This is an individual patient data meta-analysis, which is the strongest form these syntheses take: the authors obtained the raw trial records through the Vivli platform rather than working from published summaries. Sixteen phase 3 trials of tofacitinib, baricitinib and upadacitinib contributed 11,883 participants. Median BMI was 26.8 kg/m2 and 3,676 participants, about 31%, had class 1 to 3 obesity.
Response fell steadily as BMI rose. Against patients of healthy weight, the adjusted relative risk of an ACR20 response was 0.94 for overweight, 0.92 for class 1 obesity, 0.88 for class 2, and 0.78 for class 3. DAS28-CRP moved the same way, with an adjusted mean difference of 0.54 at class 3 obesity. Heterogeneity was low to moderate and risk of bias was low.
The finding that makes this actionable is the negative one: no BMI gradient appeared in the placebo groups. Heavier patients with rheumatoid arthritis do worse in general, so a gradient in the treatment arm alone could easily have been prognosis rather than pharmacology. Its absence under placebo says obesity is modifying the drug effect itself.
For practice, this moves weight out of the lifestyle conversation and into the treatment decision. A patient with class 3 obesity starting a JAK inhibitor should be told the expected benefit is smaller, and weight management should be arranged alongside the prescription rather than mentioned after it fails.
- Record BMI before starting a JAK inhibitor and use it when setting the expected response, not only when reviewing comorbidity.
- In class 2 and 3 obesity, plan the review point earlier and agree in advance what an inadequate response will trigger.
- Offer weight management as part of starting treatment, with a named route, rather than as general advice.
- Do not read this as a reason to withhold a JAK inhibitor: benefit was reduced, not absent, at every BMI band.
- Where a switch is being considered for non-response, ask whether weight was ever addressed before changing class.
The statistics, in plain English
A relative risk of 0.78 means patients with class 3 obesity were 22% less likely to reach an ACR20 response than patients of healthy weight on the same drug. The confidence interval of 0.71 to 0.85 sits entirely below 1.0, so this is unlikely to be chance. The gradient matters more than any single number: 0.94, 0.92, 0.88 and 0.78 across rising BMI bands is a dose-response pattern, and those are harder to produce by accident than a single threshold effect. I-squared of 0 to 40% means the trials broadly agreed. The most important comparison is the one that found nothing: because the placebo arms showed no BMI gradient, the effect cannot be explained by heavier patients simply having worse disease.
Read the rest in the app
You have read your two free briefings this month. The app carries all 27 specialties, every morning, free — and this finding is waiting in it.
Scan to keep reading on your phone. No account needed to start.
Tomorrow morning, before your first patient
One edition a day for rheumatology, written by the desk, every claim tied to its paper. Six minutes.
Get the app — free