Development of a risk score model for the prediction of patients needing percutaneous coronary intervention

Abstract Background The incidence of coronary heart disease (CHD) is increasing worldwide. The need for percutaneous coronary intervention (PCI) is determined by coronary angiography (CAG). As coronary angiography is an invasive and risky test for patients, it will be of great help to develop a predicting model for the assessment of the probability of PCI in patients with CHD using the test indexes and clinical characteristics. Methods A total of 454 patients with CHD were admitted to the cardiovascular medicine department of a hospital from January 2016 to December 2021, including 286 patients who underwent CAG and were treated with PCI, and 168 patients who only underwent CAG to confirm the diagnosis of CHD were set as the control group. Clinical data and laboratory indexes were collected. According to the clinical symptoms and the examination signs, the patients in the PCI therapy group were further split into three subgroups: chronic coronary syndrome (CCS), unstable angina pectoris (UAP), and acute myocardial infarction (AMI). The significant indicators were extracted by comparing the differences among the groups. A nomogram was drawn based on the logistic regression model, and predicted probabilities were performed using R software (version 4.1.3). Results Twelve risk factors were selected by regression analysis; the nomogram was successfully constructed to predict the probability of needing PCI in patients with CHD. The calibration curve shows that the predicted probability is in good agreement with the actual probability (C‐index = 0.84, 95% CI = 0.79–0.89). According to the results of the fitted model, the ROC curve was plotted, and the area under the curve was 0.801. Among the three subgroups of the treatment group, 17 indexes were statistically different, and the results of the univariable and multivariable logistic regression analysis revealed that cTnI and ALB were the two most important independent impact factors. Conclusion cTnI and ALB are independent factors for the classification of CHD. A nomogram with 12 risk factors can be used to predict the probability of requiring PCI in patients with suspected CHD, which provided a favorable and discriminative model for clinical diagnosis and treatment.


| INTRODUC TI ON
Coronary heart disease (CHD), one of the leading causes of death worldwide, is defined as atherosclerosis, thromboembolism, or spasm of the blood vessels supplying blood and oxygen to the myocardial cells, causing luminal narrowing or even occlusion, which in turn leads to myocardial ischemia and hypoxia or even necrosis. 1 The incidence of CHD has been increasing recently and has been getting more common among young individuals. According to the China Cardiovascular Health and Disease Report 2020, cardiovascular disease deaths accounted for the first cause of total deaths among urban and rural residents in China in 2018, with 330 million cardiovascular patients, including about 11.39 million with CHD.
The prevalence and mortality of CHD are continuously rising, and seriously threaten human health. 2 The fast-paced life, intense work pressure, dietary habits, obesity, and erratic lifestyle are closely related to CHD. [3][4][5][6] Pharmacologic therapy is fundamental for the stabilization or abatement of coronary atherosclerotic plaques and preventing diseases such as coronary thrombosis, acute myocardial infarction, and sudden cardiac arrest. Revascularization by percutaneous coronary intervention (PCI) or coronary artery bypass grafting (CABG) can further reduce angina, improve life quality, and increase infarct-free survival. 7 However, many studies have shown that in most patients with stable angina, the benefits of coronary revascularization are limited to improving the quality of life rather than reducing cardiovascular events. 8 Coronary angiography is a gold standard for evaluating the degree of coronary artery stenosis in coronary atherosclerotic heart disease. However, it is a risky and invasive procedure for patients.
Therefore, in this study, we reviewed the clinical characteristics and laboratory indexes of 454 patients, analyzed the relevant impact factors, and developed a model to predict whether PCI is required, hence exploring a new model to identify CHD requiring PCI economically and feasibly, which will reduce unnecessary medical consumables and alleviate the burden on the healthcare system. ESC Guidelines for the diagnosis and management of chronic coronary syndromes served as the foundation for the grouping criteria. 9 The inclusion criteria of current study are CDH diagnosed by CAG, the criteria for intervention with PCI were performed according to the Chinese Guidelines for Percutaneous Coronary Intervention (2016): I. Stable coronary heart disease (SCAD) should be based on the degree of coronary stenosis as a decision basis for whether to intervene, with direct intervention when the lumen stenosis is ≥90%;

| Study design and participants selection
when the lumen stenosis is <90%, only corresponding intervention if there is evidence of ischemia or a flow reserve fraction (FFR) ≤0.8. II.
Patients with non-ST-segment elevation acute coronary syndrome

| Data collection
General clinical data of all patients were recorded, including age, sex, history of hypertension, history of diabetes mellitus, smoking history, and alcohol consumption. After admission, patients underwent echocardiography, and their ejection fraction was recorded.
Laboratory tests of routine blood, biochemistry, glycosylated hemoglobin, cardiac markers, and coagulation function were recorded and compiled.

| Test method
Routine blood tests were performed using the Sysmex XN-9000 hematology analyzer. Biochemical indexes were detected using the Siemens ADVIA2400 biochemical analyzer with reagents from Ningbo PREB Biotechnology Company. Glycosylated hemoglobin was detected using the Bio-Rad D-100 instrument. High-sensitivity troponin-I, creatine kinase, and creatine kinase isoenzyme were detected by Johnson & Johnson VITRO5600 dry biochemical immunoassay machine. The coagulation function was tested using Wolfen ACL-TOP700 automatic hemagglutination instrument.

| Statistical analysis
SPSS 25.0 statistical software was used for data processing and analysis. The measurement data conformed to normal distribution or approximately normal distribution were expressed as x ± s, and the independent sample t test was used for comparison, Analysis of variance (ANOVA) was used for comparison among the three subgroups, and the chi-square test was used before ANOVA, and the F test was used when the variance was the same, and the Welch test was used when the variance was different; skewed distribution was expressed as M (Q R ), and rank-sum test was used for comparison; count data were expressed as the number of cases, and chi-square test was used for comparison between groups. Univariate and multifactorial analyses were performed to analyze the differences of each index in each group of patients, and then the statistically significant indexes in the analysis results as well as other meaningful indexes were included in the logistic regression analysis, the fitted model was obtained after stepwise regression using R4.1.3 software, and 12 predictors were obtained and plotted in nomogram. The probability of patients with suspected CHD requiring PCI could be obtained by the nomogram scores. p < 0.05 was considered a statistically significant difference.

| Comparison of general clinical data between the PCI treatment group and the control group
Male and Diabetic patients account for a higher proportion of PCI cases. There was no statistically significant difference in age, smoking history, drinking history, and hypertension history between the two groups (all p values >0.05). As shown in Table 1, the proportion of male and diabetic patients was significantly higher in the PCI treatment group than in the control group (all p values <0.05).

| Comparison of testing indexes between two groups of patients
Twenty-four testing indexes show significant difference between the treated and control groups before coronary angiography. The examination items and laboratory indices from patients without missing indices. The statistically significant differences were found in ejection fraction, creatine kinase, creatine kinase isoenzyme, high-sensitivity troponin I, glucose, glycosylated hemoglobin, fructosamine, alanine aminotransferase, aspartate aminotransferase, lipoprotein a, total bile acids, lactate dehydrogenase, high-sensitive C-reactive protein, white blood cell count, absolute neutrophil count, absolute monocyte count, fibrinogen, normal prothrombin time, red blood cell count, ferritin, unconjugated iron, potassium, and creatinine (all p values <0.05). For the data high-sensitivity troponin I levels, a skewed distribution and string data types exist, so the data were converted to a categorical variable, that is, 0-0.034 was normal and >0.034 was high, as shown in Table 2.

| Comparison of clinical data of the three subgroups of the treatment group
Among the three subgroups (CCS, UAP, and AMI), gender and age showed statistically significant difference between the CCS and AMI groups (p < 0.05), with lower proportion of males in the CCS group and younger patients in the AMI group ( Figure 1). This result indicates that older patients are more likely to have CCS and younger patients are more likely to have AMI.

| Comparison of laboratory indicators in the three subgroups of the treatment group
The laboratory indicators such as ejection fraction, creatine kinase, creatine kinase-MB isoenzyme, troponin I, alanine aminotransferase, aspartate aminotransferase, lactic dehydrogenase, hypersensitive-C-reactive protein, leucocyte count, neutrophil count percentage, LDL cholesterol levels, apolipoprotein B, fasting glucose, ferritin, and albumin showed statistically significant among the three subgroups (p < 0.05, Figure 2). Ejection fraction and serum albumin were lower in the AMI group than CCS group, and all other test indexes were higher in the AMI group than in CCS and UAP.

| Univariate and multifactorial binary logistic regression analysis
The indicators with significant differences among the aforementioned indicators were used as independent variables. Whether PCI was performed was set as the dependent variable. With univariate logistic regression analysis, previously proved significant indicators showing the severity of CHD but were found insignificant here, such as smoking, alcohol consumption, hypertension, homocysteine, and low-density lipoprotein, were included in the binary logistic regress.  (Figure 3). Among them, the history of diabetes mellitus, high-sensitivity troponin I, fructosamine, glucose, absolute neutrophil value, and fibrinogen may be independent influencing factors, that is, p < 0.05.

| Univariate and multifactorial multivariate logistic regression analysis
The indicators with significant differences in the univariate analysis were subjected to univariate logistic regression analysis and multivariate logistic regression analysis, and the results are shown in Table 3. The model fit was good at p < 0.05, and the comparison between the CCS and AMI groups showed significance between the two groups in age, cTnI, and ALB.

| Nomogram drawing and verification
Based on the binary logistic regression analysis model, a nomogram was drawn (Figure 4), and the scores of each index could be obtained and summed to obtain the p-value corresponding to the total score, which is the predicted probability of needing PCI in patients with suspected CHD. The calibration curve ( Figure 5) was found to be in general agreement with the reference curve, indicating that the predicted probability of occurrence was in good agreement with the actual probability of occurrence. The uncorrected C-index was 0.84 (95% CI: 0.79, 0.89) and the corrected C-index was 0.80, indicating TA B L E 2 Test indicators with statistically significant differences between the two groups of patients (n = 454).

| DISCUSS ION
Coronary heart disease is a common cardiovascular disease, and the initial diagnosis of CHD can be determined by clinical symptoms, electrocardiogram, and biomarker of myocardial injury.
However, CAG is required to confirm the diagnosis of CHD and the degree of coronary stenosis to determine whether the patient needs to undergo PCI. As known, CAG is an invasive procedure, and patients may be allergic to the contrast agent, which will cause severe hypersensitivity reactions and can be life-threatening. 11 Many scholars have explored the influencing factors related to the degree of stenosis in patients with CHD, suggesting that the history of diabetes, blood glucose level, white blood cell count, serum troponin, fibrinogen level, fibrinogen to albumin ratio, BNP level, total bilirubin, and uric acid level were independent factors on the severity of coronary artery lesions. [12][13][14][15][16][17][18] However, to date, there is still a lack of non-invasive and efficient methods to predict the need for PCI in patients with coronary artery disease. In this study, Diabetes is clinically referred to as an equivalent risk for CHD, indicating that diabetes plays an important role in the development of cardiovascular disease. 19 Numerous studies have confirmed that diabetes mellitus is an important risk factor for CHD. First, elevated blood glucose is a major determinant of arterial stiffness, and chronic hyperglycemia is associated with the accumulation of advanced glycosylation end products (AGEs), which contribute to atherosclerosis and increase the severity of coronary artery lesions.
Secondly, oxidative stress exacerbates macrovascular damage in diabetic patients, namely by inducing the production of reactive oxygen species (ROS), which subsequently damages the endothelial system. 11 Hyperglycemia in diabetic patients also promotes protein kinase C activation and diacylglycerol production, both accelerate the development of atherosclerosis by promoting inflammatory mediators and smooth muscle cell recruitment. It has been shown that the risk of developing CHD in diabetes is highest at any age and is mainly associated with insulin resistance, type 2 diabetes, and metabolic syndrome. 20 The results of this study also suggest that the history of diabetes and fasting glucose levels, fructosamine are good predictors for PCI in patients with CHD and are the main influencing factors of the severity of coronary artery disease.
Inflammatory responses occur throughout the pathophysiology of the development of coronary atherosclerotic heart disease.  are risk factors for CHD, but in this study no statistically significant differences were found between these indicators in patients who underwent PCI or not. We think that these indicators may play an important role in the process of triggering CHD but have little significance for whether to perform PCI; it may also be caused by the relatively small sample size and geographically base of this study, which may generate a possibility of bias, and in the future, we need to enlarge the sample size with multiple centers, and use a machine learning models to improve the prediction efficiency to further improve the predictive ability of PCI for CHD.

AUTH O R CO NTR I B UTI O N S
XW, YPL, and FW jointly designed this study and reviewed and revised the article. XW and YPL collected clinical data from CHD patients. XW and FW further collated and preliminarily analyzed the data and conducted statistical analysis and drew the figures and tables of the whole article. XW and YPL wrote the results section of the article, while XW wrote the rest of the article. All authors read and approved the final article.

ACK N OWLED G M ENTS
We would like to thank the Department of Laboratory Medicine

FU N D I N G S TATEM ENT
No funding.

CO N FLI C T O F I NTE R E S T S TATE M E NT
The authors declare that they have no competing interests.

DATA AVA I L A B I L I T Y S TAT E M E N T
The data that support the findings of this study are available on request from the corresponding author. The data are not publicly available due to privacy or ethical restrictions.

CO N S E NT FO R PU B LI C ATI O N
All the participants gave consent for direct quotes from their interviews to be published in this manuscript.