ACR 2026 – Day 1 Report Save

ACR Convergence was off to a big start with the year in review, posters and plenary sessions. Here are a few of my favorite presentations from todays Plenary I session at ACR Convergence. Two of five plenary abstracts were about weight loss, one was about artificial intelligence, and two offered new mechanistic insights with therapeutic implications.

• GLP-1 Weight Loss Modestly Correlates with RAPID3 Lowering in RA and PsA: (Abstract 0867) Nicholas McCormick presented a RISE registry study on the amount of weight lost in RA and PsA patients after starting semaglutide or tirzepatide. Obesity affects 44.6% of RA and 53.3% of PsA patients and worsens disease burden and lowers remission rates. Of 16,848 RA/PsA patients starting a GLP-1 (2018–2024), 1,880 had paired weight and RAPID3 data (mean age 58, 83% women, mean BMI 36.4, baseline RAPID3 13.8). At 12 months, mean weight loss was 15.6 lbs. (6.9%), 53.8% lost ≥5%, RAPID3 fell 1.3 points and 28.9% reached the RAPID3 MCID. Each 5% weight loss was associated with a 0.30-point RAPID3 improvement (equally from pain, patient global and HAQ. Real-world weight loss in RISE (5.8% semaglutide, 8.2% tirzepatide) was low compared to 13–21% seen in other GLP-1RA RCTs.  On questions the author noted that significant changes in the CDAI and CRP were not seen in their preliminary analyses. Link
• AI Platforms Still Hallucinate in Rheumatology: (Abstract 0869) Pitliya and colleagues analyzed the utility and veracity of AI tools/platforms in rheumatology case management. In particular how good was AI and complying with current guidelines and how often is it hallucinated? Her group built 60 standardized scenarios from current ACR and EULAR recommendations across RA, SLE, gout, ANCA vasculitis, osteoporosis and SpA. The scenarios covered diagnosis, biologic selection, steroid tapering, vaccination and complications. Identical case prompts went to ChatGPT (GPT-5.5), Gemini 3, Claude (Opus 4.7), Perplexity and OpenEvidence in March–April 2026. Concordance with ACR/EULAR guidelines ranged from 91% (OpenEvidence) and 89% (ChatGPT) to 85% (Claude), 78% (Gemini) and 75% (Perplexity). Hence not all AI platforms perform equally; with a 16.3-point spread (p<0.001). Hallucination rates also varied nearly five-fold: 3.8% to 18.1% and were the least with Open Evidence (3.8%) and the most with perplexity and Gemini (18% and 15.6% respectively). Not one platform was hallucination-free. OpenEvidence led on accuracy and citation reliability, while ChatGPT gave the most complete and educational answers and was fastest (17.5 s). Errors were seen on every platform, including incorrect dosing, outdated drug selection and inappropriate screening. Dr. Pitliya recommended when using AI - verify every citation, cross-check every dose and confirm the guideline version, because AI output is a draft, not a decision. Link
• TRIUMPH-4: Retatrutide for Knee OA and Obesity: (Abstract 0871) Dr. Philip Conaghan (Leeds) presented the 68-week phase 3 trial of retatrutide, an investigational once-weekly GIP/GLP-1/glucagon triple agonist (half-life ~6 days). The new 2026 ACR OA guideline conditionally recommends GLP-1 agonists for weight reduction alongside diet and exercise, but no GLP-1 is FDA-approved for knee OA pain. TRIUMPH-4 randomized 445 non-diabetic adults with BMI ≥27, moderate-to-severe knee pain (WOMAC pain 4–9) and Kellgren-Lawrence grade 2–3 OA of the knee to receive retatrutide 9 mg, 12 mg or placebo (all received lifestyle counseling and acetaminophen for rescue pain). Patients were heavy (mean BMI 40.4, weight 112.7 kg, 23 years of obesity) and in pain (baseline WOMAC pain 5.95). At week 68, WOMAC pain fell 4.48 (9 mg) and 4.41 (12 mg) versus 2.43 with placebo (p<0.001), with separation by week 16. Weight fell 26.4% and 28.7% versus 2.1%, or 29–32 kg (up to 71 lbs!!). A ≥50% pain reduction was reached by 86% and 78% versus 46% (PBO); and a ≥70% weight reduction was seen by 73% and 68% versus 26% (PBO). Retatrutide also halved hsCRP (−50% vs −25%) and cut triglycerides by ~40%. Side effects were as expected: nausea affected 38–43%, vomiting ~21% (0% on placebo), serious AEs were 9.5–10.8% versus 2.0%, and AE-related discontinuation was 12.9% and 18.2% versus 4.0%. The bottom line: Womac pain started out at 5.95/10 and with weight loss of more than a quarter of body weight dropped by nearly 4.5 points. The 9 mg dose did as well as 12 mg on pain, with fewer dropouts, possibly making it the ideal dose. Link