Specification Setting — ICH Q6A Applied to Real CMC Submissions

The most common reason FDA issues a Complete Response Letter for CMC deficiencies related to specifications is not that the acceptance criteria are wrong — it is that the submission does not contain the scientific justification for why those acceptance criteria are right.

That sentence is not a rhetorical device. It describes the factual pattern that emerges when you read FDA Complete Response Letters with CMC deficiencies involving 3.2.S.4.1 and 3.2.P.5.1. The acceptance criteria themselves are often defensible. The assay range is within ±5.0% of label claim. The dissolution criterion is set at Q ≥ 80% at 45 minutes in pH 6.8 phosphate buffer. The water content limit is consistent with the solid-state form. The degradation product limit is at or below the ICH Q3B(R2) identification threshold. Taken in isolation, none of those criteria would fail a technical review. But the submission does not show why — not the why that ICH Q6A requires, and not the why that FDA’s chemistry reviewers are trained to look for when they open the pharmaceutical development section and the Quality Overall Summary.

ICH Q6A, finalized under the ICH Step 4 process on October 6, 1999 and adopted by FDA as a Guidance for Industry under a Federal Register notice published December 29, 2000, is the foundational guideline for setting specifications for new chemical entities and new dosage forms. It is twenty-seven years old. It has been supplemented by ICH Q8(R2) on pharmaceutical development, ICH Q9(R1) on quality risk management (revised and adopted under Step 4 on January 18, 2023), and by FDA’s own body of specification-related guidance for industry — including ANDAs: Impurities in Drug Products, ANDA Submissions: Refuse-to-Receive for Lack of Justification of Proposed Specifications, and the dissolution-specific guidance for immediate-release solid oral dosage forms — which collectively consolidate agency thinking on justification documentation. Despite more than a quarter century of regulatory infrastructure, the gap between what Q6A requires and what most submissions provide remains the most consistent source of first-cycle specification deficiencies. Understanding why requires understanding what Q6A’s justification hierarchy actually demands — tier by tier — and then understanding what it means to apply those demands to each acceptance criterion in a real specification table.

The ICH Q6A Justification Hierarchy: Why All Three Tiers Matter

ICH Q6A does not instruct applicants to set acceptance criteria. It instructs applicants to justify them. The distinction sounds semantic. In practice, it defines the difference between a specification table that survives first-cycle review and one that generates a list of deficiencies.

The Q6A justification hierarchy operates on three ascending tiers, and the hierarchy is not a menu of options — it is a ranked framework in which the tier selected for a given criterion must be the highest tier for which supporting data exists. The first tier, and the highest in the hierarchy, is justification derived from safety and efficacy data. When an acceptance criterion can be linked to a clinical dataset — when there is a pharmacokinetic bridge between bioavailability and dissolution behavior, when a clinical study established the exposure threshold that defines the lower bound of the assay range, when a clinical adverse event profile established the impurity threshold that defines the upper bound of a degradation product limit — that linkage is the most defensible justification available, and ICH Q6A expects it to be documented explicitly. The criterion should reference the study number. The narrative should explain the relationship between the dataset and the criterion. The Quality Overall Summary should cross-reference the pharmaceutical development section where the data appears. This is not a theoretical standard. When FDA reviewers find a dissolution criterion of Q ≥ 85% at 30 minutes in a modified-release formulation and the submission contains an IVIVC established across the Phase 2 and Phase 3 dose range, the review is straightforward because the data speaks directly to the criterion. When the same criterion appears in a submission where the IVIVC section of 3.2.P.2 is absent or underdeveloped, the reviewer has no basis for accepting that criterion as adequately controlled.

The second tier is justification derived from compendial standards. USP, the European Pharmacopoeia, and the Japanese Pharmacopoeia establish criteria for drug substances and drug products that are the product of collaborative scientific review conducted by experts whose standard of evidence exceeds what any single sponsor is required to provide in a new drug application. When a compendial monograph exists and is applicable, Q6A anticipates that its criteria will be incorporated. This is not a mechanical exercise. The applicant must demonstrate that the compendial test is appropriate for the specific product — that the USP dissolution method, if adopted, is discriminating for the formulation at hand, that the compendial assay procedure has been shown fit for purpose for the drug substance as synthesized. But when those conditions are met and the compendial standard is adopted, the justification tier is satisfied by the reference. What sponsors fail to provide, frequently, is the demonstration that the compendial test is applicable — not because that demonstration is technically difficult, but because it requires a documented comparative analysis that sits in 3.2.P.2 and is cross-referenced in the specification section, and that cross-reference is routinely absent. USP <1>, the general chapter on injections, and the broader compendial reference framework establish that universal applicability cannot be assumed. Applicability must be shown.

The third tier — justification derived from analytical capability and batch history — is where most specification tables actually live, and it is the tier that Q6A circumscribes most carefully. Capability-based justification is legitimate for criteria that are not safety-critical. For a drug product appearance criterion, for a drug substance description limit, for a residual solvent criterion that is well below the ICH Q3C permitted daily exposure — in these cases, setting a criterion that reflects manufacturing capability demonstrated across representative batches, supported by statistical analysis of the batch distribution, and confirmed to be achievable under the validated process is appropriate Q6A practice. The ICH Q8(R2) framework on design space and the ICH Q9(R1) framework on risk-based prioritization both support this approach. What those frameworks do not support — and what Q6A explicitly prohibits at this tier — is using capability alone to justify criteria for safety-critical attributes. Assay range cannot be justified solely on the grounds that commercial batches have historically fallen between 97.0% and 103.0%. Degradation product limits cannot be justified solely on the grounds that no batch has exceeded 0.08% under ICH Q3B(R2). When the attribute has a direct patient safety implication, the justification must reach the first or second tier. Using third-tier capability data as the primary justification for a safety-critical criterion is not a documentation shortcut. It is a submission gap.

The operational consequence of this hierarchy is that specification development must be synchronized with clinical and analytical development from Phase 1 onward. The data that justifies first-tier criteria comes from clinical studies. If the clinical study protocol does not capture the exposure-response data necessary to link a dissolution criterion or an assay range to a pharmacokinetic outcome, that data cannot be recovered at the NDA stage. The submission will fall to the second or third tier for criteria that would have been more defensible at the first tier with prospective planning. This is the structural reason that ICH Q8(R2) frames pharmaceutical development as a prospective exercise — not a retrospective documentation exercise — and it is the reason that specification tables constructed from Phase 3 batch data alone, without a prospective justification framework, generate the deficiency pattern described at the outset of this article.

The documentation architecture for each criterion should follow a consistent structure. For each acceptance criterion in the specification table, the pharmaceutical development section — 3.2.P.2 for drug products, 3.2.S.2 for drug substances — should contain the proposed criterion, the supporting dataset or regulatory standard on which the criterion is based, the tier of the Q6A justification hierarchy that is being invoked, and an explicit narrative statement explaining why the criterion adequately controls the quality attribute in question. The Quality Overall Summary should replicate this structure in condensed form and cross-reference the development section. When FDA’s chemistry reviewers evaluate a specification, they read the specification table, then the QOS, then the pharmaceutical development section. A submission in which those three documents are internally consistent and each criterion can be traced through the Q6A hierarchy from the specification table to the supporting data is a submission that can be reviewed in the first cycle. A submission in which the specification table does not cross-reference to justification, the QOS summarizes criteria without tying them to data, and the pharmaceutical development section contains no structured criterion-by-criterion analysis is a submission that will generate deficiencies — not because any individual criterion is wrong, but because the architecture of justification is absent.

Applying the Q6A Decision Trees: What They Require for Polymorphism, Dissolution, and Impurities

ICH Q6A contains decision trees that are not advisory illustrations. They are documented analytical frameworks that, when applied to a specific drug substance or drug product attribute, produce an outcome with regulatory consequence. The decision trees for drug substance polymorphism, drug substance particle size, drug product dissolution, drug product disintegration, and residual solvents each require documented answers to specific questions, and the specification criterion that emerges from the decision tree analysis is only defensible if the submission shows the path through the tree.

The polymorphism decision tree begins with a question that many submissions answer implicitly rather than explicitly: does the drug substance exist in polymorphic forms? If the answer is yes — and for the majority of crystalline small molecules, systematic screening will reveal more than one form — the decision tree requires the applicant to determine whether the polymorphic form affects solubility, stability, or bioavailability. If it does, the specification must control polymorphic form, and the analytical method must be validated to distinguish the relevant forms. XRPD and DSC are the standard techniques; solid-state NMR and Raman spectroscopy are confirmatory. The specification criterion for polymorphic form — typically expressed as “Form I as determined by XRPD with a reference standard diffractogram” — must be supported by a development narrative that demonstrates the form was controlled in clinical batches, that the commercial manufacturing process produces the specified form consistently, and that method capability has been validated to distinguish the specified form from relevant polymorphic variants. Submissions that list a polymorphic form specification without the decision tree analysis — or that state in the pharmaceutical development section only that “polymorphic screening was conducted” without disclosing the forms identified, the forms observed in clinical batches, and the outcome of the decision tree — will receive a deficiency asking for the complete analysis.

The dissolution decision tree for immediate-release solid oral dosage forms requires the applicant to establish first whether the drug substance has a solubility that could limit bioavailability — the BCS classification exercise. For BCS Class I compounds, where solubility is high and permeability is high, dissolution is expected to be rapid and complete, and Q6A allows a simpler dissolution specification. For BCS Class II and IV compounds, where solubility limits absorption, the dissolution method must be discriminating — it must be capable of detecting formulation variables or manufacturing changes that would affect in vivo performance. A dissolution method that releases 100% of every batch within fifteen minutes in a BCS Class II compound is not discriminating and does not protect against formulation failures that would produce inadequate bioavailability in patients. FDA has issued deficiencies on this basis: “The proposed dissolution method does not appear to be discriminating. Provide data demonstrating the method’s ability to detect formulation variables relevant to in vivo performance.” The dissolution criterion itself — the Q value, the time point, the medium — must emerge from the decision tree analysis and must be connected to the solubility and permeability characterization in 3.2.P.2.

For impurity criteria, Q6A intersects with ICH Q3A(R2) for drug substances and ICH Q3B(R2) for drug products. The specification limit for any individual degradation product or process-related impurity must be justified at the appropriate Q6A tier. If the impurity has been qualified at the proposed limit through clinical exposure — patients in Phase 3 were exposed to the drug product at the proposed commercial manufacturing scale and the impurity was present at levels at or above the proposed specification limit with no adverse signals — that is first-tier justification. If the impurity is below the applicable ICH Q3B(R2) qualification threshold — which Attachment 1 of Q3B(R2) sets on a sliding scale by maximum daily dose, from 1.0% or 50 μg total daily intake below 10 mg, down to 0.2% or 3 mg TDI for the 100 mg–2 g dose range, and 0.15% above 2 g — and no safety concern has been identified, that pre-established ICH threshold itself provides the justification, without requiring product-specific safety data. Note that Q3B(R2)’s identification thresholds sit below and separate from its qualification thresholds — for the >10 mg–2 g dose range, the identification threshold is 0.2% or 2 mg TDI, while the corresponding qualification threshold one tier up is 0.2% or 3 mg TDI at the 100 mg–2 g dose range; conflating the identification figure with the qualification figure is a common drafting error that understates the qualification standard a sponsor must actually meet. If the limit is set on the basis of manufacturing capability and the impurity is below the threshold and has no structural alert, capability-based third-tier justification is appropriate. The error is in applying third-tier justification to an impurity that has a structural flag, or to a limit above the qualification threshold, without conducting the safety assessment that Q6A and Q3B(R2) together require.

The specification tightening obligation from Phase 3 to commercial launch is frequently overlooked. ICH Q6A expects that initial acceptance criteria, set on the basis of a limited number of clinical batches, will be tightened as commercial data accumulates. The Phase 3 batch data provides the first estimate of process capability and the first dataset for statistical derivation of acceptance criteria. As commercial batches accumulate — and the FDA guidance on drug substance and drug product specifications defines the minimum dataset that should inform commercial specification setting — the criteria should be revised to reflect actual commercial process performance. This is not a voluntary improvement. It is the expected lifecycle trajectory, and submissions for post-approval changes that propose to widen criteria that were initially set on commercial data, without a documented process change or scientific rationale, will be reviewed with heightened scrutiny. The lifecycle management program should establish, prospectively, the triggers for specification review: a defined number of commercial batches, a process capability analysis at scheduled intervals, a post-approval change that affects a quality attribute, and a systematic review of annual product quality reviews that identifies trending toward specification limits.

The XGene Specification Architecture and Justification Framework

At XGene, specification development follows a structured methodology that applies the ICH Q6A decision trees and justification hierarchy at the attribute level — not at the document level. The distinction matters. Most pharmaceutical development sections address specifications as a section-level activity: a chapter in 3.2.P.2 titled “Justification of Specifications” that covers the specification table in aggregate. The XGene approach begins at the attribute level during Phase 2, before the specification table is finalized, and builds each criterion from the decision tree up.

For each quality attribute — whether a universal test applicable to every drug product or drug substance of its type, or a specific test required by the Q6A decision tree analysis — the XGene framework documents four elements in sequence. First, the proposed acceptance criterion. Second, the supporting dataset: the clinical study, the compendial monograph, the batch analysis with statistical derivation, or the combination of these that supports the criterion. Third, the Q6A justification tier — safety/efficacy, compendial, or capability — and the explicit acknowledgment of whether the attribute is safety-critical and therefore cannot rely on capability alone. Fourth, the justification narrative: a direct, one-to-two-paragraph statement that explains why the proposed criterion adequately controls the quality attribute for patient safety and product quality.

For drug substance universal tests, this framework applies to description, identification, assay — where the ±5.0% of label claim convention must be connected to either clinical exposure-response data or a documented process capability analysis — impurity profile, water content, and residual solvents. For drug product universal tests, the framework applies to description, identification, assay with the 90.0–110.0% typical range justified at the appropriate tier, degradation products, dosage uniformity under USP <905>, and water content. For attribute-specific tests — dissolution in immediate-release and modified-release products, disintegration, hardness, polymorphic form, particle size in inhalation products — the decision tree path is documented alongside the criterion, and the exit point of the decision tree is explicitly cross-referenced to the analytical method validation and the batch data analysis.

The batch data statistical analysis component of the XGene framework addresses the capability-based third tier directly. For criteria where capability justification is appropriate, a process capability analysis is conducted on the available batch dataset using a tolerance interval approach — typically a 95% confidence, 99% coverage tolerance interval — so that the acceptance criterion reflects the underlying process distribution with statistical confidence rather than the minimum and maximum observed in a finite sample. The minimum-maximum approach, which many submissions use by default, does not reflect process capability: it reflects the range of results observed in a finite sample, which will expand as more batches are run. Statistical derivation using tolerance intervals produces a criterion that is defensible against the argument that additional commercial batches will fall outside the initially set range.

The lifecycle management program at XGene establishes specification review triggers at three points: after every fifty commercial batches, after any post-approval manufacturing change that affects a quality attribute, and as part of each annual product quality review cycle. At each trigger point, the batch data is analyzed for capability trend, the acceptance criteria are evaluated against current process performance, and a documented decision is made — with rationale — to maintain, tighten, or seek post-approval approval to modify the criteria. This creates a specification lifecycle file that, in the event of a post-approval submission or FDA inspection, demonstrates that the specification is not a static artifact from the original NDA but a living document managed in accordance with the ICH Q6A philosophy of fit-for-purpose criteria that reflect the evolving state of process understanding.

The output of the XGene Specification Architecture and Justification Framework is a specification table that functions as a cross-referenced document, not a standalone table. Each criterion carries an internal reference to the section of the pharmaceutical development report where the decision tree analysis and supporting data are documented. The Quality Overall Summary specification justification section mirrors this structure in condensed form. When FDA’s chemistry reviewers open the submission, the path from criterion to justification to supporting data is navigable in three document references. That navigability is what first-cycle review requires — not perfection in the criteria themselves, but transparency in the reasoning that produced them.