Study participants
Between April 2018 and June 2020, 826 adults with elevated office blood pressure were recruited as participants in Daping Hospital. Participants were further screened through the following two exclusion criteria: (1) antihypertensive drug users; (2) renal insufficiency (Scr ≥ 133 mol/L) [18]. The study was reviewed and approved by the Ethics Committee of Daping Hospital of the Army Medical University and registered in the Chinese Clinical Trial Registry with the registration number ChiCTR1800015507. All participants gave informed consent to the study.
General clinical Information
The age, sex, smoking history, drinking history, diabetes history, and cardiovascular and cerebrovascular disease history of the participants were collected through questionnaires. The height and weight of the selected participants were measured, and the body mass index (BMI) was calculated. The calculation formula was (BMI) = weight (kg) ÷ height 2 (m) [19]
Blood pressure measurement
After the participants took a seat rest in the clinic for 20 minutes, the blood pressure of the right brachial artery was measured 3 times by professional medical personnel with mercury sphygmomanometer. The average value of blood pressure measurement was taken as office blood pressure measurement result [20].
24-hour ambulatory blood pressure monitoring was carried out by ambulatory ECG blood pressure recorder CB-2301-A (Wuxi, China), with 6: 00-22: 00 as daytime blood pressure and 22: 00-6: 00 as nighttime blood pressure. Daytime blood pressure measurement is conducted every 30 minutes, the number of effective sphygmomanometers should be above 80%, and nighttime blood pressure requires effective blood pressure every hour.
WCH and SHT were distinguished according to office blood pressure and 24-hour ambulatory blood pressure. European Hypertension Practice Guideline Standard (2014 Edition) was adopted, and the blood pressure rise boundary point was set at office blood pressure ≥ 140/90 mmHg. The average ambulatory blood pressure was ≥ 135/85 mmHg during the daytime, ≥ 120/70 mmHg during the nighttime, and ≥ 130/80 mmHg during the whole day [21, 22]. SHT was defined as elevated office blood pressure and ambulatory blood pressure. While, WCH only showed elevated office blood pressure [23].
Biochemical detection
The total cholesterol (TC), triglyceride (TG), high-density lipoprotein cholesterol (HDL-C), low-density lipoprotein cholesterol (LDL-C), fasting blood glucose (Glu), serum creatinine (Scr) and uric acid (UA) were measured by BECKMAN AU5800 biochemical analyzer. Diabetes mellitus (DM) is defined as the previous definite diagnosis of diabetes mellitus, or the fasting blood glucose detected this time is ≥ 7.0 mmol/L [24].
Division of training set and verification set
Using RStudio software (Version 1.3. 959), the participants were divided into the training set and the verification set, and the ratio of training set to verification set was 4:1. The statistical differences of age, sex, WCH prevalence rate, and other research variables between the training set and the verification set were compared. The differences of research variables between WCH patients and SHT patients in training set and verification set were also analyzed respectively. The counting data is expressed as the number of cases (%), and the Pearson chi-square test is used for statistical analysis. The measurement data are expressed as median (IQR, interquartile range), and the statistical analysis method adopts Kruskal-Wallis rank-sum test [25, 26]. The above statistical analysis was performed by SPSS (Version 22.0) software.
Model construction
LASSO regression and univariate logistic regression were used to select the variables of the model. The study variables involved in the screening process included sex, age, body mass index (BMI), smoking history, drinking history, isolated systolic hypertension (ISH), systolic blood pressure (SBP), diastolic blood pressure (DBP), heart rate (HR), diabetes mellitus (DM), total cholesterol (TC), triglyceride (TG), high-density lipoprotein cholesterol (HDL-C), low-density lipoprotein cholesterol (LDL-C), creatinine (Scr), uric acid (UA) and cardiovascular and cerebrovascular diseases (CCVD). Firstly, the least absolute shrinkage and selection operator (LASSO) regression model was constructed in the training set. LASSO regression algorithm, which is suitable for the regression of high-dimensional data and interactive data analysis, was conducted by 10-fold cross-validation with penalty parameter tuning based on minimum criteria and 1 standard error of the minimum criteria (the 1-SE criteria) in the training set. Since increasing the number of independent variables of the model cannot significantly improve the performance of the model after the value reaches a certain value, the 1-SE criteria can give the model with excellent performance but the least number of independent variables, so we generally adopt this criteria [27]. We further carried out univariate logistic regression analysis on the research variables corresponding to the 1-SE criteria, and included the research variables with statistical differences as the final model construction variables. Multivariate logistic regression analysis was used to construct the scoring model of WCH, and the corresponding nomogram was drawn [28]. The R language packets used mainly include "readxl", "glmnet", "informationValue", "rms", "pROC" and so on.
Discrimination test
Firstly, the receiver operating curve (ROC) was used in the training set to test the discrimination degree of the scoring model in the training set, and the coefficient of determination (R2), the area under the ROC curve (AUC) and its 95% confidence interval (95% CI) are obtained. Furthermore, after applying the scoring model to the verification set, the ROC curve was used to evaluate the discrimination degree of the verification set model, and whether the discrimination degree was similar to the training set model was observed [29].
Calibration degree test
Firstly, the Bootstrap method was used to randomly select samples 1000 times in the training set to verify the calibration degree of the model. The corresponding Calibration curves are made, and the mean square error (MSE) and mean absolute error (MAE) values are calculated to evaluate the model. The lower the MAE and MSE, the better the stability of the model [30]. Furthermore, the calibration degree of the model in the verification set is also checked by the same method.
Expand application
European Hypertension Practice Guidelines (2018 Edition) recommend ambulatory blood pressure monitoring for patients with grade 1 hypertension (SBP of 140-159 mmHg and/or DBP of 90-99 mmHg) or hypertensive patients without target organ damage to screen WCH [31]. Then this study took patients with grade 1 hypertension and hypertensive patients without CCVD as the subjects for ambulatory blood pressure examination according to the guidelines, and the corresponding method was called the hypertension guideline method (Method 1). Based on our scoring model, we use participants with WCH risk higher than 0.2 as screening subjects for ambulatory blood pressure monitoring, and the corresponding method is considered as the scoring model method (Method 2). We further compared the difference between the sample participation rate and WCH missed diagnosis rate between the two methods. The formula for calculating the participation rate and missed diagnosis rate is as follows.