皇家马斯登医院

Radiogenomics aims to non-invasively predict tumour genotypes from imaging, but most studies assume molecular homogeneity by assigning a single biopsy-derived label to all lesions within a patient. This approach risks substantial label noise given well-documented interlesional heterogeneity. We investigated whether anchoring training to biopsy-confirmed lesions improves radiogenomic model performance and generalisability. We retrospectively analysed 1646 patients (11473 segmented lesions) with contrast-enhanced CT and EGFR mutation status from next-generation sequencing at the Netherlands Cancer Institute, alongside an external NSCLC radiogenomics cohort (n = 158). All visible lesions were segmented, and the exact biopsy site was matched to its segmentation. Radiomic features were extracted, and machine learning models were trained with three lesion selection strategies: all lesions, non-biopsied lesions only, and biopsy-confirmed lesions only. To disentangle label quality from sample size, we created size-matched variants (one lesion per patient) for all-lesion and non-biopsied strategies. All models achieved significant discrimination of EGFR status on internal validation (AUC = 0.62–0.68). However, performance of the all-lesion and non-biopsied models declined on external validation (AUC = 0.55–0.63), while the biopsy-anchored model maintained stable performance (AUC = 0.62), despite having only 1/10th of the training sample size. When training sets were size-matched, the biopsy-anchored approach significantly outperformed a model trained on all available lesions on external validation (p = 0.037). Radiogenomic models trained on biopsy-confirmed lesions outperform conventional all-lesion strategies in external validation, despite using an order of magnitude fewer samples. Prioritising lesion-level label fidelity can mitigate heterogeneity-driven noise, enhancing robustness and clinical translation of imaging-based genomic prediction. Question Does assigning biopsy-derived molecular labels to all lesions introduce heterogeneity-driven label noise that reduces the generalisability of radiogenomic models? Findings Models trained exclusively on biopsy-confirmed lesions demonstrated superior external generalisability compared with all-lesion approaches, despite being trained on substantially fewer samples. Clinical relevance Biopsy-anchored radiogenomics improves the reliability of non-invasive mutation prediction by accounting for tumour heterogeneity, potentially supporting clinical decision-making when tissue sampling is limited or molecular results are discordant across lesions.

Abstract Background Age is a critical driver of the worldwide increasing cancer incidence rates, and with the growing number of older adults diagnosed with cancer, there is a rising need for specialized care. However, geriatric oncology (GO) education remains underrepresented in medical curricula. The gap between demand and adequately trained providers requires the development of more robust GO training programs. This study examined the current state of GO education, and the areas of interest within GO and the expected teaching methods for this specialized content. Methods A multi-center international online survey aimed at assessing the awareness and perceived significance of GO among early-career medical students and nursing trainees was conducted. Results Among 158 participants, only 36.5% had oncology training and 19.6% geriatrics training. Despite this, GO was consistently rated as highly important, with no difference by prior GO training (p = 0.853). Participants expressed strongest interest in mental health, palliative care, and quality-of-care topics, while satisfaction with current training was moderate. Teaching was predominantly lecture-based, although more interactive approaches such as seminars and bedside teaching were preferred. Over half of participants were willing to contribute to improving GO education through feedback or volunteering. Conclusion GO remains underrepresented in current training despite its high perceived importance among trainees. Participants preferred interactive teaching formats, particularly seminars and bedside teaching, and showed strong interest in mental health and palliative care topics. These findings support integrating geriatric oncology into curricula using more practice-oriented teaching strategies.

LBA4 Background: ESR1 mutations ( ESR1 m) constitutively activate the estrogen receptor (ER) and are the most common mechanism of acquired resistance to aromatase inhibitor (AI) + CDK4/6i. Molecular monitoring by ctDNA analysis can detect the emergence of ESR1 m during 1L AI + CDK4/6i. Camizestrant, the next-generation selective ER degrader (SERD) and complete ER antagonist, has shown anti-tumor activity in pts with and without detectable ESR1 m. SERENA-6 is the first global registrational Phase 3 trial assessing a ctDNA-guided approach to detect the emergence of ESR1 m during 1L AI + CDK4/6i to inform a switch in therapy ahead of disease progression. Methods: Pts with HR+/HER2– ABC who had received ≥6 months of 1L AI (anastrozole/letrozole) + CDK4/6i (abemaciclib/palbociclib/ribociclib) were enrolled and had ctDNA tested for ESR1 m every 2–3 months, coinciding with routine imaging. At ESR1 m detection, pts without evidence of disease progression were randomized 1:1 to switch to camizestrant (75 mg) with continued CDK4/6i (type and dose maintained) + placebo for AI vs continuing AI + CDK4/6i + placebo for camizestrant. The primary endpoint was investigator-assessed PFS (per RECIST v1.1). Prespecified interim analysis data cutoff was Nov 28, 2024. Results: 3,256 eligible pts were surveilled for ESR1 m using ctDNA until 315 eligible pts were randomized to switch to camizestrant (n=157) or continue with AI (n=158). All pts remained on the same CDK4/6i. ~50% of randomized pts had ESR1 m detected at the first ctDNA test. Baseline characteristics were well balanced between treatments. After 171 PFS events, hazard ratio for PFS was 0.44 (95% CI 0.31–0.60, p<0.00001; median PFS 16.0 vs 9.2 months). PFS benefit was consistent across subgroups. PFS rate at 12 months was 60.7% (95% CI 51.1–69.0) vs 33.4% (95% CI 24.9–42.2) and at 24 months was 29.7% (95% CI 19.0–41.2) vs 5.4% (95% CI 0.7–18.2). PFS2 hazard ratio was 0.52 (95% CI 0.33–0.81; 27% maturity). OS is immature (12%). Camizestrant + CDK4/6i was well tolerated with safety consistent with the known profiles of camizestrant, and of each CDK4/6i. Rates of treatment discontinuation due to adverse events were 1.3% for camizestrant and 1.9% for AI. Conclusions: Camizestrant + CDK4/6i guided by emergence of ESR1 m during 1L AI + CDK4/6i in pts with HR+/HER2– ABC resulted in a statistically significant and clinically meaningful improvement in PFS. SERENA-6 is the first global Phase 3 trial to demonstrate clinical utility of using ctDNA to detect and treat emerging resistance, ahead of disease progression. These findings represent a potential new treatment strategy to optimize and improve 1L patient outcomes. Clinical trial information: NCT04964934 .

PURPOSE:To develop the REporting checklist for FoundatIon and large laNguagE models (REFINE), an international reporting guideline for transparent and reproducible reporting of foundation model (FM) and large language model (LLM) studies in medical research, including imaging artificial intelligence (AI) applications. METHODS:The protocol was prespecified and publicly archived. A modified Delphi process was conducted to establish reporting standards for unimodal and multimodal FM and LLM applications involving text, imaging, and structured data. The steering committee coordinated protocol development, expert recruitment, all Delphi rounds, and the harmonization phase. Decisions were made based on predefined consensus thresholds. In Rounds 1 and 2, structured ratings and free-text feedback informed iterative revisions. In the post-Delphi harmonization phase, terminology was standardized, and detailed reporting instructions were finalized. RESULTS:The REFINE development group comprised 57 contributors from 17 countries, and 54 panelists from 16 countries completed Rounds 1 and 2. The harmonization phase was completed by three expert panelists and the steering committee. The entire process produced a 44-item, six-section framework with standardized terminology and detailed reporting instructions, supported by an online platform for practical use (https://refinechecklist.github.io/refine/checklist.html). CONCLUSION:The REFINE provides a comprehensive, consensus-based reporting standard for medical FM and LLM research, including imaging AI studies. The online version facilitates practical implementation. CLINICAL SIGNIFICANCE:The REFINE enables transparent, comparable, and reproducible reporting of FM and LLM studies, supporting reliable evidence synthesis in medical and imaging-focused AI studies.