Related Substances and Unknown Impurities

Seven incidents, seven different reasons the tool available at the time missed something real. That is not a coincidence worth explaining away — it is the pattern this module names directly, before asking what a firm actually owes once a better tool exists.

Figure D1.1

D1.1  Not one story, repeated

A tidy version of impurity history says detection has simply gotten better and better, and each incident is the same story retold. Figure D1.1 argues against that. No method existed at all in 1937. A peak was seen and left unidentified in 1989. The right test ran at the wrong point in the process for Viracept. A compendial assay was gameable, not insensitive, for heparin. Nitrosamines sat below what routine HPLC-UV purity methods are built to see at all. AZBT was a process-related analogue nobody had reason to build a method for until it existed. Diethylene glycol is the oldest, best understood failure mode of the seven, and it still recurred in 2022–2025.

Practice The practical consequence: a quality system built to defend against “the next nitrosamine” specifically will still miss the next heparin, because the two failures have almost nothing in common except that both were, at the time, invisible. What generalises is not a single fix. It is the discipline of asking, for a given method, exactly what it cannot see — and this course has already shown FDA asking that question in the present tense, not as history:

“Your implemented method was unable to detect multiple unknown impurities that your CTL’s validated method identified.”Almon Healthcare Private Limited — Warning Letter 320-26-102, 13 July 2026. fda.gov

Read that one carefully, because it is a different failure from the seven in Figure D1.1. Almon's own contract testing laboratory already had a method that worked. This was not a case where no tool existed yet, or where the science had not caught up — a better method was sitting one phone call away and the firm's own implemented method fell short of it. That is a validation gap 21 CFR 211.192 already reaches, not the harder question this module is actually about.

D1.2  What actually reopens a marketed product

Requirement The harder question is this: if a firm's own process has not changed, and no contract lab is sitting on an unused better method, does the mere existence of improved instrumentation somewhere in the world create an obligation to re-examine an approved, unchanged product? ICH M7(R2) answers narrowly, in language written specifically to close off a broader reading:

“Application of this guideline may be warranted to marketed products if there is specific cause for concern. The existence of impurity structural alerts alone is considered insufficient to trigger follow-up measures, unless it is a structure in the cohort of concern…a specific cause for concern would be new relevant impurity hazard data…generated after the overall control strategy and specifications for market authorization were established…Similarly, a newly discovered impurity that is a known Class 1 or Class 2 mutagen that is present in a marketed product could also be a cause for concern.”ICH M7(R2), §4.4, Other Considerations for Marketed Products. source

Figure D1.2

Practice Two things trigger a required re-evaluation: new hazard data classifying an impurity as Class 1 or 2 mutagenic, or the discovery of an impurity that is itself already known to be Class 1 or 2. A structural alert with no cohort-of-concern status does not. Neither does a laboratory simply being able to resolve a peak that used to be noise. This is the same shape of gate Arc C built for change control and recall — a specific, named trigger, not a general one — applied here to the question of whether better instruments alone reopen a file.

That narrowness is not a loophole. It is what makes the obligation administrable at all: without it, every incremental improvement in laboratory instrumentation anywhere would create a standing duty to re-examine every approved product everywhere, on a timetable set by nobody. What it does not do is settle the professional question Module D3 returns to — a firm can be correct that nothing legally required it to look, and still have a defensible answer to why it did anyway.

Five techniques, five different reasons a conventional HPLC-UV method can miss something real. Each is tied here to a failure mode this course has already built, so the connection is concrete rather than a list of instrument names.

Figure D2.1

D2.1  Five blind spots, named

LC-HRMS. Arc B was explicit that nitrosamines sit below what routine purity methods can see, and named the fix directly:

“Typical routine tests (e.g., high performance liquid chromatography) for API purity, identity, and known impurities are unlikely to detect the presence of nitrosamine impurities.”FDA, Control of Nitrosamine Impurities in Human Drugs, Revision 2, September 2024. source

High-resolution mass spectrometry does not depend on a chromophore or a fixed detection wavelength at all — it identifies a peak by accurate mass and elemental composition, which is exactly what a purity method built around UV absorbance cannot do for a peak with no chromophore in the first place.

CAD. Arc A's own worked example depended on an assumption: that every component responds to a UV detector “in proportion to its mass, identically to the drug substance” — an assumption that holds only when the impurity shares essentially the same chromophore as the parent. Charged aerosol detection removes the assumption rather than working around it: it nebulises column effluent into an aerosol and measures particle charge, a physical principle with no dependence on light absorption, so a non-chromophoric compound — invisible to UV — is not invisible to a CAD.

2D-LC. A trace impurity co-eluting under the tail of a much larger peak can be lost to dynamic range in a single separation, the same geometry problem behind the two-panel expanded views this course has used throughout. A second, differently-selective chromatographic dimension moves the major component out of the way, so a minor component hidden in its tail becomes its own resolved peak.

qNMR. Every quantitation method this course has built depends on a reference standard or a relative response factor calibrated against one. qNMR does not: its signal is proportional to the number of magnetically equivalent nuclei, a fact of physics rather than a calibration curve, so it can quantify an impurity's purity without an authentic reference standard of that specific impurity — useful exactly when a newly discovered impurity has no standard yet to be had.

IMS. A structural isomer or a co-eluting species that shares the same mass is invisible to plain mass spectrometry, which resolves by mass-to-charge ratio and cannot distinguish two ions with the same formula and different shape. Ion mobility spectrometry adds a dimension mass spectrometry does not have: it separates ions in the gas phase by collision cross-section — effectively, shape — resolving species that are otherwise indistinguishable. AZBT and its structurally close relatives are exactly the kind of near-isomeric impurity family this distinction matters for.

D2.2  The framework keeps expanding to meet the frontier

Practice None of this is unprecedented from the regulatory side. The frameworks this course has taught — Q3A and Q3B's thresholds, Q3C's residual solvents, Q3D's elemental impurities, M7's mutagenic impurities — were each written to cover a category of impurity the previous guidelines left out.

Figure D2.2

The newest entry is still being written. ICH Q3E, covering extractables and leachables — chemicals that migrate into a drug product from packaging and delivery-device components rather than from synthesis or degradation — reached Step 2 draft status in 2025, with Step 4 finalisation targeted for 2027. It is not evidence that the system is behind. It is the same pattern this course has taught since Arc A, continuing in real time: the boundary of what is governed keeps moving to meet where detection and manufacturing complexity actually are.

A gap this predictable is not a defect to be regulated away. It is a permanent feature of a field where the tools improve faster than any guideline can be written and ratified, and it is where a quality unit's own judgement — not a checklist — does the remaining work.

Figure D3.1

D3.1  The incentive not to look

Requirement Name the incentive plainly, because a training programme that pretends it does not exist is not a credible one: a firm that never runs a more sensitive method than the one its specification requires cannot be cited for what that method would have found. Nothing in Module D1's analysis is a licence to exploit that fact. M7(R2) §4.4's narrow trigger describes what the law compels; it was never framed in this course as a description of what a defensible quality culture looks like.

Practice The distinction this course has returned to since Arc A is control, not evasion. A firm that periodically qualifies newer analytical technology against its existing methods, documents what it finds, and uses that finding to decide — with a documented rationale, the same standard Module C2 built for a specification — whether the existing control strategy still holds, is doing exactly what Figure D3.1's gap calls for. A firm that avoids qualifying newer technology specifically so that it never has to make that decision has not avoided an obligation. It has avoided finding out whether one exists.

D3.2  The clock is a preview of the argument

Module C4 already made this course's clearest version of the point, in a narrower context. EMA's guidance on nitrosamine corrective action anchors the three-year implementation clock to the date an acceptable intake was first published, and says explicitly that a later, more restrictive revision does not move that anchor. A firm cannot buy itself time by waiting for the number to settle. The same logic applies one level up, outside any single guideline's text: waiting for a framework to catch up to what current technology can already show does not pause the professional question. It only delays when the firm itself finds out the answer.

Practice None of this converts a discretionary decision into a hidden regulation, and this course will not pretend otherwise. A firm that reasonably concludes, on a documented basis, that its current methods remain fit for purpose has made a legitimate quality decision — the same kind Module C2 taught for a specification's justification. What is not defensible is never asking the question because the answer might be inconvenient. The difference between those two firms is not visible in either one's specification file. It is visible only in whether the review happened at all, and whether it was documented when it did.