USP, ICH and COA for botanicals

Why it matters

Specifications and component testing are the most cited dietary supplement GMP failures in FDA's own inspection data

Form 483 observations under 21 CFR 111.75, by paragraph, FY2024 and FY2025FDA data

Show data

21 CFR 111.75 paragraph	What was missing	FY2024	FY2025
(a)(1)(i)	At least one identity test on a dietary ingredient before use	39	66
(a)(2)	Tests, or reliance on a COA, for other components; identity of other components	25	28
(a)(2)(ii)(A)	Supplier qualified by confirming its COA results	29	45
(a)(2)(ii)(B)	COA states methods, limits and actual results	2	15
(a)(2)(ii)(C)–(E)	Qualification records, periodic re-confirmation, quality-control approval	13	18
(c), (d)	Finished-batch verification and its documented basis	41	65
(h)	Appropriate, scientifically valid methods	22	20
(b), (f), (i)	In-process monitoring, label review, corrective action plans	2	8
Total under § 111.75		173	265

Counts are observations (citations), not firms. FDA notes that the spreadsheets do not include every Form 483 issued, because some were prepared manually. Grouping of paragraphs is EUTECNIQ's; the raw counts are FDA's.FDA, Inspection Observations, FY2025 and FY2024 Excel files, Foods sheet (page dated December 16, 2025; downloaded October 22, 2026).

Five years of § 111.75 citations by paragraph: identity, supplier qualification and batch verification grew fastestFDA data

Paragraph groups as in the table above; FDA's codes for (c), (d) and (h) changed over the period, so sub-paragraphs are summed. Click a year to sort.U.S. FDA, Inspection observations FY2021–FY2025 Excel files (Foods sheet), downloaded Oct 24, 2026; summed with src/modules/qual_six/compute/p111_years.py.

Data table: § 111.75 citations, fiscal 2021–2025

Paragraph	What was missing	FY2021	FY2022	FY2023	FY2024	FY2025
§ 111.75(a)(1)(i)	Identity test on a dietary ingredient	15	22	15	39	66
§ 111.75(a)(2)	Tests or COA reliance for other components	7	10	11	25	28
§ 111.75(a)(2)(ii)(A)	Supplier qualified by confirming COA results	13	25	18	29	45
§ 111.75(a)(2)(ii)(B)	COA states methods, limits and results	0	0	4	2	15
§ 111.75(a)(2)(ii)(C)–(E)	Qualification records, re-confirmation, QC approval	4	7	8	13	18
§ 111.75(c), (d)	Finished-batch verification and its basis	22	39	25	41	65
§ 111.75(h)	Scientifically valid methods	14	25	19	22	20
§ 111.75(b), (e), (f), (i)	In-process, product and label checks, corrective action	1	1	4	2	8
Total under § 111.75		76	129	104	173	265

Academy: how identity is established, and 16 years of FDA identity citations (Course 3)

Explanation, details and sources

In FDA's fiscal year 2025 data, inspectors recorded 1,938 observations under Part 111. Specifications (§ 111.70) drew the most, and the rules on whether specifications are met (§ 111.75) came second.

The map

Part 111 binds every supplement maker, USP binds when a product claims it, and ICH guidelines are drug guidance used as reference

The three layers answer different questions, and each has a different legal weight for a botanical ingredient or supplement sold in the United States.

Binding regulation, standards that bind when claimed, and drug guidance used as a referenceStatute, regulation and guidance scope

21 CFR Part 111 and Part 117; 21 U.S.C. 343(s)(2)(D) (eCFR and Cornell LII, read October 21–22, 2026); ICH Q6A (Step 4, October 6, 1999), Q3D(R2) (April 26, 2022) and Q7 (November 10, 2000), scope sections; USP chapter numbers from the USP Herbal Medicines Compendium general chapters list.

Text version of the three layers

Statute and regulation

21 CFR Part 111

Binding on anyone who manufactures, packages, labels or holds a dietary supplement. Operations that make only bulk ingredients for further manufacture generally fall under Part 117 instead (see the regulatory overview).

Statute, through the compendium

USP–NF

A supplement is misbranded if it "is covered by the specifications of an official compendium", "is represented as conforming to" them, "and fails to so conform" (21 U.S.C. 343(s)(2)(D)). A USP general chapter is otherwise a published method or standard that a firm may adopt.

International drug guidance

ICH Q guidelines

Harmonised guidance for medicines. Some exclude herbal products in their own text. EUTECNIQ would use their concepts (risk management, validation, stability design) where they fit, and would not describe a supplement as "ICH compliant".

Chapter numbers and titles: USP Herbal Medicines Compendium, "General Chapters" list; USP–NF DOI landing pages for ⟨2021⟩, ⟨2023⟩, ⟨2040⟩, ⟨2232⟩, ⟨2740⟩ and ⟨2750⟩; USP course page for ⟨2251⟩ (all read October 22, 2026). 21 U.S.C. 343(s); 21 CFR 111.70, 111.75 (eCFR, current as of October 21, 2026). ICH Q6A (Step 4, October 6, 1999), scope; ICH Q3D(R2) (adopted April 26, 2022), scope. Expiration dating and dissolution: FDA, Small Entity Compliance Guide for 21 CFR Part 111 (December 2010; content current as of September 20, 2018), citing 72 FR 34752.

ICH guideline PDFs (database.ich.org), cover pages and scope sections; ICH Q1 EWG Work Plan, July 10, 2026; FDA guidance pages for Q2(R2), Q14 and Q9(R1) (read October 22, 2026). FDA's botanical drug guidance is covered on pharmaceutical partnerships.

Explanation, details and sources

Quality question by USP general chapter, ICH guideline and U.S. requirement
Quality question	USP general chapters	ICH guideline (drugs)	U.S. requirement for supplements
Is it the right plant and part?	⟨561⟩ Articles of Botanical Origin; ⟨563⟩ Identification of Articles of Botanical Origin; ⟨203⟩ and ⟨1064⟩ HPTLC identification	None specific to botanicals	Identity specification for each component, § 111.70(b)(1); at least one identity test for each dietary ingredient, § 111.75(a)(1)(i); plant part on the label, 21 U.S.C. 343(s)(2)(C)
How much of what?	⟨565⟩ Botanical Extracts; individual monographs; ⟨621⟩ Chromatography	Q6A sets specifications for chemical drugs, but its text says herbal products "are similarly not covered"	Purity, strength and composition specifications, § 111.70(b)(2) and (e)
Elemental contaminants	⟨2232⟩ Elemental Contaminants in Dietary Supplements; ⟨233⟩ procedures	Q3D(R2) states it "does not apply to herbal products"	Limits on contamination that may adulterate, § 111.70(b)(3) and (e); state law such as California Proposition 65 (see legal status)
Residual solvents	⟨467⟩ Residual Solvents	Q3C (residual solvents)	Contamination limits, § 111.70(b)(3) and (e)
Microbial quality	⟨2021⟩ Microbial Enumeration Tests; ⟨2022⟩ absence of specified microorganisms; ⟨2023⟩ Microbiological Attributes of Nonsterile Nutritional and Dietary Supplements	None for supplements	Contamination limits, § 111.70(b)(3) and (e)
Undeclared drugs	⟨2251⟩ Screening for Undeclared Drugs and Drug Analogues	Not applicable	Adulteration provisions of the FD&C Act; supplier risk assessment
Does the dosage form release?	⟨2040⟩ Disintegration and Dissolution of Dietary Supplements	Q6A (drug products)	Not required by Part 111, but data should support any specification set (FDA compliance guide); a supplement must meet a disintegration specification it is represented to meet, 21 U.S.C. 343(s)(2)(E)
Does it stay in specification?	Monograph storage statements	Q1A(R2) and Q1E; a consolidated Q1 is in draft	Part 111 does not require an expiration date, but FDA says "you should have data to support any such date that you place on a product label" (see QbD and control strategy)
Are the methods fit for purpose?	⟨1225⟩ Validation, ⟨1226⟩ Verification, ⟨1224⟩ Transfer of compendial procedures	Q2(R2) and Q14	"Appropriate, scientifically valid methods", § 111.75(h)(1)
How is the operation run?	⟨2750⟩ Manufacturing Practices for Dietary Supplements; ⟨2740⟩ Manufacturing Practices for Dietary Ingredients	Q7 (APIs, including extracted and herbal-extract APIs), Q9(R1), Q10	21 CFR Part 111; Part 117 for ingredient-only operations

ICH quality guidelines with status and relevance to botanicals
ICH guideline	Status as of October 2026	What it says about botanicals
Q1A(R2) Stability Testing of New Drug Substances and Products; Q1E Evaluation for Stability Data	Final Step 4, February 6, 2003. A consolidated Q1 reached Step 2 (draft) in April 2025; ICH's July 2026 work plan targets Step 3 sign-off and Assembly review for Step 4 in December 2026	Drug guidance; the storage conditions and extrapolation rules are a common design reference
Q2(R2) Validation of Analytical Procedures; Q14 Analytical Procedure Development	Final FDA final guidance, March 2024 (docket FDA-2022-D-1503)	Method concepts apply to any analyte; see analytical method development
Q3D(R2) Guideline for Elemental Impurities	Final Adopted April 26, 2022	"This guideline does not apply to herbal products"
Q6A Specifications for new drug substances and products (chemical)	Final Step 4, October 6, 1999	"herbal products and crude products of animal or plant origin are similarly not covered"
Q7 Good Manufacturing Practice Guide for Active Pharmaceutical Ingredients	Final Step 4, November 10, 2000	Covers APIs made "by extraction" or "by recovery from natural sources"; its Table 1 lists "Herbal extracts used as API"
Q8(R2) Pharmaceutical Development; Q9(R1) Quality Risk Management; Q10 Pharmaceutical Quality System; Q12 Lifecycle Management	Final Q8(R2) August 2009; Q9(R1) FDA final guidance (docket FDA-2022-D-0705); Q10 June 4, 2008; Q12 November 20, 2019	Framework concepts; see QbD and control strategy
Q11 Development and Manufacture of Drug Substances	Final Step 4, May 1, 2012	Drug substances; relevant to botanical drug partners

21 CFR 111.70 and 111.75

A dietary ingredient needs its own identity test before use; other specifications may rest on a qualified supplier's COA

This reading is EUTECNIQ's summary of the regulation, not legal advice. The red labels are FDA's FY2025 citation counts for each paragraph.

Four things a manufacturer releases, and what § 111.75 requires before each is usedRegulation

21 CFR 111.70, 111.75 and 111.80 (eCFR, current as of October 21, 2026; source 72 FR 34942, June 25, 2007, as amended). Citation counts: FDA Inspection Observations, FY2025, Foods sheet.

Step through one material, with the regulation quoted at each answer

1. What are you releasing?

Part 111 sets different steps for components, finished batches and product received only for packaging or labeling.

A dietary ingredient Another component (excipient, capsule shell)A finished batch Product received for packaging or labeling
2. Has FDA exempted you from identity testing for this ingredient by petition?

§ 111.75(a)(1)(ii): a petition under 21 CFR 10.30 must show "no material diminution of assurance, compared to the assurance provided by 100 percent identity testing".

No Yes, a petition was granted
§ 111.75(a)(1)(i) Test the identity of every dietary ingredient before use

"Conduct at least one appropriate test or examination to verify the identity of any component that is a dietary ingredient". A supplier's COA cannot replace this test. The test must be one of the listed kinds, "gross organoleptic", "macroscopic", "microscopic", "chemical", or "other scientifically valid methods" (§ 111.75(h)(2)), and appropriate for the material.

Next: the ingredient's other specifications
§ 111.75(a)(1)(ii) Test under the terms FDA set

"If FDA grants the petition, you must conduct the tests and examinations for the dietary ingredient … under the terms specified by FDA when the petition is granted".

Next: the ingredient's other specifications
3. For the remaining specifications, will you rely on the supplier's COA?

§ 111.75(a)(2) covers confirming the identity of other components and whether "other applicable component specifications established in accordance with § 111.70(b) are met": purity, strength, composition and contaminant limits.

No, we will test Yes, we want to rely on the COA
§ 111.75(a)(2)(i) Conduct appropriate tests or examinations

Methods must be "appropriate, scientifically valid methods" (§ 111.75(h)(1)). Representative samples of "each unique shipment, and of each unique lot within each unique shipment" are required (§ 111.80(a)). The sampling arithmetic is below.
4. Are all five conditions of § 111.75(a)(2)(ii) met?
1. (A) "You first qualify the supplier by establishing the reliability of the supplier's certificate of analysis through confirmation of the results of the supplier's tests or examinations"
2. (B) "The certificate of analysis includes a description of the test or examination method(s) used, limits of the test or examinations, and actual results"
3. (C) "You maintain documentation of how you qualified the supplier"
4. (D) "You periodically re-confirm the supplier's certificate of analysis"
5. (E) "Your quality control personnel review and approve the documentation setting forth the basis for qualification (and re-qualification)"
Yes, all five Not all
§ 111.75(a)(2)(ii) The COA may be relied on for those specifications

Reliance holds only while it is maintained: re-confirm periodically with independent results, and re-qualify when the supplier, site, process or method changes. In FY2025, 45 observations cited § 111.75(a)(2)(ii)(A) for suppliers that were never qualified.
§ 111.75(a)(2) The COA cannot be relied on yet

Test the material, or first qualify the supplier. A COA that lacks methods, limits or actual results fails condition (B) whatever its numbers show; FDA cited that gap 15 times in FY2025.

Test instead
5. Can every product specification be verified on the finished batch?

§ 111.75(c): "For a subset of finished dietary supplement batches that you identify through a sound statistical sampling plan (or for every finished batch), you must verify that your finished batch … meets product specifications".

Yes No valid finished-batch method exists for one specification
§ 111.75(c)(1)–(4) Select, test and document

Select specifications that "would verify that the production and process control system is producing a dietary supplement that meets all product specifications", test them, document why that choice is sufficient, and have quality control "review and approve" that documentation.
§ 111.75(d)(1) Exempt it, and document why

An exemption needs a determination that the selected tests cannot verify that specification and "there is no scientifically valid method for testing or examining such exempted product specification at the finished batch stage", plus documentation of how component and in-process controls ensure it is met. Quality control must approve (§ 111.75(d)(2)).
§ 111.75(e) and § 111.70(f) Examine and document received product

Specifications must "provide sufficient assurance that the product you receive is adequately identified and is consistent with your purchase order". Before packaging or labeling, "visually examine the product and have documentation to determine whether the specifications … are met".

Reading a COA

A reviewer checks who issued the certificate, whether each method can see the known adulterants, and whether the results are numbers

Annotated specimen certificate of analysisSpecimen, not a real lot

Certificate of analysis SpecimenCoA no. SPEC-0001 · page 1 of 1

1 Issued by: Example Analytical Laboratory (fictitious), testing for the manufacturerExample Trading Co. (fictitious), on its own letterhead Original manufacturer: Example Extracts Co. (fictitious), site address statednot stated; no original certificate attached 2 Product: Ginkgo leaf dry extract, Ginkgo biloba L., leaf illustrative Batch SPECIMEN-001 · manufactured 2026-07 · retest 2028-07

Specimen test results. All values are illustrative.
Test	Method	Acceptance criterion	Result
3 Identity	HPTLC vs botanical reference material, USP ⟨203⟩UV/Vis spectrum	Matches reference fingerprint	Conforms; plate image on fileConforms
4 Flavonol glycosides	HPLC-UV after acid hydrolysis"Total flavonoids", UV	22.0–27.0% buyer spec	24.6%26.1%
5 Free quercetin, before hydrolysis	HPLC-UV	NMT 0.5%	0.1%not tested
Rutin, before hydrolysis	HPLC-UV	NMT 4%	1.2%not tested
Quercetin : kaempferol : isorhamnetin	HPLC-UV after hydrolysis	4:4:1 to 6:5:1	5.0 : 4.4 : 1not reported
Terpene lactones	HPLC	≥ 6.0% buyer spec	6.4%Complies
6 Elemental contaminants	ICP-MS, USP ⟨233⟩"Heavy metals", USP ⟨231⟩	As (inorganic) ≤ 1.5; Cd ≤ 0.5; Pb ≤ 0.5; Hg ≤ 1.5 µg/g≤ 10 ppm	0.08; 0.02; 0.11; < 0.01 µg/g< 10 ppm
Residual ethanol	GC headspace, USP ⟨467⟩	≤ 5,000 ppm buyer spec	410 ppmComplies
Microbial counts	USP ⟨2021⟩, ⟨2022⟩	TAMC ≤ 10⁴ cfu/g; E. coli and Salmonella absent in 10 g buyer spec	TAMC 3.0 × 10² cfu/g; absent; absentComplies
Loss on drying	USP ⟨731⟩	≤ 5.0% buyer spec	2.8%

Who issued it. The laboratory's name and address, and the original manufacturer. If a trader reissues a certificate, WHO's model says it "should show the name and address of the laboratory that performed the tests and the name and address of the original manufacturer" and attach the original.
Which material. Latin binomial and plant part, batch number matching the containers and the purchase order, manufacture and retest dates. Supplement labels must name the plant part (21 U.S.C. 343(s)(2)(C)).
Identity by a method that can tell species apart. For an extract, a chromatographic fingerprint against authenticated reference material. Part 111 also requires the buyer's own identity test for every dietary ingredient lot.
Assay by a specific method. Spectrophotometric "total" assays read higher than HPLC and cannot distinguish the declared plant from cheaper material with the same chromophore.
Checks aimed at the known adulteration route. For ginkgo: free quercetin and rutin measured before hydrolysis, and the flavonol ratio after it.
Element-specific results against USP ⟨2232⟩ limits, not a single colorimetric "heavy metals" total.
Numbers, limits and methods for every test. "Complies" alone does not meet § 111.75(a)(2)(ii)(B), which asks for "actual results".
Approval. Name, signature and date of the person authorized to release the certificate; subcontracted results identified as such.

Every name, number and date on the specimen is illustrative. "Buyer spec" marks acceptance criteria a buyer would set. Rutin NMT 4%, quercetin NMT 0.5% and the 4:4:1 to 6:5:1 ratio are the USP and American Herbal Pharmacopoeia values reported by Gafner et al. (2023); confirm against the current USP monograph. Element limits are USP ⟨2232⟩ Table 2 values for individual components.WHO Technical Report Series No. 1010 (2018), Annex 4, Model certificate of analysis; 21 CFR 111.75(a)(2)(ii)(B); Gafner S, et al. J Nat Prod. 2023;86(2):460–472; USP ⟨2232⟩ as published in USP 38–NF 33.

Eight documented red flags, what each hides, and what to ask for instead

Each row is drawn from a regulation, WHO's model certificate or a peer-reviewed adulteration study.

Gafner S, Blumenthal M, Foster S, Cardellina JH, Khan IA, Upton R. J Nat Prod. 2023;86(2):460–472, doi:10.1021/acs.jnatprod.2c00929. Wohlmuth H, et al. Phytomedicine. 2014;21(6):912–918. Villani TS, et al. Food Chem. 2015;170:271–280. Gafner S, Orhan N, Kahraman Ç, Blumenthal M. Pharm Biol. 2026;64(1):87–107, doi:10.1080/13880209.2025.2606229. WHO TRS No. 1010, Annex 4. USP, "⟨232⟩ Elemental Impurities—Limits and ⟨233⟩ Elemental Impurities—Procedures" revision history page.

Explanation, details and sources

The specimen below follows WHO's model certificate of analysis and Part 111's content requirement. Switch to the red-flag version to see the same lot reported in ways that published adulteration studies warn about.

Red flag on a COA	Why it matters	Ask for	Source
Composition given only as a UV/Vis "total"	Spectrophotometric results "are sometimes the sole data on the chemical composition" on a supplier COA, and read high because non-target compounds absorb. Silymarin: 30–65% by HPLC-UV versus 65–80% by spectrophotometric tests.	A specific chromatographic assay and the method name next to each result	Gafner et al. 2023
"Complies" with no number, method or limit	The COA cannot be relied on: § 111.75(a)(2)(ii)(B) requires methods, limits and "actual results". FDA cited this 15 times in FY2025.	Numerical results for every quantitative test	21 CFR 111.75; FDA FY2025 data
Trader's certificate, original manufacturer not named	The tested lot cannot be traced to a site, and the buyer cannot judge which laboratory produced the data.	The original manufacturer's COA attached, and the testing laboratory named	WHO TRS 1010, Annex 4
Ginkgo "flavonol glycosides" from a post-hydrolysis assay only	Spiking with rutin, quercetin or flavonol-rich extracts has been "reported in over two dozen studies"; a hydrolysis assay counts added aglycones as if they were ginkgo glycosides. In 2014, three of eight retail products tested showed "compelling evidence of adulteration".	Rutin and free quercetin before hydrolysis; quercetin : kaempferol : isorhamnetin ratio	Gafner et al. 2023; Wohlmuth et al. 2014
St John's wort "hypericins" by visible absorbance	Food dyes are added so the extract appears to contain its labeled hypericin content; brilliant blue absorbs at 590 nm, "the same wavelength where hypericin is measured".	HPTLC or HPLC-UV/Vis profile	Gafner et al. 2023
Grape seed or cranberry "proanthocyanidins" by a colorimetric assay	The assays cannot tell proanthocyanidin sources apart. Nine of 21 commercial grape seed products contained peanut skin extract.	Chromatographic fingerprint against authentic extracts	Gafner et al. 2023; Villani et al. 2015
"Heavy metals ≤ 10 ppm, USP ⟨231⟩"	USP stated that ⟨231⟩ "will remain in effect until January 1, 2018", replaced by the element-specific approaches of ⟨232⟩/⟨2232⟩. A 10 ppm total says nothing about lead against a 0.5 µg/g component limit.	Arsenic, cadmium, lead and mercury by ICP-MS or equivalent	USP ⟨232⟩/⟨233⟩ revision page; USP ⟨2232⟩
Turmeric "curcuminoids 95%" with no authenticity test	Across 48 publications covering 2,235 commercial samples, the overall adulteration rate was 20.0%. Synthetic curcumin can deceive HPTLC, HPLC, IR, NIR, Raman and NMR methods.	Tests for synthetic curcumin and dyes; supplier history	Gafner et al. 2026 (scoping review); Gafner et al. 2023, Table 1

Adulteration risk

For many high-value extracts, the adulterant is chosen to pass the very assay most buyers order

Show the table (10 rows): Selected adulteration schemes and the methods at risk of being deceived (reproduced from G

Selected adulteration schemes and the methods at risk of being deceived (reproduced from Gafner et al. 2023, Table 1)
Botanical (part)	Analyte	Documented adulterant	Methods at risk of being deceived
Black cohosh (root)	Triterpenoids	Other Actaea species	HPTLC, HPLC-UV/Vis
Turmeric (root)	Curcumin	Synthetic curcumin	HPTLC, HPLC-UV/Vis, HPLC-MS, IR, NIR, Raman, NMR
Ginkgo (leaf)	Flavonoids	Rutin-rich extracts	HPTLC, HPLC-UV/Vis, HPLC-MS, IR, NIR, Raman, NMR
Goldenseal (root)	Berberine	Coptis, Oregon grape or barberry extract	HPTLC, HPLC-UV/Vis
Asian ginseng	Ginsenosides	Ginseng leaf extracts, Panax quinquefolius extract	HPTLC, HPLC-UV/Vis, HPLC-MS
Rhodiola (root)	Cinnamyl alcohol glycosides, salidroside	Other Rhodiola species	HPTLC, HPLC-UV/Vis
European elder (berry)	Anthocyanins	Black rice extract	UV/Vis, HPTLC, HPLC-UV/Vis
Saw palmetto (fruit)	Fatty acids	Vegetable oils	HPTLC, GC-FID, GC-MS
Milk thistle (fruit)	Flavonolignans	Extracts from exhausted milk thistle seed	HPTLC
Bilberry (fruit)	Anthocyanins	Black rice, blueberry or mulberry extract	UV/Vis

Reproduced Selected rows of Table 1, "Adulteration Schemes for Important Medicinal Plant Extracts", in Gafner S, et al. Botanical ingredient forensics. J Nat Prod. 2023;86(2):460–472 (open access, PMC9972475). Pharmacopeial HPLC identification alone may also miss adulterated bilberry extracts: Govindaraghavan S. Fitoterapia. 2014;99:124–138.

Explanation, details and sources

A risk screen starts from what has been documented for the botanical. The table reproduces part of a peer-reviewed review based on the ABC-AHP-NCNPR Botanical Adulterants Prevention Program's publications.

Supplier evaluation

No single document qualifies a supplier, so EUTECNIQ would score seven lines of evidence and let three pass/fail gates override any score

Weighted score with gates (management proposal)S = Σ w_i s_i5 Σ w_i × 100 if every gate passes

Example supplier

Gates (any failure means the supplier is not approved) Identity of each dietary ingredient lot confirmed by the buyer's own test Lawful origin: no OFAC-prohibited origin; CITES and Lacey documents available COA states methods, limits and actual results

Documents

Audit

Testing

Adulteration risk

Origin and legality

Continuity

Performance

Weighted score, %88.7

Sum of weights23

Decision (proposal)Approve

All gates pass. A score of 80% or more would support approval with routine re-confirmation.

Scores on the seven lines of evidenceInvented example; weights are a proposal

Rings at scores 1 to 5; the amber ring marks 4. Line width of each spoke is proportional to its weight. When a gate fails, the polygon turns red and the score is reported but not used.

Method, assumptions and limits

What it calculates: A weighted mean of seven scores, rescaled to 0–100%, and a decision band. Three gates are evaluated separately; if any gate fails the decision is "Not approved" whatever the score.
Equation: S = Σ(w_i s_i) ÷ (5 Σ w_i) × 100, with weights w_i and scores s_i integers from 0 to 5. If all weights are 0, S is reported as 0.
Bands (proposal): S ≥ 80%: approve with routine periodic re-confirmation. 60% ≤ S < 80%: conditional, with tightened testing and a shorter re-qualification interval. S < 60%: do not approve. These thresholds are EUTECNIQ's proposal, not a regulatory standard.
Assumptions: Scores are judged on a written rubric by the quality unit; the lines of evidence are treated as independent and compensatory above the gates. The examples are invented to show how gates and weights interact.
What it does not establish: It is not a supplier qualification under § 111.75(a)(2)(ii), which requires confirming the supplier's test results. It does not replace an audit, testing, or legal review of origin and sanctions.
Verification: Six cases are checked against an independent Python implementation (tests/numerics/ref/regulatory-quality-standards.py).
References: 21 CFR 111.75 and 117.410 (eCFR, October 21, 2026); Gafner et al., J Nat Prod 2023.

Explanation, details and sources

A weighted average can hide a fatal gap: a supplier with excellent paperwork and an unlawful origin still scores well. The proposed scorecard therefore separates gates from scored criteria. Weights and bands are a management proposal; the example suppliers are invented.

Line of evidence	What is examined	Basis
1. Documents and COA reliability	Specifications, COA content, change notification, regulatory status of the ingredient	Regulation 21 CFR 111.75(a)(2)(ii)(B)–(E)
2. Audit	Onsite or remote audit of the manufacturing site, not the broker's office	Regulation Onsite audits are named verification activities in 21 CFR 117.410(b) and, for supplement importers, 21 CFR 1.511(c)
3. Independent orthogonal testing	The buyer's own results by a different separation or detection principle, compared with the supplier's	Regulation Confirmation of the supplier's results, § 111.75(a)(2)(ii)(A); orthogonality is EUTECNIQ practice
4. Adulteration-risk screen	Known adulterants for the species, and whether the specification includes tests that detect them	Published evidence BAPP publications; Gafner et al. 2023
5. Origin, sanctions and protected species	Country of harvest and processing, CITES listing, Lacey Act declaration data, OFAC-prohibited origins	Statute and regulation 31 CFR 560.201; 16 U.S.C. 3372; 50 CFR 23.13 (see legal status)
6. Supply continuity	Harvest seasonality, number of sites, inventory, dependence on one region	Proposal Commercial criterion
7. Performance history	Out-of-specification results, deviations, complaints, outcome of periodic re-confirmation	Regulation Periodic re-confirmation, § 111.75(a)(2)(ii)(D)

Sampling arithmetic

Sampling √N + 1 containers finds a 10% adulterated consignment of 40 drums only about half the time

Hypergeometric detection probability (perfect per-container test), and with sensitivity sP_detect = 1 − C(N − K, n)C(N, n) P_detect(s) = 1 − Σ_k C(K, k) C(N − K, n − k)C(N, n) (1 − s)^k

Containers in the consignment, N40 Adulterated containers, K4 (10%) Containers sampled and tested individually, n7 Chance the test flags an adulterated sample, s Target detection probability

Detection probability, %55.2

Containers needed for target21

WHO n plan (1 + √N)7

WHO p plan (0.4 √N)3

WHO r plan (1.5 √N)10

Detection with r plan, %70.0

Sampling 7 of 40 containers misses all 4 adulterated containers 44.8% of the time.

Detection probability over containers sampled and adulterated fractionComputed from the stated equation

Surface for the consignment size and test sensitivity chosen. Amber: combinations that reach the target. Points: your sample and the WHO n and r plans at the chosen fraction. For drawing, K = fraction × N is allowed between integers; readouts use whole containers.

Show data: detection probability for a 40-container consignment, perfect test

Adulterated containers K (fraction)	n = 3 (p plan)	n = 7 (n plan)	n = 10 (r plan)	n = 21	Containers for 95%
1 (2.5%)	0.075	0.175	0.250	0.525	38
2 (5%)	0.146	0.323	0.442	0.781	31
4 (10%)	0.277	0.552	0.700	0.958	21
8 (20%)	0.498	0.819	0.924	0.999	12

Computed with scipy.stats.hypergeom. Under WHO's p and r plans every container is opened and identity-tested; the n and r counts here are the containers tested individually in full.

Method, assumptions and limits

What it calculates: The probability that a simple random sample of n containers from a consignment of N contains at least one of K adulterated containers and that the test flags it; the smallest n that reaches the target; and WHO's n, p and r plan sizes.
Equations: Hypergeometric: P(k) = C(K,k) C(N−K, n−k) / C(N,n). P_detect = 1 − Σ_k P(k)(1 − s)^k; with s = 1 this reduces to 1 − C(N−K, n)/C(N, n). WHO TRS 929 Annex 4: n = 1 + √N "obtained by simple rounding", every container if N ≤ 4; p = 0.4√N and r = 1.5√N, "rounding up to the next highest integer".
Variables: N: containers in the consignment, 5–200. K: adulterated containers, 0–N. n: containers sampled and tested individually, 1–N. s: probability that the test flags one adulterated container, 50–100%.
Assumptions: Containers are either adulterated or not; samples are drawn at random without replacement; each sampled container is tested separately; test results are independent. Compositing samples, as WHO's n plan does, dilutes an adulterated portion and can push it below the method's detection limit, which this model does not include.
Where WHO's table differs: WHO's Table 1 lists a few values that differ from its own formulas at boundaries (for the n plan at N = 3 and 13; for the r plan at N = 2 and 22). The calculator applies the formulas as written.
What it does not establish: It does not measure within-container heterogeneity, which WHO treats as intrinsic for "partially processed natural products … herbal (dried plants and their parts)". It does not replace the § 111.75(a)(1)(i) identity test of each dietary ingredient or a "sound statistical sampling plan" for finished batches (§ 111.75(c)).
Verification: Seven cases are checked against scipy.stats.hypergeom in tests/numerics/ref/regulatory-quality-standards.py, including WHO's own worked example (N = 40 gives n = 7, p = 3, r = 10).
References: WHO Expert Committee on Specifications for Pharmaceutical Preparations, TRS No. 929 (2005), Annex 4, sections 5.1.1–5.1.3, Table 1 and Appendix 5.

WHO, "WHO guidelines for sampling of pharmaceutical products and related materials", TRS No. 929, Annex 4 (2005). Plant extracts "may be considered uniform unless there are signs to the contrary", while partially processed herbal materials "should be treated as intrinsically non-uniform"; the r plan "may also be used for herbal medicinal products used as starting materials".

Explanation, details and sources

WHO limits its √N + 1 rule (the "n plan") to uniform material from a recognized source and says it "is not statistically based". The hypergeometric distribution gives the actual chance that a random sample includes at least one adulterated container.

Elemental contaminants

USP ⟨2232⟩ caps lead at 5 µg per day from a supplement; Proposition 65's reproductive safe harbor is ten times lower

USP ⟨2232⟩ scope sentences: USP–NF DOI landing page (official 2022 text, preview). Table values as published in USP 38–NF 33, ⟨2232⟩ (PF 39(5)); component limits "are based on a maximum daily intake of 10 g". The current official values could not be read without a USP–NF subscription and should be confirmed there. Lead safe-harbor levels: OEHHA, Proposition 65 chemical page for lead (modified January 5, 2025).

USP ⟨2232⟩ daily-intake optiondaily exposure = c × m compliant if ≤ PDE

Element Measured concentration, c0.3 µg/g Maximum labeled daily intake, m2 g/day

Daily exposure, µg/day0.6

Share of USP PDE, %12

Highest concentration at this intake, µg/g2.5

Share of Prop 65 lead MADL, %120

Within the USP ⟨2232⟩ PDE, but above California's Proposition 65 safe harbor for lead (0.5 µg/day).

Daily exposure against the limitsCalculator; values you set

Log scale. USP's daily-intake option multiplies the measured amount per serving by the maximum servings per day in the labeling; this calculator uses concentration × daily mass, which is equivalent.

Method, assumptions and limits

What it calculates: Daily exposure to one element from a supplement, its share of the USP ⟨2232⟩ permitted daily exposure, the highest concentration that keeps exposure at the PDE for this intake, and for lead its share of California's Proposition 65 maximum allowable dose level (MADL).
Equations: E = c × m (µg/day). Share of PDE = 100 E / PDE. c_max = PDE / m. Share of MADL = 100 E / 0.5.
Variables: c: measured concentration, 0.01–10 µg/g (equal to mg/kg or ppm by mass). m: maximum daily intake recommended in the labeling, 0.25–20 g/day. PDE values from ⟨2232⟩ Table 1.
Assumptions: For inorganic arsenic, ⟨2232⟩ allows a total-arsenic result as a conservative estimate; if it exceeds the limit, speciation is required. Methylmercury testing is unnecessary when total mercury is below the methylmercury limit.
What it does not establish: A Proposition 65 warning decision depends on exposure assessment rules, other chemicals and counsel's advice; the MADL comparison is a screening figure. It does not address intake from other foods.
Verification: Six cases are checked against an independent Python implementation (tests/numerics/ref/regulatory-quality-standards.py).
References: USP ⟨2232⟩ (USP 38–NF 33 text); OEHHA, Proposition 65, lead (safe harbor levels).

Academy: how a soil metal becomes a daily dose (Course 5) · limits compared across frameworks (Course 6)

Explanation, details and sources

USP ⟨2232⟩ limits arsenic, cadmium, lead and mercury in finished supplements "labeled as conforming to USP or NF standards". It "is not intended to set limits for dietary ingredients", which are set in monographs. Buyers often adopt its component limits in ingredient specifications.

USP 2232 permitted daily exposures and component limits, and the Proposition 65 lead safe-harbor level
Element	PDE, µg/day (Table 1)	Individual component limit, µg/g (Table 2)	Proposition 65 safe harbor, µg/day
Arsenic (inorganic)	15	1.5	not compared here
Cadmium	5	0.5	not compared here
Lead	5	0.5	0.5 (MADL, reproductive); 15 (NSRL, cancer)
Mercury (total)	15	1.5	not compared here
Methylmercury (as Hg)	2	0.2	not compared here

Decision rules

A result near the limit passes or fails depending on the decision rule agreed before testing

Guard band and specific risk (normal, k = 2)AL = TL − r·U
P(true > TL) = Φ((y − TL)/(U/2))

Measured result, % of the limit95.0% Expanded uncertainty U (k = 2), % of the limit10.0% Decision rule

Acceptance limit90.0%

Binary decisionFail

Non-binary statementConditional pass

Chance the true value exceeds the limit15.9%

Where the true value could lie, against the tolerance and acceptance limitsCalculator, normal model

ILAC-G8:09/2019, Guidelines on Decision Rules and Statements of Conformity, sections 4.1–4.2 and Table 1; JCGM 106:2012; ISO/IEC 17025:2017 (decision rule, 3.7; statements of conformity, 7.8.6).

Method, assumptions and limits

What it calculates: For an upper tolerance limit TL, the acceptance limit AL = TL − w with guard band w = rU, the binary decision (pass if the result is at or below AL), ILAC G8's non-binary statement (pass, conditional pass, conditional fail, fail), and the specific risk.
Equations: Standard uncertainty u = U/k with k = 2. Probability that the true value exceeds TL: Φ((y − TL)/u). For a passed item this is the probability of false acceptance; for a failed item the calculator shows the chance that it actually conforms.
Benchmarks: ILAC G8 Table 1 (single-sided specification, normal distribution): w = 3U, < 1 ppm false accept; 1.5U, < 0.16%; U, < 2.5%; 0.83U (ISO 14253-1:2017), < 5%; simple acceptance, < 50%.
Valid range: Result 50–150% of the limit; U 0.5–40% of the limit. Normal distribution assumed; results near a detection limit or with skewed uncertainty need other treatment.
Does not establish: The uncertainty of any real method. ISO/IEC 17025 requires the laboratory to evaluate measurement uncertainty and to apply a documented decision rule agreed with the customer when it states conformity (ILAC G8 §5.2).
Verification: Unit tests reproduce the Table 1 bounds; independent reference cases in tests/numerics/ref/regulatory-quality-standards.py (scipy.stats.norm).

With the ILAC G8 guard band w = U, a result accepted at the acceptance limit carries less than 2.5% risk of being out of specification.

EUTECNIQ's proposed role

EUTECNIQ would write specifications, qualify suppliers on evidence, and test where a COA cannot be trusted

Show the table (5 rows)

Activity	EUTECNIQ (proposed)	Qualified external partner
Specifications	Component and finished-product specifications tied to known adulteration routes	Client's quality unit approves
COA review	Structured review against § 111.75(a)(2)(ii)(B) and the red flags above	Client's quality unit approves any reliance on a COA
Confirmatory and orthogonal testing	Planned in-house methods once a laboratory exists	ISO/IEC 17025-accredited laboratory for tests outside EUTECNIQ's scope
Supplier audits	Audit plan, risk ranking, follow-up	Local auditors for remote sites
Origin and sanctions screening	Document collection and first-pass screening	Trade and sanctions counsel decides

A supplier's COA≠the buyer's identity test of a dietary ingredient
"Complies"≠an actual result
A high UV/Vis assay value≠a high content of the declared plant's compounds
ICH quality guidance≠a requirement for dietary supplements
Sampling √N + 1 containers≠a statistically designed plan

Reviewing a supplier or a certificate?

EUTECNIQ proposes to start from the specification and the botanical's documented adulteration routes, then decide what to test and what can be relied on.

Discuss a quality program Legal status and counsel

Specifications and component testing are the most cited dietary supplement GMP failures in FDA's own inspection data

Part 111 binds every supplement maker, USP binds when a product claims it, and ICH guidelines are drug guidance used as reference

21 CFR Part 111

USP–NF

ICH Q guidelines

A dietary ingredient needs its own identity test before use; other specifications may rest on a qualified supplier's COA

1. What are you releasing?

2. Has FDA exempted you from identity testing for this ingredient by petition?

§ 111.75(a)(1)(i) Test the identity of every dietary ingredient before use

§ 111.75(a)(1)(ii) Test under the terms FDA set

3. For the remaining specifications, will you rely on the supplier's COA?

§ 111.75(a)(2)(i) Conduct appropriate tests or examinations

4. Are all five conditions of § 111.75(a)(2)(ii) met?

§ 111.75(a)(2)(ii) The COA may be relied on for those specifications

§ 111.75(a)(2) The COA cannot be relied on yet

5. Can every product specification be verified on the finished batch?

§ 111.75(c)(1)–(4) Select, test and document

§ 111.75(d)(1) Exempt it, and document why

§ 111.75(e) and § 111.70(f) Examine and document received product

A reviewer checks who issued the certificate, whether each method can see the known adulterants, and whether the results are numbers

Eight documented red flags, what each hides, and what to ask for instead

For many high-value extracts, the adulterant is chosen to pass the very assay most buyers order

No single document qualifies a supplier, so EUTECNIQ would score seven lines of evidence and let three pass/fail gates override any score

Sampling √N + 1 containers finds a 10% adulterated consignment of 40 drums only about half the time

USP ⟨2232⟩ caps lead at 5 µg per day from a supplement; Proposition 65's reproductive safe harbor is ten times lower

A result near the limit passes or fails depending on the decision rule agreed before testing

EUTECNIQ would write specifications, qualify suppliers on evidence, and test where a COA cannot be trusted

Reviewing a supplier or a certificate?