1. 研究背景与多参数建模的核心价值
1.1 晚期肺癌免疫治疗的临床现状与疗效预测需求
在过去的十年中，肺癌治疗领域发生了重大范式转变，特别是随着对肺癌生物学理解的加深，靶向治疗和免疫治疗取得了显著进展 1。免疫检查点抑制剂（ICIs），如针对程序性死亡受体1（PD-1）/程序性死亡配体1（PD-L1）的抑制剂，已在晚期非小细胞肺癌（NSCLC）的治疗中展现出巨大的临床效益 1。目前，ICIs已被广泛应用于转移性疾病的一线治疗、不可切除局部晚期疾病放化疗后的巩固治疗，以及可切除疾病手术切除和化疗后的辅助治疗 1。例如，帕博利珠单抗（Pembrolizumab）联合铂类化疗被认为是晚期NSCLC最适合的一线免疫检查点抑制剂治疗方案 2。此外，PD-1/PD-L1抑制剂联合细胞毒性T淋巴细胞相关抗原4（CTLA-4）抗体在晚期肺癌患者中也显示出显著改善客观缓解率（ORR）、总生存期（OS）和无进展生存期（PFS）的效果 3。
然而，尽管ICIs带来了革命性的治疗突破，其疗效在患者个体间存在显著异质性 1。并非所有患者都能从免疫治疗中获益，部分患者甚至可能出现疾病进展 4。例如，约60%的广泛期小细胞肺癌（ES-SCLC）患者在治疗3个月内会复发 5。这种疗效的不可预测性对临床实践构成了巨大挑战，亟需开发更精准的预测生物标志物来筛选出最可能响应治疗的患者，从而实现个体化精准医疗 16。传统的单一生物标志物，如PD-L1表达和肿瘤突变负荷（TMB），虽然已被广泛研究用于患者选择，但它们并非完美无缺，仍存在局限性 47。因此，探索更全面、更准确的疗效预测方法，以指导临床决策并优化患者管理，是当前晚期肺癌免疫治疗领域亟待解决的关键问题。
1.2 多参数建模的定义与预测优势
面对免疫治疗疗效的高度异质性及单一生物标志物的局限性，多参数建模应运而生，并日益成为预测晚期肺癌免疫治疗疗效的重要方向。多参数建模是指通过整合来自多个维度的数据，包括生物标志物（如基因组学、蛋白质组学）、临床特征（如患者人口学信息、病理分期、既往治疗史）以及影像组学数据（如CT、PET-CT图像特征），来构建预测模型的方法。这种方法的核心优势在于其能够捕捉疾病的复杂性和异质性，远超任何单一指标所能提供的信息量。
相较于单指标预测，多参数建模具有显著的优势。首先，它能大幅提升预测的准确性和稳健性。单一生物标志物往往只能反映肿瘤生物学某一方面的特征，而多参数模型能够综合考虑多个独立或协同作用的因素，从而更全面地刻画肿瘤的复杂生物学行为和宿主免疫状态，提高预测的灵敏度和特异性 89。例如，在预测特定疾病风险或复发时，多变量模型已被证明比单变量方法具有更高的准确性 10。其次，多参数建模有助于覆盖更广泛的个体差异。肺癌患者在基因突变、免疫微环境、临床表现以及对治疗的反应上存在巨大差异，单一指标难以充分反映这些复杂的相互作用。通过整合多维度数据，模型能够识别出传统上被忽视的、但对疗效有重要影响的潜在关联，从而更好地为不同亚组患者提供个性化的预测 1112。
此外，多参数建模能够提供更具临床实用性的风险分层。例如，在预测血栓风险时，简单的多变量模型就能有效识别出短期内具有较高症状性VTE风险的患者群体 13。在肺癌免疫治疗领域，这意味着可以更精准地识别出高应答率、低应答率或可能发生严重不良事件的患者，从而优化治疗方案，避免不必要的治疗，并减少副作用。尽管机器学习模型在理论上可能表现出更优的性能，但有研究指出，在某些临床预测模型中，机器学习与传统逻辑回归模型在性能上并未显示出显著差异，这提示了在不同场景下选择合适模型的复杂性 14。然而，随着深度学习等先进算法的发展，多模态数据融合的能力不断增强，有望进一步释放多参数建模在精准预测方面的潜力 1516。通过系统性地集成和分析这些异构数据，多参数建模为实现晚期肺癌免疫治疗的精准预测和个体化管理提供了强大的工具。
2. 多参数建模的关键参数类型与特征挖掘
2.1 生物标志物参数：分子特征与免疫微环境
生物标志物在预测晚期肺癌免疫治疗疗效中扮演着核心角色，特别是分子特征和免疫微环境相关指标。它们能够从肿瘤细胞的内在特性和宿主免疫系统的相互作用两方面揭示患者对免疫治疗的潜在反应。
PD-L1表达水平是目前临床上最广泛使用的免疫治疗预测标志物之一。PD-L1在肿瘤细胞或免疫细胞上的高表达通常预示着患者对PD-1/PD-L1抑制剂有更好的反应 17。PD-L1的检测通常通过免疫组织化学（IHC）方法进行，其表达水平被划分为不同阈值（如≥1%或≥50%），以指导临床用药。例如，在一项关于新辅助免疫治疗的荟萃分析中，PD-L1表达者（≥1%）更可能获得病理完全缓解（pCR）17。然而，PD-L1表达的异质性、检测方法（抗体克隆、平台、判读标准）的差异以及动态变化使其作为单一预测指标仍存在局限性 18。部分PD-L1低表达甚至阴性的患者也能从免疫治疗中获益，而部分高表达患者却无应答，这提示了需要结合其他标志物进行更全面的评估。
肿瘤突变负荷（TMB）是另一个重要的预测生物标志物，指肿瘤基因组中体细胞非同义突变的总数。高TMB通常意味着肿瘤细胞会产生更多的新抗原，从而更容易被免疫系统识别和攻击，进而增强对免疫检查点抑制剂的敏感性 19。研究表明，TMB与多种癌症（包括肺癌）患者接受免疫治疗后的生存期改善相关 19。特别是，高TMB与PD-1联合CTLA-4阻断治疗在非小细胞肺癌患者中观察到的客观缓解、持久获益和无进展生存期改善密切相关，且TMB的预测作用独立于PD-L1表达 20。然而，不同癌种间TMB的预测阈值差异显著，且其检测标准化仍面临挑战 1921。
微卫星不稳定性（MSI-H）/错配修复缺陷（dMMR）也是重要的预测标志物，尤其在结直肠癌中。在肺癌中，MSI-H/dMMR可以作为独立的免疫治疗预测生物标志物 18。MSI-H肿瘤由于DNA错配修复功能缺陷，导致基因组高度不稳定，从而产生大量突变和新抗原，激活抗肿瘤免疫反应 22。循环肿瘤DNA（ctDNA）检测MSI-H已成为一种快速可靠的非侵入性方法，且结果显示ctDNA检测到的MSI-H患者接受免疫治疗后无进展生存期显著延长 23。
肿瘤浸润淋巴细胞（TILs）反映了肿瘤微环境中的免疫细胞浸润状态。肿瘤浸润淋巴细胞（TILs）疗法在肺癌治疗中具有巨大的潜力 24，且高密度的TILs，特别是细胞毒性T淋巴细胞（CD8+ T细胞），通常与更好的免疫治疗疗效相关。TILs的数量、表型和功能状态共同决定了肿瘤的“免疫原性”或“冷热”状态。高LMS评分的肺鳞癌患者群体更倾向于“冷”肿瘤，表现为免疫抑制和免疫排斥，而低LMS评分的患者则有更好的免疫治疗反应 25。此外，外周血淋巴细胞亚群水平（PBLSL）的异质性也能够预测晚期肺癌患者对放化疗或免疫放化疗的疗效。例如，较高水平的B细胞、较高的CD4+/CD8+ T细胞比值以及较低的CD8+ T细胞水平与更好的短期疗效相关 26。
多生物标志物联合建模是提高疗效预测准确性的关键方向。由于单一标志物的局限性，研究者正致力于整合多个生物标志物来构建更全面的预测模型。例如，有研究通过整合总新抗原负荷、代谢评分、CD8 T细胞和MHC评分等多个组学数据，构建了“免疫治疗长期生存预测评分”（iLSPS）的高斯过程回归模型，该模型在预测黑色素瘤和肺癌患者的长期生存方面优于单个生物标志物 27。另一个例子是，SMARCA4突变与低表达在肺腺癌中预示不良预后，但其对免疫治疗的敏感性取决于共突变基因，如与TP53共突变时免疫治疗效果更佳，而与KRAS或STK11共突变时效果较差，这表明分子标志物之间的相互作用需要被纳入考量 28。此外，抗移码肽抗体作为一类基于血液的新型生物标志物，通过评估血清中识别肿瘤细胞产生的变异肽段的抗体水平，有望同时预测肿瘤应答和免疫相关不良事件，准确率高达98% 29。这些研究案例都强调了多生物标志物联合建模在揭示复杂生物学机制和提高预测精度方面的巨大潜力。
2.2 临床特征参数：基线状态与治疗相关指标
在晚期肺癌免疫治疗疗效预测的多参数建模中，除了复杂的生物标志物外，一系列可易于获取的临床特征参数也发挥着不可忽视的作用。这些参数反映了患者的基线状态、疾病负担以及既往治疗情况，它们的筛选和纳入能够有效提升模型的实用性和预测能力。
患者年龄是重要的临床参数之一。尽管免疫治疗在老年患者中具有一定的耐受性，但年龄对疗效的影响仍需考量。例如，一项针对广泛期小细胞肺癌（ES-SCLC）老年患者（≥70岁）的真实世界研究显示，免疫联合化疗组与单纯化疗组相比，无进展生存期（PFS）显著延长（8.26个月 vs 6.59个月，P=0.02），但总生存期（OS）差异无统计学意义（14.20个月 vs 11.44个月，P=0.14）30。这表明在特定亚组中，年龄可能影响治疗效果的某些方面，模型在设计时应考虑其潜在的权重。
ECOG体力状态评分（ECOG PS）是评估患者日常活动能力和生活质量的重要指标，也是预测治疗耐受性和疗效的基石。ECOG PS评分越高，通常预示着患者的预后越差，对治疗的耐受性也可能较差。在晚期NSCLC的预测模型中，ECOG PS常常被纳入作为重要的协变量。例如，在IMpower150队列衍生的模型中，ECOG PS评分可以作为评估临床价值的关键指标之一，帮助区分“统计学显著”与“临床可操作性”之间的差异。一项关于ES-SCLC患者的研究显示，Karnofsky体力状态评分（KPS）≥80分与OS和PFS的显著改善相关31，这提示了良好体力状态对疗效的积极影响。另一项研究则指出，对于ECOG评分为0的NSCLC患者，Nivolumab联合Ipilimumab和化疗（Niv+Ipi+CT）可带来最佳的OS获益，而ECOG评分为1的患者，Pembrolizumab联合化疗（Pem+CT）可能最为有效32。这些发现强调了ECOG评分在患者分层和个体化治疗选择中的指导作用。
吸烟史是肺癌发病的重要危险因素，也可能影响免疫治疗的疗效。吸烟导致的基因突变累积可能增加肿瘤新抗原的产生，从而影响免疫治疗的反应。有研究表明，吸烟史的患者可能对某些免疫治疗方案有更好的响应，或在某些生物标志物组合中表现出差异化的预测价值。例如，在一项大型临床试验和真实世界数据分析中，PD-L1作为预测生物标志物对吸烟史患者的预测效果优于非吸烟者33。此外，有研究发现，对于有吸烟史的男性患者，Camrelizumab联合化疗（Cam+CT）在PFS方面提供了最佳效益32。相反，Pem+CT在女性和非吸烟患者亚组中表现出最佳的PFS和OS效益32。这些都说明了吸烟史在多参数模型中的重要贡献。
既往治疗方案，特别是化疗或靶向治疗，对后续免疫治疗的疗效预测至关重要。患者是否接受过化疗、靶向治疗，以及这些治疗的疗效和毒性，都会影响免疫系统的状态和肿瘤的生物学特性。例如，有研究发现既往放疗可以增强派姆单抗（Pembrolizumab）在非小细胞肺癌中的抗肿瘤活性，并延长无进展生存期和总生存期，且安全性可接受34。此外，化疗诱导的免疫原性细胞死亡可以增强抗肿瘤免疫反应，提高对免疫治疗的敏感性35。在特定基因突变（如KRAS G12C突变）的NSCLC患者中，既往接受过铂类化疗和PD-1或PD-L1抑制剂治疗后疾病进展的患者，Sotorasib相较于多西他赛能显著延长PFS，并具有更好的安全性36。这些信息在多参数模型中可以作为重要的输入变量，用于更精确地评估患者对免疫治疗的潜在反应。
其他基线血清学和临床指标如乳酸脱氢酶（LDH）、C反应蛋白（CRP）、白蛋白、血红蛋白以及肝转移状态等也常被纳入预测模型。高LDH和CRP水平通常提示较高的肿瘤负荷、炎症反应或不良预后，这些因素可能削弱免疫治疗的效果37 38。一项预测晚期SCLC患者免疫化疗预后的模型中，体重指数（BMI）、肝转移、红细胞分布宽度变异系数、LDH、白蛋白和CRP均被筛选为重要的预测因子，并在训练集和验证集中展现出良好的预测性能（1年OS的AUC值分别为0.760、0.732和0.722）37。另一项针对NSCLC患者免疫治疗预后的研究也发现，较高的LDH和中性粒细胞-淋巴细胞比值（NLR）与较差的PFS和OS相关39。在老年ES-SCLC患者中，基线无肝转移和女性患者是OS的有利预后因素，且对于有基线脑转移或无肝转移的患者，免疫治疗能提供生存获益30。这些实证数据充分说明了将这些易于获取的临床和血清参数整合到多参数模型中的价值。
综上所述，临床特征参数，如年龄、ECOG PS、吸烟史、既往治疗以及各种血液生化指标，为多参数建模提供了丰富且易于获取的信息。通过科学的筛选逻辑和加权贡献，这些参数与生物标志物相结合，可以构建出更全面、更具临床指导意义的预测模型，从而优化晚期肺癌免疫治疗的个体化决策。
2.3 影像组学参数：定量影像特征的提取与验证
影像组学（Radiomics）作为一种新兴的生物标志物发现方法，通过从医学影像数据中高通量提取和分析定量特征，将图像信息转化为可挖掘的高维数据，从而为疾病诊断、预后评估和疗效预测提供支持 40。在晚期肺癌免疫治疗疗效预测中，CT、PET-CT等常规影像技术所获取的图像，结合影像组学分析，能够捕捉肿瘤内部异质性以及肿瘤微环境的复杂性，提供传统形态学评估难以获得的深层信息 4142。
定量影像特征的提取方法：
影像组学特征的提取通常涉及以下几个主要步骤：
- 图像采集与预处理：获取标准化的高质量医学影像（如CT、PET-CT）。预处理包括图像配准、去噪、归一化等，以减少批次效应和提高特征的稳定性。
- 肿瘤区域勾画（Segmentation）：精确地勾画出肿瘤病灶的感兴趣区域（ROI）。这可以由经验丰富的放射科医生手动或半自动完成，也可以通过深度学习算法自动完成，以确保特征提取的准确性。
- 特征提取：从勾画出的ROI中提取大量定量特征。这些特征可分为几大类：
- 形态学特征（Shape features）：描述肿瘤的大小、形状和边缘规则性，如体积、表面积、球形度、扁平度等 4344。
- 一阶统计特征（First-order statistics）：反映ROI内像素强度分布的特征，如均值、中位数、标准差、偏度、峰度等，这些特征与肿瘤的均一性或异质性相关 4344。
- 纹理特征（Texture features）：描述像素强度空间分布的模式和关系，反映肿瘤的内部异质性。常用的纹理矩阵包括灰度共生矩阵（GLCM）、灰度游程长度矩阵（GLRLM）、灰度区域大小矩阵（GLSZM）、邻域灰度差分矩阵（NGTDM）和直方图（Histogram）等。例如，纹理特征可以揭示肿瘤的粗糙度、对比度、均匀性和复杂度等信息 4445。
- 高阶特征（Higher-order features）：通过应用小波变换、拉普拉斯滤波等数学变换后，再提取上述各类特征，以捕捉图像中更精细的、人眼难以察觉的纹理细节 44。
- 代谢参数（Metabolic parameters）：主要来源于PET-CT图像，如标准化摄取值（SUVmax, SUVmean）、代谢肿瘤体积（MTV）和总病灶糖酵解量（TLG）。这些参数反映了肿瘤的代谢活性，与肿瘤的侵袭性和增殖能力相关 434647。
基于AI辅助工具的影像组学模型构建实例：
近年来，随着人工智能（AI）和深度学习技术的发展，影像组学在预测免疫治疗疗效方面取得了显著进展，尤其是在非小细胞肺癌（NSCLC）领域。
- PD-L1表达预测：PD-L1的组织活检具有侵入性且存在肿瘤异质性问题。影像组学和深度学习提供了一种非侵入性预测PD-L1状态的途径。有研究利用深度学习模型（SResCNN）分析PET/CT图像，开发出一种深度学习评分（DLS），能够预测NSCLC患者的PD-L1表达状态，并且该DLS在预测PFS和OS方面的效用与IHC检测的PD-L1状态无异，提示其有望作为IHC的替代或补充手段 48。另一项研究则利用多模态CT图像（增强CT和非增强CT）提取深度学习特征，构建了名为LUNAI-fCT的模型，在预测NSCLC新辅助免疫化疗后的病理完全缓解（pCR）方面表现出优异性能（AUC达0.866） 49。还有研究利用PET/CT影像组学开发了基于SVM算法的模型，预测NSCLC患者的PD-L1表达，并在PD-L1阳性患者中预测PFS和OS，结果显示了良好的预测性能 50。此外，针对脑转移瘤的PD-L1表达，影像组学分类器也能实现高精度的非侵入性预测 51。
- 疗效和预后预测：除了PD-L1状态，影像组学还能直接预测患者对免疫治疗的响应和生存预后。一项研究通过对基线FDG-PET/CT图像的分析，发现年龄、临床IV期、淋巴管侵犯特征、原发肿瘤体积（PT volume）、SUVmax和TLG是NSCLC患者接受ICI治疗后长期预后的强预测因子 43。另一项研究构建了CT影像组学模型，用于预测TMB水平和免疫治疗疗效，发现特定影像组学特征与TMB相关，且能预测免疫治疗效果 52。此外，短期随访CT图像的影像组学特征（delta-radiomics）被证明可以显著提高对晚期NSCLC免疫治疗疗效的预测性能，尤其在腺癌患者中表现更优，并对PFS和OS具有预后价值 53。
- 多模态融合：将影像组学特征与临床特征、生物标志物相结合的多模态模型，往往能提供更全面的预测。例如，一个结合了PET/CT代谢参数和血液炎症标志物的列线图模型，在预测晚期NSCLC一线化疗联合免疫治疗患者的PFS和OS方面展现出良好性能 47。有研究利用深度学习对基线CT图像进行分析，构建了Deep-CT模型，能够独立于传统临床病理生物标志物预测NSCLC患者的总生存期和无进展生存期，且与传统风险因素结合后预测性能显著提升 54。AI方法也被应用于发现新的影像组学和病理组学生物标志物，并且通过整合多模态/多组学数据，正在形成复杂的基于AI的新型标志物 40。
- 免疫相关不良事件预测：影像组学不仅可以预测疗效，还能预测免疫治疗相关的毒副作用。例如，对于免疫检查点抑制剂相关的肺炎（ICIP），影像组学已被用于预测和识别，以期将临床定性问题转化为定量问题，从而改善ICIP的诊断和治疗 55。
- 肿瘤异质性量化：肿瘤内部异质性是影响治疗响应的关键因素。通过CT图像定量分析肿瘤内部异质性（如使用K-means非监督聚类方法划分肿瘤亚区域），可以构建异质性栖息地模型，在预测NSCLC新辅助免疫化疗后的病理完全缓解方面优于传统影像组学模型，为个体化治疗提供了潜在依据 45。
总的来说，影像组学通过从常规医学影像中提取定量特征，并结合AI和深度学习算法，为晚期肺癌免疫治疗疗效的精准预测提供了强大的非侵入性工具。它能够反映肿瘤的生物学特性和微环境状态，辅助临床决策，实现更精准的患者分层和个体化治疗 41。
3. 多参数建模的方法学进展与模型性能对比
3.1 传统统计模型：逻辑回归与Cox比例风险模型
在多参数建模预测晚期肺癌免疫治疗疗效的早期研究中，传统统计模型如逻辑回归（Logistic Regression）和Cox比例风险模型（Cox Proportional Hazards Model）扮演了基石性的角色。这些模型因其原理相对简单、结果易于解释而广泛应用于临床预测模型的构建。
逻辑回归模型主要用于预测二分类结果，例如患者是否对免疫治疗产生客观缓解（客观缓解 vs 无客观缓解），或疾病是否控制（疾病控制 vs 疾病进展）。它通过构建一个线性组合的预测变量，然后将其通过一个逻辑函数（Sigmoid函数）转换为介于0到1之间的概率值。在多参数整合中，逻辑回归能够将多个临床、生物或影像特征纳入模型，评估每个参数对疗效预测的独立贡献。例如，有研究利用逻辑回归分析结合影像组学特征和临床数据，开发出预测可切除非小细胞肺癌新辅助免疫化疗主要病理反应（MPR）的列线图，取得了良好的判别性能（训练集和验证集AUC分别为0.84和0.81），并被认为具有临床应用价值 56。
Cox比例风险模型则主要用于生存分析，预测患者的无进展生存期（PFS）或总生存期（OS）。该模型能够处理右删失数据（即在研究结束时部分患者尚未发生事件），并估计协变量对事件发生风险（风险比，Hazard Ratio, HR）的影响。在多参数预测中，Cox模型可以整合患者的基线特征、生物标志物数据等，以评估它们对患者生存结局的独立预测价值。例如，在一项针对非小细胞肺癌免疫治疗疗效的多维度预测模型研究中，研究者利用最小绝对收缩与选择算子（LASSO）Cox回归分析筛选出影像组学特征和免疫特征，并构建了多维度列线图。结果显示，该多维度模型在预测PFS和OS方面优于任何单一模型，C指数分别为0.721和0.727，AUC值也更高，证实了多参数整合的优势 57。此外，有研究使用多变量Cox模型来识别与晚期NSCLC患者1年总生存期相关的临床因素，并以此为基础开发了预测模型 58。
传统统计模型在多参数整合中的应用场景：
- 小样本数据：当数据集相对较小，难以支持复杂机器学习模型时，逻辑回归和Cox模型是更稳健的选择。它们对样本量的要求相对较低，且不易过拟合。
- 变量筛选和解释：这些模型能够清晰地量化每个预测变量对结果的影响（例如逻辑回归的OR值，Cox模型的HR值），这对于临床医生理解模型的预测逻辑和变量的重要性至关重要。研究者可以通过这些模型的系数来判断哪些参数是疗效的关键驱动因素。例如，在膀胱癌患者的预后预测中，虽然多参数模型纳入了多种生物标志物，但研究发现，即使是最佳的生物标志物组合，其对模型判别能力的提升也相对有限，这提示了在变量选择时需要权衡其临床实际意义 59。
- 风险分层：通过模型输出的概率或风险评分，患者可以被划分为不同风险等级，从而指导后续的治疗策略或随访方案。
局限性：
尽管传统统计模型在多参数建模中具有其优势，但也存在显著局限性：
- 线性假设：逻辑回归和Cox模型均基于变量之间存在线性关系的假设（或通过数学转换可线性化）。然而，生物医学数据中变量间的关系往往是非线性的，且存在复杂的交互作用，传统模型难以有效捕捉这些非线性关系。
- 处理高维数据的能力有限：当需要整合的参数数量非常多（即高维数据），尤其是涉及影像组学或基因组学等数据时，传统模型的性能会受到限制，容易出现多重共线性问题，导致模型不稳定或过拟合。尽管逐步回归等变量筛选方法可以尝试缓解，但其本质上仍是基于线性假设，并且可能遗漏重要的非线性交互作用。
- 对数据分布的严格要求：Cox比例风险模型要求协变量对风险的效应是随时间恒定的，即“比例风险假设”。当这一假设不成立时，模型的结果可能不准确。
- 难以处理异构数据：传统模型在整合不同类型数据（如连续型、离散型、分类型数据）时相对简单，但对于结构化程度低、非数值型的复杂数据（如原始医学图像、文本信息）的处理能力较弱。
因此，为了克服这些局限性，研究者们开始转向机器学习和深度学习等更先进的方法，以更好地处理复杂、高维和异构的多参数数据，从而构建更精准和鲁棒的预测模型。
3.2 机器学习模型：随机森林与深度学习的突破
随着数据量的爆炸式增长和计算能力的提升，机器学习（Machine Learning, ML）和深度学习（Deep Learning, DL）模型在多参数建模预测晚期肺癌免疫治疗疗效方面展现出前所未有的潜力。这些模型能够处理高维复杂数据，自动发现数据中的非线性关系和隐藏模式，从而克服了传统统计模型的局限性。
随机森林（Random Forest）和XGBoost等树模型在处理高维参数时的优势：
随机森林和梯度提升树（如XGBoost）是集成学习方法中的代表，它们通过构建大量决策树并进行集成来提高预测的准确性和鲁棒性。
- 处理高维数据和特征交互：这类模型能够有效处理包含大量特征的高维数据，而无需进行严格的特征选择或降维。它们能够自动捕获特征之间的非线性关系和高阶交互作用，这对于复杂生物学过程的建模尤为重要。例如，在分析基因表达数据时，这些模型可以识别出基因之间的复杂调控网络，而不仅仅是单个基因的影响 60。
- 对缺失数据和异常值的鲁棒性：树模型对输入数据中的缺失值和异常值不敏感，不需要进行复杂的预处理，这在临床真实世界数据中非常实用。
- 特征重要性评估：随机森林和XGBoost等模型可以输出特征重要性分数，帮助研究人员理解哪些参数对预测结果贡献最大，从而为临床决策提供可解释的依据。
- 非线性决策边界：与传统统计模型（如逻辑回归）的线性决策边界不同，树模型能够构建复杂的非线性决策边界，更好地拟合数据的内在结构，从而提高预测精度。一项研究利用随机森林模型结合CT影像组学特征和临床病理特征，成功预测了非小细胞肺癌患者免疫检查点抑制剂的临床获益，其AUC值在训练集和验证集分别达到了0.848和0.795 61。另一项研究则利用随机森林模型预测肺癌化疗反应，并指出其在分类和识别重要特征方面的优势 60。此外，在针对预测放射治疗联合免疫治疗引起症状性肺炎的研究中，基于随机森林的模型虽然表现不如深度学习模型，但也为多参数融合提供了初步探索 62。
深度学习（如卷积神经网络结合影像组学）在多模态数据融合中的性能提升案例：
深度学习，特别是卷积神经网络（Convolutional Neural Networks, CNNs），在处理图像数据方面具有无与伦比的优势。将CNNs与影像组学相结合，并进一步与其他模态数据融合，已经成为晚期肺癌免疫治疗疗效预测的热点方向。
- 自动特征提取：与传统的影像组学需要人工设计和提取特征不同，CNNs能够自动从原始图像中学习和提取层次化的、对预测任务最有用的特征，避免了特征工程的复杂性和主观性 63。这使得深度学习模型能够捕捉到人眼难以察觉的细微图像模式。
- 多模态数据融合：深度学习模型特别擅长处理和融合不同类型的数据模态，如影像数据（CT、PET-CT）、基因组学数据、病理图像和临床数据。通过不同的融合策略（如早期融合、晚期融合），深度学习可以学习不同模态之间的复杂关联，从而提升预测性能。
- 影像-生物标志物融合：一项研究提出了一种基于深度学习的多模态特征融合模型（MFMDL），该模型利用图神经网络将基因网络中的基因-基因关系映射到低维向量空间，然后融合生物通路特征和免疫细胞浸润特征来预测免疫治疗反应。该模型在包括肺癌在内的多种癌症类型上均优于传统生物标志物 64。
- 影像-临床融合：在预测肺癌放射治疗联合免疫治疗引起的症状性肺炎（SP）时，有研究开发了一种多模态融合模型，该模型整合了深度图像特征、影像组学特征和临床数据。结果显示，该模型的AUC达到了0.922，显著优于仅使用影像组学特征（AUC 0.811）或临床信息（AUC 0.711）的模型，甚至优于传统的基于DNN的omix属性和临床数据整合模型（AUC 0.872） 62。这充分展示了深度学习在多模态融合方面的强大能力。
- PET/CT融合预测PD-L1状态：PD-L1表达是免疫治疗的重要生物标志物，但其检测具有侵入性。研究发现，将PET和CT图像作为输入进行早期融合的深度学习模型，在自动预测非小细胞肺癌PD-L1状态方面表现出最佳性能，明显优于单一模态或晚期融合模型 65。这表明，在低层级融合不同模态的信息，可以捕获更复杂的模式，从而提高预测准确性。
- 病理图像与临床/影像组学融合：在结直肠癌肺转移患者术后预后预测中，一项研究构建了一个结合深度学习病理组学（从全玻片图像中提取特征）、影像组学、免疫评分和临床因素的列线图模型，在预测总生存期（OS）和无病生存期（DFS）方面表现出色（AUC分别为0.860和0.875） 66。这说明深度学习在病理图像分析中的应用，结合其他参数，能够显著提升预后预测能力。
- 弱监督深度学习预测病理缓解：在预测非小细胞肺癌新辅助化疗免疫治疗后的主要病理缓解（MPR）方面，研究者应用弱监督深度学习技术分析术前全玻片图像（WSIs）。通过DenseNet121等先进架构在微观（斑块）和宏观（玻片）层面分析WSIs，并结合创新的数据增强和归一化技术，模型在验证集和外部测试集中均表现出良好的性能，并且通过与临床数据整合，进一步提升了鉴别准确性 67。这强调了深度学习在处理复杂病理图像数据方面的优势。
综上所述，随机森林和XGBoost等树模型在高维参数处理和特征交互建模方面具有独特优势，而深度学习模型，特别是CNNs，则通过自动特征提取和强大的多模态数据融合能力，在提升晚期肺癌免疫治疗疗效预测精度方面取得了突破性进展。它们能够更全面、更深入地挖掘复杂生物学数据中的信息，为实现精准医疗提供了关键技术支撑。
3.3 多模态融合建模：跨维度数据的协同预测
多模态融合建模是多参数建模领域的核心发展方向，旨在通过整合来自不同数据源和维度的信息，实现对晚期肺癌免疫治疗疗效更全面、更精准的预测。这种策略认识到单一模态数据（如基因组学、影像组学或临床数据）往往只能捕捉肿瘤生物学和宿主免疫反应的某个侧面，而将多模态数据进行有效融合，则能揭示更深层次、更复杂的生物学关联，从而显著提升预测模型的性能和临床实用性。
跨模态融合策略：
多模态融合主要有以下几种策略：
- 早期融合（Early Fusion）：在特征提取阶段就将不同模态的原始数据或低级特征进行拼接或合并，形成统一的特征向量，然后输入到模型中进行学习。这种方法在特征层面进行融合，允许模型学习不同模态数据之间的早期交互作用。例如，有研究将PET和CT图像作为输入，进行早期融合的深度学习模型，在自动预测非小细胞肺癌（NSCLC）PD-L1状态方面表现出最佳性能，明显优于单一模态或晚期融合模型 65。这表明在较低层级融合解剖和代谢信息能捕获更复杂的模式，从而提高预测准确性 65。
- 晚期融合（Late Fusion）：也称为决策级融合，即对每种模态数据分别训练独立的模型，然后将各个模型的预测结果（如概率、分类分数）进行组合，得出最终的预测。这种方法的好处是灵活性高，可以针对不同模态选择最适合的模型，但可能无法充分捕捉模态间的深层交互。例如，一项研究尝试了早期融合和晚期融合两种策略，在预测PD-L1状态时，早期融合通常优于晚期融合，这可能与早期融合在较低层面捕获复杂模式的能力有关 65。
- 中间融合（Intermediate Fusion）/混合融合（Hybrid Fusion）：结合了早期和晚期融合的优点，在特征提取的中间阶段进行融合，或通过复杂的神经网络结构设计，在不同层级进行信息交互和融合。例如，利用图神经网络（Graph Neural Networks, GNN）将基因-基因关系映射到低维向量空间，然后融合生物通路特征和免疫细胞浸润特征，构建多模态特征融合深度学习模型（MFMDL） 64。这种模型在多个癌症类型（包括肺癌）的免疫治疗反应预测中，性能优于传统生物标志物，强调了融合不同模态特征的必要性 64。
单模态与多模态模型的性能对比：
多项研究证据表明，多模态融合模型通常能显著优于单一模态模型。这种性能提升体现在AUC（受试者工作特征曲线下面积）、准确率、敏感性、特异性等多种评估指标上：
- 生物-影像融合：在NSCLC中，利用PET/CT影像组学与临床特征相结合的模型，在预测EGFR敏感突变方面表现出良好性能 68。一项研究发现，将PET/CT图像的深层特征（通过ResNet提取）和浅层特征（传统的影像组学特征）融合后，预测EGFR敏感突变的AUC可达0.94，远高于仅使用浅层特征的模型（AUC 0.68），也高于单一CT或PET图像的模型 68。这说明影像特征的深度和广度结合能显著增强预测能力。
- 影像-临床融合：在预测肺癌放疗联合免疫治疗后的症状性肺炎（SP）方面，一项多模态融合模型整合了深度图像特征、影像组学特征和临床数据，其AUC高达0.922 62。这显著优于仅使用影像组学特征的模型（AUC 0.811）、仅使用临床信息的模型（AUC 0.711），以及仅将影像组学与临床数据简单整合的模型（AUC 0.872） 62。这有力证明了多模态数据融合，特别是结合深度学习提取的特征，能够大幅提升预测性能。
- 多组学融合：随着高通量测序技术的发展，基因组学、转录组学、蛋白质组学等多组学数据为多参数建模提供了丰富素材。例如，一项研究整合了基因表达、DNA甲基化、基因组突变和临床数据，通过多组学整合算法识别出与预后相关的LUSC亚型，并构建了基于七个关键基因的LUSC多组学特征（LMS）评分 25。该LMS评分在预测总生存期和免疫治疗反应方面优于已有的LUSC生物标志物，并发现低LMS评分的患者具有更高的总生存率和更好的免疫治疗反应 25。此外，综合多组学数据和机器学习方法，能够构建出更全面的评分系统，预测免疫检查点阻断治疗的有效性，并已验证其优于现有临床生物标志物 69。另一项研究通过整合非同义突变数据，开发了基于基因突变的分类器（GPS），其在预测免疫治疗结果方面也优于传统预测因子 70。
- 多维度数据协同预测复杂疗效模式：肺癌免疫治疗的疗效受到肿瘤异质性、免疫微环境、宿主遗传背景等多种复杂因素的共同影响。多模态融合建模能够：
- 捕捉肿瘤异质性：影像组学可以量化肿瘤内部的纹理和形状异质性，而基因组学则揭示肿瘤的分子异质性。两者的融合可以提供更全面的肿瘤画像。
- 揭示免疫微环境特征：基因表达谱（如免疫基因签名）、T细胞浸润情况（通过病理图像或流式细胞术）以及影像学特征（如淋巴结肿大、肿瘤血管新生模式）可以共同反映肿瘤微环境的免疫状态。
- 识别新的预测生物标志物：通过机器学习算法在融合数据中进行挖掘，有助于发现传统方法难以识别的新型生物标志物或生物标志物组合。例如，有研究通过整合空间多组学分析和虚拟临床试验，成功预测了免疫治疗反应并发现了新的生物标志物 71。
- 提高模型泛化能力：整合多模态数据有助于模型学习更具普遍性的模式，从而提高模型在不同患者队列和临床情境下的泛化能力。
总而言之，多模态融合建模通过汇聚和解析来自不同维度的信息，显著提升了晚期肺癌免疫治疗疗效预测的准确性和稳健性。这种跨维度数据的协同效应，使得模型能够更深入地理解疾病的复杂性，为实现真正的个体化精准治疗提供了强有力的方法学支撑 727374。
4. 模型验证体系与临床转化挑战
4.1 内部与外部验证：数据异质性的应对策略
多参数预测模型在晚期肺癌免疫治疗疗效预测中的价值，最终体现在其在真实临床环境中能否准确、稳定地指导决策。因此，严格的模型验证是其从研究阶段走向临床应用的关键环节。模型验证通常包括内部验证和外部验证，旨在评估模型在不同数据集上的性能、稳定性和泛化能力。数据异质性是模型验证中面临的主要挑战，尤其是在多中心、多队列研究中。
内部验证：
内部验证是指在构建模型所用的同一数据集内部进行验证，以评估模型在训练数据上的稳定性和是否存在过拟合。常用的内部验证方法包括：
- 交叉验证（Cross-validation）：这是一种通过将原始数据集划分为训练集和验证集来评估模型性能的技术。最常见的形式是k折交叉验证（k-fold cross-validation），即将数据集随机分成k个子集。每次训练时，选择k-1个子集作为训练数据，剩余1个子集作为验证数据，重复k次，每次使用不同的验证子集。最终的模型性能是这k次评估结果的平均值。这种方法能够充分利用有限的数据，并对模型的泛化能力给出较为可靠的估计。例如，在构建晚期NSCLC患者免疫治疗疗效的列线图模型时，常常采用五折交叉验证来评估模型的内部一致性指数（C-index）和校准度 75。
- Bootstrap重采样（Bootstrap resampling）：这是一种基于重抽样的统计方法，通过从原始数据集中有放回地随机抽取样本，生成大量的模拟数据集。然后，在每个模拟数据集上训练模型，并在原始数据集的剩余样本上进行验证。这种方法尤其适用于小样本数据集，可以估计模型系数的稳定性、预测误差和优化偏差。例如，在预测晚期黑色素瘤患者对免疫检查点抑制剂响应的模型中，采用了Bootstrap重采样进行内部验证，以调整过拟合的风险 76。
内部验证的目的是确保模型在训练数据上具有良好的性能，并减少过拟合的风险。然而，内部验证无法完全模拟模型在全新、未见过的数据上的表现，因此外部验证显得尤为重要。
外部验证：
外部验证是指在与模型构建数据来源不同（如不同机构、不同时间段或不同人群）的独立数据集上评估模型的性能。这是评估模型泛化能力和临床适用性的“金标准”。外部验证成功与否直接决定了模型能否被广泛采纳和应用。
- 多中心、多队列外部验证：理想的外部验证应在一个或多个独立的多中心队列中进行。这能够充分检验模型在不同医疗实践、不同患者特征和不同数据采集标准下的稳健性。例如，一项预测非小细胞肺癌肿瘤浸润免疫细胞的预后模型，就在两个独立的外部验证队列（GSE31210, GSE37745）中进行了评估，并取得了良好的一致性指数和校准度 77。另一项针对广泛期小细胞肺癌患者免疫化疗预后的模型，在训练集和验证集均展示了良好的预测性能（AUC值均大于0.7），表明其具有一定的泛化能力 75。一项针对Sintilimab治疗NSCLC患者的LIPI（Lung Immune Prognostic Index）模型，通过多中心数据分层为训练、内部验证和外部验证队列，模型在所有队列中均显示出优秀的判别性能和一致的校准度，证明了其在不同人群中的泛化能力 78。
数据异质性的应对策略：
数据异质性是外部验证中最常见的挑战，它可能来源于以下方面：
- PD-L1检测平台差异：PD-L1表达水平是免疫治疗的关键预测标志物，但不同抗体克隆（如22C3、28-8、SP142、SP263）、不同的检测平台、不同的判读标准和阈值设定，都会导致PD-L1检测结果的异质性。为应对此问题，需要进行检测标准化和互通性研究，例如开展跨平台对比研究，建立统一的质控体系和判读标准，或开发能够兼容不同平台数据的算法。
- 影像数据采集差异：不同扫描仪型号、扫描参数、图像重建算法等均会导致影像特征的差异。应对策略包括图像标准化（如灰度值归一化、体素大小重采样）、批次效应校正（如ComBat算法）、以及采用数据增强技术来增加模型的鲁棒性。
- 临床数据定义不一致：不同机构对某些临床变量（如ECOG PS评分、疾病分期、并发症记录）的定义或记录方式可能存在差异。解决方案是建立统一的数据字典和数据采集规范，并在数据预处理阶段进行严格的数据清洗和标准化。
- 人群特征差异：不同地域、种族、生活习惯的患者群体在遗传背景和疾病表型上可能存在差异。为提高模型泛化能力，应尽可能纳入多元化人群的临床数据进行模型训练和验证，或针对特定人群开发亚组模型。
- 治疗方案和管理差异：即使是相同的免疫治疗药物，在不同临床中心可能与不同的化疗方案、剂量调整策略、支持治疗等相结合，这也会引入异质性。解决途径包括详细记录治疗方案，并在模型中将关键治疗参数作为协变量纳入，或在验证时选择治疗方案相似的队列。
成功的外部验证和对数据异质性的有效管理是多参数模型从实验室走向临床的关键一步。只有通过严格的验证，才能确保模型的预测能力在真实世界中持续有效，从而真正为临床决策提供可靠依据。
4.2 临床实用性评估：决策曲线与阈值分析
模型的统计学性能指标，如AUC、准确率、敏感性和特异性，虽然能反映模型的判别能力，但并不能直接量化模型在临床实践中的真正价值和对患者治疗决策的影响。为了弥补这一差距，临床实用性评估变得至关重要，其中决策曲线分析（Decision Curve Analysis, DCA）和阈值分析是评估多参数模型临床价值的有效工具。
决策曲线分析（DCA）：
DCA是一种用于评估诊断工具、预测模型或干预策略临床实用性的方法，它通过计算在不同风险阈值下，使用模型所带来的“净获益”（Net Benefit）来评估其优劣。净获益的计算公式考虑了真阳性（正确识别高风险患者并进行干预）所带来的益处，以及假阳性（错误识别低风险患者并进行不必要干预）所带来的危害。
- 净获益（Net Benefit）：净获益是基于特定风险阈值，通过权衡真阳性（TP）和假阳性（FP）来量化决策策略的临床价值。其公式通常表示为：
净获益 = (TP / N) - (FP / N) * (P_t / (1 - P_t))，其中N是总样本数，P_t是临床医生选择干预的概率阈值。当DCA曲线高于“所有干预”曲线（假设所有患者都进行干预）和“无干预”曲线（假设所有患者都不进行干预）时，表明该模型在相应阈值范围内具有临床实用性。 - 应用案例：在晚期NSCLC免疫治疗疗效预测中，DCA能够帮助临床医生判断，在什么风险概率下，使用该预测模型来指导患者是否接受免疫治疗能获得最大的临床净获益。例如，如果模型预测患者对免疫治疗应答的概率较低，而DCA显示在这个概率阈值下选择不治疗可以避免不必要的毒性和经济负担，则模型具有临床实用性。一项研究构建了预测非小细胞肺癌患者免疫治疗联合抗血管生成治疗预后的列线图，通过DCA评估发现，该列线图在广泛的阈值概率范围内提供了更高的净获益，表明其在临床决策中具有良好价值 79。另一项预测晚期NSCLC免疫检查点抑制剂治疗预后的模型，同样通过DCA验证了其在临床应用中的潜在价值，表明模型能够有效识别出高获益人群 80。此外，有研究开发了增强的电子患者报告结局测量（ePROMs）预测模型，其DCA曲线显示，在25%到75%的风险阈值之间，ePROM增强模型比仅包含临床因素的模型带来更高的净获益，提示其在指导临床决策（如是否进行干预）方面具有优越性 58。
阈值分析：
阈值分析是DCA的组成部分，它涉及确定一个或多个决策点或风险概率阈值，在该阈值之上或之下，模型的预测结果会触发不同的临床行动。例如，一个模型可能建议，当预测患者对免疫治疗应答的概率高于某个特定阈值（如60%）时才给予治疗。阈值的选择需要结合治疗的成本、潜在副作用以及不治疗的后果等因素进行权衡。
区分“统计学显著”与“临床可操作性”：
这是模型评估中一个极其重要的概念：
- 统计学显著性：主要关注的是观察到的结果是否不大可能是随机发生的。例如，一个P值小于0.05的结果被认为是统计学显著的。模型性能指标如AUC、C-index等，其统计学显著性可以通过置信区间来评估。然而，一个统计学显著的结果并不一定意味着它在临床上具有实际意义或可操作性。
- 临床可操作性/临床实用性：则强调模型预测结果对实际临床决策和患者管理的影响。一个模型即使在统计学上表现优异（如AUC很高），但如果它提供的额外信息不足以改变临床医生的现有决策，或者其预测的风险差异在临床上意义不大，那么它的临床可操作性就较低。例如，IMpower150队列衍生模型的研究会说明决策曲线分析（DCA）、净获益率等指标在评估模型临床价值中的应用，区分“统计学显著”与“临床可操作性”。
DCA通过量化模型的净获益，直接回答了“使用这个模型是否比不使用（或普遍干预）更好？”这一临床核心问题，从而更好地连接了模型的统计学性能与实际临床价值。这对于将多参数预测模型成功转化为临床决策支持工具至关重要。
4.3 现存挑战：可解释性与泛化能力的瓶颈
尽管多参数建模在晚期肺癌免疫治疗疗效预测中展现出巨大潜力，但其在临床转化和实际应用过程中仍面临一系列挑战。其中，“黑箱”问题、参数共线性以及真实世界数据偏倚是限制其进一步发展的关键瓶颈。
“黑箱”问题：可解释性不足
多参数模型，尤其是基于复杂机器学习和深度学习算法的模型，往往被视为“黑箱”。这意味着模型能够提供预测结果，但其内部决策过程不透明，难以直接理解哪些输入特征以及这些特征如何共同作用促成了最终的预测结果。
- 深度学习的特征重要性难以阐释：深度学习模型（如卷积神经网络）通过多层非线性变换自动提取特征，这些特征往往是高度抽象的，与人类可理解的生物学或临床概念之间缺乏直观的对应关系。例如，一个深度学习模型可能通过分析影像学特征来预测疗效，但无法明确指出是哪些具体的纹理、形状或强度分布模式导致了这一预测。这种可解释性不足，使得临床医生难以信任和采纳模型，因为他们无法理解模型背后的逻辑，也无法在出现错误预测时进行干预或修正 81。在医疗领域，医生需要了解模型的推理过程，以便对其预测结果的正确性、可靠性及潜在偏见进行评估，特别是在涉及患者生命健康的决策时 81。
- 可解释人工智能（XAI）的兴起与局限：为了解决“黑箱”问题，可解释人工智能（Explainable AI, XAI）应运而生。XAI旨在开发能使人类理解其决策过程的AI系统 82。例如，局部可解释模型不可知解释（LIME）和SHAP（SHapley Additive exPlanations）等方法能够解释单个预测的贡献度，而注意力机制（Attention Mechanism）则可以在深度学习模型中突出对预测贡献最大的输入区域 83。然而，当前的XAI方法仍存在局限性：它们可能无法提供全局性的、易于理解的解释，或者其解释本身仍需要专业知识才能解读。此外，XAI方法在不同类型的模型和数据上的效果不一，有时甚至可能误导用户。例如，一项研究利用XAI方法识别非小细胞肺癌分类的生物标志物，虽然成功发现了与既往研究一致的生物标志物，但仍有部分新发现的生物标志物需要进一步的生物学验证 82。另一个例子是，一个用于癌症药物价值预测的XAI模型，旨在提供易于理解和解释的数据，但其有效性仍需在更广泛的场景中验证 84。
参数共线性对模型稳定性的影响
参数共线性是指模型中多个预测变量之间存在高度相关性。这在多参数建模中尤为常见，因为许多生物标志物和临床特征在生物学上是相互关联的。
- TMB与PD-L1的相关性：例如，高肿瘤突变负荷（TMB）的肿瘤通常产生更多新抗原，这可能导致更强的免疫反应，进而诱导PD-L1表达上调。因此，TMB和PD-L1表达之间可能存在一定的正相关性。当这两个高度相关的参数同时被纳入模型时，可能会导致：
- 模型不稳定：共线性的存在会使得模型系数的估计值变得不稳定，对数据中的微小扰动非常敏感，从而导致模型的可变性和不可靠性。
- 特征重要性判断困难：模型可能无法准确区分TMB和PD-L1各自对疗效预测的独立贡献，从而影响对关键生物学机制的理解。
- 过拟合风险增加：在训练数据上表现良好，但在新数据上泛化能力差。
- 应对策略：解决参数共线性的方法包括：特征选择（只选择其中一个相关性高的特征）、主成分分析（PCA）等降维技术、以及正则化方法（如Lasso、Ridge回归），这些方法可以在一定程度上减轻共线性问题对模型稳定性的影响。然而，这些方法可能会牺牲部分信息的完整性或可解释性。
真实世界数据偏倚：泛化能力的瓶颈
多参数模型在训练和验证过程中往往依赖于特定队列或机构的数据，这使得模型在应用于不同来源的真实世界数据时，其泛化能力可能受到限制。
- 选择偏倚：临床研究往往有严格的纳入和排除标准，导致入组患者的特征可能与真实世界患者群体存在差异。例如，临床试验通常排除ECOG PS评分较差、合并症较多或既往治疗史复杂的患者。当模型在这些“理想”患者数据上训练后，应用于真实世界中更复杂的患者群体时，其预测准确性可能下降。
- 数据质量与完整性：真实世界数据通常存在数据缺失、录入错误、数据定义不一致等问题，这会影响模型的训练和验证效果。
- 动态变化：肿瘤的生物学特性、免疫微环境以及治疗方案都在不断演变。基于历史数据训练的模型可能无法适应新的治疗策略或疾病进展模式。
- 监管与伦理挑战：AI模型在医疗领域的应用还需面对严格的监管审批流程，如FDA对AI诊断模型的要求，以及数据隐私和伦理方面的考量。
克服这些挑战需要多学科的协作，包括开发更具可解释性的AI算法、优化数据采集和标准化流程、构建更具代表性的多中心数据集，并持续进行模型验证和迭代更新，以最终实现多参数模型在晚期肺癌免疫治疗中的广泛临床应用。
5. 未来研究方向与展望
5.1 多组学整合与动态建模
随着生命科学技术的飞速发展，特别是高通量测序和成像技术的进步，多组学数据的获取变得更加便捷和深入。未来的多参数建模将不再局限于传统的基因组、转录组和蛋白质组，而是会更广泛地整合包括单细胞测序、空间转录组等在内的新型组学数据，并结合治疗过程中的动态生物标志物，构建更全面、更具预测力的时序模型。
多组学参数的扩展：
单细胞测序（Single-cell sequencing）的引入：
传统的批量测序（bulk sequencing）只能提供组织样本中细胞的平均信息，无法揭示肿瘤内部的细胞异质性。而单细胞测序技术，如单细胞RNA测序（scRNA-seq），能够解析单个细胞的基因表达谱，从而深入理解肿瘤微环境（TME）中不同细胞亚群的组成、状态和功能。例如，研究表明肿瘤微环境的复杂性和细胞组成对免疫治疗的反应至关重要 85。通过单细胞测序，研究人员可以识别出在免疫治疗响应或耐药过程中发挥关键作用的特定淋巴细胞表型 85、肿瘤相关巨噬细胞（TAMs）亚群 86、癌相关成纤维细胞（CAFs） 87，甚至是肿瘤细胞内部免疫应答相关基因的异质性 88。例如，有研究通过单细胞转录组分析揭示了EGFR突变型肺腺癌的免疫抑制微环境，发现其缺乏CD8+组织驻留记忆（TRM）细胞，这可能导致其对免疫治疗响应较差 89。未来，将这些单细胞层面的精细信息纳入多参数模型，有望更精准地刻画肿瘤免疫微环境的状态，从而提升疗效预测的准确性。空间转录组（Spatial transcriptomics）的价值：
空间转录组技术能够在保留组织空间信息的前提下，测量基因表达水平。这对于理解肿瘤微环境细胞间的相互作用至关重要，因为细胞的功能和相互作用往往与其空间位置紧密相关。例如，研究发现CD14+APOE+细胞与MMP7+肿瘤细胞在空间上的共定位与非小细胞肺癌免疫治疗应答不佳有关，这种共定位抑制了T细胞浸润和三级淋巴结构（TLS）的形成 90。此外，空间转录组技术还被用于揭示化疗免疫治疗后肿瘤微环境的重塑，识别关键的细胞亚群和细胞间通讯模式，如SELENOP-巨噬细胞与抗原提呈癌相关成纤维细胞（apCAFs）在空间上的共定位对治疗响应至关重要 87。将空间转录组数据纳入多参数模型，可以捕捉细胞间的空间相互作用和组织微环境的结构特征，为疗效预测提供更丰富、更具生物学意义的维度。其他新型组学的整合：
除了上述两种，其他多组学数据，如蛋白质组学、代谢组学、表观遗传学（如甲基化组）等也将被更广泛地整合。例如，一项研究通过多组学整合揭示了线粒体基因在肺腺癌预后和个体化治疗中的意义，并构建了基于线粒体相关基因的人工智能衍生预后签名（AIDPS）模型 91。另一项研究则整合了乳酸化和PANoptosis相关的标志物，构建了预测肺腺癌预后的多组学模型，发现高风险患者表现出基因组改变增加、免疫浸润减少和免疫治疗应答较差的特点 92。未来，通过多组学融合，有望从更全面的生物学视角揭示肿瘤对免疫治疗的响应机制，并发现新的预测性生物标志物组合。
动态建模的开发：基于治疗过程中动态生物标志物：
免疫治疗的疗效并非一成不变，而是随着治疗进程和肿瘤-免疫微环境的动态变化而发展。因此，基于治疗过程中动态生物标志物构建时序模型，将是未来的重要研究方向。
循环肿瘤DNA（ctDNA）的动态监测：
ctDNA是一种非侵入性、可重复采样的生物标志物，能够实时反映肿瘤负荷和分子残余病灶（MRD）的状态。研究表明，ctDNA在诊断、治疗监测和预后评估中具有巨大潜力 93。在免疫治疗中，ctDNA的动态变化能够预测疗效。例如，对于局部晚期NSCLC患者，放化疗后ctDNA检测不到的患者无论是否接受巩固性免疫检查点抑制剂（CICI）治疗，预后都非常好；而放化疗后仍存在MRD的患者，接受CICI治疗则能显著改善预后 94。此外，ctDNA在CICI治疗早期变化的模式也能识别出对巩固治疗有反应的患者 94。未来，将ctDNA的动态变化数据纳入多参数时序模型，可以实现对疗效的早期预测和实时评估，指导治疗方案的调整。外周血免疫细胞亚群的动态变化：
外周血淋巴细胞亚群（如CD4+ T细胞、CD8+ T细胞、B细胞、NK细胞等）的数量和比例，在免疫治疗过程中会发生动态变化，这些变化可能与疗效相关。通过定期监测这些免疫细胞亚群的变化，结合其他临床和影像参数，可以构建动态预测模型，及时识别潜在的响应者或耐药患者。其他动态标志物：
除了ctDNA和外周血免疫细胞，治疗期间的炎症指标（如CRP、LDH）、影像学特征（如肿瘤大小、代谢活性、纹理特征的动态变化，即“delta-radiomics”）也都是潜在的动态生物标志物。将这些在治疗过程中多点采样的动态数据整合到时序模型中，能够更精确地捕捉肿瘤对免疫治疗的反应模式，识别早期预警信号，从而实现真正的个性化动态管理。
总之，未来的多参数建模将朝着更高维度、更精细化、更动态化的方向发展，通过整合单细胞和空间组学等前沿技术，结合ctDNA等动态生物标志物，构建能够实时反映患者对免疫治疗响应和耐药机制的时序预测模型，从而为晚期肺癌患者带来更精准的个体化治疗策略。
5.2 标准化流程与开放数据库建设
在多参数建模预测晚期肺癌免疫治疗疗效的研究中，模型的稳健性、可重复性和泛化能力是其能否成功应用于临床的关键。为了实现这些目标，建立统一的标准化流程和推动开放数据库的建设变得尤为重要。
标准化流程的需求：
参数采集的标准化：
- 生物标志物检测标准化：目前，许多生物标志物，如PD-L1表达、TMB等，其检测方法、试剂、仪器平台以及判读标准仍存在较大差异。例如，PD-L1的抗体克隆不同，检测结果往往不一致；TMB的检测需要统一测序深度、生物信息学分析流程和突变阈值。缺乏统一标准导致不同研究之间的数据难以比较和整合，也限制了模型的泛化能力。因此，亟需制定详细的国际或行业标准操作规程（SOP），涵盖样本采集、处理、检测、数据质控及报告的各个环节，确保不同研究中心间数据的一致性和可比性。
- 影像数据采集标准化：医学影像数据的采集受多种因素影响，如扫描设备型号、扫描协议、重建算法等。这些差异会引入“批次效应”，影响影像组学特征的提取和模型的稳定性。未来的研究应致力于推广统一的影像采集协议，并开发先进的图像标准化和批次效应校正算法，以减少数据异质性。
- 临床数据统一：临床特征如ECOG PS、疾病分期、既往治疗史等应采用统一的定义和编码系统，确保数据在不同机构间的互操作性。
特征筛选与模型构建的标准化：
- 统一特征提取方法：特别是在影像组学中，特征提取软件和参数设置的多样性会导致不同的结果。建议推广或开发标准化的、开源的特征提取工具，并明确其参数设置。
- 模型构建方法的透明化：研究应详细报告所采用的模型算法、训练参数、交叉验证策略、过拟合处理方法等，以增加模型构建过程的透明度和可重复性。
- 预定义端点和评估指标：为避免“P值黑客”（p-hacking）和结果选择性报告，研究应在项目启动前预先定义主要和次要研究终点，并使用公认的统计学和临床实用性评估指标（如AUC、C-index、DCA等）进行模型性能评估。
模型验证标准指南：
- 强调外部验证：如前所述，内部验证只能反映模型的拟合优度，外部验证才是模型泛化能力的真正考验。未来的研究应更普遍地要求进行独立、多中心、大样本量的外部验证。
- 制定报告指南：参照PRISMA、STARD等报告指南，为预测模型研究制定专门的报告规范（如TRIPOD声明），确保报告内容的完整性和透明性，便于研究者评估模型的质量和可信度。
开放数据库建设的重要性：
开放数据库是推动多参数建模发展和提升模型泛化能力的核心动力。
- 增加训练样本量和多样性：大规模、多中心、涵盖不同种族和地理区域的开放数据库能够提供足够多样化的数据，帮助模型学习更具普遍性的规律，减少因特定队列偏倚导致的过拟合，从而显著提升模型的泛化能力。例如，TCGA-LUAD（The Cancer Genome Atlas - Lung Adenocarcinoma）等大型公共数据库，收录了大量肺腺癌患者的基因组、转录组、蛋白质组以及临床信息，为研究者提供了丰富的数据资源进行多组学整合分析和模型开发。
- 促进跨机构协作与验证：开放数据库为全球研究者提供了共享数据和验证模型的基础平台，打破了数据孤岛，促进了学术交流与合作。通过共享数据，不同团队可以独立地验证彼此的模型，从而加快模型的迭代优化和临床转化。
- 支持基准测试与性能对比：开放数据库可以作为模型性能的基准测试平台，不同的算法和模型可以在同一数据集上进行公平比较，从而推动更优预测模型的开发。
- IMvigor210队列：作为一项重要的临床试验队列，IMvigor210的公开数据（尤其是在膀胱癌免疫治疗领域）为研究免疫治疗疗效的生物标志物和预测模型提供了宝贵的资源，它包含了详细的临床信息、基因组数据和免疫学数据，是许多免疫治疗预测模型研究的重要验证队列。类似的，肺癌领域也需要更多这样高质量、可访问的开放临床试验和真实世界数据。
通过以上标准化流程的建立和开放数据库的建设，将能够显著提升多参数预测模型的质量、可信度和临床实用性，最终加速其在晚期肺癌免疫治疗领域的广泛应用，真正实现精准医疗的目标。
5.3 临床决策支持系统的落地路径
将多参数预测模型从科研成果转化为临床实践，最终目标是实现临床决策支持系统（Clinical Decision Support System, CDSS）的落地。这需要模型与现有的医疗信息系统（如电子病历系统）无缝集成，并考虑监管审批、伦理以及用户采纳等方面的复杂挑战。
模型与电子病历系统（EMR）、计算机辅助诊断（CAD）工具的集成方案：
为了让多参数预测模型真正赋能临床决策，其结果必须能够便捷地被临床医生获取和理解，并融入日常工作流程。
与电子病历系统（EMR）的集成：
EMR是临床信息的枢纽，整合了患者的诊断、治疗、用药、检查结果等所有数据。将预测模型与EMR系统集成，意味着：- 数据自动输入与提取：模型能够自动从EMR中提取所需的患者数据（如年龄、ECOG评分、既往病史、实验室检查结果、基因检测报告等），省去手动数据输入的时间和错误。例如，一些AI模型已经能够自动处理非结构化的医疗文本数据，如病理报告和放射学报告，从中提取关键信息用于癌症结局预测 9596。这种自动化能力大大提升了数据的可及性和利用效率。
- 实时预测与报告生成：模型可在患者就诊或关键决策点时，基于EMR中的最新数据进行实时预测，并以结构化、可视化的方式（如风险评分、预测概率、决策曲线图）直接呈现在EMR界面。这可以帮助医生快速评估患者风险，如血栓风险预测等。
- 警报与建议：系统可根据预测结果，在EMR中触发警报或提供个性化的治疗建议。例如，当预测患者对免疫治疗应答概率极低或副作用风险极高时，系统可弹出提示，建议医生重新评估治疗方案，或考虑其他替代疗法。
- 工作流嵌入：将预测工具嵌入到临床医生的日常工作流中，例如在开具治疗医嘱或进行多学科会诊时自动触发模型计算，确保预测信息在最需要时呈现。
- 未来发展：电子系统在临床研究中被越来越多地使用，能够简化数据传输，提高试验透明度，并改善研究文档管理 97。
与计算机辅助诊断（CAD）工具的结合：
CAD工具主要用于影像学分析，通过AI算法辅助医生进行疾病检测、诊断和量化。将多参数模型与CAD工具结合，尤其是在影像组学方面，可以实现：- 自动化影像特征提取：CAD工具可以自动从CT、PET-CT等影像中提取定量影像组学特征，并将其作为预测模型的输入。例如，深度学习模型能够直接从CT图像中检测肺部结节，并评估免疫治疗的适用性 98。
- 影像可视化与解释：预测模型的结果可以与CAD工具的影像分析结果结合呈现，例如在图像上高亮显示与预测结果最相关的区域（如热力图），增强模型的可解释性。
- 动态监测与疗效评估：在治疗过程中，CAD工具可定期对影像数据进行分析，监测肿瘤大小、形态和代谢特征的动态变化，并将这些动态影像组学特征输入预测模型，实时评估治疗响应，辅助医生调整治疗方案 99。
- 病理诊断与分型：AI在病理学中也扮演着越来越重要的角色，通过分析全玻片图像进行肿瘤识别、分级甚至分子分型，这些病理特征可以作为多参数模型的重要组成部分 100101。
监管审批（如FDA对AI诊断模型的要求）对临床转化的影响：
AI驱动的医疗器械（SaMD, Software as a Medical Device）在进入临床应用前，必须获得严格的监管审批，这通常是一个复杂且漫长的过程。
FDA的监管框架：美国食品药品监督管理局（FDA）是全球领先的医疗器械监管机构之一，其对AI/ML软件的监管框架具有借鉴意义。FDA将SaMD分为不同风险等级，预测模型通常属于较高风险类别，需要提供充分的验证数据以证明其安全性、有效性和性能。
- 可解释性与透明度：监管机构越来越关注AI模型的“可解释性”（explainability），要求开发者能够解释模型的决策逻辑，这与模型“黑箱”问题形成了冲突。
- 鲁棒性与泛化能力：模型需要证明其在不同人群、不同临床设置下的鲁棒性和泛化能力，这要求进行大规模、多中心、多样化的外部验证 102。
- 持续学习与更新：AI模型，特别是那些采用持续学习（continual learning）或自适应算法的模型，其性能可能会随数据输入而变化。FDA正在探索针对这类模型的“预先指定变更控制计划”（Pre-specified Change Control Plan），允许在预定义的安全范围内进行模型更新而无需重新提交完整审批。
- 真实世界证据（RWE）：FDA越来越重视使用真实世界证据来支持AI医疗产品的审批，这需要强大的数据整合和分析能力，例如从EMR中提取和分析数据 9596。
伦理与法律挑战：
- 数据隐私与安全：多参数模型需要整合大量敏感的患者数据，这带来了严峻的数据隐私保护（如GDPR、HIPAA法规）和网络安全挑战。
- 算法偏倚与公平性：如果训练数据存在偏倚（如缺乏特定种族或地理区域的数据），模型可能会在这些未充分代表的群体中表现不佳，甚至产生歧视性结果。确保模型的公平性是一个重要的伦理考量 103。
- 责任归属：当AI模型做出错误预测导致不良后果时，责任应归属给开发者、医生还是医院，这在法律上尚不明确。
用户采纳与信任：
总的来说，多参数预测模型要真正落地为临床决策支持系统，需要技术创新（如与EMR/CAD的深度集成）、严格的监管审批流程（强调安全性、有效性、泛化能力和可解释性）、以及对伦理、法律和社会接受度的全面考量 106107108。只有跨学科、多方协作，才能最终实现AI在晚期肺癌免疫治疗中的精准化和个性化应用。
内容由 AI 生成，仅供参考，请仔细甄别
参考文献
1Immunotherapy in Lung Cancer: Current Landscape and Future Directions.PubMed
Hirva Mamdani, Sandro Matosevic, Ahmed Bilal Khalid, et al.
Front Immunol. 2022 Feb 9;13:823618. doi: 10.3389/fimmu.2022.823618. eCollection 2022.
Over the past decade, lung cancer treatment has undergone a major paradigm shift. A greater understanding of lung cancer biology has led to the development of many effective targeted therapies as well as of immunotherapy. Immune checkpoint inhibitors (ICIs) have shown tremendous benefit in the treatment of non-small cell lung cancer (NSCLC) and are now being used as first-line therapies in metastatic disease, consolidation therapy following chemoradiation in unresectable locally advanced disease, and adjuvant therapy following surgical resection and chemotherapy in resectable disease. Despite these benefits, predicting who will respond to ICIs has proven to be difficult and there remains a need to discover new predictive immunotherapy biomarkers. Furthermore, resistance to ICIs in lung cancer is frequent either because of a lack of response or disease progression after an initial response. The utility of ICIs in the treatment of small cell lung cancer (SCLC) remains limited to first-line treatment of extensive stage disease in combination with chemotherapy with modest impact on overall survival. It is thus important to explore and exploit additional targets to reap the full benefits of immunotherapy in the treatment of lung cancer. Here, we will summarize the current state of immunotherapy in lung cancer, discuss novel targets, and explore the intersection between DNA repair defects and immunotherapy.
2Association of Survival and Immune-Related Biomarkers With Immunotherapy in Patients With Non-Small Cell Lung Cancer: A Meta-analysis and Individual Patient-Level Analysis.PubMed
Yunfang Yu, Dongqiang Zeng, Qiyun Ou, et al.
JAMA Netw Open. 2019 Jul 3;2(7):e196879. doi: 10.1001/jamanetworkopen.2019.6879.
IMPORTANCE: The beneficial role of immunotherapy and the clinical relevance of current biomarkers in non-small cell lung cancer (NSCLC) remain inconclusive; thus, appropriate strategies and reliable predictors need further definition. OBJECTIVES: To evaluate the association of clinical outcomes with immune checkpoint inhibitors, tumor vaccines, and cellular immunotherapy in patients with advanced NSCLC and to explore appropriate strategies, candidates, and predictors. DATA SOURCES: The PubMed, EMBASE, and Cochrane Central Register of Controlled Trials databases were searched from inception to June 2018, using relevant search keywords and Medical Subject Headings (MeSH) terms, including tumor vaccine, cellular immunotherapy, immune checkpoint inhibitor, cytotoxic T-lymphocyte-associated protein 4, programmed death-ligand 1, programmed death receptor 1, and non-small cell lung carcinoma. Systematic reviews, meta-analyses, references, and conference proceedings were manually searched. STUDY SELECTION: English-language randomized clinical trials with available data that measured overall survival (OS), progression-free survival (PFS), or objective response rate comparing immune checkpoint inhibitors, tumor vaccines, or cellular immunotherapy with conventional therapy for patients with advanced or metastatic NSCLC were included. Thirty-one immunotherapy randomized clinical trials were included, and multicohort data included next-generation sequencing data from patients with advanced NCSLC. DATA EXTRACTION AND SYNTHESIS: Hazard ratios and 95% CIs were pooled to estimate the survival increases in OS and PFS. Dichotomous data, such as object response rate data, were analyzed using the risk ratio. Mantel-Haenszel random-effects model was used. I2 was used to assess the heterogeneity between trials; an I2 value exceeding 50% indicated the existence of substantial heterogeneity. Analyses took place from February 1, 2018, to August 31, 2018. MAIN OUTCOMES AND MEASURES: Primary outcomes were OS and PFS. RESULTS: In total, 14 395 patients (9500 [66.0%] men) were included in the meta-analysis, and 1833 patients (mean [SD], 65.2 [9.9] years; 1063 [58.0%] men) were included in the individual patient-level study. Compared with conventional therapy, immunotherapy was associated with significantly longer OS (hazard ratio, 0.76; 95% CI, 0.71-0.82; P < .001) and PFS (hazard ratio, 0.76; 95% CI, 0.70-0.83; P < .001). The best checkpoint blockade strategy was first-line pembrolizumab with platinum-based chemotherapy. The combined predictive utility of programmed cell death ligand 1 (PD-L1) expression and tumor mutation burden (TMB) was associated with predictive prognosis (whole-exome sequencing: 1-year PFS area under the receiver operating characteristic curve [AUC], 0.829; 3-year PFS AUC, 0.839; targeted next-generation sequencing: 1-year PFS AUC, 0.826; 3-year PFS AUC, 0.948). Moreover, the addition of CD8+ T-cell tumor-infiltrating lymphocytes was associated with improved prognosis predictions for OS (3-year OS AUC, 0.659; 5-year OS AUC, 0.665). RYR1 or MGAM mutations were significantly associated with concomitantly increased durable clinical benefits (RYR1: durable clinical benefit [DCB], 12 of 51 patients [24%]; no durable benefit [NDB], 2 of 55 patients [4%]; P < .001; MGAM: DCB, 12 of 51 patients [24%]; NDB, 0 patients; P < .001), a higher TMB (RYRI: high TMB, 12 of 53 patients [23%]; low TMB, 2 of 53 patients [38%]; P < .001; MGAM: high TMB, 9 of 53 patients [17%]; low TMB, 0 patients; P < .001), and higher PD-L1 expression (RYRI: high PD-L1 expression, 8 of 30 patients [27%]; low PD-L1 expression, 6 of 85 [7.1%]; P < .001; MGAM: high PD-L1 expression, 6 of 30 patients [20%]; low PD-L1 expression, 5 of 85 patients [6%]; P < .001). CONCLUSIONS AND RELEVANCE: Immunotherapies showed promising clinical outcomes for patients with NSCLC. Pembrolizumab with platinum-based chemotherapy was found to be the most appropriate first-line immune checkpoint inhibitor regimen for advanced NSCLC, and the combined use of PD-L1 expression and TMB was found to be a promising biomarker to evaluate patients' survival and response to precision immunotherapy. The further combination of CD8+ T-cell tumor-infiltrating lymphocytes, PD-L1 expression, and TMB was associated with reliable prognosis. The predictive value of that combination needs to be prospectively validated in large-scale studies.
3Efficacy and safety of PD-1/PD-L1 plus CTLA-4 antibodies ± other therapies in lung cancer: a systematic review and meta-analysis.PubMed
Xiang Shen, Shangke Huang, Hua Xiao, et al.
Eur J Hosp Pharm. 2023 Jan;30(1):3-8. doi: 10.1136/ejhpharm-2021-002803. Epub 2021 Sep 8.
PURPOSE: To investigate the efficacy and safety of programmed cell death 1 (PD-1)/programmed cell death-ligand 1 (PD-L1) plus cytotoxic T lymphocyte antigen-4 (CTLA-4) antibodies ± other therapies in patients with advanced lung cancer. METHODS: In accordance with the retrieval strategy, we searched electronic databases for randomised controlled trials testing PD-1/PD-L1 plus CTLA-4 antibodies in patients with lung cancer; RR (for objective response rate (ORR), overall survival (OS), progression-free survival (PFS), and immune-related adverse events (irAEs)) from individual studies were calculated and pooled by using random-effects models or fixed-effects models; heterogeneity and publication bias analyses were also performed, using Review Manager 5.3 and Stata 15.1 for statistical analysis. RESULTS: We included six studies. Four different immune checkpoint inhibitors (nivolumab, pembrolizumab, durvalumab, tremelimumab) were used. Dual checkpoint inhibitors ± other therapies for advanced lung cancer showed significant improvements in ORR (RR 1.49, 95% CI 1.11 to 1.98; p=0.007), OS (HR 0.72, 95% CI 0.63 to 0.83; p<0.00001), and PFS (HR 0.72, 95% CI 0.63 to 0.82; p<0.00001). The subgroup analyses were consistent with the pooled results. The PD-L1 ≥1% (HR 0.67, 95% CI 0.54 to 0.82; p<0.0001) subgroup differences indicated a statistically significant subgroup effect, but the PD-L1 <1% subgroup (HR 0.88, 95% CI 0.75 to 1.05; p=0.15) was not statistically significant. The incidence of adverse events (grade ≥3) was lower than that of the control group (RR 0.90, 95% CI 0.80 to 1.02; p=0.09), but was not significant. CONCLUSIONS: PD-1/PD-L1 inhibitors combined with CTLA-4 inhibitors ± other therapies can improve the ORR, OS and PFS of patients with advanced or metastatic lung cancer, but the incidence of adverse reactions is high although generally tolerable. PROSPERO REGISTRATION: CRD42020149216.
4Predictive genetic biomarkers in immune checkpoint inhibitors for non-small-cell lung cancer.PubMed
Geane Tomaídes Henriques, Cleide Barbieri de Souza, Pedro Nazareth Aguiar
Immunotherapy. 2022 Mar;14(4):249-257. doi: 10.2217/imt-2021-0175. Epub 2022 Jan 25.
Immune checkpoint inhibitors improved the overall survival of patients with advanced non-small-cell lung cancer and changed the treatment since the last decade. The duration of response is longer than what is seen with chemotherapy or targeted agents; however, some patients have no benefit or even a progressive disease as best response. Immune checkpoint inhibitor plus chemotherapy combinations are a very useful strategy, but defining precisely who will benefit most from immunotherapy is still a main question. Therefore, understanding the genetics of the tumor microenvironment is a way to determine new predictive biomarkers to replace the only one currently accepted, PD-L1 expression, whose application is surrounded by uncertainties.
5Small Cell Lung Cancer: A Review.PubMed
So Yeon Kim, Henry S Park, Anne C Chiang
JAMA. 2025 Jun 3;333(21):1906-1917. doi: 10.1001/jama.2025.0560.
IMPORTANCE: Small cell lung cancer (SCLC) is a high-grade neuroendocrine carcinoma with an incidence of 4.7 cases per 100 000 individuals in 2021 in the US and a 5-year overall survival of 12% to 30%. OBSERVATIONS: Cigarette smoking is the primary risk factor for development of SCLC, as 95% of patients diagnosed with SCLC have a history of tobacco use. Patients with SCLC may present with respiratory symptoms such as cough (40%), shortness of breath (34%), hemoptysis (10%), or metastases with corresponding local symptoms (30%) such as pleuritis or bone pain; approximately 60% of patients with SCLC may be asymptomatic at diagnosis. Chest imaging may demonstrate central hilar (85%) or mediastinal lymphadenopathy (75%). At diagnosis, approximately 15% of patients have brain metastases, which may present as headache or focal weakness. Diagnosis is confirmed by biopsy of a primary lung mass, thoracic lymph node, or metastatic lesion. Small cell lung cancer is classified into limited stage (LS-SCLC; 30%) vs extensive stage (ES-SCLC; 70%) based on whether the disease can be treated within a radiation field that is typically confined to 1 hemithorax but may include contralateral mediastinal and supraclavicular nodes. For patients with LS-SCLC, surgery or concurrent chemotherapy with platinum-etoposide and radiotherapy is potentially curative in 30% of patients. More recently, median survival for LS-SCLC has reached up to 55.9 months with the addition of durvalumab, an immunotherapy. First-line treatment for ES-SCLC is combined treatment with platinum-etoposide chemotherapy and immunotherapy with the programmed cell death 1 ligand 1 (PD-L1) inhibitors durvalumab or atezolizumab followed by maintenance immunotherapy until disease progression or toxicity. Although initial rates of tumor shrinkage are 60% to 70% with platinum-etoposide and immunotherapy treatment, the median overall survival of patients treated for ES-SCLC is approximately 12 to 13 months, with 60% of patients relapsing within 3 months. Second-line therapy for patients with ES-SCLC includes the DNA-alkylating agent lurbinectedin (35% overall response rate; median progression-free survival, 3.7 months) and a bispecific T-cell engager against delta-like ligand 3, tarlatamab (40% overall response rate; median progression-free survival, 4.9 months). CONCLUSIONS AND RELEVANCE: Small cell lung cancer is a smoking-related malignancy that presents at an advanced stage in 70% of patients. Three-year overall survival is approximately 56.5% for LS-SCLC and 17.6% for ES-SCLC. First-line treatment for LS-SCLC is radiation targeting the tumor given concurrently with chemotherapy and followed by consolidation immunotherapy. For ES-SCLC, first-line treatment is chemotherapy and immunotherapy followed by maintenance immunotherapy.
6Toward personalized treatment approaches for non-small-cell lung cancer.PubMed
Meina Wang, Roy S Herbst, Chris Boshoff
Nat Med. 2021 Aug;27(8):1345-1356. doi: 10.1038/s41591-021-01450-2. Epub 2021 Aug 12.
Worldwide, lung cancer is the most common cause of cancer-related deaths. Molecular targeted therapies and immunotherapies for non-small-cell lung cancer (NSCLC) have improved outcomes markedly over the past two decades. However, the vast majority of advanced NSCLCs become resistant to current treatments and eventually progress. In this Perspective, we discuss some of the recent breakthrough therapies developed for NSCLC, focusing on immunotherapies and targeted therapies. We highlight our current understanding of mechanisms of resistance and the importance of incorporating genomic analyses into clinical studies to decipher these further. We underscore the future role of neoadjuvant and maintenance combination therapy approaches to potentially cure early disease. A major challenge to successful development of rational combination therapies will be the application of robust predictive biomarkers for clear-cut patient stratification, and we provide our views on clinical research areas that could influence how NSCLC will be managed over the coming decade.
7The cutting-edge progress of immune-checkpoint blockade in lung cancer.PubMed
Fei Zhou, Meng Qiao, Caicun Zhou
Cell Mol Immunol. 2021 Feb;18(2):279-293. doi: 10.1038/s41423-020-00577-5. Epub 2020 Nov 11.
Great advances in immune checkpoint blockade have resulted in a paradigm shift in patients with lung cancer. Immune-checkpoint inhibitor (ICI) treatment, either as monotherapy or combination therapy, has been established as the standard of care for patients with locally advanced/metastatic non-small cell lung cancer without EGFR/ALK alterations or extensive-stage small cell lung cancer. An increasing number of clinical trials are also ongoing to further investigate the role of ICIs in patients with early-stage lung cancer as neoadjuvant or adjuvant therapy. Although PD-L1 expression and tumor mutational burden have been widely studied for patient selection, both of these biomarkers are imperfect. Due to the complex cancer-immune interactions among tumor cells, the tumor microenvironment and host immunity, collaborative efforts are needed to establish a multidimensional immunogram to integrate complementary predictive biomarkers for personalized immunotherapy. Furthermore, as a result of the wide use of ICIs, managing acquired resistance to ICI treatment remains an inevitable challenge. A deeper understanding of the underlying biological mechanisms of acquired resistance to ICIs is helpful to overcome these obstacles. In this review, we describe the cutting-edge progress made in patients with lung cancer, the optimal duration of ICI treatment, ICIs in some special populations, the unique response patterns during ICI treatment, the emerging predictive biomarkers, and our understanding of primary and acquired resistance mechanisms to ICI treatment.
8Accuracies of univariate and multivariate genomic prediction models in African cassava.PubMed
Uche Godfrey Okeke, Deniz Akdemir, Ismail Rabbi, et al.
Genet Sel Evol. 2017 Dec 4;49(1):88. doi: 10.1186/s12711-017-0361-y.
BACKGROUND: Genomic selection (GS) promises to accelerate genetic gain in plant breeding programs especially for crop species such as cassava that have long breeding cycles. Practically, to implement GS in cassava breeding, it is necessary to evaluate different GS models and to develop suitable models for an optimized breeding pipeline. In this paper, we compared (1) prediction accuracies from a single-trait (uT) and a multi-trait (MT) mixed model for a single-environment genetic evaluation (Scenario 1), and (2) accuracies from a compound symmetric multi-environment model (uE) parameterized as a univariate multi-kernel model to a multivariate (ME) multi-environment mixed model that accounts for genotype-by-environment interaction for multi-environment genetic evaluation (Scenario 2). For these analyses, we used 16 years of public cassava breeding data for six target cassava traits and a fivefold cross-validation scheme with 10-repeat cycles to assess model prediction accuracies. RESULTS: In Scenario 1, the MT models had higher prediction accuracies than the uT models for all traits and locations analyzed, which amounted to on average a 40% improved prediction accuracy. For Scenario 2, we observed that the ME model had on average (across all locations and traits) a 12% improved prediction accuracy compared to the uE model. CONCLUSIONS: We recommend the use of multivariate mixed models (MT and ME) for cassava genetic evaluation. These models may be useful for other plant species.
9A Comparison of Univariate and Multivariate Forecasting Models Predicting Emergency Department Patient Arrivals during the COVID-19 Pandemic.PubMed
Egbe-Etu Etu, Leslie Monplaisir, Sara Masoud, et al.
Healthcare (Basel). 2022 Jun 16;10(6):1120. doi: 10.3390/healthcare10061120.
The COVID-19 pandemic has heightened the existing concern about the uncertainty surrounding patient arrival and the overutilization of resources in emergency departments (EDs). The prediction of variations in patient arrivals is vital for managing limited healthcare resources and facilitating data-driven resource planning. The objective of this study was to forecast ED patient arrivals during a pandemic over different time horizons. A secondary objective was to compare the performance of different forecasting models in predicting ED patient arrivals. We included all ED patient encounters at an urban teaching hospital between January 2019 and December 2020. We divided the data into training and testing datasets and applied univariate and multivariable forecasting models to predict daily ED visits. The influence of COVID-19 lockdown and climatic factors were included in the multivariable models. The model evaluation consisted of the root mean square error (RMSE) and mean absolute error (MAE) over different forecasting horizons. Our exploratory analysis illustrated that monthly and weekly patterns impact daily demand for care. The Holt-Winters approach outperformed all other univariate and multivariable forecasting models for short-term predictions, while the Long Short-Term Memory approach performed best in extended predictions. The developed forecasting models are able to accurately predict ED patient arrivals and peaks during a surge when tested on two years of data from a high-volume urban ED. These short- and long-term prediction models can potentially enhance ED and hospital resource planning.
10Multivariate linear mixture models for the prediction of febrile seizure risk and recurrence: a prospective case-control study.PubMed
Jan Papež, René Labounek, Petr Jabandžiev, et al.
Sci Rep. 2023 Oct 13;13(1):17372. doi: 10.1038/s41598-023-43599-5.
Our goal was to identify highly accurate empirical models for the prediction of the risk of febrile seizure (FS) and FS recurrence. In a prospective, three-arm, case-control study, we enrolled 162 children (age 25.8 ± 17.1 months old, 71 females). Participants formed one case group (patients with FS) and two control groups (febrile patients without seizures and healthy controls). The impact of blood iron status, peak body temperature, and participants' demographics on FS risk and recurrence was investigated with univariate and multivariate statistics. Serum iron concentration, iron saturation, and unsaturated iron-binding capacity differed between the three investigated groups (p < 0.05). These serum analytes were key variables in the design of novel multivariate linear mixture models. The models classified FS risk with higher accuracy than univariate approaches. The designed bi-linear classifier achieved a sensitivity/specificity of 82%/89% and was closest to the gold-standard classifier. A multivariate model assessing FS recurrence provided a difference (p < 0.05) with a separating sensitivity/specificity of 72%/69%. Iron deficiency, height percentile, and age were significant FS risk factors. In addition, height percentile and hemoglobin concentration were linked to FS recurrence. Novel multivariate models utilizing blood iron status and demographic variables predicted FS risk and recurrence among infants and young children with fever.
11The Heterogeneity Problem: Approaches to Identify Psychiatric Subtypes.PubMed
Eric Feczko, Oscar Miranda-Dominguez, Mollie Marr, et al.
Trends Cogn Sci. 2019 Jul;23(7):584-601. doi: 10.1016/j.tics.2019.03.009. Epub 2019 May 29.
The imprecise nature of psychiatric nosology restricts progress towards characterizing and treating mental health disorders. One issue is the 'heterogeneity problem': different causal mechanisms may relate to the same disorder, and multiple outcomes of interest can occur within one individual. Our review tackles this heterogeneity problem, providing considerations, concepts, and approaches for investigators examining human cognition and mental health. We highlight the difficulty of pure dimensional approaches due to 'the curse of dimensionality'. Computationally, we consider supervised and unsupervised statistical approaches to identify putative subtypes within a population. However, we emphasize that subtype identification should be linked to a particular outcome or question. We conclude with novel hybrid approaches that can identify subtypes tied to outcomes, and may help advance precision diagnostic and treatment tools.
12Mixture multilevel vector-autoregressive modeling.PubMed
Anja F Ernst, Marieke E Timmerman, Feng Ji, et al.
Psychol Methods. 2024 Feb;29(1):137-154. doi: 10.1037/met0000551. Epub 2023 Aug 10.
With the rising popularity of intensive longitudinal research, the modeling techniques for such data are increasingly focused on individual differences. Here we present mixture multilevel vector-autoregressive modeling, which extends multilevel vector-autoregressive modeling by including a mixture, to identify individuals with similar traits and dynamic processes. This exploratory model identifies mixture components, where each component refers to individuals with similarities in means (expressing traits), autoregressions, and cross-regressions (expressing dynamics), while allowing for some interindividual differences in these attributes. Key issues in modeling are discussed, where the issue of centering predictors is examined in a small simulation study. The proposed model is validated in a simulation study and used to analyze the affective data from the COGITO study. These data consist of samples for two different age groups of over 100 individuals each who were measured for about 100 days. We demonstrate the advantage of exploratory identifying mixture components by analyzing these heterogeneous samples jointly. The model identifies three distinct components, and we provide an interpretation for each component motivated by developmental psychology. (PsycInfo Database Record (c) 2024 APA, all rights reserved).
13Development and validation of a predictive model for chemotherapy-associated thrombosis.PubMed
Alok A Khorana, Nicole M Kuderer, Eva Culakova, et al.
Blood. 2008 May 15;111(10):4902-7. doi: 10.1182/blood-2007-10-116327. Epub 2008 Jan 23.
Risk of venous thromboembolism (VTE) is elevated in cancer, but individual risk factors cannot identify a sufficiently high-risk group of outpatients for thromboprophylaxis. We developed a simple model for predicting chemotherapy-associated VTE using baseline clinical and laboratory variables. The association of VTE with multiple variables was characterized in a derivation cohort of 2701 cancer outpatients from a prospective observational study. A risk model was derived and validated in an independent cohort of 1365 patients from the same study. Five predictive variables were identified in a multivariate model: site of cancer (2 points for very high-risk site, 1 point for high-risk site), platelet count of 350 x 10(9)/L or more, hemoglobin less than 100 g/L (10 g/dL) and/or use of erythropoiesis-stimulating agents, leukocyte count more than 11 x 10(9)/L, and body mass index of 35 kg/m(2) or more (1 point each). Rates of VTE in the derivation and validation cohorts, respectively, were 0.8% and 0.3% in low-risk (score = 0), 1.8% and 2% in intermediate-risk (score = 1-2), and 7.1% and 6.7% in high-risk (score >/= 3) category over a median of 2.5 months (C-statistic = 0.7 for both cohorts). This model can identify patients with a nearly 7% short-term risk of symptomatic VTE and may be used to select cancer outpatients for studies of thromboprophylaxis.
14A systematic review shows no performance benefit of machine learning over logistic regression for clinical prediction models.PubMed
Evangelia Christodoulou, Jie Ma, Gary S Collins, et al.
J Clin Epidemiol. 2019 Jun;110:12-22. doi: 10.1016/j.jclinepi.2019.02.004. Epub 2019 Feb 11.
OBJECTIVES: The objective of this study was to compare performance of logistic regression (LR) with machine learning (ML) for clinical prediction modeling in the literature. STUDY DESIGN AND SETTING: We conducted a Medline literature search (1/2016 to 8/2017) and extracted comparisons between LR and ML models for binary outcomes. RESULTS: We included 71 of 927 studies. The median sample size was 1,250 (range 72-3,994,872), with 19 predictors considered (range 5-563) and eight events per predictor (range 0.3-6,697). The most common ML methods were classification trees, random forests, artificial neural networks, and support vector machines. In 48 (68%) studies, we observed potential bias in the validation procedures. Sixty-four (90%) studies used the area under the receiver operating characteristic curve (AUC) to assess discrimination. Calibration was not addressed in 56 (79%) studies. We identified 282 comparisons between an LR and ML model (AUC range, 0.52-0.99). For 145 comparisons at low risk of bias, the difference in logit(AUC) between LR and ML was 0.00 (95% confidence interval, -0.18 to 0.18). For 137 comparisons at high risk of bias, logit(AUC) was 0.34 (0.20-0.47) higher for ML. CONCLUSION: We found no evidence of superior performance of ML over LR. Improvements in methodology and reporting are needed for studies that compare modeling algorithms.
15Deep-learning-enabled protein-protein interaction analysis for prediction of SARS-CoV-2 infectivity and variant evolution.PubMed
Guangyu Wang, Xiaohong Liu, Kai Wang, et al.
Nat Med. 2023 Aug;29(8):2007-2018. doi: 10.1038/s41591-023-02483-5. Epub 2023 Jul 31.
Host-pathogen interactions and pathogen evolution are underpinned by protein-protein interactions between viral and host proteins. An understanding of how viral variants affect protein-protein binding is important for predicting viral-host interactions, such as the emergence of new pathogenic SARS-CoV-2 variants. Here we propose an artificial intelligence-based framework called UniBind, in which proteins are represented as a graph at the residue and atom levels. UniBind integrates protein three-dimensional structure and binding affinity and is capable of multi-task learning for heterogeneous biological data integration. In systematic tests on benchmark datasets and further experimental validation, UniBind effectively and scalably predicted the effects of SARS-CoV-2 spike protein variants on their binding affinities to the human ACE2 receptor, as well as to SARS-CoV-2 neutralizing monoclonal antibodies. Furthermore, in a cross-species analysis, UniBind could be applied to predict host susceptibility to SARS-CoV-2 variants and to predict future viral variant evolutionary trends. This in silico approach has the potential to serve as an early warning system for problematic emerging SARS-CoV-2 variants, as well as to facilitate research on protein-protein interactions in general.
16Predictive Modeling of Alzheimer's and Parkinson's Disease Using Metabolomic and Lipidomic Profiles from Cerebrospinal Fluid.PubMed
Nathan Hwangbo, Xinyu Zhang, Daniel Raftery, et al.
Metabolites. 2022 Mar 22;12(4):277. doi: 10.3390/metabo12040277.
In recent years, metabolomics has been used as a powerful tool to better understand the physiology of neurodegenerative diseases and identify potential biomarkers for progression. We used targeted and untargeted aqueous, and lipidomic profiles of the metabolome from human cerebrospinal fluid to build multivariate predictive models distinguishing patients with Alzheimer's disease (AD), Parkinson's disease (PD), and healthy age-matched controls. We emphasize several statistical challenges associated with metabolomic studies where the number of measured metabolites far exceeds sample size. We found strong separation in the metabolome between PD and controls, as well as between PD and AD, with weaker separation between AD and controls. Consistent with existing literature, we found alanine, kynurenine, tryptophan, and serine to be associated with PD classification against controls, while alanine, creatine, and long chain ceramides were associated with AD classification against controls. We conducted a univariate pathway analysis of untargeted and targeted metabolite profiles and find that vitamin E and urea cycle metabolism pathways are associated with PD, while the aspartate/asparagine and c21-steroid hormone biosynthesis pathways are associated with AD. We also found that the amount of metabolite missingness varied by phenotype, highlighting the importance of examining missing data in future metabolomic studies.
17Neoadjuvant immunotherapy for advanced, resectable non-small cell lung cancer: A systematic review and meta-analysis.PubMed
Yajing Wu, Vivek Verma, Carl M Gay, et al.
Cancer. 2023 Jul 1;129(13):1969-1985. doi: 10.1002/cncr.34755. Epub 2023 Mar 30.
BACKGROUND: Neoadjuvant immunotherapy (nIT) is a rapidly emerging paradigm for advanced resectable non-small cell lung cancer (NSCLC). The objectives of this PRISMA/MOOSE/PICOD-guided systematic review and meta-analysis were (1) to assess the safety and efficacy of nIT, (2) to compare the safety and efficacy of neoadjuvant chemoimmunotherapy (nCIT) versus chemotherapy alone (nCT), and (3) to explore predictors of pathologic response with nIT and their association with outcomes. METHODS: Eligibility was resectable stage I-III NSCLC and the receipt of programmed death-1/programmed cell death ligand-1 (PD-L1)/cytotoxic T-lymphocyte-associated antigen-4 inhibitors before resection; other forms and modalities of neoadjuvant and/or adjuvant therapies were allowed. For statistical analysis, the Mantel-Haenszel fixed-effect or random-effect model was used, depending on the heterogeneity (I ). RESULTS: Sixty-six articles met the criteria (eight randomized studies, 39 prospective nonrandomized studies, and 19 retrospective studies). The pooled pathologic complete response (pCR) rate was 28.1%. The estimated grade ≥3 toxicity rate was 18.0%. Compared with nCT, nCIT achieved higher rates of pCR (odds ratio [OR], 7.63; 95% confidence interval [CI], 4.49-12.97; p < .001), progression-free survival (PFS) (hazard ratio [HR] 0.51; 95% CI, 0.38-0.67; p < .001), and overall survival (OS) (HR, 0.51; 95% CI, 0.36-0.74; p = .0003) but yielded similar toxicity rates (OR, 1.01; 95% CI, 0.67-1.52; p = .97). The results remained robust on sensitivity analysis when all retrospective publications were removed. pCR was associated with improved PFS (HR, 0.25; 0.15-0.43; p < .001) and OS (HR, 0.26; 95% CI, 0.10-0.67; p = .005). PD-L1 expressors (≥1%) were more likely to achieve a pCR (OR, 2.93; 95% CI, 1.22-7.03; p = .02). CONCLUSIONS: In patients with advanced resectable NSCLC, neoadjuvant immunotherapy was safe and efficacious. nCIT improved pathologic response rates and PFS/OS over nCT, particularly in patients who had tumors that expressed PD-L1, without increasing toxicities. PLAIN LANGUAGE SUMMARY: This meta-analysis of 66 studies showed that neoadjuvant immunotherapy for advanced resectable non-small cell lung cancer is safe and efficacious. Compared with chemotherapy alone, chemoimmunotherapy improved pathologic response rates and survival, particularly for patients who had tumors that expressed programmed cell death ligand-1, without increasing toxicities.
18Microsatellite Instability, Mismatch Repair, and Tumor Mutation Burden in Lung Cancer.PubMed
Oana C Rosca, Oana E Vele
Surg Pathol Clin. 2024 Jun;17(2):295-305. doi: 10.1016/j.path.2023.11.011. Epub 2023 Dec 20.
Since US Food and Drug Administration approval of programmed death ligand 1 (PD-L1) as the first companion diagnostic for immune checkpoint inhibitors (ICIs) in non-small cell lung cancer, many patients have experienced increased overall survival. To improve selection of ICI responders versus nonresponders, microsatellite instability/mismatch repair deficiency (MSI/MMR) and tumor mutation burden (TMB) came into play. Clinical data show PD-L1, MSI/MMR, and TMB are independent predictive immunotherapy biomarkers. Harmonization of testing methodologies, optimization of assay design, and results analysis are ongoing. Future algorithms to determine immunotherapy eligibility might involve complementary use of current and novel biomarkers. Artificial intelligence could facilitate algorithm implementation to convert complex genetic data into recommendations for specific ICIs.
19Tumor mutational load predicts survival after immunotherapy across multiple cancer types.PubMed
Robert M Samstein, Chung-Han Lee, Alexander N Shoushtari, et al.
Nat Genet. 2019 Feb;51(2):202-206. doi: 10.1038/s41588-018-0312-8. Epub 2019 Jan 14.
Immune checkpoint inhibitor (ICI) treatments benefit some patients with metastatic cancers, but predictive biomarkers are needed. Findings in selected cancer types suggest that tumor mutational burden (TMB) may predict clinical response to ICI. To examine this association more broadly, we analyzed the clinical and genomic data of 1,662 advanced cancer patients treated with ICI, and 5,371 non-ICI-treated patients, whose tumors underwent targeted next-generation sequencing (MSK-IMPACT). Among all patients, higher somatic TMB (highest 20% in each histology) was associated with better overall survival. For most cancer histologies, an association between higher TMB and improved survival was observed. The TMB cutpoints associated with improved survival varied markedly between cancer types. These data indicate that TMB is associated with improved survival in patients receiving ICI across a wide variety of cancer types, but that there may not be one universal definition of high TMB.
20Genomic Features of Response to Combination Immunotherapy in Patients with Advanced Non-Small-Cell Lung Cancer.PubMed
Matthew D Hellmann, Tavi Nathanson, Hira Rizvi, et al.
Cancer Cell. 2018 May 14;33(5):843-852.e4. doi: 10.1016/j.ccell.2018.03.018. Epub 2018 Apr 12.
Combination immune checkpoint blockade has demonstrated promising benefit in lung cancer, but predictors of response to combination therapy are unknown. Using whole-exome sequencing to examine non-small-cell lung cancer (NSCLC) treated with PD-1 plus CTLA-4 blockade, we found that high tumor mutation burden (TMB) predicted improved objective response, durable benefit, and progression-free survival. TMB was independent of PD-L1 expression and the strongest feature associated with efficacy in multivariable analysis. The low response rate in TMB low NSCLCs demonstrates that combination immunotherapy does not overcome the negative predictive impact of low TMB. This study demonstrates the association between TMB and benefit to combination immunotherapy in NSCLC. TMB should be incorporated in future trials examining PD-(L)1 with CTLA-4 blockade in NSCLC.
21Prediction performance of twelve tumor mutation burden panels in melanoma and non-small cell lung cancer.PubMed
Dechen Xu, Jie Li, Dong Wang, et al.
Crit Rev Oncol Hematol. 2022 Jan;169:103573. doi: 10.1016/j.critrevonc.2021.103573. Epub 2021 Dec 18.
As a potential biomarker to predict the response to immunotherapy, tumor mutation burden (TMB) which can be estimated by the cancer gene panel (CGP) has received considerable attention. However, it is not clear which CGP is better in predicting the efficacy of immunotherapy. To evaluate the twelve CGPs, we compared them on 13 datasets of melanoma and non-small cell lung cancer (NSCLC) from the perspective of gene composition, reliability of measuring TMB and prediction performance of patient treatment benefits. The larger CGPs generally performed better, but their proportions of driver genes and function densities were smaller. The CGPs performed differently on melanoma and NSCLC patients treated with two blockades. Moreover, their ability to classify and predict patients with or without long-term clinical benefits was similar but not good enough, so it is necessary to explore a higher-performance biomarker.
22Extracellular vesicles and ctDNA in lung cancer: biomarker sources and therapeutic applications.PubMed
Chengliang Huang, Sitong Liu, Xiang Tong, et al.
Cancer Chemother Pharmacol. 2018 Aug;82(2):171-183. doi: 10.1007/s00280-018-3586-8. Epub 2018 Jun 8.
Lung cancer is the leading cause of cancer death in the world. Recently, targeted therapy and anti-programmed cell death receptor 1 (PD-1) and anti-programmed cell death ligand 1 (PD-L1) immunotherapy have made great progress in treatment of lung cancer. However, responses to these therapies are variable, influenced by genetic alterations, high microsatellite instability and mismatch repair deficiency. Liquid biopsy of extracellular vesicles and circulating tumor DNA (ctDNA) emerges as a new promising non-invasive means that enables not only biomarker determination, but also continuous monitoring of cancer treatment. Notably, tumor extracellular vesicles play important roles in tumor formation and progression, and also serve as natural carriers for anti-tumor drugs and short-interfering RNA. In this review, we summarize the latest progress in understanding the relationships of extracellular vesicles and ctDNA in cancer biology, diagnosis and drug delivery. In particular, the application of extracellular vesicles and ctDNA in anti-PD-1/PD-L1 immunotherapy is discussed.
23Pan-Cancer Prevalence of Microsatellite Instability-High (MSI-H) Identified by Circulating Tumor DNA and Associated Real-World Clinical Outcomes.PubMed
Pashtoon Murtaza Kasi, Leslie A Bucheit, Jiemin Liao, et al.
JCO Precis Oncol. 2023 Sep;7:e2300118. doi: 10.1200/PO.23.00118.
PURPOSE: Immune checkpoint inhibitors are approved for advanced solid tumors with microsatellite instability-high (MSI-H). Although several technologies can assess MSI-H status, detection and outcomes with circulating tumor DNA (ctDNA)-detected MSI-H are lacking. As such, we examined pan-cancer MSI-H prevalence across 21 cancers and outcomes after ctDNA-detected MSI-H. METHODS: Patients with advanced cancer who had ctDNA testing (Guardant360) from October 1, 2018, to June 30, 2022, were retrospectively assessed for prevalence. GuardantINFORM, which includes anonymized genomic and structured payer claims data, was queried to assess outcomes. Patients who initiated new treatment within 90 days of MSI-H detection were sorted into immunotherapy included in treatment (IO) or no immunotherapy included (non-IO) groups. Real-world time to treatment discontinuation (rwTTD) and real-world time to next treatment (rwTTNT) were assessed in months as proxies of progression-free survival (PFS); real-world overall survival (rwOS) was assessed in months. Cox regression tests analyzed differences. Colorectal cancer, non-small-cell lung cancer (NSCLC), prostate cancer, gastroesophageal cancer, and uterine cancer (UC) were assessed independently; all other cancers were grouped. RESULTS: In total, 1.4% of 171,881 patients had MSI-H detected. Of 770 patients with outcomes available, rwTTD and rwTTNT were significantly longer for patients who received IO compared with non-IO for all cancers ( ≤ .05; hazard ratio [HR] range, 0.31-0.52 and 0.25-0.54, respectively) except NSCLC. rwOS had limited follow-up for all cohorts except UC (IO 39 non-IO 23 months; HR, 0.18; = .004); however, there was a consistent trend toward prolonged OS in IO-treated patients. CONCLUSION: These data support use of a well-validated ctDNA assay to detect MSI-H across solid tumors and suggest prolonged PFS in patients treated with IO-containing regimens after detection. Tumor-agnostic, ctDNA-based MSI testing may be reliable for rapid decision making.
24TIL Therapy in Lung Cancer: Current Progress and Perspectives.PubMed
Weilei Hu, Yifei Bian, Hongbin Ji
Adv Sci (Weinh). 2024 Dec;11(46):e2409356. doi: 10.1002/advs.202409356. Epub 2024 Oct 18.
Lung cancer remains the most prevalent malignant tumor worldwide and is the leading cause of cancer-related mortality. Although immune checkpoint blockade has revolutionized the treatment of advanced lung cancer, many patients still do not respond well, often due to the lack of functional T cell infiltration. Adoptive cell therapy (ACT) using expanded immune cells has emerged as an important therapeutic modality. Tumor-infiltrating lymphocytes (TIL) therapy is one form of ACT involving the administration of expanded and activated autologous T cells derived from surgically resected cancer tissues and reinfusion into patients and holds great therapeutic potential for lung cancer. In this review, TIL therapy is introduced and its suitability for lung cancer is discussed. Then its historical and clinical developments are summarized, and the methods developed up-to-date to identify tumor-recognizing TILs and optimize TIL composition. Some perspectives toward future TIL therapy for lung cancer are also provided.
25Integrative multi-omic and machine learning approach for prognostic stratification and therapeutic targeting in lung squamous cell carcinoma.PubMed
Xiao Zhang, Pengpeng Zhang, Qianhe Ren, et al.
Biofactors. 2025 Jan-Feb;51(1):e2128. doi: 10.1002/biof.2128. Epub 2024 Oct 11.
The proliferation, metastasis, and drug resistance of cancer cells pose significant challenges to the treatment of lung squamous cell carcinoma (LUSC). However, there is a lack of optimal predictive models that can accurately forecast patient prognosis and guide the selection of targeted therapies. The extensive multi-omic data obtained from multi-level molecular biology provides a unique perspective for understanding the underlying biological characteristics of cancer, offering potential prognostic indicators and drug sensitivity biomarkers for LUSC patients. We integrated diverse datasets encompassing gene expression, DNA methylation, genomic mutations, and clinical data from LUSC patients to achieve consensus clustering using a suite of 10 multi-omics integration algorithms. Subsequently, we employed 10 commonly used machine learning algorithms, combining them into 101 unique configurations to design an optimal performing model. We then explored the characteristics of high- and low-risk LUSC patient groups in terms of the tumor microenvironment and response to immunotherapy, ultimately validating the functional roles of the model genes through in vitro experiments. Through the application of 10 clustering algorithms, we identified two prognostically relevant subtypes, with CS1 exhibiting a more favorable prognosis. We then constructed a subtype-specific machine learning model, LUSC multi-omics signature (LMS) based on seven key hub genes. Compared to previously published LUSC biomarkers, our LMS score demonstrated superior predictive performance. Patients with lower LMS scores had higher overall survival rates and better responses to immunotherapy. Notably, the high LMS group was more inclined toward "cold" tumors, characterized by immune suppression and exclusion, but drugs like dasatinib may represent promising therapeutic options for these patients. Notably, we also validated the model gene SERPINB13 through cell experiments, confirming its role as a potential oncogene influencing the progression of LUSC and as a promising therapeutic target. Our research provides new insights into refining the molecular classification of LUSC and further optimizing immunotherapy strategies.
26Heterogeneity in peripheral blood immune lymphocyte subsets predicts the response of immunotherapy or chemoradiotherapy in advanced lung cancer: an analysis across different pathological types, treatment modalities and age.PubMed
Chuanwang Miao, Yuanji Chen, Hao Zhang, et al.
Front Immunol. 2024 Oct 17;15:1464728. doi: 10.3389/fimmu.2024.1464728. eCollection 2024.
BACKGROUND: The shaping of the tumor immune microenvironment does not only rely on tumor-infiltrating lymphocytes but on the recruitment of lymphocytes in peripheral blood. Monitoring peripheral blood lymphocyte subsets level (PBLSL) can predict treatment response and prognosis with immune checkpoint inhibitors. This study investigated the heterogeneity of PBLSL in response to chemoradiotherapy (CRT) or combined with immunotherapy (CRIT) in advanced lung cancer patients. METHODS: 77 patients with advanced lung cancer receiving CRT or CRIT were divided into treatment-responsive and non-responsive groups based on efficacy. The study analyzed short-term efficacy and progression-free survival (PFS) according to baseline PBLSL and explored the impact under different stratifications, including treatment modality, pathology type, and age. RESULTS: In all patients, higher levels of B cells, higher CD4+/CD8+ T cell ratios, and lower CD8+ T cell levels were associated with better short-term outcomes (P = 0.0035, P = 0.044, P = 0.022). Subgroup analysis revealed that in the CRT group, higher B cell levels correlated with improved efficacy (P = 0.011) and superior PFS (P = 0.048, HR = 0.3886, 95% CI = 0.1696 to 0.8902). In the CRIT group, higher CD4+ T cell levels, lower CD8+ T cell levels, and higher CD4+/CD8+ T cell ratios were linked to better efficacy (P = 0.038, P = 0.047, P = 0.017). For adenocarcinoma patients, higher CD4+/CD8+ T cell ratios and lower CD8+ T cell levels predicted better efficacy (P = 0.0155, P = 0.0119). B cell levels were significant in squamous cell carcinoma (P = 0.0291), while no PBLSL was predictive for small cell lung cancer. Among patients under 65, higher B cell levels were linked to improved efficacy and prolonged PFS (P = 0.0036, P = 0.0332, HR = 0.4111, 95% CI = 0.1973 to 0.8563). For patients over 65, differences in CD4+ T cell levels and CD4+/CD8+ T cell ratios were significant (P = 0.0433, P = 0.0338). CONCLUSION: PBLSL predicted efficacy and prognosis in various patient stratifications, suggesting PBLSL is a reliable predictor for CRT and CRIT in advanced lung cancer. Detecting different cellular subpopulations helps identify patients with significant treatment responses across different stratifications.
27Multi-omics indicators of long-term survival benefits after immune checkpoint inhibitor therapy.PubMed
Jie Zhao, Yiting Dong, Hua Bai, et al.
Cell Rep Methods. 2023 Oct 23;3(10):100596. doi: 10.1016/j.crmeth.2023.100596. Epub 2023 Sep 21.
Molecular indicators of long-term survival (LTS) in response to immune-checkpoint inhibitor (ICI) treatment have the potential to provide both mechanistic and therapeutic insights. In this study, we construct predictive models of LTS following ICI therapy based on data from 158 clinical trials involving 21,023 patients of 25 cancer types with available 1-year overall survival (OS) rates. We present evidence for the use of 1-year OS rate as a surrogate for LTS. Based on these and corresponding TCGA multi-omics data, total neoantigen, metabolism score, CD8 T cell, and MHC_score were identified as predictive biomarkers. These were integrated into a Gaussian process regression model that estimates "long-term survival predictive score of immunotherapy" (iLSPS). We found that iLSPS outperformed the predictive capabilities of individual biomarkers and successfully predicted LTS of patient groups with melanoma and lung cancer. Our study explores the feasibility of modeling LTS based on multi-omics indicators and machine-learning methods.
28SMARCA4 mutations and expression in lung adenocarcinoma: prognostic significance and impact on the immunotherapy response.PubMed
Yuming Zhang, Dantong Sun, Weizhong Han, et al.
FEBS Open Bio. 2024 Dec;14(12):2086-2103. doi: 10.1002/2211-5463.13899. Epub 2024 Sep 25.
The switch/sucrose non-fermenting (SWI/SNF) complex family includes important chromatin-remodeling factors that are frequently mutated in lung adenocarcinoma (LUAD). However, the role of one family member, SMARCA4, in LUAD prognosis and immunotherapy sensitivity remains unclear. In the present study, 6745 LUAD samples from the cBioPortal database were used to analyze the relationships between SMARCA4 mutations and patient prognoses and clinical characteristics. Additionally, we examined the correlation between SMARCA4 mutations and prognosis in patients treated with immunotherapy using two immune-related datasets. SMARCA4 mutations and low expression were associated with shorter survival, and mutations were associated with a high tumor mutational burden and high microsatellite instability. SMARCA4 mutations were accompanied by KRAS, KEAP1, TP53 and STK11 mutations. No significant difference was observed in the immunotherapy response between patients with and without SMARCA4 mutations. When KRAS or STK11 mutations were present, immunotherapy effectiveness was poorer; however, when both SMARCA4 and TP53 mutations were present, immunotherapy was more effective. Furthermore, low SMARCA4 expression predicted a higher immunophenoscore, and SMARCA4 expression was correlated with certain immune microenvironment features. Taken together, our results suggest that SMARCA4 mutations and low expression might be associated with poor LUAD prognosis. Additionally, immunotherapy efficacy in patients with SMARCA4 mutations depended on the co-mutant genes. Thus, SMARCA4 could be an important factor to be considered for LUAD diagnosis and treatment.
29Predicting response and toxicity to immune checkpoint inhibitors in lung cancer using antibodies to frameshift neoantigens.PubMed
Luhui Shen, Justin R Brown, Stephen Albert Johnston, et al.
J Transl Med. 2023 May 22;21(1):338. doi: 10.1186/s12967-023-04172-w.
PURPOSE: To evaluate a new class of blood-based biomarkers, anti-frameshift peptide antibodies, for predicting both tumor responses and adverse immune events to immune checkpoint inhibitor (ICI) therapies in advanced lung cancer patients. EXPERIMENTAL DESIGN: Serum samples were obtained from 74 lung cancer patients prior to palliative PD-(L)1 therapies with subsequently recorded tumor responses and immune adverse events (irAEs). Pretreatment samples were assayed on microarrays of frameshift peptides (FSPs), representing ~ 375,000 variant peptides that tumor cells can be informatically predicted to produce from translated mRNA processing errors. Serum-antibodies specifically recognizing these ligands were measured. Binding activities preferentially associated with best-response and adverse-event outcomes were determined. These antibody bound FSPs were used in iterative resampling analyses to develop predictive models of tumor response and immune toxicity. RESULTS: Lung cancer serum samples were classified based on predictive models of ICI treatment outcomes. Disease progression was predicted pretreatment with ~ 98% accuracy in the full cohort of all response categories, though ~ 30% of the samples were indeterminate. This model was built with a heterogeneous sample cohort from patients that (i) would show either clear response or stable outcomes, (ii) would be administered either single or combination therapies and (iii) were diagnosed with different lung cancer subtypes. Removing the stable disease, combination therapy or SCLC groups from model building increased the proportion of samples classified while performance remained high. Informatic analyses showed that several of the FSPs in the all-response model mapped to translations of variant mRNAs from the same genes. In the predictive model for treatment toxicities, binding to irAE-associated FSPs provided 90% accuracy pretreatment, with no indeterminates. Several of the classifying FSPs displayed sequence similarity to self-proteins. CONCLUSIONS: Anti-FSP antibodies may serve as biomarkers for predicting ICI outcomes when tested against ligands corresponding to mRNA-error derived FSPs. Model performances suggest this approach might provide a single test to predict treatment response to ICI and identify patients at high risk for immunotherapy toxicities.
30Real-world data on immunotherapy combined with chemotherapy in elderly patients with extensive-stage small cell lung cancer.PubMed
Ke Zhao, Shuangqing Lu, Jiling Niu, et al.
BMC Cancer. 2025 Mar 14;25(1):467. doi: 10.1186/s12885-025-13880-z.
BACKGROUND: Immunotherapy combined with chemotherapy has shown good results in the treatment of extensive-stage small cell lung cancer (ES-SCLC), but there are fewer clinical studies on elderly ES-SCLC patients. This study was aimed to evaluate the efficacy and safety of immunotherapy in combination with chemotherapy in elderly patients with ES-SCLC. METHODS: Elderly patients with ES-SCLC who were 70 years of age or older and were diagnosed at Shandong Cancer Hospital from May 20, 2020, to February 24, 2023, were included in this study. Overall survival (OS) and progression-free survival (PFS) were calculated via the Kaplan‒Meier method and compared via the log-rank test. In addition, the Cox regression model was used to analyze prognostic factors. RESULTS: A total of 135 patients were included in this study; 82 patients were in the immunotherapy combined with chemotherapy (IO + ChT) group, and 53 patients were in the chemotherapy alone (ChT-alone) group. The median overall survival (mOS) for the entire patient cohort was 12.89 months, whereas the median progression-free survival (mPFS) was 7.21 months. There was a significant difference in mPFS (8.26 months vs. 6.59 months, P =.02) and no statistically significant difference in mOS (14.20 months vs. 11.44 months, P =.14) between the IO + ChT and ChT-alone groups. The incidence of grade ≥ 3 adverse events in the IO + ChT group was not significantly different from that in the ChT-alone group. Moreover, we did not observe grade ≥ 3 immune-related adverse reactions. The univariate multifactorial analysis demonstrated that the absence of liver metastases at baseline and in female patients were favorable prognostic factors for OS, and the addition of immunotherapy was a favorable prognostic factor that improved overall survival in elderly patients with ES-SCLC. Subgroup analyses indicated that adding immunotherapy provided a survival benefit for patients with baseline brain metastases and baseline liver-free metastases. CONCLUSION: Immunotherapy combined with chemotherapy can provide a survival benefit, and the addition of immunotherapy does not result in significant toxicity in elderly patients. The results of this study have important clinical implications for the future treatment of elderly patients with ES-SCLC.
31Factors associated with overall survival, progression-free survival and toxicity in patients with small cell lung cancer and thoracic irradiation in a clinical real-world setting.PubMed
Marie-Theres Kassik, Dirk Vordermark, Christine Kornhuber, et al.
Radiat Oncol. 2023 Apr 18;18(1):70. doi: 10.1186/s13014-023-02252-1.
BACKGROUND: Small-cell lung cancer (SCLC) is a malignant tumor known for its poor prognosis. In addition to chemotherapy and immunotherapy irradiation plays a big role especially in inoperability. This study evaluated prognostic factors in patients with SCLC, receiving chemotherapy and thoracic irradiation, that may affect overall survival (OS), progression-free survival (PFS) and toxicity. METHODS: Patients with limited disease (LD) SCLC (n = 57) and extensive disease (ED) SCLC (n = 69) who received thoracic radiotherapy were analyzed retrospectively. The prognostic factors sex, age, Karnofsky performance status (KPS), tumor-, nodal-stage and timepoint of start of irradiation in relation to the first cycle of chemotherapy were evaluated. Start of irradiation was stratified as early ([Formula: see text] 2 cycles of chemotherapy), late (3 or 4 cycles) and very late ([Formula: see text] 5 cycles). Results were analyzed by Cox univariate and multivariate as well as logistic regression analysis. RESULTS: The median OS of LD-SCLC patients was 23.7 months in early, and 22.0 months in late start of irradiation. In very late start, median OS was not reached. PFS was 11.8, 15.2 and 47.9 months, respectively. In patients with ED-SCLC OS was 4.3 months in early, 13.0 months in late and 12.2 months in very late start of irradiation. PFS was 6.7, 13.0 and 12.2 months, respectively. Prognosis of patients with LD- or ED-SCLC receiving late or very late start of irradiation was significantly prolonged in OS and PFS compared to an early start (p < 0.05). KPS [Formula: see text] 80 shows a significant increase of OS and PFS in ED-SCLC. Female sex and smaller mean lung dose were associated with lower risk of toxicity. CONCLUSION: Late or very late start of irradiation is a prognosis-enhancing factor in LD-SCLC and ED-SCLC for OS and PFS. KPS [Formula: see text] 80 increases prognosis of OS and PFS in ED-SCLC as well. Toxicity is less common in female sex and patients with low mean lung dose in LD-SCLC.
32Comparison of efficacy and safety of PD-1/PD-L1 combination therapy in first-line treatment of advanced NSCLC: an updated systematic review and network meta-analysis.PubMed
Yanqing Yang, Wei Chen, Lixian Dong, et al.
Clin Transl Oncol. 2024 Oct;26(10):2488-2502. doi: 10.1007/s12094-024-03442-3. Epub 2024 Apr 16.
BACKGROUND: The use of immune checkpoint inhibitors has led to an increase in randomized controlled trials exploring various first-line combination treatment regimens. With the introduction of new PD-1/PD-L1 inhibitors, there are now more clinical options available. For the first time, the AK105 monoclonal antibody Penpulimab, developed in China, was included. The AK105-302 Phase III trial studied the efficacy and safety of Penpulimab combined with chemotherapy in patients with advanced or metastatic squamous NSCLC. To determine the optimal treatment options, we conducted an updated network meta-analysis to compare the effectiveness and safety of these regimens. METHODS: The system retrieves data from Chinese and English electronic databases, Clinical Trials, and the gov Clinical Trial Registration website up to September 6, 2023. The study indirectly compared the efficacy and safety of PD-1/PD-L1 combination regimens, including overall survival (OS), progression-free survival (PFS), objective response rate (ORR), all-grade adverse events, and above-grade III adverse events. Subgroup analyses were conducted based on programmed death ligand 1 (PD-L1) level, histological type, ECOG score, sex, and smoking history. RESULTS: Nineteen RCTS were included, with a total of ten thousand eight hundred patients. Penpulimab plus chemotherapy (Pen + CT) provided the best OS (HR = 0.55, 95% CI 0.38-0.81) for PD-L1 patients with non-selective advanced NSCLC. Except Nivolumab plus Ipilimumab (Niv + Ipi), other PD-1/PD-L1 combination therapies significantly extended PFS compared with CT, and Nivolumab plus Bevacizumab combined with chemotherapy (Niv + Bev + CT) (HR = 0.43, 95% CI 0.26-0.74) provided the best PFS benefit and was comparable to Pen + CT (HR = 1.0) for PFS prolongation. For ORR, except Niv + Ipi, all the other regimens significantly improved ORR compared with CT. In terms of safety, except Tor + CT, the incidence of any-grade AEs or grade ≥ 3 adverse events may be higher than those of chemotherapy. The subgroup analysis revealed that for patients with PD-L1 levels below 1%, treatment with Tor + CT resulted in the best progression-free survival (HR = 0.47, 95% CI 0.25-0.86). For patients with PD-L1 levels of 1% or higher, Sintilimab plus chemotherapy (Sin + CT) (HR = 0.56, 95% CI 0.31-0.99) and Camrelizumab plus chemotherapy (Cam + CT) (HR = 0.43, 95% CI 0.28-0.64) were associated with the best overall survival and progression-free survival, respectively. For patients with SqNSCLC, combined immunotherapy may provide greater survival benefits. For patients with Non-sqNSCLC, Niv + Bev + CT and Tor + CT were associated with optimal PFS and OS, respectively. Cam + CT provided the best PFS in male patients with a history of smoking and an ECOG score of 0. In both female and non-smoking patient subgroups, Pem + CT was associated with the best PFS and OS benefits. CONCLUSION: For patients with advanced non-selective PD-L1 NSCLC, two effective regimens are Pen + CT and Niv + Bev + CT, which rank first in OS and PFS among all patients. Cam + CT and Tor + CT have advantages for OS in patients with SqNSCLC and Non-sqNSCLC, respectively. Niv + Ipi + CT provided the best OS benefit for patients with an ECOG score of 0, while Pem + CT may be the most effective treatment for patients with an ECOG score of 1. Pem + CT has a better effect on female patients and non-smokers. Sin + CT was found to be the most effective treatment for male patients and the smoking subgroup, while Cam + CT was found to be the most effective for PFS. In addition, Tor + CT was associated with the best PFS for patients with negative PD-L1 expression. Pem + CT was found to significantly improve both PFS and OS compared to CT alone. For patients with positive PD-L1 expression, Sin + CT and Cam + CT were found to be optimal for OS and PFS, respectively. It is important to note that, with the exception of Tor + CT, the toxicity of the other combinations was higher than that of CT alone.
33Predictive biomarkers for PD-1/PD-L1 checkpoint inhibitor response in NSCLC: an analysis of clinical trial and real-world data.PubMed
WeiQing Venus So, David Dejardin, Eva Rossmann, et al.
J Immunother Cancer. 2023 Feb;11(2). doi: 10.1136/jitc-2022-006464.
BACKGROUND: Many biomarkers have been proposed to be predictive of response to anti-programmed cell death protein-1 (PD-1)/anti-programmed death ligand-1 (PD-L1) checkpoint inhibitors (CPI). However, conflicting observations and lack of consensus call for an assessment of their clinical utility in a large data set. Using a combined data set of clinical trials and real-world data, we assessed the predictive and prognostic utility of biomarkers for clinical outcome of CPI in non-small cell lung cancer (NSCLC). METHODS: Retrospective cohort study using 24,152 patients selected from 71,850 patients with advanced NSCLC from electronic health records and 9 Roche atezolizumab trials. Patients were stratified into high and low biomarker groups. Correlation with treatment outcome in the different biomarker groups was investigated and compared between patients treated with CPI versus chemotherapy. Durable response was defined as having complete response/partial response without progression during the study period of 270 days. RESULTS: Standard blood analytes (eg, albumin and lymphocyte) were just prognostic, having correlation with clinical outcome irrespective of treatment type. High expression of PD-L1 on tumors (≥50% tumor cell staining) were specifically associated with response to CPI (OR 0.20; 95% CI 0.13 to 0.30; p<0.001). The association was stronger in patients with non-squamous than squamous histology, with smoking history than non-smokers, and with prior chemotherapy than first-line CPI. Higher tumor mutational burden (TMB) (≥10.44 mut/Mb) was also specifically associated with durable response to CPI (OR=0.40; 95% CI 0.29 to 0.54; p<0.001). The combination of high TMB and PD-L1 expression was the strongest predictor of durable response to CPI (OR=0.04; 95% CI 0.00 to 0.18; p<0.001). There was no significant association between PD-L1 or TMB levels with response to chemotherapy, suggesting a CPI-specific predictive effect. CONCLUSIONS: Standard blood analytes had just prognostic utility, whereas tumor PD-L1 and TMB specifically predicted response to CPI in NSCLC. The combined high TMB and PD-L1 expression was the strongest predictor of durable response. PD-L1 was also a stronger predictor in patients with non-squamous histology, smoking history or prior chemotherapy.
34Previous radiotherapy and the clinical activity and toxicity of pembrolizumab in the treatment of non-small-cell lung cancer: a secondary analysis of the KEYNOTE-001 phase 1 trial.PubMed
Narek Shaverdian, Aaron E Lisberg, Krikor Bornazyan, et al.
Lancet Oncol. 2017 Jul;18(7):895-903. doi: 10.1016/S1470-2045(17)30380-7. Epub 2017 May 24.
BACKGROUND: Preclinical studies have found radiotherapy enhances antitumour immune responses. We aimed to assess disease control and pulmonary toxicity in patients who previously received radiotherapy for non-small-cell lung cancer (NSCLC) before receiving pembrolizumab. METHODS: We assessed patients with advanced NSCLC treated on the phase 1 KEYNOTE-001 trial at a single institution (University of California, Los Angeles, CA, USA). Patients were aged 18 years or older, had an Eastern Cooperative Oncology Group performance status of 1 or less, had adequate organ function, and no history of pneumonitis. Patients received pembrolizumab at a dose of either 2 mg/kg of bodyweight or 10 mg/kg every 3 weeks, or 10 mg/kg every 2 weeks, until disease progression, unacceptable toxicity, or other protocol-defined reasons for discontinuation. Disease response and pulmonary toxicity were prospectively assessed by Immune-related Response Criteria and Common Terminology Criteria for Adverse Events version 4.0. The primary objective of the KEYNOTE-001 trial was to assess the safety, side-effect profile, and antitumour activity of pembrolizumab. For our secondary analysis, patients were divided into subgroups to compare patients who previously received radiotherapy with patients who had not. Our primary objective was to determine whether previous radiotherapy affected progression-free survival, overall survival, and pulmonary toxicity in the intention-to-treat population. The KEYNOTE-001 trial was registered with ClinicalTrials.gov, number NCT01295827. FINDINGS: Between May 22, 2012, and July 11, 2014, 98 patients were enrolled and received their first cycle of pembrolizumab. One patient was lost to follow-up. 42 (43%) of 97 patients had previously received any radiotherapy for the treatment of NSCLC before the first cycle of pembrolizumab. 38 (39%) of 97 patients received extracranial radiotherapy and 24 (25%) of 97 patients received thoracic radiotherapy. Median follow-up for surviving patients was 32·5 months (IQR 29·8-34·1). Progression-free survival with pembrolizumab was significantly longer in patients who previously received any radiotherapy than in patients without previous radiotherapy (hazard ratio [HR] 0·56 [95% CI 0·34-0·91], p=0·019; median progression-free survival 4·4 months [95% CI 2·1-8·6] vs 2·1 months [1·6-2·3]) and for patients who previously received extracranial radiotherapy compared with those without previous extracranial radiotherapy (HR 0·50 [0·30-0·84], p=0·0084; median progression-free survival 6·3 months [95% CI 2·1-10·4] vs 2·0 months [1·8-2·1]). Overall survival with pembrolizumab was significantly longer in patients who previously received any radiotherapy than in patients without previous radiotherapy (HR 0·58 [95% CI 0·36-0·94], p=0·026; median overall survival 10·7 months [95% CI 6·5-18·9] vs 5·3 months [2·7-7·7]) and for patients who previously received extracranial radiotherapy compared with those without previous extracranial radiotherapy (0·59 [95% CI 0·36-0·96], p=0·034; median overall survival 11·6 months [95% CI 6·5-20·5] vs 5·3 months [3·0-8·5]). 15 (63%) of 24 patients who had previously received thoracic radiotherapy had any recorded pulmonary toxicity versus 29 (40%) of 73 patients with no previous thoracic radiotherapy. Three (13%) patients with previous thoracic radiotherapy had treatment-related pulmonary toxicity compared with one (1%) of those without; frequency of grade 3 or worse treatment-related pulmonary toxicities was similar (one patient in each group). INTERPRETATION: Our data suggest that previous treatment with radiotherapy in patients with advanced NSCLC results in longer progression-free survival and overall survival with pembrolizumab treatment than that seen in patients who did not have previous radiotherapy, with an acceptable safety profile. Further clinical trials investigating this combination are needed to determine the optimal treatment strategy for patients with advanced NSCLC. FUNDING: US National Institutes of Health.
35Immunogenic Chemotherapy Enhances Recruitment of CAR-T Cells to Lung Tumors and Improves Antitumor Efficacy when Combined with Checkpoint Blockade.PubMed
Shivani Srivastava, Scott N Furlan, Carla A Jaeger-Ruckstuhl, et al.
Cancer Cell. 2021 Feb 8;39(2):193-208.e10. doi: 10.1016/j.ccell.2020.11.005. Epub 2020 Dec 24.
Adoptive therapy using chimeric antigen receptor-modified T cells (CAR-T cells) is effective in hematologic but not epithelial malignancies, which cause the greatest mortality. In breast and lung cancer patients, CAR-T cells targeting the tumor-associated antigen receptor tyrosine kinase-like orphan receptor 1 (ROR1) infiltrate tumors poorly and become dysfunctional. To test strategies for enhancing efficacy, we adapted the Kras;p53 autochthonous model of lung adenocarcinoma to express the CAR target ROR1. Murine ROR1 CAR-T cells transferred after lymphodepletion with cyclophosphamide (Cy) transiently control tumor growth but infiltrate tumors poorly and lose function, similar to what is seen in patients. Adding oxaliplatin (Ox) to the lymphodepletion regimen activates tumor macrophages to express T-cell-recruiting chemokines, resulting in improved CAR-T cell infiltration, remodeling of the tumor microenvironment, and increased tumor sensitivity to anti-PD-L1. Combination therapy with Ox/Cy and anti-PD-L1 synergistically improves CAR-T cell-mediated tumor control and survival, providing a strategy to improve CAR-T cell efficacy in the clinic.
36Sotorasib versus docetaxel for previously treated non-small-cell lung cancer with KRAS mutation: a randomised, open-label, phase 3 trial.PubMed
Adrianus Johannes de Langen, Melissa L Johnson, Julien Mazieres, et al.
Lancet. 2023 Mar 4;401(10378):733-746. doi: 10.1016/S0140-6736(23)00221-0. Epub 2023 Feb 7.
BACKGROUND: Sotorasib is a specific, irreversible inhibitor of the GTPase protein, KRAS. We compared the efficacy and safety of sotorasib with a standard-of-care treatment in patients with non-small-cell lung cancer (NSCLC) with the KRAS mutation who had been previously treated with other anticancer drugs. METHODS: We conducted a randomised, open-label phase 3 trial at 148 centres in 22 countries. We recruited patients aged at least 18 years with KRAS-mutated advanced NSCLC, who progressed after previous platinum-based chemotherapy and a PD-1 or PD-L1 inhibitor. Key exclusion criteria included new or progressing untreated brain lesions or symptomatic brain lesions, previously identified oncogenic driver mutation other than KRAS for which an approved therapy is available (eg EGFR or ALK), previous treatment with docetaxel (neoadjuvant or adjuvant docetaxel was allowed if the tumour did not progress within 6 months after the therapy was terminated), previous treatment with a direct KRAS inhibitor, systemic anticancer therapy within 28 days of study day 1, and therapeutic or palliative radiation therapy within 2 weeks of treatment initiation. We randomly assigned (1:1) patients to oral sotorasib (960 mg once daily) or intravenous docetaxel (75 mg/m once every 3 weeks) in an open-label manner using interactive response technology. Randomisation was stratified by number of previous lines of therapy in advanced disease (1 vs 2 vs >2), ethnicity (Asian vs non-Asian), and history of CNS metastases (present or absent). Treatment continued until an independent central confirmation of disease progression, intolerance, initiation of another anticancer therapy, withdrawal of consent, or death, whichever occurred first. The primary endpoint was progression-free survival, which was assessed by a blinded, independent central review in the intention-to-treat population. Safety was assessed in all treated patients. This trial is registered at ClinicalTrials.gov, NCT04303780, and is active but no longer recruiting. FINDINGS: Between June 4, 2020, and April 26, 2021, 345 patients were randomly assigned to receive sotorasib (n=171 [50%]) or docetaxel (n=174 [50%]). 169 (99%) patients in the sotorasib group and 151 (87%) in the docetaxel group received at least one dose. After a median follow-up of 17·7 months (IQR 16·4-20·1), the study met its primary endpoint of a statistically significant increase in the progression-free survival for sotorasib, compared with docetaxel (median progression-free survival 5·6 months [95% CI 4·3-7·8] vs 4·5 months [3·0-5·7]; hazard ratio 0·66 [0·51-0·86]; p=0·0017). Sotorasib was well tolerated, with fewer grade 3 or worse (n=56 [33%] vs n=61 [40%]) and serious treatment-related adverse events compared with docetaxel (n=18 [11%] vs n=34 [23%]). For sotorasib, the most common treatment-related adverse events of grade 3 or worse were diarrhoea (n= 20 [12%]), alanine aminotransferase increase (n=13 [8%]), and aspartate aminotransferase increase (n=9 [5%]). For docetaxel, the most common treatment-related adverse events of grade 3 or worse were neutropenia (n=13 [9%]), fatigue (n=9 [6%]), and febrile neutropenia (n=8 [5%]). INTERPRETATION: Sotorasib significantly increased progression-free survival and had a more favourable safety profile, compared with docetaxel, in patients with advanced NSCLC with the KRAS mutation and who had been previously treated with other anticancer drugs. FUNDING: Amgen.
37Integration of clinical and blood parameters in risk prognostication for patients receiving immunochemotherapy for extensive stage small cell lung cancer: real-world data from two centers.PubMed
Xiaomi Li, Li Tong, Shan Wang, et al.
BMC Med. 2024 Sep 11;22(1):381. doi: 10.1186/s12916-024-03612-8.
BACKGROUND: Immune checkpoint inhibitors (ICIs) had modest advances in the treatment of extensive-stage small cell lung cancer (ES-SCLC) in clinical trials, but there is a lack of biomarkers for prognosis in clinical practice. METHODS: We retrospectively collected data from ES-SCLC patients who received ICIs combined chemotherapy from two centers in China, integrated clinical and blood parameters, and constructed risk prognostication for immunochemotherapy. The population was divided into high- and low-risk groups, and the performance of the model was assessed separately in the training and validation cohorts. RESULTS: Two hundred and twenty and 43 patients were included in the training and validation groups, respectively. The important predictors were screened including body mass index, liver metastases, coefficient variation of red blood cell distribution width, lactate dehydrogenase, albumin, and C-reactive protein. Predicting 1-year overall survival (OS), the AUC values under ROC for the model under training, internal validation, and external validation were 0.760, 0.732, and 0.722, respectively, and the calibration curve and clinical decision curve performed well. Applied the model to divide patients into low-risk and high-risk groups, and the median OS was 23.7 months and 9.1 months, and the median progression-free survival was 8.2 months and 4.8 months, respectively; furthermore, this ability to discriminate survival was also observed in the validation cohort. CONCLUSIONS: We constructed a novel prognostic model for ES-SCLC to predict survival employing baseline tumor burden, nutritional and inflammatory parameters, it is easily measured to screen high-risk patient populations.
38Investigation of poor predictive factors in extensive stage small cell lung cancer under etoposide-platinum-atezolizumab treatment.PubMed
Jeong Uk Lim, Hye Seon Kang, Ah Young Shin, et al.
Thorac Cancer. 2022 Dec;13(23):3384-3392. doi: 10.1111/1759-7714.14697. Epub 2022 Oct 23.
BACKGROUND: The phase III trial IMpower133 showed that platinum and etoposide plus atezolizumab was associated with improved overall survival (OS) and progression free-survival (PFS) when compared to the placebo group in treatment-naïve extensive stage (ES) small cell lung cancer (SCLC). Due to superiority in clinical outcomes, combination immunotherapy plus chemotherapy have become mainstay treatment modalities as first-line treatment in ES-SCLC. Nevertheless, real-world data are still lacking and the search for potential biomarkers is essential. This study aimed to evaluate potential predictive biomarkers applicable in ES-SCLC under combination therapy. METHODS: Patients with ES-SCLC under etoposide-platinum-atezolizumab enrolled from seven university hospitals affiliated to the Catholic University of Korea were evaluated. Pretreatment clinical parameters were evaluated for association with OS and PFS. Adverse events (AEs) during induction and maintenance phases were also evaluated. p-values below 0.05 were considered statistically significant. RESULTS: A total of 41 patients were evaluated. Six-month survival was 68.6%. As best response to treatment, 26 (63.4%) showed partial response, nine (22.0%) showed stable disease, and four (9.8%) showed progressive disease. During the induction phase, grade I-II AEs occurred in 22 (53.7%) patients, and grade III-IV AEs occurred in 26 (63.4%) patients. During the maintenance phase, nine out of 25 (36.0%) patients experienced any grade AEs. In multivariate analysis for OS, lactate dehydrogenase (LDH), c-reactive protein (CRP), and forced vital capacity (%) were significant factors. In multivariate analysis for PFS, sex, and LDH were significant. CONCLUSION: In ES-SCLC under etoposide-platinum-atezolizumab, pretreatment CRP, LDH and FVC (%) were independent predictive factors.
39Development of nomograms to predict therapeutic response and prognosis of non-small cell lung cancer patients treated with anti-PD-1 antibody.PubMed
Shijin Yuan, Yan Xia, Lihong Shen, et al.
Cancer Immunol Immunother. 2021 Feb;70(2):533-546. doi: 10.1007/s00262-020-02710-9. Epub 2020 Aug 27.
BACKGROUND: Anti-programmed death-1 (PD-1) antibody changed the treatment of non-small cell lung cancer (NSCLC), however, reliable predictive markers were lacking. We aimed to explore factors associated with response and survival, and develop predictive models. METHODS: This multicenter retrospective study included a training cohort (n = 92) and validation cohort (n = 111) with NSCLC patients received anti-PD-1 antibody monotherapy in eight Chinese hospitals, and a control cohort (n = 124) with NSCLC patients received chemotherapy. Logistic and Cox models were used to identify factors associated with response and survival respectively. Nomograms were developed based on significant factors, and evaluated by Concordance-index (C-index), area under the curve (AUC) and calibration curve. RESULT: In training cohort, smoking history (P = 0.027) and higher absolute lymphocyte count (P = 0.038) were associated with response. Female (P < 0.001), age ≥ 65 years (P = 0.004) and higher lactate dehydrogenase (LDH, P < 0.001) were associated with shorter progression-free survival (PFS). Higher LDH (P < 0.001) and derived neutrophil-to-lymphocyte ratio (P = 0.035) were associated with poorer overall survival (OS). While these factors were nonsignificant in chemotherapy cohort. Three nomograms to predict response at 6-week after treatment, PFS and OS at 6-, 12- and 18-months were developed, and validated in validation cohort. The C-indices of each nomogram in both cohorts were as follow (training vs validation): 0.706 vs 0.701; 0.728 vs 0.701; 0.741 vs 0.709; respectively. AUC showed a good discriminative ability. Calibration curves demonstrated a consistence between actual results and predictions. CONCLUSION: We developed predictive nomograms based on easily available factors to help clinicians early assess response and prognosis for NSCLC patients received anti-PD-1 antibody.
40Artificial intelligence for predictive biomarker discovery in immuno-oncology: a systematic review.PubMed
A Prelaj, V Miskovic, M Zanitti, et al.
Ann Oncol. 2024 Jan;35(1):29-65. doi: 10.1016/j.annonc.2023.10.125. Epub 2023 Oct 23.
BACKGROUND: The widespread use of immune checkpoint inhibitors (ICIs) has revolutionised treatment of multiple cancer types. However, selecting patients who may benefit from ICI remains challenging. Artificial intelligence (AI) approaches allow exploitation of high-dimension oncological data in research and development of precision immuno-oncology. MATERIALS AND METHODS: We conducted a systematic literature review of peer-reviewed original articles studying the ICI efficacy prediction in cancer patients across five data modalities: genomics (including genomics, transcriptomics, and epigenomics), radiomics, digital pathology (pathomics), and real-world and multimodality data. RESULTS: A total of 90 studies were included in this systematic review, with 80% published in 2021-2022. Among them, 37 studies included genomic, 20 radiomic, 8 pathomic, 20 real-world, and 5 multimodal data. Standard machine learning (ML) methods were used in 72% of studies, deep learning (DL) methods in 22%, and both in 6%. The most frequently studied cancer type was non-small-cell lung cancer (36%), followed by melanoma (16%), while 25% included pan-cancer studies. No prospective study design incorporated AI-based methodologies from the outset; rather, all implemented AI as a post hoc analysis. Novel biomarkers for ICI in radiomics and pathomics were identified using AI approaches, and molecular biomarkers have expanded past genomics into transcriptomics and epigenomics. Finally, complex algorithms and new types of AI-based markers, such as meta-biomarkers, are emerging by integrating multimodal/multi-omics data. CONCLUSION: AI-based methods have expanded the horizon for biomarker discovery, demonstrating the power of integrating multimodal data from existing datasets to discover new meta-biomarkers. While most of the included studies showed promise for AI-based prediction of benefit from immunotherapy, none provided high-level evidence for immediate practice change. A priori planned prospective trial designs are needed to cover all lifecycle steps of these software biomarkers, from development and validation to integration into clinical practice.
41Advances in Multimodal Imaging Techniques for Evaluating and Predicting the Efficacy of Immunotherapy for NSCLC.PubMed
Jingyi Liu, Mei Xie, Jing Shen, et al.
Cancer Manag Res. 2025 Jun 7;17:1073-1086. doi: 10.2147/CMAR.S522136. eCollection 2025.
Immunotherapy has emerged as a transformative treatment for non-small cell lung cancer (NSCLC), yet its clinical benefits remain variable among patients. Early and accurate evaluation of treatment response is critical to guide therapeutic adjustments and improve outcomes. This review synthesizes recent advancements in multimodal imaging techniques-computed tomography (CT), positron emission tomography (PET)/CT, magnetic resonance imaging (MRI), and radiomics-for evaluating and predicting immunotherapy efficacy in NSCLC. We analyze the strengths and limitations of conventional morphological criteria (eg, RECIST, iRECIST) and highlight emerging quantitative biomarkers, including CT texture analysis, metabolic parameters (MTV, TLG), and diffusion-weighted MRI metrics. Notably, radiomics demonstrates promise in decoding tumor heterogeneity, PD-L1 expression, and immune microenvironment features, while immuno-PET probes targeting immune checkpoints offer novel insights into immune activity in vivo. Challenges such as pseudo-progression, nodal immune flare, and discrepancies between imaging responses and pathological responses are critically discussed. By integrating morphological, metabolic, and microenvironmental data, multimodal imaging enhances precision in patient stratification and therapeutic monitoring. Future research should prioritize multicenter, AI-driven radiomics validation and targeted tracer development to optimize NSCLC immunotherapy management. This review provides clinicians and researchers with new directions for utilizing multimodal imaging techniques in developing personalized treatment strategies.
42Predicting response to cancer immunotherapy using noninvasive radiomic biomarkers.PubMed
S Trebeschi, S G Drago, N J Birkbak, et al.
Ann Oncol. 2019 Jun 1;30(6):998-1004. doi: 10.1093/annonc/mdz108.
INTRODUCTION: Immunotherapy is regarded as one of the major breakthroughs in cancer treatment. Despite its success, only a subset of patients responds-urging the quest for predictive biomarkers. We hypothesize that artificial intelligence (AI) algorithms can automatically quantify radiographic characteristics that are related to and may therefore act as noninvasive radiomic biomarkers for immunotherapy response. PATIENTS AND METHODS: In this study, we analyzed 1055 primary and metastatic lesions from 203 patients with advanced melanoma and non-small-cell lung cancer (NSCLC) undergoing anti-PD1 therapy. We carried out an AI-based characterization of each lesion on the pretreatment contrast-enhanced CT imaging data to develop and validate a noninvasive machine learning biomarker capable of distinguishing between immunotherapy responding and nonresponding. To define the biological basis of the radiographic biomarker, we carried out gene set enrichment analysis in an independent dataset of 262 NSCLC patients. RESULTS: The biomarker reached significant performance on NSCLC lesions (up to 0.83 AUC, P < 0.001) and borderline significant for melanoma lymph nodes (0.64 AUC, P = 0.05). Combining these lesion-wide predictions on a patient level, immunotherapy response could be predicted with an AUC of up to 0.76 for both cancer types (P < 0.001), resulting in a 1-year survival difference of 24% (P = 0.02). We found highly significant associations with pathways involved in mitosis, indicating a relationship between increased proliferative potential and preferential response to immunotherapy. CONCLUSIONS: These results indicate that radiographic characteristics of lesions on standard-of-care imaging may function as noninvasive biomarkers for response to immunotherapy, and may show utility for improved patient stratification in both neoadjuvant and palliative settings.
43Predictive Value of Baseline FDG-PET/CT for the Durable Response to Immune Checkpoint Inhibition in NSCLC Patients Using the Morphological and Metabolic Features of Primary Tumors.PubMed
Ken Kudura, Nando Ritz, Tim Kutzker, et al.
Cancers (Basel). 2022 Dec 11;14(24):6095. doi: 10.3390/cancers14246095.
Objectives: We aimed to investigate the predictive value of baseline 2-deoxy-2-[18F]fluoro-D-glucose positron emission tomography/computed tomography (FDG-PET/CT) for durable responses to immune checkpoint inhibitors (ICIs) by linking the morphological and metabolic features of primary tumors (PTs) in nonsmall cell lung cancer (NSCLC) patients. Methods: For the purpose of this single-center study, the imaging data of the patients with a first diagnosis of NSCLC and an available baseline FDG-PET/CT between 2020 and 2021 were retrospectively assessed. The baseline characteristics were collected based on clinical reports and interdisciplinary tumor board documentation. The metabolic (such as standardized uptake value SUV maximum and mean (SUVmax, SUV mean), metabolic tumor volume (MTV), total lesion glycolysis (TLG)) and morphological (such as volume, morphology, margin, and presence of lymphangiosis through imaging) features of all the PTs were retrospectively assessed using FDG-PET/CT. Overall survival (OS), progression-free survival (PFS), clinical benefit (CB) and mortality rate were used as endpoints to define the long-term response to therapy. A backward, stepwise logistic regression analysis was performed in order to define the best model for predicting lasting responses to treatment. Statistical significance was assumed at p < 0.05. Results: A total of 125 patients (median age ± standard deviation (SD) 72.0 ± 9.5 years) were enrolled: 64 men (51.2%) and 61 women (48.8%). Adenocarcinoma was by far the most common histological subtype of NSCLC (47.2%). At the initial diagnosis, the vast majority of all the included patients showed either locally advanced disease (34.4%) or metastatic disease (36.8%). Fifty patients were treated with ICIs either as a first-line (20%) or second-line (20%) therapy, while 75 patients did not receive ICIs. The median values ± SD of PT SUVmax, mean, MTV, and TLG were respectively 10.1 ± 6.0, 6.1 ± 3.5, 13.5 ± 30.7, and 71.4 ± 247.7. The median volume of PT ± SD was 13.7 ± 30.7 cm3. The PTs were most frequently solid (86.4%) with irregular margins (76.8%). Furthermore, in one out of five cases, the morphological evidence of lymphangiosis was seen through imaging (n = 25). The median follow-up ± SD was 18.93 ± 6.98 months. The median values ± SD of OS and PFS were, respectively, 14.80 ± 8.68 months and 14.03 ± 9.02 months. Age, PT volume, SUVmax, TLG, the presence of lymphangiosis features through imaging, and clinical stage IV were very strong long-term outcome predictors of patients treated with ICIs, while no significant outcome predictors could be found for the cohort with no ICI treatment. The optimal cut-off values were determined for PT volume (26.94 cm3) and SUVmax (15.05). Finally, 58% of NSCLC patients treated with ICIs had a CB vs. 78.7% of patients in the cohort with no ICI treatment. However, almost all patients treated with ICIs and with disease progression over time died (mortality in the case of disease progression 95% vs. 62.5% in the cohort without ICIs). Conclusion: Baseline FDG-PET/CT could be used to predict a durable response to ICIs in NSCLC patients. Age, clinical stage IV, lymphangiosis features through imaging, PT volume (thus PT MTV due to a previously demonstrated linear correlation), PT SUVmax, and TLG were very strong long-term outcome predictors. Our results highlight the importance of linking clinical data, as much as morphological features, to the metabolic parameters of primary tumors in a multivariate outcome-predicting model using baseline FDG-PET/CT.
44CT-based radiomics in predicting pathological response in non-small cell lung cancer patients receiving neoadjuvant immunotherapy.PubMed
Qian Lin, Hai Jun Wu, Qi Shi Song, et al.
Front Oncol. 2022 Oct 4;12:937277. doi: 10.3389/fonc.2022.937277. eCollection 2022.
OBJECTIVES: In radiomics, high-throughput algorithms extract objective quantitative features from medical images. In this study, we evaluated CT-based radiomics features, clinical features, in-depth learning features, and a combination of features for predicting a good pathological response (GPR) in non-small cell lung cancer (NSCLC) patients receiving immunotherapy-based neoadjuvant therapy (NAT). MATERIALS AND METHODS: We reviewed 62 patients with NSCLC who received surgery after immunotherapy-based NAT and collected clinicopathological data and CT images before and after immunotherapy-based NAT. A series of image preprocessing was carried out on CT scanning images: tumor segmentation, conventional radiomics feature extraction, deep learning feature extraction, and normalization. Spearman correlation coefficient, principal component analysis (PCA), and least absolute shrinkage and selection operator (LASSO) were used to screen features. The pretreatment traditional radiomics combined with clinical characteristics (before_rad_cil) model and pretreatment deep learning characteristics (before_dl) model were constructed according to the data collected before treatment. The data collected after NAT created the after_rad_cil model and after_dl model. The entire model was jointly constructed by all clinical features, conventional radiomics features, and deep learning features before and after neoadjuvant treatment. Finally, according to the data obtained before and after treatment, the before_nomogram and after_nomogram were constructed. RESULTS: In the before_rad_cil model, four traditional radiomics features ("original_shape_flatness," "wavelet hhl_firer_skewness," "wavelet hlh_firer_skewness," and "wavelet lll_glcm_correlation") and two clinical features ("gender" and "N stage") were screened out to predict a GPR. The average prediction accuracy (ACC) after modeling with k-nearest neighbor (KNN) was 0.707. In the after_rad_cil model, nine features predictive of GPR were obtained after feature screening, among which seven were traditional radiomics features: "exponential_firer_skewness," "exponential_glrlm_runentropy," "log- sigma-5-0-mm-3d_firer_kurtosis," "logarithm_skewness," "original_shape_elongation," "original_shape_brilliance," and "wavelet llh_glcm_clustershade"; two were clinical features: "after_CRP" and "after lymphocyte percentage." The ACC after modeling with support vector machine (SVM) was 0.682. The before_dl model and after_dl model were modeled by SVM, and the ACC was 0.629 and 0.603, respectively. After feature screening, the entire model was constructed by multilayer perceptron (MLP), and the ACC of the GPR was the highest, 0.805. The calibration curve showed that the predictions of the GPR by the before_nomogram and after_nomogram were in consensus with the actual GPR. CONCLUSION: CT-based radiomics has a good predictive ability for a GPR in NSCLC patients receiving immunotherapy-based NAT. Among the radiomics features combined with the clinicopathological information model, deep learning feature model, and the entire model, the entire model had the highest prediction accuracy.
45CT-based quantification of intratumoral heterogeneity for predicting pathologic complete response to neoadjuvant immunochemotherapy in non-small cell lung cancer.PubMed
Guanchao Ye, Guangyao Wu, Chunyang Zhang, et al.
Front Immunol. 2024 Jun 12;15:1414954. doi: 10.3389/fimmu.2024.1414954. eCollection 2024.
OBJECTIVES: To investigate the prediction of pathologic complete response (pCR) in patients with non-small cell lung cancer (NSCLC) undergoing neoadjuvant immunochemotherapy (NAIC) using quantification of intratumoral heterogeneity from pre-treatment CT image. METHODS: This retrospective study included 178 patients with NSCLC who underwent NAIC at 4 different centers. The training set comprised 108 patients from center A, while the external validation set consisted of 70 patients from center B, center C, and center D. The traditional radiomics model was contrasted using radiomics features. The radiomics features of each pixel within the tumor region of interest (ROI) were extracted. The optimal division of tumor subregions was determined using the K-means unsupervised clustering method. The internal tumor heterogeneity habitat model was developed using the habitats features from each tumor sub-region. The LR algorithm was employed in this study to construct a machine learning prediction model. The diagnostic performance of the model was evaluated using criteria such as area under the receiver operating characteristic curve (AUC), accuracy, specificity, sensitivity, positive predictive value (PPV), and negative predictive value (NPV). RESULTS: In the training cohort, the traditional radiomics model achieved an AUC of 0.778 [95% confidence interval (CI): 0.688-0.868], while the tumor internal heterogeneity habitat model achieved an AUC of 0.861 (95% CI: 0.789-0.932). The tumor internal heterogeneity habitat model exhibits a higher AUC value. It demonstrates an accuracy of 0.815, surpassing the accuracy of 0.685 achieved by traditional radiomics models. In the external validation cohort, the AUC values of the two models were 0.723 (CI: 0.591-0.855) and 0.781 (95% CI: 0.673-0.889), respectively. The habitat model continues to exhibit higher AUC values. In terms of accuracy evaluation, the tumor heterogeneity habitat model outperforms the traditional radiomics model, achieving a score of 0.743 compared to 0.686. CONCLUSION: The quantitative analysis of intratumoral heterogeneity using CT to predict pCR in NSCLC patients undergoing NAIC holds the potential to inform clinical decision-making for resectable NSCLC patients, prevent overtreatment, and enable personalized and precise cancer management.
46[Construction of A Nomogram Prediction Model for PD-L1 Expression in Non-small Cell Lung Cancer Based on 18F-FDG PET/CT Metabolic Parameters].PubMed
Luoluo Hao, Lifeng Wang, Mengyao Zhang, et al.
Zhongguo Fei Ai Za Zhi. 2023 Nov 20;26(11):833-842. doi: 10.3779/j.issn.1009-3419.2023.101.32.
BACKGROUND: In recent years, immunotherapy represented by programmed cell death 1 (PD-1)/programmed cell death ligand 1 (PD-L1) immunosuppressants has greatly changed the status of non-small cell lung cancer (NSCLC) treatment. PD-L1 has become an important biomarker for screening NSCLC immunotherapy beneficiaries, but how to easily and accurately detect whether PD-L1 is expressed in NSCLC patients is a difficult problem for clinicians. The aim of this study was to construct a Nomogram prediction model of PD-L1 expression in NSCLC patients based on 18F-fluorodeoxy glucose (18F-FDG) positron emission tomography/conputed tomography (PET/CT) metabolic parameters and to evaluate its predictive value. METHODS: Retrospective collection of 18F-FDG PET/CT metabolic parameters, clinicopathological information and PD-L1 test results of 155 NSCLC patients from Inner Mongolia People's Hospital between September 2016 and July 2021. The patients were divided into the training group (n=117) and the internal validation group (n=38), and another 51 cases of NSCLC patients in our hospital between August 2021 and July 2022 were collected as the external validation group according to the same criteria. Then all of them were categorized according to the results of PD-L1 assay into PD-L1+ group and PD-L1- group. The metabolic parameters and clinicopathological information of patients in the training group were analyzed by univariate and binary Logistic regression, and a Nomogram prediction model was constructed based on the screened independent influencing factors. The effect of the model was evaluated by receiver operating characteristic (ROC) curve, calibration curve and decision curve analysis (DCA) in both the training group and the internal and external validation groups. RESULTS: Binary Logistic regression analysis showed that metabolic tumor volume (MTV), gender and tumor diameter were independent influences on PD-L1 expression. Then a Nomogram prediction model was constructed based on the above independent influences. The ROC curve for the model in the training group shows an area under the curve (AUC) of 0.769 (95%CI: 0.683-0.856) with an optimal cutoff value of 0.538. The AUC was 0.775 (95%CI: 0.614-0.936) in the internal validation group and 0.752 (95%CI: 0.612-0.893) in the external validation group. The calibration curves were tested by the Hosmer-Lemeshow test and showed that the training group (χ2=0.040, P=0.979), the internal validation group (χ2=2.605, P=0.271), and the external validation group (χ2=0.396, P=0.820) were well calibrated. The DCA curves show that the model provides clinical benefit to patients over a wide range of thresholds (training group: 0.00-0.72, internal validation group: 0.00-0.87, external validation group: 0.00-0.66). CONCLUSIONS: The Nomogram prediction model constructed on the basis of 18F-FDG PET/CT metabolic parameters has greater application value in predicting PD-L1 expression in NSCLC patients.
47[Construction of a prognostic nomogram combining PET/CT metabolic parameters and blood inflammatory markers for non-small cell lung cancer treated with first-line chemotherapy].PubMed
D Kong, L Song, Y Xiang
Nan Fang Yi Ke Da Xue Xue Bao. 2023 Dec 20;43(12):2139-2144. doi: 10.12122/j.issn.1673-4254.2023.12.20.
OBJECTIVE: To investigate the prognostic value of F-FDG-PET/CT metabolic parameters and blood inflammatory markers for advanced non-small cell lung cancer (NSCLC, stage Ⅳ/ⅢB) treated with first-line chemotherapy combined with immunotherapy and construct a nomogram prediction model for NSCLC. METHODS: We retrospectively analyzed the metabolic parameters (SUVmax, MTV and TLG) and blood markers of inflammation (NLR, DNLR, PLR and SII) in 105 patients with advanced NSCLC receiving chemotherapy combined with baseline F-FDG-PET/CT prior to immunotherapy from March, 2019 to June, 2021. ROC curve was used to calculate the best cut-off points for grouping, and univariate and multivariate COX regression analyses were performed to screen the independent predictors of prognosis for a combined diagnostic analysis. The effective biomarkers were included in the prediction model, and the nomogram model was constructed using the cph function in the rms function package of R language software. RESULTS: The patients were followed up for a median of 17.5 months, and their median progression-free survival (PFS) was 16 months with a median overall survival (OS) of 13.6 months. A high PLR (≥151.050) and a high TLG (≥101.940) were significant independent prognostic factors for PFS, and a high SII (≥941.385) and a high TLG (≥101.940) were independent prognostic factors for OS. The nomogram combining PET and blood markers of inflammation showed a good performance for prognostic prediction (with C-index of 0.682 for PFS and of 0.727 for OS) and good fitting of the calibration curve. The clinical decision curve showed good clinical utility of the nomogram. CONCLUSION: The baseline PET/CT metabolic parameters and blood inflammatory markers are associated with PFS and OS of patients with advanced NSCLC receiving first-line chemotherapy, and the constructed nomogram based on these parameters has a good performance for prognostic prediction in these patients.
48Non-invasive measurement of PD-L1 status and prediction of immunotherapy response using deep learning of PET/CT images.PubMed
Wei Mu, Lei Jiang, Yu Shi, et al.
J Immunother Cancer. 2021 Jun;9(6). doi: 10.1136/jitc-2020-002118.
BACKGROUND: Currently, only a fraction of patients with non-small cell lung cancer (NSCLC) treated with immune checkpoint inhibitors (ICIs) experience a durable clinical benefit (DCB). According to NCCN guidelines, Programmed death-ligand 1 (PD-L1) expression status determined by immunohistochemistry (IHC) of biopsies is the only clinically approved companion biomarker to trigger the use of ICI therapy. Based on prior work showing a relationship between quantitative imaging and gene expression, we hypothesize that quantitative imaging (radiomics) can provide an alternative surrogate for PD-L1 expression status in clinical decision support. METHODS: F-FDG-PET/CT images and clinical data were curated from 697 patients with NSCLC from three institutions and these were analyzed using a small-residual-convolutional-network (SResCNN) to develop a deeply learned score (DLS) to predict the PD-L1 expression status. This developed model was further used to predict DCB, progression-free survival (PFS), and overall survival (OS) in two retrospective and one prospective test cohorts of ICI-treated patients with advanced stage NSCLC. RESULTS: The PD-L1 DLS significantly discriminated between PD-L1 positive and negative patients (area under receiver operating characteristics curve ≥0.82 in the training, validation, and two external test cohorts). Importantly, the DLS was indistinguishable from IHC-derived PD-L1 status in predicting PFS and OS, suggesting the utility of DLS as a surrogate for IHC. A score generated by combining the DLS with clinical characteristics was able to accurately (C-indexes of 0.70-0.87) predict DCB, PFS, and OS in retrospective training, prospective testing and external validation cohorts. CONCLUSION: Hence, we propose DLS as a surrogate or substitute for IHC-determined PD-L1 measurement to guide individual pretherapy decisions pending in larger prospective trials.
49Non-invasive multimodal CT deep learning biomarker to predict pathological complete response of non-small cell lung cancer following neoadjuvant immunochemotherapy: a multicenter study.PubMed
Guanchao Ye, Guangyao Wu, Yu Qi, et al.
J Immunother Cancer. 2024 Sep 3;12(9):e009348. doi: 10.1136/jitc-2024-009348.
OBJECTIVES: Although neoadjuvant immunochemotherapy has been widely applied in non-small cell lung cancer (NSCLC), predicting treatment response remains a challenge. We used pretreatment multimodal CT to explore deep learning-based immunochemotherapy response image biomarkers. METHODS: This study retrospectively obtained non-contrast enhanced and contrast enhancedbubu CT scans of patients with NSCLC who underwent surgery after receiving neoadjuvant immunochemotherapy at multiple centers between August 2019 and February 2023. Deep learning features were extracted from both non-contrast enhanced and contrast enhanced CT scans to construct the predictive models (LUNAI-uCT model and LUNAI-eCT model), respectively. After the feature fusion of these two types of features, a fused model (LUNAI-fCT model) was constructed. The performance of the model was evaluated using the area under the receiver operating characteristic curve (AUC), accuracy, sensitivity, specificity, positive predictive value, and negative predictive value. SHapley Additive exPlanations analysis was used to quantify the impact of CT imaging features on model prediction. To gain insights into how our model makes predictions, we employed Gradient-weighted Class Activation Mapping to generate saliency heatmaps. RESULTS: The training and validation datasets included 113 patients from Center A at the 8:2 ratio, and the test dataset included 112 patients (Center B n=73, Center C n=20, Center D n=19). In the test dataset, the LUNAI-uCT, LUNAI-eCT, and LUNAI-fCT models achieved AUCs of 0.762 (95% CI 0.654 to 0.791), 0.797 (95% CI 0.724 to 0.844), and 0.866 (95% CI 0.821 to 0.883), respectively. CONCLUSIONS: By extracting deep learning features from contrast enhanced and non-contrast enhanced CT, we constructed the LUNAI-fCT model as an imaging biomarker, which can non-invasively predict pathological complete response in neoadjuvant immunochemotherapy for NSCLC.
50Prediction of PD-L1 expression in NSCLC patients using PET/CT radiomics and prognostic modelling for immunotherapy in PD-L1-positive NSCLC patients.PubMed
M Peng, M Wang, X Yang, et al.
Clin Radiol. 2025 Jul;86:106915. doi: 10.1016/j.crad.2025.106915. Epub 2025 Apr 2.
AIM: To develop a positron emission tomography/computed tomography (PET/CT)-based radiomics model for predicting programmed cell death ligand 1 (PD-L1) expression in non-small cell lung cancer (NSCLC) patients and estimating progression-free survival (PFS) and overall survival (OS) in PD-L1-positive patients undergoing first-line immunotherapy. MATERIALS AND METHODS: We retrospectively analysed 143 NSCLC patients who underwent pretreatment F-fluorodeoxyglucose (F-FDG) PET/CT scans, of whom 86 were PD-L1-positive. Clinical data collected included gender, age, smoking history, Tumor-Node-Metastases (TNM) staging system, pathologic types, laboratory parameters, and PET metabolic parameters. Four machine learning algorithms-Bayes, logistic, random forest, and Supportsupport vector machine (SVM)-were used to build models. The predictive performance was validated using receiver operating characteristic (ROC) curves. Univariate and multivariate Cox analyses identified independent predictors of OS and PFS in PD-L1-positive expression patients undergoing immunotherapy, and a nomogram was created to predict OS. RESULTS: A total of 20 models were built for predicting PD-L1 expression. The clinical combined PET/CT radiomics model based on the SVM algorithm performed best (area under curve for training and test sets: 0.914 and 0.877, respectively). The Cox analyses showed that smoking history independently predicted PFS. SUVmean, monocyte percentage and white blood cell count were independent predictors of OS, and the nomogram was created to predict 1-year, 2-year, and 3-year OS based on these three factors. CONCLUSION: We developed PET/CT-based machine learning models to help predict PD-L1 expression in NSCLC patients and identified independent predictors of PFS and OS in PD-L1-positive patients receiving immunotherapy, thereby aiding precision treatment.
51Radiomics for the non-invasive prediction of PD-L1 expression in patients with brain metastases secondary to non-small cell lung cancer.PubMed
Anna-Katharina Meißner, Robin Gutsche, Norbert Galldiks, et al.
J Neurooncol. 2023 Jul;163(3):597-605. doi: 10.1007/s11060-023-04367-7. Epub 2023 Jun 29.
BACKGROUND: The expression level of the programmed cell death ligand 1 (PD-L1) appears to be a predictor for response to immunotherapy using checkpoint inhibitors in patients with non-small cell lung cancer (NSCLC). As differences in terms of PD-L1 expression levels in the extracranial primary tumor and the brain metastases may occur, a reliable method for the non-invasive assessment of the intracranial PD-L1 expression is, therefore of clinical value. Here, we evaluated the potential of radiomics for a non-invasive prediction of PD-L1 expression in patients with brain metastases secondary to NSCLC. PATIENTS AND METHODS: Fifty-three NSCLC patients with brain metastases from two academic neuro-oncological centers (group 1, n = 36 patients; group 2, n = 17 patients) underwent tumor resection with a subsequent immunohistochemical evaluation of the PD-L1 expression. Brain metastases were manually segmented on preoperative T1-weighted contrast-enhanced MRI. Group 1 was used for model training and validation, group 2 for model testing. After image pre-processing and radiomics feature extraction, a test-retest analysis was performed to identify robust features prior to feature selection. The radiomics model was trained and validated using random stratified cross-validation. Finally, the best-performing radiomics model was applied to the test data. Diagnostic performance was evaluated using receiver operating characteristic (ROC) analyses. RESULTS: An intracranial PD-L1 expression (i.e., staining of at least 1% or more of tumor cells) was present in 18 of 36 patients (50%) in group 1, and 7 of 17 patients (41%) in group 2. Univariate analysis identified the contrast-enhancing tumor volume as a significant predictor for PD-L1 expression (area under the ROC curve (AUC), 0.77). A random forest classifier using a four-parameter radiomics signature, including tumor volume, yielded an AUC of 0.83 ± 0.18 in the training data (group 1), and an AUC of 0.84 in the external test data (group 2). CONCLUSION: The developed radiomics classifiers allows for a non-invasive assessment of the intracranial PD-L1 expression in patients with brain metastases secondary to NSCLC with high accuracy.
52CT radiomics-based model for predicting TMB and immunotherapy response in non-small cell lung cancer.PubMed
Jiexiao Wang, Jialiang Wang, Xiang Huang, et al.
BMC Med Imaging. 2024 Feb 15;24(1):45. doi: 10.1186/s12880-024-01221-8.
BACKGROUND: Tumor mutational burden (TMB) is one of the most significant predictive biomarkers of immunotherapy efficacy in non-small cell lung cancer (NSCLC). Radiomics allows high-throughput extraction and analysis of advanced and quantitative medical imaging features. This study develops and validates a radiomic model for predicting TMB level and the response to immunotherapy based on CT features in NSCLC. METHOD: Pre-operative chest CT images of 127 patients with NSCLC were retrospectively studied. The 3D-Slicer software was used to outline the region of interest and extract features from the CT images. Radiomics prediction model was constructed by LASSO and multiple logistic regression in a training dataset. The model was validated by receiver operating characteristic (ROC) curves and calibration curves using external datasets. Decision curve analysis was used to assess the value of the model for clinical application. RESULTS: A total of 1037 radiomic features were extracted from the CT images of NSCLC patients from TCGA. LASSO regression selected three radiomics features (Flatness, Autocorrelation and Minimum), which were associated with TMB level in NSCLC. A TMB prediction model consisting of 3 radiomic features was constructed by multiple logistic regression. The area under the curve (AUC) value in the TCGA training dataset was 0.816 (95% CI: 0.7109-0.9203) for predicting TMB level in NSCLC. The AUC value in external validation dataset I was 0.775 (95% CI: 0.5528-0.9972) for predicting TMB level in NSCLC, and the AUC value in external validation dataset II was 0.762 (95% CI: 0.5669-0.9569) for predicting the efficacy of immunotherapy in NSCLC. CONCLUSION: The model based on CT radiomic features helps to achieve cost effective improvement in TMB classification and precise immunotherapy treatment of NSCLC patients.
53A short-term follow-up CT based radiomics approach to predict response to immunotherapy in advanced non-small-cell lung cancer.PubMed
Jing Gong, Xiao Bao, Ting Wang, et al.
Oncoimmunology. 2022 Jan 25;11(1):2028962. doi: 10.1080/2162402X.2022.2028962. eCollection 2022.
To develop a short-term follow-up CT-based radiomics approach to predict response to immunotherapy in advanced non-small-cell lung cancer (NSCLC) and investigate the prognostic value of radiomics features in predicting progression-free survival (PFS) and overall survival (OS). We first retrospectively collected 224 advanced NSCLC patients from two centers, and divided them into a primary cohort and two validation cohorts respectively. Then, we processed CT scans with a series of image preprocessing techniques namely, tumor segmentation, image resampling, feature extraction and normalization. To select the optimal features, we applied the feature ranking with recursive feature elimination method. After resampling the training dataset with a synthetic minority oversampling technique, we applied the support vector machine classifier to build a machine-learning-based classification model to predict response to immunotherapy. Finally, we used Kaplan-Meier (KM) survival analysis method to evaluate prognostic value of rad-score generated by CT-radiomics model. In two validation cohorts, the delta-radiomics model significantly improved the area under receiver operating characteristic curve from 0.64 and 0.52 to 0.82 and 0.87, respectively (P < .05). In sub-group analysis, pre- and delta-radiomics model yielded higher performance for adenocarcinoma (ADC) patients than squamous cell carcinoma (SCC) patients. Through the KM survival analysis, the rad-score of delta-radiomics model had a significant prognostic for PFS and OS in validation cohorts (P < .05). Our results demonstrated that (1) delta-radiomics model could improve the prediction performance, (2) radiomics model performed better on ADC patients than SCC patients, (3) delta-radiomics model had prognostic values in predicting PFS and OS of NSCLC patients.
54Predicting benefit from immune checkpoint inhibitors in patients with non-small-cell lung cancer by CT-based ensemble deep learning: a retrospective study.PubMed
Maliazurina B Saad, Lingzhi Hong, Muhammad Aminu, et al.
Lancet Digit Health. 2023 Jul;5(7):e404-e420. doi: 10.1016/S2589-7500(23)00082-1. Epub 2023 May 31.
BACKGROUND: Only around 20-30% of patients with non-small-cell lung cancer (NCSLC) have durable benefit from immune-checkpoint inhibitors. Although tissue-based biomarkers (eg, PD-L1) are limited by suboptimal performance, tissue availability, and tumour heterogeneity, radiographic images might holistically capture the underlying cancer biology. We aimed to investigate the application of deep learning on chest CT scans to derive an imaging signature of response to immune checkpoint inhibitors and evaluate its added value in the clinical context. METHODS: In this retrospective modelling study, 976 patients with metastatic, EGFR/ALK negative NSCLC treated with immune checkpoint inhibitors at MD Anderson and Stanford were enrolled from Jan 1, 2014, to Feb 29, 2020. We built and tested an ensemble deep learning model on pretreatment CTs (Deep-CT) to predict overall survival and progression-free survival after treatment with immune checkpoint inhibitors. We also evaluated the added predictive value of the Deep-CT model in the context of existing clinicopathological and radiological metrics. FINDINGS: Our Deep-CT model demonstrated robust stratification of patient survival of the MD Anderson testing set, which was validated in the external Stanford set. The performance of the Deep-CT model remained significant on subgroup analyses stratified by PD-L1, histology, age, sex, and race. In univariate analysis, Deep-CT outperformed the conventional risk factors, including histology, smoking status, and PD-L1 expression, and remained an independent predictor after multivariate adjustment. Integrating the Deep-CT model with conventional risk factors demonstrated significantly improved prediction performance, with overall survival C-index increases from 0·70 (clinical model) to 0·75 (composite model) during testing. On the other hand, the deep learning risk scores correlated with some radiomics features, but radiomics alone could not reach the performance level of deep learning, indicating that the deep learning model effectively captured additional imaging patterns beyond known radiomics features. INTERPRETATION: This proof-of-concept study shows that automated profiling of radiographic scans through deep learning can provide orthogonal information independent of existing clinicopathological biomarkers, bringing the goal of precision immunotherapy for patients with NSCLC closer. FUNDING: National Institutes of Health, Mark Foundation Damon Runyon Foundation Physician Scientist Award, MD Anderson Strategic Initiative Development Program, MD Anderson Lung Moon Shot Program, Andrea Mugnaini, and Edward L C Smith.
55Clinical applications of radiomics in non-small cell lung cancer patients with immune checkpoint inhibitor-related pneumonitis.PubMed
Yang Shu, Wei Xu, Rui Su, et al.
Front Immunol. 2023 Sep 20;14:1251645. doi: 10.3389/fimmu.2023.1251645. eCollection 2023.
Immune checkpoint inhibitors (ICIs) modulate the body's immune function to treat tumors but may also induce pneumonitis. Immune checkpoint inhibitor-related pneumonitis (ICIP) is a serious immune-related adverse event (irAE). Immunotherapy is currently approved as a first-line treatment for non-small cell lung cancer (NSCLC), and the incidence of ICIP in NSCLC patients can be as high as 5%-19% in clinical practice. ICIP can be severe enough to lead to the death of NSCLC patients, but there is a lack of a gold standard for the diagnosis of ICIP. Radiomics is a method that uses computational techniques to analyze medical images (e.g., CT, MRI, PET) and extract important features from them, which can be used to solve classification and regression problems in the clinic. Radiomics has been applied to predict and identify ICIP in NSCLC patients in the hope of transforming clinical qualitative problems into quantitative ones, thus improving the diagnosis and treatment of ICIP. In this review, we summarize the pathogenesis of ICIP and the process of radiomics feature extraction, review the clinical application of radiomics in ICIP of NSCLC patients, and discuss its future application prospects.
56Development and validation of a radiomics-based nomogram for predicting a major pathological response to neoadjuvant immunochemotherapy for patients with potentially resectable non-small cell lung cancer.PubMed
Chaoyuan Liu, Wei Zhao, Junpeng Xie, et al.
Front Immunol. 2023 Feb 16;14:1115291. doi: 10.3389/fimmu.2023.1115291. eCollection 2023.
INTRODUCTION: The treatment response to neoadjuvant immunochemotherapy varies among patients with potentially resectable non-small cell lung cancers (NSCLC) and may have severe immune-related adverse effects. We are currently unable to accurately predict therapeutic response. We aimed to develop a radiomics-based nomogram to predict a major pathological response (MPR) of potentially resectable NSCLC to neoadjuvant immunochemotherapy using pretreatment computed tomography (CT) images and clinical characteristics. METHODS: A total of 89 eligible participants were included and randomly divided into training (N=64) and validation (N=25) sets. Radiomic features were extracted from tumor volumes of interest in pretreatment CT images. Following data dimension reduction, feature selection, and radiomic signature building, a radiomics-clinical combined nomogram was developed using logistic regression analysis. RESULTS: The radiomics-clinical combined model achieved excellent discriminative performance, with AUCs of 0.84 (95% CI, 0.74-0.93) and 0.81(95% CI, 0.63-0.98) and accuracies of 80% and 80% in the training and validation sets, respectively. Decision curves analysis (DCA) indicated that the radiomics-clinical combined nomogram was clinically valuable. DISCUSSION: The constructed nomogram was able to predict MPR to neoadjuvant immunochemotherapy with a high degree of accuracy and robustness, suggesting that it is a convenient tool for assisting with the individualized management of patients with potentially resectable NSCLC.
57Clinical multi-dimensional prognostic nomogram for predicting the efficacy of immunotherapy in NSCLC.PubMed
Qian Zhao, Xiao Zhong, Xiaoqing Wang, et al.
Sci Rep. 2024 Sep 13;14(1):21380. doi: 10.1038/s41598-024-72760-x.
The advent of immunotherapy has greatly improved the prognosis of non-small cell lung (NSCLC) patients. However, given its low response rate and high cost of treatment, the search for valuable predictive markers of treatment efficacy is necessary. Considering the complexity and heterogeneity of the tumour and tumour microenvironment, the construction of a multi-dimensional prediction model is necessary. Therefore, we aimed to integrate clinical parameters, radiomic features, and immune signature data from NSCLC patients receiving immunotherapy to construct a multi-dimensional prediction model to better predict the efficacy of immunotherapy. The current study enrolled 137 NSCLC patients who received immunotherapy. We collected baseline clinical information, CT images, and tumour tissue specimens. Using 3D-Slicer software, radiomic features were extracted from patient CT images, and tumor tissue samples obtained before immunotherapy were subjected to immunohistochemical staining. Then, the least absolute shrinkage and selection operator (LASSO) Cox regression analysis was applied to downscale the data, and the radiomic features and immune signatures associated with the prognosis of immunotherapy patients were identified. The modified lung immune predictive index (mLIPI), radiomics score (Radioscore), immune score and multi-dimensional model nomogram were constructed. The C-index and area under the curve (AUC) were applied to evaluate the predictive efficacy of the models. Three radiomic features and three immune signatures that could predict the efficacy of immunotherapy were eventually screened. Multivariate analysis showed that the mLIPI, Radioscore, and immune score were independent predictive factors for PFS and OS (P < 0.05 for all models). The multi-dimensional model combining the three models showed better predictive efficacy than the mLIPI, Radioscore, and immune score (PFS: 0.721 vs. 0.662 vs. 0.610 vs. 0.610; OS: 0.727 vs. 0.661 vs. 0.601 vs. 0.602 respectively). The multi-dimensional model showed the best predictive efficacy, with C-index for PFS and OS higher than mLIPI, radioscore and immune score: 0.721 vs. 0.662 vs. 0.610 vs. 0.610 for PFS and 0.727 vs. 0.661 vs. 0.601 vs. 0.602 for OS, respectively. The AUC for the multi-dimensional model also performed better than those of the individual models: 0.771 vs. 0.684 vs. 0.715 vs. 0.711 for PFS and 0.768 vs. 0.662 vs. 0.661 vs. 0.658 for OS, respectively. The multi-dimensional model combining the three models had better predictive efficacy than any single model and was more likely to help provide patients personalized and precision medicine.
58Development, Validation, and Clinical Utility of Electronic Patient-Reported Outcome Measure-Enhanced Prediction Models for Overall Survival in Patients With Advanced Non-Small Cell Lung Cancer Receiving Immunotherapy.PubMed
Kuan Liao, Sabine N van der Veer, Fabio Gomes, et al.
JCO Clin Cancer Inform. 2024 Dec;8:e2400035. doi: 10.1200/CCI.24.00035. Epub 2024 Nov 26.
PURPOSE: Electronic patient-reported outcome measures (ePROMs) are increasingly collected routinely in clinical practice and may be prognostic for survival in adults with advanced non-small cell lung cancer (NSCLC) in addition to clinical data. This study developed ePROM-enhanced models for predicting 1-year overall survival in patients with advanced NSCLC at the start of immunotherapy. METHODS: This is a single-center study using consecutive patients from a tertiary cancer hospital in England. Using Cox proportional hazards models, we developed one clinical factor-only model and three ePROM-enhanced models, each including one of the following factors: quality of life (as measured by EuroQoL five-dimension five-level utility score) and overall symptom burden and number of moderate-to-severe symptoms (as measured by patient-reported version of Common Terminology Criteria for Adverse Events). Predictive performance was evaluated and compared through bootstrapping internal validation, and clinical utility was determined via decision curve analysis. RESULTS: The clinical factor-only model contained age, histology, performance status, and neutrophile-to-lymphocyte ratio. While calibration was similar between the clinical factor-only and ePROM-enhanced models, the latter showed improved discrimination by 0.020 (95% CI, 0.011 to 0.024), 0.024 (95% CI, 0.016 to 0.031), and 0.024 (95% CI, 0.014 to 0.029) when enhanced with ePROMs on quality of life, overall symptom burden, and number of moderate-to-severe symptoms, respectively. If care decisions are to be made at risk thresholds between 25% and 75%, the ePROM-enhanced models led to higher net benefit than the clinical factor-only model and the default strategies of intervention for all and intervention for none. CONCLUSION: The ePROM-enhanced models outperformed the clinical factor-only model in predicting 1-year overall survival for patients with advanced NSCLC receiving immunotherapy and showed potential clinical utility for informing decisions in this population. Future studies should focus on validating the models in external data sets.
59A panel of systemic inflammatory response biomarkers for outcome prediction in patients treated with radical cystectomy for urothelial carcinoma.PubMed
Victor M Schuettfort, David D'Andrea, Fahad Quhal, et al.
BJU Int. 2022 Feb;129(2):182-193. doi: 10.1111/bju.15379. Epub 2021 Apr 7.
OBJECTIVES: To determine the predictive and prognostic value of a panel of systemic inflammatory response (SIR) biomarkers relative to established clinicopathological variables in order to improve patient selection and facilitate more efficient delivery of peri-operative systemic therapy. MATERIALS AND METHODS: The preoperative serum levels of a panel of SIR biomarkers, including albumin-globulin ratio, neutrophil-lymphocyte ratio, De Ritis ratio, monocyte-lymphocyte ratio and modified Glasgow prognostic score were assessed in 4199 patients treated with radical cystectomy for clinically non-metastatic urothelial carcinoma of the bladder. Patients were randomly divided into a training and a testing cohort. A machine-learning-based variable selection approach (least absolute shrinkage and selection operator regression) was used for the fitting of several multivariable predictive and prognostic models. The outcomes of interest included prediction of upstaging to carcinoma invading bladder muscle (MIBC), lymph node involvement, pT3/4 disease, cancer-specific survival (CSS) and recurrence-free survival (RFS). The discriminatory ability of each model was either quantified by area under the receiver-operating curves or by the C-index. After validation and calibration of each model, a nomogram was created and decision-curve analysis was used to evaluate the clinical net benefit. RESULTS: For all outcome variables, at least one SIR biomarker was selected by the machine-learning process to be of high discriminative power during the fitting of the models. In the testing cohort, model performance evaluation for preoperative prediction of lymph node metastasis, ≥pT3 disease and upstaging to MIBC showed a 200-fold bootstrap-corrected area under the curve of 67.3%, 73% and 65.8%, respectively. For postoperative prognosis of CSS and RFS, a 200-fold bootstrap corrected C-index of 73.3% and 72.2%, respectively, was found. However, even the most predictive combinations of SIR biomarkers only marginally increased the discriminative ability of the respective model in comparison to established clinicopathological variables. CONCLUSION: While our machine-learning approach for fitting of the models with the highest discriminative ability incorporated several previously validated SIR biomarkers, these failed to improve the discriminative ability of the models to a clinically meaningful degree. While the prognostic and predictive value of such cheap and readily available biomarkers warrants further evaluation in the age of immunotherapy, additional novel biomarkers are still needed to improve risk stratification.
60Pipeline design to identify key features and classify the chemotherapy response on lung cancer patients using large-scale genetic data.PubMed
María Gabriela Valdés, Iván Galván-Femenía, Vicent Ribas Ripoll, et al.
BMC Syst Biol. 2018 Nov 20;12(Suppl 5):97. doi: 10.1186/s12918-018-0615-5.
BACKGROUND: During the last decade, the interest to apply machine learning algorithms to genomic data has increased in many bioinformatics applications. Analyzing this type of data entails difficulties for managing high-dimensional data, class imbalance for knowledge extraction, identifying important features and classifying individuals. In this study, we propose a general framework to tackle these challenges with different machine learning algorithms and techniques. We apply the configuration of this framework on lung cancer patients, identifying genetic signatures for classifying response to drug treatment response. We intersect these relevant SNPs with the GWAS Catalog of the National Human Genome Research Institute and explore the Regulomedb, GTEx databases for functional analysis purposes. RESULTS: The machine learning based solution proposed in this study is a scalable and flexible alternative to the classical uni-variate regression approach to analyze large-scale data. From 36 experiments executed using the machine learning framework design, we obtain good classification performance from the top 5 models with the highest cross-validation score and the smallest standard deviation. One thousand two hundred twenty four SNPs corresponding to the key features from the top 20 models (cross validation F1 mean >= 0.65) were compared with the GWAS Catalog finding no intersection with genome-wide significant reported hits. From these, new genetic signatures in MAE, CEP104, PRKCZ and ADRB2 show relevant biological regulatory functionality related to lung physiology. CONCLUSIONS: We have defined a machine learning framework using data with an unbalanced large data-set of SNP-arrays and imputed genotyping data from a pharmacogenomics study in lung cancer patients subjected to first-line platinum-based treatment. This approach found genome signals with no genome-wide significance in the uni-variate regression approach (GWAS Catalog) that are valuable for classifying patients, only few of them with related biological function. The effect results of these variants can be explained by the recently proposed omnigenic model hypothesis, which states that complex traits can be influenced mostly by genes outside not only by the "core genes", mainly found by the genome-wide significant SNPs, but also by the rest of genes outside of the "core pathways" with apparent unrelated biological functionality.
61Combination of computed tomography imaging-based radiomics and clinicopathological characteristics for predicting the clinical benefits of immune checkpoint inhibitors in lung cancer.PubMed
Bin Yang, Li Zhou, Jing Zhong, et al.
Respir Res. 2021 Jun 28;22(1):189. doi: 10.1186/s12931-021-01780-2.
BACKGROUND: In this study, we tested whether a combination of radiomic features extracted from baseline pre-immunotherapy computed tomography (CT) images and clinicopathological characteristics could be used as novel noninvasive biomarkers for predicting the clinical benefits of non-small cell lung cancer (NSCLC) patients treated with immune checkpoint inhibitors (ICIs). METHODS: The data from 92 consecutive patients with lung cancer who had been treated with ICIs were retrospectively analyzed. In total, 88 radiomic features were selected from the pretreatment CT images for the construction of a random forest model. Radiomics model 1 was constructed based on the Rad-score. Using multivariate logistic regression analysis, the Rad-score and significant predictors were integrated into a single predictive model (radiomics nomogram model 1) to predict the durable clinical benefit (DCB) of ICIs. Radiomics model 2 was developed based on the same Rad-score as radiomics model 1.Using multivariate Cox proportional hazards regression analysis, the Rad-score, and independent risk factors, radiomics nomogram model 2 was constructed to predict the progression-free survival (PFS). RESULTS: The models successfully predicted the patients who would benefit from ICIs. For radiomics model 1, the area under the receiver operating characteristic curve values for the training and validation cohorts were 0.848 and 0.795, respectively, whereas for radiomics nomogram model 1, the values were 0.902 and 0.877, respectively. For the PFS prediction, the Harrell's concordance indexes for the training and validation cohorts were 0.717 and 0.760, respectively, using radiomics model 2, whereas they were 0.749 and 0.791, respectively, using radiomics nomogram model 2. CONCLUSIONS: CT-based radiomic features and clinicopathological factors can be used prior to the initiation of immunotherapy for identifying NSCLC patients who are the most likely to benefit from the therapy. This could guide the individualized treatment strategy for advanced NSCLC.
62Multimodal data deep learning method for predicting symptomatic pneumonitis caused by lung cancer radiotherapy combined with immunotherapy.PubMed
Mingyu Yang, Jianli Ma, Chengcheng Zhang, et al.
Front Immunol. 2025 Jan 8;15:1492399. doi: 10.3389/fimmu.2024.1492399. eCollection 2024.
OBJECTIVES: The pairing of immunotherapy and radiotherapy in the treatment of locally advanced nonsmall cell lung cancer (NSCLC) has shown promise. By combining radiotherapy with immunotherapy, the synergistic effects of these modalities not only bolster antitumor efficacy but also exacerbate lung injury. Consequently, developing a model capable of accurately predicting radiotherapy- and immunotherapy-related pneumonitis in lung cancer patients is a pressing need. Depth image features extracted from deep learning, combined with radiomics and clinical characteristics, were used to create a deep learning model. This model was developed to forecast symptomatic pneumonitis (SP) (≥Grade 2) in lung cancer patients undergoing thoracic radiotherapy in combination with immunotherapy. METHODS: The prediction was based on CT scans taken prior to the start of thoracic radiotherapy. Retrospective collection of clinical data was conducted on 261 lung cancer patients undergoing a combination of thoracic radiotherapy and immunotherapy from January 2018 to May 2023. Imaging data in the form of pre-RT-CT scans were obtained for all individuals included in the study. The region of interest (ROI) in the lung parenchyma was outlined separately from the tumor volume, and standard radiomic features were obtained through the use of 3D Slicer software. In addition, the images were cropped to a uniform size of 224x224 pixels. Data augmentation techniques, including random horizontal flipping, were employed. The normalized image data was then input into a pre-trained deep residual network, ResNet34, which utilized convolutional layers and global average pooling layers for deep feature extraction. A five-fold cross-validation approach was implemented to construct the model, automatically splitting the dataset into training and validation sets at an 8:2 ratio. This process was repeated five times, and the results from these iterations were aggregated to compute the average values of performance metrics, thereby assessing the overall performance and stability of the model. RESULTS: The multimodal fusion model developed in this research, which incorporated depth image characteristics, radiomics properties, and clinical data, demonstrated an AUC of 0.922 (95% CI: 0.902-0.945, P value < 0.001). This amalgamated model surpassed the performance of the radiomic feature model (AUC 0.811, 95% CI: 0.786-0.832, P value < 0.001), the clinical information model (AUC 0.711, 95% CI: 0.682-0.753, P value < 0.001), as well as the model that integrated omics attributes with clinical data (AUC 0.872, 95% CI: 0.845-0.896, P value < 0.001) utilizing deep neural networks (DNNs). Comparatively, the radiomic feature model based on random forest (RF) yielded an AUC of 0.576, with a 95% confidence interval of 0.523-0.628. The clinical information model based on RF had an AUC of 0.525, with a 95% confidence interval of 0.479-0.572. When both radiomic features and clinical information were combined in a model based on RF, the AUC improved slightly to 0.611, with a 95% confidence interval of 0.566-0.652. CONCLUSIONS: In this study, a deep neural network-based multimodal fusion model improved the prediction performance compared to traditional radiomics. The model accurately predicted Grade 2 or higher SP in lung cancer patients undergoing radiotherapy combined with immunotherapy.
63Radiomics and artificial intelligence for precision medicine in lung cancer treatment.PubMed
Mitchell Chen, Susan J Copley, Patrizia Viola, et al.
Semin Cancer Biol. 2023 Aug;93:97-113. doi: 10.1016/j.semcancer.2023.05.004. Epub 2023 May 19.
Lung cancer is the leading cause of cancer-related deaths worldwide. It exhibits, at the mesoscopic scale, phenotypic characteristics that are generally indiscernible to the human eye but can be captured non-invasively on medical imaging as radiomic features, which can form a high dimensional data space amenable to machine learning. Radiomic features can be harnessed and used in an artificial intelligence paradigm to risk stratify patients, and predict for histological and molecular findings, and clinical outcome measures, thereby facilitating precision medicine for improving patient care. Compared to tissue sampling-driven approaches, radiomics-based methods are superior for being non-invasive, reproducible, cheaper, and less susceptible to intra-tumoral heterogeneity. This review focuses on the application of radiomics, combined with artificial intelligence, for delivering precision medicine in lung cancer treatment, with discussion centered on pioneering and groundbreaking works, and future research directions in the area.
64A muti-modal feature fusion method based on deep learning for predicting immunotherapy response.PubMed
Xiong Li, Xuan Feng, Juan Zhou, et al.
J Theor Biol. 2024 Jun 7;586:111816. doi: 10.1016/j.jtbi.2024.111816. Epub 2024 Apr 6.
Immune checkpoint therapy (ICT) has greatly improved the survival of cancer patients in the past few years, but only a small number of patients respond to ICT. To predict ICT response, we developed a multi-modal feature fusion model based on deep learning (MFMDL). This model utilizes graph neural networks to map gene-gene relationships in gene networks to low dimensional vector spaces, and then fuses biological pathway features and immune cell infiltration features to make robust predictions of ICT. We used five datasets to validate the predictive performance of the MFMDL. These five datasets span multiple types of cancer, including melanoma, lung cancer, and gastric cancer. We found that the prediction performance of multi-modal feature fusion model based on deep learning is superior to other traditional ICT biomarkers, such as ICT targets or tumor microenvironment-associated markers. In addition, we also conducted ablation experiments to demonstrate the necessity of fusing different modal features, which can improve the prediction accuracy of the model.
65Automated PD-L1 status prediction in lung cancer with multi-modal PET/CT fusion.PubMed
Ronrick Da-Ano, Gustavo Andrade-Miranda, Olena Tankyevych, et al.
Sci Rep. 2024 Jul 19;14(1):16720. doi: 10.1038/s41598-024-66487-y.
Programmed death-ligand 1 (PD-L1) expressions play a crucial role in guiding therapeutic interventions such as the use of tyrosine kinase inhibitors (TKIs) and immune checkpoint inhibitors (ICIs) in lung cancer. Conventional determination of PD-L1 status includes careful surgical or biopsied tumor specimens. These specimens are gathered through invasive procedures, representing a risk of difficulties and potential challenges in getting reliable and representative tissue samples. Using a single center cohort of 189 patients, our objective was to evaluate various fusion methods that used non-invasive computed tomography (CT) and F-FDG positron emission tomography (PET) images as inputs to various deep learning models to automatically predict PD-L1 in non-small cell lung cancer (NSCLC). We compared three different architectures (ResNet, DenseNet, and EfficientNet) and considered different input data (CT only, PET only, PET/CT early fusion, PET/CT late fusion without as well as with partially and fully shared weights to determine the best model performance. Models were assessed utilizing areas under the receiver operating characteristic curves (AUCs) considering their 95% confidence intervals (CI). The fusion of PET and CT images as input yielded better performance for PD-L1 classification. The different data fusion schemes systematically outperformed their individual counterparts when used as input of the various deep models. Furthermore, early fusion consistently outperformed late fusion, probably as a result of its capacity to capture more complicated patterns by merging PET and CT derived content at a lower level. When we looked more closely at the effects of weight sharing in late fusion architectures, we discovered that while it might boost model stability, it did not always result in better results. This suggests that although weight sharing could be beneficial when modality parameters are similar, the anatomical and metabolic information provided by CT and PET scans are too dissimilar to consistently lead to improved PD-L1 status predictions.
66Development of a novel combined nomogram model integrating deep learning-pathomics, radiomics and immunoscore to predict postoperative outcome of colorectal cancer lung metastasis patients.PubMed
Renjie Wang, Weixing Dai, Jing Gong, et al.
J Hematol Oncol. 2022 Jan 24;15(1):11. doi: 10.1186/s13045-022-01225-3.
Limited previous studies focused on the death and progression risk stratification of colorectal cancer (CRC) lung metastasis patients. The aim of this study is to construct a nomogram model combing machine learning-pathomics, radiomics features, Immunoscore and clinical factors to predict the postoperative outcome of CRC patients with lung metastasis. In this study, a total of 103 CRC patients having metastases limited to lung and undergoing radical lung resection were identified. Patch-level convolutional neural network training in weakly supervised manner was used to perform whole slides histopathological images survival analysis. Synthetic minority oversampling technique and support vector machine classifier were used to identify radiomics features and build predictive signature. The Immunoscore for each patient was calculated from the density of CD3+ and CD8+ cells at the invasive margin and the center of metastatic tumor which were assessed on consecutive sections of automated digital pathology. Finally, pathomics and radiomics signatures were successfully developed to predict the overall survival (OS) and disease free survival (DFS) of patients. The predicted pathomics and radiomics scores are negatively correlated with Immunoscore and they are three independent prognostic factors for OS and DFS prediction. The combined nomogram showed outstanding performance in predicting OS (AUC = 0.860) and DFS (AUC = 0.875). The calibration curve and decision curve analysis demonstrated the considerable clinical usefulness of the combined nomogram. Taken together, the developed nomogram model consisting of machine learning-pathomics signature, radiomics signature, Immunoscore and clinical features could be reliable in predicting postoperative OS and DFS of colorectal lung metastasis patients.
67Whole slide image-based weakly supervised deep learning for predicting major pathological response in non-small cell lung cancer following neoadjuvant chemoimmunotherapy: a multicenter, retrospective, cohort study.PubMed
Dan Han, Hao Li, Xin Zheng, et al.
Front Immunol. 2024 Sep 20;15:1453232. doi: 10.3389/fimmu.2024.1453232. eCollection 2024.
OBJECTIVE: Develop a predictive model utilizing weakly supervised deep learning techniques to accurately forecast major pathological response (MPR) in patients with resectable non-small cell lung cancer (NSCLC) undergoing neoadjuvant chemoimmunotherapy (NICT), by leveraging whole slide images (WSIs). METHODS: This retrospective study examined pre-treatment WSIs from 186 patients with non-small cell lung cancer (NSCLC), using a weakly supervised learning framework. We employed advanced deep learning architectures, including DenseNet121, ResNet50, and Inception V3, to analyze WSIs on both micro (patch) and macro (slide) levels. The training process incorporated innovative data augmentation and normalization techniques to bolster the robustness of the models. We evaluated the performance of these models against traditional clinical predictors and integrated them with a novel pathomics signature, which was developed using multi-instance learning algorithms that facilitate feature aggregation from patch-level probability distributions. RESULTS: Univariate and multivariable analyses confirmed histology as a statistically significant prognostic factor for MPR (-value< 0.05). In patch model evaluations, DenseNet121 led in the validation set with an area under the curve (AUC) of 0.656, surpassing ResNet50 (AUC = 0.626) and Inception V3 (AUC = 0.654), and showed strong generalization in external testing (AUC = 0.611). Further evaluation through visual inspection of patch-level data integration into WSIs revealed XGBoost's superior class differentiation and generalization, achieving the highest AUCs of 0.998 in training and robust scores of 0.818 in validation and 0.805 in testing. Integrating pathomics features with clinical data into a nomogram yielded AUC of 0.819 in validation and 0.820 in testing, enhancing discriminative accuracy. Gradient-weighted Class Activation Mapping (Grad-CAM) and feature aggregation methods notably boosted the model's interpretability and feature modeling. CONCLUSION: The application of weakly supervised deep learning to WSIs offers a powerful tool for predicting MPR in NSCLC patients treated with NICT.
68Fusion of shallow and deep features from F-FDG PET/CT for predicting EGFR-sensitizing mutations in non-small cell lung cancer.PubMed
Xiaohui Yao, Yuan Zhu, Zhenxing Huang, et al.
Quant Imaging Med Surg. 2024 Aug 1;14(8):5460-5472. doi: 10.21037/qims-23-1028. Epub 2024 Jan 19.
BACKGROUND: Non-small cell lung cancer (NSCLC) patients with epidermal growth factor receptor-sensitizing (EGFR-sensitizing) mutations exhibit a positive response to tyrosine kinase inhibitors (TKIs). Given the limitations of current clinical predictive methods, it is critical to explore radiomics-based approaches. In this study, we leveraged deep-learning technology with multimodal radiomics data to more accurately predict EGFR-sensitizing mutations. METHODS: A total of 202 patients who underwent both flourine-18 fluorodeoxyglucose positron emission tomography/computed tomography (F-FDG PET/CT) scans and EGFR sequencing prior to treatment were included in this study. Deep and shallow features were extracted by a residual neural network and the Python package PyRadiomics, respectively. We used least absolute shrinkage and selection operator (LASSO) regression to select predictive features and applied a support vector machine (SVM) to classify the EGFR-sensitive patients. Moreover, we compared predictive performance across different deep models and imaging modalities. RESULTS: In the classification of EGFR-sensitive mutations, the areas under the curve (AUCs) of ResNet-based deep-shallow features and only shallow features from different multidata were as follows: RES_TRAD, PET/CT . CT-only . PET-only: 0.94 . 0.89 . 0.92; and ONLY_TRAD, PET/CT . CT-only . PET-only: 0.68 . 0.50 . 0.38. Additionally, the receiver operating characteristic (ROC) curves of the model using both deep and shallow features were significantly different from those of the model built using only shallow features (P<0.05). CONCLUSIONS: Our findings suggest that deep features significantly enhance the detection of EGFR-sensitizing mutations, especially those extracted with ResNet. Moreover, PET/CT images are more effective than CT-only and PET-only images in producing EGFR-sensitizing mutation-related signatures.
69iMLGAM: Integrated Machine Learning and Genetic Algorithm-driven Multiomics analysis for pan-cancer immunotherapy response prediction.PubMed
Bicheng Ye, Jun Fan, Lei Xue, et al.
Imeta. 2025 Mar 8;4(2):e70011. doi: 10.1002/imt2.70011. eCollection 2025 Apr.
To address the substantial variability in immune checkpoint blockade (ICB) therapy effectiveness, we developed an innovative R package called integrated Machine Learning and Genetic Algorithm-driven Multiomics analysis (iMLGAM), which establishes a comprehensive scoring system for predicting treatment outcomes through advanced multi-omics data integration. Our research demonstrates that iMLGAM scores exhibit superior predictive performance across independent cohorts, with lower scores correlating significantly with enhanced therapeutic responses and outperforming existing clinical biomarkers. Detailed analysis revealed that tumors with low iMLGAM scores display distinctive immune microenvironment characteristics, including increased immune cell infiltration and amplified antitumor immune responses. Critically, through clustered regularly interspaced short palindromic repeats screening, we identified Centrosomal Protein 55 () as a key molecule modulating tumor immune evasion, mechanistically confirming its role in regulating T cell-mediated antitumor immune responses. These findings not only validate iMLGAM as a powerful prognostic tool but also propose as a promising therapeutic target, offering novel strategies to enhance ICB treatment efficacy. The iMLGAM package is freely available on GitHub (https://github.com/Yelab1994/iMLGAM), providing researchers with an innovative approach to personalized cancer immunotherapy prediction.
70Integration of multiple machine learning approaches develops a gene mutation-based classifier for accurate immunotherapy outcomes.PubMed
Run Shi, Jing Sun, Zhaokai Zhou, et al.
NPJ Precis Oncol. 2025 Feb 27;9(1):54. doi: 10.1038/s41698-025-00842-8.
In addition to traditional biomarkers like PD-(L)1 expression and tumor mutation burden (TMB), more reliable methods for predicting immune checkpoint blockade (ICB) response in cancer patients are urgently needed. This study utilized multiple machine learning approaches on nonsynonymous mutations to identify key mutations that are most significantly correlated to ICB response. We proposed a classifier, Gene mutation-based Predictive Signature (GPS), to categorize patients based on their predicted response and clinical outcomes post-ICB therapy. GPS outperformed conventional predictors when validated in independent cohorts. Multi-omics analysis and multiplex immunohistochemistry (mIHC) revealed insights into tumor immunogenicity, immune responses, and the tumor microenvironment (TME) in lung adenocarcinoma (LUAD) across different GPS groups. Finally, we validated distinct responses of different GPS samples to ICB in an ex-vivo tumor organoid-PBMC co-culture model. Overall, our findings highlight a simple, robust classifier for accurate ICB response prediction, which could reduce costs, shorten testing times, and facilitate clinical implementation.
71Integration of Clinical Trial Spatial Multiomics Analysis and Virtual Clinical Trials Enables Immunotherapy Response Prediction and Biomarker Discovery.PubMed
Shuming Zhang, Atul Deshpande, Babita K Verma, et al.
Cancer Res. 2024 Aug 15;84(16):2734-2748. doi: 10.1158/0008-5472.CAN-24-0943.
Due to the lack of treatment options, there remains a need to advance new therapeutics in hepatocellular carcinoma (HCC). The traditional approach moves from initial molecular discovery through animal models to human trials to advance novel systemic therapies that improve treatment outcomes for patients with cancer. Computational methods that simulate tumors mathematically to describe cellular and molecular interactions are emerging as promising tools to simulate the impact of therapy entirely in silico, potentially greatly accelerating delivery of new therapeutics to patients. To facilitate the design of dosing regimens and identification of potential biomarkers for immunotherapy, we developed a new computational model to track tumor progression at the organ scale while capturing the spatial heterogeneity of the tumor in HCC. This computational model of spatial quantitative systems pharmacology was designed to simulate the effects of combination immunotherapy. The model was initiated using literature-derived parameter values and fitted to the specifics of HCC. Model validation was done through comparison with spatial multiomics data from a neoadjuvant HCC clinical trial combining anti-PD1 immunotherapy and a multitargeted tyrosine kinase inhibitor cabozantinib. Validation using spatial proteomics data from imaging mass cytometry demonstrated that closer proximity between CD8 T cells and macrophages correlated with nonresponse. We also compared the model output with Visium spatial transcriptomics profiling of samples from posttreatment tumor resections in the clinical trial and from another independent study of anti-PD1 monotherapy. Spatial transcriptomics data confirmed simulation results, suggesting the importance of spatial patterns of tumor vasculature and TGFβ in tumor and immune cell interactions. Our findings demonstrate that incorporating mathematical modeling and computer simulations with high-throughput spatial multiomics data provides a novel approach for patient outcome prediction and biomarker discovery. Significance: Incorporating mathematical modeling and computer simulations with high-throughput spatial multiomics data provides an effective approach for patient outcome prediction and biomarker discovery.
72Integration of longitudinal deep-radiomics and clinical data improves the prediction of durable benefits to anti-PD-1/PD-L1 immunotherapy in advanced NSCLC patients.PubMed
Benito Farina, Ana Delia Ramos Guerra, David Bermejo-Peláez, et al.
J Transl Med. 2023 Mar 5;21(1):174. doi: 10.1186/s12967-023-04004-x.
BACKGROUND: Identifying predictive non-invasive biomarkers of immunotherapy response is crucial to avoid premature treatment interruptions or ineffective prolongation. Our aim was to develop a non-invasive biomarker for predicting immunotherapy clinical durable benefit, based on the integration of radiomics and clinical data monitored through early anti-PD-1/PD-L1 monoclonal antibodies treatment in patients with advanced non-small cell lung cancer (NSCLC). METHODS: In this study, 264 patients with pathologically confirmed stage IV NSCLC treated with immunotherapy were retrospectively collected from two institutions. The cohort was randomly divided into a training (n = 221) and an independent test set (n = 43), ensuring the balanced availability of baseline and follow-up data for each patient. Clinical data corresponding to the start of treatment was retrieved from electronic patient records, and blood test variables after the first and third cycles of immunotherapy were also collected. Additionally, traditional radiomics and deep-radiomics features were extracted from the primary tumors of the computed tomography (CT) scans before treatment and during patient follow-up. Random Forest was used to implementing baseline and longitudinal models using clinical and radiomics data separately, and then an ensemble model was built integrating both sources of information. RESULTS: The integration of longitudinal clinical and deep-radiomics data significantly improved clinical durable benefit prediction at 6 and 9 months after treatment in the independent test set, achieving an area under the receiver operating characteristic curve of 0.824 (95% CI: [0.658,0.953]) and 0.753 (95% CI: [0.549,0.931]). The Kaplan-Meier survival analysis showed that, for both endpoints, the signatures significantly stratified high- and low-risk patients (p-value< 0.05) and were significantly correlated with progression-free survival (PFS6 model: C-index 0.723, p-value = 0.004; PFS9 model: C-index 0.685, p-value = 0.030) and overall survival (PFS6 models: C-index 0.768, p-value = 0.002; PFS9 model: C-index 0.736, p-value = 0.023). CONCLUSIONS: Integrating multidimensional and longitudinal data improved clinical durable benefit prediction to immunotherapy treatment of advanced non-small cell lung cancer patients. The selection of effective treatment and the appropriate evaluation of clinical benefit are important for better managing cancer patients with prolonged survival and preserving quality of life.
73New horizons at the interface of artificial intelligence and translational cancer research.PubMed
Josephine Yates, Eliezer M Van Allen
Cancer Cell. 2025 Apr 14;43(4):708-727. doi: 10.1016/j.ccell.2025.03.018.
Artificial intelligence (AI) is increasingly being utilized in cancer research as a computational strategy for analyzing multiomics datasets. Advances in single-cell and spatial profiling technologies have contributed significantly to our understanding of tumor biology, and AI methodologies are now being applied to accelerate translational efforts, including target discovery, biomarker identification, patient stratification, and therapeutic response prediction. Despite these advancements, the integration of AI into clinical workflows remains limited, presenting both challenges and opportunities. This review discusses AI applications in multiomics analysis and translational oncology, emphasizing their role in advancing biological discoveries and informing clinical decision-making. Key areas of focus include cellular heterogeneity, tumor microenvironment interactions, and AI-aided diagnostics. Challenges such as reproducibility, interpretability of AI models, and clinical integration are explored, with attention to strategies for addressing these hurdles. Together, these developments underscore the potential of AI and multiomics to enhance precision oncology and contribute to advancements in cancer care.
74Towards the Prediction of Responses to Cancer Immunotherapy: A Multi-Omics Review.PubMed
Weichu Tao, Qian Sun, Bingxiang Xu, et al.
Life (Basel). 2025 Feb 12;15(2):283. doi: 10.3390/life15020283.
Tumor treatment has undergone revolutionary changes with the development of immunotherapy, especially immune checkpoint inhibitors. Because not all patients respond positively to immune therapeutic agents, and severe immune-related adverse events (irAEs) are frequently observed, the development of the biomarkers evaluating the response of a patient is key for the application of immunotherapy in a wider range. Recently, various multi-omics features measured by high-throughput technologies, such as tumor mutation burden (TMB), gene expression profiles, and DNA methylation profiles, have been proved to be sensitive and accurate predictors of the response to immunotherapy. A large number of predictive models based on these features, utilizing traditional machine learning or deep learning frameworks, have also been proposed. In this review, we aim to cover recent advances in predicting tumor immunotherapy response using multi-omics features. These include new measurements, research cohorts, data sources, and predictive models. Key findings emphasize the importance of TMB, neoantigens, MSI, and mutational signatures in predicting ICI responses. The integration of bulk and single-cell RNA sequencing has enhanced our understanding of the tumor immune microenvironment and enabled the identification of predictive biomarkers like PD-L1 and IFN-γ signatures. Public datasets and machine learning models have also improved predictive tools. However, challenges remain, such as the need for large and diverse clinical datasets, standardization of multi-omics data, and model interpretability. Future research will require collaboration among researchers, clinicians, and data scientists to address these issues and enhance cancer immunotherapy precision.
75Development of a prognostic model for patients with extensive-stage small cell lung cancer undergoing immunotherapy and chemotherapy.PubMed
Yunbin Gao, Lixia Zhang, Meng Yan, et al.
Front Immunol. 2025 Mar 7;16:1561333. doi: 10.3389/fimmu.2025.1561333. eCollection 2025.
PURPOSE: In this study, we aimed to develop a predictive model for patients receiving chemotherapy and immunotherapy for extensive-stage small cell lung cancer. METHODS: We retrospectively analyzed 112 extensive-stage small cell lung cancer patients treated with first-line immunotherapy and chemotherapy. The relevant clinical data were collected to evaluate the changes during the treatment. The best subset regression, univariate analysis, and LASSO regression with cross-validation were applied for variable selection and model establishment. The nomograms for 1- and 2-year survival probabilities were established, and the calibration curve was utilized to evaluate the correspondence between actual and predicted survival. The model prediction capacity was assessed using decision curve analysis, calibration curves, and receiver operating characteristic curves. Moreover, five-fold cross-validation was conducted for internal validation. According to risk score, the patients were assigned to high- and low-risk groups, and survival curves were generated for each group. RESULTS: The LASSO regression model was established based on the variables such as age, ECOG, metastatic sites, NLR, and immunotherapy cycles. This predictive model displayed robust performance, evidenced by the Area Under the Curve of 0.887 and concordance index of 0.759. The nomogram effectively predicted 1- and 2-year survival probabilities and demonstrated a high degree of calibration. The decision curve analysis displayed that the model possessed superior predictive capability. The risk stratification for patients with high- and low-risk categories facilitated more individualized survival assessment. CONCLUSION: The study successfully developed a prognostic model for extensive-stage small cell lung cancer patients undergoing immunotherapy and chemotherapy, demonstrating the good accuracy and predictability.
76A prediction model for response to immune checkpoint inhibition in advanced melanoma.PubMed
Isabella A J van Duin, Rik J Verheijden, Paul J van Diest, et al.
Int J Cancer. 2024 May 15;154(10):1760-1771. doi: 10.1002/ijc.34853. Epub 2024 Jan 31.
Predicting who will benefit from treatment with immune checkpoint inhibition (ICI) in patients with advanced melanoma is challenging. We developed a multivariable prediction model for response to ICI, using routinely available clinical data including primary melanoma characteristics. We used a population-based cohort of 3525 patients with advanced cutaneous melanoma treated with anti-PD-1-based therapy. Our prediction model for predicting response within 6 months after ICI initiation was internally validated with bootstrap resampling. Performance evaluation included calibration, discrimination and internal-external cross-validation. Included patients received anti-PD-1 monotherapy (n = 2366) or ipilimumab plus nivolumab (n = 1159) in any treatment line. The model included serum lactate dehydrogenase, World Health Organization performance score, type and line of ICI, disease stage and time to first distant recurrence-all at start of ICI-, and location and type of primary melanoma, the presence of satellites and/or in-transit metastases at primary diagnosis and sex. The over-optimism adjusted area under the receiver operating characteristic was 0.66 (95% CI: 0.64-0.66). The range of predicted response probabilities was 7%-81%. Based on these probabilities, patients were categorized into quartiles. Compared to the lowest response quartile, patients in the highest quartile had a significantly longer median progression-free survival (20.0 vs 2.8 months; P < .001) and median overall survival (62.0 vs 8.0 months; P < .001). Our prediction model, based on routinely available clinical variables and primary melanoma characteristics, predicts response to ICI in patients with advanced melanoma and discriminates well between treated patients with a very good and very poor prognosis.
77Comprehensive analyses of the heterogeneity and prognostic significance of tumor-infiltrating immune cells in non-small-cell lung cancer: Development and validation of an individualized prognostic model.PubMed
Zhaofei Pang, Xiaowei Chen, Yu Wang, et al.
Int Immunopharmacol. 2020 Sep;86:106744. doi: 10.1016/j.intimp.2020.106744. Epub 2020 Jul 2.
Understanding the role of tumor-infiltrating immune cells (TIICs) in non-small cell lung cancer (NSCLC) is critical to finding new prognostic biomarkers and improving prognostic evaluation. Herein, we aimed to comprehensively analyze tumor-infiltrating pattern of TIICs in NSCLC and build a TIICs-associated, risk-stratification prognostic model for clinical practice. We applied CIBERSORT and ESTIMATE computational methods to analyze RNA-seq samples of 852 NSCLC patients from The Cancer Genome Atlas (TCGA). Prognotic factors were identified by univariate and multivariate Cox regression analyses for overall survival (OS). A novel model was developed to predict the 1-, 3- and 5-year OS of NSCLC based on the TCGA cohort, validated by external validation cohorts (GSE31210, GSE37745), and then evaluated by C-indexes and calibration plots. Significant heterogeneity in the infiltrating patterns of TIICs was shown among various pathological subtypes of NSCLC and between different genders. Further analyses showed that abundances of naive B cells (NBCs), T cells and mast cells (MCs) were positively correlated with prognosis. Tumor samples with high T cells abundances tended to have higher expression levels of immune checkpoint genes (PD-1, PD-L1, CTLA-4). A new immune-gene related index (IGRI) was built by five immune-related differentially expressed genes (DEGs) including BTK, CCR2, CLEC10A, NCR3 and PRKCB, which were closely correlated with TIICs abundances and prognosis. Tumor stage, IGRI, abundances of NBCs, T cells, MCs and NK cells were significant independent prognostic factors and were included in the nomogram as predictors. The internal and external calibration plots of the nomogram were in excellent agreement. This study reveals that TIICs are significantly correlated with clinicopathological features and prognosis in NSCLC and thus can be potential prognostic biomarker or therapeutic target. The remarkable heterogeneity of TIICs suggests that specific infiltrating patterns of TIICs should also be taken into consideration when determining individualized immunotherapy strategies for NSCLC patients.
78Development of a lung immune prognostic index-based nomogram model for predicting overall survival and immune-related adverse events in non-small cell lung cancer patients treated with sintilimab.PubMed
Jian Xu, Tingting Peng, Kaikai Fan, et al.
Front Immunol. 2025 May 8;16:1569689. doi: 10.3389/fimmu.2025.1569689. eCollection 2025.
BACKGROUND: Sintilimab, a programmed cell death protein-1 (PD-1) inhibitor, has shown efficacy in non-small cell lung cancer (NSCLC), though response heterogeneity persists. Previous studies suggest that the Lung Immune Prognostic Index (LIPI) may predict prognosis and immune-related adverse events (irAEs) in immunotherapy. This study aimed to develop and validate LIPI-based nomograms for predicting overall survival (OS) and irAEs in NSCLC patients treated with sintilimab. METHODS: Multicenter data stratified 356 patients into training, internal validation, and external validation cohorts. Propensity score matching (PSM) balanced baseline characteristics. Multivariable Cox regression identified OS and irAEs predictors, and nomograms were constructed using significant variables. Model performance was evaluated via concordance index (C-index), time-dependent receiver operating characteristic (ROC) curves, calibration plots, and decision curve analysis (DCA). Kaplan-Meier analysis assessed risk stratification. RESULTS: Independent prognostic factors for OS include clinical stage, treatment lines, LIPI scores and albumin level. Among them, stage IV (hazard ratio [HR]=1.725, 95% confidence interval [CI] 1.529-1.902), treatment line ≥2 (HR=1.302, 95%CI: 1.125-1.569), LIPI intermediate (HR=1.736, 95%CI: 1.586-1.925), LIPI poor (HR=1.568, 95% CI: 1.361-1.637) and albumin level≥35 (HR=1.802, 95%CI: 1.698-2.023) were risk factors for OS. The OS prediction model demonstrated excellent discrimination across all cohorts, with time-dependent AUCs maintaining 0.770-0.850 for 1-2 year predictions. Consistent calibration was observed (C-index: training=0.778, internal validation=0.793, external validation=0.790). For irAEs prediction, significant predictors included age, sex, Eastern Cooperative Oncology Group performance status (ECOG PS), and LIPI scores. Similarly, the irAEs model showed robust performance (AUCs 0.754-0.835 for 1-2 year predictions; C-index: training=0.805, internal validation=0.825, external validation=0.775). Both nomograms significantly outperformed single-variable predictions in Kaplan-Meier analyses. DCA confirmed superior net clinical benefit. CONCLUSION: LIPI-based nomograms effectively predicted OS and irAEs in sintilimab-treated NSCLC patients, offering valuable tools for personalized treatment and clinical decision-making.
79Development and validation of a nomogram for evaluating the prognosis of immunotherapy plus antiangiogenic therapy in non-small cell lung cancer.PubMed
Hao Huang, Yao Chen, Xuezi Weng, et al.
Cancer Cell Int. 2022 Aug 21;22(1):261. doi: 10.1186/s12935-022-02675-y.
BACKGROUND: With the combination therapy of PD-1/PD-L1 antibody and antiangiogenic drugs used widely in clinic, a novel method to estimate the prognosis of patients is needed. We aimed to develop a nomogram to examine prognosis of anti-PD-1/PD-L1 antibody plus bevacizumab in non-small cell lung cancer (NSCLC) patients. METHODS: We developed a nomogram using the cohort involving 204 NSCLC patients who treated with immunotherapy and anti-angiogenesis therapy. The nomogram was validated under the same conditions in another cohort with 69 patients. Prognostic factors were analyzed by Cox regression analysis. The nomogram was internally validated using bootstrap resampling and then externally validated. Performance was assessed using concordance index, calibration curve and decision curve analysis. Clinical utility was evaluated using receiver operation characteristic curve. RESULTS: Pleural metastasis (P = 0.001, HR = 2.980, 95%CI 1.521-5.837), ANC (P < 0.001, HR = 5.139, 95%CI 2.081-12.691), ALC (P = 0.010, HR = 0.331, 95%CI 0.142-0.771), B cells (P = 0.005, HR = 0.329, 95%CI 0.151-0.714), Treg cells (P = 0.002, HR = 2.934, 95%CI 1.478-5.826) were independent prognostic factors. The calibration curves showed good consistency and the C-index of nomogram were 0.808, 0.741 in training and external validation cohort, respectively. The area under the curve (AUC) in receiver operation characteristic curves (ROC) are 0.833 (P < 0.001) and 0.908 (P < 0.001), respectively. CONCLUSION: We build an accurate and convenient nomogram to predict long-time overall survival (OS) of NSCLC patients treated with PD-1/PD-L1 antibody and antiangiogenic drugs and validated this nomogram. The nomogram might be helpful to clinicians to estimate long-time OS of NSCLC patients treated with PD-1/PD-L1 antibody and antiangiogenic drugs.
80Predictive value of inflammation and nutritional index in immunotherapy for stage IV non-small cell lung cancer and model construction.PubMed
Wenqian Lei, Wei Wang, Shixiang Qin, et al.
Sci Rep. 2024 Jul 30;14(1):17511. doi: 10.1038/s41598-024-66813-4.
Identifying individuals poised to gain from immune checkpoint inhibitor (ICI) therapies is a pivotal element in the realm of tailored healthcare. The expression level of Programmed Death Ligand 1 (PD-L1) has been linked to the response to ICI therapy, but its assessment typically requires substantial tumor tissue, which can be challenging to obtain. In contrast, blood samples are more feasible for clinical application. A number of promising peripheral biomarkers have been proposed to overcome this hurdle. This research aims to evaluate the prognostic utility of the albumin-to-lactate dehydrogenase ratio (LAR), the Pan-immune-inflammation Value (PIV), and the Prognostic Nutritional Index (PNI) in predicting the response to ICI therapy in individuals with advanced non-small cell lung cancer (NSCLC). Furthermore, the study seeks to construct a predictive nomogram that includes these markers to facilitate the selection of patients with a higher likelihood of benefiting from ICI therapy. A research initiative scrutinized the treatment records of 157 advanced NSCLC patients who received ICI therapy across two Jiangxi medical centers. The cohort from Jiangxi Provincial People's Hospital (comprising 108 patients) was utilized for the training dataset, while the contingent from Jiangxi Cancer Hospital (49 patients) served for validation purposes. Stratification was based on established LAR, PIV, and PNI benchmarks to explore associations with DCR and ORR metrics. Factorial influences on ICI treatment success were discerned through univariate and multivariate Cox regression analysis. Subsequently, a Nomogram was devised to forecast outcomes, its precision gauged by ROC and calibration curves, DCA analysis, and cross-institutional validation. In the training group, the optimal threshold values for LAR, PIV, and PNI were identified as 5.205, 297.49, and 44.6, respectively. Based on these thresholds, LAR, PIV, and PNI were categorized into high (≥ Cut-off) and low (< Cut-off) groups. Patients with low LAR (L-LAR), low PIV (L-PIV), and high PNI (H-PNI) exhibited a higher disease control rate (DCR) (P < 0.05) and longer median progression-free survival (PFS) (P < 0.05). Cox multivariate analysis indicated that PS, malignant pleural effusion, liver metastasis, high PIV (H-PIV), and low PNI (L-PNI) were risk factors adversely affecting the efficacy of immunotherapy (P < 0.05). The Nomogram model predicted a concordance index (C-index) of 0.78 (95% CI: 0.73-0.84). The areas under the ROC curve (AUC) for the training group at 6, 9, and 12 months were 0.900, 0.869, and 0.866, respectively, while the AUCs for the external validation group at the same time points were 0.800, 0.886, and 0.801, respectively. Throughout immunotherapy, PIV and PNI could act as prospective indicators for forecasting treatment success in NSCLC patients, while the devised Nomogram model exhibits strong predictive performance for patient prognoses.
81Examining explainable clinical decision support systems with think aloud protocols.PubMed
Sabrina G Anjara, Adrianna Janik, Amy Dunford-Stenger, et al.
PLoS One. 2023 Sep 14;18(9):e0291443. doi: 10.1371/journal.pone.0291443. eCollection 2023.
Machine learning tools are increasingly used to improve the quality of care and the soundness of a treatment plan. Explainable AI (XAI) helps users in understanding the inner mechanisms of opaque machine learning models and is a driver of trust and adoption. Explanation methods for black-box models exist, but there is a lack of user studies on the interpretability of the provided explanations. We used a Think Aloud Protocol (TAP) to explore oncologists' assessment of a lung cancer relapse prediction system with the aim of refining the purpose-built explanation model for better credibility and utility. Novel to this context, TAP is used as a neutral methodology to elicit experts' thought processes and judgements of the AI system, without explicit prompts. TAP aims to elicit the factors which influenced clinicians' perception of credibility and usefulness of the system. Ten oncologists took part in the study. We conducted a thematic analysis of their verbalized responses, generating five themes that help us to understand the context within which oncologists' may (or may not) integrate an explainable AI system into their working day.
82An explainable AI-driven biomarker discovery framework for Non-Small Cell Lung Cancer classification.PubMed
Kountay Dwivedi, Ankit Rajpal, Sheetal Rajpal, et al.
Comput Biol Med. 2023 Feb;153:106544. doi: 10.1016/j.compbiomed.2023.106544. Epub 2023 Jan 12.
Non-Small Cell Lung Cancer (NSCLC) exhibits intrinsic heterogeneity at the molecular level that aids in distinguishing between its two prominent subtypes - Lung Adenocarcinoma (LUAD) and Lung Squamous Cell Carcinoma (LUSC). This paper proposes a novel explainable AI (XAI)-based deep learning framework to discover a small set of NSCLC biomarkers. The proposed framework comprises three modules - an autoencoder to shrink the input feature space, a feed-forward neural network to classify NSCLC instances into LUAD and LUSC, and a biomarker discovery module that leverages the combined network comprising the autoencoder and the feed-forward neural network. In the biomarker discovery module, XAI methods uncovered a set of 52 relevant biomarkers for NSCLC subtype classification. To evaluate the classification performance of the discovered biomarkers, multiple machine-learning models are constructed using these biomarkers. Using 10-Fold cross-validation, Multilayer Perceptron achieved an accuracy of 95.74% (±1.27) at 95% confidence interval. Further, using Drug-Gene Interaction Database, we observe that 14 of the discovered biomarkers are druggable. In addition, 28 biomarkers aid the prediction of the survivability of the patients. Out of 52 discovered biomarkers, we find that 45 biomarkers have been reported in previous studies on distinguishing between the two NSCLC subtypes. To the best of our knowledge, the remaining seven biomarkers have not yet been reported for NSCLC subtyping and could be further explored for their contribution to targeted therapy of lung cancer.
83Prediction of Brain Metastases Development in Patients With Lung Cancer by Explainable Artificial Intelligence From Electronic Health Records.PubMed
Zhao Li, Rongbin Li, Yujia Zhou, et al.
JCO Clin Cancer Inform. 2023 Apr;7:e2200141. doi: 10.1200/CCI.22.00141.
PURPOSE: Early detection of brain metastases (BMs) is critical for prompt treatment and optimal control of the disease. In this study, we seek to predict the risk of developing BM among patients diagnosed with lung cancer on the basis of electronic health record (EHR) data and to understand what factors are important for the model to predict BM development through explainable artificial intelligence approaches accurately. MATERIALS AND METHODS: We trained a recurrent neural network model, REverse Time AttentIoN (RETAIN), to predict the risk of developing BM using structured EHR data. To interpret the model's decision process, we analyzed the attention weights in the RETAIN model and the SHAP values from a feature attribution method, Kernel SHAP, to identify the factors contributing to BM prediction. RESULTS: We developed a high-quality cohort with 4,466 patients with BM from the Cerner Health Fact database, which contains over 70 million patients from more than 600 hospitals. RETAIN uses this data set to achieve the best area under the receiver operating characteristic curve at 0.825, a significant improvement over the baseline model. We also extended a feature attribution method, Kernel SHAP, to structured EHR data for model interpretation. Both RETAIN and Kernel SHAP can identify important features related to BM prediction. CONCLUSION: To the best of our knowledge, this is the first study to predict BM using structured EHR data. We achieved decent prediction performance for BM prediction and identified factors highly relevant to BM development. The sensitivity analysis demonstrated that both RETAIN and Kernel SHAP could discriminate unrelated features and put more weight on the features important to BM. Our study explored the potential of applying explainable artificial intelligence for future clinical applications.
84An explainable AI-assisted web application in cancer drug value prediction.PubMed
Sonali Kothari, Shivanandana Sharma, Sanskruti Shejwal, et al.
MethodsX. 2024 Apr 3;12:102696. doi: 10.1016/j.mex.2024.102696. eCollection 2024 Jun.
In recent years, there has been an increase in the interest in adopting Explainable Artificial Intelligence (XAI) for healthcare. The proposed system includes•An XAI model for cancer drug value prediction. The model provides data that is easy to understand and explain, which is critical for medical decision-making. It also produces accurate projections.•A model outperformed existing models due to extensive training and evaluation on a large cancer medication chemical compounds dataset.•Insights into the causation and correlation between the dependent and independent actors in the chemical composition of the cancer cell. While the model is evaluated on Lung Cancer data, the architecture offered in the proposed solution is cancer agnostic. It may be scaled out to other cancer cell data if the properties are similar. The work presents a viable route for customizing treatments and improving patient outcomes in oncology by combining XAI with a large dataset. This research attempts to create a framework where a user can upload a test case and receive forecasts with explanations, all in a portable PDF report.
85Single-cell multimodal analysis reveals tumor microenvironment predictive of treatment response in non-small cell lung cancer.PubMed
Yuanning Zheng, Christoph Sadée, Michael Ozawa, et al.
Sci Adv. 2025 May 23;11(21):eadu2151. doi: 10.1126/sciadv.adu2151.
Non-small cell lung cancer (NSCLC) constitutes over 80% of lung cancer cases and remains a leading cause of cancer-related mortality worldwide. Despite the advent of immune checkpoint inhibitors, their efficacy is limited to 27 to 45% of patients. Identifying likely treatment responders is essential for optimizing healthcare and improving quality of life. We generated multiplex immunofluorescence (mIF) images, histopathology, and RNA sequencing data from human NSCLC tissues. Through the analysis of mIF images, we characterized the spatial organization of 1.5 million cells based on the expression levels for 33 biomarkers. To enable large-scale characterization of tumor microenvironments, we developed NucSegAI, a deep learning model for automated nuclear segmentation and cellular classification in histology images. With this model, we analyzed the morphological, textural, and topological phenotypes of 45.6 million cells across 119 whole-slide images. Through unsupervised phenotype discovery, we identified specific lymphocyte phenotypes predictive of immunotherapy response. Our findings can improve patient stratification and guide selection of effective therapeutic regimens.
86Spatial transcriptomics of macrophage infiltration in non-small cell lung cancer reveals determinants of sensitivity and resistance to anti-PD1/PD-L1 antibodies.PubMed
Mathieu Larroquette, Jean-Philippe Guegan, Benjamin Besse, et al.
J Immunother Cancer. 2022 May;10(5). doi: 10.1136/jitc-2021-003890.
BACKGROUND: Tumor-associated macrophages (TAMs) having immunosuppressive properties are one of the most abundant immune cells in the tumor microenvironment (TME). Preclinical studies have highlighted the potential role of TAMs in resistance to immune checkpoint blockers (ICBs). Here, we investigated the predictive value of TAM infiltration in patients with non-small cell lung cancer (NSCLC) treated with ICBs and characterized their transcriptomic profiles. METHODS: Tumor samples were collected from 152 patients with NSCLC before ICB treatment onset. After immunohistochemical staining and image analysis, the correlation between CD163+ cell infiltration and survival was analyzed. Spatial transcriptomic analyses were performed using the NanoString GeoMx Immune Pathways assay to compare the gene expression profile of tumors with high or low levels of CD163+ cell infiltration and to identify determinants of response to ICBs in tumors with high CD163+ infiltration. RESULTS: Low intratumoral CD163+ cell infiltration was associated with longer progression-free survival (PFS; HR 0.61, 95% CI 0.40 to 0.94, p=0.023) and overall survival (OS; HR 0.48, 95% CI 0.28 to 0.80, p=0.004) under ICB treatment. Spatial transcriptomic profiles of 16 tumors revealed the upregulation of , CD27, and in tumors with high CD163+ cell infiltration. Moreover, in tumors with high macrophage infiltration, the upregulation of genes associated with the interferon-γ signaling pathway and the M1 phenotype was associated with better responses under immunotherapy. Surprisingly, we found also a significantly higher expression of in the tumors of responders. Analysis of three independent data sets confirmed that high expression was associated with an increased durable clinical benefit rate (47% vs 6%, p=0.004), PFS (median 10.89 months vs 1.67 months, p=0.001), and OS (median 23.11 months vs 2.66 months, p<0.001) under ICB treatment. CONCLUSIONS: Enrichment of TAMs in the TME of NSCLC is associated with resistance to immunotherapy regardless of the programmed death ligand 1 status and is driven by upregulation of , and gene expression within the tumor compartment. Our transcriptomic analyses identify new potential targets to alter TAM recruitment/polarization and highlight the complexity of the CSF1R pathway, which may not be a suitable target to improve ICB efficacy.
87Single-cell and spatial transcriptomic analyses revealing tumor microenvironment remodeling after neoadjuvant chemoimmunotherapy in non-small cell lung cancer.PubMed
Xiaolu Cui, Siyuan Liu, He Song, et al.
Mol Cancer. 2025 Apr 9;24(1):111. doi: 10.1186/s12943-025-02287-w.
Non-small cell lung cancer (NSCLC) represents the most common pathological type of lung cancer, and the combination of neoadjuvant immunotherapy with chemotherapy has emerged as the first-line treatment for NSCLC. Nevertheless, the efficacy of this therapeutic approach remains variable. The present study aims to examine the impact of chemoimmunotherapy in NSCLC patients, with a view to identifying key molecules, critical cell subpopulations, communication patterns and spatial distributions that potentially correlate with therapeutic sensitivity. A total of 16 lung cancer tissue samples were collected from a cohort of 12 NSCLC patients and subjected to single-cell RNA and spatial transcriptome sequencing. Our data demonstrated that the distribution of CD4 + Treg T cells and mCAFs indicated an immunosuppressive tumor microenvironment, while the accumulation of CD4 + Th17 T cells and iCAFs could act as a positive marker for the sensitivity to chemoimmunotherapy. Furthermore, a significant high level of SELENOP-macrophages was observed in tissues from positive responders, and a strong co-localization between SELENOP-macrophages and antigen-presenting cancer associated fibroblasts (CAFs) in the tumor boundaries was identified, indicating the cooperative roles of these two cell types in response to combined therapy. Moreover, SELENOP-macrophages were observed to be accumulated in tertiary lymphoid structures, which further suggested its critical role in recruiting lymphocytes. Furthermore, analysis of cell-cell communication, based on spatial transcriptomics, suggests that the interactions between SELENOP-macrophages, apCAFs, CD4 + and CD8 + T cells were significantly enhanced in responders. In addition, SELENOP-macrophages recruited CD4 + Naïve, Helper and CD8 + Naïve T cells through pathways such as the cholesterol, interleukin, chemokine and HLA when responding to combined therapy. The present study further unveils the dynamic spatial and transcriptional changes in the tumor microenvironment of non-small cell lung cancer in response to combination therapy.
88Single-cell RNA sequencing of lung adenocarcinoma reveals heterogeneity of immune response-related genes.PubMed
Ke-Yue Ma, Alexandra A Schonnesen, Amy Brock, et al.
JCI Insight. 2019 Feb 21;4(4). doi: 10.1172/jci.insight.121387.
Immunotherapy has emerged as a promising approach to treat cancer. However, partial responses across multiple clinical trials support the significance of characterizing intertumor and intratumor heterogeneity to achieve better clinical results and as potential tools in selecting patients for different types of cancer immunotherapies. Yet, the type of heterogeneity that informs clinical outcome and patient selection has not been fully explored. In particular, the lack of characterization of immune response-related genes in cancer cells hinders the further development of metrics to select and optimize immunotherapy. Therefore, we analyzed single-cell RNA-Seq data from lung adenocarcinoma patients and cell lines to characterize the intratumor heterogeneity of immune response-related genes and demonstrated their potential impact on the efficacy of immunotherapy. We discovered that IFN-γ signaling pathway genes are heterogeneously expressed and coregulated with other genes in single cancer cells, including MHC class II (MHCII) genes. The downregulation of genes in IFN-γ signaling pathways in cell lines corresponds to an acquired resistance phenotype. Moreover, analysis of 2 groups of tumor-restricted antigens, namely neoantigens and cancer testis antigens, revealed heterogeneity in their expression in single cells. These analyses provide a rationale for applying multiantigen combinatorial therapies to prevent tumor escape and establish a basis for future development of prognostic metrics based on intratumor heterogeneity.
89Single-cell transcriptome analysis revealed a suppressive tumor immune microenvironment in EGFR mutant lung adenocarcinoma.PubMed
Lei Yang, Yun-Ting He, Song Dong, et al.
J Immunother Cancer. 2022 Jan;10(2). doi: 10.1136/jitc-2021-003534.
BACKGROUNDS: Immunotherapy is less effective in patients with epidermal growth factor receptor (EGFR) mutant non-small-cell lung cancer (NSCLC). Lower programmed cell death-ligand 1 (PD-L1) expression and tumor mutation burden (TMB) are reported to be the underlying mechanism. Being another important factor to affect the efficacy of immunotherapy, tumor microenvironment (TME) characteristics of this subgroup of NSCLC are not comprehensively understood up to date. Hence, we initiated this study to describe the specific TME of EGFR-mutant lung adenocarcinoma (LUAD) from cellular compositional and functional perspectives to better understand the immune landscape of this most common subtype of NSCLC. METHODS: We used single-cell transcriptome sequencing and multiplex immunohistochemistry to investigate the immune microenvironment of EGFR-mutant and EGFR wild-type LUADs and determined the efficacy of immunotherapy. We analyzed single cells from nine treatment-naïve samples and compared them to three post-immunotherapy samples previously reported from single cell perspective using bioinformatics methods. RESULTS: We found that EGFR-mutant malignant epithelial cells had similar characteristics to the epithelial cells in non-responders. EGFR-mutant LUAD lacked CD8 tissue-resident memory (TRM) cells, which could promote tertiary lymphoid structure generation by secreting CXCL13. In addition, other cell types, including tumor-associated macrophages and cancer-associated fibroblasts, which are capable of recruiting, retaining, and expanding CD8 TRM cells in the TME, were also deficient in EGFR-mutant LUAD. Furthermore, EGFR-mutant LUAD had significantly less crosstalk between T cells and other cell types via programmed cell death-1 (PD-1) and PD-L1 or other immune checkpoints compared with EGFR wild-type LUAD. CONCLUSIONS: Our findings provide a comprehensive understanding of the immune landscape of EGFR-mutant LUAD at the single-cell level. Based on the results, many cellular components might have negative impact on the specific TME of EGFR-mutant LUAD through influencing CD8 TRM. Lack of CD8 TRM might be a key factor responsible for the suppressive TME of EGFR-mutant LUAD.
90The co-location of CD14+APOE+ cells and MMP7+ tumour cells contributed to worse immunotherapy response in non-small cell lung cancer.PubMed
Guangyu Fan, Tongji Xie, Le Tang, et al.
Clin Transl Med. 2024 Sep;14(9):e70009. doi: 10.1002/ctm2.70009.
Intra-tumour immune infiltration is a crucial determinant affecting immunotherapy response in non-small cell lung cancer (NSCLC). However, its phenotype and related spatial structure have remained elusive. To overcome these restrictions, we undertook a comprehensive study comprising spatial transcriptomic (ST) data (28 712 spots from six samples). We identified two distinct intra-tumour infiltration patterns: immune exclusion (characterised by myeloid cells) and immune activation (characterised by plasma cells). The immune exclusion and immune activation signatures showed adverse and favourable roles in NSCLC patients' survival, respectively. Notably, CD14+APOE+ cells were recognised as the main cell type in immune exclusion samples, with increased epithelial‒mesenchymal transition and decreased immune activities. The co-location of CD14+APOE+ cells and MMP7+ tumour cells was observed in both ST and bulk transcriptomics data, validated by multiplex immunofluorescence performed on 20 NSCLC samples. The co-location area exhibited the upregulation of proliferation-related pathways and hypoxia activities. This co-localisation inhibited T-cell infiltration and the formation of tertiary lymphoid structures. Both CD14+APOE+ cells and MMP7+ tumour cells were associated with worse survival. In an immunotherapy cohort from the ORIENT-3 clinical trial, NSCLC patients who responded unfavourably exhibited higher infiltration of CD14+APOE+ cells and MMP7+ tumour cells. Within the co-location area, the MK, SEMA3 and Macrophage migration inhibitory factor (MIF) signalling pathway was most active in cell‒cell communication. This study identified immune exclusion and activation patterns in NSCLC and the co-location of CD14+APOE+ cells and MMP7+ tumour cells as contributors to immune resistance.
91Comprehensive multi-omics integration uncovers mitochondrial gene signatures for prognosis and personalized therapy in lung adenocarcinoma.PubMed
Wenjia Zhang, Lei Zhao, Tiansheng Zheng, et al.
J Transl Med. 2024 Oct 21;22(1):952. doi: 10.1186/s12967-024-05754-y.
The therapeutic efficacy of lung adenocarcinoma (LUAD), the most prevalent histological subtype of primary lung cancer, remains inadequate, with accurate prognostic assessment posing significant challenges. This study sought to elucidate the prognostic significance of mitochondrial-related genes in LUAD through an integrative multi-omics approach, aimed at developing personalized therapeutic strategies. Utilizing transcriptomic and single-cell RNA sequencing (scRNA-seq) data, alongside clinical information from publicly available databases, we first applied dimensionality reduction and clustering techniques to the LUAD single-cell dataset, focusing on the subclassification of fibroblasts, epithelial cells, and T cells. Mitochondrial-related prognostic genes were subsequently identified using TCGA-LUAD data, and LUAD cases were stratified into distinct molecular subtypes through consensus clustering, allowing for the exploration of gene expression profiles and clinical feature distributions across subtypes. By leveraging an ensemble of machine learning algorithms, we developed an Artificial Intelligence-Derived Prognostic Signature (AIDPS) model based on mitochondrial-related genes and validated its prognostic accuracy across multiple independent datasets. The AIDPS model demonstrated robust predictive power for LUAD patient outcomes, revealing significant differences in responses to immunotherapy and chemotherapy, as well as survival outcomes between risk groups. Furthermore, we conducted comprehensive analyses of tumor mutation burden (TMB), immune microenvironment characteristics, and genome-wide association study (GWAS) data, providing additional insights into the mechanistic roles of mitochondrial-related genes in LUAD pathogenesis. This study not only offers a novel approach to improving prognostic assessments in LUAD but also establishes a strong foundation for the development of personalized therapeutic interventions.
92Multi-Omics Integration of Lactylation- and PANoptosis-Based Signatures in Lung Adenocarcinoma: Prognostic Stratification and Immune Response.PubMed
Zhenhao Xu, Yisha Huang, Xiuling Yu, et al.
Int J Mol Sci. 2025 Jun 23;26(13):5999. doi: 10.3390/ijms26135999.
Lactylation and PANoptosis are emerging modes of tumor progression regulation; however, their interplay and effect on the prognosis for lung adenocarcinoma (LUAD) remain unclear. This research analyzed both bulk and single-cell transcriptomic profiles of LUAD and identified 506 potential markers related to lactylation and PANoptosis. Employing 117 machine learning approaches and 5 LUAD datasets, lactylation and PANoptosis-related signatures (LAPRS) and further predictive nomograms were constructed with 85 prognostic genes. The performance of LAPRS was validated with multifaceted validation, including Kaplan-Meier analysis, time-dependent ROC curves and comparison with 55 existing LUAD models. LAPRS enabled the stratification of LUAD patients into high- and low-risk subgroups. Through additional investigation, high-risk individuals showed elevated genomic alterations, reduced immune infiltration, and poorer immunotherapy response, while low-risk individuals showed better drug sensitivity and a higher tumor mutation burden. Further analysis via 18 models and experimental validation revealed APOL1 as a poor prognostic factor, potentially interacting with the lactylation-related gene VIM through TNF signaling. This research clarifies the mechanistic roles of lactylation and PANoptosis in LUAD and proposes APOL1 as a novel prognostic marker, offering insights for therapeutic stratification.
93Implementing circulating tumor DNA as a prognostic biomarker in resectable non-small cell lung cancer.PubMed
Chris Abbosh, Darren Hodgson, Gary J Doherty, et al.
Trends Cancer. 2024 Jul;10(7):643-654. doi: 10.1016/j.trecan.2024.04.004. Epub 2024 Jun 4.
Systemic treatment of resectable non-small cell lung cancer (NSCLC) is evolving with emerging neoadjuvant, perioperative, and adjuvant immunotherapy approaches. Circulating tumor DNA (ctDNA) detection at clinical diagnosis, during neoadjuvant therapy, or after resection may discern high-risk patients who might benefit from therapy escalation or switch. This Review summarizes translational implications of data supporting ctDNA-based risk determination in NSCLC and outstanding questions regarding ctDNA validity/utility as a prognostic biomarker. We discuss emerging ctDNA capabilities to refine clinical tumor-node-metastasis (TNM) staging in lung adenocarcinoma, ctDNA dynamics during neoadjuvant therapy for identifying patients deriving suboptimal benefit, and postoperative molecular residual disease (MRD) detection to escalate systemic therapy. Considering differential relapse characteristics in landmark MRD-negative/MRD-positive patients, we propose how ctDNA might integrate with pathological response data for optimal postoperative risk stratification.
94Circulating Tumor DNA Dynamics Predict Benefit from Consolidation Immunotherapy in Locally Advanced Non-Small Cell Lung Cancer.PubMed
Everett J Moding, Yufei Liu, Barzin Y Nabet, et al.
Nat Cancer. 2020 Feb;1(2):176-183. doi: 10.1038/s43018-019-0011-0. Epub 2020 Jan 20.
Circulating tumor DNA (ctDNA) molecular residual disease (MRD) following curative-intent treatment strongly predicts recurrence in multiple tumor types, but whether further treatment can improve outcomes in patients with MRD remains unclear. We applied CAPP-Seq ctDNA analysis to 218 samples from 65 patients receiving chemoradiation therapy (CRT) for locally advanced NSCLC, including 28 patients receiving consolidation immune checkpoint inhibition (CICI). Patients with undetectable ctDNA after CRT had excellent outcomes whether or not they received CICI. Among such patients, one died from CICI-related pneumonitis, highlighting the potential utility of only treating patients with MRD. In contrast, patients with MRD after CRT who received CICI had significantly better outcomes than patients who did not receive CICI. Furthermore, the ctDNA response pattern early during CICI identified patients responding to consolidation therapy. Our results suggest that CICI improves outcomes for NSCLC patients with MRD and that ctDNA analysis may facilitate personalization of consolidation therapy.
95Automated real-world data integration improves cancer outcome prediction.PubMed
Justin Jee, Christopher Fong, Karl Pichotta, et al.
Nature. 2024 Dec;636(8043):728-736. doi: 10.1038/s41586-024-08167-5. Epub 2024 Nov 6.
The digitization of health records and growing availability of tumour DNA sequencing provide an opportunity to study the determinants of cancer outcomes with unprecedented richness. Patient data are often stored in unstructured text and siloed datasets. Here we combine natural language processing annotations with structured medication, patient-reported demographic, tumour registry and tumour genomic data from 24,950 patients at Memorial Sloan Kettering Cancer Center to generate a clinicogenomic, harmonized oncologic real-world dataset (MSK-CHORD). MSK-CHORD includes data for non-small-cell lung (n = 7,809), breast (n = 5,368), colorectal (n = 5,543), prostate (n = 3,211) and pancreatic (n = 3,109) cancers and enables discovery of clinicogenomic relationships not apparent in smaller datasets. Leveraging MSK-CHORD to train machine learning models to predict overall survival, we find that models including features derived from natural language processing, such as sites of disease, outperform those based on genomic data or stage alone as tested by cross-validation and an external, multi-institution dataset. By annotating 705,241 radiology reports, MSK-CHORD also uncovers predictors of metastasis to specific organ sites, including a relationship between SETD2 mutation and lower metastatic potential in immunotherapy-treated lung adenocarcinoma corroborated in independent datasets. We demonstrate the feasibility of automated annotation from unstructured notes and its utility in predicting patient outcomes. The resulting data are provided as a public resource for real-world oncologic research.
96Navigating the Complexities of Artificial Intelligence-Enabled Real-World Data Collection for Oncology Pharmacovigilance.PubMed
Jack Gallifant, Leo Anthony Celi, Elad Sharon, et al.
JCO Clin Cancer Inform. 2024 May;8:e2400051. doi: 10.1200/CCI.24.00051.
This new editorial discusses the promise and challenges of successful integration of natural language processing methods into electronic health records for timely, robust, and fair oncology pharmacovigilance.
97Oncology Research: Clinical Trial Management Systems, Electronic Medical Record, and Artificial Intelligence.PubMed
Candida Barlow
Semin Oncol Nurs. 2020 Apr;36(2):151005. doi: 10.1016/j.soncn.2020.151005. Epub 2020 Apr 4.
OBJECTIVE: To discuss the implications of electronic systems and regulations regarding the use of electronic systems implemented during the conduct of a clinical trial and identify the impact of such platforms on oncology nurses' responsible for providing care to the research participant. DATA SOURCES: Peer-reviewed journal articles, internet, book chapters, and white papers. CONCLUSION: Electronic systems are being increasingly used in the conduct of clinical research. Electronic systems enable the capability to streamline data transfer, remote enrollment capabilities, greater transparency of the trial conduct, improved research documentation, and clearer audit trails. The oncology nurse is at the center of implementation of electronic systems to support the conduct of clinical research and enables safe and effective care to the research participant. IMPLICATIONS FOR NURSING PRACTICE: Oncology nurses are vital to the successful outcome of clinical research studies and are key members of the clinical research team. Electronic systems move beyond traditional data collection in clinical trials with multiple benefits. Such systems may enhance the successful completion and adherence of the clinical trial and maintain the safety of the individual consented to research trial.
98Applicability analysis of immunotherapy for lung cancer patients based on deep learning.PubMed
Wenjing Yan, Xiao Tang, Lidong Wang, et al.
Methods. 2022 Sep;205:149-156. doi: 10.1016/j.ymeth.2022.07.004. Epub 2022 Jul 6.
According to global and Chinese cancer statistics, lung cancer is the second most common cancer globally with the highest mortality rate and a severe threat to human life and health. In recent years, immunotherapy has made significant breakthroughs in the treatment of cancer patients. However, only 30% of patients are applicable and may have immune-related adverse events. The traditional immunological inspection methods have limitations and often can not obtain the expected benefits. Deep learning is a typical representation learning method that can spontaneously mine the hidden feature of effective classification from seas of data. In order to alleviate medical resources and reduce costs, this paper proposes a deep learning-based method to predict patients best suited for immune checkpoint blocking therapy from patients CT images. The deep immunotherapy analysis method proposed in this paper is divided into three steps:(1) Using LUNA16 public dataset to develop a deep learning model for nodule detection. (2) Nodule detection was performed on the Anti-PD-1_Lung dataset, and the effectiveness of immunotherapy was determined by comparing the detection results of nodules before and after immunotherapy. (3) After the data set was processed, the deep learning method trained and analyzed the Lung images. According to the experimental results and comparative analysis, the proposed deep immunotherapy analysis method has a good performance in the detection of nodules. It works for the predictions for the applicability of immunotherapy for lung cancer..
99Multiple Resolution Residually Connected Feature Streams for Automatic Lung Tumor Segmentation From CT Images.PubMed
Jue Jiang, Yu-Chi Hu, Chia-Ju Liu, et al.
IEEE Trans Med Imaging. 2019 Jan;38(1):134-144. doi: 10.1109/TMI.2018.2857800. Epub 2018 Jul 23.
Volumetric lung tumor segmentation and accurate longitudinal tracking of tumor volume changes from computed tomography images are essential for monitoring tumor response to therapy. Hence, we developed two multiple resolution residually connected network (MRRN) formulations called incremental-MRRN and dense-MRRN. Our networks simultaneously combine features across multiple image resolution and feature levels through residual connections to detect and segment the lung tumors. We evaluated our method on a total of 1210 non-small cell (NSCLC) lung tumors and nodules from three data sets consisting of 377 tumors from the open-source Cancer Imaging Archive (TCIA), 304 advanced stage NSCLC treated with anti- PD-1 checkpoint immunotherapy from internal institution MSKCC data set, and 529 lung nodules from the Lung Image Database Consortium (LIDC). The algorithm was trained using 377 tumors from the TCIA data set and validated on the MSKCC and tested on LIDC data sets. The segmentation accuracy compared to expert delineations was evaluated by computing the dice similarity coefficient, Hausdorff distances, sensitivity, and precision metrics. Our best performing incremental-MRRN method produced the highest DSC of 0.74 ± 0.13 for TCIA, 0.75±0.12 for MSKCC, and 0.68±0.23 for the LIDC data sets. There was no significant difference in the estimations of volumetric tumor changes computed using the incremental-MRRN method compared with the expert segmentation. In summary, we have developed a multi-scale CNN approach for volumetrically segmenting lung tumors which enables accurate, automated identification of and serial measurement of tumor volumes in the lung.
100New vision of HookEfficientNet deep neural network: Intelligent histopathological recognition system of non-small cell lung cancer.PubMed
Huijie Yuan, Toshitaka Kido, Masayuki Hirata, et al.
Comput Biol Med. 2024 Aug;178:108710. doi: 10.1016/j.compbiomed.2024.108710. Epub 2024 Jun 4.
BACKGROUND: Efficient and precise diagnosis of non-small cell lung cancer (NSCLC) is quite critical for subsequent targeted therapy and immunotherapy. Since the advent of whole slide images (WSIs), the transition from traditional histopathology to digital pathology has aroused the application of convolutional neural networks (CNNs) in histopathological recognition and diagnosis. HookNet can make full use of macroscopic and microscopic information for pathological diagnosis, but it cannot integrate other excellent CNN structures. The new version of HookEfficientNet is based on a combination of HookNet structure and EfficientNet that performs well in the recognition of general objects. Here, a high-precision artificial intelligence-guided histopathological recognition system was established by HookEfficientNet to provide a basis for the intelligent differential diagnosis of NSCLC. METHODS: A total of 216 WSIs of lung adenocarcinoma (LUAD) and 192 WSIs of lung squamous cell carcinoma (LUSC) were recruited from the First Affiliated Hospital of Zhengzhou University. Deep learning methods based on HookEfficientNet, HookNet and EfficientNet B4-B6 were developed and compared with each other using area under the curve (AUC) and the Youden index. Temperature scaling was used to calibrate the heatmap and highlight the cancer region of interest. Four pathologists of different levels blindly reviewed 108 WSIs of LUAD and LUSC, and the diagnostic results were compared with the various deep learning models. RESULTS: The HookEfficientNet model outperformed HookNet and EfficientNet B4-B6. After temperature scaling, the HookEfficientNet model achieved AUCs of 0.973, 0.980, and 0.989 and Youden index values of 0.863, 0.899, and 0.922 for LUAD, LUSC and normal lung tissue, respectively, in the testing set. The accuracy of the model was better than the average accuracy from experienced pathologists, and the model was superior to pathologists in the diagnosis of LUSC. CONCLUSIONS: HookEfficientNet can effectively recognize LUAD and LUSC with performance superior to that of senior pathologists, especially for LUSC. The model has great potential to facilitate the application of deep learning-assisted histopathological diagnosis for LUAD and LUSC in the future.
101Artificial intelligence in digital pathology - new tools for diagnosis and precision oncology.PubMed
Kaustav Bera, Kurt A Schalper, David L Rimm, et al.
Nat Rev Clin Oncol. 2019 Nov;16(11):703-715. doi: 10.1038/s41571-019-0252-y. Epub 2019 Aug 9.
In the past decade, advances in precision oncology have resulted in an increased demand for predictive assays that enable the selection and stratification of patients for treatment. The enormous divergence of signalling and transcriptional networks mediating the crosstalk between cancer, stromal and immune cells complicates the development of functionally relevant biomarkers based on a single gene or protein. However, the result of these complex processes can be uniquely captured in the morphometric features of stained tissue specimens. The possibility of digitizing whole-slide images of tissue has led to the advent of artificial intelligence (AI) and machine learning tools in digital pathology, which enable mining of subvisual morphometric phenotypes and might, ultimately, improve patient management. In this Perspective, we critically evaluate various AI-based computational approaches for digital pathology, focusing on deep neural networks and 'hand-crafted' feature-based methodologies. We aim to provide a broad framework for incorporating AI and machine learning tools into clinical oncology, with an emphasis on biomarker development. We discuss some of the challenges relating to the use of AI, including the need for well-curated validation datasets, regulatory approval and fair reimbursement strategies. Finally, we present potential future opportunities for precision oncology.
102Artificial Intelligence in Oncology: Current Landscape, Challenges, and Future Directions.PubMed
William Lotter, Michael J Hassett, Nikolaus Schultz, et al.
Cancer Discov. 2024 May 1;14(5):711-726. doi: 10.1158/2159-8290.CD-23-1199.
UNLABELLED: Artificial intelligence (AI) in oncology is advancing beyond algorithm development to integration into clinical practice. This review describes the current state of the field, with a specific focus on clinical integration. AI applications are structured according to cancer type and clinical domain, focusing on the four most common cancers and tasks of detection, diagnosis, and treatment. These applications encompass various data modalities, including imaging, genomics, and medical records. We conclude with a summary of existing challenges, evolving solutions, and potential future directions for the field. SIGNIFICANCE: AI is increasingly being applied to all aspects of oncology, where several applications are maturing beyond research and development to direct clinical integration. This review summarizes the current state of the field through the lens of clinical translation along the clinical care continuum. Emerging areas are also highlighted, along with common challenges, evolving solutions, and potential future directions for the field.
103Novel research and future prospects of artificial intelligence in cancer diagnosis and treatment.PubMed
Chaoyi Zhang, Jin Xu, Rong Tang, et al.
J Hematol Oncol. 2023 Nov 27;16(1):114. doi: 10.1186/s13045-023-01514-5.
Research into the potential benefits of artificial intelligence for comprehending the intricate biology of cancer has grown as a result of the widespread use of deep learning and machine learning in the healthcare sector and the availability of highly specialized cancer datasets. Here, we review new artificial intelligence approaches and how they are being used in oncology. We describe how artificial intelligence might be used in the detection, prognosis, and administration of cancer treatments and introduce the use of the latest large language models such as ChatGPT in oncology clinics. We highlight artificial intelligence applications for omics data types, and we offer perspectives on how the various data types might be combined to create decision-support tools. We also evaluate the present constraints and challenges to applying artificial intelligence in precision oncology. Finally, we discuss how current challenges may be surmounted to make artificial intelligence useful in clinical settings in the future.
104Artificial intelligence for clinical oncology.PubMed
Benjamin H Kann, Ahmed Hosny, Hugo J W L Aerts
Cancer Cell. 2021 Jul 12;39(7):916-927. doi: 10.1016/j.ccell.2021.04.002. Epub 2021 Apr 29.
Clinical oncology is experiencing rapid growth in data that are collected to enhance cancer care. With recent advances in the field of artificial intelligence (AI), there is now a computational basis to integrate and synthesize this growing body of multi-dimensional data, deduce patterns, and predict outcomes to improve shared patient and clinician decision making. While there is high potential, significant challenges remain. In this perspective, we propose a pathway of clinical cancer care touchpoints for narrow-task AI applications and review a selection of applications. We describe the challenges faced in the clinical translation of AI and propose solutions. We also suggest paths forward in weaving AI into individualized patient care, with an emphasis on clinical validity, utility, and usability. By illuminating these issues in the context of current AI applications for clinical oncology, we hope to help advance meaningful investigations that will ultimately translate to real-world clinical use.
105Deep learning in cancer diagnosis, prognosis and treatment selection.PubMed
Khoa A Tran, Olga Kondrashova, Andrew Bradley, et al.
Genome Med. 2021 Sep 27;13(1):152. doi: 10.1186/s13073-021-00968-x.
Deep learning is a subdiscipline of artificial intelligence that uses a machine learning technique called artificial neural networks to extract patterns and make predictions from large data sets. The increasing adoption of deep learning across healthcare domains together with the availability of highly characterised cancer datasets has accelerated research into the utility of deep learning in the analysis of the complex biology of cancer. While early results are promising, this is a rapidly evolving field with new knowledge emerging in both cancer biology and deep learning. In this review, we provide an overview of emerging deep learning techniques and how they are being applied to oncology. We focus on the deep learning applications for omics data types, including genomic, methylation and transcriptomic data, as well as histopathology-based genomic inference, and provide perspectives on how the different data types can be integrated to develop decision support tools. We provide specific examples of how deep learning may be applied in cancer diagnosis, prognosis and treatment management. We also assess the current limitations and challenges for the application of deep learning in precision oncology, including the lack of phenotypically rich data and the need for more explainable deep learning models. Finally, we conclude with a discussion of how current obstacles can be overcome to enable future clinical utilisation of deep learning.
106Artificial intelligence for multimodal data integration in oncology.PubMed
Jana Lipkova, Richard J Chen, Bowen Chen, et al.
Cancer Cell. 2022 Oct 10;40(10):1095-1110. doi: 10.1016/j.ccell.2022.09.012.
In oncology, the patient state is characterized by a whole spectrum of modalities, ranging from radiology, histology, and genomics to electronic health records. Current artificial intelligence (AI) models operate mainly in the realm of a single modality, neglecting the broader clinical context, which inevitably diminishes their potential. Integration of different data modalities provides opportunities to increase robustness and accuracy of diagnostic and prognostic models, bringing AI closer to clinical practice. AI models are also capable of discovering novel patterns within and across modalities suitable for explaining differences in patient outcomes or treatment resistance. The insights gleaned from such models can guide exploration studies and contribute to the discovery of novel biomarkers and therapeutic targets. To support these advances, here we present a synopsis of AI methods and strategies for multimodal data fusion and association discovery. We outline approaches for AI interpretability and directions for AI-driven exploration through multimodal data interconnections. We examine challenges in clinical adoption and discuss emerging solutions.
107Harnessing multimodal data integration to advance precision oncology.PubMed
Kevin M Boehm, Pegah Khosravi, Rami Vanguri, et al.
Nat Rev Cancer. 2022 Feb;22(2):114-126. doi: 10.1038/s41568-021-00408-3. Epub 2021 Oct 18.
Advances in quantitative biomarker development have accelerated new forms of data-driven insights for patients with cancer. However, most approaches are limited to a single mode of data, leaving integrated approaches across modalities relatively underdeveloped. Multimodal integration of advanced molecular diagnostics, radiological and histological imaging, and codified clinical data presents opportunities to advance precision oncology beyond genomics and standard molecular techniques. However, most medical datasets are still too sparse to be useful for the training of modern machine learning techniques, and significant challenges remain before this is remedied. Combined efforts of data engineering, computational methods for analysis of heterogeneous data and instantiation of synergistic data models in biomedical research are required for success. In this Perspective, we offer our opinions on synthesizing complementary modalities of data with emerging multimodal artificial intelligence methods. Advancing along this direction will result in a reimagined class of multimodal biomarkers to propel the field of precision oncology in the coming decade.
108Artificial intelligence for assisting cancer diagnosis and treatment in the era of precision medicine.PubMed
Zi-Hang Chen, Li Lin, Chen-Fei Wu, et al.
Cancer Commun (Lond). 2021 Nov;41(11):1100-1115. doi: 10.1002/cac2.12215. Epub 2021 Oct 6.
Over the past decade, artificial intelligence (AI) has contributed substantially to the resolution of various medical problems, including cancer. Deep learning (DL), a subfield of AI, is characterized by its ability to perform automated feature extraction and has great power in the assimilation and evaluation of large amounts of complicated data. On the basis of a large quantity of medical data and novel computational technologies, AI, especially DL, has been applied in various aspects of oncology research and has the potential to enhance cancer diagnosis and treatment. These applications range from early cancer detection, diagnosis, classification and grading, molecular characterization of tumors, prediction of patient outcomes and treatment responses, personalized treatment, automatic radiotherapy workflows, novel anti-cancer drug discovery, and clinical trials. In this review, we introduced the general principle of AI, summarized major areas of its application for cancer diagnosis and treatment, and discussed its future directions and remaining challenges. As the adoption of AI in clinical use is increasing, we anticipate the arrival of AI-powered cancer care.