PURPOSE Tubo-ovarian cancer (TOC) is a sentinel cancer for BRCA1 and BRCA2 pathogenic variants (PVs). Identification of a PV in the first member of a family at increased genetic risk (the proband) provides opportunities for cancer prevention in other at-risk family members. Although Australian testing rates are now high, PVs in patients with TOC whose diagnosis predated revised testing guidelines might have been missed. We assessed the feasibility of detecting PVs in this population to enable genetic risk reduction in relatives. PATIENTS AND METHODS In this pilot study, deceased probands were ascertained from research cohort studies, identification by a relative, and gynecologic oncology clinics. DNA was extracted from archival tissue or stored blood for panel sequencing of 10 risk-associated genes. Testing of deceased probands ascertained through clinic records was performed with a consent waiver. RESULTS We identified 85 PVs in 84 of 787 (11%) probands. Familial contacts of 39 of 60 (65%) deceased probands with an identified recipient (60 of 84; 71%) have received a written notification of results, with follow-up verbal contact made in 85% (33 of 39). A minority of families (n = 4) were already aware of the PV. For many (29 of 33; 88%), the genetic result provided new information and referral to a genetic service was accepted in most cases (66%; 19 of 29). Those who declined referral (4 of 29) were all male next of kin whose family member had died more than 10 years before. CONCLUSION We overcame ethical and logistic challenges to demonstrate that retrospective genetic testing to identify PVs in previously untested deceased probands with TOC is feasible. Understanding reasons for a family member's decision to accept or decline a referral will be important for guiding future TRACEBACK projects.

Forming a reliable judgement of a machine learning (ML) model's appropriateness for an application ecosystem is critical for its responsible use, and requires considering a broad range of factors including harms, benefits, and responsibilities. In practice, however, evaluations of ML models frequently focus on only a narrow range of decontextualized predictive behaviours. We examine the evaluation gaps between the idealized breadth of evaluation concerns and the observed narrow focus of actual evaluations. Through an empirical study of papers from recent high-profile conferences in the Computer Vision and Natural Language Processing communities, we demonstrate a general focus on a handful of evaluation methods. By considering the metrics and test data distributions used in these methods, we draw attention to which properties of models are centered in the field, revealing the properties that are frequently neglected or sidelined during evaluation. By studying these properties, we demonstrate the machine learning discipline's implicit assumption of a range of commitments which have normative impacts; these include commitments to consequentialism, abstractability from context, the quantifiability of impacts, the limited role of model inputs in evaluation, and the equivalence of different failure modes. Shedding light on these assumptions enables us to question their appropriateness for ML system contexts, pointing the way towards more contextualized evaluation methodologies for robustly examining the trustworthiness of ML models.CCS Concepts: • Computing methodologies → Machine learning approaches.

No abstract

No abstract

No abstract