Communications medicine最新文献

{"title":"The unprecedented Paxlovid journey from milligrams to millions of patient doses during the Covid-19 pandemic.","authors":"Weili Yu, Mahesh K Krishnan, Matt Weekly, Ravi M Shanker, Pankaj Doshi, John A Ragan, Robert A Greene, Brett Gampper, Stéphane Caron, Andrew McKillop, John Ludwig, Mikael Dolsten","doi":"10.1038/s43856-025-00798-7","DOIUrl":"https://doi.org/10.1038/s43856-025-00798-7","url":null,"abstract":"","PeriodicalId":72646,"journal":{"name":"Communications medicine","volume":"5 1","pages":"80"},"PeriodicalIF":5.4,"publicationDate":"2025-03-20","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143671735","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Acquired resistance to immune checkpoint therapy is caused by glycoprotein non-metastatic melanoma protein B signal cascade.","authors":"Jin-Sung Chung, Vijay Ramani, Lei Guo, Vinita Popat, Ponciano D Cruz, Lin Xu, Hans Hammers, Kiyoshi Ariizumi","doi":"10.1038/s43856-025-00786-x","DOIUrl":"https://doi.org/10.1038/s43856-025-00786-x","url":null,"abstract":"<p><strong>Background: </strong>Acquired resistance (AR) is a major limitation of immune checkpoint inhibitor (ICI) therapy when treating renal cell carcinoma (RCC). Understanding who will get AR is currently unknown. We hypothesized the T-cell-inhibitory glycoprotein non-metastatic melanoma protein B (GPNMB) to be a prognostic marker for patients with AR.</p><p><strong>Methods: </strong>Alongside other markers, GPNMB was measured in the blood of RCC patients (n = 39) several times after starting ICI treatment and analyzed for association with Response Evaluation Criteria in Solid Tumors (RECIST) tumor response. To better understand the role of GPNMB in AR, we created an ICI-resistant RenCa mouse kidney cancer model by repeatedly selecting the largest tumors that grew in ICI-treated mice.</p><p><strong>Results: </strong>Here we show that among patients who positively respond to ICI, a subset of patients (n = 9) acquire resistance within 2 years that coincides with an increase in serum GPNMB. Our mouse model recapitulates this elevation in GPNMB at the onset of AR which is triggered by cytoplasmic motif signaling via the Programmed cell death ligand 1 (PDL1) receptor that is known to protect tumors from Interferon-gamma (IFN-γ) cytotoxicity. This PDL1-induced signal leads to upregulation of the SRY-box transcription factor 10 (SOX10), dysregulation of the microphthalmia-associated transcription factor (MITF) pathway, and overexpression of GPNMB. Indeed, activation of SOX10-MITF signaling is present in plasma cell-free RNA from RCC patients who develop AR.</p><p><strong>Conclusions: </strong>Elevation of the SOX10-MITF-GPNMB signal cascade via the PDL1 receptor leads to AR. Therefore, GPNMB can be both a prognosticator of and a potential treatment target for overcoming AR to ICI treatment in RCC.</p>","PeriodicalId":72646,"journal":{"name":"Communications medicine","volume":"5 1","pages":"79"},"PeriodicalIF":5.4,"publicationDate":"2025-03-20","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143671732","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Titanium micro-particles are commonly found in soft tissues surrounding dental implants.","authors":"Carlotta Dionigi, Gyula Nagy, Jan Derks, Yuki Ichioka, Cristiano Tomasi, Lena Larsson, Daniel Primetzhofer, Tord Berglundh","doi":"10.1038/s43856-025-00756-3","DOIUrl":"https://doi.org/10.1038/s43856-025-00756-3","url":null,"abstract":"<p><strong>Background: </strong>Dental implants are one of the most frequently used medical devices for therapeutic purposes in dentistry. Peri-implantitis is a severe, microbial biofilm-associated condition, characterized by inflammation in peri-implant soft tissues and destruction of supporting bone. It has been suggested that metal particles originating from the implant may influence the local host response to microbial biofilms.</p><p><strong>Methods: </strong>Soft tissue biopsies were collected from implant sites with and without peri-implantitis in 21 patients. Micro Proton-induced X-ray Emission (µ-PIXE) analysis was used to localize, quantify and characterize titanium micro-particles within tissues. RNA sequencing was performed to evaluate potential associations between titanium micro-particles and gene expression profiles in peri-implantitis lesions.</p><p><strong>Results: </strong>Titanium micro-particles are consistent findings in soft tissues surrounding dental implants. Their occurrence varies across patients but not between sites with and without peri-implantitis within the same individual. Most particles reside in a 2-mm wide tissue portion close to the implant/tissue interface. The time in function of the implants does not influence the volumetric density of titanium micro-particles, while implant systems do. Fourteen differentially expressed genes are identified when comparing peri-implantitis samples with high and low densities of titanium micro-particles. The gene-set enrichment analysis reveals functions related to the regulation of the immune response and epithelial development.</p><p><strong>Conclusions: </strong>The present results indicate that titanium micro-particles are commonly found in tissues surrounding dental implants and are not associated with the occurrence of peri-implantitis.</p>","PeriodicalId":72646,"journal":{"name":"Communications medicine","volume":"5 1","pages":"78"},"PeriodicalIF":5.4,"publicationDate":"2025-03-18","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143659871","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Utility and cost-effectiveness of LiverMultiScan for MASLD diagnosis: a real-world multi-national randomised clinical trial.","authors":"Elizabeth Shumbayawonda, Marika French, Jane Elizabeth Carolan, Cayden Beyer, Paula Lorgelly, Dimitar Tonev, Rajarshi Banerjee, Michael H Miller, Christopher D Byrne, Janisha Patel, Saima Ajaz, Kosh Agarwal, Johanna Backhus, Minneke J Coenraad, Jelte J Schaapman, Andrew Fraser, Miguel Castelo Branco, Stephen Barclay, Matthias M Dollinger, Daniel J Cuthbertson, Daniel Forton, Hildo J Lamb","doi":"10.1038/s43856-025-00796-9","DOIUrl":"https://doi.org/10.1038/s43856-025-00796-9","url":null,"abstract":"<p><strong>Background: </strong>Increasing prevalence of metabolic dysfunction-associated liver disease (MASLD) and metabolic dysfunction-associated steatohepatitis (MASH) poses a growing healthcare burden. Noninvasive diagnostic tools to replace liver biopsy are urgently needed. We investigated the utility and cost-effectiveness of including multiparametric magnetic resonance imaging (mpMRI) to the management of adults with suspected MASLD multi-nationally.</p><p><strong>Methods: </strong>RADIcAL-1, a 1:1 randomised controlled trial (standard-of-care [SoC] vs. imaging arm [IA; SoC+mpMRI]) included 802 participants from Germany, Netherlands, Portugal and UK. Wilcoxon-rank tests were used to compare access to healthcare practitioners, patient assessments and proportion of patients with a diagnosis (%diagnosis). Liver fat and disease activity (corrected T1 [cT1]) were used to identify patients not requiring biopsy in the imaging arm. Primary endpoint was mpMRI cost-effectiveness and improvement in resource use (visits avoided) using mpMRI.</p><p><strong>Results: </strong>mpMRI is cost-effective with an ICER of €4968/QALY gained. 403 were randomised to IA and 399 to SoC. SoC has significantly more specialist appointments (p = 0.015) and patient assessments (p < 0.001). Across all involved hospitals, %diagnosis is significantly higher in the imaging arm (p = 0.0012). cT1 correctly classifies 50% of patients without MASH with fibrosis and can avoid biopsy. Including all costs, the imaging arm incurs higher short-term per-patient healthcare expenditure compared to the SoC arm (€1,300 vs. €830).</p><p><strong>Conclusion: </strong>Adding mpMRI to SoC for the management of adults with suspected MASLD multi-nationally is cost-effective, enhances rate of diagnosis multi-nationally and increases rate of diagnosis without increasing other liver-related health care resource use. Due to the need for standardisation of SoC, widespread use can support optimisation of the MASLD clinical pathway and improve long-term patient management.</p>","PeriodicalId":72646,"journal":{"name":"Communications medicine","volume":"5 1","pages":"74"},"PeriodicalIF":5.4,"publicationDate":"2025-03-18","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143659875","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Identifying and ranking non-traditional risk factors for cardiovascular disease prediction in people with type 2 diabetes.","authors":"Katarzyna Dziopa, Nishi Chaturvedi, Folkert W Asselbergs, Amand F Schmidt","doi":"10.1038/s43856-025-00785-y","DOIUrl":"10.1038/s43856-025-00785-y","url":null,"abstract":"<p><strong>Background: </strong>Cardiovascular disease (CVD) prediction models perform poorly in people with type 2 diabetes (T2DM). We aimed to identify potentially non-traditional CVD predictors for six facets of CVD (including coronary heart disease, ischemic stroke, heart failure, and atrial fibrillation) in people with T2DM.</p><p><strong>Methods: </strong>We analysed data on 600+ features from the UK Biobank, stratified by history of CVD and T2DM: 459,142 participants without diabetes or CVD, 14,610 with diabetes but without CVD, and 4432 with diabetes and CVD. A penalised generalized linear model with a binomial distribution was used to identify CVD-related features. Subsequently, a 20% hold-out set was used to replicate identified features and provide an importance based ranking.</p><p><strong>Results: </strong>Here we show that non-traditional risk factors are of particular importance in people with diabetes. Classical CVD risk factors (e.g. family history, high blood pressure) rank highly in people without diabetes. For individuals with T2DM but no CVD, top predictors include cystatin C, self-reported health satisfaction, biochemical measures of ill health. In people with diabetes and CVD, key predictors are self-reported ill health and blood cell counts. Unique diabetes-related risk factors include dietary patterns, mental health and biochemistry measures (e.g. oestradiol, rheumatoid factor). Adding these features improves risk stratification; per 1000 people with diabetes, 133 CVD and 165 HF cases receive a higher risk.</p><p><strong>Conclusions: </strong>This study identifies numerous replicated non-traditional CVD risk factors for people with T2DM, providing insight to improve guideline recommended risk prediction models which currently overlook these features.</p>","PeriodicalId":72646,"journal":{"name":"Communications medicine","volume":"5 1","pages":"77"},"PeriodicalIF":5.4,"publicationDate":"2025-03-14","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC11906859/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143625817","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Targeting the retinoid signaling pathway with YCT-529 for effective and reversible oral contraception in mice and primates.","authors":"Nadja Mannowetz, Sanny S W Chung, Soma Maitra, Md Abdullah Al Noman, Henry L Wong, Narsihmulu Cheryala, Akash Bakshi, Debra J Wolgemuth, Gunda I Georg","doi":"10.1038/s43856-025-00752-7","DOIUrl":"10.1038/s43856-025-00752-7","url":null,"abstract":"<p><strong>Background: </strong>The retinoic acid receptor alpha (Rarα) has been validated as a male contraceptive target by genetic knockouts resulting in male sterility. The effects on spermatogenesis in the absence of RARα resemble the loss of RAR signaling in vitamin A deficiency, and the mice are otherwise normal. The effects on spermatogenesis in animals treated orally with the dual RARα/RARγ antagonist BMS-189453 closely phenocopies the absence of RARα function. Notably, the resulting male sterility is reversible. We, therefore, wished to identify RARα-selective inhibitors for potential male non-hormonal contraception.</p><p><strong>Methods: </strong>YCT-529 was investigated for RARα selective inhibition, physicochemical characteristics, oral bioavailability, and pharmacokinetic properties in mice and non-human primates. It was assessed in mouse mating trials to determine the most effective dosing regimen to induce infertility in male mice and in male non-human primates to reduce sperm levels.</p><p><strong>Results: </strong>Characterization of YCT-529 shows suitable biochemical, physicochemical, and pharmacokinetic properties for in vivo testing. YCT-529 inhibits mouse fertility of male mice within 4 weeks of oral administration, correlating with disrupted spermatogenesis demonstrating specific inhibition of the RARα pathway. Within 6 weeks after cessation of dosing, mouse fertility reverses. Furthermore, YCT-529 inhibits sperm production in a non-human primate model within 2 weeks of oral dosing without adverse side effects. Within 10-15 weeks after cessation of dosing, non-human primates' sperm counts fully reverses.</p><p><strong>Conclusions: </strong>These results lay the groundwork for evaluating YCT-529 in human clinical trials.</p>","PeriodicalId":72646,"journal":{"name":"Communications medicine","volume":"5 1","pages":"68"},"PeriodicalIF":5.4,"publicationDate":"2025-03-13","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC11906769/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143626700","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Age- and vaccination status-dependent isolation guidelines based on simulation of SARS-CoV-2 Delta cases in Singapore.","authors":"Keisuke Ejima, Marco Ajelli, Ananya Singh, Hoong Kai Chua, Luis Ponce, Yuqian Wang, Yong Dam Jeong, Shingo Iwami, Kenji Shibuya, Kiyosu Taniguchi, Norio Ohmagari, Po Ying Chia, Sean W X Ong, Kelvin Bryan Tan, David Chien Lye, Barnaby E Young","doi":"10.1038/s43856-025-00797-8","DOIUrl":"10.1038/s43856-025-00797-8","url":null,"abstract":"<p><strong>Background: </strong>In the absence of effective pharmaceutical interventions early in an infectious disease outbreak, non-pharmaceutical measures, especially isolating infected individuals, critically limit its impact. The ongoing COVID-19 pandemic has sparked debates on optimal isolation guidelines. This study proposes a variable isolation period approach (variable-period approach), tailoring isolation durations for distinct population groups with varied viral load dynamics.</p><p><strong>Methods: </strong>To compare our variable-period approach with a fixed-period strategy, we developed a simulation model generating synthetic longitudinal SARS-CoV-2 viral load data. The data was generated from the viral dynamics model parameterized using SARS-CoV-2 Delta patient data in Singapore, accounting for age and vaccination status.</p><p><strong>Results: </strong>Findings show that age and vaccination status significantly influence viral dynamics, with younger age and vaccination linked to shorter viral shedding durations. The variable-period framework suggests longer isolation lengths for older and unvaccinated individuals. By setting the leaking risk (risk of remaining infectious at the end of isolation) below 10%, the optimal fixed-period isolation is 14 days, with an average excess isolation burden of 7.4 unnecessary days. In contrast, the variable-period guideline reduces the excess isolation burden to 6.0 days, with the optimal isolation periods ranging from 9 to 16 days, depending on the population group. We confirmed similar results when we used the effective reproduction number as an alternative to the leaking risk.</p><p><strong>Conclusions: </strong>In this case, study using the SARS-CoV-2 Delta variant, our analysis demonstrates that unnecessary time spent in isolation can be reduced by adopting variable-period guidelines based on patient characteristics.</p>","PeriodicalId":72646,"journal":{"name":"Communications medicine","volume":"5 1","pages":"76"},"PeriodicalIF":5.4,"publicationDate":"2025-03-13","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC11906760/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143626977","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Magnetic particle imaging angiography of the femoral artery in a human cadaveric perfusion model.","authors":"Viktor Hartung, Philipp Gruschwitz, Anne Marie Augustin, Jan-Peter Grunz, Florian Kleefeldt, Dominik Peter, Süleyman Ergün, Johanna Günther, Teresa Reichl, Thomas Kampf, Martin Andreas Rückert, Stefan Herz, Volker Christian Behr, Thorsten Alexander Bley, Patrick Vogel","doi":"10.1038/s43856-025-00794-x","DOIUrl":"10.1038/s43856-025-00794-x","url":null,"abstract":"<p><strong>Background: </strong>Magnetic particle imaging (MPI) allows for radiation-free visualization of tracers without background signal. With the first human-sized interventional MPI scanner being recently developed, the aim of the present study was to test its performance for guiding of endovascular procedures in a realistic perfusion model.</p><p><strong>Methods: </strong>Three fresh-frozen cadaveric legs were prepared to establish continuous circulation in the superficial femoral artery via introducer sheaths in the inguinal and infragenicular region. To facilitate vessel visualization, a mixture of a MPI tracer (Resotran® or Perimag®) and X-ray contrast agent was injected under continuous extracorporeal perfusion and imaged simultaneously with MPI angiography and digital subtraction angiography (DSA) as reference.</p><p><strong>Results: </strong>The MPI scanner integrates seamlessly into the standard operating procedures in the angiography suite and simultaneous imaging with DSA and MPI is feasible. The MPI scanner detects a tracer bolus of 2 ml Perimag® or 1.5 ml Resotran®. Imaging results are consistent and reproducible in three cadaveric leg phantoms.</p><p><strong>Conclusion: </strong>This study demonstrates, that the recently developed human-sized MPI scanner facilitates reliable radiation-free image guidance for peripheral vascular interventions in the superficial femoral artery with a tracer approved for use in humans.</p>","PeriodicalId":72646,"journal":{"name":"Communications medicine","volume":"5 1","pages":"75"},"PeriodicalIF":5.4,"publicationDate":"2025-03-13","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC11906881/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143626686","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Remission of alcohol use disorder following traumatic brain injury with focal orbitofrontal cortex hemorrhage: case report and network mapping.","authors":"Saarah Haque, Albert Bellmunt-Gil, Benjamin Davidson, Christian Lüscher, Michael D Fox, Juho Joutsa, Matthew J Burke","doi":"10.1038/s43856-025-00760-7","DOIUrl":"10.1038/s43856-025-00760-7","url":null,"abstract":"<p><strong>Background: </strong>The orbitofrontal cortex (OFC) and its role in the regulation of urges/compulsion has been identified as a critical component of circuit-based addiction models. Building on such models, it was recently shown that brain lesions disrupting addictive behavior can be mapped to a common brain circuit.</p><p><strong>Methods: </strong>We present a case of a 42-year-old woman with chronic treatment-refractory alcohol use disorder who experienced early remission following a traumatic brain injury (TBI) with focal left OFC intracerebral hemorrhage. Using a network mapping approach (normative connectome, n = 1000), functional connectivity was computed from the traced OFC lesion across all brain voxels.</p><p><strong>Results: </strong>The case lesion map topography converges on a brain lesion map previously described as disrupting addictive behavior, but with an inverse connectivity profile (spatial correlation r = -0.59). This spatial correlation is more negative than what would be expected by chance (permutation test 1-sided, p = 0.04) or by random lesion cases (1-sided, p < 0.001).</p><p><strong>Conclusions: </strong>Based on these results, we suggest that potentially just disrupting this brain network, regardless of the directionality, could facilitate remission. However, this case report cannot control for multiple psychosocial factors potentially impacting alcohol remission and caution is also needed for considering TBI as a mechanism for generating an isolated focal lesion. Overall, this case contributes to our understanding of circuit-based models of addictive behavior and could be useful in generating hypotheses for neuromodulatory treatment strategies.</p>","PeriodicalId":72646,"journal":{"name":"Communications medicine","volume":"5 1","pages":"69"},"PeriodicalIF":5.4,"publicationDate":"2025-03-12","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC11903661/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143617864","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Low responsiveness of machine learning models to critical or deteriorating health conditions.","authors":"Tanmoy Sarkar Pias, Sharmin Afrose, Moon Das Tuli, Ipsita Hamid Trisha, Xinwei Deng, Charles B Nemeroff, Danfeng Daphne Yao","doi":"10.1038/s43856-025-00775-0","DOIUrl":"10.1038/s43856-025-00775-0","url":null,"abstract":"<p><strong>Background: </strong>Machine learning (ML) based mortality prediction models can be immensely useful in intensive care units. Such a model should generate warnings to alert physicians when a patient's condition rapidly deteriorates, or their vitals are in highly abnormal ranges. Before clinical deployment, it is important to comprehensively assess a model's ability to recognize critical patient conditions.</p><p><strong>Methods: </strong>We develop multiple medical ML testing approaches, including a gradient ascent method and neural activation map. We systematically assess these machine learning models' ability to respond to serious medical conditions using additional test cases, some of which are time series. Guided by medical doctors, our evaluation involves multiple machine learning models, resampling techniques, and four datasets for two clinical prediction tasks.</p><p><strong>Results: </strong>We identify serious deficiencies in the models' responsiveness, with the models being unable to recognize severely impaired medical conditions or rapidly deteriorating health. For in-hospital mortality prediction, the models tested using our synthesized cases fail to recognize 66% of the injuries. In some instances, the models fail to generate adequate mortality risk scores for all test cases. Our study identifies similar kinds of deficiencies in the responsiveness of 5-year breast and lung cancer prediction models.</p><p><strong>Conclusions: </strong>Using generated test cases, we find that statistical machine-learning models trained solely from patient data are grossly insufficient and have many dangerous blind spots. Most of the ML models tested fail to respond adequately to critically ill patients. How to incorporate medical knowledge into clinical machine learning models is an important future research direction.</p>","PeriodicalId":72646,"journal":{"name":"Communications medicine","volume":"5 1","pages":"62"},"PeriodicalIF":5.4,"publicationDate":"2025-03-11","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC11897252/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143607444","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}