Leeds General Infirmary

Abstract Introduction Best medical therapy (BMT), including antithrombotic and lipid-lowering treatment, is the standard of care for patients undergoing major lower limb amputation (MLLA), aiming to reduce major adverse cardiac events (MACE) and mortality. We assessed whether patients undergoing MLLA are receiving appropriate BMT. Methods This single-centre study included patients undergoing MLLA with clinical or radiological evidence of atherosclerosis. Medication use at admission and discharge was recorded. Triglyceride and LDL-C levels were measured at baseline, 3 months, and 6 months postoperatively. Results A total of 116 patients (81.9% male) were included. Comorbidities: diabetes (67.2%), hypertension (60.3%), heart failure (17.2%), and chronic kidney disease (28.5%). High-intensity statin use (75.0 to 82.5%, P = 0.0209) and ezetimibe increased significantly (6.9 to 11.3%, P = 0.0253). No patients were prescribed inclisiran, evolocumab, or icosapent ethyl, despite 10 meeting potential eligibility criteria. Antithrombotic use improved: full dose DOACs increased from 29.0 to 42.0% and warfarin from 0.9 to 2.7% (P < 0.001), COMPASS regimen increased from 12.9 to 21.4% (P = 0.0067). No statistical difference in single dose antiplatelet use was observed; patients receiving no antithrombotic therapy declined significantly (32 to 14, P < 0.001). Median LDL-C levels were 1.4 mmol/L baseline and at 6 months (P = 0.5443). Triglycerides levels increased 1.4 baseline to 2.0 mmol/L at 6 months (P = 0.1876). Conclusions While BMT prescribing improved following MLLA, adjunct lipid-lowering therapies remain underutilised reflected in no overall change in biochemical parameters. Broader implementation of full BMT is necessary to optimise cardiovascular risk reduction.

Routine monitoring of surgical outcomes can improve service quality. Risk-adjusted monitoring tools for adults with congenital heart disease (CHD) in England and Wales are lacking. Using national audit data of all adult CHD surgical procedures in public hospitals from 2015 to 2022, we developed logistic regression models for mortality at 30 days and 90 days and a 30-day complications outcome. Risk factors included patient demographics and categorical derived variables for case complexity and procedure risk. Model performance was assessed by area under the receiver operating characteristic curve and calibration errors for in-sample and cross-validation data sets. Average 30-day and 90-day mortality were 1.4% (49/3502) and 1.7% (58/3493). Moderate and severe CHD complexity were strong predictors of 30-day mortality (odds ratio [95% CI], 3.5 [0.8-15.8], 8.6 [2.4-30.9]), as was high-risk procedure (OR, 3.6 [2.1-6.0]). Average 30-day complication rate was 7.5% (242/3223). Procedure risk groups (OR, 2.4 [0.9-6.0] to 12.2 [4.0-36.8]) and procedure complexity (OR, 2.5 [1.5-4.3]) were the strongest predictors. In cross-validation, 30-day and 90-day mortality models had median discrimination (interquartile range in parentheses) of 0.844 (0.84-0.85) and 0.866 (0.86-0.87), calibration slopes of 1.05 (0.60-1.13) and 1.11 (0.61-1.21), and calibration-in-the-large of 0.00 (-0.12 to 0.19) and -0.07 (-0.17 to 0.30). The 30-day complications model had cross-validation discrimination of 0.760 (0.76-0.76), calibration slope of 0.93 (0.74-1.18), and calibration-in-the-large of -0.07 (-0.13 to 0.22). The adult CHD risk models perform well for short-term mortality despite a low number of events. The risk model for 30-day complications showed reduced performance, suggesting that important risk factors are not captured by routinely collected data.

United Kingdom and European guidelines recommend that patients with atherosclerotic carotid artery disease should be commenced on optimal cardioprotective medical therapy, specifically antiplatelet agents and lipid lowering therapy (LLT). This single centre cohort study explored the peri-operative prescription and titration of guideline directed medical therapy (GDMT) alongside low density lipoprotein cholesterol (LDL-C) levels in patients undergoing carotid intervention. The study included 260 patients who underwent carotid intervention from 1 January 2017 to 31 December 2021. Electronic hospital records were checked for data on demographics, comorbidities, antiplatelet and LLT prescriptions, and cholesterol results. Data from four time points were obtained: prior to the index event; at the time of admission for carotid intervention; at discharge following intervention; and at one year. Antiplatelet therapy was indicated in one fifth and high intensity LLT in 39% of patients who were not on cardioprotective medications prior to the index event. From the time of admission for carotid intervention to discharge, the proportion of patients who were prescribed antithrombotic therapy (82% vs. 100%; p < .001) and LLT (88% vs. 94%; p = .004) statistically significantly increased. Mean LDL-C levels statistically significantly reduced from 2.2 mmol/L pre-admission to 1.8 mmol/L at one year follow up (p = .026). However, only 53% of patients achieved an LDL-C target of < 1.8 mmol/L one year following intervention. All patients who did not achieve the LDL-C target had scope for uptitration of statin or co-prescription of adjuvant LLT agents, including ezetimibe, inclisiran, and PCSK9i. Underutilisation of LLT prior to incident stroke was identified as a key tractable problem. Admission for carotid intervention is associated with an increase in GDMT and reduction in LDL-C levels. There is scope for improvement and a need for long term community based management of cardiovascular risk, specifically risk assessment, initiation and ongoing uptitration of LLT, and regular monitoring of LDL-C levels.

BackgroundImproved survivorship following myocardial infarction (MI) has resulted in transferred morbidity to other long-term conditions (LTCs). Understanding of disease accrual over time following MI has been limited by a lack of methodologies that consider real-world complexity. We characterised post-MI multimorbidity trajectories in a real-world population of individuals following MI.MethodsThis population-wide retrospective study comprised all individuals with MI in the Wales Multimorbidity e-Cohort (which included linked primary and secondary care data for 2,902,101 GP-registered residents of Wales; 2005–2019). Single-year post-MI disease clusters and multi-year latent multimorbidity trajectories were constructed from 227 chronic conditions and 62 acute conditions, using non-negative matrix factorisation (NMF). Multinomial logistic regression identified sociodemographic factors associated with single-year post-MI disease clusters. Time-updating flexible parametric survival models quantified the association between multi-year multimorbidity trajectories and long-term all-cause mortality, adjusting for age, sex, year of MI, socioeconomic deprivation and rurality.ResultsIn total, 70,529 individuals had an incident MI during the study period (median [interquartile range] age 72 [62–82] years; 40.6% female), with restricted mean post-MI survival of 8.1 years. At MI diagnosis, 67,023 (95.0%) had ≥ 2 LTCs (median 8 [interquartile range; IQR 5–12]), which increased to 99.8% at 1 year (50,633/50,737 surviving patients, median [IQR] 12 [7–17]). NMF classified individuals into one of 10 post-MI disease clusters, based on all observed acute and chronic diagnoses (n = 3,954,622) accrued over time. Individuals that followed the most adverse latent multimorbidity trajectory (n = 26,035, 36.9%) were characterised by recurrent MI, acute infections and renal disease and had an increased risk of all-cause mortality (adjusted hazard ratio 6.62; 95% CI: 6.09–7.20) compared with individuals in the least adverse trajectory.ConclusionsPatients with MI have a high pre-existing multimorbidity burden that increases post-MI. Using NMF, we were able to reduce the real-world complexity of all individual diseases accrued over time into 10 latent post-MI multimorbidity trajectories. These trajectories were characterised by specific patterns of acute and chronic conditions, with differential impact on outcomes. These trajectories may enable the implementation of targeted strategies for individuals that are at greatest risk of adverse outcomes.Supplementary InformationThe online version contains supplementary material available at 10.1186/s12916-025-04520-1.

Purpose To develop a self-supervised text-vision framework to detect abnormalities on brain MRI scans by leveraging free-text neuroradiology reports, eliminating the need for expert-labeled training datasets. Materials and Methods This retrospective and prospective multicenter study included 81 936 brain MRI examinations and corresponding radiology reports for adult patients at two UK National Health Service hospitals from January 2008 to December 2019 for training and internal testing and 1369 prospectively collected examinations between March 2022 and March 2024 from four separate National Health Service hospitals for external testing (ClinicalTrials.gov no. NCT04368481). A neuroradiology language model (NeuroBERT) was trained using self-supervised tasks to generate report embeddings. Convolutional neural networks (one per MRI sequence) were trained to map scans to embeddings by minimizing mean squared error loss. The framework then detected abnormalities in new examinations by scoring scans against query sentences using text-image similarity. Model diagnostic performance was assessed using the area under the receiver operating characteristic curve (AUC). Results The framework achieved an AUC of 0.95 (95% CI: 0.94, 0.97) for normal versus abnormal classification and generalized to external sites with examination-level AUCs of 0.90 (95% CI: 0.86, 0.93) in Bedford, 0.87 (95% CI: 0.83, 0.90) in Nottingham, 0.86 (95% CI: 0.83, 0.90) in Norwich, and 0.85 (95% CI: 0.81, 0.89) in Yeovil. In five zero-shot classification tasks-acute stroke, multiple sclerosis, intracranial hemorrhage, meningioma, and hydrocephalus-the framework achieved a mean AUC of 0.89 (range, 0.77-0.93). For visual-semantic image retrieval, mean precision was 0.84 among the top 15 images across seven pathologies. Conclusion The self-supervised text-vision framework accurately detected brain MRI abnormalities without expert-labeled datasets. Clinical trial registration no. NCT04368481 Keywords: Head and Neck, Unsupervised Learning, Convolutional Neural Network (CNN), Neuroradiology © The Author(s) 2025. Published by the Radiological Society of North America under a CC BY 4.0 license. Supplemental material is available for this article. See also commentary by Ghodasara in this issue.

Orthodontic training in the UK is evolving to address workforce shortages and meet increasing demand. A new curriculum, effective from September 2024 emphasizes evidence-based practice. Despite these advancements, challenges persist, including competitive recruitment, financial constraints and an uneven distribution of services. Possible areas for future development include the integration of primary care experience into training; and the revision of orthodontic commissioning in the NHS. Any measures introduced should aim to improve training pathways, ensure equitable service delivery, and maintain high standards of care. Addressing these issues is vital for developing a sustainable orthodontic workforce capable of meeting future healthcare needs. CPD/Clinical Relevance: This article reviews current developments in UK orthodontic training, assessment, workforce challenges, and NHS service provision, providing essential context for trainees, educators, and clinicians preparing for future practice.