S157 // COA FORENSICS
HPLC and LC-MS appear in almost all COAs on the research market. The problem is treating the two as equivalent - they're not. This article separates, in an operational way, what each method confirmswhich unconfirmedand how to identify "pretty COAs" that only look technical. At the end you'll find a S157 checklist, educational examples with Forensic Score (0-100) and direct links to your ecosystem: COA Auditor, Lexicon, Lab Tools and Peptide Database.
The aim is to educate and standardise the reading of reports. We don't accuse specific entities, we don't identify vendors and we don't do "how-to" sourcing. The focus is on reducing error and increasing consistency in auditing.
Why "pretty COA" is a risk vector
A report with a perfect layout can hide the essentials: raw datatraceability and internal coherence. Fraud rarely depends on "gross lies"; it depends on controlled ambiguity - enough information for a technical opinion, but not enough to be auditable.
- high-performance liquid chromatography measures "separation and relative purity" - it doesn't prove identity.
- MS confirms mass/fragments - but can be presented without sample context, without chain of custody and without contamination control.
- The common mistake: accepting percentages (e.g. "99%") without looking at baseline, integration, secondary peaks and method.
Forensic checklist (quick) - 12 audit points
- Batch on the bottle and PDF are identical (same format, same number, no "creative variations")?
- Laboratory is it verifiable (address, contact, accreditation, method, instrument)?
- HPLC method describes column, mobile phase, gradient, wavelength, total time?
- Chromatogram does it exist (not just a table) and does it have a "clean" and coherent baseline?
- Integration shows how areas were added up (without convenient "cuts")?
- Secondary peaks are they visible and discussed (impurities/degradation)?
- LC-MS has expected m/z + tolerance and (where applicable) MS/MS fragmentation?
- Date and time of the test match the "history" of the batch (it's not a recycled PDF)?
- Signature / responsible does it exist and is it consistent between pages?
- Endotoxins / sterility (when alleged) has method and limits (not just "Pass")?
- Units and decimals are they coherent (no impossible %, no sums >100)?
- Raw dataIs there a way to get raw data, instrument logs, or file references?
Forensic Score (0-100) - educational examples
Example A - "Minimalist COA"
"Clean" PDF with percentages and stamp, but no chromatograms, no method, no traceability.
- !No graphics: just a purity table. Without a baseline, there is no real audit.
- iOpaque labno verifiable accreditation and no instrument specified.
- !Weak batchgeneric code (e.g. "BATCH-01") that can be repeated.
Example B - "ISO quoted (opaque)"
Cites ISO/IEC 17025, but does not provide scope, method or raw data; chromatogram exists, but with unclear integration.
- iISO without scopeAccreditation exists, but the specific test may not be accredited.
- iIntegrationlack of parameters (threshold, smoothing) makes replication difficult.
- iLC-MS: shows mass, but no sample/controls details.
Example C - "Auditable evidence"
Charts, complete method, consistent batch and trail for raw data (file / contact / internal id).
- ✓Chromatogram + method: column, gradient, time, detector, consistent RT.
- ✓Contextual LC-MS: expected m/z + tolerance + fragmentation where it makes sense.
- ✓Tracecoherent batch, dates and responsible person.
How to use the score (without "false certainties")
The score doesn't prove "truth" - it just measures it auditability. Use it to prioritise what to investigate:
- 10-30: reject as evidence (insufficient).
- 231-70: ask for additional data / look for inconsistencies.
- 371-100: better signal; still validate chain of custody.
Quick illustration: anatomy of a chromatogram (and where they "hide tricks")
Visual mini-map: what an AOC should connect (no "leaps of faith")
HPLC: integration vs baseline vs secondary peaks - 6 visual red flags
| Red flag | What are you looking for in the graphic? | Why it matters | How to mitigate (educational) |
|---|---|---|---|
| RF1 baseline | Baseline with "waves", drift or abnormal noise, especially near the main peak. | Noise can mask small peaks and create "artificial purity" through faulty integration. | Require high-resolution chromatogram + integration parameters. Cross-check with LC-MS and repeat method. |
| RF2 integration | Integrated areas without marks/lines; sharp cuts; peaks "swallowed up" by the main peak. | Integration is where the percentage is born. Without transparency, "99%" is just marketing. | Look for software exports (raw + integration report) or ask for parameter justification. |
| RF3 peaks | Small secondary peaks before/after the main one, mainly at similar RT. | Impurities and degradation can exist even when the main peak dominates. | Assess whether secondary peaks add up to a relevant area. Check consistency between plots. |
| RF4 RT | Inconsistent retention time between pages/batches or "round RT" (e.g. 10.00 every time). | RT is the method's signature. Inconsistency suggests recycled template or uncontrolled method. | Compare RT with previous literature / COAs. Require column and gradient detail. |
| RF5 scale | "Chopped" Y-axis or aggressive compression that visually reduces secondary peaks. | Manipulating scale is the simplest way to hide noise/peaks without "lying" in the figures. | Ask for the same graph on a logarithmic and linear scale, or zoomed in on the baseline. |
| RF6 method | Missing, generic or incompatible method (no column, no solvents, no detector). | Without a method, there is no replication. And without replication, there is no audit. | Solid COA describes method. If it doesn't, treat it as weak evidence. |
Key Terms (Lexicon) - for quick navigation
Internal shortcuts (Lexicon/Tools/Database) for the terms that appear most often in COAs and audits.
Related Database Profiles - repeated patterns without accusing anyone
Objective: to show how a "COA template" (layout, tables, phrases) can appear repeated in multiple substances - this doesn't prove fraud, but it is a risk pattern that deserves an extra audit.
Semaglutide
Metabolic (GLP-1). Highly targeted by market volume - typical scenario for recycled COAs.
Tirzepatide
Dual (GLP-1/GIP). Attention "99%" without chromatogram and without contextual LC-MS.
Retatrutide
Triple (GLP-1/GIP/Glucagon). Excellent case for comparing RT/integration between batches.
Cagrilintide
Amylin (satiety). Useful for observing common "templates" in metabolic reports.
CJC-1295
GH class (GHRH). Often appears with "generic" COAs (incomplete method).
PT-141
Neuro/sexual health. Good for detecting identity inconsistencies when only HPLC is available.
How to turn this into S157 practice (without complicating things)
- If the COA does not have a chromatogram and method, treat it as weak evidence (even if it says "99%").
- If you have a chromatogram, audit baseline + integration + secondary peaks with the RF1-RF6.
- If you quote ISO/IEC 17025, you confirm scope and whether the specific test is covered by accreditation (don't assume).
- If the PDF doesn't link to raw data/custody, take the risk and use the COA Auditor to standardise reading.
If you want, at the end of each audit keep a snapshot of 5 items: batch, date, method, chromatogram, LC-MS. This creates consistency between analyses and reduces "feel-good" decisions.
References
- ISO/IEC 17025:2017 - General requirements for the competence of testing and calibration laboratories.
- Skoog DA, Holler FJ, Crouch SR. Principles of Instrumental Analysis. (HPLC/UV fundamentals).
- Harris DC. Quantitative Chemical Analysis. (integration, baseline, analytical error).
EN 
PT 
FR 
ES 
DE 
IT 
PL