Stem Cell Therapy Reviews: How Reliable Are Patient Success Stories?

On their own, patient success stories are the weakest kind of evidence, and 30 of them on one page do not change that. What upgrades a story is four details: a registration number with published results, whether the patient paid, a control group, and how long follow-up ran. Of 59 clinic websites read page by page, 4 could produce a registered trial for the treatment they were selling.^[1]

The 39-person randomized double-blind trial on knee osteoarthritis shows what those details buy. The placebo group got zero cells injected into the joint, and WOMAC scores still improved 25.0% at 6 months. By 12 months they were back to 0%.^[2]

The high-dose group went 83.9% and 89.5% at those two points. The low-dose group hit 51.5% and 68.2%. The researchers called their own treatment effective, at p=0.04 and p=0.02. The two groups’ cartilage thickness differed by 0.7 mm, p=0.89.^[2]

Cross-section of a knee joint showing thinned, frayed articular cartilage over the tibia with glowing stem cells arriving at the defect

The Missing Denominator

The U.S. surveys counted 351 companies and 570 clinics in 2016, compared with 1,480 companies and 2,754 clinics in March 2021. These are two historical snapshots; their difference is not a measured constant annual growth rate.

Of those 1,480 companies, 1,262 claimed they could treat pain, or 85.3%. Another 689 chased orthopedics, or 46.5%, and 339 went after sports injuries, or 22.9%. Only 56 put prices on their websites, or 3.8%.^[4]

California holds 347 of the 2,754 clinics. Florida holds 333. Texas holds 310. Together that is 990, or 35.9% of the country.^[4]

Australia bans patient testimonials in advertising. Clinics there started calling them clinical case studies. One claimed a 75% improvement in osteoarthritis pain. Nobody explained where that figure came from.^[5]

A UK team went through 24 clinic websites and rated 79% of them as problematic. Of the flagged sites, 15 carried a risk of harming patients, or 78.9%. More than two-thirds never mentioned side effects or risks. 3 mentioned infection risk. None linked to a finished clinical trial.^[6]

Only 3 of the 24 published a price. The range ran from £995 to over £7,000.^[6]

Japan prohibits the same thing from a different angle. A testimonial on a clinic’s own website, about what a treatment did or how well it worked, counts as advertising and is banned; the MHLW guideline adds that content which would otherwise be advertisable does not save it.^[7] A testimonial on someone else’s blog or review site is not advertising, unless the clinic paid for it or arranged it. A self-pay provider that wants to advertise beyond the standard items has to publish the treatment, the cost, and the main risks and side effects. Using an unapproved product adds four more disclosures: that it is unapproved, where it came from, whether an approved domestic equivalent exists, and what the safety record looks like overseas.^[7]

The FDA warning quoted in the original version concerns unapproved regenerative products in the United States. It should not be generalized into a rule that every paid cell treatment worldwide is illegal. Approved therapies and specific authorized research charges exist. FDA investigational charging rule

Why Improvement Isn’t Proof

Knee pain rises and falls on its own. Pooled across 8 randomized controlled trials and 467 patients, about 63% of the pain improvement at 6 months came from context effects, along with 61% of the functional improvement. At 12 months those figures were 50% and 66%.^[9]

What the cells add is smaller. A separate systematic review put the between-group gap at 1.2 points of pain relief out of 10. Function improved by 14.2 points out of 100.^[9]

Trial NCT02726945 lines up with that. Its high-dose arm beat placebo by 58.9 percentage points at 6 months and 89.5 points at 12. A clinic in Australia reported that 45% of treated patients had less pain at 6 months. The placebo patients reported roughly the same.^[2][5]

Symptoms can change because of natural fluctuation, regression to the mean, co-interventions and treatment context. There is no universal untreated WOMAC change of 10–20 points over 1–3 months; the instrument version and baseline severity matter.

Timelines for standard non-surgical care
Standard option	Time to work, or how long it lasts
Oral painkillers	usually within 1 to 2 weeks
Corticosteroid knee injection	usually holds for 4 to 6 weeks
Exercise and weight loss	usually shows up in 4 to 12 weeks

Citations That Don’t Match

Staff listed on those 59 websites hold credentials from more than 30 specialty boards. Not one is in cell therapy or regenerative medicine. Cell type is where the first mismatch shows up: papers get cited for a product the clinic does not sell, and autologous and allogeneic products are not interchangeable in either the evidence or the handling.^[1]

Most sites keep 10 to 30 testimonials posted. About a third list peer-reviewed papers. Of the papers listed, fewer than 2% support the same-day product the clinic sells.^[1]

ISSCR recommends rigorous preclinical evidence, independent review, appropriate trial design and transparent reporting. Uncontrolled testimonials do not establish benefit, but an approved cell therapy is not made unproven merely because it is delivered outside a trial. ISSCR guidelines

Claims that a product is legal or FDA-approved because it appears in a clinical trial database, or because a business has registered with FDA, are often false.^[8]

Payment rules depend on the jurisdiction and pathway

China drew this line back in 2015. The Stem Cell Clinical Research Administrative Measures (Trial), issued as Guo Wei Ke Jiao Fa No. 48, bar institutions from charging participants and from running ads, direct or disguised.^[11]

State Council Order No. 818, in force from May 1, 2026, keeps the ban. Article 20 prohibits fees. Article 34 allows charges only for a technology that has finished clinical research and been cleared for clinical translation by the State Council health authority.^[12]

By September 2023, China had 118 registered institutions and 140 registered projects. All of them sit at Grade 3A hospitals. Just 1 project covers diabetes, or 0.7%. Anti-aging, cancer, and cosmetics had none.^[13][12][14]

In the U.S. patients pay out of pocket. The range runs $1,200 to $28,000. The average is about $5,100.^[4]

Japan’s self-pay figures are supposed to be published, which is why treatment costs in Japan sit in the provider’s own advertising rather than in the regulator’s filing list.^[7]

Procedure duration does not establish treatment quality or effectiveness. Ask about cell collection, processing, delivery, monitoring and follow-up separately.

The NEJM report described three women with severe visual loss after intravitreal autologous adipose-derived “stem cell” injections at a clinic associated with a registered study. A public registry listing is not proof of regulatory approval or adequate ethical oversight. Original NEJM case series

Never Mentioned in the Story

In 2018, 3 U.S. states reported 12 patients infected after unapproved cord blood products. All 12 were hospitalized. Texas had 7, Florida 4, and Arizona 1.^[16]

6 unopened vials from the Texas sites all grew Enterobacter cloacae. Five of those also grew Citrobacter freundii. The contamination happened before the product shipped. Sterile technique at the injection had nothing to do with it.^[16]

Cell processing facilities in Japan must be licensed or notified, and 3,671 of them stood on the books as of June 30, 2026.^[17] Of those, 85 hold a licence and 3,586 filed a notification.

A generic joint-injection infection rate cannot estimate the safety of a particular cell product. Product contamination, processing and delivery route contribute additional risks; the absence of infection in a small series does not establish a reliably low rate.

Cross-border care added a second outbreak in 2022. 3 patients caught Mycobacterium abscessus after embryonic stem cell injections in 3 Mexican cities. 2 of the clinics sat 269 km apart, and the strains were a single clone differing by 1 nucleotide. As of March 28, 2024, all 3 were still in treatment.^[18]

3 women got fat-derived “stem cells” injected into both eyes. Their vision ranged from 20/30 to 20/200 beforehand, and from 20/200 to no light perception a year later.^[15]

FDA lists the risks the stories leave out: blindness and tumours where migrated cells grow, neurological events, reactions at the harvest and injection sites, bloodstream infections that threaten life, and contamination when a product is never screened for hepatitis B, hepatitis C, or HIV.^[19]

Providers in Japan must report a death or a suspected serious event to the regulator within 7 days. Other serious cases get 15 days. Those reporting rules bind the provider, which then has to notify the manufacturer of the cells.^[20]

Regulators and Courts Are Catching Up

Two federal appeals courts have ruled against clinics. In 2021 the Eleventh Circuit upheld a permanent injunction against a Florida clinic. On September 27, 2024, the Ninth Circuit reversed a California district court in case 22-56014 and held that stromal vascular fraction separated from fat counts as a drug under federal law.^[21][22]

21 CFR 1271.15(b) holds the exception known as the same surgical procedure. The tissue taken out and the tissue put back have to be the same thing. FDA’s guidance on the exception answers the questions that follow from that wording.^[23][24]

The Section 361 HCT/P criteria include minimal manipulation, homologous use, restrictions on combining tissues with other articles, and criteria for systemic effect/metabolic activity. There are specified exceptions, including certain autologous and close-relative uses; failure to qualify does not mean every product follows an identical approval pathway. Exact 21 CFR 1271.10 criteria

Japan runs a third route. The Act on the Safety of Regenerative Medicine sorts techniques into Class I, II and III by risk. A provider files a treatment plan only after a certified committee reviews it, and Class I plans go through a specially certified committee and the health ministry while Class II and III go to the regional bureau.^[20] Treating without a filed plan is a criminal matter: up to 1 year in prison or a 1 million yen fine for Class I, and up to 500,000 yen for Class II and III.^[27] The filed plans sit in the public record of hospitals and clinics, listed by provider and by treatment name.^[27]

The register as of June 30, 2026 holds 6,743 filed treatment plans and 85 research plans.^[17] Behind them sit 167 certified committees, 80 of them specially certified, and 3,671 cell processing facilities.^[17] Class III accounts for 4,272 of the 6,743 plans and Class II for 2,463. Class I, the highest-risk band, holds 8.^[17] Kanto-Shinetsu, the region that includes Tokyo, accounts for 3,600 of them.^[17]

Filed is not approved. Japan’s product route runs separately, and PMDA’s list of approved regenerative medical products covers April 2015 to May 2026.^[28]

Japanese provision plans, U.S. approved cell/gene products and numbers of commercial clinics count different things in different jurisdictions. Dividing these totals does not yield an approval rate or a useful measure of legitimacy.

FDA handed the industry 36 months of enforcement discretion in 2017, and COVID stretched the deadline to May 31, 2021.^[30] 4 final guidance documents sit under that framework.^[31] By October 2023 the agency had sent 28 warning letters and more than 640 notice-type letters. Against 2,754 clinics, that is 1 warning letter for every 98 clinics.^[21][4]

MARCS-CMS 695592, dated August 15, 2025, went to an Arizona clinic selling cord mesenchymal stem cells for autism. MARCS-CMS 712579, dated February 11, 2026, went to a Nevada clinic over a cord-derived stem cell product.^[32][33]

Under Order 818 the filing must arrive within 5 working days of ethics approval (Article 15), records stay on file 30 years after a study ends (Article 22), and participants get follow-up monitoring (Article 26). The institution itself must be a Grade 3A hospital (Article 11). Regulators then get 5 working days to route a filing to review and 15 working days to decide (Article 31).^[12]

For running banned research, Article 43 fines 1 to 10 million yuan when there is no illegal gain or it falls under 1 million, and 10 to 20 times the gain above that. A 5-year ban from the field comes with it. Individuals separately face 100,000 to 200,000 yuan and a 10-year-to-lifetime ban.^[12]

What to Check Before You Pay

Registration number and results. In China, look it up in the national medical research registration system. In Japan, the treatment plan is a public filing, so the exact treatment name on the provider’s registered plan has to match the name on your quote. Anywhere else, check the trial registry for the status and results sections.^[34]
How they charge. Ask whether this is research or an approved clinical application. Fee rules are jurisdiction- and pathway-specific; in China, the cited research framework prohibits relevant research charges, while other countries have defined exceptions.^[12]
Control group and follow-up. If the only data points on offer are 3 and 6 months, there is usually no long-term safety data behind them. The follow-up schedule written into the plan is the part to compare against the published trials, which the treatment process sets out step by step.
Citations and indication. Open the papers the site lists and compare the cell type and the delivery route. Then check that the approval or registered indication covers the exact condition you are treating.
Where to report a claim that overshoots. Japan’s health ministry runs a public channel for reporting exaggerated medical advertising.^[35]

A straight answer before you pay anything.

Send the treatment name and the quote you have been given. Our team checks the plan against the public filings, sets out which records and blood tests the physicians still need, and translates the consultation. If the treatment is not suitable for your case, we say so.

Request a consultation

When a Review Deserves Trust

The FDA’s official list holds 54 approved cell and gene therapy products. 9 are cord blood products. The list was last updated September 17, 2026.^[29]

The March 2021 survey identified 1,480 U.S. businesses and 2,754 clinics. It is a historical market snapshot, not the denominator for a current FDA product-approval rate.

MACI is approved for certain focal cartilage defects, not generalized knee osteoarthritis. The claim that only cord-blood stem cell products are FDA-approved is outdated: on December 18, 2024, FDA approved Ryoncil, a donor bone-marrow MSC product, for steroid-refractory acute graft-versus-host disease in children aged two months or older. This does not validate MSC treatment for arthritis, hair loss or anti-ageing. FDA Ryoncil approval

Orthopedics alone accounts for 9 unapproved conditions, from osteoarthritis and tendinitis to back pain and knee pain. Neurology accounts for 6, including multiple sclerosis, Parkinson’s disease, epilepsy, and stroke. Another 5 include autism, blindness, and chronic pain. No exosome product has been approved for anything.^[19][8]

Follow-up is intervention- and protocol-specific; neither 12–24 months for all trials nor five years for all post-market monitoring is a universal rule. U.S., Japanese and Chinese regulatory pathways differ and must be checked separately.

Registered research, IND research, and unapproved treatment
Aspect	Registered clinical research in China	Research under a U.S. IND	Unapproved commercial treatment
Who pays	regulators bar research fees for participants	Sponsor coverage varies; authorized charges and ordinary-care costs can occur	the patient pays the whole bill
Who sees the results	regulators publish the filings and project lists	data goes into the marketing application and the label	the clinic decides how much to disclose
Who answers for harm	the sponsor covers treatment costs, unless the harm is the clinic’s own fault	researchers must report adverse events under FDA rules	Reporting and liability depend on the jurisdiction, product and legal pathway; verify the actual requirements

Common Questions

Is FDA Registration the Same as Approval?

No. Registration and listing are paperwork, and FDA says an IND is not marketing authorization either. The U.S. had approved 54 cell and gene therapy products as of 2026.^[30][29]

Should a Trial Cost You Money?

In China, the cited framework prohibits charging participants for the research intervention. In Japan, ask both which legal research or treatment pathway applies and who pays each item; insurance coverage alone does not decide trial legitimacy. In the U.S., certain investigational charges require FDA authorization. 21 CFR 312.8

What a Celebrity Endorsement Tells You

Not much. Of those 24 UK clinics, exactly 1 said the treatment was still experimental, or 4.2%. Another clinic claimed there was no rejection risk and no risk of spreading infectious disease.^[6]

It Worked for Me, So What Happened?

An individual improvement can be real while its cause remains uncertain. Natural history, standard treatment, regression to the mean and contextual effects can all contribute. A pooled contextual-effect estimate does not determine what caused one person’s recovery.

Does Going Abroad Help?

A typical trip runs 3 to 7 days. The 3 patients from 2022 split 2 in Colorado and 1 in Arizona. As of March 2024 investigators still had not pinned down the source of the infection.^[18] The paperwork comes first: records, imaging and blood work are collected before travel, which the medical travel process sets out in order.

Where the Numbers Come From

Numbers require an identifiable source, population, date and endpoint. Clinic prices and scheduling estimates are not clinical evidence; a range does not repair an unsupported assertion. Cross-country counts are descriptive and should not be converted into approval or enforcement rates.

Retain denominators when interpreting results. A percentage change in a symptom score is not a percentage of patients cured, and adverse-event counts are not necessarily counts of different patients.