Risk factors for carotid plaque in type 2 diabetes mellitus: The need for more extensive data

Abstract

The risk factors for carotid plaque in type 2 diabetes mellitus (T2DM) patients are still not well-established. A recent study offers a scientific basis for integrated cerebrovascular risk assessment and early intervention in T2DM patients. However, as a single-center study involving a relatively homogenous Chinese population, the generalizability of its conclusions to other ethnic groups remains limited. Due to the lack of data on diabetes duration, medication history, and lifestyle information, the causal relationship between risk factors and plaque development cannot be determined. Future research should focus on improving research design and conducting comprehensive data collection to elucidate the vascular risk of T2DM patients. Equally important is the application of machine learning approaches to these data, which may help uncover novel biomarkers and assess the generalizability of findings beyond the original study population.

Key Words: Risk factors; Carotid plaque; Type 2 diabetes mellitus; Limitations; Future perspective

Core Tip: The precise risk factors for carotid plaque in patients with type 2 diabetes mellitus have not been clearly defined. A recent study has highlighted age, body mass index, fasting plasma glucose, glycated hemoglobin, serum creatinine, urinary albumin‑to‑creatinine ratio, and serum uric acid as key contributors to carotid atherosclerosis in this population. However, the study carries certain methodological limitations. To improve the robustness of these findings, future prospective multicenter research should incorporate diabetes duration and a full medical history. Additional use of machine learning techniques could further elucidate causal pathways and improve cerebrovascular risk stratification in individuals with type 2 diabetes mellitus.

INTRODUCTION

Type 2 diabetes mellitus (T2DM) is a chronic metabolic disease characterized by persistent hyperglycemia, and its incidence is on the rise worldwide[1]. T2DM is one of the important independent risk factors for panvascular disease, involving chronic hyperglycemia, insulin resistance, dyslipidemia, oxidative stress, and a state of longterm inflammation[2,3]. Panvascular disease is a group of systemic vascular diseases that can simultaneously affect large, medium, and microvessels, with the main target organs being the heart, brain, peripheral blood vessels, kidneys, and retina[4]. Among others, cardiovascular and cerebrovascular diseases are the leading cause of death among world population[5]. Due to the easy detection of carotid plaque and its close association with the risk of cardiovascular events, the studies of carotid plaque risk factors hold significant importance for disease stratification and early interventions[6].

The risk factors for carotid atherosclerosis are often grouped into traditional and non-traditional categories, reflecting the complex metabolic environment of diabetes mellitus. Traditional risk factors include older age, male sex, a longer time living with diabetes, hypertension, body mass index, smoking, physical inactivity, dyslipidemia, and irrational diet[7]. All of these are independent predictors of carotid plaque formation and carotid intima-media thickness (IMT) progression. Non-traditional risk factors are constantly being explored and discovered, such as glycated hemoglobin A1c (HbA1c), 1,5-anhydroglucitol, triglycerides, high-density lipoprotein cholesterol lipoprotein(a), meteorin-like protein, high-sensitivity C-reactive protein, interleukin-1 beta (IL-1β), growth factor receptor-bound protein 2, serum creatinine (Scr), estimated glomerular filtration rate, serum uric acid[2,7-10], and fibroblast growth factor 23[11-15]. Indices of glycemic variability (coefficient of variation) and reduced time in range derived from continuous glucose monitoring have emerged as significant predictors of atherosclerosis progression[16]. Novel markers, the uric acid-to-high-density lipoprotein ratio, the uric acid-to-albumin ratio, and urinary albumin-to-creatinine ratio (UACR) have been identified as potential risk factors for carotid atherosclerosis in T2DM[17-19]. Table 1 summarizes the risk factors associated with carotid plaque in T2DM.

Table 1 Summary of risk factors associated with carotid plaque in type 2 diabetes mellitus.

Classification	Category	Risk factors
Traditional	-	Older age, male sex, a longer time living with diabetes, hypertension, body mass index, smoking, physical inactivity, dyslipidemia, irrational diet
Non-traditional	Glycemic metabolism	Glycated hemoglobin A1c, 1,5-anhydroglucitol, glycemic variability (coefficient of variation), reduced time in range
	Lipid metabolism	Triglycerides, low high-density lipoprotein cholesterol, lipoprotein(a), uric acid-to-high-density lipoprotein ratio, meteorin-like protein
	Chronic inflammation	High-sensitivity C-reactive protein, interleukin-1 beta, growth factor receptor-bound protein 2
	Oxidative stress	Specific oxysterols
	Renal dysfunction	Serum creatinine, urinary albumin-to-creatinine ratio, estimated glomerular filtration rate
	Others	Serum uric acid, uric acid-to-albumin ratio, fibroblast growth factor 23

Although the related biomarkers of carotid plaque have been continuously reported, few studies can comprehensively reveal the risk factors for carotid plaque in T2DM. After reading Shi et al’s article[19] on the risk factors for carotid plaque in T2DM, we review the characteristics and limitations of this study and provide some suggestions for further research. We emphasize the use of machine learning algorithms to analyze more extensive data for identifying the relevant risk factors and improving the reference for clinical primary prevention and early intervention for carotid plaque in T2DM[20].

EVALUATION OF THE STUDY

We are pleased to have read the high-quality article by Shi et al[19], published in World Journal of Diabetes. The primary focus of this article is the comprehensive analysis of risk factors associated with carotid plaque formation in a high-risk population of patients with T2DM. The study provides a scientific basis for integrated cardiovascular risk assessment and early intervention in T2DM patients. The findings are valid and consistent with current understanding of pathophysiology but the inherent design of the study limits the reliability of the conclusions.

In T2DM care, one pressing question is how to spot high-risk patients before they experience major macrovascular events. The study of Shi et al[19] focuses on that question by examining the subclinical phase. Although carotid plaque and carotid IMT serve as critical indicators for achieving this objective[21], the results of that study (e.g., age, body mass index, fasting plasma glucose, HbA1c, Scr, UACR, and SUA) can offer earlier insights into the risk of cardiovascular and cerebrovascular events. As a cross-sectional and discriminative study of risk factors, its major contribution is not the identification of novel risk factors but the systematic validation of the independent predictive value of a variety of traditional risk factors in a well-defined and rigorously controlled T2DM patient cohort. This is consistent with the emphasis on individualized risk assessment in the current guidelines, which aims to provide more detailed indicators for risk stratification to improve routine care or preventive cardiology programs for highly heterogeneous T2DM populations.

LIMITATIONS OF RESEARCH METHODS AND DESIGN

As a single-center study based on a relatively homogeneous Chinese population, the conclusions of Shi et al’s study[19] are doubtful to be generalized to other ethnic and geographical populations, which limits the application of the research results. More importantly, the retrospective cross-sectional design can only reveal correlations between variables, but cannot determine the time sequence and causality between risk factors and plaque formation[22,23]. Although the authors identify Scr and UACR as important risk factors, it is important to note that in a cross-sectional design, we cannot determine if renal dysfunction precedes or results from atherosclerosis. Both are likely to be parallel consequences of long-term hyperglycemia and other metabolic damage. Scr and UACR have been shown to play a role in the formation of carotid plaques in previous studies[24,25], but more robust studies (e.g., prospective cohort studies or randomized controlled trials) are still needed to determine their temporal sequence and causal relationship[26].

The study excluded patients with a history of malignant tumors, autoimmune diseases, and recent cardiovascular events to isolate the effects of T2DM-related metabolic disorders and avoid strong, non-diabetic-related inflammation and vascular damage[27]. However, this may lead to the fact that the study sample cannot fully represent the complex T2DM patient population usually accompanied by multiple comorbidities. Nonetheless, the biggest issue is that the study did not account for T2DM duration, which is essential for estimating the cumulative risk of complications[28]. Its absence may lead to substantial confounding bias. For instance, patients with severe carotid plaque but a short history of diabetes may have a very different set of dominant risk factors compared to those with mild plaque but long-standing disease. Without duration data, the regression models cannot adjust for this critical variable, potentially leading to overestimation of the effects of other factors (e.g., HbA1c) or failure to detect interactions between duration and other factors.

The study collected data on glycemic control, insulin resistance, lipid metabolism, renal function, and uric acid metabolism. Still, several important variables were absent. The greatest source of potential confounding is the absence of recording of medication history. The use of medications such as statins, antihypertensives (especially angiotensin-converting-enzyme inhibitors and angiotensin-receptor blockers), and specific antidiabetic agents (e.g., sodium-glucose co-transporter 2 inhibitors and glucagon-like peptide-1 receptor agonists with clear cardiovascular benefits) directly influences all key indicators (low-density lipoprotein cholesterol, blood pressure, HbA1c, and UACR) and the primary outcome (plaque formation)[29-32]. This omission makes it impossible to distinguish whether the observed risk factors are primary or residual risks under current medical therapy. Additionally, basic lifestyle information (e.g., smoking, drinking, dietary patterns, and physical activity levels) was not collected[33]. These are the fundamental drivers of the occurrence and progression of atherosclerosis.

FUTURE PERSPECTIVES

Multi-center data analysis and validation

In order to compensate for the shortcomings of existing research, it is necessary to carry out a prospective, multi-center, community-based cohort study to conduct in-depth exploration, systematically collect potential factors at baseline, and perform regular follow-up to track the dynamic changes of carotid IMT, thereby enabling the establishment of a causal relationship[34]. Future research could make a brief synthesis of recent databases implementing comprehensive data collection (e.g., United Kingdom Biobank and China Kadoorie Biobank) to promote the globalization of the results[35,36]. At the same time, the results should be rigorously validated using large public international databases. Alternatively, external validation can be conducted on diverse population data obtained through multi-center collaboration[37]. This will confirm the generalizability of the conclusion and promote its translation into clinical practice.

Multimodal data integration

When considering risk factors, it is essential to include novel biomarkers, such as lipoprotein-associated phospholipase A2 and high-sensitivity C-reactive protein reflecting vascular inflammation[38], as well as myeloperoxidase to reflect oxidative stress and N-terminal pro-B-type natriuretic peptide to reflect cardiac stress[39,40]. Feature engineering methods should be used to construct more predictive derived features[41], for example, the uric acid-to-albumin ratio and HbA1c-to-creatinine ratio. Besides, plaque imaging characteristics should be extended beyond IMT measurement to encompass morphological features closely associated with vulnerability and should also integrate multimodal data[42]. Convolutional neural networks can process original carotid ultrasound images to automatically extract deep features and plaque texture and morphology[43-45]. Natural language processing can extract unstructured text information about lifestyle descriptions and disease course from electronic health records[46-48].

Multiomics data fusion

Furthermore, the integrated omics techniques (e.g., proteomics and metabolomics) will facilitate the screening of characteristic molecular profiles[49-51]. Machine learning algorithms are especially suitable for dealing with such high dimensional data. The algorithms including support vector machines and regularized logistic regression (least absolute shrinkage and selection operator) can screen for molecular features strongly correlated with carotid plaques[52]. Constructing combined models that integrate clinical indicators and omics biomarkers is expected to raise the area under the curve for prediction to a higher level[53]. This would enable earlier and more accurate risk warnings.

Full use of mechanical learning

For high-dimensional data, algorithms like random forest or support vector machines can help build better predictive models[54]. For example, the algorithms of XGBoost and LightGBM can automatically handle nonlinear interactions between these related features[55-57], like the synergistic effect of age and HbA1c. Machine learning algorithms can better handle both continuous variables (e.g., fasting plasma glucose and Scr) and discrete variables (e.g., sex) in similar studies, significantly improving the accuracy of carotid plaque risk prediction (i.e., area under the curve value)[58]. Methodologically, random survival forest and extreme gradient boosting are mostly suited for longitudinal risk prediction[59], while multiple imputation and latent variable model could be used to handle missing data or measurement bias in large-scale datasets[60]. After collecting longitudinal follow-up data, researchers can apply random survival forests or survival models based on deep learning to predict the time-to-event risk of carotid plaque development or progression. This would allow dynamic risk stratification and help guide the timing of intervention.

CONCLUSION

Carotid atherosclerosis in T2DM arises from multiple factors, including traditional risk factors, systemic inflammation, oxidative stress, and diabetes-related metabolic disturbances. Shi et al’s study[19] clearly identified the key factors influencing the development of carotid atherosclerosis in T2DM, such as age, obesity, hyperglycemia, renal dysfunction, and hyperuricemia. While it could not definitively describe the causal relationship between T2DM and carotid atherosclerosis due to design limitations, it provides highly valuable guidance for clinicians. The study stresses the importance of uric acid and urinary albumin management, as well as blood glucose and blood lipid control. Further study should collect more extensive data to thoroughly reveal the related factors of carotid atherosclerosis in T2DM by using machine learning for analysis. Through multifaceted risk assessment and intervention, we can truly achieve early prevention of the progression of cardiovascular events in this population.