likely to favor studies conducted in compliance with GLP and standardized test guideline". Beronius et al. (2014) proposed criteria so that relevant non-standard studies are not dismissed on very strict reliability criteria alone. Weighting keeps such studies in view at a stated weight; it does not silently drop them.

OECD ENV/MC/CHEM(98)17, OECD Principles on Good Laboratory Practice (as revised in 1997); 21 CFR 58.1 (eCFR, current as of October 28, 2026); Klimisch HJ, Andreae M, Tillmann U. Regul Toxicol Pharmacol 1997;25(1):1–5; Schneider K et al. Toxicol Lett 2009;189(2):138–144; Beronius A et al. J Appl Toxicol 2014;34(6):607–617; Beronius A et al. J Appl Toxicol 2018;38(12):1460–1470; Ingre-Khans E et al. Toxicol Res 2019;8(1):46–56; IARC Monographs Preamble (2019).

Principle 6 · Supply

A good sample is not a reliable supply

21 CFR Part 111: test the identity of every dietary ingredient, and trust a certificate only under five conditionsRegulation
21 CFR Part 111: representative samples of each unique lot; the identity of a dietary ingredient must be tested by the manufacturer; other specifications may rely on a supplier certificate only under five conditions; finished batches verified on all batches or a statistically sampled subsetA component lot arrivesCollect representative samples of each unique lot in each shipment (§ 111.80(a))Identity of a dietary ingredientConduct at least one appropriate identity test or examination(§ 111.75(a)(1)(i)). A supplier's certificate cannot substitute for it; onlyan exemption FDA grants on petition changes this.Other specifications and other componentsTest, or rely on the supplier's certificate of analysis only if all fiveconditions of § 111.75(a)(2)(ii) are met:A. Supplier qualified by confirming its test resultsB. Certificate gives methods, limits and actual resultsC. Qualification documentedD. Certificate periodically re-confirmedE. Quality control approved the basisFinished batchesVerify product specifications on every batch, or on a subset chosen by a sound statisticalsampling plan (§ 111.75(c))after manufacture21 CFR Part 111: representative samples of each unique lot; the identity of a dietary ingredient must be tested by the manufacturer; other specifications may rely on a supplier certificate only under five conditions; finished batches verified on all batches or a statistically sampled subsetA component lot arrivesCollect representative samples of each unique lot in eachshipment (§ 111.80(a))Identity of a dietary ingredientConduct at least one appropriate identity test or examination(§ 111.75(a)(1)(i)). A supplier's certificate cannot substitutefor it; only an exemption FDA grants on petition changesthis.Other specifications and other componentsTest, or rely on the supplier's certificate of analysis only if allfive conditions of § 111.75(a)(2)(ii) are met:A. Supplier qualified by confirming its test resultsB. Certificate gives methods, limits and actual resultsC. Qualification documentedD. Certificate periodically re-confirmedE. Quality control approved the basisFinished batchesVerify product specifications on every batch, or on a subsetchosen by a sound statistical sampling plan (§ 111.75(c))
21 CFR 111.75(a) and (c) and 111.80(a), eCFR current as of October 28, 2026. Schematic of the regulatory text; paraphrased conditions.
Across 25 commercial ginseng products, marker content varied 15- to 200-foldPeer-reviewed survey
Fold variation in marker compounds across 25 commercial ginseng products: ginsenosides 15-fold in capsules and 36-fold in liquids; eleutherosides 43-fold and 200-foldGinsenosides, capsulesGinsenosides, capsules: marker content varied 15-fold between products15×Ginsenosides, liquidsGinsenosides, liquids: marker content varied 36-fold between products36×Eleutherosides, capsulesEleutherosides, capsules: marker content varied 43-fold between products43×Eleutherosides, liquidsEleutherosides, liquids: marker content varied 200-fold between products200×1×10×100×1,000×Highest ÷ lowest marker content between products (log scale)Fold variation in marker compounds across 25 commercial ginseng products: ginsenosides 15-fold in capsules and 36-fold in liquids; eleutherosides 43-fold and 200-foldGinsenosides, capsulesGinsenosides, capsules: marker content varied 15-fold between products15×Ginsenosides, liquidsGinsenosides, liquids: marker content varied 36-fold between products36×Eleutherosides,capsulesEleutherosides, capsules: marker content varied 43-fold between products43×Eleutherosides, liquidsEleutherosides, liquids: marker content varied 200-fold between products200×1×10×100×1,000×Highest ÷ lowest marker content betweenproducts (log scale)
Harkey MR et al. Am J Clin Nutr 2001;73(6):1101–1106 (HPLC and LC-MS/MS; products from one health food store). Product-to-product variation; lot-to-lot data from Gurley et al. (2000) are in the panel below.
Show data: fold variation
Fold variation in marker compounds
Markers, dosage formVariation between products
Ginsenosides, capsules15-fold
Ginsenosides, liquids36-fold
Eleutherosides, capsules43-fold
Eleutherosides, liquids200-fold

Interactive calculator Change the inputs and the result updates. An educational simulation with stated inputs.

Binomial acceptance probability (attributes plan)P(accept) = Σk=0..c C(n,k)
· pk (1 − p)n−k
Lot accepted
At least one bad unit in the sample

Operating characteristic curve: how often a lot of a given quality passesEducational calculation
Show data
Acceptance probability for the chosen plan at selected lot qualities
Bad units in lot, %P(accept), %
0100
Calculated from the binomial model behind single attributes sampling plans; benchmark: Codex CAC/GL 50-2004, section 2.2 (lot size and sample size), where plan 1 (n = 5, c = 0) and plan 2 (n = 50, c = 3) each accept lots at their limiting quality (36.9% and 12.9%) about 10% of the time. An educational calculation, not a sampling plan for any product.
Method, assumptions and limits: sampling-plan explorer
What it calculates
For a single sampling plan by attributes (test n units, accept the lot if at most c are nonconforming), the probability of accepting a lot whose true share of nonconforming units is p, and the probability that the sample contains at least one nonconforming unit.
Equations
P(accept) = Σk=0c C(n,k) pk(1 − p)n−k; P(at least one) = 1 − (1 − p)n.
Variables and units
n: sample units (1–80); c: acceptance number (0–10, held below n); p: share of nonconforming units in the lot, 0–60% in steps of 0.1.
Assumptions
Units drawn at random from a homogeneous lot that is large compared with the sample (Codex: when the sample is at most 10% of a homogeneous lot, the absolute sample size is what matters); no measurement error.
Valid range
Inputs are clamped to the slider ranges; c is limited to n − 1.
What it does not establish
It does not apply to heterogeneous bulk such as aflatoxin in nuts or spices, which Codex CAC/GL 50 excludes and EU rules sample with many pooled increments, and it is not a plan for any product.
References
Codex CAC/GL 50-2004, sections 1.1, 1.4, 2.2.12, 2.2.14 and 2.2.17 with the lot size and sample size passage that follows it.
Verification
Both Codex benchmarks reproduce: plan 1 accepts 10.0% of lots at 36.9%; plan 2 accepts 9.9% at 12.9%.

Part 111 requires samples from each unique lot, an identity test on every dietary ingredient, and five conditions before a supplier's certificate replaces other tests. Whether one bad sub-sample condemns a lot depends on the plan: with c = 0, as in pathogen criteria, it does.

Go deeper: lot-to-lot variation, sampling plans and adulteration, with sources

Lots differ Peer-reviewed surveys

Gurley et al. (2000) assayed 20 ephedra-containing supplements and found significant lot-to-lot variation in alkaloid content for four products; in one, lot-to-lot differences in (−)-ephedrine, (+)-pseudoephedrine and (−)-methylephedrine exceeded 180%, 250% and 1,000%. Harkey et al. (2001) found every one of 25 ginseng products correctly identified to genus, while marker content differed significantly from labeled amounts.

When one bad unit decides, and when it does not Codex and EU rules

Codex defines the acceptance number as "the maximum number of nonconforming units... allowed in the sample if the lot is to be accepted", and warns that "no sampling plan can ensure that every item in a lot conforms." EU microbiological criteria set Salmonella in ready-to-eat foods containing raw egg at n = 5, c = 0, absence in 25 g: the result is unsatisfactory "if the presence of the bacterium is detected in any of the sample units."

Aflatoxins behave differently. The EU sampling rules adopted in 2006, Regulation (EC) No 401/2006, stated that they are "very heterogeneously distributed in a lot". Under that regulation, a lot of spices of 15 metric tons or more was split into sublots of 25 tonnes, each sampled with 100 incremental samples of about 100 g combined into a 10 kg aggregate sample, and the lot was accepted if the laboratory sample conformed to the limit after correction for recovery and measurement uncertainty. Implementing Regulation (EU) 2023/2782 repealed and replaced it from April 1, 2024. Reading a result near a limit: certificate-of-analysis science.

Identity, certificates and adulteration 21 CFR 111

A manufacturer must "conduct at least one appropriate test or examination to verify the identity of any component that is a dietary ingredient" (§ 111.75(a)(1)(i)). A supplier's certificate may be relied on for other specifications only after the supplier is qualified by confirming its results, the certificate states methods, limits and actual results, qualification is documented, the certificate is periodically re-confirmed, and quality control approves the basis (§ 111.75(a)(2)(ii)(A)–(E)). Market surveys of substitution and the identity methods that catch it: botanical identity course.

Gurley BJ, Gardner SF, Hubbard MA. Am J Health Syst Pharm 2000;57(10):963–969; Harkey MR et al. Am J Clin Nutr 2001;73(6):1101–1106; Codex Alimentarius CAC/GL 50-2004, General Guidelines on Sampling; Commission Regulation (EC) No 2073/2005, Annex I (text as adopted); Commission Regulation (EC) No 401/2006, recitals and Annex I part E (text as adopted); Commission Implementing Regulation (EU) 2023/2782; 21 CFR 111.75 and 111.80 (eCFR, current as of October 28, 2026).

Checklist

Six questions to ask about any safety claim

Screenshot thisBefore accepting or dismissing a safety claim, ask:

  1. Species. Was it seen in people, and does an animal mechanism apply to humans?
  2. Dose. How does the tested dose compare with what people actually take?
  3. Material. Which plant part, extract, concentration and chemotype were tested?
  4. Purity. Could a contaminant or a substituted plant explain the effect?
  5. Quality. Was the study well designed, fully reported and scored for reliability?
  6. Supply. Is every lot identity-tested, and was the certificate verified?

Next, the numbers behind a safety decision

The toxicology and dose course turns these questions into NOAELs, benchmark doses, uncertainty factors and margins of exposure.

Health information notice

These statements have not been evaluated by the Food and Drug Administration. Nothing on this site is intended to diagnose, treat, cure or prevent any disease. The information is for general education only. It is not medical advice and does not replace advice, diagnosis or treatment from a qualified health professional, or any medication a clinician has prescribed.

Do not start, stop or change any medicine, supplement or herbal product because of something you read here. Talk to your doctor or pharmacist first, especially if you are pregnant or breastfeeding, have a medical condition, take any medicines, are preparing for surgery, or are caring for a child. Research findings, traditional uses and safety data are reported as their sources state them; they are not recommendations or claims about any product. If you think you may have a medical emergency, call your doctor or 911 at once.

EUTECNIQ Sciences sells no supplements or health products today and receives nothing from any product or ingredient mentioned on this site; any commercial interest it later has in an ingredient will be stated on that ingredient’s page.

Ready to check what you take?

Free tools built on published reference values, each with its sources and its limits. Your lists stay in your browser.