Pathology

NLP extracts background breast features but adds no independent prognostic value

In 7,603 Stage I–III breast cancers, NLP extracted background features from 9,754 reports with >90% accuracy. Hierarchical clustering identified three background‑feature categories that correlated with tumor subtype and biology. Benign proliferative changes associated with better survival in age/year‑adjusted models (HR 0.91), but this effect disappeared after adjusting for stage and subtype. Conclusion: NLP is feasible for large‑scale feature extraction but these background features do not independently alter prognosis beyond established variables. 1

IDC‑P and tertiary Gleason pattern 5 may predict limited salvage radiotherapy benefit

Central review of 167 JCOG0401 patients found adverse features linked to shorter TTF, including IDC‑P and tertiary GP5. Exploratory interactions suggested SRT benefit was seen without IDC‑P (HR 0.33) but not with IDC‑P (HR 0.77). Similarly, absence of tertiary GP5 associated with a marked SRT benefit (HR 0.099) whereas benefit was absent with tertiary GP5 (HR 0.76). These findings are hypothesis‑generating and require prospective validation before guiding postoperative treatment changes. 2

MRI uptake reduced biopsies and increased detection of GG≥2 prostate cancer nationwide

Nationwide Dutch pathology data (127,856 biopsies, 2015–2021) show diagnostic biopsies fell by 8.7% while MRI terminology rose from 8.2% to 51%. Cancer‑negative biopsies declined from 49% to 29% and GG≥2 detection increased from 30% to 53%. Biopsy‑to‑prostatectomy Grade Group concordance improved from 51% to 60% over time. Interpretation: MRI adoption coincides with fewer unnecessary biopsies and better detection of clinically significant prostate cancer, though confounding by PSA screening and adjunct tests cannot be excluded. 3