In a large multicenter real-world cohort, we aimed to evaluate outcomes of FLAG-Ida salvage therapy for relapsed/refractory (R/R) acute myeloid leukemia (AML) and validated the SALFLAGE prognostic score. We analyzed 1079 adults with R/R AML treated across 112 PETHEMA institutions over 26 years (1998-2024), including patients with primary refractory disease (36.9%) and first relapse episode (63.1%), with a median age of 52 years. Complete remission composite (CRc) was achieved 56.8%, including complete remission (CR) in 51.0%, CR with incomplete recovery in 4.0%, and morphological-free-state in 1.8%, enabling 35.2% of patients and 62% of responders to proceed to allogeneic transplantation without morphological disease. With median follow-up of 50.9 months, median overall survival (OS) was 10.2 months, with 5-year OS rate of 21.6%. Prior allogeneic transplantation (HR 0.54; p < 0.001) and relapse-free interval ≥ 1 year (HR 0.75; p = 0.024) independently predicted improved OS, whereas modified high-risk cytogenetics including t(8; 21) (HR 3.58; p < 0.001), FLT3-ITD mutation at primary diagnosis (HR 1.61; p < 0.001), and age ≥ 60 (HR 1.43; p < 0.001) conferred inferior OS. Validation of the SALFLAGE score demonstrated moderate discrimination (C-index 0.67), with 5-year survival of 38.4%, 27.2%, and 12.7% across risk categories (p < 0.001). Outcomes improved over periods (1998-2005 vs. 2006-2016 vs. 2017-2024): 30-day mortality was 6.9% vs. 9.3% vs. 5.0%, respectively (p = 0.030), and median OS was 7.8 versus 9.4 versus 11.1 months, respectively (p = 0.16). We confirm FLAG-Ida as a reference salvage regimen in fit R/R AML and validate the SALFLAGE score in this setting.
Timely characterization of ischemic stroke is crucial for providing endovascular or pharmacological treatment. Current stroke care protocols include NCCT as first imaging support but exhibit low differentiation between healthy and affected tissue. Complementary, some protocols include MRI ADC/DWI finding observations, but this modality has relatively low availability. To generate synthetic ADC and DWI sequences from NCCT studies, maintaining textural characteristics of ischemic lesions. From a cohort of ischemic stroke patients, we introduced a multi-domain generative representation to approximate ADC/DWI sequences from NCCT onset studies. The proposed approach explores a cycle consistency loss over an architecture of stacked U-Net blocks that guarantee textural stroke lesion preservation. The proposed approach was validated in a retrospective study of 96 patients, recovering synthetic MRI sequences with an average SSIM/PSNR score of 0.81/26.19. Also, a specific validation around stroke lesion achieved an average SSIM/PSNR score of 0.51/28.8. The proposed approach outperformed a standard CycleGAN and a StarGAN architecture, evidencing significant statistical differences ( p<0.01 ). The proposed method is able to recover an approximation to MRI (ADC/DWI) sequences from NCCT scans, while potentially avoiding the current delay in clinical practice and providing a first approximation of lesion patterns observed from ADC and DWI studies.
Postoperative delirium is a common complication in older surgical patients. Intraoperative EEG burst suppression may predict its occurrence. This study aimed to evaluate the association between burst suppression and postoperative delirium exclusively in non-cardiac surgery through systematic review and meta-analysis. A Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA)-guided systematic review was conducted using PubMed, Cochrane Library, and ScienceDirect. Eligible studies included adult non-cardiac surgical patients assessing EEG burst suppression and postoperative delirium. Risk of bias was evaluated using RoB 2, the Newcastle-Ottawa Scale, and the Joanna Briggs Institute tools. Meta-analysis employed a random-effects Mantel-Haenszel model, with narrative synthesis for studies lacking extractable data. Across the included studies, the incidence of postoperative delirium (POD) ranged from 8.9% to 26%. Meta-analysis demonstrated that intraoperative burst suppression was significantly associated with POD occurrence (OR: 1.73, 95% CI: 1.03-2.90; I² = 38%). Patients with POD had significantly longer burst-suppression duration (mean difference of 25.31 min; p < 0.00001; I² = 0%). The pooled adjusted analysis confirmed an independent association between burst suppression and POD (OR: 2.69, 95% CI: 1.90-3.81; I² = 31%). Individual studies supported a dose-dependent relationship between the duration of suppression and the risk of delirium. Anesthetic agents influenced the suppression burden - propofol was associated with longer durations, whereas desflurane was associated with increased POD risk, independent of suppression. Preoperative cognitive impairment and frailty were identified as significant effect modifiers, increasing susceptibility to burst suppression and subsequent delirium. Intraoperative EEG burst suppression is significantly associated with postoperative delirium in non-cardiac surgery. Duration and patient vulnerability influence risk, highlighting the potential of duration as a modifiable intraoperative predictor.
El síndrome de Mowat-Wilson es un trastorno genético raro caracterizado por presentar manifestaciones clínicas que afectan diversos sistemas y órganos. En 2025 se reportó el primer caso de este síndrome en Santander, Colombia; se trata de un paciente masculino de 2 años y 9 meses, con antecedentes de hipospadias y criptorquidia. En su primer mes de vida, fue diagnosticado con enfermedad de Hirschsprung y retraso en el neurodesarrollo. A los 10 meses, se confirma el diagnóstico del síndrome a través de un estudio de exoma trío que identifica una variante patogénica de novo en el gen ZEB2. Posteriormente, a los 21 meses, fue diagnosticado con epilepsia de difícil control. El manejo clínico incluyó diferentes intervenciones quirúrgicas, tratamiento de complicaciones y seguimiento especializado por los servicios de endocrinología y neurología, entre otros. La identificación temprana de las características clínicas puede mejorar significativamente el pronóstico y la calidad de vida del paciente.
Ischemic stroke is one of the leading causes of death and disability worldwide. Lesion segmentation from diffusion-weighted MRI (DW-MRI) is crucial for guiding clinical interventions and estimating patient prognosis. This process relies on expert diffusion-weighted imaging (DWI) interpretation to identify potential lesion tissue, followed by confirmation using apparent diffusion coefficient (ADC) maps. Computational stroke segmentation methods combine DWI and ADC but rely on implicitly learned multimodal relationships, neglecting foundational information provided by hyperintense DWI signals. In this work, we introduce a dual-decoder architecture to segment DWI hyperintensities alongside stroke lesions, providing structured guidance through a multisegmentation loss. This setup regularizes feature learning by incorporating preliminary localization and lesion shape information, improving segmentation accuracy and enabling modality-specific interpretability. Our proposed approach can generate both expert- and threshold-confirmed lesion masks, providing an intrinsic quality control mechanism that identifies discordant predictions and explicitly visualizes sequential clinical reasoning. Trained and evaluated on 180 DW-MRI studies annotated by three expert neuroradiologists, the proposed approach showed significant Dice improvements of 2.0% on expert-confirmed masks and 4.5% on threshold-confirmed masks over conventional approaches. These gains were especially notable in challenging cases with small and subtle lesions, highlighting the clinical and methodological value of auxiliary hyperintensity learning.
