143174 research outputs found
Sort by
Investigating estimand considerations in adaptive trials: a systematic review
No full textBackground:
Randomised controlled trials (RCTs) are the gold standard for evaluating treatment effects, with the results informing policy and clinical practice. To ensure appropriate methods are utilised and to avoid misinterpretation of the results of a clinical trial, it is vital that we understand the research question a trial aims to answer. However, there is often ambiguity in how trialists define their research questions. In 2019, an addendum to the international trial regulatory guidelines (ICH E9 (R1)) introduced the estimand framework to combat this. A review of protocols published in 2020 investigated the early adoption of the estimand framework and found no uptake as well as a lack of clarity on key items such as the handling of intercurrent events. The aim of this review was to identify the current application of the estimand framework specifically to trials with an adaptive design.
Methods:
The search strategy aimed to identify trial protocols and statistical analysis plans that described RCTs published in two journals (BMJ Open and Trials) in 2023. Articles were eligible if they related to phase 2–4 trials with an adaptive design. A pre-piloted data extract form was used to extract data relating to study details, intercurrent events and estimands.
Results:
One thousand five hundred and forty-one articles were identified by the initial search. Following screening, 146 articles were identified as meeting the eligibility criteria. Of the eligible articles, five (3%) stated their primary estimand, and of these, three (2%) stated all five estimand attributes. Ninety-four (64%) articles described one or more intercurrent events; these included a total of two hundred and thirty-two intercurrent events described. Fifty-two (36%) articles did not describe any intercurrent events. No articles specified the estimand for any planned interim analyses or considered the implications of adaptations on the primary estimand.
Conclusions:
This review provides evidence that there is still a lack of uptake of the estimand framework in RCTs. Wider application of the estimand framework would ensure clarity in the reporting and interpretation of clinical trial results. In addition, clear guidance on how to implement the estimand framework to trials with an adaptive design is needed
Ultra-wide-field, deep, adaptive two-photon microscopy for multiscale neuronal imaging
No full textObserving the activity patterns of large neural populations throughout the brain is essential for understanding brain function. However, capturing neural interactions across widely distributed brain regions from both superficial and deep cortical layers remains challenging with existing microscopy technologies. Here, we introduce a state-
of-the-art two-photon microscopy system, ULTRA, capable of single-cell resolution imaging across an ultra-large field of view (FOV) exceeding 50 mm², enabling deep and very wide field in vivo imaging. To demonstrate its capabilities, we conducted a series of experiments under multiple imaging conditions, successfully visualizing brain
structures and neuronal activities spanning a spatial range of over 7 mm from superficial layers to depths of up to 900 μm, while covering a volume of 45.24 mm3 in the mouse brain. This versatile imaging platform overcomes traditional spatial constraints, providing a powerful tool for comprehensive exploration of neuronal circuitry over extensive spatial scales with cellular resolution
Photogrammetry measurements of blunt body dynamics in a supersonic wind tunnel
No full textAbstract
This paper presents free-oscillation experiments of a blunt body conducted in a high-speed wind tunnel, with the model motion measured using photogrammetry. A faceted blunt model, mounted on a spherical air bearing, is free to rotate in
roll, pitch, and yaw in response to the freestream flow (M= 2). Four synchronised high-speed cameras capture the model from multiple angles, and the unique coded targets printed on the model’s surface are reconstructed as points in 3D space, achieving accuracy within 1◦
for both static and dynamic measurements. The Kabsch algorithm is used to find the optimal rotation between two point clouds, hence allowing reconstruction of the angular motion over the entire run. The method
shows promise for free-oscillation tests in high-speed ground facilities, offering advantages over ballistic range and free flight tests such as a constant freestream velocity and hundreds of oscillation cycles. This capability enables the observation of dynamic instabilities that develop over extended timescales, thus revealing a precessional instability previously reported
only for slender bodies at hypersonic Mach numbers
The value of a peer-to-peer teaching community in medical education
No full textIntroduction
The Medical Education Society (MedED) at Imperial College School of Medicine (ICSM) offers near-peer educational opportunities across all years of medical school. Near-peer education has demonstrated significant benefits in medical education. However, studies have yet to explore the value of establishing a peer-to-peer teaching community.
Methods
Medical students who participated in MedED as student-attendees or student-teachers during the academic year 2022-23 were invited to participate in a survey and follow-up interview, exploring their experiences within the Society. Survey data was collected anonymously through Qualtrics, and interviews were held on Microsoft Teams. Quantitative survey data was analysed using descriptive statistics, while interview transcripts and free-text survey responses underwent inductive thematic analysis.
Results
A total of 66 students completed the survey, with 19 (28.8%) from years 1-2 and 47 (71.2%) from years 3-6. Early-year students had higher lecture attendance rates (79%) compared to later-year students (34%), and both groups preferred online rather than in-person lectures (both >50%). For student-attendees, benefits of participating in MedED included improving knowledge, motivation and sense of community. Among student-teachers, main motivations for teaching included helping others and developing transferrable skills.
A total of 13 participants were interviewed, including 5 who were both student-attendees and student-teachers. Three themes emerged: academic value, highlighting knowledge and skills gained through MedED; career prospects, focusing on long-term professional benefits; and sense of community and wellbeing, emphasising the positive social interactions and support networks fostered through MedED.
Conclusion
MedED provides student-led teaching initiatives that supplement the formal curriculum, enhancing student confidence and inclusivity, and fostering a sustainable community of peer-education. Beyond immediate academic values, this community has also created longer-term, personal and professional impacts on students, including broadening career aspirations. This work highlights opportunities for further development through student-staff collaborations and the role of peer communities in supporting student wellbeing
A standardized definition of Rapid Evidence Assessment for environmental applications
No full textPreprint versio
Impact of the COVID-19 pandemic on cancer screening in Europe: a systematic review of disruptions, barriers, and policy responses
No full textBackground
Cancer screening is a cornerstone of cancer control, but the COVID-19 pandemic caused unprecedented disruption to preventive healthcare worldwide. In Europe, national screening programmes were severely affected, with consequences extending beyond screening to diagnosis, treatment, and equity. While several country-specific studies exist, cross-regional syntheses remain scarce. Understanding the scale, determinants, and outcomes of these disruptions is crucial to building resilient, equiTable screening systems.
Aim/objective
This systematic review synthesises evidence on the impact of the COVID-19 pandemic on cancer screening across Europe, examining differences by cancer type, screening modality, and national context. It also explores downstream effects, barriers, enablers, and policy responses to guide future preparedness.
Methods
Following PRISMA guidelines, six databases and grey literature sources were searched for studies published between December 2019 and January 2025. Eligible studies included quantitative and qualitative analyses of screening activity during the pandemic. Data were extracted on study characteristics, outcomes, and contextual factors. Given the heterogeneity of measures, findings were summarised using descriptive statistics and thematic synthesis.
Result
Forty-five studies from 18 European countries revealed a 30-60% reduction in screening participation at peak disruption, varying by cancer type and country. Consequences included delayed diagnoses, stage migration, increased projected mortality, and widening inequalities. Major barriers included service suspension, staff redeployment, and fear of infection. Enablers comprised adaptive communication, safety protocols, and digital innovations.
Conclusion
COVID-19 caused substantial and uneven disruption to European cancer screening. Protecting continuity, institutionalising innovations, and addressing inequities are critical to enhancing resilience for future health crises
Artificial intelligence in virtual fracture clinics: a systematic review of imaging and clinical-text tools
No full textBackground
Virtual fracture clinics (VFCs) are a well-established component of acute orthopedic management pathways.
Artificial intelligence (AI) healthcare tools are increasingly sophisticated and have the potential to disrupt current practices. The aim of this review was to determine the opportunities, performance and readiness of AI systems that integrate clinical-text and imaging data for the triage or management of patients in VFCs.
Methods
A search of MEDLINE and Embase was performed between January 2010 and July 2025. The review included primary research studies investigating AI for fracture detection via X-rays and natural language processing (NLP) models for clinical documentation. A random-effects meta-analysis was conducted to calculate pooled sensitivity and specificity, stratified by anatomical region and developer type (commercial vs. researcher-developed).
Results
We included 54 studies: 52 imaging/X-ray studies and 2 NLP/clinical-text studies. Among the imaging studies,
13 evaluated commercial tools, and 39 evaluated researcher-developed models. There were 2 NLP models, both of which interpreted radiology reports rather than text summaries of clinical assessments. No studies that included the use of NLP models in acute orthopedic care could be found. A meta-analysis of commercial tools (n=11) demonstrated a pooled sensitivity across both multiregional "Limb" tools of 92.58% (95% CI: 90.61–94.17%) and anatomy-specific "Wrist" tools of 89.95% (95% CI: 72.18–96.86%). Wrist-specific commercial tools demonstrated higher specificity (96.08%; 95% CI: 90.12–99.01%) compared to general limb tools (89.69%; 95% CI: 84.02–93.51%), suggesting that anatomical targeting may reduce the number of false positives. Researcher-developed models (n=32) often reported superior metrics for sensitivity (e.g., Limbs 95.11%; 95% CI: 91.83–
97.11%) compared to commercial tools sensitivity.
Conclusions
VFCs require the integration of information from imaging and patient records. Multiple image interpretation
tools are available with high performance in fracture identification. The development and integration of NLP
tools to interpret clinical documentation from emergency departments and urgent care centers are necessary
for AI-VFC
Anticoagulation, Recurrent Thrombosis and Major bleeding in antiphospholipid syndrome: UK multicentre observational study
No full textAntiphospholipid syndrome (APS) is an acquired autoimmune disease characterised by presence of thrombosis and/or pregnancy morbidity with persistently positive antiphospholipid antibodies(aPL). Due to the heterogeneity and relative rarity of this disorder clinical practice in the management of APS remains varied. In this retrospective UK-wide multicentre study, we aimed to delineate clinical practice and outcomes in thrombotic APS to improve the areas of limited knowledge, particularly anticoagulant practice, thrombotic recurrence and bleeding. Anticoagulation in Antiphospholipid Syndrome (A2 PLS) is the largest multicentre observational study to date spanning 20 national health service Trusts in the UK and including 500 adult patients (≥18years) with thrombotic APS, on or off anticoagulation during the years 2012-2021. Thrombotic APS is primarily treated with vitamin K antagonists; however, the rates of recurrent thrombosis remain high. In the last decade, recurrent thrombosis occurred in 34.4% (43/125), 32.6% (31/95) and 39.8% (37/93) of single, dual and triple positive aPL patients respectively with a recurrence rate of 46%. There was no difference in the probability of recurrent thrombosis based on the number of positive aPL tests (p=0.82) especially in the first three years. However, the probability of recurrent thrombosis was significantly higher in patients with lupus anticoagulant (p<0.01) compared to presence of other antibodies. There was a higher probability of recurrence in patients with arterial than venous thrombosis (p=0.03). Overall, 10-year bleeding rate was 22.0% with 6.7% patients having major bleeding. Identifying APS patients at higher risk of
recurrent thrombosis remains a challenge and current risk stratification is not adequate
How forced intervention facilitates AI adoption
No full textProblem definition: Whereas artificial intelligence (AI) technologies are increasingly becoming powerful and useful in operations, human workers often resist adopting algorithms, known as algorithm aversion. This aversion can undermine the algorithm performance in practice. Whereas numerous studies explore short-term mitigation strategies for such aversion, this paper investigates whether and why forced interventions can promote AI adoption and reduce algorithm aversion in practice. Methodology/results: Data from a leading online education company reveal that sales workers underutilize a new matching algorithm and often selectively use it on low-quality leads. The company conducted a field experiment in which sales workers were forced to use or not use the algorithm for three weeks. Experimental results show that forcing workers to use the algorithm during the experiment causally increases their algorithm usage over the month after the experiment by 15.8 percentage points. We develop a theoretical model to derive empirical strategies for exploring the mechanisms behind this improvement. Contrary to the traditional literature focusing on habit formation, our findings suggest learning is a key driver for algorithm adoption among workers over the month after the experiment. Specifically, forced algorithm use allows workers to experience the unbiased algorithm performance and positively adjust their beliefs about it. Consequently, after the experiment, workers use the algorithm not only more frequently but also more on high-quality leads. Managerial implications: The study empirically shows that forced intervention can effectively improve persistent algorithm use after the intervention, which is crucial for continuous development of the algorithm. More importantly, forced intervention breaks the vicious cycle of biased beliefs and selective usage by enabling workers to form unbiased evaluation of the algorithm efficacy and mitigate selective adoption on low-quality cases. This suggests that firms can implement extrinsic interventions or educational programs to help workers recognize the benefits of algorithms and develop unbiased beliefs about their capabilities, thus facilitating sustained algorithm usage.
Supplemental Material: The online appendix is available at https://doi.org/10.1287/msom.2024.1137
Jacobian Scopes: token-level causal attributions in LLMs
No full textPreprint versionLarge language models (LLMs) make nexttoken predictions based on clues present in their context, such as semantic descriptions and incontext examples. Yet, elucidating which prior tokens most strongly influence a given prediction remains challenging due to the proliferation of layers and attention heads in modern architectures. We propose Jacobian Scopes, a suite of gradient-based, token-level causal attribution methods for interpreting LLM predictions. Grounded in perturbation theory and information geometry, Jacobian Scopes quantify how input tokens influence various aspects of a model's prediction, such as specific logits, the full predictive distribution, and model uncertainty (effective temperature). Through case studies spanning instruction understanding, translation, and in-context learning (ICL), we demonstrate how Jacobian Scopes reveal implicit political biases, uncover word-and phrase-level translation strategies, and shed light on recently debated mechanisms underlying in-context time-series forecasting. To facilitate exploration of Jacobian Scopes on custom text, we open-source our implementations and provide a cloud-hosted interactive demo at https://huggingface.co/spaces/ Typony/JacobianScopes