Citations

Background Cardiac involvement by extranodal natural killer/T-cell lymphoma (ENKTL) and aggressive NK-cell leukaemia (ANKL) is exceedingly rare and presents with heterogeneous manifestations, often leading to diagnostic uncertainty and delay. We contribute three additional biopsy- or autopsy-proven cases and place them in the context of a literature review of previously reported cases classified according to the 5th edition of the World Health Organization (WHO) Classification of Hematolymphoid Tumors. Case series Three men (ages 58–72) presented with diverse cardiac manifestations. A 72-year-old with progressive dyspnoea and cardiogenic shock had cMRI findings interpreted as myocarditis on an infiltrative substrate; endomyocardial biopsy confirmed cardiac infiltration by ENKTL. Despite brentuximab vedotin (BV)-based therapy he died shortly thereafter. A 60-year-old with encephalopathy, biventricular dysfunction, and widespread nodal/multiorgan disease had nodal pathology consistent with EBV-negative ANKL according to the criteria of the 5th edition of the WHO Classification of Hematolymphoid Tumors; he declined chemotherapy and died within six days, with autopsy confirming multiorgan (including myocardial and bone marrow) infiltration. A 58-year-old with prior sinonasal ENKTL treated with four cycles of SMILE developed a new right atrial mass one month later; the mass was diagnosed by cMRI/PET-CT concordance and an extracardiac (EBUS-lung) tissue confirmation of the same recurrence. BV followed by gemcitabine/oxaliplatin failed to control progression and he died approximately one month after cardiac spread. Across cases, CD56 was positive in all, EBER in two, and CD30 was expressed variably. Conclusions Cardiac ENKTL/ANKL is a rapidly fatal entity with protean features. Multimodal imaging aids early recognition and tissue targeting, but definitive diagnosis rests on tissue confirmation. L-asparaginase-based chemotherapy remains the treatment backbone; CD30- and PD-L1-guided agents such as BV, PD-1 blockade, and daratumumab may offer transient control after biomarker-guided selection but lack randomized data in cardiac disease. Across the reported experience, survival after cardiac involvement was uniformly short, typically days to a few months. Our three cases add to this literature by illustrating the diversity of cardiac presentations, the variable CD30 expression relevant to brentuximab vedotin use, and the contribution of multimodal imaging to diagnosis. Findings from our series and the pooled literature are descriptive rather than constituting a formal survival analysis, and underscore the need for earlier diagnosis and prospective investigation.

Aims: The primary aim of this study was to determine whether CYP2D6 phenotypes are associated with atomoxetine treatment outcomes in children and adolescents with attention-deficit hyperactivity disorder. Patients and Methods: Patients 21 years and younger with a known CYP2D6 genotype and a prescription for atomoxetine were included in this electronic health record review. The primary outcome was defined as discontinuation of atomoxetine due to a lack of efficacy or due to toxicity. CYP2D6 activity scores were adjusted to account for phenoconversion if patients were concomitantly taking a CYP2D6 inhibitor medication. Results: Of 108 patients that met inclusion criteria, including 13 poor, 30 intermediate, 61 normal, and 4 ultra-rapid CYP2D6 metabolizers after adjusting for phenoconversion. Normal/ultrarapid metabolizers were significantly more likely to discontinue atomoxetine due to lack of efficacy (subdistribution hazard ratio [sHR] = 1.9, p = 0.04), while poor/intermediate metabolizers were significantly more likely to stop due to toxicity (sHR 0.4, p = 0.03). A significant nonlinear effect of adjusted activity score on likelihood of discontinuation due to lack of efficacy ( p < 0.001) or toxicity ( p < 0.001) was identified, where higher CYP2D6 activity scores were associated with more discontinuation due to lack of efficacy and lower activity scores were associated with more discontinuation due to toxicity. When CYP2D6 genotype was available prior to prescribing atomoxetine in intermediate/poor metabolizers, lower median initial doses were observed (0.5 mg/kg/day) compared to those who underwent CYP2D6 genotyping after the initial atomoxetine prescription (0.8 mg/kg/day). Conclusion: Patients with normal or increased CYP2D6 phenotypes may be more likely to discontinue atomoxetine due to a lack of efficacy, while those with reduced function CYP2D6 phenotypes may be more likely to discontinue atomoxetine due to toxicity.

Abstract Deep learning models enable the prediction of clinical endpoints from whole-slide images (WSIs), but many such models function as “black boxes”, lacking transparency about whether and which histomorphological patterns drive their predictions, hindering interpretability and clinical adoption. Here we propose a human-in-the-loop explanation framework, MorphoXAI, which provides both local and global interpretability for deep learning models by incorporating human-expert interpretations. At the global level, it reveals the histomorphological patterns on which the model consistently relies to distinguish between classes of WSIs, as well as the patterns associated with confusion between classes. At the local level, it indicates which of these patterns are used in the prediction of an individual WSI and which regions within the slide correspond to such patterns. We validated our method on a deep learning model trained for ovarian tumor histologic subtype prediction. The results show that our framework generates explanations that accurately reflect the histomorphology underlying the model’s predictions at both global and local levels. For interpretability and clinical utility in diagnostic contexts, human evaluation results showed that our explanations were easy to interpret, rich in diagnostic features, and directly helpful for diagnostic decision-making, thereby enhancing pathologist-AI collaboration. Our work highlights that unifying global and local explanations and grounding them in expert-interpreted morphology enhances the interpretability and verifiability of deep learning models, thereby facilitating the transparent deployment of such models in clinical practice.

Abstract Background The risks of breast cancer in African American (AA) women associated with inherited mutations in breast cancer predisposition genes are not well defined. Thus, whether multigene germline hereditary cancer testing panels are applicable to this population is unknown. We assessed associations between mutations in panel-based genes and breast cancer risk in 5054 AA women with breast cancer and 4993 unaffected AA women drawn from 10 epidemiologic studies. Methods Germline DNA samples were sequenced for mutations in 23 cancer predisposition genes using a QIAseq multiplex amplicon panel. Prevalence of mutations and odds ratios (ORs) for associations with breast cancer risk were estimated with adjustment for study design, age, and family history of breast cancer. Results Pathogenic mutations were identified in 10.3% of women with estrogen receptor (ER)-negative breast cancer, 5.2% of women with ER-positive breast cancer, and 2.3% of unaffected women. Mutations in BRCA1, BRCA2, and PALB2 were associated with high risks of breast cancer (OR = 47.55, 95% confidence interval [CI] = 10.43 to &gt;100; OR = 7.25, 95% CI = 4.07 to 14.12; OR = 8.54, 95% CI = 3.67 to 24.95, respectively). RAD51D mutations were associated with high risk of ER-negative disease (OR = 7.82, 95% CI = 1.61 to 57.42). Moderate risks were observed for CHEK2, ATM, ERCC3, and FANCC mutations with ER-positive cancer, and RECQL mutations with all breast cancer. Conclusions The study identifies genes that predispose to breast cancer in the AA population, demonstrates the validity of current breast cancer testing panels for use in AA women, and provides a basis for increased referral of AA patients for cancer genetic testing.