ONEGUILD  ·  A MONTHLY LETTER · SEPTEMBER 2026

A multicentre, double-blind randomised controlled trial published in *Clinical Infectious Diseases* (Lagrand et al., March 2026) enrolled 84 adults with diabetic foot osteomyelitis (DFO). Every patient underwent both a percutaneous bone biopsy and an ulcer bed biopsy; participants were then randomised so their antibiotic regimen was guided by one culture or the other. Percutaneous bone biopsy has long been the reference standard, but ulcer bed biopsy is what many clinicians actually use. At twelve months, remission rates were 31.4% in the bone biopsy group (95% CI 18.1–48.7) and 39.4% in the ulcer bed biopsy group (95% CI 24.2–57.0). Not superior, not inferior — the same clinical outcome from a less invasive procedure. This is a small trial, and the authors are honest about it. But the finding cuts against decades of guideline language and has generated an active correspondence in the journal through July 2026. For clinicians who have been struggling to justify the logistics of routine bone biopsy in DFO — and for patients who would rather not have one — this is the first randomised evidence that the simpler approach may be enough.

Autonomous artificial intelligence for diabetic retinopathy screening has been FDA-authorized since 2018, and since 2021 it has had its own Medicare billing code (CPT 92229). A retrospective analysis of Medicare Part B data published in *Ophthalmic Epidemiology* (Satheesh, Calotti and Bair, August 2026) tracked what happened next, from 2021 to 2023. Use grew quickly, spreading across billing providers, provider types, and states. But the authors conclude that absolute use remained small: early billing uptake rather than meaningful population penetration. It echoes an earlier *JAMA Ophthalmology* cohort (Shah et al., 2024) of nearly five million US patients with diabetes, in which only 4.2% received any of the targeted retinal imaging over five years, and AI-based screening accounted for a small share of it.

This is not a story about AI performance. The technology is validated, reimbursed, and deployable in primary care, and it remains largely unused. It is a story about the gap between what medicine can do and what medicine actually does — the same gap that Frank Ross wrote about in this space in July, in a different key. For an audience working across specialties, the pattern is worth noticing: the frontier of care is not always scientific. Sometimes it is workflow, reimbursement, awareness, or professional inertia. Sometimes it is all four.

Read the 2026 Medicare analysis →

A cost-effectiveness analysis published in the *International Wound Journal* (Hitchman et al., May 2026) modelled the economics of extracorporeal shockwave therapy for diabetes-related foot ulcers from the perspective of the UK's National Health Service. The model was built on data from SOLEFUL, a three-arm double-blinded pilot randomised controlled trial of 74 patients, in which median healing time was 54 days with high-dose ESWT (500 shocks/cm²), 78.5 days with low dose, and 83 days with sham. The analysis tracked a modelled cohort of 1,000 people with unhealed DFUs through healing, major lower limb amputation, or death over a one-year horizon. Cost-effectiveness analyses rarely make it into anyone's reading list, and that is a problem. The question of whether a health system will fund a therapy is often the question that determines whether patients ever receive it — more decisive, in practice, than the trial that showed the therapy works. This is a pilot-based model, not definitive economics, and a fully powered ESWT trial is still needed. But it is the kind of work that turns a promising result into a fundable one, and it deserves more attention than it usually gets.

At the American Diabetes Association Scientific Sessions in New Orleans on 7 June 2026, Denise Faustman of Massachusetts General Hospital presented phase II trial results on the Bacillus Calmette–Guérin (BCG) vaccine in autoimmune diabetes. In the first arm, 34 adults with childhood-onset type 1 diabetes received BCG and 24 received placebo; after five or more years of follow-up, the BCG group showed measurable improvements in glycemic control and a statistically significant reduction in insulin use, with participants spending substantially more time in target glucose range than before the trial. A second arm reported similar findings in patients with latent autoimmune diabetes in adults (LADA). *Nature* covered the results in its news pages on 9 June. A caveat that matters: these are conference-presented data, not a full peer-reviewed publication yet. The BCG hypothesis in T1D has had a long and contested history, with some earlier studies finding effects and others not. But the mechanism being proposed — that a live-attenuated vaccine can modulate autoimmunity off-target — is one of the more interesting frontiers of immunometabolism, and this is the largest and longest positive trial to date. Worth tracking as the peer-reviewed paper appears.