Endocrinology

Systematic review: AI-based nutrition interventions give modest, short-term metabolic gains.

This systematic review included 16 studies with 10,863 participants evaluating AI-driven nutrition interventions in humans. Several randomized trials reported short-term improvements in glycemic control, weight, fat mass, lipids, or symptom scores favoring AI-supported programs. Most positive effects occurred within multimodal programs, limiting attribution to the AI component alone. Risk of bias and methodological heterogeneity were moderate-to-high, so clinical effectiveness remains preliminary. 1

JADE-PRISM: precision-biologic algorithm improved targets but not 3‑year complications in young-onset diabetes.

In 884 Chinese adults with young-onset diabetes, the JADE-PRISM precision program raised attainment of three or more treatment targets (23.8% vs 12.2%, P < 0.001). The intervention improved proxies of β-cell function and reduced emotional distress compared with JADE-only care. The primary composite of incident or progressive diabetes complications occurred similarly (26.3% vs 28.2%; OR 0.908, 95% CI 0.675–1.221). The program is feasible in a technology-enabled specialty setting but did not reduce complications over three years. 2

Phase 2 monotherapy trials in solid cancers show lower efficacy and similar toxicity versus phase 3.

Meta-analysis of 130 phase 2 and 52 phase 3 monotherapy arms (6,665 vs 18,694 patients) found pooled objective response 7% in phase 2 versus 24% in phase 3 (p < 0.0001). Median PFS and OS were shorter in phase 2 (3.23 vs 5.43 months and 9.46 vs 14.44 months respectively; both p ≤ 0.0001). Drug-related grade 3–4 adverse events were not lower in phase 2 (30% vs 25%). Authors argue phase 2 exposure should be justified by research aims, not assumed therapeutic value. 3