Radiation Oncology

MR-guided split-course hypo-CCRT plus consolidative immunotherapy in stage III NSCLC shows promising survival

Unresectable stage IIIA-C NSCLC patients (n=104) received MR-guided split-course hypofractionated RT, 30 Gy in 6 fractions twice, with weekly docetaxel and cisplatin. Most patients completed chemoradiation, and 71.2% proceeded to consolidative immunotherapy. Median PFS was 27.5 months, with 2-year PFS 54.7% and 2-year OS 65.3%, indicating favorable disease control. Grade 3-4 lymphopenia was frequent, but clinically significant pneumonitis, esophagitis, and fatal hemoptysis were uncommon. 1

Five-day 30 Gy preoperative RT achieves durable local control in high-risk soft tissue sarcoma

This phase 2, single-center trial delivered 30 Gy in 5 preoperative fractions to 110 patients with high-risk extremity or trunk soft tissue sarcoma. At 2 years, 18.9% of evaluable patients had grade 2 or higher radiation toxic effects, with lower rates in the expansion cohort. Major wound complications occurred in 30% of patients, and prolonged wound closure was concentrated among those needing local flaps. Two-year local control, accounting for competing mortality, was 92.4%, with few fractures and amputations observed. 2

Fluciclovine or PSMA PET–guided dose-escalated salvage RT improves outcomes versus prior nonescalated PET-guided RT

EMPIRE-2 randomized men with biochemical recurrence after prostatectomy and negative conventional imaging to [18F]-fluciclovine or [68Ga]-PSMA-11 PET–guided salvage RT. PET findings directed target selection and dose escalation up to 76 Gy to the bed and 56 Gy to the pelvis. Two-year event-free survival for EMPIRE-2 overall exceeded that of the EMPIRE-1 fluciclovine-guided cohort lacking systematic dose escalation, including after propensity weighting. Within EMPIRE-2, 2-year event-free survival was essentially identical for fluciclovine and PSMA-11, with no significant difference. 3

Hybrid AI-driven automated planning generates fast, high-quality plans across multiple disease sites

A hybrid automated planning strategy combined deep-learning dose prediction with clinical-goal–driven inverse optimization to produce deliverable plans within about five minutes. Models were trained at one institution and retrospectively tested at three institutions for five cancer sites with site-specific goal priorities. Over 80% of 250 AI-generated plans met clinical criteria, and 60% were preferred over manual plans in blinded review. Dosimetric analyses showed automated plans matched or exceeded manual plans for target coverage and organ-at-risk sparing. 4