A method for establishing a risk prediction model for senile dementia

Abstract

The present invention provides a method for establishing a risk prediction model for the onset of senile dementia, the method comprising: determining a risk factor scoring method; a Cox regression model probability method using the onset probability P of the Cox regression model to represent the risk of individual dementia; and evaluating two prediction models. The area under the ROC curve of the risk factor scoring method is larger than that of the Cox regression model probability method, and the specificity and sensitivity of the risk factor scoring method are also better than those of the Cox regression model probability method. The results suggest that the risk factor scoring method is more effective in predicting the risk of senile dementia.

Description

Technical Field

[0001] The present invention relates to the field of risk prediction, and in particular to a method for establishing a risk prediction model for senile dementia. Background Art

[0002] Alzheimer's disease is a common neurodegenerative disease among the elderly. The progression of the disease varies from person to person. Some patients may develop cognitive impairment due to Alzheimer's disease as the disease progresses, which seriously affects the patient's quality of life and increases the burden on individuals, families and society.

[0003] Mild cognitive impairment in Alzheimer's disease is the early stage of PD-CI. Studies have shown that about 20% of PD patients have MCI symptoms when diagnosed, and 5 years after PD is diagnosed, the symptom ratio rises to 40-50%. PD-MCI is generally considered to be a transitional stage between normal cognitive function and Alzheimer's disease dementia. Therefore, early prediction and intervention of PD-MCI is of great significance to slow down or prevent dementia in patients.

[0004] In the related technologies, there are relatively few studies on PD-CI. Some studies have found that mutations in the GBA, APOE, RIMS2, TMEM108 and WWOX genes are associated with PD-CI, and the polygenic risk scores calculated from mutations in these five gene sites can be used to predict the progression of PDD; other studies have pointed out that the downregulation of cerebrospinal fluid beta amyloid protein (Aβ42) concentration is associated with PD-CI; and the downregulation of peripheral blood synuclein gene (SNCA) expression is associated with PD cognitive deterioration; the upregulation of peripheral blood serum exosome miRNAhas-miRNA-451a is associated with PD-CI. However, there are no reports that point out effective markers for the dynamic prediction of PD-MCI, and use them to detect or predict the occurrence of PD-CI earlier, increasing the effectiveness of treatment. Summary of the invention

[0005] In view of the above problems, the present invention is proposed to provide a method for establishing a risk prediction model for senile dementia that overcomes the above problems or at least partially solves the above problems.

[0006] According to one aspect of the present invention, a method for establishing a risk prediction model for senile dementia is provided, the method comprising:

[0007] Determine the risk factor scoring method;

[0008] The Cox regression model probability method uses the incidence probability P of the Cox regression model to represent the risk of individual dementia;

[0009] Two prediction models were evaluated.

[0010] Optionally, the risk factor scoring method specifically includes:

[0011] The occurrence of dementia was taken as the dependent variable, and the statistically significant risk factors screened out in the multivariate Cox regression analysis were taken as the independent variables;

[0012] The significance of the regression coefficient in the Cox regression analysis output is the relative risk of dementia for each level increase in the independent variable;

[0013] According to this principle, the regression coefficient of each variable is determined as the score of each variable, and the sum of all variable scores is the risk index RI, that is, RI = β1X1 + β2X2 + ... βiXi. The more risk factors an individual has, the higher his RI is, and theoretically the greater his risk of dementia.

[0014] Optionally, the Cox regression model probability method uses the incidence probability P of the Cox regression model to represent the risk of individual dementia incidence, specifically including:

[0015] The Cox regression model probability method uses the incidence probability P of the Cox regression model to represent the risk of individual dementia. The higher the individual's probability of onset, the greater the risk of dementia.

[0016] P=1-ho(t) exp(RI / 100) , where ho(t) represents the risk function at time t when X1=X2=…=Xi=0.

[0017] Optionally, the evaluating the two prediction models specifically includes:

[0018] The effects of the two methods were evaluated using sensitivity, specificity and the area under the ROC curve. The prediction thresholds of the two methods were based on the value when the Youden index was the largest, Youden index = sensitivity + specificity - 1.

[0019] Optionally, the Cox regression model analysis of risk factors for senile dementia specifically includes:

[0020] The PH assumption of the selected variables was tested using the time covariate method;

[0021] All selected variables met the PH assumption.

[0022] The present invention provides a method for establishing a risk prediction model for the onset of senile dementia, the method comprising: determining a risk factor scoring method; a Cox regression model probability method using the onset probability P of the Cox regression model to represent the risk of individual dementia; and evaluating two prediction models. The area under the ROC curve of the risk factor scoring method is larger than that of the Cox regression model probability method, and the specificity and sensitivity of the risk factor scoring method are also better than those of the Cox regression model probability method. The results suggest that the risk factor scoring method is more effective in predicting the risk of senile dementia.

[0023] The above description is only an overview of the technical solution of the present invention. In order to more clearly understand the technical means of the present invention, it can be implemented according to the contents of the specification. In order to make the above and other purposes, features and advantages of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are listed below. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying creative work.

[0025] Figure 1 This is a ROC curve diagram of the RI provided in the embodiment of the present invention for screening the risk of dementia. DETAILED DESCRIPTION

[0026] The exemplary embodiments of the present disclosure will be described in more detail below with reference to the accompanying drawings. Although the exemplary embodiments of the present disclosure are shown in the accompanying drawings, it should be understood that the present disclosure can be implemented in various forms and should not be limited by the embodiments set forth herein. On the contrary, these embodiments are provided to enable a more thorough understanding of the present disclosure and to fully convey the scope of the present disclosure to those skilled in the art.

[0027] The terms "comprises" and "having" and any variations thereof in the description embodiments, claims and drawings of the present invention are intended to cover non-exclusive inclusions, for example, including a series of steps or units.

[0028] The technical solution of the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments.

[0029] like Figure 1 As shown in the figure, the prediction models were established using the risk factor scoring method and the Cox regression model probability method.

[0030] 1) Risk factor scoring method

[0031] Whether dementia occurs is used as the dependent variable (yes = 1, no = 0), and the statistically significant risk factors screened out in the multivariate Cox regression analysis are used as independent variables (P < 0.05 is the significance limit for screening risk factors). The significance of the regression coefficient in the output results of the Cox regression analysis is the relative risk of dementia increased by an individual for each level increase in the independent variable. According to this principle, we determine the regression coefficient of each variable (the regression coefficient is taken to one decimal place) as the score of each variable, and the accumulation of all variable scores is the risk index (RI), that is, RI = β1X1 + β2X2 + ... βiXi. The more risk factors an individual has, the higher his RI is, and theoretically, the greater his risk of dementia is.

[0032] 2) The Cox regression model probability method uses the incidence probability P of the Cox regression model to represent the risk of individual dementia. The higher the individual's incidence probability, the greater the risk of dementia.

[0033] P=1-ho(t) exp(RI / 100) , where ho(t) represents the risk function at time t when X1=X2=…=Xi=0.

[0034] 3) Evaluation of the two prediction models

[0035] The effects of the two methods were evaluated using sensitivity, specificity, and the area under the ROC curve (AUC). The prediction thresholds of the two methods were based on the value at which the Youden index was the largest, Youden index = sensitivity + specificity - 1.

[0036] step:

[0037] 1. Cox regression model analysis of risk factors for senile dementia

[0038] 1) Proportional Hazards (PH) Assumption

[0039] The PH assumption of the selected variables was tested using the time covariate method, and the results are shown in Table 1. After testing, all selected variables met the PH assumption.

[0040] Table 1 PH hypothesis test based on covariate model

[0041]

[0042] 2. Univariate Cox regression analysis of senile dementia

[0043] The assignment methods of each research factor and its dummy variable are shown in Table 2. Univariate analysis showed that old age, female, low education level, widowhood or unmarried, smoking, hypertension, history of stroke, low COGSCORE score, high RELSCORE score, larger waist-to-hip ratio and lack of physical exercise are risk factors for senile dementia, and the statistical test differences are significant.

[0044] Table 2 Risk factors and their value assignment methods

[0045]

[0046] 3. Multivariate Cox regression analysis of senile dementia

[0047] The significant variables in the univariate analysis: gender, education level, marital status, hypertension, history of stroke, COGSCORE score, RELSCORE score, waist-to-hip ratio and lack of physical exercise were included in the multivariate Cox regression model. The analysis results showed that old age (90+, HR=8.03, 95%CI, 2.36-14.23; 70-74, HR=1.06, 95%CI, 0.60-1.88, P<0.001), current smoking (HR=1.98, 95%CI, 1.53-3.12, P=0.007), history of stroke (HR=1.65, 95%CI, 1.19-2.25), larger WHR (HR= 1.82, 95% CI, 1.22-4.51) is a risk factor for senile dementia, while more physical exercise (HR=0.98, 95% CI, 0.96-0.99) and higher COGSCORE score (HR=0.96, 95% CI, 0.93-0.99) are protective factors for senile dementia. The effects of gender, marital status, hypertension and RELSCORE score on the incidence of dementia are no longer statistically significant.

[0048] 5. Risk prediction model for senile dementia

[0049] 1) Risk factor scoring method

[0050] The six significant variables screened out in the above Cox multivariate analysis model were scored, and the significance of the regression coefficient β in the Cox regression analysis output was the increase in dementia for each level increase in the independent variable.

[0051] Relative risk of disease. According to this principle, we determined that the regression coefficient of each variable would be used as the basis for scoring each variable. For the convenience of calculation, 10 times the regression coefficient (the regression coefficient is taken to one decimal place) is used as the score Si of each variable. The more risk factors an individual has, the higher the cumulative score, and theoretically the greater the risk of dementia. Therefore, the cumulative risk score (∑Si) of each individual, that is, the risk index, can be used as a continuous indicator to represent the risk of dementia for an individual. See Table 1-18 for specific value assignments.

[0052] The formula for calculating the individual risk index is: RI = β1X1 + β2X2 + ... βiXi, where β is the regression coefficient, X1, X2, ...Xi are risk factors, and the score of each risk factor in this study is 10 times β. According to the risk factor score, the formula for calculating the RI of senile dementia is as follows: RI = 2 (70-74 years old) or 10 (75-79 years old) or 13 (80-84 years old) or 16 (85-89) or 17 (90 years old and above) + 8 (history of stroke) + 5 (smoking) - 5 (physical exercise) + 32 × waist-hip ratio - 6 × COGSCORE, categorical variables are based on the presence of risk factors.

[0053] If a situation occurs, the corresponding factor score will be added, and if it does not occur, it will be 0; for continuous variables, the factor score and the variable value are directly multiplied.

[0054] The RI score range is -283-129. For the convenience of calculation, 400 is added to the RI value, and the final range is 75-429. According to the RI percentile 35% and 70% grouping, the number of samples in each group is relatively balanced. The 1835 samples were divided into three groups according to the RI score: low, medium, and high risk. The risk of dementia increased with the increase of RI. The trend chi-square test showed significant difference in incidence (P<0.001). The results suggest that the RI index stratifies the risk of dementia in the subjects of this study and screens the high-risk population.

[0055] Figure 1 This is the ROC curve of RI for dementia screening, with an AUC of 0.76, a standard error of 0.02, 95% CI: 0.73-0.80, an AUC area greater than 0.5, and a non-parametric test with statistical significance (P<0.001), suggesting that RI can be used to predict the risk of dementia. When RI=246 is used as the discrimination threshold (when the Youden index is the largest), the screening sensitivity is 72.31% and the specificity is 63.39%.

[0056] 2) Cox regression model probability method

[0057] A Cox regression model was established based on six statistically significant risk factors. The expression of the Cox model is h(t)=h0(t)exp(β1X1+β2X2+…+βiXi, where X1, X2…Xi are covariate influencing factors, h(t) is the hazard function of an individual with covariates X1, X2…Xi at time t, which approximately represents the probability of the individual at time t to develop the disease within a unit period after time t; h0(t) is an unknown function of t, that is, the hazard function at time t when X1=X2=X3=…=Xi=0, which is called the baseline hazard function, and β1, β2…βi are the regression coefficients corresponding to each covariate. The formula for calculating the individual's risk of developing the disease since the baseline (absolute risk) is: P=1-ho(t)exp(RI / 100). The larger the P value, the greater the probability of the risk of developing dementia. As shown in Table 1-20, with the increase of the probability of developing the disease, the incidence of dementia increases (trend chi-square test, P<0.001).

[0058] The ROC curve of the Cox regression model probability method for dementia screening showed that the AUC was 0.73, the standard error was 0.02, 95% CI: 0.69-0.77, the AUC area was greater than 0.5, and the non-parametric test was statistically significant (P<0.001). When RI=0.09 was used as the discrimination threshold (when the Youden index was the largest), the screening sensitivity was 71.62% and the specificity was 61.23%.

[0059] Beneficial effect: In summary, the area under the ROC curve of the risk factor scoring method is larger than that of the Cox regression model probability method, and the specificity and sensitivity of the risk factor scoring method are also better than those of the Cox regression model probability method. This result suggests that the risk factor scoring method is more effective in predicting the risk of senile dementia.

[0060] The above specific implementation methods further illustrate the objectives, technical solutions and beneficial effects of the present invention in detail. It should be understood that the above are only specific implementation methods of the present invention and are not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A method for establishing a risk prediction model for senile dementia, characterized in that: The establishment method comprises: Determine the risk factor scoring method, including: The occurrence of dementia was taken as the dependent variable, and the statistically significant risk factors screened out in the multivariate Cox regression analysis were taken as the independent variables; The significance of the regression coefficient in the Cox regression analysis output is the relative risk of dementia for each level increase in the independent variable; According to this principle, the regression coefficient of each variable is determined as the score of each variable. The sum of all variable scores is the risk index RI, that is, RI = β1X1 + β2X2 + ... βiXi. The more risk factors an individual has, the higher his RI is, and theoretically, the greater his risk of dementia is. RI=β1X1+β2X2+…βiXi, where β is the regression coefficient, X1, X2, …Xi are risk factors, and the score of each risk factor is 10 times of β; According to the risk factor score, the RI calculation formula for senile dementia is as follows: RI = 2 (70-74 years old) or 10 (75-79 years old) or 13 (80-84 years old) or 16 (85-89) or 17 (90 years old and above) + 8 (history of stroke) + 5 (smoking) - 5 (physical exercise) + 32 × waist-hip ratio - 6 × COGSCORE. Categorical variables are scored according to the presence of risk factors. If they appear, the corresponding factor score is added, and if they do not appear, it is 0. For continuous variables, the factor score and the variable value are directly multiplied. The Cox regression model probability method uses the incidence probability P of the Cox regression model to represent the risk of individual dementia, including: The Cox regression model probability method uses the incidence probability P of the Cox regression model to represent the risk of individual dementia. The higher the individual's probability of onset, the greater the risk of dementia. P=1-ho(t) exp(RI / 100) , where ho(t) represents the risk function at time t when X1=X2=…=Xi=0; Two prediction models were evaluated, including; The effects of the two methods were evaluated using sensitivity, specificity and the area under the ROC curve. The prediction thresholds of the two methods were based on the value when the Youden index was the largest, Youden index = sensitivity + specificity - 1.

2. The method for establishing a risk prediction model for senile dementia according to claim 1, characterized in that: The Cox regression model analysis of risk factors for senile dementia specifically includes: The PH assumption of the selected variables was tested using the time covariate method; All selected variables met the PH assumption.

Images