Step 1: Is it the same product and the same condition?
The most common trick is citing a study of one product for a different one. A trial of culture-expanded bone marrow MSCs applied surgically to a cartilage defect does not support an injection of umbilical cord cells into an arthritic knee. Check: cell source, processing (same-day vs expanded), dose, delivery route, and the exact condition and severity studied. If any differ from what you are being offered, the citation is decoration.
Step 2: What kind of study is it?
| Design | What it can show | Our grade |
|---|---|---|
| Systematic review of randomized trials | The best available answer | L1 |
| Randomized controlled trial, adequately sized, blinded | Cause and effect | L1-L2 |
| Small randomized trial, or non-blinded | A signal worth confirming | L2-L3 |
| Cohort or case-control study | Association | L3 |
| Case series, case report, registry without control | Safety and feasibility, not efficacy | L3-L4 |
| Animal or laboratory study | Mechanism; nothing about humans | L4 for clinical claims |
Full framework in evidence levels explained.
Step 3: Is there a control group, and what did it get?
Placebo response in osteoarthritis injection trials is large: saline injections routinely produce 20 to 40 percent pain improvement. A study without a control group cannot distinguish the cells from the needle, the attention, and time. A study comparing cells to nothing exaggerates the benefit; comparing cells to saline or to hyaluronic acid is fairer; comparing to PRP is the most useful for a patient deciding what to buy.
Step 4: How many patients, and how many finished?
Trials with 20 patients per arm can produce impressive percentages by chance. Look for the number randomized and the number analyzed at the final time point. A 40 percent dropout rate means the survivors are not representative.
Step 5: What was measured, and was it the primary outcome?
Pain and function scores (WOMAC, KOOS, VAS) are patient-reported and meaningful; MRI cartilage volume is objective but may not track symptoms; "satisfaction" is nearly meaningless. Check whether the outcome the clinic highlights was the trial's pre-registered primary outcome or a secondary finding picked after the fact. ClinicalTrials.gov registration numbers let you check.
Step 6: How long was follow-up?
Six months is common and short. Benefit that disappears at 12 months is a different product from benefit that persists at 3 years. For structural claims (cartilage regrowth), anything under 2 years is preliminary.
Step 7: Statistical vs clinical significance
A 4-point change on a 100-point pain scale can be statistically significant in a large trial and imperceptible to a patient. Look for the minimal clinically important difference (MCID), which for WOMAC pain is roughly 10 to 15 points, and ask whether the treatment effect exceeds it.
Step 8: Who paid, and who wrote it?
Industry funding does not invalidate a study, but check whether the authors sell the product. Studies published in journals owned by clinic groups, or in journals that do not appear in PubMed, carry less weight. A quick PubMed search of the title tells you a lot.
A worked example
A clinic cites "a 2023 study showing 85 percent of patients improved." You find it: a single-center case series of 40 patients, autologous adipose cells, no control group, 6-month follow-up, outcome was patient-reported improvement on a yes/no question, published in a journal not indexed on PubMed, authored by the clinic's medical director. Grade: L4. The citation tells you the clinic's patients were asked whether they felt better and most said yes, which is what happens after any injection with attention and time. Compare with interpreting clinic success percentages.
Where to look things up
PubMed (pubmed.ncbi.nlm.nih.gov) for indexed studies, ClinicalTrials.gov for registered trials and their pre-specified outcomes, and the Cochrane Library for systematic reviews. If a cited study is in none of the three, treat it as marketing. We apply this method to every claim on this site; see the 2026 research roundup.
Regulatory status: FDA vs INVIMA
United States (FDA): No stem cell product is FDA-approved for orthopedic, neurologic, or systemic use. Minimally manipulated same-day autologous procedures are marketed under a narrow exemption; culture-expanded cells and most umbilical products are unapproved biologics and the FDA's authority over them was reaffirmed in federal court in 2024. Florida's 2025 law permits certain treatments under state rules but does not create FDA approval.
Colombia (INVIMA): Cell therapies are regulated as advanced-therapy products. Clinics require INVIMA habilitation for the facility and, for laboratory-processed products, product-level authorization or an approved research protocol. Ask to see the INVIMA registration number and verify the treating physician on ReTHUS. "INVIMA registered" for the clinic is not the same as an authorized product.
Frequently asked questions
What is the single most important thing to check in a stem cell study?
Whether there was a control group and what it received. Uncontrolled improvement after a joint injection is mostly placebo and time.
Are animal studies useful?
For mechanism, yes. For deciding whether to pay for a human treatment, no. Most animal successes never translate.
How do I find whether a study is real?
Search the title on PubMed and the registration number on ClinicalTrials.gov. If it appears on neither, be skeptical.
Send us the studies a clinic gave you.
We will grade them using this method and tell you what they actually support, in plain language, before you pay anyone.
WhatsApp Andy in MedellínRequest a quote