Every number in this study went the right way, and the authors wrote a sentence in their own methods that stops any of it being read as the drug’s doing. That sentence is worth more than the results, and it is the thing the randomized knee trial has and this does not.
What was measured
A retrospective, self-controlled cohort of 93 adults with knee osteoarthritis and overweight or obesity, treated in routine rheumatology care. [1] 88 completed six months. Pain on a visual analog scale, WOMAC function, anthropometry, inflammatory markers, glycated hemoglobin, lipids, dose exposure and safety were all tracked.
At six months, mean weight had fallen 10.88 kg, pain by 2.47 points, WOMAC total by 22.50 points, C-reactive protein by 2.67 mg/L and sedimentation rate by 7.62 mm/h — the last two being inflammation markers rather than symptoms, and so the easiest things here to move.
The phenotypes
The authors also ran an exploratory K-means clustering across dose, weight loss, body mass index change, pain and WOMAC improvement, C-reactive protein, sedimentation rate and glycated hemoglobin, to sort patients into response patterns.
K-means always returns clusters. Give it eight variables and 88 people and it will partition them, whether or not the groups mean anything. The authors call the analysis exploratory and examined how stable the solutions were under resampling, which is the right precaution — and a set of phenotypes found once in 88 people is a hypothesis, not a taxonomy, in the same way a subgroup effect found once is.
What it is good for
Describing what happened to a real clinic’s patients over six months, which is genuinely useful and is not the same as showing what caused it.
If you have knee osteoarthritis and are considering one of these drugs, the pain and function evidence worth weighing comes from the trial with a placebo arm. This cohort adds inflammatory markers and a picture of routine care, and it adds them with the caveat its own authors attached — which is more than most real-world reports manage.