Life Sciences / Regulatory Review 🧬

📌 Navigate

📋 Exec Summary

Q3 2022 established the defining structural dynamic of medical AI: regulatory throughput and capital markets decoupled entirely. FDA maintained steady AI/ML clearances while biotech capital markets hit their deepest modern drawdown (NBI down ~30%, IPOs tracking toward 21 vs. 104 in 2021). PCCP moved through Congress embedded in FDORA, and radiology AI reached a category maturation inflection on the snapshot-dependent public list with majority concentration.

📊 What Moved

📈 Trend Arcs

Arc 1: FDA AI/ML Device Clearance Volume — Compounding Through Winter

Velocity: Accelerating (regulatory throughput); Decoupled from (capital markets)

The FDA's AI/ML-enabled medical device authorization count compounded through Q3 2022 independent of the biotech capital environment. The pace of steady new clearances per month was steady and in some sub-specialties accelerating. Radiology remained dominant on the public list at mid-year, but the base was broad: chest X-ray triage, cardiac CT analysis, mammography detection, bone density assessment, fundus imaging. The 510(k) substantial equivalence pathway accounted for the overwhelming majority of clearances; the predicate library in radiology was deep enough that new submissions could move efficiently through the review process.

The arc's trajectory from July to September showed continued additions across multiple radiology subspecialties, with cardiology and pathology growing their share. Apple Watch AFib History added a consumer AI device to the cleared list. Aidoc's expanding portfolio of X-ray and CT triage tools continued receiving clearances for new intended uses. The cumulative clearance count was on track to approach the next major milestone by year-end — and the FDA's calendar year 2022 clearance pace would ultimately be the highest single-year total to that point.

The decoupling from capital markets was structurally important. Companies that had built regulatory affairs infrastructure and submission pipelines during 2020–2021 continued submitting and receiving clearances in 2022 regardless of whether they could raise a new round. The 510(k) process operates on administrative timelines, not funding timelines. The winter killed Series B fundraising for AI drug discovery platforms; it did not kill 510(k) clearances for AI SaMD with established predicates.

Where it stands at quarter close: The cumulative FDA AI/ML device authorization count is approaching the next major milestone on the public list. The pipeline for Q4 2022 is robust. The limitation is depth rather than breadth: radiology concentration means the cleared list overstates AI's penetration across medical practice relative to what the numbers imply.

Arc 2: PCCP — From Concept to Legislative Threshold

Velocity: Accelerating

The Predetermined Change Control Plan arc ran the full quarter at accelerating velocity without reaching resolution — that resolution is a Q4 event. The concept had been articulated in FDA's January 2021 AI/ML SaMD Action Plan as a framework to address a fundamental regulatory mismatch: traditional medical device law was built for static devices, but AI/ML models improve continuously through retraining and real-world performance data. Without a change control mechanism, cleared AI devices would be regulatory frozen at the version that cleared — unable to incorporate algorithm improvements without filing a new premarket submission for each material change.

Through Q3 2022, the PCCP framework was under active development in two parallel tracks. On the regulatory side, FDA's CDRH was developing draft guidance on what a compliant PCCP submission would look like, what evidence standards would apply, and how post-market monitoring obligations would be structured. On the legislative side, language for what would become FDORA's Section 515C was embedded in the broader government funding package negotiations, with year-end as the expected legislative vehicle.

The industry was tracking this closely. For any company building adaptive AI in a medical device — algorithms that were intended to update as they accumulated real-world data — the PCCP pathway was the difference between a clearance that locked in Q3 2022 performance and one that could evolve through a pre-approved modification plan. The companies that were designing PCCP into their development architecture in Q3 2022 were positioning for the regulatory environment that would formally exist after December 29.

Where it stands at quarter close: Not yet codified. FDORA has not passed. PCCP remains a proposed framework in FDA guidance documents and draft legislative language. The signal for builders is unambiguous: design for PCCP now. The Q4 resolution is expected and will not surprise anyone who has been following the legislative calendar. What changes at codification is certainty, not direction.

Arc 3: Biotech Capital Winter — Structural Reset, Not Cyclical Dip

Velocity: Decelerating (biotech funding), Accelerating (scope of contraction)

The biotech capital contraction in Q3 2022 was increasingly recognized as structural rather than cyclical. The NASDAQ Biotechnology Index had fallen roughly 30% from its 2021 peak. Cash reserves at NBI-listed companies declined 53% from Q4 2021 to Q3 2022 — not because companies were burning cash faster, but because they could not raise replacement capital. Biotech IPOs tracked toward 21 for full-year 2022, down from 104 in 2021. Crossover investors — the institutional capital that bridges private biotech to public markets — had largely withdrawn from new positions.

The contraction progressed through Q3 with increasing severity. July saw continued layoff announcements across mid-cap biotechs. August and September saw the pipeline-cut decisions that mid-cap and small-cap biotechs make when cash runway compresses: R&D head-count reductions, discontinuation of earlier-stage programs, deferral of IND filings. Agios cutting its R&D team by roughly half was among the most visible examples. Eisai's complete shutdown of H3 Biomedicine — a fully-resourced oncology research subsidiary — was a harder signal.

AI drug discovery platforms were not immune. The 2020–2021 capital thesis that rewarded platform scale and data moat narratives had produced large Series B rounds for companies like Recursion, Insilico Medicine, and Exscientia. By Q3 2022, the crossover investors and established biotech VCs that had written those checks had pulled back. The bar for new investment had reset to clinical-stage evidence. Platform AI without a compound in clinical development was effectively unfundable.

Where it stands at quarter close: The winter has not reached bottom — the funding environment will remain difficult through 2022 and into 2023. The decoupling between regulatory progress (advancing) and capital market access (contracting) is fully established and structural. Companies built on platform narrative rather than clinical-stage evidence are at acute risk.

🗺️ Landscape Shift

The most important structural shift in the landscape: the decoupling of regulatory trajectory from capital availability. FDA's clearance machine does not run on venture capital. The predicate system, the administrative review process, and CDRH's Digital Center of Excellence are operating on government administrative timelines that are not correlated with LIBOR or S&P 500 multiples. Companies that positioned for regulatory advancement during the capital winter — maintaining regulatory affairs investment while cutting R&D overhead — will emerge from the winter with cleared products that competitors chose not to file during.

🧭 Regulatory Direction of Travel

The FDA's direction of travel in Q3 2022 was clear across multiple signals:

Clearance velocity and mix: The 510(k) substantial equivalence pathway dominated. De Novo authorizations were rare in AI/ML SaMD but meaningful when they occurred, establishing new regulatory classifications that other devices could then use as predicates. The mix heavily favored radiology, with cardiology growing slowly and pathology and ophthalmology emerging. FDA's efficiency in processing radiology AI submissions reflected the depth of the predicate library and the regulatory expertise that had accumulated on both sides of the review — at FDA and in industry's regulatory affairs functions.

PCCP trajectory: FDA's public statements and guidance development through Q3 2022 consistently signaled that the PCCP framework was not a question of if but when. CDRH leadership treated the framework as established policy in advance of statutory codification — developers who engaged FDA informally were receiving guidance consistent with the draft framework even before FDORA passed.

Evidence standard evolution: FDA was asking more from submissions in 2022 than it had in 2018–2019. The agency's increased attention to dataset representativeness — whether training and validation data reflected the intended use population, including underrepresented patient groups — reflected both internal policy development and external pressure from academic research documenting generalizability failures in cleared AI devices. The standard for what constitutes adequate clinical validation was rising, even if the formal guidance had not yet been updated.

International alignment: The IMDRF (International Medical Device Regulators Forum) principles on Machine Learning-Based Software as a Medical Device — published October 2021 — continued to be the international reference framework through Q3 2022. FDA, Health Canada, and UK MHRA were aligned on the IMDRF principles. The EU AI Act's high-risk classification of medical AI was directionally consistent with IMDRF's risk framework but operationally different in its conformity assessment requirements. Geographic divergence between FDA pathways and EU AI Act obligations was becoming a planning consideration for companies operating in both markets.

Guidance pipeline: The key anticipated guidance at Q3 close was the PCCP implementing guidance expected to follow FDORA's passage. FDA's Digital Health Center of Excellence was also known to be working on updated AI/ML SaMD guidance that would incorporate learnings from the clearance volume accumulated since the 2019 discussion paper. No formal draft was published in Q3; publication was expected in 2023.

💰 Funding & Deal Pattern

The broad biotech venture capital environment in Q3 2022 was in active contraction. Mega-rounds (>$100M) had fallen from 2021 peaks.

The structural read: capital in Q3 2022 was rewarding regulatory and commercial clarity over platform ambition. The market had internalized that platform AI promises required more time and risk tolerance than the 2021 vintage implied, and was resetting to shorter de-risking cycles.

🔍 The Counter-Narrative

• The consensus: A high cumulative count proves AI has matured in medicine. The reality: Academic publications in 2022 documented that cleared radiology AI devices frequently failed to generalize across patient populations, imaging equipment manufacturers, and clinical settings not in their validation datasets. FDA clearance establishes safety and effectiveness for a defined intended use — it does not guarantee real-world deployment performance.

• The consensus: The biotech funding winter is uniformly negative for all companies. The reality: Companies that maintained regulatory investment while cutting R&D overhead were positioning for structural advantage. The clearance pipeline doesn't pause for venture cycles. Work done in Q3 2022 produces clearances in 2023-2024, after capital improves. Regulatory advancement is a lagging function — and the companies that protected it would emerge ahead.

📐 Builder's Benchmark

Median clearance timelines (510(k), AI/ML SaMD, approximate Q3 2022 data):

• 510(k) total review time: approximately 130–180 days from submission receipt to decision for AI/ML devices with complete submissions
• Pre-submission meetings (Q-Sub) before 510(k) filing: 3–4 months lead time; strongly recommended for novel AI device types
• De Novo pathway: 12–18 months from submission to authorization; used when no predicate exists
• PMA pathway: rare for AI/ML SaMD; reserved for highest-risk devices; 180-day review clock with frequent major amendments

Reimbursement coverage rates:

• FDA-cleared AI devices with standalone CPT codes: a small fraction of the public-list clearances
• New Category III CPT codes (emerging technologies): available; do not confer automatic coverage
• Medicare coverage with evidence development (CED): being explored for some high-visibility AI diagnostic tools
• Commercial coverage: device-by-device; highly variable; most cleared AI devices at Q3 2022 were reimbursed as part of professional service billing, not as separately billable AI

Clinical validation study sizes accepted by FDA:

• Retrospective reads from archived imaging studies: accepted for lower-risk, well-characterized intended uses
• Prospective multi-site studies: required for higher-risk intended uses or novel patient populations
• Sample sizes: highly variable by indication; 300–500 per site for retrospective studies in well-established radiology applications; 1,000+ for prospective studies in higher-risk settings
• Performance metrics: sensitivity, specificity, and AUC for diagnostic aids; reader study comparisons (AI vs. physician) increasingly requested

PCCP adoption rates:

• Formal PCCP submissions: near-zero at Q3 2022 close — the statutory framework did not yet exist
• Informal PCCP design in development pipelines: increasing; companies engaging FDA via pre-sub meetings were receiving guidance on modification planning
• Expected post-FDORA: first formal PCCP submissions anticipated in 2023 after implementing guidance publication

👀 What to Watch

1. FDORA passage and Section 515C codification — The year-end government spending package is the expected legislative vehicle. PCCP statutory authority changes the development architecture calculus for every company building adaptive AI in a medical device. The question to watch at passage: how quickly FDA follows with draft implementing guidance. Timeframe: December 2022.

2. FDA total AI/ML device count crossing the next major milestone — The count approaching the next milestone is significant less as a milestone and more as a baseline inflection — it will mark the point at which radiology AI's dominance becomes harder to sustain as a share of the total, as multi-specialty clearances compound. Watch the specialty mix shift in Q4 and Q1 2023. Timeframe: Q4 2022.

3. EU AI Act next formal session — The European Council's general approach was adopted mid-2022. Parliament's Committee on the Internal Market and Civil Liberties (IMCO) was working on its report. Trilogue between Council, Parliament, and Commission would follow the Parliament position in 2023, not Q3 2022. The outcome on how "high-risk AI in medical devices" is operationalized will determine whether the EU and FDA frameworks converge or diverge. Timeframe: Q4 2022 through 2023.

4. Insilico Medicine ISM001-055 Phase 2 initiation — Insilico's idiopathic pulmonary fibrosis candidate, designed using an AI-generated target and AI-synthesized molecule, completed Phase 1 safety and PK data collection through Q4 2022. Phase 2 first patient dosing is expected in 2023. This will be the first clinical efficacy readout attributable to an end-to-end AI drug discovery process. Timeframe: First half 2023.

5. FDA Digital Health Center of Excellence guidance activity on generative AI and LLMs — The November 30 ChatGPT launch (a Q4 event) will force FDA to begin receiving and responding to developer inquiries about LLM clinical applications. Watch for any informal FDA communication — Q-Sub responses, conference presentations, published statements — that begins to frame whether and how LLM-based clinical decision support tools will be regulated. This is the early-signal arc that will define a major regulatory frontier for 2023 and beyond. Timeframe: Q4 2022 – Q1 2023.

📎 Sources

Key references for this quarter. Links provided where available; historical entries may reference publications by title and date.