ChiCTR2600116526 版本V1.0 版本创建时间2026/01/12 11:49:11 中国临床试验注册中心

审核状态:

Project audit state:

通过审核

Successful

注册号:

Registration number:

ChiCTR2600116526 

最近更新日期:

Date of Last Refreshed on:

2026-01-12 11:49:06 

注册时间:

Date of Registration:

2026-01-12 00:00:00 

注册号状态:

预注册

Registration Status:

Prospective registration

注册题目:

基于机器学习的“双引擎+强制安全护栏”胰岛素泵初始剂量决策支持系统在住院2型糖尿病患者中的多中心混合影子验证与反事实模拟研究

Public title:

A Multicenter Hybrid Shadow Validation and Counterfactual Simulation Study of a Dual-Engine AI Decision Support System with Mandatory Safety Guardrails for Acute-Phase Insulin Pump Therapy in T2D Inpatients

注册题目简写:

English Acronym:

研究课题的正式科学名称:

基于机器学习的“双引擎+强制安全护栏”胰岛素泵初始剂量决策支持系统在住院2型糖尿病患者中的多中心混合影子验证与反事实模拟研究

Scientific title:

A Multicenter Hybrid Shadow Validation and Counterfactual Simulation Study of a Dual-Engine AI Decision Support System with Mandatory Safety Guardrails for Acute-Phase Insulin Pump Therapy in T2D Inpatients

研究课题代号(代码):

Study subject ID:

在二级注册机构或其它机构的注册号:

The registration number of the Partner Registry or other register:

申请注册联系人:

荣曦 

研究负责人:

荣曦 

Applicant:

Xi Rong 

Study leader:

Xi Rong 

申请注册联系人电话:

Applicant telephone:

+86 152 8969 0808

研究负责人电话:

Study leader's telephone:

+86 152 8969 0808

申请注册联系人传真 :

Applicant Fax:

研究负责人传真:

Study leader's fax:

申请注册联系人电子邮件:

Applicant E-mail:

tonyellison@163.com

研究负责人电子邮件:

Study leader's E-mail:

tonyellison@163.com

申请单位网址(自愿提供):

Applicant website(voluntary supply):

研究负责人网址(自愿提供):

Study leader's website(voluntary supply):

申请注册联系人通讯地址:

中国广西壮族自治区南宁市青秀区双拥路6号

研究负责人通讯地址:

中国广西壮族自治区南宁市青秀区双拥路6号

Applicant address:

No. 6 Shuangyong Road, Qingxiu District, Nanning, Guangxi Zhuang Autonomous Region, China

Study leader's address:

No. 6 Shuangyong Road, Qingxiu District, Nanning, Guangxi Zhuang Autonomous Region, China

申请注册联系人邮政编码:

Applicant postcode:

530021

研究负责人邮政编码:

Study leader's postcode:

530021

申请人所在单位:

广西医科大学第一附属医院

Applicant's institution:

The First Affiliated Hospital of Guangxi Medical University

研究负责人所在单位:

广西医科大学第一附属医院

Affiliation of the Leader:

The First Affiliated Hospital of Guangxi Medical University

是否获伦理委员会批准:

是/Yes

Approved by ethic committee:

Yes

伦理委员会批件文号:

Approved No. of ethic committee:

2025-K0604

伦理委员会批件附件:

Approved file of Ethical Committee:

查看附件View

批准本研究的伦理委员会名称:

广西医科大学第一附属医院医学伦理委员会

Name of the ethic committee:

Medical Ethics Committee of the First Affiliated Hospital of Guangxi Medical University

伦理委员会批准日期:

Date of approved by ethic committee:

2025-12-15 00:00:00

伦理委员会联系人:

陈丽

Contact Name of the ethic committee:

Li Chen

伦理委员会联系地址:

中国广西壮族自治区南宁市青秀区双拥路6号

Contact Address of the ethic committee:

No. 6 Shuangyong Road, Qingxiu District, Nanning, Guangxi Zhuang Autonomous Region, China

伦理委员会联系人电话:

Contact phone of the ethic committee:

+86 771 530 9187

伦理委员会联系人邮箱:

Contact email of the ethic committee:

研究实施负责(组长)单位:

广西医科大学第一附属医院

Primary sponsor:

The First Affiliated Hospital of Guangxi Medical University

研究实施负责(组长)单位地址:

中国广西壮族自治区南宁市青秀区双拥路6号

Primary sponsor's address:

No. 6 Shuangyong Road, Qingxiu District, Nanning, Guangxi Zhuang Autonomous Region, China

试验主办单位(项目批准或申办者):

Secondary sponsor:

国家:

中国

省(直辖市):

广西壮族自治区

市(区县):

南宁

Country:

China

Province:

Guangxi Zhuang Autonomous Region

City:

Nanning

单位(医院):

广西医科大学第一附属医院

具体地址:

中国广西壮族自治区南宁市青秀区双拥路6号

Institution
hospital:

The First Affiliated Hospital of Guangxi Medical University

Address:

No. 6 Shuangyong Road, Qingxiu District, Nanning, Guangxi Zhuang Autonomous Region, China

经费或物资来源:

广西医科大学第一附属医院临床研究攀登计划青年科技启明星项目(YYZS2023010)

Source(s) of funding:

Clinical Research"Climbing" Program of the First Affiliated Hospital of Guangxi Medical University (YYZS2023010)

Target disease:

Type 2 Diabetes Mellitus

Target disease code:

E11

研究类型:

观察性研究

Study type:

Observational study

研究所处阶段:

其它 

Study phase:

N/A

研究设计:

队列研究 

Study design:

Cohort study 

研究目的:

主要目的:验证AI趋势引擎在真实临床路径下对血糖预测的准确性(数字孪生效度);评估剂量安全约束与熔断系统(安全护栏)对真实低血糖事件的预警灵敏度。 次要目的:通过反事实模拟,评估基础 AI 策略(双引擎+护栏回滚)相对于医生处方的安全性获益(非劣效性);评估ISDO硅基优化器在低危人群中提升血糖达标率(TIR)的潜在价值;分析医生剂量与AI推荐剂量的偏离度及其与临床结局的关联。  

Objectives of Study:

Primary Objectives: To verify the predictive validity ("Digital Twin" fidelity) of the AI Trend Engine in real-world clinical pathways; and to evaluate the sensitivity of the Dose Safety Constraint & Circuit Breaker System (Safety Guardrails) in detecting true hypoglycemia events. Secondary Objectives: To estimate the safety benefit (non-inferiority) of the baseline AI strategy compared to physician prescriptions via counterfactual simulation; to evaluate the potential clinical value of the In Silico Dose Optimizer (ISDO) in improving Time in Range (TIR) for low-risk patients; and to analyze the discordance between physician doses and AI recommendations and its association with clinical outcomes.

药物成份或治疗方案详述:

本研究为前瞻性、多中心、观察性影子验证研究,无干预措施。 1、常规治疗:所有受试者均接受常规住院胰岛素泵(CSII)强化治疗。临床医生完全依据现行指南及个人经验,独立制定初始胰岛素剂量(含24小时基础率及餐前大剂量),并根据患者血糖监测结果进行常规调整。所用药物为已上市的速效胰岛素类似物(如门冬胰岛素、赖脯胰岛素等)。 2、AI 系统运行:研究对象“双引擎+强制安全护栏”AI决策支持系统部署于医院信息系统后台,处于“静默运行”状态。系统根据患者基线特征计算推荐剂量并预测血糖趋势,但结果不向临床医生展示,不用于指导治疗,仅用于事后数据比对分析。 

Description for medicine or protocol of treatment in detail:

This is a prospective, multicenter, observational shadow validation study with no intervention. 1. Standard of Care: All participants receive standard inpatient Continuous Subcutaneous Insulin Infusion (CSII) therapy. Physicians independently determine the initial insulin doses (basal rates and meal boluses) based on current clinical guidelines and experience, and adjust them according to routine glucose monitoring. The medications used are marketed rapid-acting insulin analogs (e.g., Insulin Aspart, Insulin Lispro). 2. AI System Operation: The "Dual-Engine + Mandatory Safety Guardrails" AI decision support system operates in "Silent Mode" in the background. It calculates recommended doses and predicts glucose trends based on patient baseline characteristics, but the results are blinded to physicians and not used for treatment decisions, serving only for retrospective data analysis. 

纳入标准:

(1)年龄:≥18 岁; (2)诊断:诊断明确为2型糖尿病,符合中华医学会糖尿病学分会发布的《中国糖尿病防治指南》(2024版),糖尿病的诊断标准; (3)治疗时机:住院期间由医生决定首次启用胰岛素泵治疗; (4)饮食状态:能够正常进食糖尿病餐(接受标准医学营养治疗),且规律进食一日三餐者; (5)观察窗口:预计CSII 持续治疗时间≥24 小时,且预期能够完成至少一次“晚餐后–睡前–次日清晨空腹”的完整血糖监测周期; (6)数据完整性:入泵前基线资料完整,包含模型所需的全部输入变量(人口学特征、人体测量指标、病史、实验室检查结果等); (7)知情同意:受试者本人或其法定代理人自愿签署书面知情同意书。

Inclusion criteria

(1) Age: >=18 years old; (2) Diagnosis: the patient was diagnosed with type 2 diabetes mellitus in accordance with the diagnostic criteria of "Guidelines for the Prevention and Treatment of Diabetes in China" (2024 edition) issued by the Diabetes Society of the Chinese Medical Association. (3) Timing of treatment: doctors decided to start insulin pump therapy for the first time during hospitalization. (4) Dietary status: those who were able to eat diabetic meals normally (receiving standard medical nutrition treatment) and regularly ate three meals a day; (5) Observation window: the duration of CSII treatment was expected to be >=24 hours, and it was expected to complete at least one complete blood glucose monitoring cycle of "after dinner, before bed, and fasting the next morning". (6) Data integrity: baseline data were complete, including all input variables needed for the model (demographic characteristics, anthropometric indicators, medical history, laboratory test results, etc.). (7) Informed consent: subjects or their legal representatives voluntarily signed written informed consent.

排除标准:

1. 特殊疾病与生理状态: (1)糖尿病类型为1型糖尿病、妊娠期糖尿病或其他特殊类型糖尿病; (2)急性应激状态:入院时处于糖尿病酮症酸中毒(DKA)、高血糖高渗状态(HHS)、伴有全身炎症反应综合征(SIRS)的严重感染(如合并脓毒症、坏死性筋膜炎等评估达Wagner 3-5级的糖尿病足)、外伤或大手术围手术期,需静脉胰岛素治疗且尚未转为皮下CSII稳定期者; (3)特殊进食模式:无法经口进食,需依赖鼻饲(胃管)、肠外营养(静脉营养)或需要禁食状态的患者; (4)严重肝肾功能衰竭:处于透析阶段的肾脏病患者或诊断为肝衰竭患者。 (5)导入期干预:对于上午入院且安装胰岛素泵治疗的患者,若在 T0(晚餐前)之前已经发生了低血糖(<3.9 mmol/l)导致医生被迫大幅修改了原定的基础率方案,则该患者不再适合作为“初始剂量预测”的验证对象,应予剔除。 2. 干扰性药物与治疗史: (1)糖皮质激素使用:入组前4周内或当前正在接受全身性糖皮质激素治疗者(如泼尼松、地塞米松等); (2)长效制剂残留:入组前4周内曾注射过超长效周制剂胰岛素(如伊柯胰岛素)者。注:使用 GLP-1RA 周制剂者不排除。 (3)非初始胰岛素泵治疗状态:入院前正在持续使用自带胰岛素泵的患者。 3. 依从性与伦理: (1)无法配合常规指尖血糖监测,或因皮肤/精神问题不能稳定佩戴胰岛素泵者; (2)主治医生认为其存在严重精神障碍、明显认知功能障碍,或其他可能影响研究依从性与安全性的情况。

Exclusion criteria:

1. Special diseases and physiological states: (1) Type of diabetes: type 1 diabetes mellitus, gestational diabetes mellitus or other special type of diabetes mellitus; (2) Acute stress state: Patients were admitted to the hospital in diabetic ketoacidosis (DKA), hyperglycemic hyperosmolar state (HHS), severe infection with systemic inflammatory response syndrome (SIRS) (e.g., diabetic foot with Wagner grades 3-5, such as sepsis and necrotizing fasciitis), trauma, or perioperative period of major surgery. Patients who need intravenous insulin therapy and have not transitioned to stable subcutaneous CSII; (3) special feeding mode: patients who cannot eat by mouth, rely on nasal feeding (gastric tube), parenteral nutrition (intravenous nutrition), or need fasting state; (4) Severe liver and kidney failure: patients with kidney disease in the dialysis stage or patients diagnosed with liver failure. (5) Run-in intervention: patients who were admitted to the hospital in the morning and received insulin pump therapy should be excluded if hypoglycemia (<3.9 mmol/l) had occurred before T0 (before dinner), which forced the doctor to significantly modify the original basal rate plan. 2. Interfering medications and treatment history: (1) Glucocorticoid use: patients receiving systemic glucocorticoids (such as prednisone, dexamethasone, etc.) within 4 weeks before enrollment or currently; (2) long-acting insulin residue: patients who had been injected with ultra-long-acting insulin (e.g., insulin icoides) within 4 weeks before enrollment. Note: Use of GLP-1RA weekly preparations was not excluded. (3) Non-initial insulin pump therapy: patients with continuous use of their own insulin pump before admission. 3. Compliance and Ethics: (1) Unable to cooperate with routine fingertip blood glucose monitoring, or unable to wear insulin pump stably due to skin or mental problems; (2) patients with severe mental disorders, obvious cognitive impairment, or other conditions that may affect the compliance and safety of the study according to the attending doctors.

研究实施时间:

Study execute time:

From 2026-02-01 00:00:00 To 2026-12-31 00:00:00  

征募观察对象时间:

Recruiting time:

From 2026-02-01 00:00:00 To 2026-12-31 00:00:00  

干预措施:

Interventions:

组别:

观察组(2型糖尿病胰岛素泵治疗队列)

样本量:

520

Group:

Observation group (type 2 diabetes mellitus insulin pump therapy cohort)

Sample size:

干预措施:

干预措施代码:

N/A

Intervention:

None

Intervention code:

研究实施地点:

Countries of recruitment and research settings:

国家:

中国

省(直辖市):

广西壮族自治区 

市(区县):

南宁 

Country:

China 

Province:

Guangxi Zhuang Autonomous Region 

City:

Nanning 

单位(医院):

广西医科大学第一附属医院 

单位级别:

三甲 

Institution
hospital:

The First Affiliated Hospital of Guangxi Medical University

Level of the institution:

Tertiary A

国家:

中国

省(直辖市):

广西壮族自治区 

市(区县):

南宁 

Country:

China 

Province:

Guangxi Zhuang Autonomous Region 

City:

Nanning 

单位(医院):

广西医科大学第二附属医院 

单位级别:

三甲 

Institution
hospital:

The Second Affiliated Hospital of Guangxi Medical University

Level of the institution:

Tertiary A

国家:

中国

省(直辖市):

广西壮族自治区 

市(区县):

柳州 

Country:

China 

Province:

Guangxi Zhuang Autonomous Region 

City:

Liuzhou 

单位(医院):

柳州市人民医院 

单位级别:

三甲 

Institution
hospital:

Liuzhou People's Hospital

Level of the institution:

Tertiary A

国家:

中国

省(直辖市):

广西壮族自治区 

市(区县):

梧州 

Country:

China 

Province:

Guangxi Zhuang Autonomous Region 

City:

Wuzhou 

单位(医院):

广西壮族自治区桂东人民医院 

单位级别:

三甲 

Institution
hospital:

Guidong People's Hospital of Guangxi Zhuang Autonomous Region

Level of the institution:

Tertiary A

国家:

中国

省(直辖市):

广西壮族自治区 

市(区县):

柳州 

Country:

China 

Province:

Guangxi Zhuang Autonomous Region 

City:

Liuzhou 

单位(医院):

柳州市柳江区人民医院 

单位级别:

二甲 

Institution
hospital:

Liujiang District People's Hospital of Liuzhou

Level of the institution:

Secondary A

国家:

中国

省(直辖市):

广东 

市(区县):

清远 

Country:

China 

Province:

Guangdong 

City:

Qingyuan 

单位(医院):

佛冈县人民医院 

单位级别:

二甲 

Institution
hospital:

People's Hospital of Fogang County

Level of the institution:

Secondary A

国家:

中国

省(直辖市):

广西壮族自治区 

市(区县):

南宁 

Country:

China 

Province:

Guangxi Zhuang Autonomous Region 

City:

Nanning 

单位(医院):

马山县加方乡卫生院 

单位级别:

一级 

Institution
hospital:

Mashan County Jiafang Township Health Center

Level of the institution:

Primary

国家:

中国

省(直辖市):

广西壮族自治区 

市(区县):

南宁 

Country:

China 

Province:

Guangxi Zhuang Autonomous Region 

City:

Nanning 

单位(医院):

马山县古寨瑶族乡卫生院 

单位级别:

一级 

Institution
hospital:

Mashan County Guzhai Yao Township Health Center

Level of the institution:

Primary

测量指标:

Outcomes:

指标中文名:

血糖趋势引擎的预测准确性

指标类型:

主要指标

Outcome:

Predictive Accuracy of Glucose Trend Engine

Type:

Primary indicator

测量时间点:

CSII 启动后 24 小时内 (T0-T3)

测量方法:

计算预测值与实测值之间的平均绝对误差 (MAE)、均方根误差 (RMSE) 及 95% 预测区间覆盖率 (CP)

Measure time point of outcome:

Within 24 hours of CSII initiation (T0-T3)

Measure method:

Calculate MAE, RMSE, and 95% Coverage Probability (CP) between predicted and observed glucose values.

指标中文名:

安全护栏对低血糖事件的预警灵敏度

指标类型:

主要指标

Outcome:

Sensitivity of Safety Guardrails for Hypoglycemia Detection

Type:

Primary indicator

测量时间点:

CSII 启动后 24 小时内 (T0-T3)

测量方法:

以真实低血糖事件(血糖<3.9 mmol/L或防御性干预)为金标准,计算护栏预警(预测下限<3.9 mmol/L)的灵敏度与提前预警时间

Measure time point of outcome:

Within 24 hours of CSII initiation (T0-T3)

Measure method:

Calculate sensitivity and lead time of guardrail alerts (predicted lower bound <3.9 mmol/L) against true hypoglycemia events (BG <3.9 mmol/L or rescue intervention)

指标中文名:

模拟低血糖风险降低率 (反事实推演)

指标类型:

次要指标

Outcome:

Simulated Reduction in Hypoglycemia Risk (Counterfactual)

Type:

Secondary indicator

测量时间点:

CSII 启动后 24 小时内

测量方法:

在全分析集(FAS)中,通过反事实模拟比较 AI 策略(C1及C2)与医生真实路径(A)的预测低血糖事件发生率。重点进行预设亚组分析:验证 C1 策略在高危人群(H1)中的风险降低优势

Measure time point of outcome:

Within 24 hours of CSII initiation

Measure method:

Compare the incidence of predicted hypoglycemia between AI strategies (C1 & C2) and physician's real path (A) in the Full Analysis Set (FAS) via counterfactual simulation. Key pre-specified subgroup analysis will focus on verifying the risk reduction of the C1 strategy in the high-risk (H1) population.

指标中文名:

模拟血糖达标率 / TIR (反事实推演)

指标类型:

次要指标

Outcome:

Simulated Time in Range / TIR (Counterfactual)

Type:

Secondary indicator

测量时间点:

CSII 启动后 24 小时内

测量方法:

在全分析集(FAS)中,通过反事实模拟比较 AI 策略(C1及C2)与医生真实路径(A)的空腹血糖达标率及目标范围内时间比率 (TIR)。重点进行预设亚组分析:验证 ISDO 策略(C2)在低危人群(H0)中的疗效提升优势

Measure time point of outcome:

Within 24 hours of CSII initiation

Measure method:

Compare the fasting glucose target achievement rate and Time in Range (TIR) between AI strategies (C1 & C2) and physician's real path (A) in the Full Analysis Set (FAS). Key pre-specified subgroup analysis will focus on verifying the efficacy improvement of the ISDO strategy (C2) in the low-risk (H0) population

采集人体标本:

Collecting sample(s)
from participants:

标本中文名:

组织:

Sample Name:

None

Tissue:

人体标本去向

其它  

说明

本研究仅收集常规诊疗产生的临床数据(如血糖数值),不进行任何额外的生物样本采集、处理或冻存

Fate of sample:

0thers  

Note:

This study only collects clinical data generated from routine care (e.g., blood glucose values) and does not involve any additional collection, processing, or biobanking of biological samples.

征募研究对象情况:

Recruiting status:

尚未开始

Not yet recruiting

年龄范围:

Participant age:

最小 Min age 18 years
最大 Max age years

性别:

男女均可

Gender:

Both

随机方法(请说明由何人用什么方法产生随机序列):

Randomization Procedure (please state who generates the random number sequence and by what method):

N/A

是否公开试验完成后的统计结果:

Calculated Results after the Study Completed public access:

公开/Public

盲法:

Blinding:

None

试验完成后的统计结果(上传文件):

Calculated Results after
the Study Completed(upload file):

是否共享原始数据:

IPD sharing

Yes

共享原始数据的方式(说明:请填入公开原始数据日期和方式,如采用网络平台,需填该网络平台名称和网址):

研究发表后,去标识化的个体参与者数据 (IPD) 可通过向主要研究者发送电子邮件申请获取,或上传至公共数据存储平台(如 Zenodo/GitHub)

The way of sharing IPD”(include metadata and protocol, If use web-based public database, please provide the url):

De-identified Individual Participant Data (IPD) will be available upon reasonable request to the Principal Investigator via email, or uploaded to public repositories (e.g., Zenodo/GitHub) after publication

数据采集和管理(说明:数据采集和管理由两部分组成,一为病例记录表(Case Record Form, CRF),二为电子采集和管理系统(Electronic Data Capture, EDC),如ResMan即为一种基于互联网的EDC:

纸质病例记录表 (Paper CRF) + 电子数据采集 (EDC)。 说明:各分中心使用统一印制的纸质 CRF 收集源数据,定期由临床协调员 (CRC) 核查并录入加密的电子数据库(Excel/EDC系统)进行管理

Data collection and Management (A standard data collection and management system include a CRF and an electronic data capture:

Paper Case Report Form (CRF) + Electronic Data Capture (EDC). Description: Source data is collected via standardized paper CRFs at each site, then verified and transcribed into a secured electronic database by Clinical Research Coordinators (CRC)

数据与安全监察委员会:

Data and Safety Monitoring Committee:

暂未确定/Not yet

注册人:

Name of Registration:

 2026-01-12 11:49:06