FDA's newest stem-cell warning letter quotes the company's own "What is Regenerative Medicine?" web page as proof the product is an unapproved drug
September 22, 2026 · 5 items
FDA's newest stem-cell warning letter quotes the company's own "What is Regenerative Medicine?" web page as proof the product is an unapproved drug
FDA — Warning Letters · September 11, 2026Regenerative medicine
NexCell Scientific (Irvine, CA) makes an umbilical cord blood-derived total nucleated cell product for allogeneic use. FDA inspected it in Dec 9–12, 2025 and Feb 3–5, 2026. CBER's letter says the product is an unapproved new drug and an unlicensed biological product.
The legal hinge is homologous use. A cord-blood product only avoids premarket review if it does what cord blood does in the donor (rebuilding the blood and immune system). FDA says this product was pitched for "inflammation, injuries, neurological conditions and liver cirrhosis" instead. That means it no longer qualifies for the lighter-touch "361 tissue" rules.
The evidence came from marketing. FDA quotes the company's website: "When injected into a damaged tissue, the cells specialize to carry out the necessary function of that particular area." It also cites a vendor "white paper" listing musculoskeletal, cardiovascular, neurological and autoimmune uses.
On top of that are manufacturing failures: no validated process for identity, strength, quality and purity, and no validated aseptic process for a product sold as sterile. The company has 15 working days to respond. If it doesn't fix things, it risks seizure and/or injunction.
Why it matters for an independent practice: This is an MMR problem as much as a regulatory one. The copy on a clinic's "how stem cells work" page is exactly what FDA reads to decide what a product is intended for. Two moves this week: (1) audit every client page that describes cord blood, MSC or "allograft" products with words like "specialize," "repair," "regenerate" or a list of diseases. (2) Ask each clinic which supplier's products it administers, and whether that supplier's own marketing claims more than the label. FDA — Warning Letter to NexCell Scientific Inc, CBER 26-730612 Date note: signed Sep 11, published to FDA's index Sep 22. The public posting is the news. Twelve days sit between the two dates, and both are stated here on purpose.
A patient-access company bought an AI voice agent — and says the agent is useless without the 260-rule scheduling rulebook underneath it
HIT Consultant · September 21, 2026Regenerative medicine
DexCare (spun out of Providence in 2021, $146M raised) has acquired Mila Health. Mila's AI agents call, text and chat with patients to book overdue visits, check eligibility, walk them through pre-procedure prep and follow up after discharge. Terms were not disclosed. This is DexCare's second acquisition.
Mila's reported results versus traditional call centers: 54% response rate on outbound outreach (said to be 80% better than manual call-center benchmarks), no-shows cut in half, and 18% net revenue expansion.
The part to take away is the architecture. The agents run inside DexCare's "access data model," one standardized rulebook built from the EHR, scheduling rules, subspecialty requirements and provider preferences. At Tampa General, DexCare codified 260 separate scheduling policies and found more than 50 operational gaps along the way. Every booking has "traceable decision logic" behind it.
CEO Matt Blosl: "Any agent can sound incredibly human, but ours understands the health system it speaks for."
:warning: Disclosure: every performance number is vendor-reported, from the acquisition announcement. HIT Consultant is a Candidate source.
Why it matters for an independent practice: This is the AI-receptionist build with the missing step named. Before an n8n/Vapi voice agent books anything for a regen clinic, somebody has to write down the clinic's rules: which provider does which procedure, what needs imaging first, which payers get which slots, and when a caller goes to a human. That rulebook is the deliverable. The voice is the easy part. A practice (or MMR) that writes the rulebook first gets the no-show gains. One that skips it gets a friendly agent booking the wrong patient into the wrong slot. HIT Consultant — DexCare acquires AI care coordination platform Mila Health
Google now freezes ALL new reviews on a Business Profile when it spots a "spike in spam reviews" — and a clinic with a good review-request habit just got caught by it
Search Engine Roundtable / Barry Schwartz · September 22, 2026Patient acquisition
Google is emailing Business Profile owners: "Google has detected a spike in spam reviews on your Business Profile and removed them so they do not affect your Google rating."
The same email says new ratings, reviews and other user contributions are temporarily paused, "typically within a few days" until Google judges the risk is handled. No action is required, but owners who think real reviews were removed by mistake can request a review.
A false positive is already on record. Local-SEO consultant Amy Toman described a former client that had just opened a new location and is "always very good about getting a steady stream of reviews": "I think Google questioned their methods and slapped this on… this filter, along with many of the new ones, needs to be refined a bit."
Screenshot first shared on LinkedIn by Antoine Cameron. Google has made no formal announcement, so this is observed behavior, not documented policy.
Why it matters for an independent practice: Review velocity is one of the main local-ranking levers MMR works on for clinics, and bursts are exactly what automated review-request campaigns produce (a text after every visit, a blast to the whole patient list). Three concrete moves: spread review requests out instead of batching them. Be extra careful at new locations, which is where the false positive above happened. And make sure every client profile has email notifications on, so a freeze is seen the day it happens rather than found a month later as a flat review count. Search Engine Roundtable — Google Business Profiles email: detected a spike in spam reviews (Candidate source: first Lens-C item on this board not from a Semrush-owned outlet since Sep 18.)
Heidi, the AI scribe used in 2.8 million visits a week, just raised $340M to stop only writing the note and start acting on it — "supervised action"
Business Wire — Heidi press release · September 22, 2026Practice operations
Heidi (Melbourne) closed a $100M Series C led by Blackbird at a $900M valuation, plus a $240M growth investment from General Catalyst's Customer Value Fund. Total: $340M.
Scale, as the company states it: 2.8 million patient visits a week in 190 countries and 110 languages; 175 million visits supported in total (up from 73M at Series B); ARR from $1M to $50M (April 2026) in two years; enterprise activation rate around 62%. US reference customer: Beth Israel Lahey Health.
The direction is the story. The money goes to "moving from documentation into supervised action" and "agentic capabilities to support the work around patient visits," plus building "the regulatory and clinical safety foundations those capabilities require." CEO Dr. Thomas Kelly: "from documenting care to helping clinicians act on it."
:warning: Disclosure: this is a funding announcement, a shape that usually earns :-1: here. It runs because the product change (scribe → agent) affects tools independent practices already use. Every figure is company-reported.
Why it matters for an independent practice: Many independent and regen practices first adopt AI as a scribe, and that scribe is now turning into an agent that acts: orders, follow-ups, patient messages. The question that was "is the note accurate?" becomes "what is it allowed to do, and who signs off?" Before a practice turns on any post-visit action feature in its scribe, it should write down which actions need a clinician click and check that the vendor's BAA covers the agent, not just the transcript. Business Wire — Heidi secures US$340M to scale agents across health systems globally
Nate B Jones went inside OpenAI and came back arguing that AI adoption at work is about access, not aptitude, with chapters on a proactive "9am report" and on what you actually pay for in tokens
Nate B Jones · September 22, 2026Buildable AI
Dated from the watch page's upload stamp: 2026-09-22T07:00:21-07:00, under a day old at run time. The description gives one sentence as the thesis: "AI adoption at work is less about aptitude than access."
Built as conversations with two OpenAI staffers, introduced in the chapters only as Andrew and Akshay. The chapters on the watch page: Why I was wrong about computer use (6:11) · Connectors, access, and the context problem (10:17) · How each function crossed over, coding then legal (13:20) · ChatGPT sites and a new kind of artifact (16:08) · The skills worth betting on (18:55) · Token efficiency and what you actually pay for (22:48) · Proactive AI and the 9am report (25:35) · From individual productivity to team productivity (36:16).
:warning: Thin on specifics, stated honestly. The description has no transcript or numbers beyond the chapter list, so the bullets above are the structure, not the findings. Runs under the standing rule that a fresh, date-verified, on-topic video earns the slot when the thinness is disclosed.
Why it matters for an independent practice: Zero medical content, which is the preferred shape. Two chapters map straight onto a practice. "Connectors, access, and the context problem" is the question every clinic faces when an agent needs the EHR, the calendar and the inbox: the AI is only as useful as what it can reach, and what it can reach is exactly the PHI exposure. "Proactive AI and the 9am report" is the morning huddle, built once: yesterday's no-shows, today's unconfirmed bookings, reviews waiting for replies, all on the front desk's screen before the doors open. YouTube — Nate B Jones: I Stopped Knowing What My Computer Was Doing. Then I Asked OpenAI Why.
