Platform · Demonstration 3 · The consistency problem
A passing assay does not make a consistent ingredient
A published study measured 11 constituents and a chromatographic fingerprint in 51 batches of ginkgo tablets and capsules from 25 manufacturers. Of the 38 batches inside a marker range, 14 fail the authors' fingerprint threshold.
This page lets you set the specification, add tests one at a time and see which batches survive and how much of each constituent a serving would deliver.
Proposed capabilityCompany in formation (what exists today). Every value here is a published measurement by Liu et al. (2022); the specification, decision rules, serving size and blends are hypothetical.
Marker content against fingerprint similarity, 51 batches
Horizontal: sum of six flavonol glycosides, mg per 40 mg extract. Vertical: similarity of each batch's UPLC fingerprint to the common pattern (1.0 = identical).
In brief · Outcomes
Of 51 batches, 38 pass the marker assay and 3 pass every check; each failure opens an investigation, not a finding of adulteration
The checks below are applied in order at the default settings: a hypothetical specification of 5.0–7.0 mg six-glycoside sum per 40 mg extract, then the further tests one at a time. The statistics and the data behind each count follow in sections 02 and 03.
| Check, in the order applied | Fails this check on its own | Still accepted after adding it |
|---|---|---|
| Targeted assay, six-glycoside sum within 5.0–7.0 mg | 13 of 51 | 38 of 51 |
| Free quercetin at most 0.2 mg | 1 (L2, 0.272 mg), already outside the range | 38 |
| Rutin at most 25% of the six glycosides | 12, including A1 at similarity 1.00 | 30 (8 removed) |
| Fingerprint similarity at or above 0.85 | 19; 14 of them inside the marker range | 18 (12 more removed) |
| Supplier qualification | Of the 12 manufacturers with several batches, only G has no batch failing an authenticity check (quercetin, rutin ratio or fingerprint) | 3, all from G; 6 held for lack of history; 42 rejected |
What would trigger an investigation
- Any failed check. A failed check marks a batch for investigation; it does not prove adulteration.
- A break in a source's own pattern. Within manufacturer K, batches K3 and K4 contain 2.2 times the rutin of K1, K2 and K5 while the other five glycosides differ by at most 16%: a pattern worth an investigation, not a finding.
- A source with one batch on record. Its batch is held, not accepted, until the source has a history.
- The disposition that follows. Under Part 111 a component that fails its identity specification must be rejected (21 CFR 111.77(b)); other failures may be corrected only by a treatment, adjustment or reprocessing that quality control approves after a material review (111.77(a), 111.90). EUTECNIQ proposes to write the fingerprint and authenticity checks into identity. Proposed capability
Measured data Hypothetical example Counts computed from Liu et al. 2022, Table 2. The marker range, the 25% rutin limit and the supplier rule are hypothetical; the 0.2 mg quercetin limit and the 0.85 similarity cut are as stated by the authors. Table 3 of the paper lists 17 batches below 0.85, Table 2 gives 19, and this page uses Table 2 (section 02). Change any setting in section 03; each test's limits are in section 06.
01 · What a specification can see
Marker, class, fingerprint and activity are four different measurements
A botanical extract contains hundreds of compounds and a specification tests a few of them. Each kind of test answers its own question, and a pass on one says little about the others.
How much of one named compound
A targeted assay quantifies one compound against its reference standard, such as rutin in mg per 40 mg extract. Compounds nobody looked for stay invisible.
One number for a group
The compendial ginkgo assay hydrolyzes flavonol glycosides and measures the released aglycones, so added aglycones or cheap glycosides raise the result. The six-glycoside sum used here adds six intact glycosides and is a different number.
Does the whole pattern match a reference
Liu et al. compared 59 common chromatographic peaks with the common pattern of their own 51 batches. High similarity means close to this market, which is not the same as close to authenticated leaf extract.
What the material does
Activity needs a validated bioassay or a clinical study. No test in this dataset measures it, and a consistent marker shows consistent chemistry for that marker only.
A marker is not automatically the active. The European Medicines Agency separates active markers, “generally accepted to contribute to the therapeutic activity,” from analytical markers, which “serve solely for analytical purposes.” Published evidence
EMA/HMPC, Reflection paper on markers, EMEA/HMPC/253629/2007 (2008); Liu et al., Chin Med 2022 (introduction and methods); EMA/HMPC/321095/2012 (Ph. Eur. ginkgo extract specification, summarized in section 02). Next: the 51 profiles.
02 · The published data
Fifty-one batches, 25 manufacturers, 11 measured constituents
The flavonol glycosides vary least between batches. Organic acids, which no ginkgo specification on this page limits, vary up to 141-fold, and a few batches have profiles unlike the rest.
The 51 profiles: each constituent relative to the 51-batch mean
Rows: organic acids, flavonol glycosides and free quercetin. Columns: batches grouped by manufacturer. Warm shades: below the mean; cool shades: above.
What was measured
- Material
- 51 batches of ginkgo extract tablets and capsules bought on the market in China, 1 to 7 batches from each of 25 manufacturers (A–Y); 13 manufacturers contributed one batch.
- Method
- UPLC with diode-array detection at 260 nm; 11 constituents quantified against reference standards; recovery 97.5–107% (Table 1 of the paper).
- Units
- mg per 40 mg of ginkgo extract, normalized to the labeled extract content of each product.
- Fingerprint
- Correlation of each chromatogram with the common pattern of the 51 batches (59 common peaks, ChemPattern 2.0).
- Not measured
- Terpene lactones, ginkgolic acids, compendial total flavonol glycosides and any biological activity.
Sampling limits to keep in mind
- A market purchase in one country, not a random or representative sample. These are preparations sold in China, used here as a published example of batch variation, not as a description of U.S. products.
- One laboratory and one method; no measurement uncertainty is reported for individual values.
- The authors note that preparation makers can buy registered extract from other companies, so a manufacturer code does not identify the extract source.
- The similarity reference is derived from the same 51 batches. The authors report that the Chinese Pharmacopoeia 2020 fingerprint, which focuses on flavonol glycosides, gave similarity above 0.9 for all samples (data not shown in the paper).
- Table 3 of the paper lists 17 batches below 0.85. Table 2's values give 19 (U1 at 0.78 and X1 at 0.83 are missing from Table 3); this page uses Table 2.
Our reading of Table 2 Within manufacturer K, batches K3 and K4 contain 2.2 times the rutin of K1, K2 and K5, while the other five glycosides differ by at most 16%, a pattern that warrants investigation. Rutin is a cheap, widely available flavonol glycoside, and the authors name its illegal addition as a known concern.
The European Pharmacopoeia monograph “Ginkgo dry extract, refined and quantified” sets flavone glycosides at 22.0–27.0%, terpene lactones at 5.0–7.0% (ginkgolides A, B and C 2.8–3.4%, bilobalide 2.6–3.2%) and ginkgolic acids at no more than 5 ppm (EMA/HMPC/321095/2012). Those analytes were not measured here, so these batches cannot be judged against it. Next: apply a specification.
Every value: Table 2 of Liu et al. 2022, with three computed columns
Values in mg per 40 mg extract as printed; n.d. = not detected (“–” in the paper). The printed total was checked against the 11 columns for every batch; the column maxima and averages match the summary rows of Table 2. Decision column: all tests on, default specification (section 03).
| Batch | 6-HKA | pHA | GA | PA | KR | KRCG | QDG | QCG | Rutin | IR | Quercetin | Total (printed) | Similarity | Six glycosides (computed) | Rutin share (computed) | Decision, all tests on (computed) |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| A1 | 0.048 | 0.033 | 0.014 | 0.365 | 0.818 | 0.820 | 1.009 | 0.886 | 2.075 | 0.644 | 0.038 | 6.749 | 1.00 | 6.252 | 33% | Rutin share |
| B1 | 0.077 | 0.051 | 0.053 | 0.715 | 0.814 | 0.786 | 1.070 | 0.885 | 2.116 | 0.632 | 0.101 | 7.298 | 0.93 | 6.303 | 34% | Rutin share |
| C1 | 0.007 | 0.016 | 0.003 | 0.070 | 0.924 | 1.040 | 1.080 | 1.240 | 2.196 | 0.844 | 0.091 | 7.511 | 0.89 | 7.324 | 30% | Specification |
| D1 | 0.020 | 0.025 | 0.005 | 0.141 | 0.756 | 0.947 | 0.675 | 1.194 | 0.774 | 0.751 | 0.056 | 5.343 | 0.90 | 5.097 | 15% | Held |
| E1 | 0.010 | 0.037 | n.d. | 0.090 | 0.724 | 1.193 | 0.990 | 1.311 | 0.773 | 0.563 | 0.045 | 5.736 | 0.86 | 5.554 | 14% | Held |
| F1 | 0.022 | 0.061 | 0.065 | 0.908 | 0.831 | 0.913 | 1.117 | 1.111 | 1.964 | 0.762 | 0.023 | 7.779 | 0.86 | 6.698 | 29% | Rutin share |
| F2 | 0.007 | 0.065 | 0.049 | 0.955 | 0.897 | 1.007 | 1.188 | 1.215 | 1.071 | 0.835 | 0.027 | 7.317 | 0.87 | 6.213 | 17% | Supplier |
| G1 | 0.007 | 0.026 | n.d. | 0.213 | 0.791 | 0.692 | 0.886 | 0.856 | 0.805 | 0.786 | 0.018 | 5.080 | 0.92 | 4.816 | 17% | Specification |
| G2 | 0.006 | 0.035 | n.d. | 0.183 | 0.858 | 0.815 | 0.864 | 1.135 | 0.887 | 0.796 | 0.020 | 5.598 | 0.86 | 5.355 | 17% | Accepted |
| G3 | 0.008 | 0.031 | n.d. | 0.187 | 0.803 | 0.803 | 0.885 | 1.119 | 0.869 | 0.741 | 0.024 | 5.471 | 0.86 | 5.220 | 17% | Accepted |
| G4 | 0.004 | 0.037 | n.d. | 0.215 | 0.878 | 0.893 | 1.019 | 1.204 | 0.915 | 0.818 | 0.026 | 6.009 | 0.91 | 5.727 | 16% | Accepted |
| H1 | 0.007 | 0.026 | n.d. | 0.068 | 0.744 | 1.101 | 1.116 | 1.384 | 0.809 | 0.704 | 0.038 | 5.997 | 0.86 | 5.858 | 14% | Held |
| I1 | 0.057 | 0.027 | 0.019 | 0.421 | 0.960 | 0.989 | 0.892 | 1.240 | 1.144 | 0.886 | 0.022 | 6.656 | 0.98 | 6.111 | 19% | Held |
| J1 | 0.029 | 0.074 | 0.047 | 0.692 | 0.925 | 1.042 | 1.206 | 1.292 | 0.859 | 0.833 | 0.022 | 7.020 | 0.93 | 6.157 | 14% | Held |
| K1 | 0.018 | 0.032 | n.d. | 0.056 | 0.598 | 1.026 | 0.842 | 1.227 | 0.693 | 0.507 | 0.056 | 5.055 | 0.75 | 4.893 | 14% | Specification |
| K2 | 0.018 | 0.035 | n.d. | 0.062 | 0.627 | 1.011 | 0.831 | 1.206 | 0.700 | 0.536 | 0.052 | 5.078 | 0.86 | 4.911 | 14% | Specification |
| K3 | 0.020 | 0.034 | 0.009 | 0.066 | 0.612 | 0.970 | 0.908 | 1.135 | 1.636 | 0.521 | 0.052 | 5.963 | 0.74 | 5.782 | 28% | Rutin share |
| K4 | 0.019 | 0.034 | n.d. | 0.032 | 0.613 | 1.003 | 0.961 | 1.173 | 1.675 | 0.526 | 0.048 | 6.086 | 0.73 | 5.951 | 28% | Rutin share |
| K5 | 0.018 | 0.028 | n.d. | 0.058 | 0.597 | 1.002 | 0.932 | 1.170 | 0.833 | 0.511 | 0.078 | 5.227 | 0.85 | 5.045 | 17% | Supplier |
| L1 | 0.009 | 0.079 | 0.054 | 1.132 | 0.880 | 1.026 | 1.127 | 1.301 | 0.823 | 0.835 | 0.021 | 7.288 | 0.83 | 5.992 | 14% | Fingerprint |
| L2 | 0.005 | 0.030 | n.d. | 0.100 | 0.491 | 0.898 | 0.614 | 1.121 | 0.501 | 0.492 | 0.272 | 4.523 | 0.71 | 4.117 | 12% | Specification |
| M1 | 0.009 | 0.022 | n.d. | 0.105 | 1.011 | 1.324 | 1.152 | 1.559 | 0.973 | 0.947 | 0.048 | 7.150 | 0.82 | 6.966 | 14% | Fingerprint |
| N1 | 0.011 | 0.025 | 0.001 | 0.055 | 0.797 | 1.105 | 1.010 | 1.308 | 0.797 | 0.726 | 0.060 | 5.895 | 0.84 | 5.743 | 14% | Fingerprint |
| N2 | 0.014 | 0.027 | n.d. | 0.067 | 0.808 | 1.071 | 0.829 | 1.240 | 0.790 | 0.735 | 0.053 | 5.634 | 0.87 | 5.473 | 14% | Supplier |
| O1 | n.d. | 0.009 | n.d. | 0.008 | 0.706 | 1.098 | 0.339 | 1.373 | 0.820 | 0.618 | 0.075 | 5.046 | 0.78 | 4.954 | 17% | Specification |
| P1 | 0.016 | 0.034 | 0.002 | 0.200 | 1.108 | 1.034 | 1.056 | 1.124 | 1.207 | 1.182 | 0.072 | 7.035 | 0.88 | 6.711 | 18% | Held |
| Q1 | 0.009 | 0.039 | 0.002 | 0.142 | 0.859 | 0.875 | 0.650 | 1.112 | 0.800 | 0.717 | 0.030 | 5.233 | 0.90 | 5.013 | 16% | Supplier |
| Q2 | 0.023 | 0.038 | 0.010 | 0.195 | 0.708 | 0.932 | 0.910 | 1.244 | 1.873 | 0.593 | 0.018 | 6.545 | 0.92 | 6.260 | 30% | Rutin share |
| Q3 | 0.004 | 0.035 | 0.005 | 0.112 | 0.773 | 0.840 | 0.735 | 1.119 | 1.725 | 0.654 | 0.042 | 6.043 | 0.86 | 5.846 | 30% | Rutin share |
| R1 | 0.002 | 0.023 | n.d. | 0.039 | 0.696 | 0.982 | 0.805 | 1.160 | 0.615 | 0.701 | 0.142 | 5.167 | 0.83 | 4.959 | 12% | Specification |
| R2 | 0.008 | 0.029 | n.d. | 0.037 | 0.752 | 1.071 | 0.943 | 1.256 | 0.669 | 0.758 | 0.148 | 5.672 | 0.73 | 5.449 | 12% | Fingerprint |
| R3 | 0.009 | 0.029 | n.d. | 0.038 | 0.761 | 1.075 | 0.925 | 1.248 | 1.358 | 0.772 | 0.155 | 6.370 | 0.74 | 6.139 | 22% | Fingerprint |
| S1 | 0.034 | 0.050 | 0.030 | 0.697 | 1.296 | 1.160 | 0.935 | 1.404 | 2.906 | 1.185 | 0.029 | 9.727 | 0.96 | 8.886 | 33% | Specification |
| S2 | 0.058 | 0.045 | 0.024 | 0.669 | 1.333 | 1.135 | 1.118 | 1.362 | 1.247 | 1.232 | 0.032 | 8.255 | 0.93 | 7.427 | 17% | Specification |
| T1 | 0.002 | 0.027 | n.d. | 0.063 | 0.700 | 1.053 | 0.904 | 1.207 | 0.630 | 0.708 | 0.151 | 5.445 | 0.71 | 5.202 | 12% | Fingerprint |
| U1 | 0.003 | 0.015 | n.d. | 0.037 | 0.935 | 1.224 | 0.904 | 1.419 | 0.873 | 0.859 | 0.127 | 6.395 | 0.78 | 6.214 | 14% | Fingerprint |
| V1 | 0.029 | 0.046 | n.d. | 0.049 | 0.913 | 1.496 | 1.440 | 1.749 | 2.507 | 0.790 | 0.082 | 9.101 | 0.45 | 8.895 | 28% | Specification |
| V2 | 0.015 | 0.045 | 0.052 | 0.033 | 0.783 | 0.059 | 0.086 | 0.368 | 1.314 | 1.827 | 0.041 | 4.624 | 0.86 | 4.437 | 30% | Specification |
| V3 | 0.005 | 0.023 | n.d. | 0.022 | 0.818 | 0.896 | 0.724 | 1.070 | 0.920 | 0.759 | 0.036 | 5.272 | 0.84 | 5.187 | 18% | Fingerprint |
| W1 | 0.004 | 0.012 | n.d. | 0.018 | 0.947 | 1.115 | 0.912 | 1.262 | 0.877 | 0.906 | 0.020 | 6.071 | 0.78 | 6.019 | 15% | Fingerprint |
| W2 | 0.003 | 0.013 | n.d. | 0.018 | 0.961 | 1.135 | 0.981 | 1.319 | 1.090 | 0.923 | 0.022 | 6.465 | 0.86 | 6.409 | 17% | Supplier |
| W3 | 0.012 | 0.019 | 0.011 | 0.059 | 0.807 | 1.019 | 0.734 | 1.172 | 0.915 | 0.859 | 0.095 | 5.701 | 0.88 | 5.506 | 17% | Supplier |
| X1 | 0.035 | 0.032 | 0.020 | 0.554 | 0.809 | 1.123 | 1.107 | 1.363 | 0.819 | 0.756 | 0.021 | 6.638 | 0.83 | 5.977 | 14% | Fingerprint |
| X2 | 0.064 | 0.027 | 0.019 | 0.439 | 0.868 | 1.060 | 1.050 | 1.350 | 1.171 | 0.803 | 0.024 | 6.875 | 0.92 | 6.302 | 19% | Supplier |
| X3 | 0.050 | 0.025 | 0.016 | 0.367 | 0.713 | 1.005 | 0.712 | 1.407 | 2.190 | 0.730 | 0.019 | 7.234 | 0.95 | 6.757 | 32% | Rutin share |
| X4 | 0.047 | 0.018 | 0.015 | 0.349 | 0.704 | 1.109 | 0.929 | 1.434 | 1.066 | 0.647 | 0.024 | 6.341 | 0.85 | 5.889 | 18% | Supplier |
| X5 | 0.039 | 0.020 | 0.034 | 0.651 | 1.064 | 1.614 | 1.204 | 2.104 | 1.132 | 1.018 | 0.025 | 8.905 | 0.93 | 8.136 | 14% | Specification |
| X6 | 0.134 | 0.048 | 0.031 | 0.524 | 1.116 | 1.348 | 1.800 | 1.805 | 1.507 | 1.045 | 0.029 | 9.384 | 0.95 | 8.621 | 17% | Specification |
| X7 | 0.056 | 0.028 | 0.016 | 0.425 | 0.793 | 1.031 | 1.042 | 1.357 | 1.139 | 0.766 | 0.021 | 6.673 | 0.95 | 6.128 | 19% | Supplier |
| Y1 | 0.002 | 0.028 | n.d. | 0.028 | 0.665 | 1.105 | 1.090 | 1.285 | 0.896 | 0.543 | 0.048 | 5.690 | 0.73 | 5.584 | 16% | Fingerprint |
| Y2 | 0.003 | 0.033 | n.d. | 0.031 | 0.656 | 0.936 | 0.753 | 1.318 | 0.808 | 0.627 | 0.074 | 5.237 | 0.53 | 5.098 | 16% | Fingerprint |
03 · Apply a specification
Each added test removes batches that the assay accepted
Start with the targeted assay alone, then add a free-aglycone limit, a profile ratio, the fingerprint and supplier qualification. Every decision is computed live from the published values.
Shown at the default settings: six-glycoside sum 5.0–7.0 mg per 40 mg, targeted assay only, 80 mg extract per serving. The table below the tool gives the result as each test is added.
- Accepted
- 38 of 51
- Held
- 0
- Rejected
- 13
- Six-glycoside sum delivered, accepted batches
- 10.0–13.9 mg per serving (1.4-fold)
- Protocatechuic acid delivered (not specified)
- 0.036–2.26 mg per serving (63-fold)
Decision for each batch, grouped by manufacturer (A–Y)
- A1: Accepted; six-glycoside sum 6.252 mg per 40 mg, similarity 1.00
- B1: Accepted; six-glycoside sum 6.303 mg per 40 mg, similarity 0.93
- C1: Outside the marker specification; six-glycoside sum 7.324 mg per 40 mg, similarity 0.89
- D1: Accepted; six-glycoside sum 5.097 mg per 40 mg, similarity 0.90
- E1: Accepted; six-glycoside sum 5.554 mg per 40 mg, similarity 0.86
- F1: Accepted; six-glycoside sum 6.698 mg per 40 mg, similarity 0.86
- F2: Accepted; six-glycoside sum 6.213 mg per 40 mg, similarity 0.87
- G1: Outside the marker specification; six-glycoside sum 4.816 mg per 40 mg, similarity 0.92
- G2: Accepted; six-glycoside sum 5.355 mg per 40 mg, similarity 0.86
- G3: Accepted; six-glycoside sum 5.220 mg per 40 mg, similarity 0.86
- G4: Accepted; six-glycoside sum 5.727 mg per 40 mg, similarity 0.91
- H1: Accepted; six-glycoside sum 5.858 mg per 40 mg, similarity 0.86
- I1: Accepted; six-glycoside sum 6.111 mg per 40 mg, similarity 0.98
- J1: Accepted; six-glycoside sum 6.157 mg per 40 mg, similarity 0.93
- K1: Outside the marker specification; six-glycoside sum 4.893 mg per 40 mg, similarity 0.75
- K2: Outside the marker specification; six-glycoside sum 4.911 mg per 40 mg, similarity 0.86
- K3: Accepted; six-glycoside sum 5.782 mg per 40 mg, similarity 0.74
- K4: Accepted; six-glycoside sum 5.951 mg per 40 mg, similarity 0.73
- K5: Accepted; six-glycoside sum 5.045 mg per 40 mg, similarity 0.85
- L1: Accepted; six-glycoside sum 5.992 mg per 40 mg, similarity 0.83
- L2: Outside the marker specification; six-glycoside sum 4.117 mg per 40 mg, similarity 0.71
- M1: Accepted; six-glycoside sum 6.966 mg per 40 mg, similarity 0.82
- N1: Accepted; six-glycoside sum 5.743 mg per 40 mg, similarity 0.84
- N2: Accepted; six-glycoside sum 5.473 mg per 40 mg, similarity 0.87
- O1: Outside the marker specification; six-glycoside sum 4.954 mg per 40 mg, similarity 0.78
- P1: Accepted; six-glycoside sum 6.711 mg per 40 mg, similarity 0.88
- Q1: Accepted; six-glycoside sum 5.013 mg per 40 mg, similarity 0.90
- Q2: Accepted; six-glycoside sum 6.260 mg per 40 mg, similarity 0.92
- Q3: Accepted; six-glycoside sum 5.846 mg per 40 mg, similarity 0.86
- R1: Outside the marker specification; six-glycoside sum 4.959 mg per 40 mg, similarity 0.83
- R2: Accepted; six-glycoside sum 5.449 mg per 40 mg, similarity 0.73
- R3: Accepted; six-glycoside sum 6.139 mg per 40 mg, similarity 0.74
- S1: Outside the marker specification; six-glycoside sum 8.886 mg per 40 mg, similarity 0.96
- S2: Outside the marker specification; six-glycoside sum 7.427 mg per 40 mg, similarity 0.93
- T1: Accepted; six-glycoside sum 5.202 mg per 40 mg, similarity 0.71
- U1: Accepted; six-glycoside sum 6.214 mg per 40 mg, similarity 0.78
- V1: Outside the marker specification; six-glycoside sum 8.895 mg per 40 mg, similarity 0.45
- V2: Outside the marker specification; six-glycoside sum 4.437 mg per 40 mg, similarity 0.86
- V3: Accepted; six-glycoside sum 5.187 mg per 40 mg, similarity 0.84
- W1: Accepted; six-glycoside sum 6.019 mg per 40 mg, similarity 0.78
- W2: Accepted; six-glycoside sum 6.409 mg per 40 mg, similarity 0.86
- W3: Accepted; six-glycoside sum 5.506 mg per 40 mg, similarity 0.88
- X1: Accepted; six-glycoside sum 5.977 mg per 40 mg, similarity 0.83
- X2: Accepted; six-glycoside sum 6.302 mg per 40 mg, similarity 0.92
- X3: Accepted; six-glycoside sum 6.757 mg per 40 mg, similarity 0.95
- X4: Accepted; six-glycoside sum 5.889 mg per 40 mg, similarity 0.85
- X5: Outside the marker specification; six-glycoside sum 8.136 mg per 40 mg, similarity 0.93
- X6: Outside the marker specification; six-glycoside sum 8.621 mg per 40 mg, similarity 0.95
- X7: Accepted; six-glycoside sum 6.128 mg per 40 mg, similarity 0.95
- Y1: Accepted; six-glycoside sum 5.584 mg per 40 mg, similarity 0.73
- Y2: Accepted; six-glycoside sum 5.098 mg per 40 mg, similarity 0.53
- Accepted
- Outside the marker specification
- Free quercetin above 0.2 mg per 40 mg extract
- Rutin above 25% of the six glycosides
- Fingerprint similarity below the threshold
- Supplier not qualified: another of its batches failed a check
- Held: only one batch on record, so the supplier cannot be qualified
What a serving delivers
Each dot is one batch. Values: Liu et al. 2022, Table 2, scaled to the serving.
Measured data Hypothetical example Values: Liu et al. 2022, Table 2. Specification, ratio limit, supplier rule and serving are hypothetical; the 0.2 mg quercetin limit and the 0.85 similarity cut are as stated by the authors.
Batches accepted as tests are added in order (hypothetical specification 5.0–7.0 mg six-glycoside sum per 40 mg extract; serving 80 mg extract)
| Tests applied | Accepted | Held | Rejected | Six glycosides per serving, accepted (mg) | Protocatechuic acid per serving, accepted (mg) |
|---|---|---|---|---|---|
| Targeted assay against the specification | 38 | 0 | 13 | 10.0–13.9 | 0.036–2.264 |
| + free-quercetin limit | 38 | 0 | 13 | 10.0–13.9 | 0.036–2.264 |
| + profile ratio (rutin share) | 30 | 0 | 21 | 10.0–13.9 | 0.036–2.264 |
| + fingerprint similarity ≥ 0.85 | 18 | 0 | 33 | 10.0–13.4 | 0.036–1.910 |
| + supplier qualification | 3 | 6 | 42 | 10.4–11.5 | 0.366–0.430 |
At the default specification the assay accepts 38 batches. The rutin ratio removes 8, the fingerprint 12 more, and supplier qualification leaves 3, all from manufacturer G, with 6 more held for lack of history.
Method, assumptions and limits
What it calculates. For each batch, the first enabled test it fails, in this order: marker specification (S), free quercetin (Q), rutin share (R), fingerprint similarity (F); then, if supplier qualification is on, U for a batch whose manufacturer has any batch failing Q, R or F, and H (held) for a manufacturer with a single batch.
Equations. Six-glycoside sum = KR + KRCG + QDG + QCG + rutin + IR (mg per 40 mg extract). Rutin share = rutin ÷ six-glycoside sum. Dose per serving = content × serving ÷ 40 mg. Standardized fill = serving × (lower + upper limit) ÷ 2 ÷ marker content, in mg extract; any other constituent then delivers its content ÷ 40 × fill.
Assumptions. Values are used as printed, without measurement uncertainty; a not-detected value counts as zero, as in the authors' totals. Limits are inclusive. The serving and the standardization are arithmetic on a hypothetical label, not dosing advice and not a formulation.
Valid range. Limits from 0 to 20 mg per 40 mg; a lower limit above the upper limit is swapped and the page says so.
What it does not establish. Authenticity, safety or activity of any product. A failed check marks a batch for investigation; it does not prove adulteration. The 25% rutin limit was placed at a gap in these same data, which a real specification must not do: it would be set from authenticated reference lots.
Verification. The default results were checked against an independent calculation and against the totals, maxima and averages printed in Table 2.
04 · What a serving delivers
A fixed serving of extract delivers a variable dose
At a hypothetical label of 80 mg extract per serving, the same labeled amount delivers different amounts of each constituent. A specification narrows the constituent it names and leaves the rest as they come.
Six-glycoside dose across all 51 batches
8.2 to 17.8 mg per 80 mg serving, from the same labeled extract amount.
Protocatechuic acid across all 51
0.016 to 2.26 mg per serving. No specification in this demonstration covers it.
Six-glycoside dose inside the specification
10.0 to 13.9 mg per serving in the 38 accepted batches: a 5.0–7.0 range allows up to 1.4-fold by design.
Protocatechuic acid after every test
0.37 to 0.43 mg per serving in the 3 batches still accepted, all from one manufacturer.
These are batch-to-batch differences. Unit-to-unit uniformity within one batch is a separate test that needs per-unit data, which the study does not report. Hypothetical example
Computed from Liu et al. 2022, Table 2, for a hypothetical 80 mg serving; change the serving in section 03.
05 · Process correction
Blending corrects a marker, and it can also hide a failed batch
Mixing lots by mass is linear, so a blend's content of each measured constituent can be computed from published values. A fingerprint similarity cannot be computed this way; it would have to be measured on the blend.
Correcting strength
S2 is above and G1 below the hypothetical range, and both pass every authenticity check. A blend of about 45% S2 and 55% G1 gives 6.0 mg.
Correcting strength: two batches that pass every authenticity check, one above and one below the specification, blended to 6.0 mg (values per 40 mg extract)
| Quantity | S2 (measured) | G1 (measured) | Blend 45% S2 + 55% G1 (computed) |
|---|---|---|---|
| Six glycosides | 7.427 mg | 4.816 mg | 6.000 mg |
| Free quercetin | 0.032 mg | 0.018 mg | 0.024 mg |
| Protocatechuic acid | 0.669 mg | 0.213 mg | 0.420 mg |
| Rutin share of six glycosides | 17% | 17% | 17% |
| Fingerprint similarity | 0.93 | 0.92 | not computable |
Hiding a failure
L2 exceeds the free-quercetin limit and has the lowest glycoside content. Mixed 1:1 with X6, the blend passes both the marker range and the limit while half of it is the failed batch.
Hiding a failure: a batch over the free-quercetin limit, blended 1:1 with a rich batch (values per 40 mg extract)
| Quantity | L2 (measured) | X6 (measured) | Blend 50% L2 + 50% X6 (computed) |
|---|---|---|---|
| Six glycosides | 4.117 mg | 8.621 mg | 6.369 mg |
| Free quercetin | 0.272 mg | 0.029 mg | 0.151 mg |
| Protocatechuic acid | 0.100 mg | 0.524 mg | 0.312 mg |
| Rutin share of six glycosides | 12% | 17% | 16% |
| Fingerprint similarity | 0.71 | 0.95 | not computable |
Measured data Hypothetical example Batch values: Liu et al. 2022, Table 2. Blends are computed by mass balance, not measured lots; not-detected values count as zero.
Standardizing the serving fixes one number
With every batch-level test on and the fill per serving adjusted so each of the 18 accepted batches delivers 12.0 mg of the six glycosides, the extract per serving ranges from 72 to 96 mg. Protocatechuic acid still ranges from 0.034 to 1.84 mg per serving, 55-fold, against 53-fold before the adjustment.
Under Part 111, a component that fails its identity specification must be rejected and not used (21 CFR 111.77(b)). Other specification failures may be corrected only by a treatment, in-process adjustment or reprocessing that quality control approves after a material review (111.77(a), 111.90). EUTECNIQ proposes to write the fingerprint and authenticity checks into identity and to blend only lots that pass them. Proposed capability
eCFR 21 CFR 111.77 and 111.90 (version of October 28, 2026). Whether a fingerprint failure is an identity failure depends on how the manufacturer writes its identity specification under 111.70(b)(1).
06 · What each test catches here
No single test sorted these batches correctly on its own
Run on the 51 published batches at the default settings, each test removes a different set, and two kinds of evidence a full decision needs are not in the dataset at all.
| Test | What it measures | What it cannot tell | In these 51 batches |
|---|---|---|---|
| Targeted assay, six-glycoside sum | Six named glycosides against reference standards | Anything not among the six; whether rutin was added | 13 outside 5.0–7.0 mg; 38 accepted |
| Free-quercetin limit | Free aglycone, a marker of hydrolysis or added aglycone | Added glycosides such as rutin | One batch above 0.2 mg (L2, 0.272), already outside the range |
| Profile ratio, rutin share | Whether one glycoside is out of proportion to the others | Its own limit, without authenticated reference lots | 12 batches above 25%, including A1 at similarity 1.00 |
| Fingerprint similarity | The whole UPLC pattern against the market's common pattern | Whether the reference itself is authentic | 19 below 0.85; 14 of them inside the marker range |
| Supplier qualification | Whether a source's batches pass consistently over time | Anything about a source with one batch on record | Only G, of 12 multi-batch manufacturers, has no batch failing an authenticity check |
| Non-targeted LC-HRMS screening | Compounds nobody specified, by accurate mass | Quantity without standards; identity without confirmation | Not performed in this study |
| Bioassay or clinical evidence | What the material does | Chemical identity | Not performed in this study |
Counts computed from Liu et al. 2022, Table 2. For non-targeted screening and confidence levels see targeted and non-targeted analysis; for a case where only isotope ratios separated synthetic from natural curcuminoids see discovery and characterization; for decision rules near a limit see certificate-of-analysis science.
07 · What EUTECNIQ proposes
Connect sourcing, specification, process and records in one decision
A contract testing laboratory reports results against the methods and limits its client sends. Every outcome on this page turned on choices made before a test ran: which limits, which supplier, which blending rule. EUTECNIQ proposes to make those choices with its clients and keep the reasons on record.
Qualify the extract source
Qualify the plant material and the extraction process behind an ingredient, since a brand or manufacturer code need not identify it. Keep authenticated reference lots of documented origin.
Set limits from many lots
Combine a marker range, an aglycone limit, profile ratios and a fingerprint compared with authenticated references, each limit justified from multi-lot data and documented.
Correct strength only in lots that pass identity
Write blending and standardization rules in advance, apply them only to lots that pass identity and authenticity, and state which constituents the correction leaves variable.
One record per lot
Link the supplier's certificate, its periodic confirmation (21 CFR 111.75(a)(2)(ii)), every test result, the decision and its reason, and the dose a serving delivers.
Proposed capabilityEUTECNIQ has not built or operated this system. The roles above describe what it proposes to offer.
08 · Sources
Sources for this demonstration
Consistency demonstration: dataset, markers and specifications
- Liu L-n, Jin H-y, Ke Z, Xu W-y, Sun L, Ma S-c. A strategy for quality control of ginkgo biloba preparations based on UPLC fingerprint analysis and multi-component separation combined with quantitative analysis. Chin Med. 2022;17:72. Open access, CC BY 4.0 (PMC9202209). Table 2 values (51 batches, 11 constituents, similarity) used throughout this page; full text read November 1, 2026. doi.org/10.1186/s13020-022-00618-3
- European Medicines Agency, Committee on Herbal Medicinal Products. Assessment report on Ginkgo biloba L., folium. EMA/HMPC/321095/2012, final, January 28, 2014 (herbal preparations: Ph. Eur. monograph 1827, Ginkgo dry extract, refined and quantified, p. 4). Read November 1, 2026. ema.europa.eu/en/documents/herbal-report/final-assessment-report-ginkgo-biloba-l-folium_en.pdf
- EMA Committee on Herbal Medicinal Products. Reflection paper on markers used for quantitative and qualitative analysis of herbal medicinal products and traditional herbal medicinal products. EMEA/HMPC/253629/2007, July 15, 2008 (definitions of active and analytical markers). Read November 1, 2026. ema.europa.eu/en/documents/scientific-guideline/reflection-paper-markers-used-quantitative-qualitative-analysis-herbal-medicinal-products_en.pdf
Consistency demonstration: U.S. dietary supplement CGMP (21 CFR Part 111)
- Electronic Code of Federal Regulations. 21 CFR 111.70, What specifications must you establish? (paragraphs (b) and (e)). eCFR version of October 28, 2026, read November 1, 2026. ecfr.gov/current/title-21/chapter-I/subchapter-B/part-111/subpart-E/section-111.70
- Electronic Code of Federal Regulations. 21 CFR 111.75, What must you do to determine whether specifications are met? (paragraph (a): identity testing; reliance on a qualified supplier's certificate of analysis). eCFR version of October 28, 2026, read November 1, 2026. ecfr.gov/current/title-21/chapter-I/subchapter-B/part-111/subpart-E/section-111.75
- Electronic Code of Federal Regulations. 21 CFR 111.77, What must you do if established specifications are not met? eCFR version of October 28, 2026, read November 1, 2026. ecfr.gov/current/title-21/chapter-I/subchapter-B/part-111/subpart-E/section-111.77
- Electronic Code of Federal Regulations. 21 CFR 111.90, What requirements apply to treatments, in-process adjustments, and reprocessing? eCFR version of October 28, 2026, read November 1, 2026. ecfr.gov/current/title-21/chapter-I/subchapter-B/part-111/subpart-E/section-111.90
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, and nothing here is medical advice.
Read the full notice
The information is for general education only. It 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.