FDA Submission Automation: Coming To A Cloud Near You

By Matthew Pillar, Editor, Bioprocess Online

I was moderating a panel discussion on best practices for IND submission and clinical hold avoidance a few weeks back. At one point in the conversation, NDA Partners’ Daniela Drago, Ph.D. and Umoja Biopharma VP of Regulatory Affairs Helen Kim both emphasized the importance of expressing individuality and personality in the submission itself. Your submission, they concurred, is one of hundreds of complex documents an FDA team might review in a given quarter, and one of thousands it might review in a given year. It’s read by humans. Respect those humans. Make it readable, accessible, and digestible.

The point resonated with our audience. Comments during and feedback gathered after the event suggested the point couldn’t possibly be overstated. Particularly poignant was the affirmation of their suggestion from former FDA reviewers.

But then, a few days later I was talking with Brian Kirk, Ph.D. Lead Author, Global Regulatory Affairs CMC for Plasma-Derived Therapies at Takeda. Dr. Kirk and I were making plans for the upcoming Bioprocess Online Live panel discussion Data-Driven CMC For Process Development Speed and Regulatory Efficiency (February 28 at 11 AM ET, hit that link to register, it’s free). Kirk’s the guy responsible for writing a good portion of the submission doc types I discussed with Drago and Kim. He doesn’t disagree with their suggestion, but he’s here to tell you it’s hard. Really hard. These are big, data-intensive documents.

Unstructured Data = Inefficiency

If writing submission documentation is hard work, so is reviewing it, in no small part because there’s not much in the way of submission standardization. IND submissions are typically unstructured PDFs. Theoretically, they follow a common order as prescribed by the Agency:

• Table of Contents
• Introductory Statement and General Investigational Plan
• Investigator Brochure
• Clinical Components
  • Protocol
  • Summary of Previous Human Experience with the Investigational Drug
• Nonclinical Components
  • Animal Pharmacology and Toxicology (PT)
• Chemistry, Manufacturing and Controls (CMC)
• Other information as necessary

But how they’re written, and how critical PQ and CMC data is presented, leaves a lot of room for creative interpretation for both authors and reviewers. The Office of Pharmaceutical Quality (OPQ) reviews around 3,000 INDs, 240 NDAs/BLAs, 900 ANDAs, and 10,000 supplements annually. In the current unstructured PDF submission format, the bulk of those submissions contain freestyle narrative. That means applicants’ risk mitigation approaches, for instance, are dispersed in pages upon pages of text, making comparative assessment and evaluation against precedents a manual and tedious process for reviewers. When unstructured applications are evaluated in individual silos, there’s no efficient way for reviewers to make apples-to-apples comparisons to relevant approved drug products and facilities within the FDA’s repertoire.

Enter KASA: The Knowledge-aided Assessment & Structured Application

The FDA’s Knowledge-aided Assessment & Structured Application (KASA) initiative seeks to solve these problems. The initiative has many goals, the most immediate of them for biopharmaceutical companies being standardization of data packages submitted to the Office of Biotechnology Products (OBP). That’s the part of CDER’s OPQ that’s charged with review, regulation, and research of biological products and biosimilar biological products. While KASA has been live for oral solid dosage submissions since February 2021, its next initiative seeks to apply its standardized submission criteria to drug substances (for new and generic drugs) this year (KASA 4.0), and liquid-based dosage forms for generics, INDs, NDAs, and BLAs (KASA 5.0) over a span of three years beginning in 2024.

To do this, the Agency is working on structured electronic data standards that enable more efficient capture of critical information from sources including:

• Its Quality Surveillance Dashboard, a framework for consistent evaluation of facilities and potential quality signals within a product’s lifecycle,
• Its multi-disciplinary Integrated Quality Assessment team, which is charged with ensuring effective and efficient assessment of drug applications,
• ICH M4Q(R2), the most recent version of the ICH Common Technical Document on Quality Guideline,
• and PQ/CMC, for which the agency is developing electronic, structured standards and a data exchange standard for PQ/CMC data submission.

An approach to risk assessment that’s more systematic and standardized than it is manual and arbitrary, the Agency says, will improve the speed and quality of review and decision-making. On the back end, structured data submissions in KASA will yield automated assessments of the following by way of established rules and algorithms:

• Risk associated with how a product is designed and manufactured,
• Risk mitigation and control strategies through product design features and applied quality standards,
• and manufacturing controls and the demonstrated capability of facilities involved.

How effective the Agency will be at expediting review of the tens of thousands of electronic submission documents it will begin receiving via the cloud as the KASA initiative rolls out is dependent on the industry’s acceptance of and conformation to the new standards. That’s easier said than done in a dynamic biopharma development environment where one size won’t fit all.

Some Data Types Are Easier To Standardize And Monitor Than Others

While drug product risk assessment is a relatively easy nut to crack using structured data and AI, I anticipate it will be more difficult to consistently monitor facilities and manufacturing risks, such as a site’s capability to manufacture various dosage forms, CGMP history, and approved control strategies for unit operations. That skepticism is rooted in the new variables introduced every time tech is transferred, facilities are commissioned, and new biologic modalities are discovered and developed, all of which are happening at a mindboggling pace. Human relief valves will have to step in when users on both sides of the initiative realize “we don’t have a field for that.”

The global nature of today’s biopharmaceutical business will also present challenges to the KASA initiative. Efforts toward regulatory harmonization (the process by which technical guidelines are developed to be uniform across participating authorities) and regulatory convergence (the process whereby regulatory requirements across countries or regions become more similar or aligned over time) are slow, hampered by politics, legality, and regulatory maturity.

What About Differentiating Prose?

On one hand, the KASA initiative creates significant opportunities for efficiency gains. In a regulatory environment that moves at a snail’s pace by design, we can all agree that a reduction in needless delays in the review process would be welcome. Likewise, authors and architects of submission documents would likely embrace a more prescriptive approach. It’s not a stretch to imagine an automated data capture and entry paradigm, whereby PQ and CMC data once collected, copied, and pasted into submissions is populated in submission documents directly from smart processing and manufacturing systems. A single version of truth, after all, is the Holy Grail of data management, variability and human error be damned.

On the other hand, the only way to eliminate human error is to eliminate the human who caused it, and the agency has stated that it intends to “provide a structured assessment that radically eliminates text-based narratives and summarization of information from the applications.” That intent seems to fly in the face of intentional insertion of personality in FDA submissions, minimizing the filer’s communicative opportunities and creating concern about the reduction of submissions to mere electronic transactions. But then, how necessary is human language if submissions are eventually reviewed, assessed, and ultimately decided on by machines?

Preparing For KASA

With the KASA initiative moving forward with a full head of steam, there is some onus on biopharma companies to adopt digitization of PQ and CMC data in a format that makes data transfer to cloud-based submission documents seamless. There’s advantage in being better prepared than your competition to hand the baton (your submission) off to a runner (the Agency) that’s operating at a faster pace with new tools. To that end, the industry has a long way to go and a short time to get there. A very recent article in Quality Magazine reports that only 3 percent of life sciences companies self-assess as “intelligent” in the context of digitization adoption. Pay close attention to the KASA initiative as it rolls out to determine how its submission technologies and frameworks might influence your internal quality data collection and reporting infrastructure improvements. It seems the talking part’s done—the more seamless the transfer of digital data, the faster your timelines.

More detail on the KASA plan can be found in these slides presented at the Agency’s Meeting of the Pharmaceutical Science and Clinical Pharmacology Advisory Committee in November, 2022. Have a look, and in the meantime, let me know what you think about KASA and how you’re preparing for it. I’m at matt.pillar@lifescienceconnect.com.