Introduction
A report from the national cardiovascular data registry of the American College of Cardiology showed that only 41% of patients undergoing elective coronary angiography (CAG) are diagnosed as having obstructive coronary artery disease (OCAD); hence, a better risk stratification to increase pretest probability of coronary artery disease (CAD) appears warranted [
1]. A recent meta-analysis (33 studies, 120,548 participants) suggested that OCAD in patients with acute coronary syndrome has a significantly higher cardiovascular risk at baseline and a higher likelihood of death or major cardiovascular events [
2].
The classic risk stratification tool for CAD was the Framingham score system, which predicts the 10-year risk of coronary heart disease; however, the association of Framingham score with plaque burden is less robust [
3,
4]. Although non-invasive diagnostic technology advancements, such as stress testing and computed tomography (CT) scanning adopted to increase the pretest probability of CAD in most tertiary hospitals, are available, high costs and unavailability limit their application in daily clinical practice. Ibrahim et al. recently established a new clinical and biomarker score with high accuracy for predicting the presence of anatomically significant CAD (≥70% stenosis), which included clinical variables (male sex and previous percutaneous coronary intervention (PCI)) and four biomarkers (midkine, adiponectin, apolipoprotein C-I, and kidney injury molecule-1), among patients with known CAD (e.g., patients with previous acute myocardial infarctions (MI), who had PCI, or who underwent coronary artery bypass grafting (CABG)). However, whether this model could predict CAD in patients presenting at primary-level hospitals or clinics is unknown [
5].
Therefore, this study aimed to establish a new simple prediction model, including traditional Framingham risk factors for OCAD, based on a continuously recruited at-risk cohort from an observational database investigating acute kidney injuries following elective CAG [
6]. We hypothesized that the addition of new contemporary predictors to traditional Framingham risk factors could increase the accuracy of predicting anatomically significant CAD and, consequently, the novel model could be used in a broader population with suspected CAD.
Discussion
In this study, Framingham risk score showed lower discrimination for OCAD prevalence in patients without known CAD undergoing elective CAG, whereas our new simple MFS model showed better performance in estimating the pretest probability of OCAD and identified more than two thirds of patients at high risk for OCAD.
Framingham risk score including traditional risk factors, which is the classic CAD risk-prediction model, showed less association with OCAD in the present study, which is consistent with a previous Canadian study that investigated the association of Framingham risk score with computed tomography angiography (CTA) measures of coronary atherosclerosis. Coronary atherosclerosis was present in 63.5% of the patients, which suggested a high prevalence similar to that in our study (74.9% of OCAD). Nevertheless, OCAD was diagnosed using CAG in our study, which may have more accuracy and significance for clinical decision than CTA [
4].
Our simple modified Framingham risk score (FRS) might outperform FRS by ROC analysis, while we could not further calculate the net reclassification improvement to show how much improvement by using modified FRS to predict OCAD. Because FRS was to predict the 10-year risk of CAD, whose endpoint rate was lower than 15%, however, the modified FRS was to predict the OCAD with higher rate (more than 50%) for patients with suspected CAD, we could not find out ideal overlapping parts, ever after adopting all the cut-off value of endpoint rate [
3].
This study established a novel, simple risk stratification method, which could be a useful tool in most primary-level hospitals or clinics and requires no large equipment or expensive examination in identifying patients with high risk for OCAD. Genders et al. established prediction models with a more accurate estimation of the pretest probability of OCAD in lower prevalence populations than that of the Duke clinical score, which is recommended by an American College of Cardiology guideline. However, to improve the model, coronary calcium score by CT scanning and classification of chest pain symptoms (i.e., typical, atypical, or non-specific) are necessary, which requires medical knowledge and experience or equipment and entails costs [
16,
17].
Moreover, our results indicated that the use of MFS model suggested the yield of testing (i.e., the proportion of individuals referred for testing who have abnormal results) among patients with a high pretest probability [
18]. More expensive or limited diagnostic tests in the risk stratification in a large community population can be addressed by the MFS model. In other word, the simple MFS model will improve the precision of risk stratification for more physicians, to increase invasive procedure among high risk patients, but to reduce invasive procedure among low risk patients.
For OCAD diagnosis, the modified Framingham score had a final area under the ROC curve of 0.703 in the validation set. Although the area under the ROC curves was less than that of previous studies, the performance of the MFS model was good [
5,
18]. The small area under the ROC curve in our study could be attributed to the following: first, our study lacked more novel biomarkers, which were selected from nearly one hundred candidate variables [
5]; second, the sample size was smaller than that of previous studies [
16,
18]; third, we excluded subgroups with essential variables (e.g., PCI) or missed some significant variables (e.g., types of chest pain) [
16].
One recent novel prediction score model in America for significant CAD showed better performance (c-statistic, 0.87 in the validation set); at the optimal cut-point, the score was both highly sensitive (77%) and specific (84%) for CAD diagnosis. The model included four biomarkers (midkine, adiponectin, apolipoprotein C-I, and kidney injury molecule-1). However, including these biomarkers in the routine examination in primary-level hospitals or clinics may not be suitable. In addition, the model was established in patients with or without known coronary disease and included previous PCI as a predictor. Our simple model includes traditional risk factors or routine examination without high cost. The development setting was also different, i.e., our simple model was based mostly on Chinese individuals; in the American model, Whites.
The modified Framingham score has three additional variables (i.e., hs-CRP, LVEF, and anemia), which could be associated with both anatomically OCAD and coronary events. A previous meta-analysis including 160,309 people without a history of vascular disease suggested that CRP concentration has continuous associations with the risk of CAD, ischemic stroke, vascular mortality, and death from several cancers and lung diseases that are each of broadly similar size [
19]. A Japanese observational study with a median follow-up period of 6.5 years showed that hs-CRP was associated with higher incidence of major adverse cardiac events or all-cause mortality in patients with established CAD and undergoing PCI [
20]. The Atherosclerosis Risk in Communities (ARIC) study, which included 14,410 subjects (between 45 and 64 years) without CVD and had a follow-up duration of 6.1 years, showed that anemia is an independent risk factor for CVD outcomes [
21]. Another recent cohort study of outpatients with stable CAD (21,829 with baseline hemoglobin levels) showed that anemia is a powerful predictor of cardiovascular and non-cardiovascular mortality [
22]. Our study may be the first to identify anemia as a CAD predictor. A prospective study including 100 diagnostic coronary catheterization candidates found that the overall accuracy of akinesia/hypokinesia and LVEF < 55% in predicting abnormal CAG (≥50% stenosis) was poor [
23]. Another study on 182 patients undergoing exercise Tl-201 gated single-photon emission computed tomography suggested that worsening of the LVEF during exercise has the potential to detect multivessel CAD among patients without multivessel patterns of reversible defects [
24]. In our study, the LVEF measured by average echo could also be a CAD predictor. The predictive value of two new simple predictors (LVEF and anemia) needs further external validation in larger studies. Similarly, the new simple score model also requires further evaluation in relation to risk stratification. In addition, we would investigate the risk factors of dangerous culprit lesions with culprit plaque rupture (CPR) and thin-cap fibro-atheroma (TCFA), such as hypertension, advanced ages, diabetes mellitus or hyperlipidemia, which were evaluated by optical coherence tomography (OCT) or intravenous ultrasound (IVUS) in the futures [
25].
Limitations
Our study has several limitations. First, our study adds significant information to the current literature on pretest probability of OCAD for stable patients without known CAD; however, this is a single-center prospective observational study with a limited sample size. Second, angina types (e.g., classification of chest pain symptoms: typical, atypical, or non-specific) were not considered; thus, the accuracy of the OCAD risk model may have been affected. Thirdly, some bias in patient selection especially in CAG possibly existed; cardiologists tended to recruit patients with more baseline risk factors but with a stable condition for elective CAG. Thus, our findings may not be applicable to patients with low risk for OCAD or with emergent conditions. Fourthly, our study only focused on the anatomical result of angiography without kind of plaque evaluated with imaging (OCTs or IVUS), such as culprit plaque rupture (CPR) and thin-cap fibro-atheroma (TCFA), which could trigger acute coronary syndrome (ACS). Lastly, the high loss of follow-up rate possibly affected the quality of long-term prognosis and the predictive value of the modified Framingham score.