ATTD: Glucose → Predictions
At the ATTD conference experts argued the field must shift from raw glucose streams to AI‑driven predictions to improve Type 2 diabetes management — moving from hindsight to actionable foresight for glycemic control. The conversation highlights model‑based forecasting as the next step for making CGM data clinically useful in T2D care ATTD conference spotlight — shift to AI predictions for T2D | X.
ATTD 2026 ran Mar 11–14 in Barcelona) and staged a symposium titled "Beyond Glucose: The Integrated Future of CGM, CKM, and AI" where SIBIONICS laid out AI-driven insight tools). SIBIONICS presented GS3 AI insights) while device makers including Dexcom said they would share new Type 2 diabetes outcome data and product roadmaps at the meeting). Academic groups highlighted new forecasting architectures, citing a 2025 npj paper on a large sensor foundation model pretrained on CGM data) and recent preprints for state‑space (SSM‑CGM) and transformer (AttenGluco) approaches on arXiv). Speakers and reviewers flagged practical barriers: a 2025 Frontiers review listed data governance, algorithm optimization, and ethical challenges), and the SSM‑CGM team stressed clinical interpretability as a design requirement). Conference retrospectives pointed to measurable benefits already seen with automation and CGM in Type 2 care — Tandem’s Control‑IQ+ study at ATTD 2025 reported a 24% increase in time‑in‑range and a 0.9% A1c drop over 13 weeks reported by diatribe.org), while Dexcom has highlighted a 15‑day G7 sensor with an overall MARD of ~8.0% in its reporting investor release). Organizers and presenters called for prospective validation: the ATTD program lists multiple AI and forecasting sessions) and researchers urged testing models on Type 2 cohorts such as AI‑READI and AIREADI to prove real‑world benefit before clinical adoption as described in SSM‑CGM and AttenGluco papers).
