Continuous Manufacturing — FDA Quality by Design Expectations and the CMC Submission Package

Continuous manufacturing has crossed from demonstration project to regulatory expectation. The FDA has approved more than a dozen drug products manufactured by continuous processes, issued ICH Q13 as a finalized guideline, and integrated continuous manufacturing review competency into CDER’s Office of Pharmaceutical Quality. What has not changed is the quality of CMC packages that reach reviewers — many still treat continuous manufacturing as batch manufacturing with an uninterrupted conveyor, and the deficiency letters show it.

The strategic stakes are straightforward: a continuous manufacturing (CM) program that reaches NDA submission with a CMC package built on batch-process logic will not fail because the technology doesn’t work — the process performance data is often excellent. It fails because the submission cannot answer the specific regulatory questions ICH Q13 requires a CM package to answer, and those questions have no batch-manufacturing analog. Companies investing in CM as a manufacturing modernization strategy need a CMC regulatory architecture built for continuous processes from the start, not retrofitted onto a batch submission template after a deficiency letter arrives.

The ICH Q13 Framework and What FDA Has Actually Required in Approved CM Submissions

ICH Q13, adopted by the ICH Assembly in November 2022 and issued as final FDA guidance in March 2023, defines continuous manufacturing as a process in which input materials are continuously charged and output materials are continuously discharged — and it establishes two regulatory categories that determine how a CMC package must be structured: drug substance continuous manufacturing, covering integrated reaction, workup, crystallization, and drying operations, and drug product continuous manufacturing, covering continuous direct compression, continuous twin-screw wet granulation, and continuous hot-melt extrusion. The regulatory record establishes that Orkambi (lumacaftor/ivacaftor, Vertex Pharmaceuticals, NDA 206038, approved July 2015) was the first FDA-approved drug product manufactured by continuous direct compression, and the CDER chemistry review for that program established the residence time distribution characterization precedent and time-based batch definition approach that subsequent CM submissions have referenced since.

Symdeko (tezacaftor/ivacaftor, Vertex, NDA 210491, approved February 2018) extended that CM platform to a two-API combination product, and FDA review documents confirm that the batch definition approach for this second-generation program expanded to a mass-based boundary rather than the time-based approach used for Orkambi — evidence that FDA evaluates batch definition on a program-specific basis rather than applying a single universal rule. Separately, the public FDA enforcement and technology-adoption record documents Janssen’s 2016 transition of Prezista (darunavir) manufacturing from batch to continuous production at its Gurabo, Puerto Rico facility — the first FDA-approved batch-to-CM post-approval manufacturing change — cited in FDA Emerging Technology Program materials as the reference case for the site change supplement classification decision (CBE-30 versus prior-approval supplement) that CM transitions require.

The FDA Guidance for Industry on advancing emerging technology applications — issued in draft in December 2015 under the title “Advancement of Emerging Technology Applications to Modernize the Pharmaceutical Manufacturing Base” and finalized in September 2017 under the retitled “Advancement of Emerging Technology Applications for Pharmaceutical Innovation and Modernization” — established the Emerging Technology Program pathway that most CM sponsors still use today to engage CDER’s Emerging Technology Team before NDA submission. Public ETP meeting records confirm a documented engagement pattern of roughly two to three meetings, with an initial response timeline of about 60 days, before a CM-based NDA submission — a cadence that gives sponsors an opportunity to resolve RTD characterization and diversion logic questions with reviewers before they become deficiencies.

RTD, Diversion Logic, and Batch Definition — The Three CM-Specific CMC Elements Most Packages Get Wrong

Material traceability is the foundational CM critical quality attribute, and it is defined operationally by residence time distribution (RTD) characterization — the ability to identify which input material volume produced which output material volume at any point in the process. RTD is measured through tracer experiments, using a colored or UV-absorbing tracer injected as a step change or pulse, with the resulting mean residence time (tau) and variance (sigma-squared) characterizing the mixing and plug-flow behavior of the line; published continuous manufacturing characterization literature reports mean residence times of roughly 30 to 90 seconds for continuous direct compression lines and roughly 15 to 45 seconds for twin-screw granulation lines, with startup diversion volumes typically set at three to five times the RTD mean based on that published process characterization. The most frequent deficiency pattern at this stage is a 3.2.P.3 section that describes the line as a “continuous process” without ever defining what constitutes a releasable batch — leaving the reviewer unable to determine what unit of material the specification and release testing actually apply to.

RTD characterization is not an academic exercise; it is the technical foundation for diversion logic, and diversion logic is where FDA deficiencies concentrate most heavily. Diversion logic is the CMC regulatory obligation that arises when an in-process control — for example, an NIR prediction of API concentration falling below the acceptance criterion — triggers a computational model that uses the RTD characterization to back-calculate exactly which volume of output material was affected, so that volume can be diverted or rejected rather than the entire campaign. The deficiency pattern FDA reviewers cite most consistently is a submission in which this mathematical model is described narratively but never validated or included as a submitted computational tool — meaning the sponsor is asserting a diversion capability that the review division cannot independently verify.

Batch definition is the third element, and FDA requires sponsors to establish it before NDA submission, not derive it after the fact from however the process happened to run: the two accepted approaches are a time-defined batch, such as an eight-hour manufacturing campaign constituting one batch, or a mass-defined batch, such as 500 kilograms of output constituting one batch, and whichever definition is chosen must be applied consistently across every process validation batch. A recurring deficiency is a Stage 2 process performance qualification package that includes only three commercial-scale batches with final-batch summary statistics, when the continuous nature of the process means FDA reviewers expect trending data demonstrating process stability across the full duration of all three campaigns — not simply a pass/fail conclusion at the end of each run.

Real-Time Release Testing and Design Space for Continuous Processes — The Regulatory Evidence Architecture

Real-time release testing (RTRT) replaces traditional end-product testing as the primary release mechanism in most CM programs, and it requires three elements working together under ICH Q10’s process performance and product quality monitoring expectations: a PAT measurement method validated per ICH Q2(R2) analytical procedure validation principles, a multivariate calibration model — typically PLS or PCA — validated against reference laboratory methods, and an ongoing model maintenance protocol with defined update triggers. For blend and content uniformity, published RTRT practice uses NIR or Raman spectroscopy at the blend or granule stream with a relative standard deviation of 3.0 percent or better as the acceptance criterion for content uniformity prediction, with the PLS model’s acceptance criterion matched to the underlying specification release limit for API content. The deficiency pattern FDA cites here is an RTRT model submitted as an appendix with calibration statistics but no protocol for what triggers a model update or recalibration — leaving reviewers unable to assess whether the model will remain valid as raw material lots or equipment change over the product’s commercial life.

Design space for a continuous process is structurally different from a batch design space under ICH Q8(R2): it must capture process dynamics, not just steady-state ranges of critical process parameters, because a CM line has startup and shutdown transitions during which the process is, by definition, not at steady state, and the mathematical relationship between in-process control deviations and product quality must be defined across that entire operating window. The normal operating range must be explicitly justified as nested within the proven acceptable range, which is in turn nested within the overall design space, in the 3.2.P.2 pharmaceutical development section — and the deficiency FDA reviewers raise most often is a design space defined entirely from steady-state multivariate DOE work, with no data addressing whether off-specification material generated during startup can carry into, or contaminate, steady-state output.

FDA’s three-stage process validation framework from the 2011 Process Validation Guidance applies to CM programs, but each stage carries CM-specific requirements: Stage 1 process design must include RTD characterization and RTRT model development as deliverables, not just process parameter ranges; Stage 2 process performance qualification must demonstrate that the process produces conforming material across the entire defined batch boundary, not just at a single sampling point; and Stage 3 continued process verification for CM is inherently more rigorous than batch CPV because a continuous line generates continuous data streams that must feed real statistical process control charts, under 21 CFR 211.110’s in-process sampling and testing requirements, rather than the periodic batch-release sampling that governs conventional manufacturing.

The XGene Continuous Manufacturing CMC Architecture Building a 3.2.P.2/P.3 Package That Survives FDA Review

The XGene Continuous Manufacturing CMC Architecture is a structured CMC regulatory strategy framework built specifically for continuous manufacturing NDA and BLA submissions, integrating RTD characterization, diversion logic validation, RTRT strategy, batch definition, and design space documentation into the complete 3.2.P.2/3.2.P.3 package FDA reviewers expect.

Step 1 — RTD Characterization and Diversion Logic Validation Protocol: Design and execute tracer studies at startup, steady state, and campaign end to establish the mean residence time and variance for each unit operation, then build and validate the computational diversion model as a submittable tool — not a narrative description — that back-calculates affected material volume from any in-process control exceedance.

Step 2 — Batch Definition and Process Validation Architecture: Establish the time- or mass-defined batch boundary before process validation begins, and design the Stage 1 through Stage 3 process validation program so that RTD and RTRT model development are explicit Stage 1 deliverables and Stage 3 continued process verification is built on real-time SPC monitoring rather than periodic batch sampling.

Step 3 — RTRT Model Development and Regulatory Submission Strategy: Build the PAT measurement method, validate the multivariate calibration model against reference methods under ICH Q2(R2) principles, and document the ongoing model maintenance protocol with defined update triggers so the submission demonstrates lifecycle control of the release testing model, not just its initial validation state.

Step 4 — Design Space Documentation Across Process Dynamics: Define and justify the design space, proven acceptable range, and normal operating range not only at steady state but across startup and shutdown transitions, with explicit data addressing whether transitional material can affect steady-state product quality.

The output of the XGene Continuous Manufacturing CMC Architecture is a submission-ready 3.2.P.2/3.2.P.3 evidence package that maps each ICH Q13 regulatory expectation to a specific validated model, dataset, or protocol — not a gap list, but a close-out package built to survive FDA chemistry review on the first cycle.

Companies that treat continuous manufacturing as a manufacturing-floor initiative, without building the CMC regulatory architecture ICH Q13 actually requires, are deferring the hardest part of the program to the moment it is most expensive to discover: after a deficiency letter, on a PDUFA clock, with a review division that has already formed its impression of the sponsor’s technical readiness. The RTD, diversion logic, batch definition, and design space elements described here are not optional refinements to a batch-manufacturing CMC template — they are the specific evidentiary architecture FDA reviewers have told the industry, through every approved CM precedent and every published deficiency pattern, that a continuous manufacturing submission must contain. The cost of building that architecture correctly before submission is a fraction of the cost of a Complete Response Letter cycle built around resolving it after the fact.

For your continuous manufacturing program, can you identify today whether your 3.2.P.3 submission includes a validated RTD model with documented diversion logic tested against known process disturbances, and a defined batch definition with the rationale for the time or mass boundary that constitutes a releasable unit?