How Do Scientists Measure Whether a Stem Cell Therapy "Worked"? A Guide to Clinical Trial Endpoints

What This Article Covers

Every clinical trial is built around a specific, pre-agreed question — not "did people feel better?" but "did this precisely defined measurement change by this much, in this timeframe, compared to a control group?" That measurement is called an endpoint, and it's the single most important concept for understanding whether a stem cell trial actually succeeded. This article explains what endpoints are, how primary and secondary endpoints differ, walks through real endpoints from real stem cell trials in heart failure, Parkinson's disease, spinal cord injury, and osteoarthritis, and explains why a trial can technically "fail" on paper while still containing a genuinely encouraging result — the distinction patients need before getting excited about a "promising" trial.

What Is an Endpoint, and Why Not Just Ask Patients If They Feel Better?

An endpoint is a specific outcome that a trial is designed, in advance, to measure and statistically analyze in order to determine whether a treatment worked. The National Institutes of Health's NCATS toolkit defines it as "a targeted outcome of a clinical trial that is statistically analyzed" to assess efficacy and safety. Endpoints have to be locked in before a single patient is enrolled, precisely so researchers can't quietly redefine "success" after seeing how the data turned out.

"How do you feel?" isn't a usable endpoint because it's subjective, shaped by expectation, and impossible to compare consistently across hundreds of patients. A patient who paid tens of thousands of dollars and traveled overseas for treatment has every psychological incentive to report improvement, regardless of whether their heart, spine, or joint actually changed. That's why researchers use objective, standardized instruments instead: an imaging measurement like left ventricular ejection fraction, a validated symptom scale like the MDS-UPDRS for Parkinson's motor symptoms, or a structured neurological exam like the ASIA Impairment Scale — tools built specifically so a "7-point improvement" means the same thing in every trial site that uses them.

Primary vs. Secondary Endpoints

Every well-designed trial designates one (occasionally a couple of) primary endpoint — the single outcome the trial is statistically powered to detect, and the one regulators like the FDA weigh most heavily when deciding whether a therapy works. Per ClinicalTrials.gov's own definitions, the primary outcome measure is "the outcome measure(s) of greatest importance," the one used to calculate how many patients the trial needs to enroll for a reliable answer.

Secondary endpoints are everything else the trial also tracks — additional outcomes offering supportive or exploratory evidence but not the trial's core statistical target. A heart failure trial might name "reduction in heart failure events" as primary while also tracking ejection fraction and hospitalization rate as secondary. Secondary endpoints matter and can generate real hypotheses for future study — but under FDA guidance on multiple endpoints, they're inherently a lower bar of evidence, partly because testing many outcomes at once raises the odds that something looks significant purely by chance.

Real Examples From Stem Cell Trials

• Heart failure — ejection fraction and cardiac events: The Phase 3 DREAM-HF trial tested an allogeneic mesenchymal precursor cell therapy in over 500 chronic heart failure patients, with its primary endpoint set as recurrent nonfatal decompensated heart failure events, published in the Journal of the American College of Cardiology.
• Parkinson's disease — MDS-UPDRS motor score: BlueRock Therapeutics' Phase 1 trial of the stem-cell-derived therapy bemdaneprocel tracked the MDS-UPDRS Part III, the standard motor-symptom rating scale for Parkinson's, as a secondary efficacy measure alongside its primary safety endpoint. At one year, the high-dose group showed a median 13-point reduction (improvement) in off-medication motor scores, plus roughly two additional hours per day of good symptom control ("ON time").
• Spinal cord injury — the ASIA Impairment Scale: A Mayo Clinic Phase 1 trial of intrathecal adipose-derived mesenchymal stem cells, published in Nature Communications, used safety as its primary endpoint and the ASIA Impairment Scale — a standardized five-level neurological exam — as a secondary measure. Seven of ten participants improved by at least one ASIA grade; three showed no change. The authors cautioned these functional results "are to be interpreted with caution given limits of phase 1 trials."
• Orthopedics — pain and function scores: A 2025 meta-analysis in Stem Cell Research & Therapy pooled eight randomized controlled trials (502 patients) of mesenchymal stem cell injections for knee osteoarthritis, finding statistically significant improvements in WOMAC pain/function scores versus control at 6 and 12 months, with no significant increase in adverse events.

Why a Trial Can "Fail" and Still Contain Good News

DREAM-HF is the clearest illustration of this in stem cell medicine. Its primary endpoint — recurrent nonfatal heart failure events — was not met; event rates were statistically similar between the cell therapy and placebo groups. By the strict letter of the trial's design, DREAM-HF failed. But several pre-specified secondary and subgroup endpoints told a different story. Patients receiving the cell therapy had significantly fewer nonfatal heart attacks and strokes (4.6% vs. 13.0%), and a composite of cardiac death, heart attack, or stroke was also significantly lower in the treatment group. Cardiology press coverage summed it up well: the trial "misses primary endpoint, but signals a path forward." That's the nuance patients need to hold onto — the therapy didn't do the one thing the trial was designed to prove, but it did something else that showed up consistently, which is now being tested as a primary endpoint in follow-up research.

This pattern — a missed primary endpoint alongside a genuinely significant secondary one — recurs often enough in stem cell trials that it deserves its own rule of thumb: a "positive" secondary result from a trial that missed its primary endpoint is a reason for cautious interest and further study, not a reason to conclude the therapy works.

Statistical Significance vs. Clinical Significance

Even a "successful," statistically significant endpoint doesn't automatically mean the effect matters to a patient's actual life. Statistical significance (usually a p-value under 0.05) means a result is unlikely to be due to chance alone. Clinical significance is a separate question: is the size of that effect large enough to matter to someone's health, function, or quality of life? As a health journalism glossary puts it, a pain medication could be statistically significant yet only reduce pain by 0.4 points on a 10-point scale — a real effect, but arguably too small to justify the cost or risk. This is why methodologists increasingly urge trials to report effect sizes and confidence intervals alongside p-values, comparing results against a "minimal clinically important difference" — the smallest change patients themselves would notice — rather than reporting significance alone. A press release that leads only with "statistically significant" and omits the size of the effect is worth reading twice.

Bottom Line

Endpoints are the difference between a real answer and an anecdote. A trial's primary endpoint is the specific, pre-registered test it was designed to pass or fail — and that's the number that should carry the most weight when deciding how seriously to take a "promising" headline. Secondary endpoints and subgroup findings can be genuinely informative, as DREAM-HF and the bemdaneprocel trial both show — but they're a lower tier of evidence, more prone to chance findings, and should be read as "worth watching," not "proven." Before trusting news about any stem cell therapy, ask: What was the primary endpoint, and did the trial meet it? How big was the effect, not just whether it was "significant"? And was this result pre-specified, or found after the fact?