FDA KASA Technical Architecture — How the Knowledge-Aided Assessment Works

FDA KASA Technical Architecture — How the Knowledge-Aided Assessment Works

FDA’s Knowledge-Aided Assessment and Structured Application system — KASA — is the structured, risk-based quality assessment platform that Office of Pharmaceutical Quality reviewers now use in place of freestyle narrative review, and understanding how it is actually built, and where it connects to the separate PQ-CMC structured-data-exchange standard, helps pharmaceutical companies understand what their submissions need to achieve to benefit from the faster, more consistent review that KASA is designed to deliver.

That framing matters because the most common mischaracterization of KASA I encounter in CMC teams is one of two extremes: either dismissing it as a distant future-state system irrelevant to current submission planning, or describing it as an automated engine that directly parses sponsor-submitted FHIR files and issues findings without reviewer involvement. Neither description is accurate, and both lead to the same operational mistake — failing to design CMC submissions, narrative and structured alike, with KASA’s actual functional requirements in mind.

KASA is a data-based platform for structured quality assessments that supports FDA’s knowledge management across a product’s lifecycle. It captures and manages CMC knowledge, applies predefined risk-assessment algorithms, performs computer-aided analyses across FDA’s repository of prior assessments, and replaces narrative summarization with a structured assessment format. It is not a black box, and it is not an autonomous decision engine — reviewers, not KASA, make the regulatory determination. Its value to both FDA and sponsors depends on the completeness and quality of the CMC information a reviewer has to work with, and, increasingly, on the completeness and conformance of the structured PQ-CMC data submitted alongside it. Understanding what KASA actually does — and how it differs from the PQ-CMC data standard sponsors submit against — is the prerequisite for understanding what your structured CMC submissions need to deliver.

What KASA Does: Structured, Risk-Based Assessment — and How PQ-CMC Data Feeds It

KASA does not have a single “five-component ingestion-to-registry” architecture as FDA has documented it. What FDA has consistently described, from the 2018 Pharmaceutical Science and Clinical Pharmacology Advisory Committee vote through the December 2023 CDER Digital Transformation Symposium, are four core objectives and a Structured Review Framework built around them. Sponsors who understand that framework understand exactly what a KASA-informed review actually evaluates — and why data quality, not file format alone, is what determines whether a submission benefits from it.

The first objective is knowledge capture and management across the product lifecycle. KASA’s Structured Review Framework organizes an application’s foundational entities — drug substance, drug product, manufacturing, facilities, biopharmaceutics — and links each assessment to the precedent assessments FDA has already completed. As of FDA’s December 2023 program update, OPQ reported more than 1,000 ANDA reviews completed in KASA across drug product, drug substance, manufacturing, facilities, and biopharmaceutics disciplines. Rather than a reviewer reconstructing what specification ranges or stability precedents apply to a comparable drug substance class from memory or a prior narrative review, KASA surfaces that history in structured, searchable form.

The second objective is risk-based structured assessment. KASA applies a predefined ranking table together with a failure mode, effects, and criticality analysis (FMECA) approach to estimate the inherent risk associated with a given critical quality attribute, and then captures — through structured dropdown fields rather than narrative paragraphs — which control strategy the applicant used to mitigate that risk (product design, in-process control, release and stability testing) and what residual risk remains after that control strategy is applied. This structured risk capture, not automated document parsing, is the mechanism FDA has pointed to as replacing freestyle narrative review with a decision architecture reviewers can compare across applications.

The third objective is computer-aided, cross-application analysis. Because every assessment is entered into the same structured fields, KASA’s analytics layer can compare a pending application’s control strategy against every comparable control strategy already captured in the system — surfacing, for example, a manufacturing facility with no demonstrated capability for a proposed unit operation, or a specification that is an outlier relative to its peer group of approved products at the same facility. This cross-application, cross-facility comparison is the specific efficiency gain FDA has cited for KASA over the pre-2021 narrative-PDF review model, and it is not achievable by a reviewer working application-by-application in isolation.

The fourth objective is the structured assessment interface itself — the KASA Assessment Dashboard and its underlying Drug Product, Manufacturing Integrated, Biopharmaceutics, and, since the KASA 4.0 release in 2023, Drug Substance assessment modules. Reviewers work inside these modules rather than a Word document, with a side-by-side audit view that color-codes exactly what changed, by revision, across an application’s review cycle — a lifecycle-tracking capability FDA built into the platform from its earliest 2021 release rather than adding later. KASA does not replace the CMC reviewer; the reviewer applies scientific and regulatory judgment to the structured risk picture KASA presents, confirms or challenges the algorithmically suggested risk ranking, and makes the final determination.

It is important for sponsors to keep two related but architecturally distinct FDA initiatives straight, because the article’s framework box below depends on the distinction. KASA — live for generic solid oral dosage form ANDAs since February 2021, expanded to drug substance assessment in 2023, and on FDA’s published roadmap to extend to INDs, NDAs, BLAs, and post-approval supplements through 2025–2027 — is the internal structured-assessment platform FDA reviewers use. PQ-CMC, the separate HL7 FHIR Implementation Guide effort now published as version 2.0.0 (STU2), is the sponsor-facing data standard defining how CMC content should be structured for electronic submission; it is currently scoped to solid oral dosage forms and, in its published stages, covers drug substance and drug product general information, control of materials, specification, composition, batch formula, and substance characterization content. The two programs are explicitly connected in FDA’s own stated architecture: OPQ’s published vision is that structured PQ-CMC data submitted in Module 3 will eventually populate the CMC review template directly, reducing the manual re-entry that today separates a sponsor’s eCTD submission from a reviewer’s KASA assessment. That direct data flow is not yet fully automated, and FDA has been explicit that additional CTD Module 3 subdomains — including drug product and drug substance manufacturing process content — remain under active development rather than published. Sponsors preparing PQ-CMC structured data today should treat FHIR Bundle completeness and profile conformance against the current published stages as building the on-ramp to more efficient KASA-informed review, even while the direct machine-to-machine ingestion path continues to mature.

The practical implication for sponsors is direct. Data completeness and internal consistency are submission quality metrics — not technical compliance details separate from scientific content. A PQ-CMC FHIR Bundle that is scientifically accurate but structurally incomplete presents the reviewer with the same problem a narrative submission with missing sections presents: the information needed to populate the structured risk assessment is not available in usable form, and the reviewer must reconstruct it manually or issue a query. Every FHIR profile validation failure, every acceptance criterion submitted as an unstructured text string instead of a discrete operator-value-units triplet, and every controlled-terminology mismatch is a data gap that today still requires manual reviewer work-around, and that FDA’s stated future-state architecture is designed to eliminate as the PQ-CMC IG’s scope expands.

FDA’s PDUFA VII Performance Goals and Procedures for FY 2023–2027 include several CMC-specific efficiency commitments — a “Four-Part Harmony” structure for CMC information requests, the CMC Development and Readiness Pilot (CDRP) for accelerated programs, and a strategy document on advancing innovative manufacturing technology — but the published goals letter does not name KASA or PQ-CMC as an explicit PDUFA VII deliverable. The connection between KASA/PQ-CMC modernization and PDUFA-era review efficiency is real but indirect: both are OPQ initiatives aimed at the same underlying goal of faster, more consistent CMC review, and sponsors should not assume a specific PDUFA VII milestone is contingent on their PQ-CMC data quality. What is directly within a sponsor’s control is whether the structured data they submit is complete and conformant enough to support the KASA-informed review FDA’s reviewers are already conducting.

Structured data quality determines how much manual reconstruction a KASA reviewer has to perform. And how much manual reconstruction a reviewer has to perform determines whether your submission benefits from the more consistent, better-documented review KASA was built to deliver.

Before your next PQ-CMC structured data submission, run your FHIR Bundle through the HL7 FHIR validator loaded against the current PQ-CMC Implementation Guide package — identify every profile validation error and resolve each before submission, because unresolved validation errors become reviewer work-arounds and information requests that add weeks to your CMC review timeline.