Reading the Research: A Patient's Primer

Every regenerative medicine clinic, including ours, will cite "the research." It's worth knowing how to read the research yourself, because the same studies can be summarized very differently depending on who's reading them. This guide is the longer version of how a clinician should evaluate a regenerative medicine paper, and how you can do the same when a clinic hands you a glossy citation list.

You don't need a stats background. You need a handful of mental filters, a willingness to read past the abstract, and the patience to look at the methods section even when it's hard going.

Why this matters

The regenerative medicine space sits in an unusual epistemic position. The biology is real. The evidence base is genuinely growing. But the marketing has consistently run a few years ahead of the literature, and individual clinics frequently cite studies in ways that don't match what the studies actually showed.

If you're being asked to make a multi-thousand-dollar decision about your own joint, you should be able to evaluate the studies cited to you with the same care you'd evaluate any other expensive purchase. That doesn't require a research degree. It requires knowing what questions to ask of a paper.

The questions below are roughly the order I run through them when I read a new study.

Filter 1: What kind of study is it?

In rough order of how much weight to give a single study:

Multi-center randomized controlled trial (RCT). Gold standard. Patients randomly assigned to treatment vs control, multiple sites, ideally double-blinded. The double-blinding is important because both the patient and the clinician's expectations can bias outcomes; in a well-designed regenerative trial, neither knows whether the syringe contained the active intervention or the placebo.

Single-center RCT. Still strong, smaller scale, more vulnerable to site-specific effects.

Prospective cohort study. Patients enrolled before treatment, followed forward. No randomization but real-time data collection.

Retrospective chart review. Looking backward at treated patients. Cheap to run, useful for hypothesis generation, vulnerable to selection bias.

Case series. A handful of patients reported in detail. Hypothesis-generating only. Not evidence of efficacy.

Case report. A single patient. Anecdote with a citation.

A clinic that builds its evidence base out of case reports and series, with no RCTs in the pile, is showing you what's available, not what's strong. Sometimes that's all that exists for a particular indication. That's a fine reason to be cautious about the indication. It's not a reason to weight the case series as if it were a Cochrane review.

Filter 2: What did they actually measure?

Three kinds of outcomes matter in regenerative medicine, and they don't always agree:

Pain. Usually measured on a Visual Analog Scale (VAS, 0 to 10) or a Numeric Pain Rating Scale. Subjective, but the thing patients care most about. The "minimum clinically important difference" (MCID) for VAS in chronic pain is typically considered around 2 points; smaller changes can be statistically significant without being clinically meaningful.

Function. Usually measured with validated joint-specific scores: WOMAC for knee OA, KOOS for knee, ASES for shoulder, ODI for back, FAAM for foot and ankle. These add information pain scores miss; a patient can have unchanged pain but markedly improved function, or unchanged function with reduced pain, and both matter.

Structure. Imaging-based. Has the cartilage actually changed? Has the tendon thickened? These take longer to show up, and most regenerative protocols show modest structural change even when symptoms improve significantly. Structural endpoints typically use MRI-based scoring systems (MOAKS, WORMS for knee; Sugaya for rotator cuff; Pfirrmann for disc).

Quality of life and global outcomes. Some studies report Patient Global Impression of Change (PGIC), SF-36, or other instruments. These can capture meaningful effects that joint-specific scores miss.

Composite endpoints. Increasingly common: a single endpoint that requires improvement on multiple measures simultaneously (often called a "treatment responder" composite). These are harder to game and more clinically meaningful, but they make the responder rate look smaller because the bar is higher.

Pay attention to which the study measured. Pain-only studies tell you about symptoms. Function-only studies tell you about capacity. Structural studies tell you whether the tissue actually changed. The best studies report all three.

Also pay attention to which outcome was pre-specified as the primary endpoint. If a trial pre-registered pain as the primary outcome and the primary outcome was negative, but the authors emphasize a secondary functional outcome in the abstract, that's a common form of motivated reporting that you should weight accordingly.

Filter 3: How long did they follow patients?

This is the filter that disqualifies the most enthusiastic-sounding studies.

Cartilage doesn't rebuild in 12 weeks. Tendon remodeling takes months. The published evidence on MSC therapy in knee OA shows the strongest effects between 6 and 24 months. If a paper reports "remarkable improvement at 3 months" and stops there, you're seeing noise, regression to the mean, and placebo, not the actual signal.

When evaluating a study, find the longest follow-up timepoint and ask: did the effect persist? If a study shows a strong 3-month effect that flattens to noise at 6 months, the protocol probably isn't doing what the headline suggests.

For PRP, the meaningful endpoint is usually 6 to 12 months. Anything shorter is mostly capturing the early signaling response, which is real but doesn't tell you whether the protocol produces durable effect.

For MSC therapy, 12 months is the typical primary endpoint. 24 months is increasingly the secondary durability endpoint, and the better trials are now reporting 5-year data.

A specific pattern to watch for: trials that crossed over their control arm to active treatment after a few months. The cross-over makes the long-term comparison meaningless. The headline may say "improvement sustained at 12 months," but if the control arm received active treatment at month three, you're not comparing apples to apples after that point.

Filter 4: Responder rate, not average change

Most patients reading a study look at the average pain reduction. "Patients improved by 2.4 points on a 10-point pain scale." That sounds great.

It isn't, necessarily. An average is a smoothed-out number. A 2.4-point average could come from everyone improving by 2.4 points, or from half the patients improving by 5 and half not changing at all. The second scenario is far more useful to know about as a candidate.

Look for the responder rate: the percentage of patients who hit a defined meaningful threshold. The standard is usually a 50% reduction in pain or a clinically meaningful change on a function score. A protocol with a 65% responder rate is one where roughly two of three patients hit a real result. A protocol that reports only "average improvement of 2.4 points" is hiding the distribution.

A related concept is "number needed to treat" (NNT). NNT is the number of patients you'd have to treat with the active intervention to produce one additional good outcome compared to the control. A treatment with an NNT of 3 (one in three patients benefits beyond placebo) is meaningfully different from one with an NNT of 10. NNT is rarely reported in regenerative trials but can sometimes be calculated from the published results.

Filter 5: What was the control?

A good control for a regenerative trial is either:

• Saline (true placebo)
• Hyaluronic acid (current standard injectable for OA, which has its own modest effect)
• Sham procedure (when feasible)

A bad control is "no treatment" or "natural history." Without a real control arm, regression to the mean and placebo effect (which in OA can be 20 to 30 percent) will make almost any injection look good.

The 2021 JAMA paper on PRP for knee OA used saline as a control and found PRP no better than saline on the primary outcome. Many earlier PRP studies, comparing PRP to hyaluronic acid without a saline arm, had shown clear PRP superiority. Both findings are real. The interpretation is that PRP probably beats hyaluronic acid but doesn't blow saline out of the water, which lands somewhere honest in the middle.

A subtle issue with sham controls in regenerative medicine: a "sham" procedure that still involves needle insertion into the joint, even without injecting active product, may itself produce some local effect (joint capsule disruption, low-grade inflammatory response). True placebo control is harder to achieve in injection studies than in drug trials.

Filter 6: Conflict of interest

Every legitimate study includes a conflict-of-interest disclosure. Read it. If the senior author owns equity in the cell-product company or is a paid consultant, that's worth knowing. Conflict doesn't invalidate findings, but it sets a higher bar.

The same logic applies to clinic websites that cite "the literature." If every cited paper is from the same small group of authors or the same product manufacturer, treat the case as suggestive rather than settled.

Common forms of conflict in this space:

The senior author is a paid consultant to the cell product manufacturer.

The study was funded by the manufacturer.

The trial was designed and analyzed by the manufacturer's contract research organization.

The primary outcome was negative but the abstract emphasizes a secondary positive outcome.

The trial protocol was amended mid-study to redefine the primary outcome.

None of these makes the study worthless. Some combination of them should make you treat the findings as preliminary rather than definitive.

Filter 7: Is it a meta-analysis?

A well-conducted meta-analysis pools multiple studies of the same question and gives you a more stable estimate than any single trial. The recent MSC-for-knee-OA meta-analyses, pooling 15 to 20 RCTs each, are some of the most useful evidence available right now. They tell you the average effect across multiple labs, with multiple protocols, in thousands of patients.

A bad meta-analysis pools dissimilar studies. If a "meta-analysis" includes case reports, observational studies, and small RCTs with different cell products and different outcome measures, the average is noise dressed up as signal.

A well-done meta-analysis will include:

A pre-registered protocol (often on PROSPERO).

Adherence to PRISMA reporting standards.

A clear inclusion and exclusion criteria description.

A risk-of-bias assessment for each included study (often using Cochrane's RoB or RoB 2 tools).

Heterogeneity analysis (I² statistic). High heterogeneity (I² > 75%) means the studies disagree with each other enough that pooling them is harder to justify.

Subgroup and sensitivity analyses to see if the conclusion depends on which studies are included.

A funnel plot to assess publication bias.

If a meta-analysis is missing several of these elements, weight it accordingly.

Effect size, in plain English

A common source of confusion: studies report statistical results in ways that aren't intuitive.

Statistical significance (p < 0.05) means the observed effect is unlikely to be due to chance alone. It does not mean the effect is large or clinically important. A trivial change can be statistically significant in a large enough study.

Effect size (often reported as Cohen's d for continuous outcomes or odds ratio for binary outcomes) describes how big the effect actually is. Cohen's d of 0.2 is small, 0.5 is medium, 0.8 is large. Most regenerative interventions show medium effect sizes when they work.

Confidence intervals show the range of plausible true effects. A 95% CI of "1.2 to 2.4" is precise. A 95% CI of "0.3 to 5.6" is imprecise, even if the point estimate of the effect was the same.

Absolute vs relative risk reduction. "50% relative reduction in pain" sounds bigger than "the average patient improved from 7 to 3.5 on a 10-point scale." Both can describe the same finding.

The right way to evaluate effect size for a patient decision: think about whether the magnitude of effect you'd plausibly experience is worth the cost, time, and risk of the protocol. Statistical significance alone doesn't answer that.

Where to find the studies

A few practical resources:

PubMed (pubmed.ncbi.nlm.nih.gov). The default. Free, comprehensive, and indexed. Search with specific terms ("mesenchymal stem cell knee osteoarthritis randomized") and filter for trial type and date.

Cochrane Library (cochranelibrary.com). Systematic reviews on most clinical questions. The reviews are notoriously cautious in their conclusions but the underlying analysis is high quality.

ClinicalTrials.gov. Registry of ongoing and completed trials. Useful for seeing what's in progress and what the registered primary outcomes were before the paper was written.

Google Scholar. Useful for cited-by searching and finding gray literature.

The original journal. Many regenerative medicine papers are open access. If you find a paper of interest, look for the PDF link.

When reading, read past the abstract. The abstract is written to summarize favorably; the methods and discussion sections contain the limitations and caveats.

How to read a Cochrane review

Cochrane reviews are systematic syntheses of evidence on a clinical question, conducted under strict methodology. They typically include:

A defined question (population, intervention, comparator, outcome, study type; PICOS).

A pre-registered protocol.

A comprehensive search of multiple databases.

Risk-of-bias assessment for each included study.

A meta-analysis where studies are similar enough to pool.

A summary-of-findings table that grades the quality of evidence for each outcome (GRADE methodology).

Cochrane reviews are useful because they aggregate the available evidence in a transparent way. They tend to be cautious; "more evidence is needed" is a frequent conclusion. That caution is a feature, not a bug.

A worked example

Here's roughly how I'd evaluate a hypothetical paper titled "Intra-articular mesenchymal stem cells for knee osteoarthritis: a randomized controlled trial."

First, the basics. Single center or multi-center? RCT or observational? How many patients? Published in what journal?

Second, the methods. What was the intervention? What was the control? How were patients randomized and blinded? What were the inclusion and exclusion criteria? Who was the population (mild vs severe OA, age range, prior treatments)?

Third, the primary outcome. What was it pre-specified to be? Was it pain (VAS, WOMAC pain) or function (WOMAC function, KOOS function) or composite? What was the follow-up duration?

Fourth, the results. What was the effect size on the primary outcome? Was the confidence interval wide or narrow? Did the effect persist at the longest follow-up? What was the responder rate? Were there pre-specified subgroup analyses?

Fifth, the discussion and limitations. What did the authors acknowledge as limitations? Did they discuss why their findings differ from prior studies (if they do)? What are the relevant comparisons?

Sixth, the disclosures. Who funded the study? What are the authors' financial relationships?

After all of that, I form a view: does this study's finding meaningfully shift my prior on this intervention? Often the answer is "marginally"; single studies rarely overturn an evidence base. Sometimes the answer is "yes, this is a high-quality trial whose result requires me to update."

A short version of the framework

When a clinic hands you a paper, ask:

What kind of study is it?

What did they measure, and how?

How long did they follow patients?

What was the responder rate?

What was the control?

Who funded this and who are the authors connected to?

Did the effect persist at the longest follow-up?

What did the authors acknowledge as limitations?

Eight questions. They'll separate the marketing from the medicine faster than any other filter.

What a strong evidence base looks like, for a given indication

Concretely, here's what I want to see before recommending a protocol for a specific condition:

At least one well-controlled RCT (saline or hyaluronic acid control).

12 months minimum follow-up.

Both pain and function outcomes reported.

A clear responder rate.

Reasonable consistency across multiple studies.

A defensible biological mechanism that explains why it should work in this tissue.

Either independent replication or a strong meta-analysis pooling multiple studies.

For mesenchymal stem cell therapy in knee osteoarthritis, that bar is met. For PRP in lateral epicondylitis, that bar is met. For BMAC in cartilage defects, the bar is partially met. For stem cell IV therapy as a "wellness" intervention in healthy patients, the bar is not met, and we don't offer it as one.

What the absence of evidence does and doesn't mean

A common misuse of literature: "there's no evidence that X doesn't work, so it might work."

Absence of evidence is not evidence of effect. For interventions that haven't been studied in your indication, the honest position is "we don't know whether this works for this." That doesn't mean it doesn't work; it means we shouldn't pretend it does.

Conversely, "there's evidence it works in indication A, so it probably works in indication B" is also a leap that gets made too casually. The biology might generalize. It might not. Specific evidence in your specific indication is what should drive the recommendation.

How we use the literature at Apex

Practical translation:

For indications where we have strong evidence (moderate knee OA, lateral epicondylitis, certain tendinopathies), we'll cite specific studies and explain what they showed.

For indications where the evidence is emerging but reasonable, we'll say so and discuss it as a trial-of-treatment with realistic expectations.

For indications where the evidence is thin or absent, we don't offer the protocol off-label. That's part of why we treat a narrower list of conditions than some other regenerative clinics.

For your specific case, the question we try to answer during consultation is: where does your indication sit on this spectrum, and what's the realistic expected response?

A practical takeaway

If you're researching stem cell therapy for knee osteoarthritis, the question is no longer "is there evidence?" The answer to that is yes. The question is whether your specific case fits the population the evidence supports. That's a workup question, not a marketing one.

To find out where your case falls, request a consultation or call us at (972) 768-2328. We'll look at your imaging, evaluate the indication honestly, and tell you whether the evidence supports a protocol for you.