AI reclassifies one in nine immune-low breast tumours, and esketamine fails its anxiety trial

The edition in brief

Stromal tumour-infiltrating lymphocytes were rescored in 4,262 tumour samples from the APHINITY trial by manual, automated digital and AI-based methods. Manual scoring was reproducible between five pathologists (intraclass correlation 0.84) but agreed only modestly with the automated approaches, and AI reclassified 120 of 1,035 node-positive tumours from immune-low to immune-high — a group whose survival curves separated more with pertuzumab than tumours both methods called immune-low. Higher lymphocyte levels were associated with better invasive disease-free survival by every method, and the largest six-year absolute gain from pertuzumab, 12.1 percentage points, was in node-positive disease with manual scores of 70% or more. A meta-analysis of 12 studies of minimally invasive nipple-sparing mastectomy found longer operating times but less blood loss, shorter incisions and lower rates of overall complications, skin necrosis, seroma and haematoma, with no difference in nipple or flap necrosis or in recurrence — though long-term oncological follow-up remains thin. A phase 1 trial of the nectin-4 antibody-drug conjugate SHR-A2102 in 395 pretreated Chinese patients found no maximum tolerated dose up to 10 mg/kg, grade 3-4 treatment-related events in 54%, and two treatment-related deaths; no response rates are given in this report. The practice-changer is negative: in 180 patients with preoperative anxiety having major abdominal cancer surgery, esketamine did not improve the anxiety response rate at day 3 (39.5% vs 42.5%, risk ratio 0.93) or anxiety scores at day 30.