Integration is where purity is decided
Rule AX-R12 is added. The uncertainty budget shows that the choice between two integration conventions is larger than every other error source combined.
The index published its first uncertainty budget for a purity figure in 2026-Q1, and the exercise reordered the priorities.
The budget
| Source | Standard uncertainty (pp) |
|---|---|
| Relative response factors assumed to be unity | 0.10–0.30 |
| Integration boundary on an unresolved shoulder | 0.05–0.45 |
| Baseline placement | 0.08 |
| Peaks below the reporting threshold | 0.05 (systematic, one-directional) |
| Detector non-linearity | 0.06 |
| Column ageing across a sequence | 0.05 |
| Injection repeatability | 0.04 |
| Sample-preparation concentration error | 0.03 |
Combined standard uncertainty for a well-resolved main peak comes out at 0.24 pp, giving an expanded uncertainty of roughly 0.48 pp at 95 % confidence. With an unresolved shoulder it roughly doubles to 1.0 pp.
Three conclusions
- Differences under 0.3 pp are not findings. They are inside the expanded uncertainty of either measurement. The index does not treat them as findings and readers should not either.
- The empirical and theoretical numbers agree. Measured inter-lab limits of agreement run ±0.44 to ±0.52 pp for the high-volume laboratories, close to the calculated expanded uncertainty. That agreement is reassuring about both.
- Rule AX-R05's threshold is justified. Its 1.5 pp trigger is roughly three times the expanded uncertainty of a well-behaved measurement, which is why the index is willing to call a divergence above it a finding.
Rule AX-R12
The largest controllable term in the budget is the integration boundary, so the index added a rule for the one condition that makes that boundary undrawable: main-peak asymmetry above 1.50 at 10 % of peak height. Above that the trailing edge is placed by judgement rather than by geometry, the reported purity becomes a range, and the record is marked integration-limited.
The rule was drafted with a second limb — resolution below 1.20 to the nearest named impurity — and that limb was withdrawn before it ever adjudicated anything. How close the nearest impurity elutes is a property of the compound and of the gradient the laboratory chose, not a statement about the system or the material, and the index sets no threshold on it. Resolution is computed and printed on every record that has one, as a reported figure with no verdict attached. Asymmetry, by contrast, is a column finding a laboratory can act on: 13 records in the current index exceed 1.50, and a re-injection on a fresh column has cleared it every time it has been tried.
| Compound | Records | Median purity | Suppliers |
|---|---|---|---|
| Semaglutide | 106 | 99.03 | 8 |
| Retatrutide | 95 | 98.88 | 8 |
| Tirzepatide | 78 | 98.55 | 10 |
| Semax | 56 | 99.56 | 7 |
| TB-500 | 51 | 99.67 | 11 |
| BPC-157 | 43 | 99.59 | 9 |
| ARA-290 | 33 | 99.42 | 4 |
| CJC-1295 (no DAC) | 30 | 98.97 | 5 |
| Selank | 26 | 99.81 | 4 |
| Epitalon | 24 | 99.28 | 3 |
| CJC-1295 with DAC | 23 | 99.41 | 7 |
| Ipamorelin | 20 | 99.16 | 7 |