【作者】 高洁；

【导师】 王克芳； 顾俊莲；

【作者基本信息】 山东大学 ， 护理学， 2025， 博士

【摘要】 研究背景下尿路症状(Lower urinary tract symptoms,LUTS)已成为威胁全球公共健康的重大问题。流行病学研究显示,女性LUTS的患病率介于12.6%至89.6%之间。LUTS严重影响患者的日常生活,给个人和社会带来沉重的医疗经济负担。女性护士作为职业女性群体,因长时间站立、轮班工作和高强度工作压力等职业因素,可能面临更高的LUTS风险。尽管LUTS的临床表现复杂多样,但目前针对中国女性,尤其是女性护士群体LUTS临床亚型的系统性研究较为有限。深入剖析LUTS的流行特征及其亚型分布,对于制定精准的防控策略和优化医疗资源配置至关重要。此外,LUTS的发生涉及生理、心理及环境等多重因素的复杂交互,但现有研究对这些因素的协同作用机制探索有限,这在很大程度上阻碍了对疾病机制的深入理解。建立精确的LUTS风险预测模型,对于及早筛查易感人群具有重要价值,有助于推动针对性预防策略的开展。近年来,机器学习模型在疾病预测领域展现出优于传统统计方法的潜力,但基于前瞻性队列数据的女性LUTS预测模型仍较匮乏,其在该领域的预测性能和临床实用价值尚需深入探索与验证。同时,系统识别关键预测因子,将为女性护士 LUTS的早期筛查和精准干预提供科学依据。此外,LUTS的发病机制尚未完全阐明,而尿液微生态与代谢组学研究的兴起为揭示其病理生理过程提供了新视角。尿液微生物群作为尿液微环境的重要组成部分,其失调可能通过引发慢性炎症或代谢紊乱参与多种泌尿系统疾病的发生发展。同样,尿液代谢组学分析有助于发现与LUTS相关的生物标志物及代谢通路。然而,基于人群水平的女性LUTS微生物组与代谢组研究较为有限。系统探讨这些特征,不仅有望发掘新型生物标志物以支持精准诊断,还可能揭示潜在治疗靶点,为女性LUTS的综合管理提供新思路。当前针对女性护士 LUTS的流行特征、危险因素及相关病理机制的研究仍较为有限。基于女性护士群体的系统研究,既有助于阐明LUTS的复杂病因,也能为这一高风险职业人群提供科学的预防干预依据,减轻公共卫生负担。同时,由于女性护士在年龄、生理特征及生活方式等方面具有一定的人群代表性,其研究结果亦可为广大女性人群LUTS的防控与机制研究提供重要参考。研究目的本研究旨在通过整合多维度数据以及多组学分析,深入分析LUTS在女性护士中的流行特征,系统探究LUTS的影响因素及其交互作用模式,并构建可靠的风险预测模型。同时,通过代谢组学和微生物组学的联合分析,揭示该疾病的潜在致病机理。具体目标包括:(1)基于大样本数据识别女性护士 LUTS的流行特征,并通过亚型分析阐明症状模式及识别各亚型的驱动因素,系统评估LUTS的风险因素及其交互效应。(2)基于机器学习算法构建女性护士 LUTS风险预测模型,全面评估其预测性能与临床应用价值,并评估各预测因子的重要性。(3)探索女性LUTS相关的尿液代谢物谱特征,鉴定关键代谢物及其相关代谢通路,揭示潜在的代谢机制,并分析环境预测因子与代谢物谱的关联。(4)探索女性LUTS相关的尿液微生物群特征,鉴定特征菌群并进行功能预测,分析微生物组成变化与环境因素的关联,整合代谢组学数据探讨潜在关联机制和生物标志物,为揭示LUTS发病机制提供新证据。研究方法第一部分:女性护士 LUTS流行特征及影响因素研究(1)研究对象调查对象源自研究团队开展的女性护士下尿路健康促进项目(Nurse Urinary Related Health Study,NURS)队列的基线和随访调查数据。该队列采用分层整群抽样的方法,随机选取山东省内20家医院的女性护士进行调查,调查范围涵盖人口统计学特征、生活行为方式、健康状况、下尿路症状等多个维度。利用基线(2020～2021年)及随访(2023～2024年)数据,共纳入8959名同时参与基线和随访调查并且数据完整的受试者进行分析。(2)统计学分析通过潜在类别分析识别女性护士 LUTS临床亚型,并比较其特征、影响因素及转归;采用广义估计方程模型评估LUTS风险因素及其交互作用(基于相乘与相加尺度);利用自回归交叉滞后路径分析探讨关键风险因素与LUTS的因果时序关系。第二部分:基于机器学习算法的女性护士 LUTS预测模型构建与评价(1)研究对象研究对象同第一部分,将基线无LUTS症状的4236名女性护士纳入研究,追踪其在随访期的LUTS患病情况。(2)统计学分析基于前瞻性队列数据,应用logistic回归(logistic Regression,LR)、k近邻法(k-Nearest Neighbor,kNN)、决策树(Decision Tree,DT),随机森林(Random Forest,RF),极限梯度提升(eXtreme Gradient Boosting,XGBoost),光梯度提升机(Light Gradient Boosting Mechine,LightGBM),支持向量机(Support Vector Machine,SVM),神经网络(Neural Network,NN)八种机器学习算法构建女性护士 LUTS预测模型,通过受试者工作曲线下面积(Area under the curve,AUC)、灵敏度、F1值等指标对模型性能进行综合评估,借助沙普利加和解释(SHapley Additive exPlanations,SHAP)方法解析预测因子的相对重要性,并以第一部分得出的风险因素为特征纳入模型进行敏感性分析,验证模型稳健性。第三部分:女性LUTS相关尿液代谢组学机制研究(1)研究对象基于NURS队列数据,募集32名女性护士 LUTS患者和26名健康对照者作为研究对象,采集晨尿(中段尿)样本用于代谢组学检测。(2)统计学分析采用超高效液相色谱-四极杆飞行时间质谱(UPLC-Q-TOF-MS)进行非靶向代谢组学分析,比较两组样本尿液代谢特征的差异,筛选LUTS相关的差异代谢物,应用随机森林算法筛选潜在生物标志物,并通过ROC曲线评估潜在生物标志物的预测效能。结合冗余分析探索关键环境因素(基于第一、二部分筛选特征)与差异代谢物的关联,基于京都基因与基因组百科全书(Kyoto Encyclopedia of Genes and Genomes,KEGG)数据库解析LUTS相关的关键代谢通路。第四部分:女性LUTS相关尿液菌群机制研究及与代谢组学关联研究(1)研究对象同第三部分,采集32名LUTS患者和26名健康对照者晨尿进行16S rDNA检测。(2)统计学分析通过香农指数及辛普森指数评估菌群Alpha多样性,采用主坐标分析法测定Beta多样性特征。采用DESeq2及线性判别分析筛选组间差异菌种,结合KEGG数据库预测菌群功能。通过冗余分析方法阐明微生物群落与环境变量之间的关联性。运用Pearson相关性分析及冗余分析揭示尿液中差异菌群与代谢产物之间的潜在交互机制。基于多组学数据构建LUTS预测模型,采用logistic回归、岭回归、LASSO回归和随机森林算法,通过留一交叉验证策略进行模型训练和性能评估,并筛选重要生物标志物。研究结果1.基于NURS队列8959名女性护士的数据,53%的护士报告下尿路症状,其中尿急患病率最高(27.36%)。LUTS的患病率随年龄增长而增加,且症状表现具有年龄依赖特征。潜在类别分析识别出4种LUTS亚型:多重症状型、尿失禁型、尿急-排尿踌躇型、夜尿型。GEE分析显示,年龄(比值比[Odds ratio,OR]=1.018,95%置信区间[Confidence interval,CI]:1.012～1.025)、婚姻状况(OR=1.561,95%CI:1.389～1.755)、分娩次数(OR=1.124,95%CI:1.008～1.254)、肥胖(OR=1.606,95%CI:1.405～1.835)、延迟排尿(OR=1.603,95%CI:1.487～1.727)、限制饮水(OR=1.542,95%CI:1.437～1.654)、吸烟(OR=1.671,95%CI:1.014～2.756)、慢性病(OR=1.278,95%CI:1.137～1.437)、慢性便秘(OR=1.323,95%CI:1.223～1.432)、睡眠障碍(OR=1.548,95%CI:1.431～1.674)、抑郁(OR=1.198,95%CI:1.051～1.364)、感知压力(OR=1.055,95%CI:1.042～1.069)、每周工作时长(OR=1.074,95%CI:1.002～1.151)以及多种排尿行为(憋尿[OR=1.190,95%CI:1.136～1.247],排尿地点选择[OR=0.899,95%CI:0.867～0.931],无尿意排尿[OR=1.128,95%CI:1.087～1.172],用力排尿[OR=1.275,95%CI:1.230～1.323])以及床护比(床护比 0.4-0.6[OR=0.804,95%CI:0.675～0.958],床护比>0.6[OR=0.767,95%CI:0.642～0.916])均为女性护士 LUTS的显著风险因素。交互作用分析显示,憋尿与限制饮水、肥胖与慢性病、限制饮水与感知压力等多个变量间存在显著的相乘和相加交互效应。交叉滞后路径分析显示,排尿行为、感知压力、睡眠障碍、焦虑、抑郁、延迟排尿、限制饮水、慢性便秘和慢性病与LUTS存在双向因果时序关系,肥胖仅表现为单向因果效应(肥胖→LUTS)。2.基于基线问卷变量,结合传统logistic回归和7种机器学习算法构建的女性护士LUTS风险预测模型在测试集上表现良好。其中XGBoost模型预测效果(AUC值65.5%)与logistic回归模型(AUC值64.0%)接近,优于其他算法。与传统logistic回归相比,各类机器学习算法并未实现预测能力的显著提升。在纳入模型的预测因子中,基线储尿症状得分、失禁症状得分、感知压力、分娩次数、睡眠障碍、限制饮水等贡献显著。3.通过尿液非靶向代谢组学分析,鉴定出18种具有明确化学注释的差异代谢物。其中 27-Deoxy-5b-cyprinol,Tetrahydroaldosterone-3-glucuronide 和 Koeniginequinone B表现出较强的诊断性能,可作为潜在生物标志物。冗余分析显示,体质指数,限制饮水、感知压力和用力排尿显著影响代谢物组成,共计解释约63%的变异。KEGG通路富集分析揭示,差异代谢物主要富集于代谢调控、信号转导、能量消耗及神经系统调节相关通路。4.尿液微生物组学分析表明,LUTS组与Control组在微生物多样性指标上无显著差异,但在菌群分布模式上存在显著差异。门水平上,LUTS组Gemmatimonadetes,Fibrobacteres,Deinococcus_Thermus,Candidatus_Saccharibacteria 的丰度显著降低;属水平上,LUTS组乳酸菌属丰度下降,而加德纳菌属丰度升高,差异菌属还包括ligilactobacillus、pseudoalteromonas、limosilactobacillus、sphingosinicella 等。憋尿行为、限制饮水和感知压力显著影响菌群组成,解释约57.58%的菌属变异度。功能预测显示,LUTS组菌群在ErbB信号通路、轴突引导、TRP通道炎症介质调节等通路功能显著增强,而在光合作用和四环素生物合成通路功能减弱。尿液微生物-代谢物的交互作用分析发现,细菌分类群(Limosilactobacillus,Helcococcus,Butyrivibrio,Anaerotruncus,Schwartzia,Blastomonas,Selenomonas,Ruminococcus 等)与炎症相关的代谢物(2,3-Dinorthromboxane B1、18-carboxy dinor Leukotriene B4)及Yohimbic acid等呈正相关。在冗余分析中,发现 2,3-Dinorthromboxane B1、18-carboxy dinor Leukotriene B4、Yohimbic acid、Koeniginequinone B 以及 1,2,3,4-Tetrahydro-l-phenyl-4-(2-phenylethyl)naphthalene 对菌群变异的解释度较高。多组学联合模型展现出最优的预测性能,基于四种不同算法构建的多组学联合模型AUC值均达到96%,显著优于单一组学模型。Ligilactobacillus,Faecalibacterium,Bradyrhizobium,27-Deoxy-5b-cyprinol,Kineothrix,Tetrahydroaldosterone-3-glucuronide 等在预测 LUTS 方面发挥 了关键作用。研究结论1.女性护士 LUTS患病率处于较高水平,患病率为53%,可分为多重症状型、尿失禁型、尿急-排尿踌躇型和夜尿型四种临床亚型,各症状的流行特征及亚型分布呈现年龄依赖性。LUTS与年龄、肥胖、分娩次数、排尿行为、限制饮水、抑郁、睡眠障碍、感知压力等多维度风险因素显著相关,因素间存在复杂的交互作用,且排尿行为、感知压力等因素与LUTS呈现双向因果关系,反映了其发生发展的动态性和多因素协同特点。这些发现为LUTS流行病学提供新证据,为制定防治策略提供科学依据。2.基于前瞻性队列数据构建了女性护士 LUTS预测模型,筛选出性能最优的推荐模型(XGBoost模型),并开发了便于临床应用的在线网络计算器。模型中贡献较大的预测因子包括基线储尿症状得分、失禁症状得分、感知压力、分娩次数、睡眠障碍、限制饮水等,为实现LUTS的个体化风险评估提供了实用工具。3.非靶向代谢组学筛选出18种差异代谢物,包括炎症反应、激素代谢、胆汁酸代谢、能量代谢的相关产物。多条代谢通路可能与LUTS相关,涉及代谢调控、信号转导、能量代谢、神经系统调控相关通路。此外,BMI、感知压力、限制饮水以及用力排尿等环境因素对尿液代谢物谱具有显著影响,揭示了 LUTS的潜在代谢特征及环境因素的生物学关联,为精准诊断和干预提供了科学依据。4.LUTS患者尿液微生物群在多个分类水平上显著改变,受憋尿、限制饮水、感知压力等环境因素驱动。尿液微生物可能通过调控炎症、神经功能和能量代谢参与LUTS病理进程,并与代谢组在信号通路中存在显著交互。炎症相关代谢物与特定菌群呈正相关,提示宿主-微生物代谢网络可能通过激活炎症信号通路参与LUTS发病。基于多组学数据构建的LUTS预测模型表现出良好的预测性能,识别出多种潜在生物标志物(Ligilactobacillus、Faecalibacterium、Bradyrhizobium、27-Deoxy-5b-cyprinol、Tetrahydroaldosterone-3-glucuronide 等),揭示 了女性 LUTS 精准诊疗的新方向。更多还原

【Abstract】 BackgroundLower urinary tract symptoms(LUTS)represent a significant global public health challenge.Epidemiological studies indicate that the prevalence of LUTS among women ranges from 12.6%to 89.6%.LUTS significantly affect patients’ quality of life and imposes substantial economic burdens on both individuals and society.Female nurses,as an occupational cohort,may face elevated LUTS risk due to occupational factors including prolonged standing,shift work,and high-intensity work pressure.Despite the complex clinical manifestations of LUTS,systematic research on clinical subtypes among Chinese women,particularly female nurses,remains scarce.Comprehensive analysis of LUTS epidemiological characteristics and subtype distribution is crucial for developing targeted prevention and control strategies and optimizing healthcare resource allocation.Furthermore,while LUTS development involves complex interactions among physiological,psychological,and environmental factors,existing research has provided limited exploration of the synergistic mechanisms of these factors,substantially hindering deeper understanding of disease pathogenesis.Constructing accurate LUTS risk prediction models is essential for early identification of high-risk individuals,facilitating implementation of targeted preventive and interventional measures.In recent years,machine learning algorithms have demonstrated potential superiority compared to traditional statistical methods in disease prediction.However,female LUTS prediction models based on prospective cohort data remain scarce,and their predictive performance and clinical applicability in the LUTS field require further validation.Concurrent systematic identification of key predictive factors will provide scientific evidence for early screening and precision intervention of LUTS among female nurses.Additionally,the pathogenesis of LUTS remains incompletely elucidated,while emergent research in urinary microecology and metabolomics offers new perspectives for revealing its pathophysiological processes.The urinary microbiome,as a critical component of the urinary microenvironment,may participate in the development of various urological disorders through mechanisms involving chronic inflammation or metabolic dysregulation.Similarly,urinary metabolomic analysis contributes to the discovery of LUTS-associated biomarkers and metabolic pathways.However,population-level studies on female LUTS microbiome and metabolome remain limited.Systematic exploration of these characteristics may not only uncover novel biomarkers supporting precise diagnosis but also reveal potential therapeutic targets,providing new approaches for the comprehensive management of female LUTS.Current research on epidemiological characteristics,risk factors,and related pathological mechanisms of LUTS among female nurses remains limited.Systematic research based on the female nurse population will both contribute to elucidating the complex etiology of LUTS and provide scientific evidence for prevention and intervention in this high-risk occupational group,thereby reducing public health burden.Moreover,given that female nurses possess certain population representativeness in terms of age,physiological characteristics,and lifestyle,research findings can also provide important reference for LUTS prevention,control,and mechanistic research among the broader female population.ObjectivesThis study aims to systematically investigate the epidemiological characteristics,influencing factors,and underlying mechanisms of LUTS among female nurses,develop risk prediction models,and elucidate the role of urinary metabolomics and microbiome in LUTS pathogenesis.The specific objectives are as follows:1.Identify the epidemiological characteristics of LUTS among female nurses based on large-sample data,clarify symptom patterns through subtype analysis and identify driving factors for each subtype,and systematically evaluate LUTS risk factors and their interactive effects.2.Construct LUTS risk prediction models for female nurses based on machine learning algorithms,comprehensively evaluate their predictive performance and clinical application value,and assess the importance of individual predictive factors.3.Explore urinary metabolite profile characteristics associated with female LUTS,identify key metabolites and their related metabolic pathways,reveal potential metabolic mechanisms,and analyze associations between environmental predictive factors and metabolite profiles.4.Investigate urinary microbiome characteristics associated with female LUTS,identify LUTS-associated characteristic bacterial communities and their potential metabolic pathways,investigate associations between microbiome composition changes and environmental predictive factors,and integrate metabolomics data to explore potential associative mechanisms and biomarkers,thereby providing novel evidence for elucidating LUTS pathogenesis.MethodsPart 1:Epidemiological characteristics and influencing factors of LUTS among female nurses(1)ParticipantsThe research participants were derived from the baseline and follow-up surveys of the NURS cohort.This cohort adopted a stratified cluster sampling method to select female nurses from 20 hospitals in Shandong Province for investigation.The survey content included demographic information,lifestyle,health status,environment factors,lower urinary tract symptoms,etc.Using the baseline(2020-2021)and follow-up(2023-2024)data,a total of 8959 subjects who participated in both the baseline and follow-up surveys and had complete data were included for analysis.(2)Statistical AnalysisLatent class analysis was used to identify the clinical subtypes of LUTS in women,and their characteristics,influencing factors,and outcomes were compared.Generalized estimating equation(GEE)models were applied to assess the risk factors for LUTS and their interactions(based on multiplicative and additive scales).Autoregressive cross-lagged panel model(CLPM)was utilized to explore the causal temporal relationship between key risk factors and LUTS.Part 2:Development and Evaluation of machine learning-based prediction models for LUTS among female nurses(1)ParticipantsParticipants were the same as in Part 1.A total of 4236 female nurses without LUTS symptoms at baseline were included in the study,and their LUTS status during the follow-up period was tracked.(2)Statistical AnalysisBased on the prospective cohort data,eight machine learning algorithms,including logistic regression(LR),k-nearest neighbor(kNN),decision tree(DT),random forest(RF),extreme gradient boosting(XGBoost),light gradient boosting machine(LightGBM),support vector machine(SVM),and neural network(NN),were used to construct a prediction model for LUTS in women.The performance of the model was evaluated using indicators such as the Receiver Operating Characteristic(ROC)curve,area under the curve(AUC)value,sensitivity,and F1 score.The SHapley Additive exPlanations(SHAP)method was used to analyze the importance of the predictive factors.Sensitivity analysis was conducted by incorporating the risk factors identified in Part 1 as features into the model to verify the robustness of the model.3.Part Three:Research on the urinary metabolomics mechanism related to LUTS in female nurses(1)ParticipantsBased on the NURS cohort data,32 female nurses with LUTS and 26 healthy controls were recruited as study subjects.Morning urine samples(midstream urine)were collected for metabolomics analysis.(2)Statistical AnalysisNon-targeted metabolomics analysis was conducted using ultra-performance liquid chromatography-quadrupole time-of-flight mass spectrometry(UPLC-Q-TOF-MS)to compare the differences in urinary metabolic characteristics between the two groups,and screen for LUTS-related differential metabolites.Random forest algorithm was applied to identify key biomarkers,and receiver operating characteristic(ROC)curves were used to evaluate the predictive efficacy of biomarkers.Redundancy analysis was utilized to explore associations between key environmental factors(based on characteristics selected from Parts I and II)and differential metabolites.LUTS-related key metabolic pathways were interpreted based on the Kyoto Encyclopedia of Genes and Genomes(KEGG)database.4.Part Four:Research on the urinary microbiome mechanism related to LUTS in female nurses and its association with metabolomics(1)ParticipantsConsistent with Part III,morning urine samples from 32 LUTS patients and 26 healthy controls were collected for 16S rDNA sequencing.(2)Statistical AnalysisMicrobial alpha diversity was assessed using Shannon and Simpson diversity indices,while beta diversity was evaluated through principal coordinate analysis.DESeq2 and linear discriminant analysis were employed to identify differential bacterial species between groups,with microbial function predicted using the KEGG database.Redundancy analysis was conducted to explore associations between microbiome and environmental factors.Pearson correlation analysis and redundancy analysis were utilized to reveal potential interactive mechanisms between urinary differential bacterial communities and metabolites.LUTS prediction models were constructed based on multi-omics data using logistic regression,ridge regression,LASSO regression,and random forest algorithms.Model training and performance evaluation were conducted through leave-one-out cross-validation strategy,with identification of important biomarkers.Results1.Based on the data of 8959 female nurses from the NURS cohort,53%of the nurses reported have LUTS,with urine urgency having the highest prevalence(27.36%).The prevalence of LUTS increased with age,and the symptom manifestations showed age-related differentiation.Latent class analysis identified four LUTS subtypes:multiple severe cluster,urinary incontinence cluster,urine urgency-hesitancy cluster,and nocturia cluster.GEE analysis revealed that age(odds ratio[OR]=1.018,95%confidence interval[CI]:1.012-1.025),marital status(OR=1.561,95%CI:1.389-1.755),parity(OR=1.124,95%CI:1.008-1.254),obesity(OR=1.606,95%CI:1.405-1.835),delayed voiding(OR=1.603,95%Cl:1.487-1.727),fluid restriction(OR=1.542,95%CI:1.437-1.654),smoking(OR=1.671,95%CI:1.0142.756),chronic diseases(OR=1.278,95%CI:1.137-1.437),chronic constipation(OR=1.323,95%CI:1.223-1.432),sleep disorders(OR=1.548,95%CI:1.431-1.674),depression(OR=1.198,95%CI:1.051-1.364),perceived stress(OR=1.055,95%CI:1.042-1.069),weekly working hours(OR=1.074,95%CI:1.002-1.151),and toileting behaviors(holding urine[OR=1.190,95%CI:1.136～1.247],place preference for voiding[OR=0.899,95%CI:0.867～0.931],premature voiding[OR=1.128,95%CI:1.087～1.172],straining to void[OR=1.275,95%CI:1.230～1.323])and nurse-to-bed ratio(0.4-0.6[OR=0.804,95%CI:0.675～0.958],>0.6[OR=0.767,95%CI:0.642～0.916])were all significant risk factors for LUTS in women.Interaction analysis showed significant multiplicative and additive interaction effects among multiple variables such as holding urine and fluid restriction,obesity and chronic diseases,and fluid restriction and perceived stress.CLPM indicated that toileting behaviors,perceived stress,sleep disorders,anxiety,depression,delayed voiding,fluid restriction,chronic constipation,and chronic diseases had bidirectional causal temporal relationships with LUTS,while obesity only showed a unidirectional causal effect(obesity→ LUTS).2.Based on baseline questionnaire variables,LUTS risk prediction model for female nurses constructed using traditional logistic regression and seven machine learning algorithms performed well on the test dataset.The XGBoost model achieved predictive performance(AUC 65.5%)comparable to the logistic regression model(AUC 64.0%)and superior to other algorithms.However,machine learning algorithms did not significantly enhance predictive performance compared to traditional logistic regression.Among the predictive factors incorporated in the models,baseline storage symptom score,incontinence symptom score,perceived stress,parity,sleep disorder,and fluid restriction demonstrated significant contributions.3.Through non-targeted metabolomics analysis of urine,18 distinct metabolites with clear chemical annotations were identified as potential biomarkers.Among them,27-Deoxy-5bcyprinol(cholesterol metabolite),Tetrahydroaldosterone-3-glucuronide(aldosterone metabolite)and Koeniginequinone B(alkaloid derivative)showed strong predictive properties.Analysis of environmental factors showed that body mass index(BMI),fluid restriction,perceived stress,and straining to void significantly affected metabolite composition,collectively explaining about 63%of the variation.KEGG pathway enrichment analysis revealed that differential metabolites were mainly concentrated in pathways related to metabolic regulation,signal transduction,energy consumption and nervous system regulation.4.Urinary microbiome analysis revealed no significant differences in microbial diversity indices between the LUTS and control groups,but significant differences were observed in bacterial community distribution patterns.At the phylum level,the LUTS group showed significantly reduced abundance of Gemmatimonadetes,Fibrobacteres,Deinococcus_Thermus,and Candidatus_Saccharibacteria.At the genus level,the LUTS group exhibited decreased abundance of Lactobacillus and increased abundance of Gardnerella,with additional differential genera including Ligilactobacillus,Pseudoalteromonas,Limosilactobacillus,and Sphingosinicella.Delayed voiding,fluid restriction,and perceived stress significantly influenced bacterial community composition,explaining approximately 57.58%of genus-level variation.Functional prediction demonstrated that the LUTS group exhibited significantly enhanced bacterial community functions in ErbB signaling pathway,axon guidance,and TRP channel inflammatory mediator regulation,while functions in photosynthesis and tetracycline biosynthesis pathways were diminished.Urinary microbiome-metabolite interaction analysis revealed positive correlations between bacterial taxa and inflammation-related metabolites(2,3-Dinorthromboxane B1,18carboxy dinor Leukotriene B4)as well as Yohimbic acid.In redundancy analysis,2,3Dinorthromboxane B1,18-carboxy dinor Leukotriene B4,Yohimbic acid,Koeniginequinone B,and 1,2,3,4-Tetrahydro-l-phenyl-4-(2-phenylethyl)naphthalene demonstrated high explanatory power for bacterial community variation.Multi-omics integrated models exhibited optimal predictive performance,with AUC values reaching 96%across four different algorithms,significantly outperforming single-omics models.Ligilactobacillus,Faecalibacterium,Bradyrhizobium,27-Deoxy-5b-cyprinol,Kineothrix,and Tetrahydroaldosterone-3-glucuronide played crucial roles in LUTS prediction.Conclusions1.53%of female nurses suffered from LUTS,which could be divided into four clinical clusters:multiple severe cluster,urinary incontinence cluster,urine urgency-hesitancy cluster,and nocturia cluster.The epidemic characteristics and cluster distribution were age-dependent.LUTS is significantly correlated with multiple risk factors such as age,obesity,parity,toileting behavior,fluid restriction,depression,sleep disorders,perceived stress,etc.There is a complex interaction among these factors,and there is a two-way causal relationship between toileting behaviors,perceived stress,sleep disorders,anxiety,depression,delayed voiding,fluid restriction,chronic constipation,chronic diseases and LUTS,reflecting the dynamic and complex development of LUTS.Subtype identification and risk factor analysis provide new evidence for the epidemiology of LUTS,contribute to in-depth understanding of the etiological mechanism of female nurses’ LUTS,and provide scientific reference for the prevention and treatment strategy of LUTS.2.Based on the prospective cohort data,we constructed a predictive model for LUTS,identified the optimal performing model(XGBoost model)as the recommended approach,and developed an online web-based calculator for clinical application.Baseline storage symptom score,incontinence symptom score,perceived stress,parity,sleep disorders,and fluid restriction are the core predictors of LUTS,providing a practical tool for individualized risk assessment of LUTS.3.Non-targeted metabolomics identified 18 differential metabolites,including related products of inflammatory response,hormone metabolism,bile acid metabolism,and energy metabolism,which can be used as potential biomarkers of LUTS.Differences in metabolite composition are driven by environmental factors such as BMI,fluid restriction,and perceived stress.Multiple metabolic pathways may be associated with LUTS,involving metabolic regulation,signal transduction,energy expenditure,nervous system regulation and other pathways,providing new insights into therapeutic strategies for LUTS in women.4.The urinary microbiome in LUTS patients exhibited significant alterations across multiple taxonomic levels,driven by environmental factors including delayed voiding,fluid restriction,and perceived stress.Urinary microorganisms may participate in LUTS pathological processes through regulation of inflammation,neurological function,and energy metabolism,demonstrating significant interactions with the metabolome in signaling pathways.Inflammation-related metabolites showed positive correlations with specific bacterial taxa,suggesting that the host-microbiome metabolic network may contribute to LUTS pathogenesis through activation of inflammatory signaling pathways.The LUTS prediction model constructed based on multi-omics data demonstrated robust predictive performance and identified multiple potential biomarkers(Ligilactobacillus,Faecalibacterium,Bradyrhizobium,27-Deoxy-5b-cyprinol,and Tetrahydroaldosterone-3-glucuronide,etc.),revealing novel directions for precision diagnosis and treatment of female LUTS.更多还原