24 July 2019 in JMIR mHealth and uHealth
Journal of Medical Internet Research - International Scientific Journal for Medical Research, Information and Communication on the Internet
5 March 2025 in Journal of Medical Internet Research
Background: Molecular tumor boards (MTBs) require intensive manual investigation to generate optimal treatment recommendations for patients. Large language models (LLMs) can catalyze MTB recommendations, decrease human error, improve accessibility to care, and enhance the efficiency of precision oncology. Objective: In this study, we aimed to investigate the efficacy of LLM-generated treatments for MTB patients. We specifically investigate the LLMs’ ability to generate evidence-based treatment recommendations using PubMed references. Methods: We built a retrieval augmented generation pipeline using PubMed data. We prompted the resulting LLM to generate treatment recommendations with PubMed references using a test set of patients from an MTB conference at a large comprehensive cancer center at a tertiary care institution. Members of the MTB manually assessed the relevancy and correctness of the generated responses. Results: A total of 75% of the referenced articles were properly cited from PubMed, while 17% of the referenced articles were hallucinations, and the remaining were not properly cited from PubMed. Clinician-generated LLM queries achieved higher accuracy through clinician evaluation than automated queries, with clinicians labeling 25% of LLM responses as equal to their recommendations and 37.5% as alternative plausible treatments. Conclusions: This study demonstrates how retrieval augmented generation–enhanced LLMs can be a powerful tool in accelerating MTB conferences, as LLMs are sometimes capable of achieving clinician-equal treatment recommendations. However, further investigation is required to achieve stable results with zero hallucinations. LLMs signify a scalable solution to the time-intensive process of MTB investigations. However, LLM performance demonstrates that they must be used with heavy clinician supervision, and cannot yet fully automate the MTB pipeline.
1 June 2014 in Human Gene Therapy Clinical Development
Oncolytic virotherapy with measles vaccine virus (MeV) already has been demonstrated to be safe. However, early clinical results pointed out the necessity for an enhancement of oncolytic effectiveness of MeV-based virotherapeutics. In our work, we are developing an armed measles vaccine virus (MeV-SCD) encoding a suicide fusion gene of yeast cytosine deaminase/uracil phosphoribosyltransferase, converting the nontoxic prodrug 5-fluorocytosine (5-FC) to the chemotherapeutic drug 5-fluorouracil (5-FU). To preclinically investigate what an optimal prodrug-assisted therapeutic regimen might look like, we added 5-FC at various time points after infection with MeV-SCD and either let the prodrug remain in the tumor cell culture medium continuously for various time periods (“continuous” 5-FC application) or applied it only temporarily for defined shorter periods of time (“pulsed” 5-FC application); we also varied the time point at which 5-FC was added after infection with MeV-SCD. As a result, addition of the prodrug at early times postinfection (e.g., at 3 hr postinfection) was found to be inferior concerning the overall oncolytic effectiveness when compared with addition of 5-FC at later time points (e.g., at 24 hr postinfection). Next, oncolytic effectiveness was found to correlate positively with the overall duration of incubation of MeV-infected tumor cells with 5-FC. Of note, this was true despite our finding that addition of the prodrug could also exert an inhibitory effect on the generation of infectious progeny viral particles, that is, on virus replication. These findings should be helpful for the rational design of further trials (preclinical, clinical) using suicide gene armed virotherapeutics, such as MeV-SCD.
12 November 2019 in JMIR Medical Informatics
Background: An electronic health record (EHR) is the state-of-the-art method for ensuring all data concerning a given patient are up to date for use by multidisciplinary hospital teams. Therefore, medical students need to be trained to use health information technologies within this environment from the early stages of their education. Objective: As little is known about the effects of specific training within the medical curriculum, this study aimed to develop a course module and evaluate it to offer best practice teaching for today’s students. Moreover, we looked at the acceptance of new technologies such as EHRs. Methods: Fifth-year medical students (N=104) at the University of Tübingen took part in a standardized two-day training procedure about the advantages and risks of EHR use. After the training, students performed their own EHR entries on hypothetical patient cases in a safe practice environment. In addition, questionnaires—standardized and with open-ended questions—were administered to assess students’ experiences with a new teaching module, a newly developed EHR simulator, the acceptance of the health technology, and their attitudes toward it before and after training. Results: After the teaching, students rated the benefit of EHR training for medical knowledge significantly higher than before the session (mean 3.74, SD 1.05). However, they also had doubts about the long-term benefit of EHRs for multidisciplinary coworking after training (mean 1.96, SD 0.65). The special training with simulation software was rated as helpful for preparing students (88/102, 86.2%), but they still did not feel safe in all aspects of EHR. Conclusions: A specific simulated training on using EHRs helped students improve their knowledge and become more aware of the risks and challenges of such a system. Overall, students welcomed the new training module and supported the integration of EHR teaching into the medical curriculum. Further studies are needed to optimize training modules and make use of long-term feedback opportunities a simulated system offers.
4 February 2026 in Nature
Hepatocellular carcinoma (HCC) is the fastest growing cause of cancer-related mortality and there are limited therapies 1 . Although endoplasmic reticulum (ER) stress and the unfolded protein response (UPR) are implicated in HCC, the involvement of the UPR transducer ATF6α remains unclear 2 . Here we demonstrate the function of ATF6α as an ER-stress-inducing tumour driver and metabolic master regulator restricting cancer immunosurveillance for HCC, in contrast to its well-characterized role as an adaptive response to ER stress 3 . ATF6α activation in human HCC is significantly correlated with an aggressive tumour phenotype, characterized by reduced patient survival, enhanced tumour progression and local immunosuppression. Hepatocyte-specific ATF6α activation in mice induced progressive hepatitis with ER stress, immunosuppression and hepatocyte proliferation. Concomitantly, activated ATF6α increased glycolysis and directly repressed the gluconeogenic enzyme FBP1 by binding to gene regulatory elements. Restoring FBP1 expression limited ATF6α-activation-related pathologies. Prolonged ATF6α activation in hepatocytes triggered hepatocarcinogenesis, intratumoural T cell infiltration and nutrient-deprived immune exhaustion. Immune checkpoint blockade (ICB) 4 restored immunosurveillance and reduced HCC. Consistently, patients with HCC who achieved a complete response to immunotherapy displayed significantly increased ATF6α activation compared with those with a weaker response. Targeting Atf6 through germline ablation, hepatocyte-specific ablation or therapeutic hepatocyte delivery of antisense oligonucleotides dampened HCC in preclinical liver cancer models. Thus, prolonged ATF6α activation drives ER stress, leading to glycolysis-dependent immunosuppression in liver cancer and sensitizing to ICB. Our findings suggest that persistently activated ATF6α is a tumour driver, a potential stratification marker for ICB response and a therapeutic target for HCC.
15 October 2025 in Nature Medicine