Life Sciences / Regulatory Brief 🧬

📌 Navigate

📊 Exec Summary

February's regulatory baseline still defined the March window: QMSR compliance had just gone mandatory, FDA had already cleared the first AI tool that simultaneously detects and characterizes lung cancer on CT, and guidance had already widened the door for AI-powered clinical decision support while closing it on black-box implementations — a coordinated deregulatory move with embedded enforcement teeth.

Four things moved in life sciences / regulatory this week:

The pattern: FDA is simultaneously lowering the barrier to enter the market for lower-risk AI tools (CDS enforcement discretion, wellness wearable expansion) and raising the compliance floor for everyone already in it (QMSR mandatory, transparency mandates for AI logic) — the net effect is a more stratified device landscape where documentation quality becomes the primary differentiator.

1️⃣ eyonis LCS 510(k): First CADe+CADx AI Cleared for Lung Cancer Screening

TL;DR: Median Technologies received 510(k) clearance on February 9 for eyonis LCS — the first AI SaMD cleared to both detect and characterize pulmonary nodules on LDCT, targeting 14.5 million US-eligible screening patients. NT-APC 1508 is an existing Medicare reimbursement code that covers the procedure, removing the need for a new coverage determination at commercial launch. (Clearance: confirmed. Performance metrics: vendor-reported from manufacturer testing. Reimbursement rate and code: vendor-reported, consistent with OPPS 2026 data.)

What happened

• 510(k) clearance granted February 9, 2026
• Device: eyonis LCS; classification: AI/ML-powered CADe/CADx (computer-aided detection + diagnosis) SaMD
• Indication: detection and characterization of parenchymal pulmonary nodules on low-dose CT (LDCT) scans
• Regulatory distinction: first device cleared for the combined CADe+CADx function in lung cancer screening — prior clearances were detection-only
• Performance (manufacturer testing, lung cancer screening reference population): sensitivity 93.3%, specificity 92.4%, NPV 99.9%
• US screening population: ~14.5 million eligible individuals (lung cancer is the leading cause of cancer death in the US)
• Reimbursement: existing NT-APC 1508 code covers the procedure — no new code or coverage determination required; Medicare reimbursement rate ~$601–$700 per exam per OPPS 2026
• Commercialization: first US sites expected Q3 2026; CE marking expected Q2 2026

📊 Key facts (from Median Technologies / BusinessWire)

🔗 Primary source → Median Technologies press release

🔍 The non-obvious point

The CADe+CADx combined clearance is the regulatory novelty — prior AI lung screening tools were cleared for detection only. Full characterization in a single device changes the workflow economics: radiologists get one integrated read instead of two sequential tool outputs, which reduces inter-reader variability and creates a cleaner billing unit.

• The 99.9% NPV is the adoption unlock: high NPV in a screening population means the tool is safe to use as a negative filter, reducing unnecessary follow-up workload on the ~86% of LDCT screens that are negative
• NT-APC 1508 already existing means hospital systems can add eyonis LCS to existing billing workflows without a new coverage determination — but commercial adoption still requires local coverage policies, contracting, and clinical protocol integration that the code alone does not resolve
• For SaMD builders: the CADe+CADx combination in a single 510(k) — rather than two separate clearances — is the submission strategy to watch for other multi-function AI imaging tools

👀 What to watch

• First US commercial site launches Q3 2026 will generate real-world performance data that either validates or challenges the manufacturer reference population performance metrics — watch for post-market study publications at RSNA 2026 (November).

2️⃣ QMSR Effective Feb 2: ISO 13485 by Reference, QSIT Retired

TL;DR: The FDA Quality Management System Regulation went live on February 2, replacing the 1996 Quality System Regulation with a framework that incorporates ISO 13485:2016 by reference — aligning US CGMP with EU MDR/global standards — and simultaneously retired the QSIT inspection method in favor of a risk-based compliance program.

What happened

• Effective date: February 2, 2026; applies to all finished device manufacturers distributing in the US
• Regulation: 21 CFR Part 820 (Quality Management System Regulation — QMSR)
• Key change: incorporates ISO 13485:2016 Medical Devices — Quality Management Systems and ISO 9000:2015 by reference into 21 CFR 820
• Harmonizes US CGMP with EU MDR/global quality system standards — a manufacturer compliant with EU MDR is now largely compliant with FDA CGMP
• Simultaneously on Feb 2: FDA retired QSIT (Quality System Inspection Technique), replacing it with the updated risk-based Inspection of Medical Device Manufacturers Compliance Program
• New inspection approach evaluates compliance with QMSR and emphasizes areas posing greatest risk to patients or device users
• Scope: all Class II and III finished device manufacturers; risk management now integrated throughout the regulation

📊 Key facts (from FDA CDRH / ECA Academy)

🔗 Primary source → FDA QMSR

🔍 The non-obvious point

QMSR effectively harmonizes FDA's inspection posture with the EU MDR quality system framework — the single most practical outcome for any company building for both US and EU markets. A gap analysis against ISO 13485 is now the single compliance exercise that covers both regulators.

• The QSIT retirement on the same day is equally important: FDA's inspection approach is now risk-stratified, meaning compliant manufacturers with strong documentation and risk management records face less inspection burden — and manufacturers with gaps face more scrutiny, not less
• AI SaMD manufacturers specifically: QMSR's integrated risk management requirements apply directly to post-market surveillance and algorithm change protocols — expect CDRH to cite QMSR during AI device inspections as the quality framework expectation
• Ropes & Gray note: FDA adopted the new inspection approach the same day the rule went live — no grace period on the inspection methodology shift

👀 What to watch

• First post-QMSR FDA Form 483 observations and Warning Letters will reveal how inspectors are applying the new risk-based framework; expected to appear in public docket Q3–Q4 2026.

3️⃣ FDA CDS Guidance Update: Enforcement Discretion and AI Transparency Mandate

TL;DR: FDA's January 6 CDS software guidance update expands the non-device CDS definition and provides enforcement discretion for single-output AI recommendations — but the deregulatory framing is incomplete without the transparency mandate: AI/algorithmic CDS that cannot document its logic is now explicitly at risk of device classification. These are not two separate moves; the enforcement discretion and the transparency requirement are conditional on each other.

What happened

• Guidance issued January 6, 2026; supersedes 2022 CDS guidance (industry response and advisory coverage peaked late January–February)
• Key expansion: software identifying patients in the indicated population for a chemotherapeutic agent now excluded from device definition — previously classified as a device
• Enforcement discretion: FDA will not enforce device requirements for CDS software that provides a single output or recommendation when all other Section 520(o)(1)(E) criteria are met
• Transparency mandate: AI/algorithmic CDS must document data inputs, underlying logic, and recommendation generation methods — treating the software as a "black box" explicitly triggers device classification risk per the guidance
• Wellness update: noninvasive wearables estimating physiologic parameters (heart rate, blood glucose proxies) qualify as general wellness if no disease claims are made and minimal risk is posed

📊 Key facts (from FDA CDRH / Arnold & Porter / Faegre Drinker)

🔗 Primary source → Arnold & Porter advisory

🔍 The non-obvious point

The deregulatory framing and the enforcement mandate are not separate signals — they are one policy move. The enforcement discretion for single-recommendation AI CDS is conditional on meeting the transparency requirements. A builder who reads the deregulatory headline without reading the transparency condition has misread the guidance.

• Deregulatory move: FDA will not enforce device requirements for CDS software that provides a single output where all Section 520(o)(1)(E) criteria are met — this is a genuine expansion of the non-device lane.
• Enforcement condition: That expansion applies only to tools that can document their data inputs, underlying logic, and recommendation generation methods. Black-box AI CDS tools that cannot produce this documentation do not qualify for enforcement discretion and face device classification risk per the explicit guidance language.
• For SaMD and clinical AI builders: the transparency requirement is a de facto pre-submission documentation checklist. Data lineage, model architecture summary, and recommendation logic must be producible on demand — not as post-hoc reconstruction.
• The wellness wearable expansion is more significant for the non-regulated digital health market than for regulated device builders — but it opens the door for on-device AI health monitoring to scale without FDA pre-market review if positioned correctly and no disease claims are made.
• Watch for FDA's forthcoming AI-enabled device guidance (expected H2 2026) to codify these transparency expectations as formal submission requirements.

👀 What to watch

• FDA AI-enabled device guidance update expected H2 2026 will likely convert the CDS transparency language into formal 510(k) and De Novo submission requirements — affected builders should begin documentation alignment now.

4️⃣ 2025 AI/ML 510(k) Baseline: 295 Clearances, 62% SaMD

TL;DR: Innolitics' 2025 year-in-review quantifies the AI/ML device clearance environment going into the QMSR and CDS guidance year: 295 AI/ML-enabled device clearances in 2025, with SaMD representing 62% — the baseline for understanding what 2026 compliance changes mean at scale.

What happened

• 2025 AI/ML 510(k) clearances: 295 total (up from prior years)
• SaMD share: 62% of all AI/ML device clearances — majority are software-only, not hardware-embedded
• Published Innolitics analysis February 2026, timed with QMSR effective date
• Provides baseline velocity: if 2026 holds similar rates, ~25+ AI/ML devices clear per month on average
• QMSR (effective Feb 2) and revised CDS guidance (Jan 6) will both affect the review environment for submissions in flight in 2026

📊 Key facts (from Innolitics)

🔗 Primary source → Innolitics 2025 Year in Review

🔍 The non-obvious point

295 AI/ML clearances in 2025 — combined with QMSR going live and CDS guidance tightening on transparency — means 2026 is the first year where a large cohort of cleared AI devices must also demonstrate QMSR-compliant quality systems under inspection. The compliance gap between clearance and quality system maturity is where enforcement risk concentrates.

• 62% SaMD means the majority of AI device makers are software companies, not traditional hardware device manufacturers — QMSR's risk management integration requirements may hit them harder as they typically have less robust quality system infrastructure
• The CDS enforcement discretion expansion could suppress some 510(k) filings for lower-risk AI tools (manufacturers choosing the non-device path) — watch whether 2026 clearance numbers diverge from the 2025 baseline
• eyonis LCS (Item 1 above) is a high-water mark for what AI diagnostic SaMD can achieve — 93.3% sensitivity CADe+CADx at launch sets a performance expectation for future imaging AI submissions

👀 What to watch

• Q1 2026 AI/ML 510(k) clearance counts (expected public FDA data Q2 2026) will reveal whether QMSR compliance burden or CDS enforcement discretion expansion is the stronger force on AI device submission velocity.

📊 The pattern

FDA's February moves form a single coherent posture: raise the quality floor (QMSR mandatory, risk-based inspections), widen the regulatory door for lower-risk AI tools (CDS enforcement discretion, wellness wearable expansion), and embed transparency as the non-negotiable condition for remaining outside device classification. The eyonis LCS clearance demonstrates what the cleared side of this landscape looks like — a combined detection-characterization AI SaMD with documented performance, an existing reimbursement pathway, and a clear commercialization timeline. Builders who treat QMSR as a compliance exercise and CDS guidance as a green light without reading the transparency requirements are misreading the signal: documentation quality is now the primary differentiator between device and non-device status for AI tools in clinical settings.

👀 Watchlist

📎 Sources

Sources of truth

Also consider reading