Importance:A unified salivary gland carcinoma (SGC)-specific tumor-node-metastasis (TNM) classification can enhance prognostic accuracy, support clinical decision-making, and improve the quality of patient care. Objective:To derive and validate an SGC-specific pTNM classification with improved prognostic accuracy and optimized stage distribution for version nine of the American Joint Committee on Cancer/Union for International Cancer Control staging protocol. Design, Setting, and Participants:This retrospective prognostic cohort study derived a novel pTNM classification using data from the National Cancer Database (NCDB) of patients with surgically treated major SGC (2012-2017) and validated it in an international major SGC cohort (2008-2021) and a single-institution minor SGC cohort (Memorial Sloan Kettering Cancer Center; 1985-2016). Data were analyzed from June to November 2024. Exposures:Surgery with or without postoperative radiotherapy or chemoradiotherapy. Main Outcomes and Measures:The primary end point was overall survival (OS). Cox proportional hazards multivariable analysis was used to confirm the prognostic importance of pathologically positive lymph node (LN) number and extranodal extension (pENE) and derive an optimal pTNM classification. Results:The NCDB dataset included 8409 patients with SGC: 7659 with M0 disease (5748 with pN0 disease and 1911 with pN+ disease) and 750 with M1 disease. Among the 7659 patients with M0 disease, the median (IQR) age was 60 (48-71) years, and 3861 (50.4%) were male. The median (IQR) follow-up was 88.4 (72.3-108.5) months. The 5-year OS was 87.2% (95% CI, 86.3-88.0) for N0 disease, 68.2% (95% CI, 63.9-72.8) for 1 positive LN without pENE, 60.2% (95% CI, 53.5-67.5) for 2 positive LNs without pENE, 68.4% (95% CI, 58.0-76.6) for 3 positive LNs without pENE, 47.5% (95% CI, 41.6-52.8) for more than 3 positive LNs without pENE, and 41.4% (38.1-44.8) for pENE-positive LNs. Multivariable analysis confirmed the independent prognostication of LN count compared with pN0 disease (1 positive LN: adjusted hazard ratio [aHR], 1.70; 95% CI, 1.44-2.01; 2 positive LNs: aHR, 1.61; 95% CI, 1.31-1.98; 3 positive LNs: aHR, 2.10; 95% CI, 1.65-2.68; 4 positive LNs : aHR, 2.46; 95% CI, 1.87-3.24; more than 4 positive LNs: aHR, 2.07; 95% CI, 2.08-2.91) and pENE-positive LNs compared with pENE-negative LNs (aHR, 1.27; 95% CI, 1.10-1.48). The proposed pN classification were pN1 for 1 to 3 positive LNs and pENE negativity and pN2 for more than 3 positive LNs or pENE positivity. Model fit improved with the proposed pN classification vs the current pN classification (Akaike Information Criterion, 26 442 vs 26 483). Based on the aHR model, the following stage groups were proposed: stage I: T1N0 (1 [reference]); stage II: T2N0 (aHR, 1.34; 95% CI, 1.11-1.61); stage IIIA: T1-2N1 or T3-4N0 (aHR, 2.36; 95% CI, 1.99-2.80); stage IIIB: T1-2N2 or T3-4N1-2 (aHR, 5.15; 95% CI, 4.38-6.06); and stage IV: M1 disease (aHR, 13.61; 95% CI, 11.37-16.29). The C index values were similar (proposed classification: 0.792; current classification: 0.790), while the AIC improved slightly (proposed classification: 26 441; current classification: 26 482). Stage-specific OS differences were evident in both the international major SGC cohort (n = 1015) and Memorial Sloan Kettering Cancer Center minor SGC cohort (n = 444). Conclusions and Relevance:This unified, SGC-specific staging system improved prognostic accuracy and sample size balance and was applicable to both major and minor SGCs.
Abstract Background Generative artificial intelligence (AI) technologies might offer new possibilities for the peer review process; however, AI models' possible vulnerability to hidden nudges designed to elicit positive reviews raises concerns about manipulation susceptibility, which remains unexplored. We aimed to evaluate AI model susceptibility to hidden nudges in peer review. Methods This quasi-experimental study was conducted between July and December 2025. Four commercial AI models were evaluated simultaneously: GPT-4 (OpenAI), Gemini 2.5 Flash (Google), DeepSeek-V3 (DeepSeek), and Claude Opus 4 (Anthropic). We used 90 pre-print and 90 published manuscripts in critical care and cardiology to feed the AI models. All manuscripts were converted to individual Microsoft Word files, with identifying information removed, to mimic a manuscript submitted to a journal for peer review. Each manuscript underwent three independent evaluations per model using standardized prompts requesting evaluation and recommendation on whether to accept or reject it for publication. First, we evaluated the manuscript without any nudge. Second, we inserted a hidden nudge opposing the initial recommendation (e.g., a negative nudge if initially accepted). Finally, we evaluated the nudged manuscripts using a modified prompt warning about potential hidden nudges. All recommendations were categorized as accept or reject. The main outcomes were the change rates in recommendations after nudge insertion compared to initial recommendations, and after nudge insertion with the modified prompt, analyzed separately for each AI model. Results Across all AI models tested, nudge insertion led to a change in the recommendation in 84.4% of the time (608/720), with Deepseek being the most susceptible model (100% of change), followed by Gemini (97.8% of change), Chat GPT (82.8% of change) and Claude (57.2% of change). Using a specific prompt to warn AI models about potential malicious nudge injections in the manuscripts did not substantially alter the results. Recommendations were still modified in 76.8% of cases (553/720). Conclusions In this quasi-experimental study, all tested AI models were highly susceptible to hidden nudge insertions in manuscripts during simulated peer review. Importantly, explicitly warning AI models about potential nudge injections does not meaningfully reduce their susceptibility to manipulation.
Background:Several genes encoding proteins essential for normal bile production have been associated with progressive familial intrahepatic cholestasis in children, but there are few studies evaluating the frequency of variants in these genes among adults with chronic cholestatic disease. The aim of this study was to assess frequency of variants in progressive familial intrahepatic cholestasis-associated genes in adults with idiopathic or episodic cholestatic chronic liver disease (cCLD).Methods:Patients with cCLD of unknown cause followed in 4 different reference centers in Brazil were genotyped for ABCB11, ABCB4, ABCC2, ATP8B1, CFTR, JAG1, KIF12, LSR, MYO5B, NR1H4, PPM1F, SERPINA1, TJP2, USP53, VIPAS39, VPS33B, PEX26, and WDR83OS gene variants using Illumina platforms. Primary biliary cholangitis, primary sclerosing cholangitis, and other causes of cCLD were excluded.Results:Sixty-five patients (40 females; mean +/- SD age at disease onset, 26.8 +/- 13.4 y) were included. Most (65%) had either a family history of cholestatic liver disease or previous signs and symptoms of intrahepatic cholestasis of pregnancy or low phospholipid-associated cholelithiasis. Fifty-seven (88%) had cCLD, whereas 8 (12%) reported recurrent episodic cholestasis. Some had evidence at diagnosis of cirrhosis (n=20) or hepatocellular carcinoma (n=2) or had undergone liver transplantation (n=10). Sequencing revealed 28 variants in ABCB4 (n=25) and ABCB11 (n=3) genes in 42 patients [65%; heterozygous (n=31), homozygous (n=4), and compound heterozygous (n=7)]. Only 17 (61%) were previously reported. Most variants (57%) were classified as pathogenic or likely pathogenic.Conclusions:More than half of the patients with cCLD of unknown cause exhibited pathogenic or likely pathogenic variants in bile transporter genes, particularly ABCB4. Genotyping may identify most of those patients as late-onset patients with multidrug resistance protein 3 deficiency.
The KEYNOTE-522 regimen is the standard of care for stage II-III triple-negative breast cancer (TNBC). However, older patients were underrepresented in the pivotal trial. We evaluated the effectiveness and safety of this regimen in patients aged ≥65 years enrolled in the Neo-Real/GBECAM-0123 multicenter real-world study conducted across institutions in Brazil and Argentina. Among 724 patients, 80 (11%) were aged ≥65 years and presented distinct baseline characteristics, including lower frequencies of grade 3 tumors, Ki67 ≥ 50%, and germline BRCA1/2 mutations, alongside a higher prevalence of impaired performance status. The pathologic complete response (pCR) rate in older patients was 54.9% in comparison with 64.5% in younger patients, although age was not independently associated with pCR in multivariable analysis, including other relevant baseline variables. Older patients experienced a significantly higher toxicity burden, with increased rates of treatment discontinuation, dose reductions, treatment delays, hospitalizations, and grade ≥3 neutropenia. Taken together, these data indicate that older patients with TNBC harbor distinct biological and clinical features with numerically lower pCR rates, and that the increased toxicity burden underscores the need for personalized treatment strategies and dedicated research in this population.
High platelet reactivity (HPR) in patients with coronary artery disease receiving P2Y12 inhibitors increases ischemic risk. Chronic kidney disease (CKD) is an established contributor to HPR during clopidogrel therapy. The objective of the study was to assess whether CKD influences platelet reactivity (PR) in patients treated with clopidogrel or ticagrelor. This double-blind, double-dummy study enrolled 106 stable patients more than one year after an acute coronary syndrome, with or without CKD. Participants were matched by age and sex and randomized to clopidogrel or ticagrelor. PR was measured using the VerifyNow™ P2Y12 assay, and HPR was defined as P2Y12 reaction units (PRU) ≥ 208. Median glomerular filtration rates were 80 mL/min/1.73 m2 in non-CKD patients and 41 mL/min/1.73 m2 in CKD patients (p < 0.01). Ticagrelor produced similarly low PR in both groups (36 vs. 35 PRU; p = 0.61). Clopidogrel resulted in a numerically higher PR in CKD patients (209 vs. 180 PRU; p = 0.07). The magnitude of PR reduction with ticagrelor relative to clopidogrel was greater in CKD patients (p-interaction = 0.09). HPR was markedly more common with clopidogrel, particularly in CKD (difference 37%; adjusted OR 4.42; p = 0.01). In conclusion, CKD significantly impairs clopidogrel responsiveness but does not affect ticagrelor, resulting in a greater relative advantage of ticagrelor in patients with CKD.
