Support Treatment Decisions Research Articles

Leveraging Explainable Artificial Intelligence (XAI) for Expert Interpretability in Predicting Rapid Kidney Enlargement Risks in Autosomal Dominant Polycystic Kidney Disease (ADPKD)

Autosomal dominant polycystic kidney disease (ADPKD) is the predominant hereditary factor leading to end-stage renal disease (ESRD) worldwide, affecting individuals across all races with a prevalence of 1 in 400 to 1 in 1000. The disease presents significant challenges in management, particularly with limited options for slowing cyst progression, as well as the use of tolvaptan being restricted to high-risk patients due to potential liver injury. However, determining high-risk status typically requires magnetic resonance imaging (MRI) to calculate total kidney volume (TKV), a time-consuming process demanding specialized expertise. Motivated by these challenges, this study proposes alternative methods for high-risk categorization that do not rely on TKV data. Utilizing historical patient data, we aim to predict rapid kidney enlargement in ADPKD patients to support clinical decision-making. We applied seven machine learning algorithms—Random Forest, Logistic Regression, Support Vector Machine (SVM), Light Gradient Boosting Machine (LightGBM), Gradient Boosting Tree, XGBoost, and Deep Neural Network (DNN)—to data from the Polycystic Kidney Disease Outcomes Consortium (PKDOC) database. The XGBoost model, combined with the Synthetic Minority Oversampling Technique (SMOTE), yielded the best performance. We also leveraged explainable artificial intelligence (XAI) techniques, specifically Local Interpretable Model-Agnostic Explanations (LIME) and Shapley Additive Explanations (SHAP), to visualize and clarify the model’s predictions. Furthermore, we generated text summaries to enhance interpretability. To evaluate the effectiveness of our approach, we proposed new metrics to assess explainability and conducted a survey with 27 doctors to compare models with and without XAI techniques. The results indicated that incorporating XAI and textual summaries significantly improved expert explainability and increased confidence in the model’s ability to support treatment decisions for ADPKD patients.

Wild-type MIC distributions and epidemiological cutoff values for 5-flucytosine and Candida species as determined by EUCAST broth microdilution.

EUCAST has established clinical breakpoints and epidemiological cutoff values (ECOFFs) for Candida spp. However, limited data are available for 5-flucytosine (5-FC). We assessed the in vitro susceptibility of 5-FC against a large collection of clinical Candida species using EUCAST methodology and determined the associated ECOFFs. A total of 5622 Candida isolates were collected from patients across the Netherlands between 2008 and 2024. 5-FC MICs were determined using the EUCAST microbroth dilution reference method. Furthermore, MICs were extracted from the EUCAST website. The MICs from this study and those extracted were used to determine ECOFFs and local ECOFFs (L-ECOFFs). 5-FC exhibited potent in vitro activity against C. albicans, N. glabratus and C. parapsilosis, while decreased susceptibility was observed for C. tropicalis, Pichia species, K. marxianus, Y. lipolytica, and C. auris. The ECOFFs (mg/L) and the percentages of WT isolates for 5-FC were: C. albicans: 0.5 (97.2%), N. glabratus: 0.5 (96.6%), C. parapsilosis: 0.5 (99.5%) and P. kudriavzevii: 8 (99.4%). The L-ECOFF (mg/L) and the percentages of WT isolates for 5-FC were: C. dubliniensis: 0.25 (96.8%), C. tropicalis: 0.25 (67.2%), K. marxianus: 0.25 (48.0%), C. lusitaniae: 0.25 (86.5%), M. guillermondii: 0.125 (95.9%) and P. norvegiensis: 8 (94.2%). 5-FC remains a valuable drug to manage difficult-to-treat invasive Candida infections. In vitro susceptibility cannot be predicted based on species identification for most Candida species, but requires MIC-testing. ECOFFs will help to interpret the MICs to support treatment decisions.

Diabetes prediction model based on GA-XGBoost and stacking ensemble algorithm.

Diabetes, as an incurable lifelong chronic disease, has profound and far-reaching effects on patients. Given this, early intervention is particularly crucial, as it can not only significantly improve the prognosis of patients but also provide valuable reference information for clinical treatment. This study selected the BRFSS (Behavioral Risk Factor Surveillance System) dataset, which is publicly available on the Kaggle platform, as the research object, aiming to provide a scientific basis for the early diagnosis and treatment of diabetes through advanced machine learning techniques. Firstly, the dataset was balanced using various sampling methods; secondly, a Stacking model based on GA-XGBoost (XGBoost model optimized by genetic algorithm) was constructed for the risk prediction of diabetes; finally, the interpretability of the model was deeply analyzed using Shapley values. The results show: (1) Random oversampling, ADASYN, SMOTE, and SMOTEENN were used for data balance processing, among which SMOTEENN showed better efficiency and effect in dealing with data imbalance. (2) The GA-XGBoost model optimized the hyperparameters of the XGBoost model through a genetic algorithm to improve the model's predictive accuracy. Combined with the better-performing LightGBM model and random forest model, a two-layer Stacking model was constructed. This model not only outperforms single machine learning models in predictive effect but also provides a new idea and method in the field of model integration. (3) Shapley value analysis identified features that have a significant impact on the prediction of diabetes, such as age and body mass index. This analysis not only enhances the transparency of the model but also provides more precise treatment decision support for doctors and patients. In summary, this study has not only improved the accuracy of predicting the risk of diabetes by adopting advanced machine learning techniques and model integration strategies but also provided a powerful tool for the early diagnosis and personalized treatment of diabetes.

Molecular subtypes and nomogram for predicting the prognosis of cervical cancer based on a matrix-immune signature

Cervical cancer is a kind of tumor related to chronic HPV infection. Currently, the treatment of cervical cancer is guided mainly by clinicopathological factors. The role of tumor microenvironment in the prognosis and treatment of cervical cancer has been ignored. We aimed to use bioinformatics to identify the molecular subtypes in cervical cancer and construct a predictive nomogram combining a matrix-immune signature (MIS) and clinicopathological factors to support treatment decisions. Two cervical cancer subtypes with different prognoses were identified based on matrix- and immune-genes in TCGA-CESC. The MIS was developed using Cox regression and Lasso algorithm and verified in the Cancer Genome Characterization Initiative (CGCI) using time-dependent receiver operating characteristic (ROC) curve analysis. Multivariable analysis identified lymph node metastases, lymphovascular space invasion, and the MIS as independent prognostic factors, which were used to construct the predictive nomogram. The areas under the ROC curve of the model were 0.872, 0.879, and 0.803 for the 1-, 3-, and 5-year periods, respectively. The C-index was 0.845. Calibration curves confirmed the excellent prognosis prediction of the nomogram. The nomogram indicted a 3-year survival rate of > 90% in patients with a total score > 110.1. The constructed predictive nomogram has significant implications for prognostic assessment and treatment selection in cervical cancer.

229TiP A phase II trial of a neural network-based treatment decision support tool in patients (pts) with refractory solid organ malignancies

229TiP A phase II trial of a neural network-based treatment decision support tool in patients (pts) with refractory solid organ malignancies

Radiomics on slice-reduced versus full-chest computed tomography for diagnosis and staging of interstitial lung disease in systemic sclerosis: A comparative analysis

PurposeThe purpose of this study was to evaluate the efficacy of radiomics derived from slice-reduced CT (srCT) scans versus full-chest CT (fcCT) for diagnosing and staging of interstitial lung disease (ILD) in systemic sclerosis (SSc), considering the potential to reduce radiation exposure. Material and methodsThe fcCT corresponded to a standard high-resolution full-chest CT whereas the srCT consisted of nine axial slices. 1451 radiomic features in two dimensions from srCT and 1375 features in three dimensions from fcCT scans were extracted from 166 SSc patients. The study included first- and second-order features from original and wavelet-transformed images. We assessed the predictive performance of quantitative CT (qCT)-based logistic regression (LR) models relying on preselected features and machine learning workflows involving LR and extra-trees classifiers with data-driven feature selection. The area under the receiver operating characteristic curve (AUC) was used to estimate model performance. ResultsThe best models for diagnosis and staging ILD achieved AUC=0.85±0.08 and AUC=0.82±0.08 with srCT, and AUC=0.83±0.06 and AUC=0.76±0.08 with fcCT, respectively. srCT-based models showed slightly superior performance over fcCT-based models, particularly in 2D-radiomic analyses when interpolation resolution closely matched the original in-plane resolution. For diagnosis, the LR outperformed qCT-models, whereas for staging, the best results were obtained with a qCT-based model. ConclusionsRadiomics from srCT is an effective and preferable alternative to fcCT for diagnosing and staging SSc-ILD. This approach not only enhances predictive accuracy but also minimizes radiation exposure risks, offering a promising avenue for improved treatment decision support in SSc-ILD management.

Vignette-based comparative analysis of ChatGPT and specialist treatment decisions for rheumatic patients: results of the Rheum2Guide study

BackgroundThe complex nature of rheumatic diseases poses considerable challenges for clinicians when developing individualized treatment plans. Large language models (LLMs) such as ChatGPT could enable treatment decision support.ObjectiveTo compare treatment plans generated by ChatGPT-3.5 and GPT-4 to those of a clinical rheumatology board (RB).Design/methodsFictional patient vignettes were created and GPT-3.5, GPT-4, and the RB were queried to provide respective first- and second-line treatment plans with underlying justifications. Four rheumatologists from different centers, blinded to the origin of treatment plans, selected the overall preferred treatment concept and assessed treatment plans’ safety, EULAR guideline adherence, medical adequacy, overall quality, justification of the treatment plans and their completeness as well as patient vignette difficulty using a 5-point Likert scale.Results20 fictional vignettes covering various rheumatic diseases and varying difficulty levels were assembled and a total of 160 ratings were assessed. In 68.8% (110/160) of cases, raters preferred the RB’s treatment plans over those generated by GPT-4 (16.3%; 26/160) and GPT-3.5 (15.0%; 24/160). GPT-4’s plans were chosen more frequently for first-line treatments compared to GPT-3.5. No significant safety differences were observed between RB and GPT-4’s first-line treatment plans. Rheumatologists’ plans received significantly higher ratings in guideline adherence, medical appropriateness, completeness and overall quality. Ratings did not correlate with the vignette difficulty. LLM-generated plans were notably longer and more detailed.ConclusionGPT-4 and GPT-3.5 generated safe, high-quality treatment plans for rheumatic diseases, demonstrating promise in clinical decision support. Future research should investigate detailed standardized prompts and the impact of LLM usage on clinical decisions.

Understanding how a personalized risk prediction tool (VALUE-PERSARC) supports informed treatment decisions of soft-tissue sarcomas patients in daily clinical practice – A mixed methods study

IntroductionRisk prediction models (RPM) can help soft-tissue sarcoma(STS) patients and clinicians make informed treatment decisions by providing them with estimates of (disease-free) survival for different treatment options. However, it is unknown how RPMs are used in the clinical encounter to support decision-making. This study aimed to understand how a PERsonalised SARcoma Care (PERSARC) RPM is used to support treatment decisions and which barriers and facilitators influence its use in daily clinical practice. MethodsA convergent mixed-methods design is used to understand how PERSARC is integrated in the clinical encounter in three Dutch sarcoma centers. Data were collected using qualitative interviews with STS patients (n = 15) and clinicians (n = 8), quantitative surveys (n = 50) and audiotaped consultations (n = 30). Qualitative data were analyzed using thematic analysis and integrated with quantitative data through merging guided by the SEIPS model. ResultsPERSARC was generally used to support clinicians’ proposed treatment plan and not to help patients weigh available treatment options. Use of PERSARC in decision-making was hampered by clinician’s doubts about whether there were multiple viable treatment options,the accuracy of risk estimates, and time constraints. On the other hand, use of PERSARC facilitated clinicians to estimate and communicate the expected benefit of adjuvant therapy to patients. ConclusionPERSARC was not used to support informed treatment decision-making in STS patients. Integrating RPMs into clinical consultations requires acknowledgement of their benefits in facilitating clinicians' estimation of the expected benefit of adjuvant therapies and information provision to patients, while also considering concerns regarding RPM quality and treatment options' viability.

Best-Practice Biomarker Testing of Oesophago-Gastric Cancer in the UK: Expert Consensus Recommendations Developed Using a Modified Delphi

AimsOesophago-gastric cancers (OGCs) are amongst the most commonly diagnosed malignancies worldwide and are associated with high disease-related mortality. Predictive biomarkers are molecules that can be objectively measured and used to indicate a likely response to therapeutic intervention, thus facilitating individualised cancer therapy. However, there remains variation in uptake and implementation of biomarker testing across the UK. Materials and methodsWe conducted a modified Delphi study to formulate consensus recommendations for best-practice biomarker testing of OGC in the UK. We employed two rounds of online questionnaires followed by a virtual consensus meeting. Biomarkers for discussion included HER2, MSI/MMR, and PD-L1. Topics comprised the overall biomarker pathway, pre-analytical, analytical, and post-analytical considerations, including challenges in current practice. ResultsTwenty-six and eighteen participants completed the first and second round Delphi questionnaire, respectively, with an even split of pathologists and oncologists from across the UK. There was consensus (>80% agreement) across several topics, including the requirements for standardisation of the pathway, which must include coordination throughout the tissue journey, requirements for a quality-assured process to ensure accuracy and validity of testing, plus the need for clear, detailed information on the pathology report to support treatment decisions. There was consensus amongst oncologists regarding reflex testing of all biomarkers depending on histology; however, concerns over capacity in relation to workload and availability of pathologists were evident among the pathologists. Overall, participants were in the opinion that reflex testing improves the speed of treatment decisions and improves patient care. ConclusionThe recommendations reflect best-practices and should be implemented to support rapid multidisciplinary team decision-making within oesophago-gastric cancer. Results reflect the need for standardisation and demonstrate the challenges faced in clinical practice by those requesting and testing biomarkers for oesophago-gastric cancer, suggesting significant concerns relating to pathologist capacity.

Causal Effect Estimation on Imaging and Clinical Data for Treatment Decision Support of Aneurysmal Subarachnoid Hemorrhage.

Aneurysmal subarachnoid hemorrhage is a medical emergency of brain that has high mortality and poor prognosis. Causal effect estimation of treatment strategies on patient outcomes is crucial for aneurysmal subarachnoid hemorrhage treatment decision-making. However, most existing studies on treatment decision-making support of this disease are unable to simultaneously compare the potential outcomes of different treatments for a patient. Furthermore, these studies fail to harmoniously integrate the imaging data with non-imaging clinical data, both of which are useful in clinical scenarios. In this paper, we estimate the causal effect of various treatments on patients with aneurysmal subarachnoid hemorrhage by integrating plain CT with non-imaging clinical data, which is represented using structured tabular data. Specifically, we first propose a novel scheme that uses multi-modality confounders distillation architecture to predict the treatment outcome and treatment assignment simultaneously. With these distilled confounder features, we design an imaging and non-imaging interaction representation learning strategy to use the complementary information extracted from different modalities to balance the feature distribution of different treatment groups. We have conducted extensive experiments using a clinical dataset of 656 subarachnoid hemorrhage cases, which was collected from the Hospital Authority Data Collaboration Laboratory in Hong Kong. Our method shows consistent improvements on the evaluation metrics of treatment effect estimation, achieving state-of-the-art results over strong competitors. Code is released at https://github.com/med-air/TOP-aSAH.

Determination of endometrial cancer molecular subtypes using a whole exome-sequencing based single-method approach

AimEndometrial cancer (EC) is heterogeneous with respect to epidemiology, clinical course, histopathology and tumor biology. Recently, The Cancer Genome Atlas (TCGA) network has identified four molecular subtypes with distinct clinical courses by an integrated multi-omics approach. These subtypes are of critical importance in the clinical management of EC. However, determination of TCGA molecular subtypes requires a complex methodological approach that is resource intensive and difficult to implement in diagnostic routine procedures. In this context, Talhouk et al. reported the precise determination of modified subtypes based on molecular surrogates obtained by a two-method approach comprising immunohistochemistry and DNA-sequence analysis (Proactive Molecular Risk Classifier for Endometrial Cancer; ProMisE). In this study, we aimed to identify EC molecular subtypes in analogy to TCGA and ProMisE applying an innovative whole exome-sequencing (WES) based single-method approach.MethodsWES was performed in a cohort comprising N = 114 EC patients. WES data were analyzed using the oncology treatment decision support software MH Guide (Molecular Health, Heidelberg, Germany) and EC molecular subtypes in analogy to TCGA and ProMisE were determined. Results from both classifications were compared regarding their prognostic values using overall survival and progression-free survival analyses.ResultsApplying a single-method WES-approach, EC molecular subtypes analogue to TCGA and ProMisE were identified in the study cohort. The surrogate marker-analogue classification precisely identified high-risk and low-risk EC, whereas the TCGA-analogue classification failed to obtain significant prognostic values in this regard.ConclusionOur data demonstrate that determination of EC molecular subtypes analogue to TCGA and ProMisE is feasible by using a single-method WES approach. Within our EC cohort, prognostic implications were only reliably provided by applying the surrogate marker-analogue approach. Designation of molecular subtypes in EC will be increasingly important in routine clinical practice. Thus, the single-method WES approach provides an important simple tool to tailor therapeutic decisions in EC.

SECRET: Statistical Emulation for Computational Reverse Engineering and Translation with applications in healthcare

There have been impressive advances in the physical and mathematical modelling of complex physiological systems in the last few decades, with the potential to revolutionise personalised healthcare with patient-specific evidence-based diagnosis, risk assessment and treatment decision support using digital twins. However, practical progress and genuine clinical impact hinge on successful model calibration, parameter estimation and uncertainty quantification, which calls for novel innovative adaptions and methodological extensions of contemporary state-of-the-art inference techniques from Statistics and Machine Learning. In the present study, we focus on two computational fluid-dynamics (CFD) models of the blood systemic and pulmonary circulation. We discuss state-of-the-art emulation techniques based on deep learning and Gaussian processes, which are coupled with established inference techniques based on greedy optimisation, simulated annealing, Markov Chain Monte Carlo, History Matching and rejection sampling for computationally fast inference of unknown parameters of the CFD models from blood flow and pressure data. The inference task was set as a competitive challenge which the participants had to conduct within a limited time frame representative of clinical requirements. The performance of the methods was assessed independently and objectively by the challenge organisers, based on a ground truth that was unknown to the method developers. Our results indicate that for the systemic challenge, in which an idealised case of noise-free data was considered, the relative deviation from the ground-truth in parameter space ranges from 10−5% (highest-performing method) to 3% (lowest-performing method). For the pulmonary challenge, for which noisy data was generated, the performance ranges from 0.9% to 7% deviation for the parameter posterior mean, and from 35% to 570% deviation for the parameter posterior variance.

Machine learning-based prediction of clinical outcomes after traumatic brain injury: Hidden information of early physiological time series.

To assess the predictive value of early-stage physiological time-series (PTS) data and non-interrogative electronic health record (EHR) signals, collected within 24 h of ICU admission, for traumatic brain injury (TBI) patient outcomes. Using data from TBI patients in the multi-center eICU database, we focused on in-hospital mortality, neurological status based on the Glasgow Coma Score (mGCS) motor subscore at discharge, and prolonged ICU stay (PLOS). Three machine learning (ML) models were developed, utilizing EHR features, PTS signals collected 24 h after ICU admission, and their combination. External validation was performed using the MIMIC III dataset, and interpretability was enhanced using the Shapley Additive Explanations (SHAP) algorithm. The analysis included 1085 TBI patients. Compared to individual models and existing scoring systems, the combination of EHR and PTS features demonstrated comparable or even superior performance in predicting in-hospital mortality (AUROC = 0.878), neurological outcomes (AUROC = 0.877), and PLOS (AUROC = 0.835). The model's performance was validated in the MIMIC III dataset, and SHAP algorithms identified six key intervention points for EHR features related to prognostic outcomes. Moreover, the EHR results (All AUROC >0.8) were translated into online tools for clinical use. Our study highlights the importance of early-stage PTS signals in predicting TBI patient outcomes. The integration of interpretable algorithms and simplified prediction tools can support treatment decision-making, contributing to the development of accurate prediction models and timely clinical intervention.

The Quantitative Evaluation of Automatic Segmentation in Lumbar Magnetic Resonance Images.

This study aims to overcome challenges in lumbar spine imaging, particularly lumbar spinal stenosis, by developing an automated segmentation model using advanced techniques. Traditional manual measurement and lesion detection methods are limited by subjectivity and inefficiency. The objective is to create an accurate and automated segmentation model that identifies anatomical structures in lumbar spine magnetic resonance imaging scans. Leveraging a dataset of 539 lumbar spinal stenosis patients, the study utilizes the residual U-Net for semantic segmentation in sagittal and axial lumbar spine magnetic resonance images. The model, trained to recognize specific tissue categories, employs a geometry algorithm for anatomical structure quantification. Validation metrics, like Intersection over Union (IOU) and Dice coefficients, validate the residual U-Net's segmentation accuracy. A novel rotation matrix approach is introduced for detecting bulging discs, assessing dural sac compression, and measuring yellow ligament thickness. The residual U-Net achieves high precision in segmenting lumbar spine structures, with mean IOU values ranging from 0.82 to 0.93 across various tissue categories and views. The automated quantification system provides measurements for intervertebral disc dimensions, dural sac diameter, yellow ligament thickness, and disc hydration. Consistency between training and testing datasets assures the robustness of automated measurements. Automated lumbar spine segmentation with residual U-Net and deep learning exhibits high precision in identifying anatomical structures, facilitating efficient quantification in lumbar spinal stenosis cases. The introduction of a rotation matrix enhances lesion detection, promising improved diagnostic accuracy, and supporting treatment decisions for lumbar spinal stenosis patients.

FCER1G as a novel immune-associated blood biomarker in cardiogenic stroke

BackgroundCardioembolic stroke (CE) exhibits the highest recurrence rate and mortality rate among all subtypes of cerebral ischemic stroke (CIS), yet its pathogenesis remains uncertain. The immune system plays a pivotal role in the progression of CE. Growing evidence indicates that several immune-associated blood biomarkers may inform the causes of stroke. The study aimed to identify new immune-associated blood biomarkers in patients with CE and create an online predictive tool in distinguishing CE from noncardioembolic stroke (non-CE) in CIS. MethodsGene expression profiles that were publicly available were obtained from the Gene Expression Omnibus (GEO). The identification of differentially expressed genes (DEGs) was conducted using the Limma package. The hub module and hub genes were identified through the application of weighted gene coexpression network analysis (WGCNA). In order to identify potential diagnostic biomarkers for CE, both the random forest (RF) model and least absolute shrinkage and selection operator (LASSO) regression analysis were employed. Concurrently, the CIBERSORT algorithm was employed to evaluate the infiltration of immune cells in CE samples and examine the correlation between the biomarkers and the infiltrating immune cells. The diagnostic gene expression in blood samples was confirmed using qRT-PCR in a self-constructed dataset. Univariate and multiple logistic regression analyses were used to identify the risk factors for CE. Subsequently, the mathematical model of the nomogram was employed via Java's "Spring Boot" framework to develop the corresponding online tool, which was then deployed on a cloud server utilizing "nginx". ResultsEleven differentially expressed genes (DEGs) that were upregulated and seven DEGs that were downregulated were identified. Through bioinformatics analysis and clinical sample verification, it was discovered that Fc Fragment of IgE Receptor Ig (FCER1G) could serve as a novel potential blood biomarker for CE. FCER1G, along with other risk factors associated with CE, were utilized to develop a nomogram. The training and validation sets, which consisted of 65 CIS patients, yielded areas under the curve (AUCs) of 0.9722 and 0.9689, respectively. These results indicate a high level of precision in risk delineation by the nomogram. Furthermore, the associated online predictive platform has the potential to serve as a more efficacious and appropriate predictive instrument (https://www.origingenetic.com/CardiogenicStroke-FCER1G) for distinguishing between CE and non-CE. ConclusionBlood biomarker FCER1G has the potential to identify patients who are at a higher risk of cardioembolism and direct the search for occult AF.The utilization of this online tool is anticipated to yield significant implications in terms of distinguishing between CE and non-CE, as well as enhancing the optimization of treatment decision support.

PET/CT deep learning prognosis for treatment decision support in esophageal squamous cell carcinoma

ObjectivesThe clinical decision-making regarding choosing surgery alone (SA) or surgery followed by postoperative adjuvant chemotherapy (SPOCT) in esophageal squamous cell carcinoma (ESCC) remains controversial. We aim to propose a pre-therapy PET/CT image-based deep learning approach to improve the survival benefit and clinical management of ESCC patients.MethodsThis retrospective multicenter study included 837 ESCC patients from three institutions. Prognostic biomarkers integrating six networks were developed to build an ESCC prognosis (ESCCPro) model and predict the survival probability of ESCC patients treated with SA and SPOCT. Patients who did not undergo surgical resection were in a control group. Overall survival (OS) was the primary end-point event. The expected improvement in survival prognosis with the application of ESCCPro to assign treatment protocols was estimated by comparing the survival of patients in each subgroup. Seven clinicians with varying experience evaluated how ESCCPro performed in assisting clinical decision-making.ResultsIn this retrospective multicenter study, patients receiving SA had a median OS 9.2 months longer than controls. No significant differences in survival were found between SA patients with predicted poor outcomes and the controls (p > 0.05). It was estimated that if ESCCPro was used to determine SA and SPOCT eligibility, the median OS in the ESCCPro-recommended SA group and SPOCT group would have been 15.3 months and 24.9 months longer, respectively. In addition, ESCCPro also significantly improved prognosis accuracy, certainty, and the efficiency of clinical experts.ConclusionESCCPro assistance improved the survival benefit of ESCC patients and the clinical decision-making among the two treatment approaches.Critical relevance statementThe ESCCPro model for treatment decision-making is promising to improve overall survival in ESCC patients undergoing surgical resection and patients undergoing surgery followed by postoperative adjuvant chemotherapy.Key PointsESCC is associated with a poor prognosis and unclear ideal treatments.ESCCPro predicts the survival of patients with ESCC and the expected benefit from SA.ESCCPro improves clinicians’ stratification of patients’ prognoses.Graphical

GPT-4 as a biomedical simulator

BackgroundComputational simulation of biological processes can be a valuable tool for accelerating biomedical research, but usually requires extensive domain knowledge and manual adaptation. Large language models (LLMs) such as GPT-4 have proven surprisingly successful for a wide range of tasks. This study provides proof-of-concept for the use of GPT-4 as a versatile simulator of biological systems. MethodsWe introduce SimulateGPT, a proof-of-concept for knowledge-driven simulation across levels of biological organization through structured prompting of GPT-4. We benchmarked our approach against direct GPT-4 inference in blinded qualitative evaluations by domain experts in four scenarios and in two quantitative scenarios with experimental ground truth. The qualitative scenarios included mouse experiments with known outcomes and treatment decision support in sepsis. The quantitative scenarios included prediction of gene essentiality in cancer cells and progression-free survival in cancer patients. ResultsIn qualitative experiments, biomedical scientists rated SimulateGPT's predictions favorably over direct GPT-4 inference. In quantitative experiments, SimulateGPT substantially improved classification accuracy for predicting the essentiality of individual genes and increased correlation coefficients and precision in the regression task of predicting progression-free survival. ConclusionThis proof-of-concept study suggests that LLMs may enable a new class of biomedical simulators. Such text-based simulations appear well suited for modeling and understanding complex living systems that are difficult to describe with physics-based first-principles simulations, but for which extensive knowledge is available as written text. Finally, we propose several directions for further development of LLM-based biomedical simulators, including augmentation through web search retrieval, integrated mathematical modeling, and fine-tuning on experimental data.

YouTube as a reliable source of information on antibody drug conjugates: A qualitative analysis.

9050 Background: YouTube is an open-access source of information which can serve a platform for health information and misinformation. Antibody Drug Conjugates (ADC) are a novel therapy that utilizes monoclonal antibodies to target antigens expressed on cancer cells, thus delivering cytotoxic agents specifically to malignant cells and minimizing toxic effects on healthy cells. Our study sought to assess the quality and engagement parameters of the 100 most viewed videos on YouTube regarding ADCs. Methods: A YouTube search was conducted with the keywords “Antibody Drug Conjugate” and “cancer”. The videos were sorted by the highest number of views. Duplicates and videos not in the English language were excluded. A total of 100 videos were included in the final analysis. They were organized into three categories: academic, commercial, and media. Next, they were assessed for quality using two validated quality instruments, the DISCERN and PEMAT scoring systems. The number of views, comments, and likes were also evaluated. A Pearson correlation and descriptive statistics using R Studio v2002.12.0-353 was then used to generate data. Results: The total number of views on the videos on YouTube regarding ADC was 95,969 with a (median number of views of 723.5). The videos were categorized into 71% from academic sources, 28% from commercial, and 1% from media sources. The mean DISCERN sum was 41.13. The number of YouTube videos that were rated as poor or very poor were 44%, followed by 33% that was rated as fair, and only 23% were rated as good or excellent. 21% videos discussed risks and benefits, while only 6 % of the videos provided support in shared decision making. 63% of the videos stated the aims clearly. Based on the PEMAT score, 87% of videos made their purpose evident, 34% summarized the videos, and 31% explained how to use the medication. There was a significant positive correlation in number of likes on videos and DISCERN score (r 0.253, 95% CI 0.06-0.428, p 0.011), PEMAT understandability score (r 0.197, 95% CI 0.00063-0.379, p 0.0494), and PEMAT accountability (r 0.2, 95% CI 0.00368-0.381, p 0.0461). There was also a positive correlation between length of video and DISCERN score (r 0.32, 95% CI 0.13-0.49, p<0.001), PEMAT understandability score (r 0.23, 95% CI 0.04-0.41, p 0.01), and PEMAT actionability score (r 0.4, 95% CI 0.31-0.62, p <0.001). Conclusions: Our analysis revealed that the overall quality of videos was suboptimal, with the majority categorized as poor or fair. Important aspects such as risks, benefits and shared decision-making were inadequately addressed, indicating a potential gap in patient education and informed decision-making. Although longer videos have positive correlation, it would be impactful to keep the duration of the video shorter. Efforts to promote high-quality, patient-centered educational content on ADCs are needed to ensure access to reliable information to support treatment decisions.

Treatment patterns among HCPs and concordance with expert recommendations in HER2-negative early breast cancer: Analysis of an interactive online treatment decision support tool.

e13601 Background: Management of HER2- EBC continues to evolve with expansion of targeted therapies, including abemaciclib, olaparib, and pembrolizumab, in the adjuvant setting for eligible pts. We sought to identify contemporary practice patterns among HCPs for HER2- EBC and compare them with those of experts using an online interactive decision support tool. Methods: An online decision support tool was developed with input from 5 BC experts to provide tx recommendations for 19 unique pt case scenarios based on considerations including subtype, tx history, and recurrence risk. HCPs entered specific patient criteria along with their intended management for that case, and then were shown 5 expert recommendations. HCPs were asked if the recommendations impacted their intended approach. An analysis of expert recommendations and HCP-selected tx was performed. Results: Between Oct 2023 and Jan 2024, 83 HCPs entered 105 cases. Of the 97 cases where HCPs noted a definitive tx plan, recommendations were discordant from experts in 29%. The most commonly queried cases, expert recommendations, and discordance between HCPs and experts are outlined in the table. Discord with experts was similar among HCP tx plans for cases of TNBC (33%) and HR+/HER2- EBC (27%). HCPs predominantly sought guidance for pts with HR+/HER2- EBC, no g BRCAm, and high-risk disease per monarchE trial criteria (79%), with 22% discordance from expert recommendations. Regardless of BC subtype, the highest level of discord between HCPs and experts (67%) was observed in cases where pts had received neoadjuvant tx and had no residual disease following surgery, although this represents a small subset (n = 12). When asked if the expert recommendations changed their tx choice, 7 of 27 (26%) HCPs who responded indicated that expert recommendations changed their intended tx, whereas 15% noted that there were barriers to implementing those recommendations. Conclusions: These data suggest ongoing challenges with incorporating the newest targeted therapies for high-risk EBC management. Continued education and development of resources for HCPs, including online decision support tools, may be increasingly important as the treatment of EBC continues to evolve. [Table: see text]

Skin fragility and wound management in Ehlers- Danlos Syndrome: a report by the Ehlers Danlos Syndrome society skin working group.

The Ehlers-Danlos Syndromes (EDS) are a heterogenous group of heritable connective tissue disorders, characterised by joint hypermobility, skin hyperextensibility and generalised tissue fragility. In all types of EDS skin wound healing is impaired to a variable degree. Additional support through wound management plans may help to improve these outcomes, however, there is paucity of evidence regarding clinical management of skin fragility and wounds in EDS. This paper aims to review current evidence and provide recommendations for management of skin wounds in EDS types. Preventative measures to avoid skin injury are strongly recommended, including avoidance of high impact sport and use of appropriate protection such as shin guards. Bruising is common and some types of EDS are associated with haematoma formation with management including compression bandages and consideration of pharmacological therapy. Skin fragility and tears should be managed with a focus on protection of remaining tissue, avoidance of wound tension and low adherence dressings to avoid further injury. This paper provides clear recommendations to address skin management for this group of patients. It highlights the lack of good quality published data to support treatment decisions.