Semaglutide (Wegovy) Is Not a Knee Pain Drug

The STEP-9 trial of semaglutide for knee arthritis was positive but it had serious limitations.

Nearly a thousand news outlets covered the recent publication of the positive results of the STEP-9 trial of semaglutide vs placebo for the treatment of knee pain due to osteoarthritis in patients with obesity. Twitter overflowed with positive messages after the New England Journal of Medicine publication.

I am not sold on the trial. STEP-9 had many flaws but one fatal flaw. I will make the case that this trial does not support use of this drug for knee arthritic pain.

The glucagon-like peptide-1 receptor agonists (GLP1a) improve hard clinical outcomes in patients with diabetes and obesity. Multiple trials prove that the drugs induce weight loss in patients with obesity. And they may be disease-modifiers in patients with established heart disease.

We also know that weight loss in obese patients reduces pain and improves mobility. Patients often tell me that their orthopedist told them to lose weight before considering surgery. The idea of using the most potent weight reducers available to treat patients makes sense. And the randomized controlled trial is the only way to answer this question.

The problem with studying knee pain is the placebo effect. This meta-analysis of 198 placebo-controlled trials find that the effect size of placebo is 0.51. Pause there. In cardiology, we get excited of relative risk reductions of 20% and placebo-for-knee-pain can double that of a typical cardiac drug. Here’s a study showing saline injections not only work, but have long-lasting effects.

The most important aspect of a trial of knee pain has to be blinding.

The STEP-9 Trial

In multiple centers, 407 obese patients (average age 56 years, 80% female, mean BMI 40) were randomized (2:1) to either semaglutide or placebo injections.

I would normally write something like this next: “The authors chose a primary endpoint of x….” The first problem with STEP-9 is captured in this copy and paste:

The sponsor (Novo Nordisk) designed the trial, prepared the protocol and statistical analysis plan, and performed the statistical analyses.

The authors interpreted the aggregated data, participated in writing the first and subsequent drafts of the manuscript (with assistance from a medical writer funded by the sponsor…

This does not disqualify the study, but it would be wrong to write that the authors chose the primary endpoint.

The sponsor chose as co-primary endpoints weight loss and change in the Western Ontario and McMaster Universities Osteoarthritis Index (WOMAC) pain score (on a scale of 0 to 100, with higher scores reflecting worse outcomes) from baseline to week 68. An internet search identified a change in 10 points as clinically meaningful.

The primary endpoints were positive. Both groups lost weight. The semaglutide arm lost 10% more than the placebo arm. This was statistically significant—and expected. (I think we can stop adding weight loss as a primary endpoint in GP1a trials.)

The mean change in the WOMAC pain score at week 68 was −41.7 points with semaglutide and −27.5 points with placebo (P<0.001). The difference was 14.2 and that puts it in the clinically meaningful category. Notable though was the massive (nearly 3x) improvement from placebo.

The study team also measured numerous other qualitative outcomes which generally favored semaglutide.

The authors and sponsor then concluded (and NEJM editors let them):

The findings support the use of once-weekly subcutaneous semaglutide at a dose of 2.4 mg for weight management and treatment of pain in persons with obesity and moderate-to-severe pain due to knee osteoarthritis.

My Objections:

Here at Sensible Medicine, we often delve into the data and its analysis. But the more I learn about appraisal of evidence, the more I learn that the bias and problems with trials occur before the first patient is randomized.

The core problem with this trial is blinding. Or lack of it. The authors write in their limitations section that “perceived trial-group assignment and the effect of such perception were not assessed…”

Patients on semaglutide surely could identify their treatment assignment. The drug changes appetite and food handling. The authors and sponsors must know this.

And not assessing blinding—when hundreds of trials have documented the strong placebo effect in assessing knee pain—induces cynical thoughts in my brain.

The authors justified not measuring blinding by writing that “the magnitude and consistency of treatment benefit with semaglutide across outcomes suggests that perceived assignment was unlikely to account for the improvements observed.”

What’s wild is that they cite a paper strongly warning scientists that unblinding is a huge problem in arthritis trials. The conclusion of this paper (as if no one would have looked it up) is this: (emphasis mine)

Additional efforts to ensure adequate blinding, assess the degree to which unblinding has occurred in rheumatology studies, and interpret trial evidence considering this potential threat to trial validity should be undertaken.

The reason that unblinding is so crucial is that both the semaglutide and placebo groups improved massively from baseline in the pain score. That the semaglutide improved more could easily have been from additional placebo effect.

To be fair, I am not sure how you would manage unblinding in a trial of subjective endpoints comparing a GLP1a vs placebo. But not measuring the amount of unblinding is a real problem. A fatal flaw.

It makes me sad that such a flawed experiment led the authors—many of whom have financial relationships with the drug maker—to conclude that semaglutide should be a treatment of knee osteoarthritis.

The other sad aspect of this story is that so many news outlets and Twitter docs promoted this story without critical analysis.

---

Sensible Medicine aims to be the place where you can read critical appraisal of important medical evidence. Few places for this type of work exist. Please consider joining the thousands of Sensible Medicine supporters. JMM