Predicting multi-vascular diseases in patients with coronary artery disease

Introduction

Atherosclerosis underlines some of the major cardiovascular diseases which are the leading cause of mortality worldwide. Due to its systemic nature, atherosclerosis in the coronary arteries may indicate occurrences in other arteries.¹ Coronary arterial disease (CAD) is associated with peripheral artery disease (PAD), carotid artery stenosis (CAS), and abdominal aortic aneurysm (AAA). There is an increasing prevalence of vascular diseases in patients with a more severe CAD, especially high-risk CAD with metabolic syndromes and three-vessel disease.^2–4 Therefore, the screening in patients with CAD is justifiable in assessing the risk of atherosclerosis occurrences in other arteries.

CAD is one of the leading causes of mortality worldwide with a substantial incidence rate. According to the Centers for Disease Control and Prevention (CDC) in the United States, a total of 20.1 million adults aged 20 years old and older have CAD.⁵ Screening for PAD, carotid artery stenosis, and AAA enable early detection, risk stratification, and early cardiovascular treatment; all in favor of reducing morbidity and mortality.^6,7 However, screening program targeting all CAD patients is highly ineffective, costly, and requires multiple resources such as specific instruments and trained technicians. On the other hand, screening for early detection of some vascular diseases, e.g., PAD, in patients with significant CAD was not shown to be more beneficial compared to a routine medical checkup.⁸ Therefore, targeted screening in patients with CAD requires careful consideration based on risk factors for atherosclerosis in other arteries.

Targeted screening contributes to early detection while being more time and cost-effective than general screening; it can be beneficial for diseases previously deemed not necessary to screen. A study specifically assessed targeted screening for AAA was already done involving a small subset of indigenous people in Borneo.⁹ No research has been conducted to assess the predictors for developing multi-vascular diseases, including PAD, CAS, and AAA in patients with CAD. Furthermore, no studies assessed the number needed to screen (NNS) of screening for multi-vascular diseases in CAD patients and the impact of the clinical risk scoring system for said NNS. We hypothesize that a risk-scoring tool to measure the risk of other vascular diseases in CAD patients is feasible to construct and can be applied for targeted screening; in return, the risk-scoring tool can reduce the NNS for asymptomatic multi-vascular diseases. Therefore, this study aims to investigate factors predicting the occurrence of multi-vascular diseases in patients with CAD while constructing a predictive model and scoring system to enable targeted screening for future uses.

Methods

This study was conducted based on the STROBE guideline for observational studies.¹⁰ The study protocol was approved by the National Cardiovascular Center Harapan Kita Hospital committee of ethics (ethics approval number: LB.02.01/VII/509/KEP005/2021). All patients gave written informed consent prior to the recruitment of the study and were free to decline participation.

Results

A total of 1314 patients with CAD were identified; 203 (15.4%) patients have multi-vascular disease. All patients had complete medical record data and, therefore, no missing data in this study. Sociodemographic and clinical data are shown in Table 1. Amongst the variables in patient’s demographics, there was a significant difference in the proportion of patients with cerebrovascular disease, CAD three-vessel disease (CAD3VD), and CAD left main disease (CAD-LM). The prevalence of PAD, CAS, and AAA in patients with CAD were 143 (10.9%), 59 (4.5%), and 19 (1.4%), respectively. The overlap between vascular diseases is shown in Figure 1.

Figure 1. Vascular disease prevalence in CAD.

There is a difference in prevalence of vascular diseases in CAD patients with one, two, and three-vessel disease (Table 2).

Univariate analysis demonstrated individuals with older age, diabetes mellitus, cerebrovascular disease, CAD3VD, and left main disease were more likely to have multivascular disease (Table 3). After a multivariate analysis, four variables were retained to form the final clinical model: ages of ≥60 years (OR: 1.579; 95% CI: 1.153-2.164), diabetes mellitus (OR: 1.412; 95% CI: 1.036-1.924), cerebrovascular disease (OR: 3.656; 95% CI: 2.326-5.747), and CAD3VD (OR: 1.960; 95% CI: 1.250-3.073). All of the predictors remain significant after bootstrap internal validation (Table 3).

The area under the curve of the model was 0.659 (95% CI: 0.617-0.700) and was well calibrated (Hosmer-Lemeshow test p=0.379;). Using bootstrap validation, the optimism-corrected area under the curve was 0.653 (95% CI: 0.610-0.695), which represents the predictive ability of the model (Figure 2).

Figure 2. Logistic regression model’s validation (A) Discrimination using Receiver-operator Curve (ROC) analysis; (B) Calibration plot analysis.

Those four variables were assigned weighted points scores for the final clinical prediction of multi-vascular disease based on the magnitude of the β coefficient (Table 4). We designated a score of 0–2 as low probability (9% of chance), 3–5 as moderate probability (21%), and 6–8 as high probability (42%). The discriminatory ability of the scoring system was moderate (AUC: 0.649; 95% CI: 0.610-0.687).

Based on our data, targeted screening for patients with moderate risk or higher decreases the number needed to screen (NNS) from six to five while targeted screening for patients with high risk reduces the NNS from six to three. Targeted screening for patients with moderate risk or higher had the most balanced results of diagnostic values with positive predictive value (PPV), negative predictive value (NPV), sensitivity, and specificity of 45,1%, 86,2%, 15,8%, and 96,5% respectively.

Discussion

This is the first study in Indonesia, to investigate factors predicting the occurrence of multi-vascular diseases in patients with CAD. Our findings show that three vascular diseases were all encountered in CAD patients namely; PAD with prevalence of 10.9%, CAS with prevalence of 4.5% and lastly AAA with prevalence of 1.4%. Our final model to predict multi-vascular disease were as the following: older age (≥ 60 years) [OR = 1.58 (1.15 – 2.16; p = 0.04], diabetes mellitus [OR = 1.41 (1.04 – 1.92); p = 0.029], cerebrovascular disease [OR = 3.66 (2.32 – 5.74); p < 0.001] and CAD3VD [OR = 1.96 (1.25 – 3.07; p = 003]. Furthermore, we constructed the scoring system as well, consisting of age ≥ 60 years old (1 point), diabetes mellitus (1 point), cerebrovascular disease (4 point) and CAD-3 vessels (2 points). Remarkably, the Hosmer-Lemeshow test with a p value of 0.38 demonstrated that the calibration of this final model was highly accepted, and in addition, utilising bootstrap validation, the AUC was 0.65 (95% CI: 0.61 – 0.69) reflecting the model’s capacity to make a sufficiently accurate discrimination.

The aforementioned predictors seem to be less prevalent than that in other studies, carried out elsewhere. In a study by Poredos et al. (2007), screening with the ankle-brachial index (ABI) revealed that 42% of CAD patients also suffered PAD.² Tanimoto et al. (2015) found that CAS prevalence in patients with CAD ranged from 7.0% to 36.0%.³ Durieux et al. (2014) demonstrated that 2.5% to 14.4% of all CAD also had asymptomatic AAA, hence confirming the presence of AAA in CAD.⁴ The lower prevalence of the aforementioned predictors in this findings might be due to the younger patients recruited in our study, compared to others. It is acknowledged that the incidence of PAD, CAS and AAA increases along with age.^17–19 The shift of peak prevalence towards younger subjects in Indonesia might be caused by the higher frequency of CAD risk factors encountered in younger generation. This is supported by the study of Hussain et al. (2016), which demonstrated a higher population attributable risk proportion (PAR) percentages of smoking habits, hypertension, excess body weight, diabetes mellitus and hypercholesterolemia in CAD subjects less than 55 years old compared to older population.²⁰

The association between the severity of CAD and the extent of atherosclerosis was also observed in other vasculatures. The extent of fat deposit along coronary vasculatures is proportional to the severity of atherosclerosis, as demonstrated by the increased expression of interleukin-6 (IL-6) and leptin and the decreased expression of serum adiponectin.²¹ Our data showed no significant increased AAA prevalence in relation with CAD severities. This might be due to a small prevalence of AAA, in which affected to no statistical significance.

Given the final model in our study, Honda T, et al in the Hisayama study (2022) conducted a prospective observational study and constructed a risk prediction model for the development of ASCVD in Japanese adults. They found that age, sex, systolic blood pressure, diabetes mellitus, proteinuria, smoking habit and regular exercise are all predictors of ASCVD occurrence.²² The Hisayama study, in contrast to ours, used Japanese people who had no prior history of CVD. Consequently, they conducted a prospective observational study and followed the subjects for 24 years.

In order to construct risk prediction model, risk scoring was made from β-coefficient obtained from the multivariate analysis. The calibration of the final model fulfilled the goodness of fit, utilizing the Hosmer-Lemeshow test. Determinant capability of risk scoring was shown to be moderately good with AUC of 65.9%. Likewise, this determinant was quite comparable with another risk prediction model in those with stable CAD, carried out by Badheka et al (2011). They established the predictors such as: history of hypertension, smoking, age and history of diabetes with AUC of 68.6%.²³ Additionally, our study demonstrated that general screening of all CAD patients resulted in 15.4% cases with multi-vascular disease, and number needed to screen (NNS) was 6. Meanwhile, targeted screening for patients with high-risk based on our own risk scoring would reduce the NNS from 6 to 3 identifying 45% CAD patients suffered other vascular disease. Consequently, our clinical risk scoring proved to have a very high specificity of 96.5%. Although choosing an NNS of 6 is arbitrary, but the decrement of NNS value from 6 to 3 was quite impressive.

Conclusion

Patients with CAD who have diabetes mellitus, cerebrovascular disease, CAD3VD, and above 60 years old are associated with increased odds of multi-vascular disease. By using risk scoring tool made from these risk factors, targeted screening in high-risk patients decrease the number needed to screen in half with high specificity.

Author contributions

Suko Adiarto: Conceptualization, Methodology, Writing – Original Draft, Writing – Review & Editing, Data Curation, Visualization. Luthfian Aby Nurachman: Formal Analysis, Writing – Original Draft. Raditya Dewangga: Formal Analysis, Investigation, Suci Indriani: Investigation, Methodology Taofan: Investigation, Methodology Amir Aziz Alkatiri: Investigation, Methodology Doni Firman: Investigation, Methodology Anwar Santoso: Conceptualization, Validation, Writing – Review & Editing, Supervision, Funding Acquisition All authors read and approved the final paper.