Population genomic screening enables the identification of individuals at high risk of medically actionable conditions before disease onset, yet real-world feasibility studies are lacking. Informed by prior cost-effectiveness modelling, we conducted a prospective nationwide pilot targeting young adults in Australia (aged 18–40 years), offering genomic screening for ten genes linked to hereditary breast and ovarian cancer, Lynch syndrome and familial hypercholesterolaemia. Of 30,017 registrants, 18,573 were invited and 10,263 completed genomic screening (median age 31.9 years, 45.5

PURPOSE:Angiosarcoma and epithelioid hemangioendothelioma (EHE) are two rare vascular sarcomas with limited therapeutic options. Prior reports have shown sensitivity to microtubule-targeting agents in these histologies. We report the efficacy and safety of eribulin in these vascular sarcomas in a pooled analysis of two parallel phase 2 studies. PATIENTS AND METHODS:Patients more than age 18 years with metastatic or recurrent angiosarcoma or EHE were treated with eribulin (1.4 mg/m2 on days 1 and 8 of a 21-day cycle) until progression or unacceptable toxicity. The primary endpoint was objective response rate (ORR) by RECIST 1.1. RESULTS:Twenty-nine patients were accrued to the study, with 25 (85%) having had prior taxane exposure. We observed an ORR of 17% for angiosarcomas, with 6 of 23 (26%) patients achieving disease stability for greater than 6 months, and an ORR of 33% (2/6) for EHE, with two of six continuing treatment for over 12 months. Five patients experienced a >1.3-fold time to progression ratio (TTP2/TTP1) on eribulin compared with the immediately prior therapy. Eribulin tolerability was consistent with published data. CONCLUSIONS:Eribulin showed clinical activity in this largely taxane-pretreated population. Future studies will be needed to confirm activity. See related commentary by Chen, p. 2130.

The phased framework of oncology trials is designed to ensure patient safety and conserve resources by advancing only promising therapies from early- to late-phase testing. Despite decades of refinement, overall trial success rates-defined by the proportion of studies ultimately supporting regulatory approval-remain low, with failures increasingly occurring in late-phase studies. These failures are often contributed to by methodological shortcomings, including suboptimal end point selection, restrictive eligibility criteria, and inefficient trial designs. Although traditional approaches to biomarker discovery, outcome validation, and eligibility refinement have yielded transformational advances, increasing molecular subclassification of tumors into rare subgroups results in the conventional drug development framework being no longer fit for purpose. Artificial intelligence (AI) offers opportunities to enhance the efficiency, precision, and patient-centeredness of oncology trials. Deep learning systems integrate and analyze large data sets to uncover complex patterns often inaccessible to conventional methods. AI has potential applications in patient-trial matching, optimization of eligibility criteria, statistical modeling of survival outcomes, and the identification of novel surrogate end points although these applications remain largely investigational and are not yet established for routine use. This scoping review provides a structured overview of AI applications in oncology trials, with emphasis on outcome selection and surrogate end point evaluation. We also highlight emerging areas with potential for immediate implementation, such as patient selection, biomarker identification and synthetic control arms, to accelerate development and enhance clinical care. However, broader harmonization is needed to ensure reproducibility, transparency, and regulatory confidence before implementation. Ultimately, early and sustained collaboration between trialists, AI developers, and regulators will be essential to ensure that AI delivers meaningful advances in the design, evaluation, and delivery of new medicines.

Osteosarcoma, the most common childhood bone tumor, can occur in rare cancer predisposition syndromes; however, most are sporadic with no known predisposing factors. We investigated the frequency of SMARCAL1 putative pathogenic variants in our large ongoing study of 2119 osteosarcoma patients, their relation to patient characteristics, and the population prevalence. Our analysis uncovered a higher frequency of SMARCAL1 pathogenic variants across 3 osteosarcoma patient sets (1.8%, n = 2119) than in 2625 comparably sequenced cancer-free individuals (0.3%; P < .001). Patients with SMARCAL1 pathogenic variants had statistically significantly improved overall survival compared with patients without these variants (hazard ratio [HR] = 0.36, 95% confidence interval [CI] = 0.14 to 0.96; P = .034). In the UK Biobank (469 557 exomes), there was a 33-fold increased risk of osteosarcoma in individuals with SMARCAL1 pathogenic variants. These results identify SMARCAL1 as a new osteosarcoma predisposition gene and thus warrant follow-up to identify the mechanisms by which SMARCAL1 contributes to the etiology of osteosarcoma.

Introduction:Prospective randomized trials are lacking in extensive-stage or metastatic extra-pulmonary small cell carcinoma (EPSCC). As a result, treatment for EPSCC is largely extrapolated from SCLC studies. Methods:This single-arm phase II trial enrolled patients with EPSCC to investigate the activity of first-line durvalumab (1500 mg every 3 wk) and 4 to 6 cycles of chemotherapy (either cisplatin or carboplatin) with etoposide, followed by maintenance durvalumab 1500 mg every 4 weeks until disease progression. All had comprehensive genomic profiling of an archival tumor specimen, using Illumina TruSight Tumour 170, Illumina TSO500, or Foundation Medicine. The primary end point was progression-free survival (PFS) at 12 months. Key secondary end points were objective response rate, overall survival, and prognostic biomarkers, including tumor mutational burden (ACTRN12621001225808). Results:A total of six participants were enrolled (target: n = 16). The trial closed early due to slow accrual. The primary disease sites were the bladder, colon, gall bladder, esophagus, and pancreas. After a median follow-up of 20.5 months, the rate of PFS at 12 months was 17% (95% confidence interval [CI]: 1%-52%), median PFS was 4.9 months (95% CI: 2.0-12.5 mo), and median overall survival was 8.8 months (95% CI: 6.41-not-estimable). The objective response rate was 50% (95% CI: 19%-81%). No new safety signals were observed. Common mutations were microsatellite stability, RB1 and TP53 mutations. Tumor mutational burden ranged from 2.4 to 20.5 mutations per megabase. Conclusions:In this small prospective study, chemotherapy-durvalumab produces responses comparable to historical chemotherapy outcome, but without clear improvement in durability. Early closure due to slow accrual highlights the challenges of conducting trials in EPSCC. These findings should be considered descriptive and hypothesis generating. This study reveals that durvalumab plus chemotherapy is safe and has some activity in rare small cell cancers outside the lung, but benefits are modest and short lived. Current treatment remains largely chemotherapy based, highlighting the urgent need for larger, collaborative research efforts to find more effective options for these patients.

Indigenous peoples remain under-represented in genomic research and clinical trials. This reflects historical exclusion, mistrust and health system barriers. In this article, we describe a single Australian institution's experience in improving engagement and enrolment of Aboriginal and Torres Strait Islander patients in a cancer genomic sequencing study. We reflect on challenges related to rapport, consent and enrolment and outline practical strategies including relationship-building, targeted resources and a tailored enrolment pathway. We contend that equitable participation in genomic research requires sustained, trust-based and culturally specific engagement, rather than reliance on purely technical or procedural solutions.

For the first time, an anti-cancer therapy-the combination of nivolumab and ipilimumab-has been recommended for tumour-agnostic reimbursement without biomarker requirements. In a bold step, Australia's Pharmaceutical Benefits Advisory Committee (PBAC) proposes that prescribing should be guided by "clinical judgement" for undefined "immunotherapy-sensitive" advanced or metastatic cancers. This decision raises several concerns. Firstly, tumour-agnostic approvals have typically depended on molecular markers such as MSI-H/dMMR and high tumour mutational burden (TMB), which predict substantially higher response rates. In contrast, activity in biomarker-low populations is overall limited. Second, immune-related toxicities, including their potential for long-lasting impact on quality of life, are rather unique to immunotherapy, and particularly significant with dual checkpoint inhibition. Third, a broad tumour-agnostic listing risks incentivising use in tumour types even where evidence is negative, potentially displacing effective therapies and reducing enrolment in clinical trials. Collectively, such authorisation would increase unwarranted variation in clinical practice, expose many patients to serious adverse events without benefit, and undermine the evidence generation needed for sustainable reimbursement. We recommend managed access programs that define explicit eligibility criteria, use biomarker-based patient selection when supported by evidence, and implement structured prospective registries to systematically track patient outcomes, adverse events, and cost-effectiveness. This approach would ensure that data collected can directly inform future evidence-based reimbursement and coverage decisions.

e13683 Background: Mutational patterns offer diagnostic value for cancer type determination, particularly in patients with cancer of unknown primary site (CUP) and synchronous metastases from multiple primaries. Despite increasing adoption of comprehensive genomic profiling (CGP), systematic evaluation of mutational pattern analysis as a standalone diagnostic classifier for cancer type determination remains necessary. Methods: From the MoST pan-cancer program (ACTRN12616000908437) through December 2023, we developed machine learning (ML) cancer type classifiers incorporating age, sex, and CGP data (including TSO500 or FoundationOne CDx panels), pathogenic variants, tumour mutational burden and microsatellite status. Variants were stratified by type and functional impact as features. L2-regularised logistic regression models estimated posterior probabilities in binary classification, implementing a 1-vs-rest strategy across cancer types and hierarchical levels with n≥10 cases. Stratified 10-fold cross-validation was used, with area under the receiver operating characteristic curve (AUC) as the primary metric. Concordance between model predictions and pathology review was assessed in the MoST CUP cohort, where cases with posterior probability or likelihood ratio (LR) methods (threshold > 2.0) generated differential diagnoses and were compared against a pathologist's review. Results: The cohort consisted of 4,990 patients with solid tumours, from which 209 cancer types and hierarchical subtype models were developed. Median AUC for classification across all cancer types was 0.857 (bootstrapped 95% CI: 0.844-0.875), with 70 cancer types (33%) achieving AUC > 0.9. Models demonstrated robust performance for major cancer types: breast (AUC 0.959, 95% CI: 0.922-0.965), colorectal (0.967, 95% CI: 0.946-0.967), prostate (0.953, 95% CI: 0.924-0.964), pancreatic (0.926, 95% CI: 0.905-0.948), gynaecologic (0.884, 95% CI: 0.877-0.890), and non-small-cell lung (0.878, 95% CI: 0.804-0.899) cancers. Analysis of the CUP cohort (n = 153) revealed that model-generated differential diagnoses showed concordance with pathologist assessment in 129 cases (84.3%, 95% CI: 77.6-89.7) for ≥1 broad cancer category and 104 cases (68.0%, 95% CI: 60.0-75.3) for specific diagnostic classifications. The LR method showed concordance in 97 cases (73.9%, 95% CI: 66.1-80.6) for broad categories and 113 cases (63.4%, 95% CI: 55.2-71.0) for specific classifications. Conclusions: Cancer type classification using CGP demonstrated high discriminative performance in selected tumour types. The concordance study suggests that leveraging mutational patterns through ML could provide information beyond pathognomonic alterations, supplementing multidisciplinary assessment in diagnostically challenging cases, particularly CUPs.

e13763 Background: The Australian-led MoST program (ACTRN12616000908437 ) expands access to comprehensive genomic profiling (CGP) for rare, advanced, and treatment resistant cancers. CGP identifies actionable alterations to provide cancer patients with opportunities to access novel therapies through clinical trials. However, the actionability of alterations may depend on trial availability and local patterns of recruitment. This study compares whether CGP actionability for participants (pts) in the MoST-NZ cohort, would have differed if they lived in Australia. Methods: Pts enrolled in MoST-NZ who underwent successful CGP between 2021 and 2024 were included. Pts were discussed at a weekly joint Molecular Tumor Board (MTB) to define actionable alterations and recommend matched trials and targeted therapies. Data collection for the MoST-NZ cohort included pt demographics, actionable alterations, match for trials available in NZ and Australia, matches to targeted therapies, indication for germline genetic referrals, and barriers to trial access. Two analyses were conducted: a retrospective match of MoST-NZ pts to contemporaneous trial availability in NZ or Australia on the day of MTB presentation, using the MoST genomic-trial-matching database; and an audit of actual outcomes for the MoST-NZ pts. Results: The median age of the 190 MoST-NZ patients were 57.1 years (range 18–104), with majority being gastrointestinal (n = 50), sarcoma (n = 41), and gynaecological (n = 32) cancers. Analysis of c trial matching showed 141 (74.2%) of the MoST-NZ pts had alterations that would be matched to ≥1 trial in Australia, compared to 62 (32.6%) matched to NZ. This was similar to real world matching rate in the MoST-NZ cohort, with 51 (26.8%) pts matched to ≥1 trial. 46 of these 51 pts were matched to immunotherapy (IO) or combination IO trials, primarily based on tumor mutational burden (n = 20) and ARID1A loss (n = 20). Regarding final trial recruitment, 16 of the 51 pts (31.4%) were enrolled in trial that aligned with MTB recommendation; a similar recruitment rate previously reported in the Australian cohort (27.2%). The main barriers to trial recruitment in MoST-NZ were protocol ineligibility (n = 11), trial unavailability when eventually indicated (n = 9), clinical deterioration or pt preference (n = 7). Conclusions: CGP provides NZ pts with trial options beyond standard care, but CGP actionability is lower in NZ than Australia due to less availability of clinical trials. However, when trials are available, the rate of recruitment is similar in NZ. These findings underscore the intertwined value proposition between CGP and oncology trials. Locally, NZ would gain more value from CGP by increasing access to clinical trials.

This single-arm phase II trial (ACTRN12620000918921) evaluated the clinical activity of tremelimumab (10 mg/kg intravenously every 4 weeks for 6 cycles) in advanced cancers with a high tumour mutational burden (TMB), defined as TMB > 10 mutations/megabase (mut/Mb) on standard platforms (TSO500 panel, F1CDx), or >20 mut/Mb on the TST170 panel. The primary objective was 6-month progression-free survival rate (PFS6) by iRECIST. Secondary objectives included objective response; the ratio of time to progression (TTP) on study to TTP on prior therapy (TTP2:TTP1); overall survival (OS); and safety. After minimum followup of 12 months, the PFS6 was 6 ≥ 1.3.

Sample sizes of Sites 1-4 across ten experiments

Artificial intelligence applications in biomedicine face major challenges from data privacy requirements. To address this issue for clinically annotated tissue proteomic data, we developed a federated deep learning approach (ProCanFDL), training local models on simulated sites containing data from a pan-cancer cohort (n = 1,260) and 29 cohorts held behind private firewalls (n = 6,265), representing 19,930 replicate data-independent acquisition mass spectrometry runs. Local parameter updates were aggregated to build the global model, achieving a 43% performance gain on the hold-out test set (n = 625) in 14 cancer subtyping tasks compared with local models and matching centralized model performance. The approach's generalizability was demonstrated by retraining the global model with data from two external, data-independent acquisition mass spectrometry cohorts (n = 55) and eight acquired by tandem mass tag proteomics (n = 832). ProCanFDL presents a solution for internationally collaborative machine learning initiatives using proteomic data, for example, for discovering predictive biomarkers or treatment targets while maintaining data privacy. SIGNIFICANCE:A federated deep learning approach applied to human proteomic data, acquired using two distinct proteomic technologies from 40 tumor cohorts across eight countries, enabled accurate cancer histopathologic subtyping while preserving data privacy. This approach will enable the privacy-compliant development of large-scale proteomic artificial intelligence models, including foundation models, across institutions globally.

Distribution of cancer subtypes across Sites 2-4 and Experiments 1-10

Cell type enrichment analysis of top SHAP-ranked proteins across cancer subtypes

Rare cancers account for a quarter of cancer diagnoses in Europe yet clinical research, diagnosis, treatment access, and survival outcomes lag significantly behind common cancers. Despite the potential of precision oncology, the consistent implementation of comprehensive genomic profiling in routine clinical practice and robust evidence-generation remains a challenge in this population, compounded by regulatory hurdles and a lack of investment in drug development. A concerted effort across all stakeholders is required to optimise diagnostics, including access to molecular profiling, to expedite clinical trials and treatment access, and to gather high-quality data, including patient-reported outcomes, in rare cancers. Some initiatives are already showing promise including the establishment of national expert reference centres and European Reference Networks such as EURACAN. However, further collaboration is required to speed up the diagnostic trajectory so that rare cancer patients present with less late-stage disease, and to facilitate clinical trials leading to wider access to precision oncology drugs shown to be safe and effective. In the context of so many hurdles (diagnosis, treatment, research, development and regulatory), there is an even greater role for patient and clinical trial organisations and funders to help fill the aforementioned gaps. Innovative solutions are urgently required to address the high unmet medical need for patients with rare cancers.