ICH Q11 Doesn't Want Your Specification List — It Wants the Justification Behind It

Ask a CMC team what their control strategy is, and most will hand over a specification table: assay range here, impurity limit there, a dissolution or particle-size cutoff further down. That table is not a control strategy — it is a list of tests. ICH Q11, Development and Manufacture of Drug Substances, asks a harder question underneath it: why does each control sit exactly where it does, and not somewhere else in the process? Q11 picks up once ICH Q7's starting-material boundary is settled — it governs everything from that starting material to the finished drug substance, and it expects the answer to be process understanding, not habit.

Where Q11 picks up from Q7

ICH Q7 answers a boundary question: which step in a synthesis or fermentation counts as the point where GMP begins, because the material introduced there becomes a significant structural fragment of the finished drug substance. That question has to be settled first, because it decides where Q11's scope starts. Q11 does not re-litigate the starting-material boundary — it takes over from it, covering the manufacturing process description, the identification of steps and parameters that matter, and the control strategy that ties the whole thing together. Treating Q7 and Q11 as the same question is a common shortcut, and it produces a development package that answers only the first of the two questions a reviewer will ask.

What a control strategy actually has to show

Q11 defines the control strategy as the planned set of controls, derived from current product and process understanding, that assures process performance and product quality. That definition does the work: it is a system, assembled from a specific evidence base, not a specification sheet inherited from an earlier phase of development. Four kinds of controls make up that system, and Q11 expects each one to be placed where the process understanding says it belongs:

• Material attributes. Properties of raw materials, starting materials, and intermediates that the risk assessment shows actually influence the drug substance's quality.
• Process parameters. Sorted by the Q9 risk assessment into what actually needs tight control versus what can be managed more loosely without affecting quality.
• In-process tests and specifications. Placed where testing adds real information about the process's performance — not duplicated at every step by default.
• Procedural and facility controls. Training, equipment qualification, and environmental controls that hold the rest of the system together.

A specification tells a reviewer where testing happens. A control strategy has to tell them why it happens there, and nowhere else — and Q11 expects the sponsor to show that work, not assert it. Why a specification list is not a control strategy

Traditional versus enhanced — and the design space trap

Q11 does not force a single development philosophy. The traditional approach sets fixed process parameters and acceptance ranges based on a more limited data set, with less mechanistic justification demanded in return — the sponsor accepts a narrower, more literal operating range in exchange for a lighter development burden. The enhanced approach invests more up front: a deliberate, quality-by-design effort to build multivariate process understanding, often supporting a design space that gives room to move within a defined, pre-justified boundary without a new submission. Sponsors are free to mix the two across a single process — enhanced for the steps where flexibility pays for itself, traditional for the steps where it does not — provided each choice is justified on its own terms rather than applied as a single label to the whole filing.

The trap sits inside the enhanced approach, not outside it. A design space claim needs multivariate data — studies that vary combinations of parameters together and show how the combination affects the resulting quality attributes, with the proposed boundary independently verified. A set of one-factor-at-a-time studies, each varying a single parameter while holding the others fixed, does not generate that evidence, no matter how many of them a team runs. Sponsors who assemble a design space claim from single-factor data after development is essentially finished are treating design space as a label to apply, not a boundary to prove — and that gap is exactly what a reviewer's questions will find.

None of this is exotic once the sequencing is right: settle the Q7 boundary, run the risk assessment, decide the development approach step by step, and plan any design space claim as the multivariate study it has to be from day one rather than a label added at filing. Companies that treat Q11 as a documentation exercise — writing up whatever specifications the lab already had — pass an internal review and still collect an information request. Companies that treat it as the product-and-process understanding question it actually is are the ones whose control strategy survives the reviewer's second question, not just the first.