Each nonclinical safety study should answer a defined question about the intended exposure, and its result is judged against that exposure
Which studies a botanical safety case needs, how a NOAEL or BMDL is derived and compared with intake, and what GLP means. EUTECNIQ would design such programs with qualified laboratories; it has run none.
Status. EUTECNIQ Sciences is in formation, with no laboratory, study data or testing agreements. This page describes how programs would be designed and the rules and methods that govern them. Educational, not legal or toxicological advice.
Academy fundamentals: Course 9, Toxicology and dose
Related: dose and exposure in human studies · lessons learned from documented failures
A safety case starts with the material and the exposure, and adds studies only where a question remains
Text version of the five steps
- Characterize the test materialIdentity, composition, contaminants and stability of the exact material dosed
- Estimate exposureIntake per serving and per day, users, duration, vulnerable groups
- Weigh existing evidenceHistory of use, published toxicology, case reports, read-across
- Close the gaps that remainIn silico and in vitro first; animal studies only when a concern is unresolved and allowed
- Compare dose with exposureReference point (NOAEL or BMDL) against intake: margin of exposure or a health-based guidance value
FDA's guidance says firms are "not required to complete and submit all the studies listed in a given row" of its standard safety testing table.
Explanation, details and sources
No law prescribes a fixed battery for a new dietary ingredient. FDA's October 2026 final guidance encourages new approach methods where they are scientifically valid and fit for purpose.
Four dose groups give a BMD of 2.29 to 3.31 mg/kg per day depending on the model; the margin depends on intake
| Model | AIC | BMD | BMDL |
|---|
EFSA's Bayesian model average for 10% extra risk: BMD 4.21, 90% credible interval 2.15–8.31 mg/kg bw per day.
Method, assumptions and limits
- What it calculates
- Maximum-likelihood fits of three quantal dose–response models with a background term to the four dose groups, the benchmark dose (BMD) at the chosen extra risk, its one-sided 95% lower bound (BMDL) and the Akaike information criterion.
- Equations
- Log-logistic P(d) = g + (1 − g)/(1 + e−(a + b ln d)); Weibull P(d) = g + (1 − g)(1 − e−c·dk); log-probit P(d) = g + (1 − g)Φ(a + b ln d). Extra risk = (P(d) − g)/(1 − g). BMD solves extra risk = BMR. BMDL: smallest BMD whose profile log-likelihood lies within χ²(0.90, 1)/2 = 1.353 of the maximum. AIC = 2 × 3 − 2 ln L.
- Assumptions
- Binomial responses; slopes and shape parameters unrestricted except positive; no model averaging. EFSA's 2022 guidance recommends Bayesian model averaging, so its interval (shown) differs from these single-model, frequentist results.
- What it does not establish
- A reference point for any real substance. The EFSA example substance is not named, and the explorer reproduces a method on published numbers.
- Verification
- Six cases (three models, 10% and 5%) match an independent SciPy implementation (tests/numerics/ref/regulatory-toxicology.py) within 1% for BMD and 2% for BMDL.
- References
- EFSA Scientific Committee (2022), sections 2.3 and 2.6 and Appendix D.
Method, assumptions and limits
- What it calculates
- Estimated daily intake EDI = (mg per serving × servings per day) ÷ body weight; margin of exposure MOE = reference point ÷ EDI; ADI = NOAEL ÷ safety factor; and the ratio EDI ÷ ADI.
- Valid range
- 0 to 5,000 mg per serving, 30 to 120 kg. Intakes are for one constituent at the stated amounts; no other dietary sources are added.
- Assumptions
- The NOAEL of 3 mg/kg bw per day and the BMDL of 2.149 come from EFSA's example rat study; the safety factors follow FDA's NDI guidance (usually 100 for a chronic study, 1,000 or 2,000 when shorter studies are extrapolated to chronic use).
- What it does not establish
- Safety of any product. A margin "indicates a level of concern and does not quantify risk" (EFSA 2012), and the 10,000 benchmark applies only to substances that are both genotoxic and carcinogenic.
- Verification
- Five cases checked against tests/numerics/ref/regulatory-toxicology.py.
On the same data the BMD at 10% extra risk ranges from 2.29 (Weibull) to 3.31 (log-probit) mg/kg per day, which is why EFSA prefers model averaging.
Toxicology, efficacy and interaction research answer different questions, and evidence for one never substitutes for another
Each question has its own methods and standard of proof: NDI safety case, clinical research, measured interactions.
Explanation, details and sources
A clean toxicology package says nothing about benefit. A strong cell assay says nothing about safety or about effects in people. Each question has its own methods and its own standard of proof.
Is it safe at the intended exposure?
Hazard identification with OECD and ICH test guidelines, a reference point such as a NOAEL or BMDL, and comparison with human exposure. This page covers it.
How a safety case fits an NDI notification →Does it do what is claimed, in people?
Discovery assays and animal models generate hypotheses. Claims about people rest on human studies; FTC guidance generally expects randomized, controlled trials for health benefit claims.
Clinical research →Does it change the effect of something else?
Pharmacokinetic interactions, where one substance changes another's blood levels, are measured with enzyme and transporter studies. They are a different phenomenon from synergy at a target.
Measured herb–drug interactions →EUTECNIQ would design the program and characterize the test material; accredited laboratories would run definitive studies
Text version: the division of work as a table
| Activity | EUTECNIQ (proposed) | Qualified external partner |
|---|---|---|
| Test-material characterization | Identity, composition and stability of the lot to be dosed | Accredited analytical laboratory where independent testing is needed |
| Exposure assessment | Intake estimates for each intended use and population | Consumption-survey data providers |
| Literature review and read-across | Systematic review, in silico screening, data-gap analysis | Independent toxicologist review |
| Study design | Study selection, dose setting and protocol input | Study director at the testing facility finalizes the protocol |
| Definitive studies | Sponsor oversight and test-article supply | GLP-compliant or otherwise qualified contract laboratory |
| Reference point and margins | Dose–response analysis, BMD modeling, margin of exposure | Independent expert panel where a GRAS conclusion is sought |
| Safety narrative and dossier | Drafting and evidence mapping | Regulatory counsel; the notifier submits |
- GLP compliance≠a well-designed or relevant study
- No adverse effect at the doses tested≠safe at any dose
- A negative Ames test≠no genotoxicity of any kind
- MOE ≥ 10,000≠a general safety benchmark
- An acknowledged NDI notification≠FDA approval
- A long history of use≠safety at a new dose or in a new form
Explanation, details and sources
Definitive studies need facilities, animal welfare oversight and, for some submissions, a GLP quality system. EUTECNIQ proposes to own the design and the dossier and to place studies with qualified laboratories.
Study selection, test guidelines, dose–response, absorption and interactions, and Good Laboratory Practice
Exploratory studies guide decisions; definitive studies support a safety conclusion
Why exploratory and definitive studies differ
The two serve different purposes and are held to different standards. An exploratory result that looks reassuring is not a substitute for a definitive study when the safety conclusion depends on it.
Show the table (5 rows)
| Aspect | Exploratory (non-definitive) | Definitive |
|---|---|---|
| Purpose | Rank candidates, find hazards early, set doses, choose a formulation | Characterize hazard and dose–response for the safety conclusion |
| Typical studies | In silico alerts, screening genotoxicity, cytotoxicity, dose-range-finding | Guideline studies such as OECD TG 471, TG 487, TG 408, conducted as written or with justified deviations |
| Test material | May be a research lot | A characterized lot representative of the commercial ingredient |
| Quality system | Documented, but outside 21 CFR Part 58: "basic exploratory studies" are excluded from its definition of a nonclinical laboratory study | GLP when the study supports a submission that calls for it (for example an IND or a food additive petition); otherwise documented to an equivalent standard |
| Use of the result | Internal decisions; may be summarized as supporting information | Basis of the reference point and the safety narrative |
21 CFR 58.1 and 58.3 (eCFR, current as of October 21, 2026); FDA, NDI Notifications and Related Issues: Identity and Safety Information About the NDI (final guidance, October 2026; docket FDA-2025-D-6501).
Testing climbs from existing knowledge to computer and cell models, and reaches animals only when a concern remains unresolved
How each tier passes a question on
Each tier either resolves the question, so the program documents the conclusion and stops, or passes a defined concern to the next tier. Ames (TG 471), micronucleus (TG 487) and repeat-dose (TG 407 or 408) studies are common, but each is chosen for a reason, not by default. In the EU, animal tests for cosmetic purposes are banned outright, so cosmetic ingredients stop at the non-animal tiers.
- Tier 0
Existing knowledge
- Identity and composition
- History of use and exposure
- Published toxicology and case reports
- Read-across from close analogues
- Tier 1
In silico
- (Q)SAR models and structural alerts
- Read-across with documented similarity
- Exposure estimate for each intended use
- Used to target tests, not to replace them
- Tier 2
In vitro and in chemico
- Ames, TG 471
- Micronucleus TG 487; aberration TG 473
- Skin sensitization TG 442C/D/E; TG 497
- Skin and eye TG 439, 492, 437
- Phototoxicity TG 432
- hERG (ICH S7B); liver cell models
- Tier 3 EU cosmetics: banned
In vivo, if warranted
- Genotoxicity follow-up TG 474, 489
- Acute oral TG 420, 423, 425
- Repeat dose TG 407, 408
- Reproduction TG 421/422, 414, 443
- Chronic and cancer TG 452, 451, 453
- Tier 4
Human evidence
- Tolerability and PK under IRB oversight
- Safety monitoring within efficacy trials
- Post-market adverse-event review
What FDA's 2026 NDI guidance says about the test battery and history of use
What FDA's NDI guidance says about testing
The guidance discusses a genetic toxicity battery of two or three studies, repeat-dose, subchronic, reproductive and developmental studies and human studies, with a standard safety testing table in Appendix A. Firms are "not required to complete and submit all the studies listed in a given row", and FDA states that the NDI safety standard differs from those for food additives and drugs.
FDA, NDI identity and safety guidance (October 2026), sections VII and Appendix A; 91 FR 66336 (October 20, 2026).History of use has a defined meaning
If a notifier relies on history of safe use alone, FDA "generally recommends 25 years of widespread use to be the minimum"; fewer years can be supported with additional information. The relevant history is use of the same material at comparable intake, which a concentrated extract or a new delivery form may not share.
FDA, NDI identity and safety guidance (October 2026), section VI.What is typically expected depends on the product category, and the legal basis differs for each
Scope of this matrix
A summary of typical expectations, not legal advice; every program needs its own gap analysis.
Text version: the matrix with the basis for each cell, and the legal basis of each column
| Test family | Dietary supplementnew dietary ingredient | Food additiveFDA Redbook | Cosmetic, EUReg. (EC) 1223/2009 | Cosmetic, U.S.MoCRA | DrugIND / NDA |
|---|---|---|---|---|---|
| Genotoxicity battery | Case-by-caseSelected when history of use does not cover the intended exposure | ExpectedRecommended at every concern level | Non-animal onlyIn vitro tests only, e.g., TG 471 and 487 | Case-by-caseChosen by the safety assessor | ExpectedStandard battery under ICH S2(R1) |
| Acute oral toxicity | Case-by-caseOnly if a specific question requires it | Not usually relevantNot in the Redbook minimum battery | No animal testsNo new animal tests for cosmetic purposes | Case-by-caseChosen by the safety assessor | Case-by-caseCase-by-case under ICH M3(R2) |
| Repeat-dose (28 or 90 days) | Case-by-caseWhen exposure exceeds what history of use supports | ExpectedShort-term at level I; subchronic from level II | No animal testsNo new animal tests; existing data and non-animal approaches | Case-by-caseChosen by the safety assessor | ExpectedDuration matched to the clinical program (ICH M3(R2)) |
| Chronic and carcinogenicity | Case-by-caseOnly for a specific concern | ExpectedConcern level III | No animal testsNo new animal tests | Case-by-caseChosen by the safety assessor | ExpectedFor drugs intended for long-term use (ICH S1A) |
| Reproductive and developmental | Case-by-caseFor example when the target population includes women who may become pregnant | ExpectedConcern levels II and III | No animal testsNo new animal tests | Case-by-caseChosen by the safety assessor | ExpectedICH S5(R3), timed per ICH M3(R2) |
| Skin sensitization | Not usually relevantNot relevant to an oral product | Not usually relevantNot relevant to food use | Non-animal onlyTG 442C/D/E and defined approaches, TG 497 | Case-by-caseMost relevant to leave-on products | Case-by-caseFor topical drugs |
| Skin and eye irritation | Not usually relevantNot relevant to an oral product | Not usually relevantNot relevant to food use | Non-animal onlyTG 439, 492 and 437 | Case-by-caseChosen by the safety assessor | Case-by-caseFor topical and ophthalmic drugs |
| Phototoxicity | Case-by-caseIf constituents absorb light and reach the skin | Not usually relevantNot in the Redbook minimum battery | Non-animal onlyTG 432 | Case-by-caseFor products used on sun-exposed skin | Case-by-caseWhen indicated under ICH S10 |
| Cardiac safety (hERG/QT) | Case-by-caseIf constituents have known cardiac activity | Not usually relevantNot in the Redbook minimum battery | Not usually relevantNot a cosmetic endpoint | Not usually relevantNot a cosmetic endpoint | ExpectedICH S7B |
| Human studies | Case-by-caseHistory of use may supply the human evidence | Case-by-caseConcern level III, if indicated by available data | Case-by-caseNot covered by the animal-testing ban; chosen by the safety assessor | Case-by-caseChosen by the safety assessor | ExpectedPhases 1–3 under an IND (21 CFR 312) |
- Dietary supplement (new dietary ingredient). The notification must give the basis on which the notifier concluded the ingredient will reasonably be expected to be safe (21 U.S.C. 350b(a)(2); 21 CFR 190.6). No test battery is prescribed.
- Food additive (FDA Redbook). Recommended minimum tests rise with the concern level assigned from structure and exposure (FDA, Summary Table of Recommended Toxicological Testing for Additives Used in Food, June 2006).
- Cosmetic, EU. Testing finished cosmetics and cosmetic ingredients on animals is prohibited, as is marketing products or ingredients tested on animals; the marketing ban applies to all endpoints since March 11, 2013 (European Commission).
- Cosmetic, U.S. Under MoCRA the responsible person must keep records adequately substantiating product safety; the law does not prescribe a test battery.
- Drug (IND, NDA). The nonclinical program and its timing follow ICH M3(R2), with ICH S2(R1), S5(R3), S1A, S7B and S10 guidance; definitive studies are GLP studies under 21 CFR Part 58.
Four common designs, drawn from the guideline texts
About these designs
Group structure, dose levels and timing at each guideline's stated minimums. A laboratory's protocol sets the actual numbers.
The standard test catalog: what each guideline measures and in what system
GLP and screening versions
A definitive study supporting an application to FDA is a nonclinical laboratory study under 21 CFR Part 58 (GLP); screening versions run without GLP. Use the buttons to filter by family.
Show the table (27 rows)
| Family | Guideline | Title | What it detects | Test system | Exposure |
|---|---|---|---|---|---|
| Genotoxicity | TG 471 | Bacterial Reverse Mutation Test | Gene mutations | Salmonella typhimurium and E. coli strains, with and without metabolic activation (S9) | In vitro |
| Genotoxicity | TG 473 | In Vitro Mammalian Chromosomal Aberration Test | Structural chromosome damage | Cultured mammalian cells | In vitro |
| Genotoxicity | TG 487 | In Vitro Mammalian Cell Micronucleus Test | Chromosome breakage and loss (clastogens, aneugens) | Cultured mammalian cells | In vitro |
| Genotoxicity | TG 474 | Mammalian Erythrocyte Micronucleus Test | Chromosome damage in a whole animal | Rodent bone marrow or peripheral blood | In vivo follow-up |
| Genotoxicity | TG 489 | In Vivo Mammalian Alkaline Comet Assay | DNA strand breaks in target tissues such as liver or stomach | Rodent tissues | In vivo follow-up |
| Acute | TG 420 | Acute Oral Toxicity: Fixed Dose Procedure | Signs of toxicity after a single dose; hazard classification | Rodent | Single dose |
| Acute | TG 423 | Acute Oral Toxicity: Acute Toxic Class Method | Hazard classification with stepwise dosing | Rodent | Single dose |
| Acute | TG 425 | Acute Oral Toxicity: Up-and-Down Procedure | LD50 estimate with sequential dosing of single animals | Rodent | Single dose |
| Repeat dose | TG 407 | Repeated Dose 28-day Oral Toxicity Study in Rodents | Target organs, NOAEL for short-term exposure | Rodent | 28 days |
| Repeat dose | TG 408 | Repeated Dose 90-Day Oral Toxicity Study in Rodents | Target organs, NOAEL for subchronic exposure | Rodent | 90 days |
| Chronic & cancer | TG 452 | Chronic Toxicity Studies | Effects of long-term exposure | Rodent (non-rodent possible) | Normally 12 months |
| Chronic & cancer | TG 451 | Carcinogenicity Studies | Tumor incidence | Rodent | Most of the lifespan |
| Chronic & cancer | TG 453 | Combined Chronic Toxicity/Carcinogenicity Studies | Chronic effects and tumors in one design | Rodent | Chronic and lifetime arms |
| Reproductive | TG 421 | Reproduction/Developmental Toxicity Screening Test | Screen of fertility, pregnancy and early offspring | Rodent | Before mating to early lactation |
| Reproductive | TG 422 | Combined Repeated Dose Toxicity Study with the Reproduction/Developmental Toxicity Screening Test | Repeat-dose toxicity plus the reproduction screen | Rodent | Before mating to early lactation |
| Reproductive | TG 414 | Prenatal Developmental Toxicity Study | Effects on the embryo and fetus | Pregnant rodents or rabbits | During gestation |
| Reproductive | TG 443 | Extended One-Generation Reproductive Toxicity Study | Fertility and development, with optional neuro- and immunotoxicity cohorts | Rodent | Parents through offspring maturity |
| Sensitization | TG 442C | In Chemico Skin Sensitisation (peptide reactivity assays) | Key event 1: binding to skin proteins | Synthetic peptides | In chemico |
| Sensitization | TG 442D | In Vitro Skin Sensitisation: ARE-Nrf2 Luciferase Test Method | Key event 2: keratinocyte activation | Keratinocyte reporter cell lines | In vitro |
| Sensitization | TG 442E | In Vitro Skin Sensitisation: key event on activation of dendritic cells | Key event 3: dendritic-cell activation | Human cell lines (e.g., h-CLAT) | In vitro |
| Sensitization | TG 497 | Defined Approaches on Skin Sensitisation | Hazard and potency from fixed combinations of the assays above | Data-interpretation procedures, no animals | Non-animal |
| Irritation | TG 439 | In Vitro Skin Irritation: Reconstructed Human Epidermis Test Method | Skin irritation | Reconstructed human epidermis | In vitro |
| Irritation | TG 492 | Reconstructed human Cornea-like Epithelium (RhCE) test method | Identifies materials not requiring eye-irritation classification | Reconstructed corneal epithelium | In vitro |
| Irritation | TG 437 | Bovine Corneal Opacity and Permeability Test Method | Serious eye damage | Isolated bovine corneas from abattoirs | Ex vivo |
| Phototoxicity | TG 432 | In Vitro 3T3 NRU Phototoxicity Test | Toxicity that appears only with light exposure | Balb/c 3T3 fibroblasts, with and without UVA/visible light | In vitro |
| Cardiac & liver | ICH S7B | Nonclinical evaluation of the potential for delayed ventricular repolarization (QT prolongation) | hERG (IKr) channel block and QT effects | hERG-expressing cells; in vivo QT assay | Drugs |
| Cardiac & liver | No single guideline | Hepatotoxicity models | Liability for drug- or chemical-induced liver injury | Primary human hepatocytes, HepaRG cells, 3D spheroids, microphysiological systems | In vitro |
OECD Test Guidelines (Section 4: Health Effects), titles as published by OECD; ICH S7B (FDA guidance); Weaver RJ et al. Nat Rev Drug Discov. 2020;19:131–148 (liver models). Exposure lengths are the dosing periods the guidelines define; total study time is longer.
A reference point summarizes the dose–response; a margin compares it with human exposure
Source of the definitions
Definitions follow EFSA's 2022 benchmark dose guidance, which recommends the BMD approach over the NOAEL approach.
Definitions in full
- NOAEL
- No-observed-adverse-effect level: the highest dose tested in a study without evidence of an adverse effect. It depends on the doses chosen and on the study's statistical power; low-power studies tend to give higher NOAELs.
- LOAEL
- Lowest-observed-adverse-effect level: the next higher dose, at which a (statistically) significant adverse effect appears.
- BMR
- Benchmark response: a specified change in response relative to the control group. For quantal data EFSA prefers extra risk, (π(BMD) − π0)/(1 − π0); 10% is the usual value.
- BMD
- Benchmark dose: the dose, estimated from the fitted dose–response curve, at which the response reaches the BMR. It uses all the dose groups, not only the tested doses.
- BMDL, BMDU
- The lower and upper bounds of the BMD's 90% credible interval. The BMDL is normally used as the reference point; the BMDU/BMDL ratio expresses the uncertainty.
- Health-based guidance value
- An acceptable or tolerable daily intake (ADI, TDI) set by dividing the reference point by uncertainty factors; EFSA's default is 100, 10 for differences between species and 10 for variability among people.
- Margin of exposure (MOE)
- The reference point divided by the estimated human exposure, in the same units (mg/kg body weight per day). It describes a level of concern; it is not itself a pass or fail test.
- NDI terms (FDA, Section IX)
- For NDI notifications, FDA's October 2026 final guidance uses a margin of safety (NOAEL compared with the estimated daily intake, EDI) and an ADI equal to the NOAEL divided by a safety factor: usually 100 for a chronic study, 1,000 or 2,000 when shorter studies are extrapolated to chronic use. EDI/ADI ≤ 1 supports a conclusion that the proposed use will reasonably be expected to be safe. The NDIN page works through that calculation.
EFSA Scientific Committee. Guidance on the use of the benchmark dose approach in risk assessment. EFSA J. 2022;20(10):e07584 (sections 2.3 and 2.6). EFSA. EFSA J. 2005;3(10):282 and EFSA Scientific Committee. EFSA J. 2012;10(3):2578 (MOE). EFSA Scientific Committee. EFSA J. 2012;10(3):2579 (default values). FDA, NDI identity and safety guidance (October 2026), section III definitions and section IX (safety calculations).
In EFSA's worked example, the NOAEL is 3 mg/kg per day but the BMDL10 is about 2, because the NOAEL can only be a dose that was tested
About the example data
The data come from a two-year rat study used as an example in EFSA's 2022 guidance (the substance is not named). The control and the lowest dose show the same incidence, so the NOAEL is 3 mg/kg body weight per day. EFSA's model-averaged BMD analysis places the dose for a 10% extra risk at about 4.2 mg/kg per day, with a lower bound near 2.
Show data
| Dose (mg/kg bw per day) | Affected / group | Incidence (95% CI) | One-sided Fisher test vs control |
|---|---|---|---|
| 0 | 6 / 50 | 12.0% (4.5–24.3) | control |
| 3 | 6 / 50 | 12.0% (4.5–24.3) | p = 0.62 |
| 12 | 34 / 50 | 68.0% (53.3–80.5) | p < 0.001 |
| 30 | 42 / 50 | 84.0% (70.9–92.8) | p < 0.001 |
| Reference point (mg/kg bw per day) | Value | Basis |
|---|---|---|
| NOAEL / LOAEL | 3 / 12 | Derived here from the published counts (Fisher exact test, one-sided, p < 0.05) |
| BMD (90% credible interval), Laplace | 4.21 (2.15–8.31) | EFSA 2022, Appendix D, model averaging over 8 models |
| BMD (90% credible interval), MCMC | 4.68 (2.18–7.67) | EFSA 2022, Appendix D |
| BMDL–BMDU, EFSA BMD web app | 1.65–5.86 | EFSA 2022, Appendix D (PROAST 70.0) |
| BMDL, EFSA 2017 analysis | 1.5 | As reported in EFSA 2022, Appendix D (PROAST 62.3 with bootstrap) |
Why software and approximations change the BMDL
Botanical exposure problems
The same data give a NOAEL of 3 and a BMDL of 1.5 to 2.2 mg/kg per day depending on software and approximation, so a margin from either can differ about twofold. A margin also applies only to the extract studied: a delivery form that raises absorption at the same dose shrinks it. The NDI calculation is on the NDIN page.
Exposure, not dose, links an animal finding to a person
Why exposure, not dose, matters
Absorption, distribution, metabolism and excretion (ADME) decide how much of each constituent reaches blood and tissue. Toxicokinetics in the toxicity studies (ICH S3A) show what exposure produced an effect, which is what a margin should compare.
Text version: the steps, the M12 enzyme and transporter list, and the botanical caveats
- In vitro ADMESolubility, permeability, metabolic stability, plasma protein binding
- ToxicokineticsPlasma exposure (Cmax, AUC) of key constituents in the toxicity species (ICH S3A)
- Human pharmacokineticsExposure at the intended dose, under IRB oversight
- Exposure comparisonAnimal exposure at the reference point versus human exposure
Interaction screening
ICH M12 (2024) sets the drug-development approach: in vitro inhibition and induction data, static or physiologically based pharmacokinetic (PBPK) models, and clinical studies when the models cannot exclude a relevant interaction. Its scope is investigational drugs. For botanicals, the NaPDI Center's recommended approaches add the problems of variable composition and poorly known constituent pharmacokinetics.
| System | Inhibition | Induction |
|---|---|---|
| Cytochrome P450 enzymes | CYP1A2, 2B6, 2C8, 2C9, 2C19, 2D6 and 3A, reversible and time-dependent | CYP3A4, 2B6 and 1A2; CYP2C8, 2C9 and 2C19 in some cases |
| Transporters | P-gp, BCRP, OATP1B1, OATP1B3, OAT1, OAT3, OCT2, MATE1 and MATE2-K | Not applicable |
The precipitant is a mixture
Paine et al. (2018) list the recurring problems in natural product–drug interaction research: composition that varies between and within batches, limited knowledge of the pharmacokinetics of the constituents that cause interactions, and animal or in vitro models that may not extrapolate to people. A characterized test lot and constituent-level exposure data address the first two.
Paine MF, Shen DD, McCune JS. Drug Metab Dispos. 2018;46(7):1041–1045.A characterized test lot and constituent-level exposure data address the first two botanical problems.
Measured interaction magnitudes
Measured interaction magnitudes, such as St John's wort lowering indinavir exposure by 57% and goldenseal raising midazolam exposure by about 60%, are on the evidence and pharmacokinetics page.
Good Laboratory Practice is a quality system for how a study is run and recorded; it does not decide which study is needed
OECD definition of GLP
The OECD defines GLP as "a quality system concerned with the organisational process and the conditions under which" nonclinical safety studies are planned, performed, monitored, recorded, archived and reported. 21 CFR Part 58 is the U.S. equivalent.
Text version: where Part 58 applies, and by program stage
Where Part 58 applies
Part 58 covers nonclinical laboratory studies that support or are intended to support applications for research or marketing permits. Its list of covered submissions (§ 58.3(e)) includes INDs, NDAs, food and color additive petitions and certain GRAS procedures. NDI notifications are not on the list, and the October 2026 NDI guidance does not mention GLP.
Why GLP still matters for botanicals
Under the OECD's 1981 Council Decision on Mutual Acceptance of Data, data generated under OECD Test Guidelines and GLP are accepted in other member countries. A pivotal study run to GLP can therefore support dossiers in several jurisdictions and later drug development.
How EUTECNIQ would choose
Pivotal toxicology studies would be designed to GLP standards where the data may support a drug, food additive or multi-jurisdiction submission. Exploratory studies would be documented but not run under GLP. Neither choice changes what a study can show scientifically.
| Stage of a program | Quality system | What decides it |
|---|---|---|
| Screening versions of in vitro assays; dose-range finding | Documented, non-GLP | "Basic exploratory studies" are excluded from Part 58's definition of a nonclinical laboratory study (21 CFR 58.3) |
| Definitive guideline study supporting a covered FDA submission | GLP, 21 CFR Part 58 | The submission is on the § 58.3(e) list, for example an IND, NDA, or food additive petition |
| Definitive study for an NDI notification | GLP is a sponsor's choice | NDI notifications are not on the § 58.3(e) list; GLP data are reusable across OECD countries under Mutual Acceptance of Data |
| Human study | IRB review and informed consent (21 CFR Parts 50 and 56); IND if disease endpoints | See clinical research |
21 CFR 58.1 and 58.3 (eCFR, current as of October 21, 2026); OECD Council Decision C(81)30(Final). The same guideline can be run in either column; GLP status follows from how the data will be used, not from the guideline number.
- FDA requires GLP studies for NDI notifications≠what the regulation or the 2026 guidance says
- GLP (study conduct)≠GMP (manufacturing) or ISO/IEC 17025 (laboratory competence)
OECD Principles of Good Laboratory Practice (as revised in 1997), ENV/MC/CHEM(98)17; OECD Council Decision C(81)30(Final) on Mutual Acceptance of Data; 21 CFR 58.1 and 58.3 (eCFR, current as of October 21, 2026); FDA, NDI identity and safety guidance (October 2026).
Academy: NOAEL and benchmark dose · threshold of toxicological concern · human-equivalent dose calculator (Course 9)
Guideline studies run from two days to two years, so a program runs short studies before it commits to long ones
The durations below are taken from the test guideline texts themselves. Total study time is longer: acclimatization, pathology, statistics and reporting add months.
Show data
| Guideline | In-life period, as stated in the guideline | Version read |
|---|---|---|
| TG 487 In Vitro Mammalian Cell Micronucleus Test | Short treatment "for 3-6 hours", sampling "at a time equivalent to about 1.5 – 2.0 normal cell cycle lengths after the beginning of treatment" | Adopted July 19, 2016; corrected July 4, 2023 |
| TG 471 Bacterial Reverse Mutation Test | "after two or three days of incubation, revertant colonies are counted" | Adopted July 21, 1997; corrected June 26, 2020 |
| TG 420, 423 and 425, acute oral | Single dose; animals observed "for a total of 14 days" (TG 420 also: "observed for at least 14 days") | 420 and 423 adopted December 17, 2001; 425 adopted October 16, 2008, corrected June 30, 2022 |
| TG 414 Prenatal Developmental Toxicity Study | "from implantation (e.g. day 5 post mating) to the day prior to scheduled caesarean section", killed "one day prior to the expected day of delivery" | Adopted June 25, 2018 |
| TG 407 Repeated Dose 28-day Oral Toxicity Study in Rodents | "daily 7 days each week for a period of 28 days"; satellite group "for at least 14 days post treatment" | Adopted June 25, 2025 |
| TG 443 Extended One-Generation Reproductive Toxicity Study | P pre-mating "for a minimum of two weeks" and "a two-week mating period"; P males "treated for a minimum of 10 weeks"; P females 8–10 weeks; F1 "from weaning to adulthood" | Adopted June 25, 2025 |
| TG 408 Repeated Dose 90-Day Oral Toxicity Study in Rodents | "daily seven days each week for at least 90 days"; recovery period "fixed appropriately with regard to the effects observed" | Adopted June 25, 2025 |
| TG 452 Chronic Toxicity Studies | "normally for a period of 12 months"; shorter (6 or 9 months) or longer (18 or 24 months) by regulatory regime | Adopted June 25, 2018 |
| TG 451 Carcinogenicity Studies | "normally be 24 months for rodents"; for some mouse strains "a duration of 18 months may be more appropriate" | Adopted June 25, 2018 |
| TG 453 Combined Chronic Toxicity/Carcinogenicity Studies | "12 months (chronic phase) or 24 months (carcinogenicity phase)" | Adopted June 25, 2018 |
A rodent carcinogenicity study uses at least 50 animals per sex per group; for drugs, ICH S1B(R1) first asks whether it adds value
For a dietary ingredient, a carcinogenicity study is case-by-case and rare; it answers a specific concern such as a genotoxic finding or a structural alert.
Text version: the guideline text, quoted
TG 451 design
- "The objective of a long-term carcinogenicity study is to observe test animals for a major portion of their life span for the development of neoplastic lesions"
- Each dose group and concurrent control "should therefore contain at least 50 animals of each sex"
- "At least three dose levels and a concurrent control should be used"
- Duration "normally 24 months for rodents"; 18 months for some mouse strains
- TG 453 combines a 12-month chronic phase with the 24-month carcinogenicity phase in one study
ICH S1B(R1) weight of evidence
The 2022 addendum adds "specific weight of evidence (WoE) criteria that inform whether or not a 2-year rat study is likely to add value to a human carcinogenicity risk assessment". The assessment may conclude that carcinogenic potential in humans is:
- "likely, such that a 2-year rat carcinogenicity study would not add value"
- "unlikely, such that a 2-year rat carcinogenicity study would not add value"
- "uncertain, such that a 2-year rat carcinogenicity study would add value"
It also adds an exposure-ratio approach for the high dose in the rasH2-Tg mouse model, and "is not intended to replace the existing S1B Guideline".
ICH S1B(R1), Step 4 endorsement August 4, 2022; EMA/774371/2022, in effect in the EU from March 16, 2023.Academy: NTP botanical bioassays (Course 9)
The acute toxic class method assigns a hazard category with three animals per step; it says nothing about a safe daily intake
Text version: the GHS acute oral categories
| Category | Oral LD50 or acute toxicity estimate, mg/kg body weight | U.S. OSHA Hazard Communication Standard |
|---|---|---|
| Category 1 | ≤ 5 | Adopted |
| Category 2 | > 5 and ≤ 50 | Adopted |
| Category 3 | > 50 and ≤ 300 | Adopted |
| Category 4 | > 300 and ≤ 2000 | Adopted |
| Category 5 | > 2000 and ≤ 5000 | Not in the HCS table |
Categories 1–4: 29 CFR 1910.1200, Appendix A, Table A.1.1 (eCFR, current as of October 21, 2026). Category 5 range: OECD TG 423, Annex 2. TG 423: testing in the Category 5 range "is discouraged and should only be considered when there is a strong likelihood that results of such a test have a direct relevance for protecting human or animal health or the environment".
Step 1: dose 300 mg/kg to three animals. How many were found dead or humanely killed as moribund?
Choose the outcome of the current step.
- No animals dosed yet.
The rules, from TG 423 Annex 2
- "Three animals are used for each step." Fixed doses: 5, 50, 300 and 2000 mg/kg.
- Two or three dead at a dose: test the next lower dose, unless the lower dose was already cleared, which fixes the category.
- None or one dead: dose three more animals at the same level. None or one again: go to the next higher dose, unless that dose already caused deaths.
- Deaths at 5 mg/kg: Category 1. Two clear groups at 2000 mg/kg: Category 5, or "Category 5 or Unclassified" if no animal died.
Method, assumptions and limits
- What it calculates
- The sequence of doses and the GHS acute oral category that OECD TG 423's flow charts (Annex 2a–2d) assign for the outcomes you enter, and the number of animals used.
- Rules
- At a dose D: 2–3 deaths in the first group lead to the next lower dose; 0–1 lead to a second group at D. In the second group, 2–3 deaths lead to the next lower dose and 0–1 to the next higher dose. A category is assigned when deaths at D sit above a cleared lower dose (Category of D), when a cleared dose sits below a dose with deaths (category of the higher interval), at 5 mg/kg with deaths (Category 1), or after two clear groups at 2000 mg/kg (Category 5).
- Assumptions
- "Deaths" means animals found dead or humanely killed as moribund, as in the annex. The 5000 mg/kg limit test of Annex 3 is not included.
- What it does not establish
- A hazard category is not a reference point for intake. It gives no NOAEL, no target organ information from repeated exposure and no basis for a daily dose; those come from repeat-dose studies and the margin calculations above.
- Verification
- Nine outcome sequences, at least one from each annex, are checked against an independent Python implementation (tests/numerics/ref/regulatory-toxicology.py).
- References
- OECD TG 423, Acute Oral Toxicity – Acute Toxic Class Method (adopted December 17, 2001), paragraphs 9, 18–21 and Annex 2; 29 CFR 1910.1200 Appendix A.
Explanation, details and sources
OECD TG 423 doses three animals of one sex, normally females, at one of four fixed doses and moves up or down depending on how many die or must be humanely killed. The outcome is a category of the Globally Harmonized System (GHS).
Building a safety case for an ingredient?
EUTECNIQ proposes to start with a gap analysis: what is known about the material and its exposure, and which studies, if any, would close the remaining questions.