基于可解释机器学习的预测模型:评估CDAI 0~1级克罗恩病患者的腹部手术风险及生物制剂疗效

谢铠岭 ,  孙琪 ,  蒋智贤 ,  姚恒昌 ,  杨理超 ,  张亚炜 ,  刘浩 ,  姚宝佳 ,  吴强 ,  张丹 ,  袁联文

中南大学学报(医学版) ›› 2025, Vol. 50 ›› Issue (10) : 1703 -1715.

PDF (1493KB)
中南大学学报(医学版) ›› 2025, Vol. 50 ›› Issue (10) : 1703 -1715. DOI: 10.11817/j.issn.1672-7347.2025.250348
人工智能在临床医学中的应用专题

基于可解释机器学习的预测模型:评估CDAI 0~1级克罗恩病患者的腹部手术风险及生物制剂疗效

作者信息 +

Interpretable machine learning-based predictive model for assessing abdominal surgery risk and biologic therapy efficacy in CDAI 0 to 1 level Crohn disease patients

Author information +
文章历史 +
PDF (1527K)

摘要

目的 克罗恩病患者存在高手术风险的临床问题。本研究旨在开发一种能够预测克罗恩病活动指数(Crohn disease activity index,CDAI)为0~1级的克罗恩病患者腹部手术风险的预测模型。 方法 回顾性纳入2016至2022年期间在中南大学湘雅二医院就诊的克罗恩病患者。通过随机种子法,以5꞉5的比例将总样本分为训练集和验证集。采用多因素Cox向后逐步回归法筛选最终纳入预测模型的变量,并估算与手术风险相关因素的风险比(hazard ratio,HR)及其95%置信区间(confidence interval,CI)。此外,本研究引入8种机器学习预测模型:最小绝对收缩和选择算子(least absolute shrinkage and selection operator,LASSO)-Cox回归、随机生存森林(random survival forest,RSF)、梯度提升机(gradient boosting machine,GBM)、CoxBoost、生存支持向量机(survival support vector machine,SurvivalSVM)、极端梯度提升(extreme gradient boosting,XGBoost)、监督主成分回归(supervised principal component regression,SuperPC)和部分最小二乘回归Cox(partial least squares Cox regression,PLSR-Cox)。选择在验证集上C指数表现最优的模型作为最终决策的预测模型。采用C指数和受试者操作特征(receiver operating characteristic,ROC)曲线的曲线下面积(area under the curve,AUC)评估预测模型的区分度,校正曲线评估预测模型的校准度,并通过决策曲线分析(decision curve analysis,DCA)评估预测模型的临床净收益。采用沙普利加性解释(Shapley additive explanations,SHAP)分析解释机器学习预测模型的特征对预测结果的贡献,从而提高预测模型的可解释性。采用Kaplan-Meier曲线描述累积无手术率,Log-Rank检验比较预测模型风险分层或有无生物制剂治疗队列的组间差异。 结果 本研究共纳入615名研究对象,训练集307例,验证集308例。通过Cox多因素逐步向后回归分析,筛选出4个重要变量:C反应蛋白(HR=1.07,95% CI 1.01~1.14,P=0.025)、白蛋白(HR=0.69,95% CI 0.46~1.04,P=0.075)、纤维蛋白原(HR=0.65,95% CI 0.51~0.84,P<0.001)和蒙特利尔B分型(HR=2.26,95% CI 1.23~4.17,P=0.009)。通过比较9个预测模型在验证集上的C指数,结果显示:CoxBoost预测模型在验证集中性能最佳,C指数为0.746(95% CI 0.683~0.809)。在训练集中,CoxBoost预测模型的1年、3年和5年AUC分别为0.778、0.749和0.772;在验证集上分别为0.761、0.797和0.751。校正曲线表明:CoxBoost预测模型预测手术概率与实际手术概率一致性较好,Brier评分均小于0.25。DCA分析结果显示:CoxBoost预测模型在一定的预测概率范围内能提供较好的临床净收益。SHAP分析结果显示:蒙特利尔B分型是模型预测结果的主要影响因素。Log-Rank检验结果显示:CoxBoost预测模型划分的低、中、高风险组的累积无手术率存在显著性差异(P<0.001)。在中高风险组中,使用生物制剂显著降低了手术率(HR=0.54,95% CI 0.33~0.89,P=0.014),而在低风险组中则未见显著效果(HR=1.01,95% CI 0.51~2.00,P=0.970)。 结论 本研究开发的CoxBoost预测模型能够有效预测克罗恩病患者的腹部手术风险,并为患者是否采用生物制剂治疗的临床决策提供依据。

Abstract

Objective Crohn disease (CD) patients face a clinically significant high risk of abdominal surgery. This study aims to develop a predictive model for estimating abdominal surgery in CD patients with Crohn Disease Activity Index (CDAI) 0-1. Methods CD patients treated at the Second Xiangya Hospital of Central South University between 2016 and 2022 were retrospectively enrolled. Using a fixed random seed, the full cohort was randomly split into a training set and a validation set at a 5:5 ratio. Final predictors were selected using multivariable backward stepwise Cox regression, and hazard ratios (HRs) with 95% confidence intervals (CIs) were estimated for surgery-associated factors. In addition, 8 machine learning survival prediction models were introduced, including least absolute shrinkage and selection operator (LASSO)-Cox regression, random survival forest (RSF), gradient boosting machine (GBM), CoxBoost, survival support vector machine (SurvivalSVM), extreme gradient boosting (XGBoost), supervised principal component regression (SuperPC), and partial least squares Cox regression (PLSR-Cox). The model with the best concordance index (C-index) performance in the validation set was chosen as the final decision-support model. Model discrimination was evaluated using the C-index and time-dependent ROC curves with corresponding area under the curve (AUCs), while calibration was assessed using calibration curves. Clinical net benefit was quantified using decision curve analysis (DCA). Shapley additive explanations (SHAP) was applied to interpret the contribution of model features to prediction, improving model explainability. Kaplan-Meier curves were used to describe cumulative surgery-free probability, and the Log-Rank test was used to compare differences across predicted risk strata and between biologic-exposed and non-exposed cohorts. Results A total of 615 patients were included in the study, comprising 307 patients in the training set and 308 patients in the validation set. Multivariable backward stepwise Cox regression identified 4 key variables significantly associated with abdominal surgery risk in CD patients with CDAI 0-1, including C-reactive protein (CRP, HR=1.07, 95% CI 1.01 to 1.14, P=0.025), albumin (ALB, HR=0.69, 95% CI 0.46 to 1.04, P=0.075), fibrinogen (Fg, HR=0.65, 95% CI 0.51 to 0.84, P<0.001), and Montreal B behavior classification (HR=2.26, 95% CI 1.23 to 4.17, P=0.009). Among 9 candidate predictive models, CoxBoost achieved the best performance in the validation set, yielding a C-index of 0.746 (95% CI 0.683 to 0.809). Time-dependent ROC analysis demonstrated that in the training set, the 1-, 3-, and 5-year AUCs were 0.778, 0.749, and 0.772, respectively, while in the validation set, the corresponding AUCs were 0.761, 0.797, and 0.751. Calibration curves showed good agreement between predicted and observed surgery risk, with Brier scores <0.25 at all evaluated time points. DCA showed that CoxBoost provided clinical net benefit within a meaningful probability range. SHAP analysis result indicates that Montreal B subtype is the primary factor influencing the modeled predictive outcomes. Risk stratification generated by the CoxBoost model revealed that the cumulative surgery-free probability differed significantly among low-, moderate-, and high-risk groups (Log-Rank P<0.001). In the moderate-to-high-risk group, exposure to biologic therapy was associated with a significant reduction in abdominal surgery risk (HR=0.54, 95% CI 0.33 to 0.89, P=0.014). In contrast, no significant effect of biologic therapy was observed in the low-risk group (HR=1.01, 95% CI 0.51 to 2.00, P=0.970). Conclusion The CoxBoost model developed in this study effectively predicts abdominal surgery risk in CD patients with CDAI 0-1, and supports clinical decision-making regarding biologic therapy, providing evidence for personalized treatment planning.

Graphical abstract

关键词

炎症性肠病 / 克罗恩病 / 腹部手术风险 / 生物制剂 / 人工智能

Key words

inflammatory bowel disease / Crohn disease / abdominal surgery risk / biological agent / artificial intelligence

引用本文

引用格式 ▾
谢铠岭,孙琪,蒋智贤,姚恒昌,杨理超,张亚炜,刘浩,姚宝佳,吴强,张丹,袁联文. 基于可解释机器学习的预测模型:评估CDAI 0~1级克罗恩病患者的腹部手术风险及生物制剂疗效[J]. 中南大学学报(医学版), 2025, 50(10): 1703-1715 DOI:10.11817/j.issn.1672-7347.2025.250348

登录浏览全文

4963

注册一个新账户 忘记密码

克罗恩病(Crohn disease,CD)是一种慢性胃肠道炎症性疾病[1-2],全球发病率呈现显著的地域差异并逐年上升[3],对患者生活和医疗系统带来重大负担[4-6]。其病因复杂,涉及遗传、免疫异常和环境因素共同作用,导致免疫失调与肠道屏障功能紊乱[7-8]。CD的典型病理特征为不连续、斑片状的肠道全层性炎症,主要发生在末端回肠和近端结肠[9],临床表现为腹泻、腹痛、便血等[10]。手术干预通常标志着CD进展至复杂阶段[11]。尽管生物制剂治疗减少了手术率[12-13],但超过70%的CD患者仍需接受手术[14-17]。因此,如何评估手术适应证、预测手术风险及治疗反应,成为管理CD的核心挑战。
克罗恩病活动指数(Crohn disease activity index,CDAI)是最常用的疾病活动度评估工具[18]。对于CDAI为0~1级的患者(症状缓解或轻度活动期),尽管其短期手术风险较低,20%~30%的患者仍会在5~10年内因并发症进展而需要手术[19]。目前的争议在于,尽管症状缓解期患者通常采用氨基水杨酸、激素和免疫抑制剂的阶梯治疗,但这一策略可能导致治疗延误。对于轻度活动期患者,生物制剂治疗是否足以延缓结构性损伤、减少手术需求,仍缺乏充分证据。现有的评分系统如Mayo风险评分和Harvey-Bradshaw指数,虽能部分预测手术风险[20-22],但对于CDAI为0~1级的患者的预测尚不精准。传统统计方法难以有效解析复杂的变量交互作用,且无法捕捉多维度风险因素的关联。近年来,机器学习算法如随机森林和梯度提升树等,能够更好地构建预测模型[23-25]。因此,本研究旨在结合传统临床变量和机器学习技术,构建CDAI 0~1级CD患者腹部手术风险及生物制剂疗效的预测模型,以优化个体化治疗方案,改善患者预后。

1 对象与方法

1.1 伦理声明

本研究已经获得中南大学湘雅二医院(以下简称“我院”)伦理委员会批准(审批号:LYEC2025-0156),所有程序均遵循赫尔辛基宣言,研究过程中对患者相关信息进行匿名化处理。

1.2 对象

回顾性纳入2016年至2022年就诊于我院的CD患者。纳入标准:1)根据中国炎症性肠病诊断与治疗的共识意见[26-27]诊断为CD的患者;2)初始诊断在我院;3)CDAI为0~1级的患者。排除标准:1)在我院诊断或治疗前已进行腹部肠道切除手术(阑尾切除除外);2)主要分析变量中整体缺失比例超过20%的患者;3)随访资料缺失;4)因其他疾病或自然原因死亡。

研究终点为首次接受腹部手术。手术指征包括复发性肠出血、急性穿孔、慢性肠瘘、复发性肠梗阻、恶性肿瘤、药物治疗无效、诊断不明确及儿童生长衰竭。

1.3 方法

通过医院电子病历系统收集患者基线资料,包括性别,吸烟史,体重指数(body mass index,BMI),白细胞计数(white blood cell count,WBC),C反应蛋白(C-reactive protein,CRP),血红蛋白(hemoglobin,Hb),红细胞沉降率(erythrocyte sedimentation rate,ESR),白蛋白(albumin,ALB),凝血酶原时间(prothrombin time,PT),活化部分凝血活酶时间(activated partial thromboplastin time,APTT),纤维蛋白原(fibrinogen,Fg),D-二聚体(D-dimer,DD),蒙特利尔分型,生物制剂治疗史,肠外表现,诊断时的CDAI分级。通过电话随访和住院电子病历确定患者是否接受手术治疗。随访间隔为3个月,患者随访时间定义为首次诊断至接受手术的时间或最后1次随访的时间。参考既往的临床研究[28-31],根据患者是否使用生物制剂[包括英夫利西单抗(infliximab,IFX)、阿达木单抗(adalimumab,ADA)、维得利珠单抗(vedolizumab,VDZ)、乌司奴单抗(ustekinumab,UST)、乌帕替尼(upadacitinib,ABT-494)]超过3个月,定义其是否使用生物制剂。

1.4 统计学处理

采用mice包对临床缺失数据进行多重插补,使用预测均值匹配方法生成5个插补数据集。插补模型纳入的变量包括:基本资料(吸烟史、BMI)、实验室指标(WBC、CRP等)、疾病相关指标[肛周脓肿、蒙特利尔分型(A、B、L型各亚型)]、临床表现[肠外表现及其具体类型(口腔溃疡、肝胆病变、皮肤病变、眼部病变、骨关节炎)],以及治疗相关指标(生物制剂使用情况)。纳入插补模型有助于提高缺失数据推断的合理性和稳健性。插补前后变量分布基本一致(附图1https://doi.org/10.57760/sciencedb.34353),插补完成后,对分类变量采用众数汇总,连续变量采用均值汇总,并合并5个插补数据集以获得最终分析数据集,用于后续统计分析和建模。连续变量采用Shapiro-Wilk检验评估正态性。符合正态分布的数据以均值±标准差表示,采用独立样本t检验进行组间比较;不符合正态分布的数据以中位数(第1四分位数,第3四分位数)表示,采用Mann-Whitney U检验进行组间比较。分类变量以频数和百分比表示,根据样本量和期望频数分别采用皮尔逊卡方检验、连续性校正卡方检验或Fisher精确检验进行组间比较。

通过随机种子法按5꞉5的比例将总体样本划分为训练集和验证集。采用Cox比例风险回归模型估算腹部手术风险相关因素的风险比(hazard ratio,HR)及其95%置信区间(confidence interval,CI)。在多因素Cox回归模型中,纳入单因素Cox回归模型中P值小于0.1的变量,并基于向后逐步回归法筛选最终纳入预测模型的变量。比例风险假设采用Schoenfeld残差检验验证。采用限制性立方样条(restricted cubic spline,RCS)评估连续变量与手术风险之间的剂量-反应关系,在多因素Cox回归模型中设置4个结点(第5、35、65和95百分位),并通过Wald检验将仅包含线性项的模型与同时包含线性和立方项的模型进行比较,以评估非线性效应。

此外,引入了8种机器学习生存预测模型,包括最小绝对收缩和选择算子(least absolute shrinkage and selection operator,LASSO)-Cox回归、随机生存森林(random survival forest,RSF)、梯度提升机(gradient boosting machine,GBM)、CoxBoost、生存支持向量机(survival support vector machine,SurvivalSVM)、极端梯度提升(extreme gradient boosting,XGBoost)、监督主成分回归(supervised principal component regression,SuperPC)和部分最小二乘Cox回归(partial least squares Cox regression,PLSR-Cox)[32-33]。通过比较验证集上的C指数,选择性能最优模型作为最终预测模型。在建模前,对RSF、GBM、SurvivalSVM和XGBoost模型的数据进行标准化处理,以减少极端值的影响。

预测模型区分度主要采用C指数评估,并通过训练集内5折交叉验证结合网格搜索进行参数优化。采用受试者操作特征(receiver operating characteristic,ROC)曲线的曲线下面积(area under the curve,AUC)评估预测模型在不同时间点的区分能力,采用校准曲线和决策曲线分析(decision curve analysis,DCA)分别评估预测模型的校准度和临床净收益。不同预测模型间C指数的比较基于cschange函数完成。采用沙普利加性解释(Shapley additive explanations,SHAP)方法解释模型预测结果,通过量化并可视化各特征对预测结果的贡献度来提升模型的可解释性,SHAP值的绝对值越大,表示该特征对模型风险预测的贡献越大。预测风险值的分层通过X-tile软件完成。采用Kaplan-Meier(KM)生存曲线比较不同风险分层或有无使用生物制剂治疗患者的生存差异,组间比较采用Log-Rank检验。为便于临床应用,开发了基于网页的预测工具。统计分析与数据可视化分别采用R 4.2.0、Python 3.8和GraphPad Prism 10.1.2软件完成,分析代码已提供(https://doi.org/10.57760/sciencedb. 34354)。P<0.05为差异具有统计学意义。

2 结 果

2.1 基线资料

本研究流程可见于图1,共纳入615名患者,其中男性456例(74.1%),女性159例(25.9%)。在这些患者中,114例(18.5%)有吸烟史,208例(33.8%)存在肛周脓肿。患者的蒙特利尔分型分布如下:A型中1型77例(12.5%),2型467例(75.9%),3型71例(11.6%);B型中1型312例(50.7%),2型179例(29.1%),3型124例(20.2%);L型中1型162例(26.3%),2型37例(6.0%),3型413例(67.2%),4型3例(0.5%)。81例(13.2%)患者存在肠外表现,具体为:42例(6.8%)存在口腔溃疡,9例(1.5%)存在肝脏和胆囊病变,16例(2.6%)存在皮肤病变,7例(1.1%)存在眼部病变,15例(2.4%)存在骨关节炎。379例(61.6%)患者使用生物制剂治疗。将患者分为训练集(307例)和验证集(308例),训练集和验证集之间除蒙特利尔B型分布比例外,其余基线资料均无显著差异(均P>0.05,表1)。

2.2 单因素和多因素Cox回归

单因素Cox回归分析结果见附表1(https://doi. org/10.57760/sciencedb.xbyxb.00152)。最终,多因素Cox回归预测模型(表2)纳入以下变量:CRP(HR=1.07,95% CI 1.01~1.14,P=0.025),ALB(HR=0.69,95% CI 0.46~1.04,P=0.075),Fg(HR=0.65,95% CI 0.51~0.84,P<0.001),蒙特利尔B分型(HR=2.26,95% CI 1.23~4.17,P=0.009)。Schoenfeld残差检验表明:多因素Cox回归模型中各个变量和整体模型均满足等比例风险假设(附图2A,https://doi.org/10.57760/sciencedb. xbyxb.00152)。基于多因素Cox回归的RCS分析提示:CRP(P for nonlinear=0.110)、ALB(P for nonlinear=0.573)及Fg(P for nonlinear=0.903)与手术风险呈线性剂量反应关系(附图2B,https://doi.org/10.57760/sciencedb. xbyxb.00152)。

2.3 模型性能评估

9个模型的C指数比较显示:虽然GBM(0.858,95% CI 0.816~0.900)、XGBoost(0.874,95% CI 0.834~0.913)和RSF(0.770,95% CI 0.708~0.832)在训练集上均表现出较高的预测性能,但其在验证集上的表现明显下降:GBM为0.734(95% CI 0.657~0.811)、XGBoost为0.728(95% CI 0.651~0.804)、RSF为0.731(95% CI 0.666~0.797),提示上述模型可能存在一定程度的过拟合。相比之下,CoxBoost预测模型虽然在训练集上的表现并非最优(0.745,95% CI 0.678~0.813),但在验证集上获得了最高的C指数(0.746,95% CI 0.683~0.809),且显著优于传统Cox预测模型(P=0.046,表3),显示出更优的泛化能力。时间依赖性AUC结果与C指数一致:尽管GBM、XGBoost和CoxBoost在训练集中均表现出较高的判别能力,且GBM、XGBoost和RSF预测模型在1年、3年和5年的AUC均高于CoxBoost预测模型(GBM:0.901、0.890、0.894;XGBoost:0.904、0.900、0.924;RSF:0.797、0.782、0.820;CoxBoost:0.778、0.749、0.772),但在验证集中,GBM、XGBoost和RSF预测模型的AUC均出现明显下降。相比之下,CoxBoost预测模型在验证集中于1年、3年和5年均保持了相对稳定且更优的AUC表现(0.761、0.797和0.751),优于GBM(0.733、0.792、0.646)、XGBoost(0.713、0.780、0.663)和RSF(0.754、0.788、0.650),进一步支持CoxBoost预测模型具有更强的泛化能力(附图3,https://doi.org/10.57760/sciencedb. 34345)。此外,通过校准曲线的Brier评分对模型校准度进行评估,结果显示CoxBoost预测模型在训练集中排名前3(1年、3年和5年的Brier评分分别为0.076、0.091和0.107),但在验证集中表现最佳,具有最低的Brier评分(1年、3年和5年分别为0.060、0.092和0.142;附图4,https://doi.org/10.57760/sciencedb. 34346),提示其在预测概率校准方面同样具有优势。在DCA曲线分析中,CoxBoost在1年、3年和5年的随访节点中均展现出较优的临床净收益(20%~80%;附图5,https://doi.org/10.57760/sciencedb.34349)。因此,CoxBoost预测模型被选为最终应用模型,本研究基于此开发了一个在线预测工具(https://surgica.shinyapps.io/cdai0-1/)。

SHAP分析结果显示:蒙特利尔B分型对模型预测结果的贡献最大,其平均SHAP值(>0.2)显著高于其他特征。Fg和CRP对模型预测亦具有较大影响,其平均SHAP值介于0.10~0.15之间。相比之下,ALB的影响相对较小,平均SHAP值低于0.1(附图6A~6B,https://doi.org/10.57760/sciencedb.34350)。单一样本的SHAP瀑布图分析进一步显示:蒙特利尔B分型对该样本的预测结果产生了0.22的正向贡献,而Fg表现为-0.04的负向贡献,CRP和ALB的贡献较小,合计为-0.01(附图6C,https://doi.org/10.57760/sciencedb. 34350)。此外,SHAP依赖图表明:蒙特利尔B分型和CRP与SHAP值呈正相关,而Fg和ALB与模型预测输出呈负相关(附图6D~6G,https://doi.org/10. 57760/sciencedb.34350)。

2.4 CoxBoost预测模型的风险分层及其对生物治疗疗效的影响

为了优化临床决策,对CoxBoost预测模型的预测风险值进行了风险分层(低风险<0.149,中风险层≥0.149且≤0.389,高风险层>0.389)。无论是在训练集还是验证集中,3组间累积无手术率差异均有统计学意义(均P<0.001,图2A、2B)。进一步评估不同风险分层中使用生物制剂对手术率的影响。在低风险组中,使用生物制剂未能显著降低手术率(HR=1.01,95% CI 0.51~2.00,P=0.970,图2C);而在中高风险组中,使用生物制剂显著降低了手术率(HR=0.54,95% CI 0.33~0.89,P=0.014,图2D)。

3 讨 论

本研究开发并验证了基于CoxBoost机器学习模型的CD腹部手术风险预测工具,发现CRP、Fg、ALB和蒙特利尔B分型是关键手术风险因素。与传统Cox模型相比,CoxBoost模型在预测效能、校准性和临床净收益上具有显著优势[34-35]。该预测模型通过风险分层,成功区分了低、中、高风险组,并发现生物制剂治疗在不同风险层级中的效果差异,为临床个性化治疗提供了依据。

蒙特利尔B分型在CD患者腹部手术风险预测中至关重要,特别是B2型和B3型患者相比B1型,手术风险更高[36]。B2型患者的肠道狭窄通常由慢性炎症引起,易导致肠梗阻和腹痛,严重时可能需要手术解除梗阻或切除部分肠道[37-38]。B3型患者则有穿透性疾病,常伴有瘘管、脓肿等并发症,部分病例可能会出现肠穿孔,导致急性腹痛和腹膜炎,需要紧急手术[37-38]。本研究发现,B2型和B3型患者的手术风险明显高于B1型,进一步证明了蒙特利尔B分型在预测手术风险中的重要性,强调了其在CD患者手术风险评估中的价值。

本研究发现CRP是CD患者腹部手术风险的危险因素。CRP作为急性期反应蛋白,其升高通常反映肠道的慢性炎症[39-40],这种炎症可能导致肠道损伤和并发症,增加手术需求[41]。高CRP水平与病情活动度密切相关,提示患者面临更高的手术风险。将CRP纳入模型验证了其作为生物标志物的有效性,帮助医生评估疾病活动度并为早期干预提供依据,降低手术发生率。

Fg是由肝合成的血浆蛋白,主要通过转化为纤维蛋白参与凝血过程,其升高通常被认为反映炎症活动增强[42-45]。本研究中,较高的Fg水平与较低的腹部手术风险相关,但这一结果更可能体现疾病表型之间的流行病学相关性,而非明确的因果性保护作用。在CD背景下,Fg和/或纤维蛋白参与止血与组织修复过程,凝血酶介导的纤维蛋白形成有助于屏障修复和细胞迁移,其降解产物亦可调控免疫反应[44, 46-50]。然而,Fg升高可能反映炎症负担、组织损伤,或与营养状况、感染及药物使用等因素相关,存在残余混杂和反向因果的可能。因此,该关联可能具有阈值效应,现有结果尚不足以支持Fg对手术风险具有因果性降低作用,仍需在独立队列和机制研究中进一步验证。

ALB是体内主要的血浆蛋白,参与维持血浆胶体渗透压及物质转运[51],其水平常反映患者的营养状况和肝功能,降低时多与营养不良及疾病进展相关[52-53]。本研究发现:较高的ALB水平是腹部手术的保护因素,提示患者具有较好的营养状态和较强的恢复能力,从而降低手术干预需求。

本研究仍存在一定局限性。首先,作为回顾性分析,难以完全避免选择偏倚;且随着诊疗水平、生物制剂及手术策略的持续演进,治疗决策缺乏统一标准,可能影响模型预测的稳健性。其次,受数据条件所限,未能开展系统的定量敏感性分析,结果的泛化性需谨慎解读。再次,研究未区分急诊与择期手术,可能增加研究终点的异质性,进而影响模型在不同临床情境下的适用性。接着,单中心研究设计导致模型尚缺乏外部验证,其推广能力有待进一步评估。随后,开发的在线工具仅基于回顾性数据,仍需前瞻性研究和真实世界验证。最后,部分潜在重要变量(如生活质量、药物依从性、免疫状态及并发症等)未能纳入分析,未来需在多中心前瞻性队列中完善模型并验证其临床实用性。

参考文献

[1]

Torres J, Mehandru S, Colombel JF, et al. Crohn’s disease[J]. Lancet. 2017, 389(10080): 1741-1755.

[2]

Baumgart DC, Sandborn WJ. Crohn’s disease[J]. Lancet, 2012, 380(9853): 1590-605.

[3]

Hracs L, Windsor J, Gorospe J, et al. Global evolution of inflammatory bowel disease across epidemiologic stages[J]. Nature, 2025, 642(8067): 458-466.

[4]

Kaplan G. The global burden of IBD: from 2015 to 2025[J]. Nat Rev Gastroenterol Hepatol, 2015, 12(12): 720-727.

[5]

Solitano V, Bernstein CN, Dotan I, et al. Shaping the future of inflammatory bowel disease: a global research agenda for better management and public health response[J]. Nat Rev Gastroenterol Hepatol, 2025, 22(6): 438-452.

[6]

赵宝宁, 杨丽艳, 齐建丽, . 克罗恩病患者体力活动强度与免疫炎症状态及生活质量的相关性[J]. 临床与病理杂志, 2022, 42(12): 3017-3025.

[7]

ZHAO Baoning, YANG Liyan, QI Jianli, et al. Correlation of physical activity intensity with immune inflammatory status and quality of life in patients with Crohn’s disease[J]. Journal of Clinical and Pathological Research, 2022, 42(12): 3017-3025.

[8]

Dolinger M, Torres J, Vermeire S. Crohn’s disease[J]. Lancet, 2024, 403(10432): 1177-1191.

[9]

Nagayama M, Gogokhia L, Longman RS. Precision microbiota therapy for IBD: premise and promise[J]. Gut Microbes, 2025, 17(1): 2489067.

[10]

Roda G, Chien Ng S, Kotze PG, et al. Crohn’s disease[J]. Nat Rev Dis Primers, 2020, 6: 22.

[11]

Veauthier B, Hornecker JR. Crohn’s disease: diagnosis and management[J]. Am Fam Physician, 2018, 98(11): 661-669.

[12]

Bernstein CN, Loftus E, Ng SC, et al. Hospitalisations and surgery in Crohn’s disease[J]. Gut, 2012, 61(4): 622-629.

[13]

Hisamatsu T, Miyoshi J, Oguri N, et al. Inflammation-associated carcinogenesis in inflammatory bowel disease: clinical features and molecular mechanisms[J]. Cells, 2025, 14(8): 567.

[14]

姚嘉茵, 彭翔, 刘涛, . 维得利珠单抗治疗炎症性肠病的短期疗效与安全性:单中心真实世界研究[J]. 临床与病理杂志, 2022, 42(8): 1841-1846.

[15]

YAO Jiayin, PENG Xiang, LIU Tao, et al. Short-term efficacy and safety of vedolizumab in the treatment of inflammatory bowel disease: A single-center real-world study[J]. Journal of Clinical and Pathological Research, 2022, 42(8): 1841-1846.

[16]

Toh JW, Stewart P, Rickard MJ, et al. Indications and surgical options for small bowel, large bowel and perianal Crohn’s disease[J]. World J Gastroenterol, 2016, 22(40): 8892-8904.

[17]

Patel KV, Darakhshan AA, Griffin N, et al. Patient optimization for surgery relating to Crohn’s disease[J]. Nat Rev Gastroenterol Hepatol, 2016, 13(12): 707-719.

[18]

Samnani S, Ray C M, Gill P, et al. Diagnostic accuracy of noninvasive biomarkers and imaging for evaluating postoperative recurrence in Crohn’s disease[J/OL]. Clin Gastroenterol Hepatol, 2025, 23(13): 2433-2447.e9[2025-06-20].

[19]

Alós R, Hinojosa J. Timing of surgery in Crohn’s disease: a key issue in the management[J]. World J Gastroenterol, 2008, 14(36): 5532-5539.

[20]

Lee JS, Kim HJ, Cho HM, et al. The importance of the Crohn’s disease activity index in surgery for small bowel Crohn’s disease[J]. J Visc Surg, 2016, 153(5): 339-345.

[21]

Meima-van Praag EM, Buskens CJ, Hompes R, et al. Surgical management of Crohn’s disease: a state of the art review[J]. Int J Colorectal Dis, 2021, 36(6): 1133-1145.

[22]

Chirra P, Sharma A, Bera K, et al. Integrating radiomics with clinicoradiological scoring can predict high-risk patients who need surgery in Crohn’s disease: a pilot study[J]. Inflamm Bowel Dis, 2023, 29(3): 349-358.

[23]

Dong Y, Xu L, Fan Y, et al. A novel surgical predictive model for Chinese Crohn’s disease patients[J/OL]. Medicine (Baltimore), 2019, 98(46): e17510[2025-06-20].

[24]

Liu RQ, Liu SL, Yi L, et al. Development and validation of multiparametric models based on computed tomography enterography to determine endoscopic activity and surgical risk in patients with Crohn’s disease: a multi-center study[J/OL]. Heliyon, 2023, 9(9): e19942[2025-06-20].

[25]

Lang BM, Biedermann L, van Haaften W, et al. Genetic polymorphisms associated with smoking behaviour predict the risk of surgery in patients with Crohn’s disease[J]. Aliment Pharmacol Ther, 2018, 47(1): 55-66.

[26]

Greener JG, Kandathil SM, Moffat L, et al. A guide to machine learning for biologists[J]. Nat Rev Mol Cell Biol, 2022, 23(1): 40-55.

[27]

周靖, 丁寿鹏, 蔡子涵. 基于机器学习优化的血液指标肺结核诊断模型:多中心研究[J]. 临床与病理杂志, 2025, 45(1): 25-37.

[28]

ZHOU Jing, DING Shoupeng, CAI Zihan. Blood-based diagnostic model for pulmonary tuberculosis optimized by machine learning: A multicenter study[J]. Journal of Clinical and Pathological Research, 2025, 45(1): 25-37.

[29]

Inflammatory Bowel Disease Group, Chinese Society of Gastroenterology, Chinese Medical Association. Chinese consensus on diagnosis and treatment in inflammatory bowel disease (2018, Beijing)[J]. J Dig Dis, 2021, 22(6): 298-317.

[30]

Gomollón F, Dignass A, Annese V, et al. 3rd European evidence-based consensus on the diagnosis and management of Crohn’s disease 2016: part 1: diagnosis and medical management[J]. J Crohns Colitis, 2017, 11(1): 3-25.

[31]

Sandborn W, Hanauer S, Rutgeerts P, et al. Adalimumab for maintenance treatment of Crohn’s disease: results of the CLASSIC II trial[J]. Gut, 2007, 56(9): 1232-1239.

[32]

Loftus E, Panés J, Lacerda A, et al. Upadacitinib induction and maintenance therapy for Crohn’s disease[J]. N Engl J Med, 2023, 388(21): 1966-1980.

[33]

Lichtenstein GR, Loftus EV, Isaacs KL, et al. ACG clinical guideline: management of Crohn’s disease in adults[J]. Am J Gastroenterol, 2018, 113(4): 481-517.

[34]

Gros B, Blackwell J, Segal J, et al. Harms with placebo in trials of biological therapies and small molecules as maintenance therapy in inflammatory bowel disease: a systematic review and meta-analysis[J]. Lancet Gastroenterol Hepatol, 2024, 9(11): 1030-1040.

[35]

周雄辉, 桂嵘, 刘竞, 高萌. 结合多组学和机器学习构建多发性骨髓瘤疗效预测模型[J]. 中南大学学报(医学版), 2025, 50(4): 531-544.

[36]

ZHOU Xionghui, GUI Rong, LIU Jing, GAO Meng. Construction of a treatment response prediction model for multiple myeloma based on multi-omics and machine learning[J]. Journal of Central South University. Medical Science, 2025, 50(4): 531-544.

[37]

段洪伟, 刘怀政, 孙传政, 漆靖. 列线图和机器学习预测脓毒症合并深静脉血栓患者的院内死亡发生风险[J]. 中南大学学报(医学版), 2025, 50(6): 1013-1029.

[38]

DUAN Hongwei, LIU Huaizheng, SUN Chuanzheng, QI Jing. Nomogram and machine learning models for predicting in-hospital mortality in sepsis patients with deep vein thrombosis[J]. Journal of Central South University. Medical Science, 2025, 50(6): 1013-1029.

[39]

Xiang Y, Yang F, Yuan F, et al. Development and validation of a multimodal machine learning model for diagnosing and assessing risk of Crohn’s disease in patients with perianal fistula[J]. Aliment Pharmacol Ther, 2025, 61(5): 824-834.

[40]

蒋泽楠, 宋珑, 梁春水, 张昊, 刘立明. 基于机器学习的慢性瓣膜病合并心房颤动患者行Cox迷宫IV手术后心房颤动复发风险预测模型[J]. 中南大学学报(医学版), 2023, 48(7): 995-1007.

[41]

JIANG Zenan, SONG Long, LIANG Chunshui, ZHANG Hao, LIU Liming. Prediction model of atrial fibrillation recurrence after Cox-Maze IV procedure in patients with chronic valvular disease and atrial fibrillation based on machine learning algorithm[J]. Journal of Central South University. Medical Science, 2023, 48(7): 995-1007.

[42]

Yang LC, Yao BJ, Jiang ZX, et al. Epidemiological investigation of different types of Crohn’s disease patients undergoing first abdominal surgery in the era of biological agents: a population-based cohort study[J]. Int J Colorectal Dis, 2024, 39(1): 150.

[43]

Silverberg MS, Satsangi J, Ahmad T, et al. Toward an integrated clinical, molecular and serological classification of inflammatory bowel disease: report of a Working Party of the 2005 Montreal World Congress of Gastroenterology[J/OL]. Can J Gastroenterol, 2005, 19(): 5A-36A[2025-06-20].

[44]

Satsangi J, Silverberg MS, Vermeire S, et al. The Montreal classification of inflammatory bowel disease: controversies, consensus, and implications[J]. Gut, 2006, 55(6): 749-753.

[45]

Zhou H, Tang Y, Xu T, et al. C-reactive protein: structure, function, regulation, and role in clinical diseases[J]. Front Immunol, 2024, 15: 1425168.

[46]

Yanai H, Levine A, Hirsch A, et al. The Crohn’s disease exclusion diet for induction and maintenance of remission in adults with mild-to-moderate Crohn’s disease (CDED-AD): an open-label, pilot, randomised trial[J]. Lancet Gastroenterol Hepatol, 2022, 7(1): 49-59.

[47]

Mosli MH, Zou GY, Garg SK, et al. C-reactive protein, fecal calprotectin, and stool lactoferrin for detection of endoscopic activity in symptomatic inflammatory bowel disease patients: a systematic review and meta-analysis[J]. Am J Gastroenterol, 2015, 110(6): 802-819.

[48]

Fu L, Liu Z, Liu Y. Fibrinogen-like protein 2 in inflammatory diseases: a future therapeutic target[J]. Int Immunopharmacol, 2023, 116: 109799.

[49]

Lin HH, Chen RP, Jiang XP, et al. Elevated fibrinogen-like protein 2 in TNBS-induced colitis mice: Association with Th17 and regulatory T cells[J]. Mol Med Rep, 2017, 16(3): 3445-3454.

[50]

Hoppe B. Fibrinogen and factor XIII at the intersection of coagulation, fibrinolysis and inflammation[J]. Thromb Haemost, 2014, 112(4): 649-658.

[51]

Chen XF, Zhao Y, Guo Y, et al. Predictive value of fibrinogen in identifying inflammatory bowel disease in active stage[J]. BMC Gastroenterol, 2021, 21(1): 472.

[52]

Kattula S, Byrnes JR, Wolberg AS. Fibrinogen and fibrin in hemostasis and thrombosis[J/OL]. Arterioscler Thromb Vasc Biol, 2017, 37(3): e13-e21[2025-06-20].

[53]

Kearney KJ, RASAriëns, MacRae FL. The role of fibrin(ogen) in wound healing and infection control[J]. Semin Thromb Hemost, 2022, 48(2): 174-187.

[54]

Seltana A, Cloutier G, Reyes Nicolas V, et al. Fibrin(ogen) is constitutively expressed by differentiated intestinal epithelial cells and mediates wound healing[J]. Front Immunol, 2022, 13: 916187.

[55]

Bayer IS. Advances in fibrin-based materials in wound repair: a review[J]. Molecules, 2022, 27(14): 4504.

[56]

Owczarek D, Cibor D, Sałapa K, et al. Reduced plasma fibrin clot permeability and susceptibility to lysis in patients with inflammatory bowel disease: a novel prothrombotic mechanism[J]. Inflamm Bowel Dis, 2013, 19(12): 2616-2624.

[57]

Molitoris BA, Sandoval RM, Yadav SPS, et al. Albumin uptake and processing by the proximal tubule: physiological, pathological, and therapeutic implications[J]. Physiol Rev, 2022, 102(4): 1625-1667.

[58]

Garcia-Martinez R, Caraceni P, Bernardi M, et al. Albumin: pathophysiologic basis of its role in the treatment of cirrhosis and its complications[J]. Hepatology, 2013, 58(5): 1836-1846.

[59]

Caraceni P, O’Brien A, Gines P. Long-term albumin treatment in patients with cirrhosis and ascites[J]. J Hepatol, 2022, 76(6): 1306-1317.

基金资助

国家自然科学基金(82270590)

湖南省自然科学基金(2025JJ50584)

湖南省标准化项目(湘市监标函〔2024〕24号)

中南大学中央高校基本科研业务费专项(2024ZZTS0959┫。This work was supported by the National Natural Science Foundation┣82270590)

the Natural Science Foundation of Hunan Province(2025JJ50584)

the Standardization Project of Hunan Province (Xiangshijian Biaohan〔2024〕No. 24), and the Fundamental Research Funds for the Central Universities of Central South University(2024ZZTS0959)

RIGHTS & PERMISSIONS

开放获取(Open access):本文遵循知识共享许可协议,允许第三方用户按照署名-非商业性使用-禁止演绎4.0(CC BY-NC-ND 4.0)的方式,在任何媒介以任何形式复制、传播本作品(https://creativecommons.org/licenses/by-nc-nd/4.0/)。

AI Summary AI Mindmap
PDF (1493KB)

333

访问

0

被引

详细

导航
相关文章

AI思维导图

/