Entrectinib

A 295,000+ patient study revealed variable efficacy among cancers for patients treated with tissue-agnostic drugs and uncovered the potential for expanding approvals to other drugs of the same class IRVING, Texas, March 20, 2025 /PRNewswire/ -- Caris Life Sciences® (Caris), a leading next-generation AI TechBio company and precision medicine pioneer, today announced the publication of "Real-world evidence provides clinical insights into tissue-agnostic therapeutic approvals" in Nature Communications. Utilizing the largest real-world clinical and genomic dataset of tissue-agnostic indications reported to date across more than 295,000 patients, the study provides a comprehensive evaluation of real-world outcomes for patients eligible for these powerful "pan-cancer" therapies, such as pembrolizumab and larotrectinib. These drugs are categorized as tumor-agnostic therapies, targeting specific genetic or molecular features of tumors rather than their anatomical location or histology. The study focused on the use and clinical benefits of tissue-agnostic therapies, which target specific molecular biomarkers across numerous cancer types. Key findings include: More than 20 percent of patients are candidates for a tissue-agnostic drug. Approximately five percent of patients who lacked any tumor-specific indication were found to carry a tissue-agnostic indication and became eligible for tissue-agnostic drug treatment. Tissue-agnostic therapy uptake was poor for rare indications, illustrating the need for enhanced clinical education. Current tissue-agnostic indications could potentially be expanded, as evidenced by clinical benefits in tumor types and drugs of the same class that were not investigated in the original clinical trials. "By harnessing the power of Caris' large clinico-genomic dataset, we have shown that tissue-agnostic drugs produce different outcomes in different tissues," said George W. Sledge, Jr., MD, EVP and Chief Medical Officer of Caris and study author. "We additionally found evidence of clinical benefit beyond approved indications, illustrating the enormous potential of real-world data to inform the clinical use of tissue-agnostic drug approvals, potentially improving patient outcomes." "Caris has generated one of the most extensive clinico-genomic datasets in oncology through years of comprehensive molecular profiling," said Caris President and study author David Spetzler, MS, PhD, MBA. "Our advanced analyses of real-world data now enable us to uncover findings of clinical value for large patient populations. One in five patients in our study were eligible for a tissue-agnostic drug, which demonstrates the widespread impact of our findings." Detailed Scientific Information This study used a real-world database of 295,316 molecularly profiled tumor samples with associated clinical outcomes data across 57 tumor types to investigate the utility and uptake of six tissue-agnostic therapies (pembrolizumab, larotrectinib, entrectinib, dostarlimab, dabrafenib and selpercatinib). At least one current FDA-approved tissue-agnostic indication (TMB-High, MSI-High/dMMR, BRAFV600E mutations, and NTRK and RET fusions) was detected in 21.5% of patients, including 5.4% lacking any tumor-specific indication. In patients with rare NTRK fusions, there was poor clinical uptake of the clinically successful targeted therapies larotrectinib and entrectinib, underlining the need for oncologist education on rare indications. Strikingly, significant differences in pembrolizumab-associated outcomes were observed across tumor types for the most common indications (TMB-High and MSI-High/dMMR), demonstrating that the effects of these therapies are not tissue-agnostic. Clinical benefits were also observed in tumor types and drugs of the same class (such as nivolumab) that were not investigated in pivotal clinical trials. This suggests the possible expansion of therapeutic avenues for a given tissue-agnostic indication. These findings demonstrate the power of Caris' extensive real-world dataset for generating valuable clinical insights to inform the clinical implementations of tissue-agnostic drug approvals and improve outcomes for patients with cancer. Designed to capture and analyze molecular information from tissue and blood in a comprehensive manner, the proprietary Caris Platform is a powerful set of precision medicine solutions, providing Whole Exome Sequencing (WES) and Whole Transcriptome Sequencing (WTS) as standard practice. Caris sequences at sector-leading depth of coverage, which directly correlates to increased accuracy and detection of low-frequency biomarkers of relevance. Caris also evaluates protein biomarkers through an extensive menu of immunohistochemical (IHC) tests analyzed in a tumor-type specific manner, which in combination with WES and WTS, provides a comprehensive view of a patient's disease across DNA, RNA and proteins. The Caris approach reveals an individualized molecular blueprint of a patient's disease, providing actionable, personalized treatment pathways and driving superior clinical outcomes for patients. About Caris Life Sciences Caris Life Sciences® (Caris) is a leading next-generation AI TechBio company and precision medicine pioneer that is actively developing and delivering innovative solutions to revolutionize healthcare and improve the human condition. Through comprehensive molecular profiling (Whole Exome and Whole Transcriptome Sequencing) and the application of advanced AI and machine learning algorithms, Caris has created the large-scale, multimodal database and computing capability needed to analyze and further unravel the molecular complexity of disease. This convergence of sequencing power, big data and AI technologies provides a differentiated platform to deliver the next generation of precision medicine tools for early detection, diagnosis, monitoring, therapy selection and drug development. Caris was founded with a vision to realize the potential of precision medicine in order to improve the human condition. Headquartered in Irving, Texas, Caris has offices in Phoenix, New York, Cambridge (MA), Tokyo, Japan and Basel, Switzerland. Caris or its distributor partners provide services in the U.S. and other international markets. To learn more, please visit CarisLifeSciences.com. Caris Life Sciences Media: Corporate Communications [email protected] 214.294.5606 SOURCE Caris Life Sciences WANT YOUR COMPANY'S NEWS FEATURED ON PRNEWSWIRE.COM? 440k+ Newsrooms & Influencers 9k+ Digital Media Outlets 270k+ Journalists Opted In GET STARTED

2025年2月26日，来自罗氏的研究人员在Drug Discovery Today上发表文章Embracing the future of medicine with virtual patients，阐述了虚拟患者技术如何变革医疗研发，并通过研发价值链中的实际案例展示其整合应用的变革潜力。 虚拟患者技术正通过整合多维度数据与AI建模，重构医疗健康生态。该技术通过采集患者个体生理参数、基因特征及生活方式数据，构建动态数字孪生体，实现个性化治疗方案的精准模拟。在患者层面，虚拟患者可预测药物疗效与副作用，支持以患者为中心的精准用药策略，提升治疗有效性与安全性。 社会层面，虚拟患者通过构建多样化群体仿真模型，推动包容性临床试验，减少对动物实验的依赖，同时提升药物可及性，助力健康公平。其应用可降低药物研发成本，缩短周期，通过预测模型优化剂量方案，减少药物相关死亡率。 制药领域，虚拟患者整合器官芯片、类器官技术及真实世界数据，革新研发流程。结合AI算法与电子健康记录，实现药物靶点验证、临床试验设计及疗效预测的智能化，加速候选药物筛选与审批进程。监管机构正积极制定政策框架，美国FDA与欧盟EMA已启动虚拟患者在药物开发中的合规性探索，标志着技术转化的关键突破。 随着技术迭代，虚拟患者将深度融入临床决策支持系统，推动精准医学发展，优化资源配置，最终实现从经验医疗向数据驱动医疗的范式转变。 理解数字孪生与虚拟患者的区别 “虚拟患者”与“数字孪生”常被混用，但二者范围不同。虚拟患者是基于群体数据参数分布构建的仿真模型，代表具有生物学合理特征的假设患者，用于药物开发和监管决策中模拟不同患者对治疗的反应。这些仿真通过优化药物剂量、疗效与安全性，帮助研究者探索多种场景而无需大量真实试验。 数字孪生则是基于个体患者详细数据生成的个性化模型，通过持续更新数据提供精准治疗策略。本文主要讨论广义的虚拟患者，但具体案例中会提及数字孪生。 虚拟患者推动制药研发的演进 虚拟患者越来越多地用于药物发现与开发，以模拟药物行为、预测疗效并加速进程。通过生成式AI、机制模型和先进体外系统，虚拟患者可建模从细胞到全身的系统，助力疾病理解、生物标志物发现及药物开发。其应用包括加速药物开发、预测患者反应及支持个性化医疗。 通过在虚拟空间中测试多种可能性，虚拟患者允许更多实验以虚拟方式完成，随后在实际研发中验证结果并反馈迭代。这一过程提高了选择正确分子、优化临床试验患者筛选及减少传统试错法的概率。 图1 通过虚拟患者与数字孪生加强药物发现与开发 在加速研发的背景下，数字孪生通过提供个体患者实时响应的高精度模拟补充虚拟患者。虚拟患者基于群体数据模拟广泛反应，而数字孪生通过真实个体数据细化模型，从而提升虚拟临床试验的预测能力。构建有效的数字孪生需长期监测以捕捉疾病进展，利用电子健康记录（EHR）、患者注册库及历史临床试验的真实世界证据（RWE）提供基线数据。 虚拟患者通过详细模拟优化临床试验招募的定位与分层，数字孪生则通过实时健康数据维护个体化档案，将患者匹配至相关试验。 数字孪生在个性化医疗中通过持续监测患者反应并实时调整治疗方案展现显著优势，其动态模型能反映当前健康状态，从而提升疗效并减少不良反应。 对研发效率的潜在影响 虚拟患者技术可通过提升技术成功概率（PTS）、降低成本并缩短周期，彻底改变药物开发流程。 表1 虚拟患者技术对技术成功概率（PTS）、成本及研发周期时间的潜在影响示例 值得注意的是，AI发现的药物在I期试验中成功率显著高于传统方法。近期分析显示，AI药物的I期成功率达80–90%，远高于行业历史平均值（40–65%）。若后期试验保持此趋势，制药行业的开发能力可能翻倍。安全性是关键，但AI药物的I期成功也归因于优化的ADME（吸收、分布、代谢、排泄）特性及初步疗效。 尽管虚拟患者潜力巨大且已有成功案例，其系统化应用尚未普及。本文通过现有案例探讨其如何改变药物开发流程。 虚拟患者技术的应用案例 虚拟患者技术显著提升药物研发价值链的多个环节，包括靶点识别、患者特异性治疗计划、药物相互作用（DDI）模拟、患者反应预测及临床结果预测，从而优化决策与开发流程。 图2 展示虚拟患者技术优势的应用案例 当前在药物发现中的应用 虚拟患者技术通过整合人工智能与多组学数据，在药物研发早期阶段展现出显著优势。在靶点发现环节，其基于患者特征的预测模型可加速阿尔茨海默病等复杂疾病的靶点验证，通过个体层面的基因 / 蛋白敲除模拟，已成功识别疾病驱动基因并优化药物作用通路，显著降低 II 期临床试验失败率（传统研发痛点）。基于生理药代动力学（PBPK）模型的虚拟评估系统，可快速预测药物体内行为，将候选药物筛选效率提升40%以上，同时减少60%的动物实验需求。 在转化医学领域，芯片肺等体外系统结合虚拟患者建模，实现了呼吸系统疾病的机制研究突破；患者源性微型结肠模型则为结直肠癌微环境的长期动态解析提供了高分辨率平台。最新研究表明，机制模型与虚拟患者的协同应用可优化实验设计参数，使药代动力学（PK）和药物相互作用（DDI）预测精度提升35%，为创新疗法的临床转化奠定关键基础。这些技术的整合应用正推动药物研发向精准化、高效化方向加速演进。 当前在临床试验中的应用 实现精准剂量与疗程优化：在II/III期研究中优化药物剂量和疗程是开发关键。虚拟患者通过整合临床前数据、生物标志物及早期临床数据，可预测药物安全性、有效性及药代动力学（PK），助力制定更可靠的给药策略。例如在fazpilodemab治疗系统性红斑狼疮的研究中，通过动态模拟和马尔可夫模型整合群体PK与不良事件数据，使剂量选择精度提升40%，显著减少试错成本。 个体化给药方案：1）液体活检整合：Rostami-Hodjegan团队通过25例肾功能差异患者的血浆外泌体检测，结合PBPK模型建立虚拟孪生，使个体化给药的暴露变异性降低64%（五倍对比十四倍）。2）肝硬化患者剂量调整：基于PBPK的虚拟患者模型在6种药物评估中，对肝硬化患者的PK预测精度达92%，为脆弱人群用药安全提供新范式。 个性化安全评估：1）心脏毒性预测：西酞普兰（citalopram）的定量系统毒理学模型通过整合hERG通道抑制数据，使心脏毒性预测准确率提升35%。2）细胞因子释放综合征预警：glofitamab治疗DLBCL的预测模型以94%真阳性率识别高危患者，结合PBPK模型实现CRS相关DDI风险量化。 特殊人群模拟创新：1）儿童用药开发：奥司他韦婴儿适应症获批中，PBPK模型替代幼年狨猴重复给药毒性试验，节省12个月研发时间 ；2）罕见病外推：entrectinib儿科模型成功预测CYP3A4抑制剂联用暴露变化，支持无需额外临床试验的监管批 肿瘤试验效率革命：在mosunetuzumab I期试验中，基于QSP的数字孪生技术揭示药物暴露与肿瘤缓解率的剂量-效应关系；发现肿瘤大小、增殖率及T细胞浸润等应答生物标志物；使剂量优化决策效率提升50%。 临床决策支持系统：1）kML模型：结合治疗中血液标志物动态与机器学习，使晚期NSCLC患者生存预测精度提升28%。2）ROPRO评分：整合27个变量的多因素预后模型，使临床试验患者筛选效率提高40%，样本量需求减少30%。 减少对照组规模：Unlearn AI与强生合作显示，数字孪生可将阿尔茨海默病III期试验对照组规模减少33%。 超越药物的解决方案 以连续血糖监测（CGM）为例，数字孪生整合患者血糖、胰岛素剂量与饮食数据，可提前预测高/低血糖事件并发出预警。在COVID-19期间，基于智能手机的远程精准剂量系统（如PERSONAL-CovidBP试验）实现了高血压患者的个性化用药，减少了线下就诊需求。 实施虚拟患者的推动因素及挑战 虚拟患者技术在医疗领域的应用前景广阔，但仍面临多重技术与实施挑战 。 患者异质性建模 药物反应因个体生理差异而异，模型需精准捕捉生物过程复杂性及药效/安全/药代动力学的动态变化。结合体外模型与液体活检等技术，可量化患者异质性，提升预测精度。 模型验证体系 通过回顾性分析验证模型预测能力是临床转化的核心。对比历史数据（如DDI预测）可优化模型参数，建立与监管机构的信任基础，加速审批流程。 数据治理困境 高质量多源数据（EHR、可穿戴设备等）是建模基础，但数据孤岛、隐私保护及标准化缺失制约发展。需构建跨机构协作机制，采用联邦学习、区块链等技术保障数据安全共享。 监管与伦理挑战 需通过真实世界证据（RWE）证明临床价值，同时符合GDPR等法规要求。伦理层面需关注算法偏见（如特殊人群代表性不足），建立动态伦理审查框架。 技术生态构建 跨学科协作（生物学家+工程师+临床医生）与数据互操作性是关键。开发标准化接口与开源工具包，降低技术应用门槛，推动从研发到临床的全链条渗透。 未来发展方向 随着AI与数字孪生技术的融合，虚拟患者将向实时动态预测演进。通过整合多组学数据与强化学习，其在个体化治疗、罕见病研究及公共卫生决策中的价值将持续释放。 结论 虚拟患者技术通过模拟患者特异性场景，正在革新医疗保健与制药研发。其优势包括提升个性化治疗、优化临床试验、减少动物实验及促进包容性研究。尽管面临数据、隐私与法规挑战，跨学科协作与技术创新将推动其迈向更高效、以患者为中心的未来。 参考资料： https://doi.org/10.1016/j.drudis.2025.104322 --------- End --------- 感兴趣的读者，可以添加小邦微信加入读者实名讨论微信群。添加时请主动注明姓名-企业-职位/岗位或姓名-学校-职务/研究方向。