What is in Arc B
- Spectroscopic and chromatographic identification — USP ⟨197⟩, ⟨1119⟩, and why retention time alone is a weak sole ID test — worked as a real correlation coefficient
- Case study: when a Certificate of Analysis replaces an identity test — the 2023–2025 diethylene glycol / ethylene glycol contamination wave
Each module ends with a knowledge check. A cumulative assessment covering Arc A and Arc B together is issued separately.
This arc assumes Arc A's vocabulary and lifecycle. An identification method still has to be developed and validated or verified under USP ⟨1225⟩/⟨1226⟩ like any other method — what changes here is the specific parameter that matters most: Requirement ICH Q2(R2) requires identification methods to demonstrate specificity against structurally related materials, and that single requirement is the spine of this entire arc.
Spectroscopic and chromatographic identification
An identification test exists to answer one question: is this actually the substance the label says it is? Not every technique that produces a plausible-looking number can answer that question with confidence.
USP General Chapter ⟨197⟩, “Spectroscopic Identification Tests,” governs UV-Vis and infrared identification; USP ⟨1119⟩ covers near-infrared; and a fourth, very common approach — matching an HPLC peak's retention time — is not a dedicated identification chapter at all, which turns out to matter.
B1.1 USP ⟨197⟩, and a 2019 change worth knowing about
Requirement USP General Chapter ⟨197⟩, “Spectroscopic Identification Tests,” governs UV-Vis and infrared identification testing. A revision effective May 1, 2019 reorganized the chapter under a new “Identification Methodology” framework and renamed its two major sub-sections (Infrared Absorption became Infrared Spectroscopy; Ultraviolet Absorption became Ultraviolet-Visible Spectroscopy). The substantive change, not just the renaming, is this: the revised chapter permits comparing a sample's spectrum against a previously recorded, archived reference-standard spectrum, rather than requiring a fresh reference standard to be run side-by-side every single time.
Practice That is a genuine convenience — and it is also a real data-integrity question the moment it's adopted. An archived spectral library is now part of the identification method's chain of trust: who can add or edit a reference spectrum, whether every match against it is captured in an audit trail, and whether the library itself sits on IQ/OQ/PQ-qualified software all become questions with the same weight as running the instrument correctly on the day.
B1.2 USP ⟨1119⟩: near-infrared, fast but less specific
Requirement USP ⟨1119⟩, “Near-Infrared Spectroscopy,” covers the theory and instrumentation behind rapid, non-destructive identification — NIR needs little or no sample preparation and can often read straight through packaging, which is why it shows up in raw-material receipt and 100%-of-container ID programs. FDA's “Development and Submission of Near Infrared Analytical Procedures” guidance (finalized August 2021) explicitly covers NIR methods submitted for identity, strength, quality, purity, and potency evaluation, building on the same Q2(R1)/(R2) validation expectations as any other method.
Practice NIR's overtone-and-combination-band spectra are real and specific to a compound, but they are broader and less sharply resolved than mid-infrared fingerprint-region spectra. That is a genuine trade — speed and non-destructiveness against resolving power — not a flaw, as long as the method's validation matches the specificity claim actually being made.
B1.3 Why retention time alone is a weak sole identification test — worked as a real number
Requirement ICH Q2(R2) §3.1.2.1 states the identification requirement directly: “The identification test should be applied to materials structurally similar to or closely related to the analyte to confirm that a positive result is not obtained.” Separately, 21 CFR 211.84(d)(1) requires at least one test to verify a component's identity, and that “specific identity tests, if they exist, shall be used.” Retention time, used alone, generally does not clear that bar — and the reason is worth working as a real, computed comparison rather than just asserting it.
The scenario: a target compound has a retention time of 6.20 minutes. Practice A common convention allows a ±2% relative tolerance window around that time (6.08–6.32 minutes) for a peak to count as a match. The genuine reference standard, run again, lands at 6.21 minutes — comfortably inside the window, as expected. But a structurally related substance — sharing much of the same backbone, differing at one or two functional groups — happens to elute at 6.26 minutes, which is also inside that same window. Checking retention time alone would accept it as a match. It is not the same substance.
The full infrared spectrum tells a different story, because it carries far more information than one number. Computing the Pearson correlation coefficient between the target's spectrum and the reference standard's spectrum gives r = 0.997 — a near-perfect match, as expected for the same compound run twice. The same calculation between the target and the structurally related interferent gives r = 0.658 — well below a 0.98 matching threshold, correctly flagging it as a different substance, because the two spectra visibly diverge at several fingerprint-region peaks that a single retention-time check could never see.
Practice This is the computed version of exactly what ICH Q2(R2) §3.1.2.1 asks for: challenge the identification test against something structurally similar, and confirm it does not produce a false positive. A retention-time-only method was never challenged this way — it has no mechanism to fail that challenge, because it was never given enough information to distinguish the two substances in the first place.
None of this means retention time is useless — it is a fast, cheap, useful screening signal, and it remains part of a proper identification package (a diode-array UV spectrum collected alongside the retention time, for instance, adds back real specificity). The point is narrower and more specific: retention time by itself, with no spectral or orthogonal confirmation, is not a specific identity test in the sense 21 CFR 211.84(d)(1) requires, because coincidental matches like the one worked above are a real, quantifiable risk rather than a theoretical one.
Module B1 — spectroscopic and chromatographic identification
Six questions.
Case study: when a Certificate of Analysis replaces an identity test
Three warning letters, one recurring failure, and a contamination wave with real casualties. This is a deliberately brief callback to the Related Substances course's own DEG history, not a re-telling of it.
Between 2023 and 2025, FDA cited multiple firms for a variant of the same finding: trusting a supplier's paperwork instead of running their own identity test on incoming glycerin or propylene glycol — components that, when contaminated with diethylene glycol (DEG) or ethylene glycol (EG), have caused fatal poisonings going back to the 1937 Elixir Sulfanilamide disaster and recurring in the 2022–2025 wave covered in the Related Substances course. This module does not re-tell that history; it applies Arc B's identification framework to three of the firms cited during it.
B2.1 Lex Inc. — no specific identity test for DEG/EG contamination
“You lacked sufficiently specific identity tests for potential diethylene glycol (DEG) and ethylene glycol (EG) contamination of glycerin and propylene glycol ingredients used in the manufacturing of drug products.”Lex Inc. — Warning Letter, August 17, 2023. fda.gov
The same letter cites the firm for accepting supplier Certificate of Analysis (COA) results “instead of testing each component lot.” Read against Arc B1: a COA is, at best, a report of a test someone else ran. It is not an identity test the receiving firm performed, and 21 CFR 211.84(d) does not treat those as interchangeable.
B2.2 Intercos Europe S.p.A. — an unqualified supplier and a missing test
“You also relied on your suppliers' certificate of analysis (COA) without establishing the reliability of their test analyses at appropriate intervals… you lacked an identity test to detect diethylene glycol (DEG) and ethylene glycol (EG) in all containers of all lots of glycerin before determining acceptability.”Intercos Europe S.p.A. — Warning Letter, August 15, 2024. fda.gov
Practice This letter adds a second layer worth naming: even a firm that occasionally verifies a supplier's COA against its own testing has to do so “at appropriate intervals,” not once and never again. A supplier's reliability is not a permanent property established at qualification and never revisited — it is itself a claim that periodically needs re-checking, the same way a method's own performance needs re-verifying when anything about it changes.
B2.3 Medinatura New Mexico, Inc. — a non-specific test standing in for a specific one
“You failed to perform adequate identity testing on each shipment of each lot of incoming components… you relied on your suppliers' certificates of analysis (COAs) without establishing the reliability of your component suppliers' test analyses at appropriate intervals.”Medinatura New Mexico, Inc. — Warning Letter, December 10, 2025. fda.gov
Requirement What makes this letter the sharpest illustration of Arc B1's point: the firm's identity testing was limited to specific gravity — a single physical measurement, not chemically specific to glycerin at all, let alone capable of detecting DEG or EG contamination. It is the real-world, higher-stakes version of this module's retention-time example: a test that produces a plausible number, passes routinely, and was never capable of catching the actual failure mode it needed to catch.
All three firms share the same underlying gap Arc B1 built toward: a test (or no test at all) that could not have distinguished the genuine article from a dangerous look-alike, because it was never challenged — or never designed — to make that distinction. A Certificate of Analysis describes what someone else's method found. It does not, and cannot, stand in for a method that was validated in your own hands to actually detect the thing you need to rule out.
Module B2 — the DEG/EG case studies
Six questions.