Research Literacy
What Makes a Control Group Good — or Useless?
The short answer
A control group is only as good as its ability to isolate a treatment's real effect from everything else that makes people report feeling better — expectation, natural healing, extra attention from a clinician. In 2021, the largest sham-controlled trial ever run on platelet-rich plasma (PRP) injections randomized 240 patients with chronic Achilles tendinopathy to a real PRP injection or a sham injection (a needle inserted under the skin but never into the tendon), and found no meaningful difference at 6 months (adjusted score difference -2.7, 95% CI -8.8 to 3.3). Two decades earlier, a 180-patient trial did the same for knee arthroscopy, randomizing patients to real surgery or placebo surgery involving only skin incisions, and found the fake surgery worked just as well as the real one for two full years. Both trials share the same design principle: a control condition convincing enough that the patient — and often the person scoring the outcome — cannot tell which treatment they actually received.
Quick answers
- What is it?
- A control group exists to isolate a treatment's real biological effect from everything else that can make people report feeling better — expectation, natural healing, or extra clinical attention.
- Why are people interested in it?
- Many popular treatments look good in weak studies but fail once tested against a convincing fake treatment, so readers need to know what separates a strong control from a weak one.
- What has actually been studied?
- A 2021 sham-injection trial of PRP for Achilles tendinopathy (240 patients), a 2015 Cochrane review of earlier, smaller injection-therapy trials, and the landmark 2002 sham-surgery trial of knee arthroscopy for osteoarthritis (180 patients).
- How strong is the evidence?
- Two well-documented, rigorously controlled randomized trials directly demonstrate the point: both PRP and knee arthroscopy performed no better than a convincing sham once tested with proper blinding, despite looking effective in earlier, weaker-controlled evidence.
- What don't we know yet?
- Whether PRP has a real effect too small for a 240-patient trial to detect, whether a different sham design would change the Achilles result, why the earlier smaller injection-therapy literature showed a murkier signal, and whether current sham designs fully capture patient expectation for "cutting-edge" treatments.
- Where can I read the original research?
- Sources at the bottom of this page: Kearney RS, Ji C, Warwick J, et al., JAMA, 2021; and Moseley JB, O'Malley K, Petersen NJ, et al., New England Journal of Medicine, 2002.
The short answer: Not every comparison group does the same job. A placebo pill, a sham procedure, a waitlist, and an active comparator all technically count as "controls," but they answer different questions and resist bias to different degrees. The strongest ones share one feature: the patient — and ideally the person scoring the result — cannot tell which group they're actually in. Two real trials, run two decades apart on completely different treatments, show what that looks like when it's done well, and what gets exposed when it finally is.
What the research shows
Why a weak control fails quietly
A control group exists to answer one question: compared to what would have happened anyway, did the treatment do anything? Several things can make a person report feeling better that have nothing to do with a treatment's biological effect — natural healing over time, regression to the mean (people often enroll in a trial exactly when their symptoms are at their worst, and symptoms tend to drift back toward average afterward regardless of treatment), extra clinical attention, and plain expectation. A control group that isn't blinded, isn't convincing, or is compared against nothing at all cannot separate any of that from a real treatment effect. The failure mode is quiet: the trial still produces a number, the number can still look positive, and nothing in the result itself announces that it might be measuring belief rather than biology.
A sham injection that isolated the real effect
The clearest recent example comes from a 2021 trial in JAMA. Platelet-rich plasma (PRP) injections had been used for years to treat tendon pain, largely on the strength of smaller, often unblinded studies. Researchers at 24 sites designed a trial specifically to remove that ambiguity: 240 adults with chronic midportion Achilles tendinopathy — tendon pain lasting more than three months, confirmed by imaging — were randomized to a single intratendinous PRP injection (121 patients) or a sham injection: a needle inserted under the skin, but never advanced into the tendon itself (119 patients). Patients were kept blind to which injection they received. The primary outcome, a validated tendon-function score (VISA-A, 0–100, higher is better, with a minimal clinically important difference of 12 points), came in at 54.4 for the PRP group and 53.4 for the sham group at 6 months — an adjusted difference of -2.7 points (95% CI -8.8 to 3.3). That interval comfortably includes zero and falls nowhere near the 12-point threshold the trial itself had set as clinically meaningful. Ninety-two percent of patients completed the trial.
This doesn't prove PRP has no biological activity whatsoever — it means that whatever activity it has didn't outperform a needle stick and the belief that came with it, at this dose, in this condition. That's precisely the distinction a sham control is built to draw.
The murkier picture before the sham trial existed
The same lead investigator had already reviewed the smaller, earlier evidence on injection therapies for Achilles tendinopathy in a 2015 Cochrane review — 18 trials, 732 participants, covering multiple injectate types, not PRP alone. That review found "low quality evidence of a lack of significant or clinically important differences" between injection and control at 6 weeks, 3 months, and 6-to-12 months, but also flagged "very low quality evidence in favour of the injection therapy group" for short-term pain relief — a genuinely murkier signal, built from smaller trials of uneven blinding quality and mixed injectates lumped together. The 2015 review's own conclusion called for exactly the kind of large, sham-controlled trial that followed six years later. That trial didn't just add data — it resolved an ambiguity the earlier, weaker-controlled evidence couldn't.
Fake surgery, tested against real surgery
The same logic applies outside injections. A landmark 2002 trial in the New England Journal of Medicine tested arthroscopic knee surgery for osteoarthritis the same way: 180 patients were randomized to arthroscopic debridement, arthroscopic lavage, or placebo surgery — skin incisions only, staged to look and feel like a real operation, with no arthroscope ever inserted. Both patients and the staff assessing outcomes were kept blind to group assignment. Across five self-reported pain and function scales and one objective walking test, tracked for 24 months, there was no meaningful difference between the placebo group and either real-surgery group. At two years, for example, scores on the Knee-Specific Pain Scale were 51.6±23.7 for placebo, 53.7±23.7 for lavage (p=0.64), and 51.4±23.2 for debridement (p=0.96) — and the study's own confidence intervals excluded any clinically important gap between groups.
That trial changed real-world behavior in a way weaker evidence hadn't. A later analysis of a state ambulatory-surgery database found that after the trial's publication, the share of knee arthroscopies performed specifically for an osteoarthritis diagnosis dropped from 10.6% to 7.2% of cases — roughly 4,000 fewer such procedures per year — even as the overall volume of knee arthroscopy for other reasons continued to grow. A well-controlled sham trial did something a less rigorous body of evidence hadn't managed: it gave surgeons a specific, hard-to-dismiss reason to stop.
What separates a good control from a useless one
Four features recur across both trials. The sham has to be genuinely convincing — a needle stick that doesn't enter the tendon, an incision without an arthroscope — not a comparison so obviously different that patients can guess their assignment. The people scoring the outcome need to be blinded too, not just the patients, especially for subjective measures like pain scales. The trial needs enough patients to rule out a moderate placebo response with a reasonably narrow confidence interval — many of the smaller injection-therapy trials in the 2015 Cochrane pool did not clear that bar on their own. And the primary outcome should be set in advance, so a disappointing result can't be quietly reframed around whichever secondary measure happened to look better afterward.
What we don't know yet
Whether PRP has a real effect too small for a 240-patient trial to detect, or genuinely none. The Kearney trial rules out a clinically meaningful benefit at this size and dose — it does not rule out a smaller one.
Whether a different sham design would change the Achilles result. No retrieved trial has tested whether more closely matching injection volume, pressure, or post-injection sensation shifts the outcome.
Why the earlier, smaller injection-therapy literature showed a murkier signal than the definitive sham trial. Whether that reflects publication bias, weaker blinding, heterogeneous injectates being pooled together, or some combination is not resolved by the retrieved evidence.
Whether patient expectation differs in magnitude between a "cutting-edge" biologic injection or a named surgical procedure and a standard placebo pill, and whether current sham designs fully capture that difference. Neither retrieved trial was built to measure it directly.
Sources
- Kearney RS, Ji C, Warwick J, et al. Effect of Platelet-Rich Plasma Injection vs Sham Injection on Tendon Dysfunction in Patients With Chronic Midportion Achilles Tendinopathy: A Randomized Clinical Trial. JAMA, 2021;326(2):137-144. DOI
- Moseley JB, O'Malley K, Petersen NJ, et al. A controlled trial of arthroscopic surgery for osteoarthritis of the knee. New England Journal of Medicine, 2002;347(2):81-88. DOI
- Kearney RS, Parsons N, Metcalfe D, Costa ML. Injection therapies for Achilles tendinopathy. Cochrane Database of Systematic Reviews, 2015;2015(5):CD010960. DOI
- Amin NH, Hussain W, Ryan J, Morrison S, Miniaci A, Jones MH. Changes Within Clinical Practice After a Randomized Controlled Trial of Knee Arthroscopy for Osteoarthritis. Orthopaedic Journal of Sports Medicine, 2017;5(4):2325967117698439. DOI
Article records maintained via PubMed. Peptide Proof reports what studies found. It does not provide medical advice, dosing guidance, or treatment recommendations.
Sources
- Effect of Platelet-Rich Plasma Injection vs Sham Injection on Tendon Dysfunction in Patients With Chronic Midportion Achilles Tendinopathy: A Randomized Clinical Trial
- JAMA, 2021 · n=240 · DOI
- A controlled trial of arthroscopic surgery for osteoarthritis of the knee
- New England Journal of Medicine, 2002 · n=180 · DOI
- Injection therapies for Achilles tendinopathy
- Cochrane Database of Systematic Reviews, 2015 · n=732 · DOI
- Changes Within Clinical Practice After a Randomized Controlled Trial of Knee Arthroscopy for Osteoarthritis
- Orthopaedic Journal of Sports Medicine, 2017 · DOI
Get the evidence, not the hype.
New research briefs, plainly written and fully sourced. No dosing advice, ever.