技术领域Technical Field
本发明属于医疗服务技术领域，具体涉及一种基于蒙特卡洛Markov法构建的糖尿病经济学模型及评价方法。The present invention belongs to the technical field of medical services, and in particular relates to a diabetes economics model and an evaluation method based on the Monte Carlo Markov method.
背景技术Background Art
2型糖尿病(Diabetes Mellitus Type 2，T2DM)是糖尿病的主要类型，占糖尿病人群的90％以上，其并发症复杂，包括微血管并发症(如影响肾脏、视网膜和神经系统的并发症)和大血管并发症(心血管疾病等)，临床上常需要口服降糖药物及口服药物和注射降糖药的联合治疗。成本-效果分析在国内日益受到重视。目前国际上针对2型糖尿病人群已经开发和验证了许多模型，这些药物经济学模型大部分通过个案模拟的方式，采用蒙特卡洛结合Markov的方式构建模型，将微血管并发症和大血管并发症纳入其考虑范畴，对患者终生的疾病发生、进展以及产生的成本和健康结果进行模拟。Type 2 diabetes (T2DM) is the main type of diabetes, accounting for more than 90% of the diabetic population. Its complications are complex, including microvascular complications (such as complications affecting the kidneys, retina and nervous system) and macrovascular complications (cardiovascular diseases, etc.). Clinically, oral hypoglycemic drugs and combined treatment with oral and injected hypoglycemic drugs are often required. Cost-effectiveness analysis is gaining increasing attention in China. At present, many models have been developed and verified internationally for people with type 2 diabetes. Most of these pharmacoeconomic models are constructed through case simulation, using Monte Carlo combined with Markov methods, taking microvascular complications and macrovascular complications into consideration, and simulating the occurrence, progression, cost and health outcomes of the disease throughout a patient's life.
现有经典药物经济学模型的风险预测模型中，大血管并发症的应用多考虑UKPDS风险方程和FRAMINGHAM研究，而对于微血管并发症而言，疾病风险预测数据来源于多项不同流行病学研究。目前主流的风险预测方程依托于早期大型研究，但随时代变化，临床护理水平、治疗水平以及生活习惯的发生较大改变，此类风险方程可能会高估并发症和死亡率的发生，因此现有模型对于并发症考虑不全面，结果可能存在一定的偏倚；同时，对于近年来降糖药的新机制，现有药物经济学模型中所采用的生物标记物方法无法完全解释其对于预后的获益，例如胰岛β功能、血糖波动等在药物经济学领域受到的关注不足，无法准确体现创新药的疗效。除此之外，目前国际上众多的药物经济学评价模型异质性大，其底层逻辑、技术及标准完全不同，作为一种应用于实践的标准化模型，不仅需要其能够标准、准确地测算结果，也需要开放给评审者透明使用，以进行公正地评审。In the risk prediction models of existing classic pharmacoeconomic models, the application of macrovascular complications mostly considers the UKPDS risk equation and FRAMINGHAM study, while for microvascular complications, the disease risk prediction data comes from a number of different epidemiological studies. The current mainstream risk prediction equations rely on early large-scale studies, but with the changes of the times, the clinical care level, treatment level and living habits have changed greatly. Such risk equations may overestimate the occurrence of complications and mortality. Therefore, the existing models do not consider complications comprehensively, and the results may be biased. At the same time, for the new mechanisms of hypoglycemic drugs in recent years, the biomarker methods used in the existing pharmacoeconomic models cannot fully explain their benefits for prognosis. For example, pancreatic β function and blood sugar fluctuations have not received enough attention in the field of pharmacoeconomics, and cannot accurately reflect the efficacy of innovative drugs. In addition, many pharmacoeconomic evaluation models in the world are highly heterogeneous, and their underlying logic, technology and standards are completely different. As a standardized model applied in practice, it is not only necessary for it to be able to measure the results in a standard and accurate manner, but also to be open to reviewers for transparent use for fair review.
发明内容Summary of the invention
本发明的目的在于提供一种基于蒙特卡洛Markov法构建的糖尿病经济学模型及其评价方法，其通过EXCEL和VBA语言构建模型，建模逻辑合理，表面效度良好，符合临床实践特点；经过内部验证和交叉验证，显示模型内部效度和交叉效度良好，作为药物经济学评价工具而言，结果可信，外推性好。The purpose of the present invention is to provide a diabetes economics model constructed based on the Monte Carlo Markov method and an evaluation method thereof. The model is constructed by EXCEL and VBA languages, and the modeling logic is reasonable, the surface validity is good, and it conforms to the characteristics of clinical practice. After internal validation and cross validation, it is shown that the internal validity and cross validity of the model are good. As a pharmacoeconomic evaluation tool, the results are credible and the extrapolation is good.
本发明的目的是通过以下技术方案来实现：The purpose of the present invention is to be achieved through the following technical solutions:
一种基于蒙特卡洛Markov法的糖尿病评价模型，包括数据输入模块、疗效计算模块、并发症计算模块以及结果输出模块；A diabetes evaluation model based on Monte Carlo Markov method, including a data input module, an efficacy calculation module, a complication calculation module and a result output module;
所述数据输入模块用于输入患者基线人口学数据、病史数据、患者进入模拟时的生物标志物基线数据、治疗臂不同线程疗效参数和漂移数据、切换线程固定年限/HbA1c阈值；The data input module is used to input patient baseline demographic data, medical history data, biomarker baseline data when the patient enters the simulation, efficacy parameters and drift data of different threads in the treatment arm, and fixed years/HbA1c threshold for switching threads;
所述疗效计算模块用于计算模型模拟过程中生物标志物变化的轨迹，其中该模块模拟患者生理指标随着模拟时间增加而发生改变；The efficacy calculation module is used to calculate the trajectory of biomarker changes during the model simulation process, wherein the module simulates the changes in the patient's physiological indicators as the simulation time increases;
所述并发症计算模块用于预测模拟在模拟时限内各周期并发症的发生情况；在并发症计算模块内模拟各并发症在各周期转移情况，并通过各周期转移情况获取队列中患者在各周期下各并发症患病率及累积发病率，并储存记录；其中所述并发症计算模块包括微血管并发症和大血管并发症的共计7类并发症Markov子模型；其中微血管并发症包括眼部疾病、下肢疾病、肾部疾病，大血管并发症包括缺血性心脏病、心肌梗死、中风、心力衰竭；The complication calculation module is used to predict and simulate the occurrence of complications in each cycle within the simulation time limit; the transfer of each complication in each cycle is simulated in the complication calculation module, and the prevalence and cumulative incidence of each complication in the patients in the cohort in each cycle are obtained through the transfer of each cycle, and the records are stored; wherein the complication calculation module includes a total of 7 types of complication Markov sub-models of microvascular complications and macrovascular complications; wherein microvascular complications include eye diseases, lower limb diseases, and kidney diseases, and macrovascular complications include ischemic heart disease, myocardial infarction, stroke, and heart failure;
所述结果输出模块用于计算并输出干预药物方案最终带来的平均成本、效果和临床结局，所述临床结局包括各周期生物标志物结果、各并发症患病率及累积发病率。The result output module is used to calculate and output the average cost, effect and clinical outcome ultimately brought about by the intervention drug regimen, wherein the clinical outcome includes the biomarker results of each cycle, the prevalence of each complication and the cumulative incidence rate.
进一步的，所述疗效计算模块的模拟方式为手动设定每年固定漂移值，或者使用基于真实世界临床数据拟合得到的风险引擎进行患者生化标志物变化轨迹模拟。Furthermore, the simulation method of the efficacy calculation module is to manually set a fixed drift value each year, or to use a risk engine fitted based on real-world clinical data to simulate the change trajectory of the patient's biochemical markers.
进一步的，所述疗效计算模块还包括患者糖尿病缓解依从性机制模块和β保护机制模块；Furthermore, the efficacy calculation module also includes a patient diabetes remission compliance mechanism module and a beta protection mechanism module;
所述患者糖尿病缓解依从性机制模块用于模拟依从性较差的患者服用一年药物后再停药再继续服药的情景，当开启该机制时，在疗效计算之前将患者分为持续用药队列和间断用药队列分别计算两者临床结局；The patient diabetes remission compliance mechanism module is used to simulate the scenario of a patient with poor compliance taking medication for one year, then stopping the medication and then continuing to take the medication. When the mechanism is turned on, the patients are divided into a continuous medication cohort and an intermittent medication cohort before the efficacy calculation, and the clinical outcomes of the two are calculated separately;
当模型所模拟药物在临床上被证实具有β保护功能时，开启所述β保护机制模块，在每一周期疗效计算时，该模块设置β保护药物糖化血红蛋白漂移值为不存在β保护药物漂移值的80％。When the drug simulated by the model is clinically confirmed to have β-protection function, the β-protection mechanism module is turned on. When calculating the efficacy of each cycle, the module sets the glycosylated hemoglobin drift value of the β-protection drug to 80% of the drift value of the drug without β-protection.
进一步的，所述疗效计算模块支持当患者血压血脂体重生物标志物达到一定条件时，引入高血压药物治疗、高血脂药物治疗、体重控制药物治疗参与药物经济性评价计算。Furthermore, the efficacy calculation module supports the introduction of hypertension drug treatment, hyperlipidemia drug treatment, and weight control drug treatment to participate in the drug economic evaluation calculation when the patient's blood pressure, blood lipids, and weight biomarkers meet certain conditions.
进一步的，在并发症计算模块中，每周期首先对各并发症的发生顺序进行随机排序，发生在前的并发症对后发生的并发症产生风险影响；Furthermore, in the complication calculation module, the occurrence order of each complication is first randomly sorted in each cycle, and the complication that occurs first has a risk impact on the complication that occurs later;
基于疗效计算模块获得的各周期生物标志物结果以及病史状态预测，通过所选用的风险引擎模拟个案各周期并发症发生概率：所述风险引擎选自UKPDS82、BRAVO或CHIME中的任一个；根据公开发表的文献风险影响因素以及分布类型构建的风险方程计算患者在时间t的疾病生存率，并将疾病生存率转化为并发症发生概率，公式如下：Based on the biomarker results of each cycle and the prediction of medical history status obtained by the efficacy calculation module, the probability of complications in each cycle of the case is simulated by the selected risk engine: the risk engine is selected from any one of UKPDS82, BRAVO or CHIME; the risk equation constructed based on the risk influencing factors and distribution types of publicly published literature calculates the disease survival rate of the patient at time t, and converts the disease survival rate into the probability of complications, the formula is as follows:
H为风险概率，即并发症发生概率；S为生存概率；t0为周期间隔，本模型中周期间隔1年；t为并发症发生时间；H is the risk probability, i.e., the probability of complications; S is the survival probability;t0 is the cycle interval, which is 1 year in this model; t is the time of complication occurrence;
当计算得到患者的并发症发生概率后，然后采用计算机生成随机数，当随机数小于当前并发症发生概率时，判断患者健康状态发生转移并记录该患者健康状态；其中以一年为周期模拟个案的并发症发生情况，模拟时限为40年。After the patient's complication probability is calculated, a computer is used to generate a random number. When the random number is less than the current complication probability, the patient's health status is judged to have changed and the patient's health status is recorded. The occurrence of complications in individual cases is simulated in a one-year cycle, and the simulation time limit is 40 years.
在每一周期的模拟中，当7种并发症模拟结束后，通过死亡Markov模块进行死亡模拟，所述Markov模块用于预测判断各患者在模拟周期是否死亡，首先确定转移概率，其次确定死亡状态发生情况。In the simulation of each cycle, after the simulation of the seven complications is completed, the death simulation is performed through the death Markov module. The Markov module is used to predict whether each patient dies in the simulation cycle. First, the transfer probability is determined, and then the occurrence of the death state is determined.
进一步的，所述并发症计算模块还包括血糖波动改善机制模块；Furthermore, the complication calculation module also includes a blood sugar fluctuation improvement mechanism module;
当模型所模拟药物在临床上被证实具有改善血糖波动效果时，开启所述血糖波动改善机制模块，在每一周期并发症发病率计算结束后，再通过血糖波动改善机制对特定并发症发病率及死亡率进行校正；具体地，所述血糖波动改善机制根据已有流行病学研究使用公式对视网膜病变类并发症、外周血管疾病、截肢事件、大量白蛋白尿、死亡发生概率进行校正，校正公式为：P＝P0*HRXi–X1，其中P为校正后的发病率，P0为校正前的发病率，HR为校正值，Xi为试验组的TIR值，X1为对照组的TIR值。When the drug simulated by the model is clinically proven to have the effect of improving blood sugar fluctuations, the blood sugar fluctuation improvement mechanism module is turned on. After the calculation of the complication incidence in each cycle is completed, the blood sugar fluctuation improvement mechanism is used to correct the incidence and mortality of specific complications. Specifically, the blood sugar fluctuation improvement mechanism uses a formula based on existing epidemiological studies to correct the probability of retinopathy complications, peripheral vascular disease, amputation events, massive albuminuria, and death. The correction formula is: P=P0 *HRXi-X1 , where P is the incidence after correction,P0 is the incidence before correction, HR is the correction value, Xi is the TIR value of the experimental group, and X1 is the TIR value of the control group.
本发明还提供利用上述的基于蒙特卡洛Markov法的糖尿病评价模型进行药物经济性评价的评价方法，包括以下步骤：The present invention also provides an evaluation method for drug economic evaluation using the above-mentioned diabetes evaluation model based on the Monte Carlo Markov method, comprising the following steps:
S1、基于蒙特卡洛Markov方法使用EXCEL、VBA代码构建模型基础，确定模型成本参数及效用参数，并填入模型中；S1. Use EXCEL and VBA code to build the model foundation based on the Monte Carlo Markov method, determine the model cost parameters and utility parameters, and fill them into the model;
S2、输入患者基线人口学数据、病史数据、患者进入模拟时的生物标志物基线数据、治疗臂不同线程疗效参数和漂移数据、切换线程固定年限/HbA1c阈值；S2. Input the patient's baseline demographic data, medical history data, biomarker baseline data when the patient enters the simulation, efficacy parameters and drift data of different threads in the treatment arm, and fixed years/HbA1c threshold for switching threads;
S3、确定模型基础设置；S3, determine the basic settings of the model;
S4、基于S2输入的患者基线人口学数据、病史数据、患者进入模拟时的生物标志物基线数据生成虚拟患者，然后进行药物疗效计算，记录虚拟患者各周期生物标志物数据；S4, generating a virtual patient based on the patient's baseline demographic data, medical history data, and biomarker baseline data when the patient enters the simulation input by S2, then performing drug efficacy calculations and recording biomarker data of each cycle of the virtual patient;
S5、模拟当周期各并发症的转移概率，判断患者当周期各并发症的发生情况；S5. Simulate the transfer probability of each complication in the current period and determine the occurrence of each complication in the current period of the patient;
S6、模型更新，进入下一周期进行模拟，直到该患者死亡或到达模拟时限，模型设置计数器将患者上周期所有发生的疾病状态以及病史情况进行记录；S6. The model is updated and enters the next cycle for simulation until the patient dies or the simulation time limit is reached. The model sets a counter to record all the disease states and medical history of the patient in the previous cycle;
S7、患者完成周期全部模拟后，计数器在原本计数的基础上加上该患者疾病发生情况；然后模型对下一患者重复步骤S4-S6，直到模拟个案数达到所设置的个案模拟个数；S7. After the patient completes all simulations of the cycle, the counter adds the occurrence of the patient's disease to the original count; then the model repeats steps S4-S6 for the next patient until the number of simulated cases reaches the set number of case simulations;
S8、计算干预药物方案最终带来的平均成本、效果及临床结局。S8. Calculate the average cost, effectiveness, and clinical outcomes of the intervention drug regimen.
进一步的，步骤S1中确定模型成本参数及效用参数时，用户根据证据等级择优选择成本参数及效用参数；当成本参数及效用参数不存在上下限，采用上下浮动20％的方式模拟成本参数及效用参数的上下限。Furthermore, when determining the model cost parameters and utility parameters in step S1, the user selects the cost parameters and utility parameters based on the level of evidence; when there are no upper and lower limits for the cost parameters and utility parameters, the upper and lower limits of the cost parameters and utility parameters are simulated by floating up and down by 20%.
进一步的，步骤S3中所述的模型基础设置包括个案模拟个数、模型模拟患者换药次数、模拟时限、预测患者健康状态转移概率的风险引擎、模拟患者生物标志物漂移情况、是否开启β保护机制模块、是否开启血糖波动改善机制模块以及是否开启患者糖尿病缓解依从性机制模块，根据药物经济学指南设置适宜的成本效用贴现率。Furthermore, the basic model settings described in step S3 include the number of case simulations, the number of times the model simulates patients changing medications, the simulation time limit, a risk engine for predicting the probability of a patient's health status transition, simulation of the patient's biomarker drift, whether to enable the β protection mechanism module, whether to enable the blood sugar fluctuation improvement mechanism module, and whether to enable the patient's diabetes remission compliance mechanism module, and setting an appropriate cost-effectiveness discount rate according to pharmacoeconomic guidelines.
进一步的，步骤S7中所述的个案模拟个数为1000000。Furthermore, the number of case simulations described in step S7 is 1,000,000.
与现有2型糖尿病药物经济性评价模型相比，本发明的有益之处在于：Compared with the existing type 2 diabetes drug economic evaluation model, the present invention is beneficial in that:
1、本发明提供的糖尿病患者远期成本效用预测模型经过专家咨询后显示，其表面效度良好，建模逻辑合理，符合临床实践特点。经过内部验证和交叉验证，显示模型内部效度和交叉效度良好，作为药物经济学评价工具而言，结果可信，外推性良好。1. The long-term cost-effectiveness prediction model for diabetic patients provided by the present invention has good face validity, reasonable modeling logic, and is in line with the characteristics of clinical practice after expert consultation. After internal validation and cross-validation, it is shown that the model has good internal validity and cross-validation. As a pharmacoeconomic evaluation tool, the results are credible and have good extrapolation.
2、本模型通过采用EXCEL和VBA语言构建模型，进行模型结果运算，透明度良好，适用于专家评审需要，解决了计算机建模常被诟病的“黑箱”问题。2. This model uses EXCEL and VBA languages to construct the model and perform model result calculations. It has good transparency and is suitable for expert review needs, solving the "black box" problem that is often criticized in computer modeling.
3、本模型对于内置的风险引擎进行了更新，除了常用的UKPDS82外，还设置多种风险引擎可供选择，更适用于当前患者临床实践特征。3. This model has updated the built-in risk engine. In addition to the commonly used UKPDS82, it also provides a variety of risk engines to choose from, which are more suitable for the current clinical practice characteristics of patients.
4、本发明比起经典的CORE模型等个案模拟计算速度更快，在CPU为CORE i7的配置下，模拟100万个案仅需要40分钟。4. The present invention has a faster case simulation calculation speed than the classic CORE model. When the CPU is configured as CORE i7, it only takes 40 minutes to simulate 1 million cases.
5、基于既往2型糖尿病药物经济性评价模型难以完全捕获新药的临床获益，本发明将2型糖尿病临床治疗中重点关注的β保护机制、血糖波动改善机制、患者病情缓解机制纳入模型考虑中，能够充分体现新机制新药干预下为患者带来的生命质量改善情况与对于患者终生治疗成本的影响，为胰岛β细胞保护功能、血糖波动控制水平与糖尿病缓解方面疗效有显著改善的新药药物经济学评价提供更为合理、准确的模型工具。5. Based on the fact that previous drug economic evaluation models for type 2 diabetes are difficult to fully capture the clinical benefits of new drugs, the present invention incorporates the β protection mechanism, blood sugar fluctuation improvement mechanism, and patient condition relief mechanism, which are the focus of clinical treatment of type 2 diabetes, into the model considerations, which can fully reflect the improvement in the quality of life of patients brought about by the intervention of new mechanisms and new drugs and the impact on the patients' lifelong treatment costs, and provide a more reasonable and accurate model tool for the pharmacoeconomic evaluation of new drugs that have significantly improved the efficacy in pancreatic β cell protection function, blood sugar fluctuation control level, and diabetes relief.
附图说明BRIEF DESCRIPTION OF THE DRAWINGS
图1为本发明的糖尿病患者远期成本效用预测模型的模型结构示意图；FIG1 is a schematic diagram of the model structure of the long-term cost-effectiveness prediction model for diabetic patients of the present invention;
图2为本发明中所述的各类并发症的Markov状态进展关系示意图；FIG2 is a schematic diagram of the Markov state progression relationship of various complications described in the present invention;
图3为本发明实施例中爬坡试验所得到的模拟成本稳定性结果；FIG3 is a simulation cost stability result obtained by a ramp test in an embodiment of the present invention;
图4为本发明实施例中爬坡试验所得到的模拟效果稳定性结果；FIG4 is a simulation result of the stability of the slope test obtained in an embodiment of the present invention;
图5为本发明实施例中爬坡试验所得到的模拟ICER稳定性结果；FIG5 is a simulated ICER stability result obtained from a hill climbing test in an embodiment of the present invention;
图6为本发明实施例中爬坡试验所得到的NMR稳定性结果；FIG6 is an NMR stability result obtained from a ramp test in an embodiment of the present invention;
图7为本发明实施例中交叉验证背景视网膜病变累计发病率结果；FIG7 is a result of cross-validation of the cumulative incidence of background retinopathy in an embodiment of the present invention;
图8为本发明实施例中交叉验证背景视网膜病变患病率结果；FIG8 is a cross-validation result of the background retinopathy prevalence in an embodiment of the present invention;
图9为本发明实施例中内部验证背景视网膜病变累计发病率结果；FIG9 is a result of internal verification of the cumulative incidence of background retinopathy in an embodiment of the present invention;
图10为本发明实施例中内部验证背景视网膜病变患病率结果。FIG. 10 is the result of internal verification of the background retinopathy prevalence in an embodiment of the present invention.
具体实施方式DETAILED DESCRIPTION
为了相关技术领域人员更好的理解本发明专利的内容，下面对本发明的实施例作详细说明，本实施例在以本发明技术方案为前提下进行实施，给出了详细的实施方式和具体的操作过程，但本发明的内容不限于下述的实例。In order to help people in the relevant technical field better understand the content of the patent of the present invention, the embodiments of the present invention are described in detail below. This embodiment is implemented on the premise of the technical solution of the present invention, and a detailed implementation method and specific operation process are given, but the content of the present invention is not limited to the following examples.
实施例1Example 1
本发明采用受到国际上广泛认可的2型糖尿病药物经济性评价模型思路，使用队列研究开始、国际认可度较高的风险引擎预测患者疾病转移状态，参考中国药物经济学评价指南构建经济性评价模型。本实施例公开了模型结构以及基于所述蒙特卡洛Markov法构建的糖尿病经济学模型的评价方法，如图1所示，评价方法具体包括如下步骤：The present invention adopts the idea of the economic evaluation model of type 2 diabetes drugs that is widely recognized internationally, uses the risk engine that has been widely recognized internationally since the beginning of the cohort study to predict the disease transfer status of patients, and builds an economic evaluation model with reference to the Chinese Pharmacoeconomic Evaluation Guide. This embodiment discloses the model structure and the evaluation method of the diabetes economic model constructed based on the Monte Carlo Markov method, as shown in Figure 1, and the evaluation method specifically includes the following steps:
步骤一、基于蒙特卡洛Markov方法使用EXCEL、VBA代码构建模型基础，确定模型成本参数及效用参数，并填入模型中。Step 1: Use EXCEL and VBA code to build the model foundation based on the Monte Carlo Markov method, determine the model cost parameters and utility parameters, and fill them into the model.
根据需要选择药物临床试验方案，获取药物临床试验方案的患者基线人口学数据和干预药物的疗效参数；若出现模型所需数值但临床未汇报的情况，使用META等获取相关标准数据；同时通过文献综述确定模型成本参数及效用参数，本模型将计算患者三个部分的成本：患者背景治疗成本、用药成本、并发症处置成本。患者背景治疗成本主要为患者为治疗糖尿病就诊开具处方等为治疗药物所产生的非直接治疗成本；并发症处置成本分为：事件成本以及状态成本，其中事件成本使用并发症发生率计算、状态成本使用并发症流行率计算。Select the drug clinical trial program as needed, obtain the patient baseline demographic data and the efficacy parameters of the intervention drug of the drug clinical trial program; if the values required by the model appear but are not reported clinically, use META to obtain relevant standard data; at the same time, determine the model cost parameters and utility parameters through literature review. This model will calculate the cost of three parts of the patient: patient background treatment cost, medication cost, and complication management cost. The patient background treatment cost is mainly the indirect treatment cost incurred by the patient for prescriptions for treatment of diabetes; the complication management cost is divided into: event cost and state cost, among which the event cost is calculated using the complication incidence rate and the state cost is calculated using the complication prevalence rate.
步骤二、在上述模型中输入患者基线人口学数据、病史数据、患者进入模拟时的生物标志物基线数据、治疗臂不同线程疗效参数和漂移数据、切换线程固定年限/HbA1c阈值。Step 2: Input the patient's baseline demographic data, medical history data, biomarker baseline data when the patient enters the simulation, efficacy parameters and drift data of different threads in the treatment arm, and fixed years/HbA1c threshold for switching threads into the above model.
步骤三、确定模型基础设置：设置个案模拟个数、模型模拟患者换药次数、模拟实现(即模拟最长时间)、预测患者健康状态转移概率的风险引擎、模拟患者生物标志物漂移情况以及是否开启β保护机制模块、血糖波动改善机制模块和患者糖尿病缓解依从性机制模块，根据药物经济学指南设置适宜的成本效用贴现率；其中β保护机制模块、血糖波动改善机制模块、患者糖尿病缓解依从性机制模块参数主要来源于相关流行病学文献数据整理，且在进行药学专家、临床专家、药物经济学专家的咨询后设置。在模型的基础设置中，当模型在疗效计算模块选用基于真实世界临床数据拟合得到的风险引擎进行患者生化标志物变化轨迹模拟时，会根据用户选择的疗效漂移方程变动生物标志物并记录统计后续风险引擎中所需的一切指标(如CHIME风险引擎中需要评估患者是否使用胰岛素以及降低血脂类药物)；随后使用蒙特卡洛Markov方法模拟单个患者病史基线使其进入周期的状态模拟中；通过文献查找风险引擎数据，根据生存风险方程将文献汇报的风险引擎公式加载入模型中；模型设计UKPDS82、BRAVO、CHIME等多套风险引擎可选择，进入模拟时将会根据患者选择的风险引擎计算每个周期并发症的发生概率，当生成随机数小于发生概率时则认为患者当年发生该疾病；多次进行模拟将患者发生疾病的次数记录下来，汇总计算转换为疾病发生概率数据，利用该数据最终输出患者预期生存年、质量生命年、成本等药学经济学数据。Step 3. Determine the basic settings of the model: set the number of case simulations, the number of times the model simulates the patient's medication changes, the simulation realization (i.e., the longest simulation time), the risk engine that predicts the probability of the patient's health status transition, simulate the patient's biomarker drift, and whether to turn on the β protection mechanism module, the blood sugar fluctuation improvement mechanism module, and the patient's diabetes remission compliance mechanism module; set an appropriate cost-effectiveness discount rate according to the pharmacoeconomic guidelines; the parameters of the β protection mechanism module, the blood sugar fluctuation improvement mechanism module, and the patient's diabetes remission compliance mechanism module are mainly derived from the relevant epidemiological literature data, and are set after consulting with pharmaceutical experts, clinical experts, and pharmacoeconomics experts. In the basic settings of the model, when the model uses the risk engine obtained by fitting based on real-world clinical data to simulate the change trajectory of patient biochemical markers in the efficacy calculation module, the biomarkers will be changed according to the efficacy drift equation selected by the user and all the indicators required in the subsequent risk engine will be recorded and counted (such as the CHIME risk engine needs to evaluate whether the patient uses insulin and lipid-lowering drugs); then the Monte Carlo Markov method is used to simulate the medical history baseline of a single patient to put it into the state simulation of the cycle; the risk engine data is searched through the literature, and the risk engine formula reported in the literature is loaded into the model according to the survival risk equation; the model is designed with multiple risk engines such as UKPDS82, BRAVO, and CHIME to choose from. When entering the simulation, the probability of complications in each cycle will be calculated according to the risk engine selected by the patient. When the generated random number is less than the probability of occurrence, it is considered that the patient has the disease that year; multiple simulations are performed to record the number of times the patient has the disease, and the data is summarized and converted into disease occurrence probability data, and the data is used to finally output the patient's expected survival years, quality life years, cost and other pharmaceutical economics data.
(1)β保护机制模块(1) β protection mechanism module
胰岛β细胞功能是指胰岛β细胞分泌胰岛素来维持血糖稳定的能力，分泌胰岛素有慢性分泌波动和快速脉冲分泌波动两种形式，对维持全天血糖平稳起着至关重要的作用。恢复胰岛β细胞功能及体量对治疗糖尿病具有重要意义。新药较传统药物在胰岛β功能上存在疗效上的优势。HbA1c是2型糖尿病患者血糖控制水平的金标准。UKPDS研究在超过20年的长期前瞻性队列中观察到，随着患者持续应用降糖药物，其在用药第二年开始，将出现HbA1c的持续上升，平均增幅为0.15％，即HbA1c的逐年漂移。以往研究中，对于模型的类似校正如心血管保护机制主要通过以下5种方法进行校正，见表1。Pancreatic β-cell function refers to the ability of pancreatic β-cells to secrete insulin to maintain blood sugar stability. Insulin secretion has two forms: chronic secretion fluctuations and rapid pulse secretion fluctuations. It plays a vital role in maintaining blood sugar stability throughout the day. Restoring pancreatic β-cell function and volume is of great significance for the treatment of diabetes. New drugs have therapeutic advantages over traditional drugs in pancreatic β-function. HbA1c is the gold standard for blood sugar control in patients with type 2 diabetes. The UKPDS study observed in a long-term prospective cohort of more than 20 years that as patients continued to use hypoglycemic drugs, their HbA1c would continue to rise starting in the second year of medication, with an average increase of 0.15%, that is, the year-on-year drift of HbA1c. In previous studies, similar corrections to the model, such as cardiovascular protection mechanisms, were mainly corrected through the following 5 methods, see Table 1.
表1心血管保护校正方法Table 1 Cardiovascular protection correction method
本模型假设：a.胰岛β细胞功能稳定时，由于其能促进胰岛素的正常代偿性分泌，血糖水平将不会上升。当降糖药物胰岛β细胞功能存在一定保护效果时，其血糖逐年增幅将会有所放缓。b.本模型假设其下降20％，出于保守的研究角度，设置该效果维持5年，随后血糖按照正常增幅进行漂移。This model assumes: a. When the pancreatic β-cell function is stable, blood sugar levels will not rise because they can promote normal compensatory secretion of insulin. When hypoglycemic drugs have a certain protective effect on pancreatic β-cell function, the annual increase in blood sugar will slow down. b. This model assumes that it will drop by 20%. From a conservative research perspective, this effect is set to last for 5 years, and then blood sugar will drift at a normal rate.
模型中β保护机制算法设计：所述模型加载β保护功能，模拟个案HbA1c年漂移值为正常无β保护漂移值的80％。即：The algorithm design of the β protection mechanism in the model: The model is loaded with the β protection function, and the simulated case HbA1c annual drift value is 80% of the normal drift value without β protection. That is:
第i年HbA1c值＝第i-1年HbA1c值+0.8*正常无β保护HbA1c年漂移值。HbA1c value in year i = HbA1c value in year i-1 + 0.8*normal annual drift value of HbA1c without β protection.
(2)血糖波动改善机制模块(2) Blood sugar fluctuation improvement mechanism module
TIR(Time in range)是动态血糖监测里的核心指标，指24h内葡萄糖在目标范围(3.9～10.0mmol/L)内的时间或所占的百分比。《TIR国际共识》推荐TIR作为动态血糖监测(CGM)临床应用的十个最主要参数指标之一。当前大部分研究采用动态血糖监测作为血糖波动的评价方法。此外，关于TIR与并发症预后相关性的研究较为丰富，为其作为血糖波动机制的量化指标提供一定的临床证据。因此，本模型考虑将TIR作为血糖波动改善机制的效果评价指标。TIR (Time in range) is the core indicator in dynamic blood glucose monitoring, which refers to the time or percentage of glucose in the target range (3.9-10.0mmol/L) within 24 hours. The "TIR International Consensus" recommends TIR as one of the ten most important parameter indicators for the clinical application of dynamic blood glucose monitoring (CGM). Most current studies use dynamic blood glucose monitoring as an evaluation method for blood glucose fluctuations. In addition, there are relatively rich studies on the correlation between TIR and the prognosis of complications, which provide certain clinical evidence for its use as a quantitative indicator of the mechanism of blood glucose fluctuations. Therefore, this model considers using TIR as an effect evaluation indicator for the mechanism of improving blood glucose fluctuations.
目前已有多项研究证明，TIR与微血管并发症如视网膜病变、神经病变、肾脏病变发生减少有关，同时与心血管死亡率与全死亡率独立相关，见表2。Several studies have shown that TIR is associated with a reduction in microvascular complications such as retinopathy, neuropathy, and nephropathy, and is independently associated with cardiovascular mortality and all-cause mortality (see Table 2).
表2TIR与并发症之间关系Table 2 Relationship between TIR and complications
本实施例中，所述模型加载血糖波动改善功能，根据已有流行病学研究使用公式对视网膜病变类并发症、外周血管疾病、截肢事件、大量白蛋白尿、死亡发生概率进行矫正。矫正公式如下：In this embodiment, the model is loaded with a blood sugar fluctuation improvement function, and the probability of occurrence of complications of retinopathy, peripheral vascular disease, amputation events, macroalbuminuria, and death is corrected using a formula based on existing epidemiological studies. The correction formula is as follows:
(3)患者糖尿病缓解依从性机制模块(3) Patient diabetes remission compliance mechanism module
当开启患者糖尿病缓解依从性机制模块时，在疗效计算之前将患者分为持续用药队列和间断用药队列分别计算两者临床结局情况，在间断用药队列对依从性较差的患者服用一年药物后再停药再继续服药的情景进行模拟。When the patient diabetes remission compliance mechanism module is turned on, the patients are divided into a continuous medication cohort and an intermittent medication cohort before the efficacy calculation, and the clinical outcomes of the two are calculated separately. In the intermittent medication cohort, a scenario is simulated in which patients with poor compliance take medication for one year, then stop taking the medication, and then continue taking the medication.
步骤四、基于上述步骤二输入的患者基线人口学数据、病史数据、患者进入模拟时的生物标志物基线数据生成虚拟患者，然后通过患者进入模拟时的生物标志物基线数据(或模拟过程中各上周期生物标志物数据)及本线程干预药物疗效参数进行当周期的疗效模拟，计算各当周期药物疗效，记录患者各当周期生物标志物数据。Step 4. Generate a virtual patient based on the patient's baseline demographic data, medical history data, and biomarker baseline data when the patient enters the simulation input in step 2 above. Then, use the biomarker baseline data when the patient enters the simulation (or the biomarker data of each previous cycle during the simulation process) and the intervention drug efficacy parameters of this thread to simulate the efficacy of the current cycle, calculate the drug efficacy of each current cycle, and record the patient's biomarker data of each current cycle.
模型使用蒙特卡洛抽样模拟患者从进入研究队列后每个周期(一年)生物标志物变化趋势以及病史状态。本实施例中对于用药第一年疗效采用RCT试验数据，将其作为用药/换药第一年的疗效数据进行计算，后续模拟年限中，由于难以对患者进行数据观测，需进行预测模拟，本实施例根据步骤三所确定的模型基础设置，采用手动设定每年固定漂移值的方式进行生物标志物的模拟，对于HbA1c，采用UKPDS33和UKPDS59的数据，假设每年增长0.15％；或者也可以根据所选择的风险方程相应的漂移方程进行生物标志物的模拟。The model uses Monte Carlo sampling to simulate the biomarker change trend and medical history status of patients in each cycle (one year) after entering the study cohort. In this embodiment, RCT test data is used for the efficacy of the first year of medication, and it is calculated as the efficacy data of the first year of medication/drug change. In the subsequent simulation years, due to the difficulty in observing patient data, predictive simulation is required. This embodiment uses the model basic settings determined in step 3 to manually set a fixed drift value each year to simulate biomarkers. For HbA1c, UKPDS33 and UKPDS59 data are used, assuming an annual increase of 0.15%; or the biomarker simulation can also be performed according to the drift equation corresponding to the selected risk equation.
步骤五、基于上述患者基线人口学数据、病史数据(或上周期患者人口学数据、病史数据)以及上述步骤四所得到的患者各当周期生物标志物数据，模拟当周期各并发症的发生情况。Step 5: Based on the above patient baseline demographic data, medical history data (or the demographic data and medical history data of the patient in the previous cycle) and the biomarker data of the patient in each cycle obtained in the above step 4, simulate the occurrence of various complications in the current cycle.
模型中患者并发症包含微血管并发症和大血管并发症，其中微血管并发症包括眼部疾病(包含6个Markov状态)、下肢疾病(包含5个Markov状态)和肾部疾病(包含4个Markov状态)；大血管并发症包括缺血性心脏病(包含2个Markov状态)、心肌梗死(包含5个Markov状态)、中风(包含5个Markov状态)、心力衰竭(包含2个Markov状态)，上述各类并发症的Markov状态进展关系如图2所示。Patient complications in the model include microvascular complications and macrovascular complications. Microvascular complications include eye diseases (including 6 Markov states), lower limb diseases (including 5 Markov states) and kidney diseases (including 4 Markov states); macrovascular complications include ischemic heart disease (including 2 Markov states), myocardial infarction (including 5 Markov states), stroke (including 5 Markov states), and heart failure (including 2 Markov states). The Markov state progression relationship of the above complications is shown in Figure 2.
模型在并发症模拟部分，每周期首先对于各并发症的发生顺序进行随机排序，发生在前的并发症可对于后发生的并发症产生风险影响；其次通过模型所选用的风险引擎基于各周期生物标志物结果以及病史状态预测模拟个案各周期的并发症发生概率；当计算得到患者的并发症发生概率后，在模型中随机生成一个0-1之间的数字，当该数字小于并发症发生概率后，判断该疾病发生，同时模型对于患者的发病状态和患病状态分别进行记录。In the complication simulation part, the model first randomly sorts the order of occurrence of each complication in each cycle. Complications that occur earlier can have a risk impact on complications that occur later. Secondly, the risk engine selected by the model predicts the probability of complication occurrence in each cycle of the simulated case based on the biomarker results and medical history status of each cycle. After the patient's complication probability is calculated, a number between 0 and 1 is randomly generated in the model. When the number is less than the complication probability, the disease is judged to have occurred. At the same time, the model records the patient's onset status and disease status respectively.
步骤六、模型更新，进入下一周期进行模拟，直到该患者死亡或达到模拟最长时间40年，模型设置计数器将患者上周期所有发生的疾病状态以及病史情况进行记录以便于后续并发症发生率以及流行率的计算，同时为患者本周期疾病状态模拟提供病史依据。Step 6: Update the model and enter the next cycle for simulation until the patient dies or the maximum simulation time of 40 years is reached. The model sets a counter to record all the disease conditions and medical history of the patient in the previous cycle to facilitate the calculation of the incidence and prevalence of subsequent complications, and at the same time provide a medical history basis for the simulation of the patient's disease status in this cycle.
步骤七、患者完成周期全部模拟后，计数器在原本计数的基础上加上该患者疾病发生情况；然后模型对下一患者重复上述步骤四-六，直到模拟个案数达到所设置的个案模拟个数100万。Step 7. After the patient completes all simulations of the cycle, the counter adds the patient's disease occurrence to the original count; then the model repeats steps 4-6 for the next patient until the number of simulated cases reaches the set number of case simulations of 1 million.
步骤八、计算并输出干预药物方案最终带来的平均成本、效果及临床结局；所述临床结局包括各周期生物标志物结果、各并发症患病率及累积发病率。Step 8. Calculate and output the average cost, effect and clinical outcomes of the intervention drug regimen; the clinical outcomes include the biomarker results of each cycle, the prevalence of each complication and the cumulative incidence.
实施例2：成本-效用分析结果Example 2: Cost-utility analysis results
(一)选取药物干预方案临床试验，通过临床文献综述获取药物干预方案临床试验的患者基线人口学数据、生物标志物基线数据以及干预药物疗效参数(I) Select clinical trials of drug intervention programs and obtain baseline patient demographic data, biomarker baseline data, and intervention drug efficacy parameters from clinical trials of drug intervention programs through clinical literature review
本实施例选取于2021年获批用于治疗2型糖尿病的新药西格列他钠进行药物经济学评价。In this example, sitagliptin, a new drug approved in 2021 for the treatment of type 2 diabetes, was selected for pharmacoeconomic evaluation.
检索获取西格列他钠临床随机对照试验的文献，其中有一篇文献汇报了患者接受干预方案后的TIR改变量(time in range患者一天内血糖正常的概率，用于反应血糖波动)，选择该文献数据用于药物经济性评价。Literature on randomized controlled clinical trials of sitagliptin was retrieved, and one of the articles reported the change in TIR (the probability of a patient having normal blood sugar in time in range within a day, used to reflect blood sugar fluctuations) after the patients received the intervention program. The data from this article were selected for drug economic evaluation.
文献汇报患者基线情况如表3，其中低密度脂蛋白、高密度脂蛋白、心率、白蛋白计数指标文献并未汇报，但考虑到风险引擎需要使用到这些参数，故采用正常指标范围中值代替。患者并发症病史情况并未汇报，同样假设入组患者无并发症病史情况。The baseline conditions of the patients reported in the literature are shown in Table 3. The low-density lipoprotein, high-density lipoprotein, heart rate, and albumin count indicators were not reported in the literature, but considering that the risk engine needs to use these parameters, the median of the normal indicator range was used instead. The history of complications of the patients was not reported, and it was also assumed that the enrolled patients had no history of complications.
表3西格列他钠临床试验患者基线情况Table 3 Baseline conditions of patients in clinical trials of Siglitactam
同时确定文献汇报患者的模型可用主要疗效为HbA1C、TIR，患者接受西格列他钠干预治疗后，患者HbA1C改善了1.37％，TIR提升了17.01％；同样接受西格列汀治疗的患者HbA1C改善了1.35％，TIR提升了37.50％。两者在HbA1C改善程度上差距不大，若使用传统糖尿病模型将出现低估西格列汀在TIR上的疗效。同样的，西格列他钠改善了患者的胰岛素抵抗，但使用西格列汀干预的患者的胰岛素抵抗出现了恶化，传统糖尿病模型也会低估西格列他钠在胰岛素抵抗上的改善。At the same time, it was determined that the main efficacy of the model reported in the literature was HbA1C and TIR. After the patients received sitagliptin intervention treatment, their HbA1C improved by 1.37% and TIR increased by 17.01%. Similarly, the HbA1C of patients treated with sitagliptin improved by 1.35% and TIR increased by 37.50%. There was not much difference between the two in the degree of improvement in HbA1C . If the traditional diabetes model is used, the efficacy of sitagliptin on TIR will be underestimated. Similarly, sitagliptin improved the insulin resistance of patients, but the insulin resistance of patients treated with sitagliptin worsened, and the traditional diabetes model would also underestimate the improvement of sitagliptin on insulin resistance.
(二)根据药物干预方案临床试验结果确定个案模拟的状态转移模型的基础设置(II) Determine the basic settings of the state transition model for case simulation based on the clinical trial results of the drug intervention program
模型模拟最长时间选择为40年；为稳定结果模拟个案次数选择1000000人；选择国际认可度较高的UKPDS82风险引擎模拟患者大血管并发症发生概率以及死亡概率。同时，也选择与UKPDS82配套的生物标志物漂移方程UKPDS90模拟患者40年中生物标志物变化轨迹。根据临床研究提供的药物疗效结果，可打开模型中β保护机制模块和血糖波动改善机制模块的设置。The longest simulation time of the model is selected as 40 years; the number of simulated cases is selected as 1,000,000 people to stabilize the results; the internationally recognized UKPDS82 risk engine is selected to simulate the probability of large vascular complications and the probability of death in patients. At the same time, the biomarker drift equation UKPDS90, which is matched with UKPDS82, is also selected to simulate the biomarker change trajectory of patients in 40 years. According to the drug efficacy results provided by clinical studies, the settings of the β protection mechanism module and the blood sugar fluctuation improvement mechanism module in the model can be opened.
(三)将上述干预药物疗效填写入模型中；同时设置临床更换药物方案的条件；后将所获得的患者基线数值和各健康状态治疗成本参数及效用参数填写入模型中，所述各健康状态治疗成本参数及效用参数经文献汇总如下表4、表5所示。(III) Enter the efficacy of the above-mentioned intervention drugs into the model; at the same time, set the conditions for clinically changing the drug regimen; and then enter the obtained patient baseline values and the treatment cost parameters and utility parameters of each health state into the model. The treatment cost parameters and utility parameters of each health state are summarized in the literature and are shown in Tables 4 and 5 below.
本实施例参考糖尿病治疗指南设置当患者HbA1C达到8％时模拟个案更换用药方案；更换后方案统一使用甘精胰岛素治疗可改善HbA1C1.5％。根据药物临床文献/说明书，计算药物年干预成本：西格列他钠为2131.6元、西格列汀为2587.1元、甘精胰岛素为1492.88元。This example refers to the diabetes treatment guidelines to simulate the case change medication regimen when the patient's HbA1C reaches 8%; after the change, the regimen uses glargine insulin treatment to improve HbA1C by 1.5%. According to the drug clinical literature/instructions, the annual drug intervention cost is calculated: sitagliptin is 2131.6 yuan, sitagliptin is 2587.1 yuan, and glargine insulin is 1492.88 yuan.
表4成本参数Table 4 Cost parameters
表5效用参数Table 5 Utility parameters
(四)、当模型所需所有设置被填写入模型中，运行内置VBA程序，计算获取患者在接受药物干预后各周期各并发症的累计发生率以及患者处于各健康状态的概率。使用健康状态的效用成本参数以及药物干预的成本参数计算周期成本效用，最终加总获取药物干预方案的总成本与总效用。(IV) When all the settings required by the model are filled in, run the built-in VBA program to calculate the cumulative incidence of each complication in each cycle after the patient receives drug intervention and the probability of the patient being in each health state. Use the utility cost parameters of the health state and the cost parameters of the drug intervention to calculate the cycle cost utility, and finally add up the total cost and total utility of the drug intervention plan.
模型计算结果如下表6所示：The model calculation results are shown in Table 6 below:
表6模型计算结果Table 6 Model calculation results
利用EXCEL的制图功能绘制生化标志物漂移曲线、患者累计发病概率、患者生存率便于用户直观的观察。The mapping function of EXCEL is used to draw the drift curve of biochemical markers, the cumulative incidence probability of patients, and the survival rate of patients for users to observe intuitively.
由最终结果可知，虽然西格列他钠与西格列汀在HbA1C的改善上几乎一致，但是本发明在引入血糖波动改善这一机制，由于西格列汀在血糖波动改善上远优于西格列他钠，因此患者在特定并发症上损失的健康效用更低，生存时长也更优于西格列他钠。不过，由于西格列他钠直接干预成本对比西格列汀更低，即便西格列汀在效果上略微由于西格列他钠，但依然不具有成本-效果优势。From the final results, it can be seen that although sitagliptin and sitagliptin are almost the same in improvingHbA1C , the present invention introduces the mechanism of improving blood sugar fluctuations. Since sitagliptin is far superior to sitagliptin in improving blood sugar fluctuations, the patient's health utility loss due to specific complications is lower, and the survival time is also better than sitagliptin. However, since the direct intervention cost of sitagliptin is lower than that of sitagliptin, even if sitagliptin is slightly better than sitagliptin in effect, it still does not have a cost-effectiveness advantage.
实施例3：模型稳定性Example 3: Model stability
爬坡测试通过不断修改患者模拟次数并观察不同梯度上模拟次数结果稳定程度以判断在该模拟个数时模型结果是否稳定且收敛，旨在得到最优样本预测。The ramp test aims to obtain the optimal sample prediction by continuously modifying the number of patient simulations and observing the stability of the simulation results at different gradients to determine whether the model results are stable and convergent at the number of simulations.
本实施例患者基线特征，干预药物疗效参数、成本参数、效用参数使用与实施例2相同数据。The patient baseline characteristics, intervention drug efficacy parameters, cost parameters, and utility parameters of this example use the same data as those in Example 2.
本实施例中模型基础设置除生物标志物漂移选择手动设定每年固定漂移值、模拟个案次数外，其余设置与实施例2相同。The basic model settings in this embodiment are the same as those in Embodiment 2 except that the biomarker drift selection manually sets the fixed drift value per year and the number of simulated cases.
本实施例中对于模型进行10000*1000次模拟的个案阶梯爬坡，结果显示在进行100万个个案的基础模拟时，增量成本、增量效用、ICER、NMB趋于稳定(见图3、图4、图5、图6)，基于此结果，本模型基础模拟设置为100万个案，运行时间约40分钟。In this embodiment, the model is simulated 10000*1000 times in a step-by-step manner. The results show that when a basic simulation of 1 million cases is performed, the incremental cost, incremental utility, ICER, and NMB tend to be stable (see Figures 3, 4, 5, and 6). Based on this result, the basic simulation of this model is set to 1 million cases, and the running time is about 40 minutes.
实施例4：模型验证Example 4: Model Validation
本实施例患者基线特征，干预药物疗效参数、成本参数、效用参数使用与实施例2相同数据。The patient baseline characteristics, intervention drug efficacy parameters, cost parameters, and utility parameters of this example use the same data as those of Example 2.
本实施例中模型基础设置除生物标志物漂移选择手动设定每年固定漂移值外，其余设置与实施例2相同。The basic model settings in this embodiment are the same as those in Embodiment 2 except that the biomarker drift is selected to manually set a fixed drift value each year.
内部验证：对同一个模型采用不同的建模方式，最终观察两种模型结果是否一致，两种模型同时采用同一组参数数据模拟患者并发症累计发生率以及预期生存年结果。本实施例中采用相同的数据分别在以EXCEL公式建模及VBA建模中模拟进行内部验证，其中VBA内置模型中采用先生成随机数将其储存在临时数组中，后续模型中直接调用数组中的随机数，考虑到VBA模型生成时间较长选择复杂化生存随机数公式的方式来确保模型结果正确性，其公式为：Rnd(Int(Rnd()*10000))，该公式在模型极端值验证下结果较为接近真实情况(即研究对照组参数一致结果一致)。本实施例以背景视网膜病变为例，计算得EXCEL公式建模和VBA建模的累积发病率、患病率如图9、图10所示，可以看到两模型结果基本一致，模型内部效度良好。Internal validation: different modeling methods are used for the same model, and the two model results are finally observed to be consistent. The two models use the same set of parameter data to simulate the cumulative incidence of complications and the expected survival year results of patients. In this embodiment, the same data is used to simulate the internal validation in EXCEL formula modeling and VBA modeling, wherein the VBA built-in model adopts the first generation of random numbers and stores them in a temporary array, and the random numbers in the array are directly called in the subsequent model. Considering that the VBA model generation time is long, the method of selecting the complicated survival random number formula is used to ensure the correctness of the model results. Its formula is: Rnd (Int (Rnd () * 10000)), and the formula is closer to the actual situation under the model extreme value verification (i.e., the parameters of the control group are consistent). In this embodiment, background retinopathy is taken as an example, and the cumulative incidence and prevalence of EXCEL formula modeling and VBA modeling are calculated as shown in Figures 9 and 10. It can be seen that the results of the two models are basically consistent, and the internal validity of the model is good.
交叉验证：使用另一种已有成熟模型(即验证模型)模拟患者并发症累计发生率以及预期生存年结果。当模拟参数基本一致时，观察被验证模型结果与验证模型结果一致性程度。本实施例中通过与IHE模型采用相同的数据模拟进行交叉验证。以背景视网膜病变为例，计算得本模型和IHE模型的累积发病率、患病率如图7、图8所示，二者大致重合，表明模型交叉效度良好。Cross-validation: Use another existing mature model (i.e., validation model) to simulate the cumulative incidence of complications and expected survival years of patients. When the simulation parameters are basically the same, observe the consistency between the results of the validated model and the validation model. In this embodiment, cross-validation is performed by using the same data simulation as the IHE model. Taking background retinopathy as an example, the cumulative incidence and prevalence of this model and the IHE model are calculated as shown in Figures 7 and 8, and the two roughly overlap, indicating that the model has good cross-validity.
模型验证参数主要使用决定系数(R2)越接近1拟合优度越好、对称平均，绝对百分比误差(SMAPE)越接近0拟合优度越好、均方根百分比误差(RMSPE)越接近0拟合优度越好，其运行公式分别为：The model validation parameters mainly use the determination coefficient (R2 ) closer to 1, the better the goodness of fit, symmetrical average, the absolute percentage error (SMAPE) closer to 0, the better the goodness of fit, and the root mean square percentage error (RMSPE) closer to 0, the better the goodness of fit. The operating formulas are:
在上述背景视网膜疾病中，内部验证结果为：R2为0.977、SMAPE为0.046、RMSPE为0.002；交叉验证结果为：R2为0.999、SMAPE为0.009、RMSPE为0.005。验证结果显示，当输入参数一致、选择风险引擎一致时，模型验证效度较高，结果可信。In the above background retinal diseases, the internal validation results were: R2 was 0.977, SMAPE was 0.046, and RMSPE was 0.002; the cross-validation results were: R2 was 0.999, SMAPE was 0.009, and RMSPE was 0.005. The validation results showed that when the input parameters were consistent and the risk engine was selected consistently, the model validation validity was high and the results were credible.
实施例4：单/多因素、概率敏感性分析Example 4: Single/multi-factor, probabilistic sensitivity analysis
考虑到政策的调整，药品价格可能会有所变化；对于并发症发生治疗成本本研究均采用各类处置项目打包式成本计算，真实世界患者的治疗成本可能会有较大差异；此外，关于健康效用值，本研究均取自国内外相关文献，且由于国内相关研究匮乏，相关健康效用值可能与我国患者实际情况有一定差异。Taking into account policy adjustments, drug prices may change; for the treatment costs of complications, this study used packaged costs of various treatment items, and the treatment costs of real-world patients may vary greatly; in addition, regarding health utility values, this study was taken from relevant domestic and foreign literature, and due to the lack of relevant domestic research, the relevant health utility values may be somewhat different from the actual situation of patients in my country.
为检验模型结果稳定性，本模型采用二阶蒙特卡洛法对以下变量进行敏感性分析：实验组和对照组各个状态下涉及的健康效用值，实验组和对照组药品成本、并发症成本及其他综合项目组成本，不良反应和低血糖处理成本，贴现率。无真实世界数据浮动上下限的参数，浮动范围设置为基线上下20％。同时绘制单因素敏感性分析旋风图、概率敏感性分析散点图、增量成本-效果可接受度分析曲线图，观察到填写入模型参数值发生变化后输出结果是否仍然稳定。In order to test the stability of the model results, this model uses the second-order Monte Carlo method to perform sensitivity analysis on the following variables: health utility values involved in each state of the experimental group and the control group, drug costs, complication costs and other comprehensive project group costs, adverse reactions and hypoglycemia treatment costs, and discount rates. There are no parameters with real-world data floating upper and lower limits, and the floating range is set to 20% above and below the baseline. At the same time, a single-factor sensitivity analysis tornado chart, a probability sensitivity analysis scatter plot, and an incremental cost-effect acceptability analysis curve chart are drawn to observe whether the output results are still stable after the input model parameter values change.
以上所述仅为本发明的较佳实施例，并不以上述实施方式为限，但凡本领域普通技术人员根据本发明所揭示内容所作的等效修饰、等同替换和改进等，皆应纳入权利要求书中记载的保护范围。The above description is only a preferred embodiment of the present invention and is not limited to the above implementation mode. Any equivalent modifications, equivalent substitutions and improvements made by ordinary technicians in this field based on the contents disclosed by the present invention should be included in the protection scope recorded in the claims.
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310793410.5ACN117253618B (en) | 2023-06-30 | 2023-06-30 | Diabetes evaluation model and method based on Monte Carlo Markov method |
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310793410.5ACN117253618B (en) | 2023-06-30 | 2023-06-30 | Diabetes evaluation model and method based on Monte Carlo Markov method |
| Publication Number | Publication Date |
|---|---|
| CN117253618Atrue CN117253618A (en) | 2023-12-19 |
| CN117253618B CN117253618B (en) | 2024-09-13 |
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202310793410.5AActiveCN117253618B (en) | 2023-06-30 | 2023-06-30 | Diabetes evaluation model and method based on Monte Carlo Markov method |
| Country | Link |
|---|---|
| CN (1) | CN117253618B (en) |
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20030065669A1 (en)* | 2001-10-03 | 2003-04-03 | Fasttrack Systems, Inc. | Timeline forecasting for clinical trials |
| CN101443780A (en)* | 2004-12-30 | 2009-05-27 | 普罗文蒂斯公司 | Methods, system, and computer program products for developing and using predictive models for predicting a plurality of medical outcomes, for evaluating intervention strategies, and for simultaneously |
| US20110093249A1 (en)* | 2009-10-19 | 2011-04-21 | Theranos, Inc. | Integrated health data capture and analysis system |
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20030065669A1 (en)* | 2001-10-03 | 2003-04-03 | Fasttrack Systems, Inc. | Timeline forecasting for clinical trials |
| CN101443780A (en)* | 2004-12-30 | 2009-05-27 | 普罗文蒂斯公司 | Methods, system, and computer program products for developing and using predictive models for predicting a plurality of medical outcomes, for evaluating intervention strategies, and for simultaneously |
| US20110093249A1 (en)* | 2009-10-19 | 2011-04-21 | Theranos, Inc. | Integrated health data capture and analysis system |
| Title |
|---|
| LU J等: "Effects of continuous glucose monitoring on glycaemic control in type 2 diabetes: a systematic review and network meta‐analysis of randomized controlled trials", 《DIABETES, OBESITY AND METABOLISM》, vol. 26, no. 01, 12 October 2023 (2023-10-12), pages 362 - 372* |
| YUAN S等: "Effectiveness and cost-effectiveness of six GLP-1RAs for treatment of Chinese type 2 diabetes mellitus patients that inadequately controlled on metformin: a micro-simulation model", 《FRONTIERS IN PUBLIC HEALTH》, vol. 11, 6 September 2023 (2023-09-06), pages 1201818* |
| 刘海娇等: "2型糖尿病治疗药物经济学评价模型的分析研究", 《中国药房》, vol. 31, no. 19, 31 October 2020 (2020-10-31), pages 2392 - 2398* |
| 叶秋绵等: "钠-葡萄糖共转运蛋白2抑制剂与二肽基肽酶4抑制剂分别联合二甲双胍治疗2型糖尿病患者的长期药物经济学评价", 《中国药物经济学》, vol. 15, no. 05, 31 May 2020 (2020-05-31), pages 5 - 14* |
| 王俊: "基于Markov模型的糖尿病患者行为干预经济学评价", 《中国优秀硕士学位论文全文数据库 医药卫生科技辑(月刊)》, no. 12, 15 December 2018 (2018-12-15), pages 065 - 63* |
| Publication number | Publication date |
|---|---|
| CN117253618B (en) | 2024-09-13 |
| Publication | Publication Date | Title |
|---|---|---|
| US8756043B2 (en) | Blood glucose meter and computer-implemented method for improving glucose management through modeling of circadian profiles | |
| McEwan et al. | Evaluation of the costs and outcomes from changes in risk factors in type 2 diabetes using the Cardiff stochastic simulation cost-utility model (DiabForecaster) | |
| US20210256872A1 (en) | Devices, systems, and methods for predicting blood glucose levels based on a personalized blood glucose regulation model | |
| US20150379229A1 (en) | Computer-Implemented System And Method For Improving Glucose Management Through Modeling Of Circadian Profiles | |
| Deutsch et al. | Computer-assisted diabetic management: a complex approach | |
| CN110289094A (en) | A decision-making method for precise insulin dosing based on expert rules | |
| Palmer et al. | Cost-effectiveness of switching to biphasic insulin aspart in poorly-controlled type 2 diabetes patients in China | |
| Hu et al. | Long-term cost-effectiveness comparison of catheter ablation and antiarrhythmic drugs in atrial fibrillation treatment using discrete event simulation | |
| CN117766083A (en) | Method for constructing warfarin anticoagulation quality and adverse reaction prediction model based on machine learning algorithm and application device thereof | |
| CN117253618B (en) | Diabetes evaluation model and method based on Monte Carlo Markov method | |
| WO2006129375A1 (en) | Medication support program, medication support apparatus, recording medium recording medication support program, and medication support system | |
| Al Hayek et al. | Improvement of glycemia risk index and continuous glucose monitoring metrics during Ramadan Fasting in type 1 diabetes: a real-world observational study | |
| US20150347708A1 (en) | Blood Glucose Meter And Computer-Implemented Method For Improving Glucose Management Through Modeling Of Circadian Profiles | |
| Nichols et al. | Contemporary analysis of secondary failure of successful sulfonylurea therapy | |
| CN118471504A (en) | A method for constructing a prediction model for thyroid damage after tumor immunotherapy | |
| Vergès et al. | High efficacy of screening for diabetes and prediabetes in cardiac rehabilitation after an acute coronary syndrome (ACS). The REHABDIAB study | |
| CN115985458A (en) | Insulin dosage recommendation system integrating deep learning and medical knowledge | |
| CN114974500A (en) | Intestinal cancer patient nutrition treatment and prognosis prediction evaluation model based on TPN control system | |
| Senthil et al. | Early Prediction of Diabetes and its Risk Factors based on ARIMA-ELMAN ANN Network | |
| Lozhkina et al. | Individualized Therapy Optimization for Type 2 Diabetes | |
| CN118983103B (en) | A diabetes screening data processing management system and method thereof | |
| Gyuk et al. | Diabetes lifestyle support with improved glycemia prediction algorithm | |
| CN120413079B (en) | CRRT (China radio remote control) unscheduled off-line early warning and decision support information system | |
| WO2024261336A1 (en) | Treatment decision support for treatment of type 2 diabetes mellitus | |
| Johnston et al. | Use of regression modeling to simulate patient-specific decision analysis for patients with nonvalvular atrial fibrillation |
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |