Citations

ABSTRACT Background Evaluating fairness in artificial intelligence (AI) is essential before clinical implementation to prevent unintended harm across diverse patient populations. This study assesses performance differences of an image-based surgical site infection (SSI) detection model across age, gender, geography, race, and skin color. Methods Model development/validation study using retrospective, multicenter data across surgical subspecialties from adults who submitted an image of their incision via the electronic health record within 30 days of their procedure between 2019-2024. Area under receiver operating curve (AUROC) was compared by age (<55 vs. ≥55 years old), gender (male vs. female), geography (Rochester, Minnesota vs. Jacksonville, Florida vs. Scottsdale, Arizona vs. regional Minnesota/Wisconsin areas), race (White/Caucasian vs. non-White/Caucasian), and skin color (Fitpatrick I-III vs. IV-VI and Monk A-E vs. F-J). Results Training data contained 13,702 images from 4,274 patients between 2019-2022 (50.1% ≥55 years old; 63.2% female; 91.6% White/Caucasian; 58.5% Rochester, 17.3% Florida, 10.5% Arizona; 96.3% light skin tone), of which 1,348 images (9.8%) had an SSI. No differences in SSI detection existed for age, gender, race, or skin color (AUROC 0.83-0.87, p>0.09), however geographical differences were significant: lowest AUROC 0.83 (95% CI 0.77-0.88) vs. highest AUROC 0.87 (95% CI 0.84-0.89, p=0.01). Temporal validation using 4,242 images from 1,353 patients between 2022-2023 showed intact fairness across subgroups (ΔAUROC ≤0.06) except geography (ΔAUROC ≤0.16). Conclusions The model behaved fairly across age, gender, race, and skin color, however geographical differences persisted. These findings have practical implications underscoring the importance of evaluating bias in clinical AI before deployment.

ABSTRACT Aim To systematically review and meta‐analyze cohort studies evaluating associations of gingival inflammation, periodontitis, and tooth loss with cognitive decline and incident mild cognitive impairment or dementia. Methods PROSPERO‐registered searches of MEDLINE, Embase, Scopus, Web of Science, and the Cochrane Library identified cohort studies from inception through October 10, 2025, with a PubMed update on June 24, 2026. Random‐effects meta‐analyses pooled adjusted hazard ratios (HRs) for dementia/MCI; cognitive function outcomes were summarized qualitatively. Risk of bias (RoB) was assessed using the Newcastle–Ottawa Scale. Results Fifty‐nine studies met inclusion criteria; 28 contributed to dementia/MCI meta‐analyses. Severe periodontitis (HR: 1.16; 95% CI: 1.06–1.27; I 2 = 0%; Q ‐test p = 0.49), periodontitis of unspecified severity (HR: 1.45; 95% CI: 1.21–1.73; I 2 = 98%; Q ‐test p < 0.001), edentulism (HR: 1.26; 95% CI: 1.19–1.33), and low tooth count (HR: 1.27; 95% CI: 1.14–1.42; I 2 = 94%; Q ‐test p < 0.001) were associated with elevated dementia/MCI risk. RoB was low in 31 studies (53%), moderate in 25 (42%), and high in 3 (5%); inadequate confounder adjustment was common. Meta‐regression showed stronger estimates for ICD‐code periodontitis definitions (fold change (FC) = 1.44 95% CI: 1.04–2.01), older cohorts (FC per year = 1.02 95% CI: 1.00, 1.05), and weaker estimates with lower RoB (FC per NOS star = 0.81 95% CI: 0.73–0.92). Follow‐up duration, cohort age, and RoB were not associated with tooth‐count estimates. Conclusion Periodontitis and tooth loss are associated with modestly higher dementia/MCI risk and poorer cognitive function, but heterogeneity, residual confounding, and bias temper causal interpretation. Longitudinal studies with rigorous baseline cognition assessment and intervention studies are needed.

Abstract Circumscribed meningeal melanocytic neoplasms (CMMNs) represent a spectrum of rare central nervous system (CNS) tumors. Knowledge of their clinical behavior and molecular correlates remains limited. In this single‐institution retrospective study, we analyzed 31 patients (14 male, 17 female; median age, 59 years) diagnosed with CMMN between 2004 and 2025. In this patient cohort, based on the current World Health Organization Central Nervous System Tumor Classification (WHO CNS5) criteria, the majority of cases (18, 58.1%) fell into the intermediate‐grade melanocytic tumor (IMT) category based on histology, with either increased mitotic activity, CNS invasion, or both; 7 (22.6%) were melanocytomas, 5 (16.1%) melanomas, and 1 (3.2%) indeterminate case because of limited tissue. By next‐generation sequencing, mutations were identified in GNAQ/GNA11 in 21/24 cases, BAP1 in 5/23 (2 IMT, 3 melanoma), EIF1AX in 9/21 (1 melanocytoma, 8 IMT), and SF3B1 in 3/23 (2 IMT, 1 melanoma). Primary sites were spinal (20), posterior fossa (6), and supratentorial (5). During follow‐up, progression was observed in 5/6 melanocytomas, 9/13 IMTs, and 4/5 melanomas, suggesting a high frequency of progression across all groups. Two‐year progression‐free survival was 83.3% (95% confidence interval [CI], 53.5%–100.0%) for melanocytoma, 55.9% (95% CI, 26.7%–85.2%) for IMT, and 25.0% (95% CI, 0.0%–67.4%) for melanoma. Five‐year overall survival (OS) was 80.0% (95% CI, 44.9%–100.0%) for melanocytoma, 65.3% (95% CI, 37.0%–93.6%) for IMT, and 0.0% for melanoma. OS was significantly worse in melanoma compared with melanocytoma ( p = 0.002). BAP1 mutation correlated with worse OS (hazard ratio 8.73, p = 0.006). Upfront maximal safe resection appeared to be associated with improved outcomes in selected patients, whereas radiotherapy and systemic therapy did not demonstrate clear additional benefit in this cohort. In summary, our study shows that all CMMNs, irrespective of grade, have a high propensity to recur and melanoma histology and BAP1 mutation are associated with significantly worse OS.

Transcranial magnetic stimulation can provide noninvasive measures of cortical excitability and plasticity, but relationships between these measures and clinically observable features of autism are not fully characterized. Nineteen autistic participants aged 15 to 40 years were included in the analysis of a left primary motor cortex continuous theta-burst stimulation (cTBS) biomarker protocol. Motor evoked potentials were measured at baseline and at seven assessments from 5 to 60 min after stimulation. Linear mixed-effects models evaluated whether clinical measures moderated the post-stimulation motor evoked potential log-response ratio over time. Aberrant Behavior Checklist (ABC) and Attenuated Behavior Questionnaire (ABQ) analyses were restricted to the same 15 participants with caregiver-informant assessments. Catatonia severity, social impairment, ABQ Motor Total, ABQ Total, and cognitive ability did not significantly moderate the post-stimulation response. ABC Inappropriate Speech was associated with progressively greater post-stimulation facilitation (standardized Time-by-Inappropriate Speech interaction: beta = +0.688, SE = 0.210, 95% CI +0.276 to +1.099; Holm-adjusted P = .008 across eight informant-rated models). Two individual speech-related items ("Talks excessively" and "Talks to self loudly") survived false-discovery-rate correction. An exploratory eight-item ABC phenotype showed a large descriptive in-sample association (beta = +0.904, SE = 0.200, P less than .001) and directionally positive held-out performance. However, full-pipeline permutation tests were nonsignificant for both Pearson (r = .357, two-sided empirical P = .360) and Spearman correlations (rho = .446, two-sided empirical P = .261). Caregiver-rated inappropriate speech may be associated with altered post-cTBS motor cortical facilitation in autism. The candidate phenotype remains hypothesis-generating and requires independent validation.

Abstract BRCA1 -deficient cancers experience persistent replication stress and structural genome instability yet retain the capacity for sustained proliferation, implying reliance on compensatory genome-maintenance mechanisms. Here, we establish BRCA1–FANCM synthetic lethality in human BRCA1 -deficient breast cancer and exploit temporally controlled FANCM depletion to capture genome evolution over successive cell divisions before declining cellular fitness becomes limiting. We show that FANCM restrains genome-wide structural variation in BRCA1 -deficient breast cancer cells under endogenous replication stress. FANCM loss amplifies the characteristic BRCA1 -associated short tandem duplication (TD) phenotype while permitting larger, including megabase-scale, TDs and diverse rearrangements to emerge. Newly emerged TDs preferentially associate with Pol II-occupied regions, and FANCM depletion increases proximity between the replication machinery and elongating RNAPII in BRCA1 -mutant breast cancer cells, linking FANCM-mediated genome protection to transcription–replication encounters. BRCA1 -altered human tumors with low FANCM expression recapitulate key features of this phenotype, while genome–transcriptome integration links newly emerged SVs to configuration-dependent local transcriptional changes. Together, these findings establish FANCM as a replication-stress safeguard coupling survival to restraint of structural genome evolution.

ABSTRACT The cohesin complex, composed of SMC1, SMC3, RAD21, and STAG1/STAG2, is essential for chromosome cohesion, DNA repair, and transcriptional regulation. Pathogenic variants in cohesin components cause cohesinopathies. The classical characteristics of cohesinopathies include developmental delay (DD), intellectual disability (ID), feeding difficulties, hypotonia, short stature, hearing loss, and dysmorphic features. Here, we present a 5‐year‐old boy with classical cohesinopathy features, including DD/ID and feeding difficulties, along with non‐classical features such as hypoglycemia, recurrent fever, and immunodeficiency. Trio exome sequencing identified a novel de novo missense variant of uncertain significance (NM_005862.3:c.643G>A(p.Val215Ile)) in the STAG1 gene. The variant localizes to the RAD21 interaction interface, and molecular dynamics (MD) simulations revealed conformational changes comparable to other STAG1 variants reported as likely pathogenic in patients, supporting a deleterious effect which may disrupt the STAG1‐RAD21 interaction interface. This case expands the phenotypic and molecular spectrum of STAG1‐related cohesinopathy and advances our understanding of the disease mechanism.

Prolonged durations of hypotension during septic shock promote organ injury and mortality. Empirical evidence suggests that earlier vasopressin initiation may improve outcomes in septic shock. The objective of this study was to determine the effect of a ready-to-use vasopressin formulation availability in intensive care unit automated dispensing cabinets on outcomes in septic shock. This was a pre-post quasi experimental study of adults admitted for septic shock who received vasopressin as a second line vasopressor after norepinephrine. Inverse probability of treatment weighting with propensity scores was used to balance baseline covariates between those who received the ready-to-use vasopressin product and the traditional pharmacy-compounded vasopressin product. Outcomes included time from vasopressin order to administration, time to target mean arterial pressure, length of stay, and mortality. A total of 1104 patients with septic shock were included. After propensity score adjustment, the ready-to-use group was associated with faster time to vasopressin administration (17 vs 41 min, P < 0.001), faster time to target mean arterial pressure (hazard ratio 1.29, 95% CI 1.14 to 1.47, P < 0.001), and a shorter length of intensive care unit stay (hazard ratio 1.23, 95% CI 1.06 to 1.43, P = 0.006) as the compounded group. Mortality was not significantly different. In patients with septic shock, availability of a ready-to-use vasopressin product in the intensive care unit automated medication dispensing cabinet was associated with faster time to administration, improved time to target mean arterial pressure, and reduced duration of intensive care.

Abstract Senescent cell accumulation and fibrosis play a fundamental role in kidney ageing. Our aim was to determine whether the angiotensin IV/insulin-regulated aminopeptidase (IRAP) axis of the renin-angiotensin system contributes to renal function decline, fibrosis, and senescent cell accumulation in aged mice. We studied the role of IRAP in kidney ageing using IRAP knockout (KO) mice, as well as in wild-type (WT) mice treated with an IRAP inhibitor. Complementary proximal tubule and collecting duct cell models were used to determine the mechanisms by which IRAP inhibition exerts protective effects. Glomerular filtration rate (GFR) declined by 25% in WT mice between 3 and 23 months of age, concomitant with marked increases in albumin excretion, glomerulosclerosis, tubulointerstitial fibrosis, and accumulation of senescent cells in the proximal tubules of the kidney. IRAP KO mice were protected against age-related decline in GFR, renal fibrosis, and senescence burden compared with age-matched WT mice. IRAP inhibition for 4 weeks (HFI-419, 500 ng/kg/min) in 28-month-old WT mice similarly prevented age-related decline in GFR and attenuated expression of cellular senescence markers. Using an in vitro model of cultured proximal tubule cells, IRAP inhibition mitigated the pro-senescent effects of oxidative stress. This was associated with shifts in the metabolic phenotype of proximal tubular cells, indicating the senoprotective effects of IRAP inhibition may be mediated in part via mitochondrial function. These studies provide evidence that IRAP deficiency is protective against chronic renal cellular senescence, fibrosis, and functional decline. Therefore, IRAP inhibition has therapeutic potential to attenuate pathological cardiorenal ageing through senoprotective mechanisms.