Incremental predictive value of liver fat fraction based on spectral detector CT for major adverse cardiovascular events in T2DM patients with suspected coronary artery disease

Abstract

Background

The purpose of this study was to explore the incremental predictive value of liver fat fraction (LFF) in forecasting major adverse cardiovascular events (MACE) among patients with type 2 diabetes mellitus (T2DM).

Methods

We prospectively enrolled 265 patients with T2DM who presented to our hospital with symptoms of chest distress and pain suggestive of coronary artery disease (CAD) between August 2021 and August 2022. All participants underwent both coronary computed tomography angiography (CCTA) and upper abdominal dual-layer spectral detector computed tomography (SDCT) examinations within a 7-day interval. Detailed clinical data, CCTA imaging features, and LFF determined by SDCT multi-material decomposition algorithm were meticulously recorded. MACE was defined as the occurrence of cardiac death, acute coronary syndrome (ACS), late-phase coronary revascularization procedures, and hospital admissions due to heart failure.

Results

Among 265 patients (41% male), 51 cases of MACE were documented during a median follow-up of 30 months. The LFF in T2DM patients who experienced MACE was notably higher compared to those without MACE (p < 0.001). The LFF was divided into tertiles using the cutoffs of 4.10 and 8.30. Kaplan-Meier analysis indicated that patients with higher LFF were more likely to develop MACE, regardless of different subgroups in framingham risk score (FRS) or coronary artery calcium score (CACS). The multivariate Cox regression results indicated that, compared with patients in the lowest tertile, those in the second tertile (hazard ratio [HR] = 3.161, 95% confidence interval [CI] 1.163–8.593, P = 0.024) and third tertile (HR = 4.372, 95% CI 1.591–12.014, P = 0.004) had a significantly higher risk of MACE in patients with T2DM. Even after adjusting for early revascularization, both LFF tertile and CACS remained independently associated with MACE. Moreover, compared with the traditional FRS model, the model that included LFF, CACS, and FRS showed stable clinical net benefit and demonstrated better predictive performance, with a C-index of 0.725, a net reclassification improvement (NRI) of 0.397 (95% CI 0.187–0.528, P < 0.01), and an integrated discrimination improvement (IDI) of 0.100 (95% CI 0.043–0.190, P < 0.01).

Conclusions

The elevated LFF emerged as an independent prognostic factor for MACE in patients with T2DM. Incorporating LFF with FRS and CACS provided incremental predictive power for MACE in patients with T2DM.

Graphical abstract

Research insights

What is currently known about this topic?

T2DM is associated with increased MACE rates, underscoring the need for improved risk prediction. CACS is a well-established tool for MACE risk assessment but may not capture all risk factors. Hepatic steatosis is a common comorbidity in metabolic syndrome and T2DM.

What is the key research question?

Does the incorporation of LFF derived from SDCT into existing risk prediction models enhance the accuracy of MACE forecasting in patients with T2DM?

What is new?

SDCT-LFF measurement introduces a more accurate method for assessing hepatic steatosis. LFF as an independent predictor of MACE in T2DM patients is a novel finding. The study presents LFF as an additional tool for risk stratification, complementing FRS and CACS.

How might this study influence clinical practice?

Study findings may guide personalized prevention for T2DM patients at higher MACE risk.

Background

Type 2 diabetes mellitus (T2DM) is a systemic metabolic disorder that precipitates a spectrum of microvascular and macrovascular complications. Cardiovascular diseases stand as the leading cause of mortality among T2DM patients [1, 2]. Consequently, the accurate forecasting of major adverse cardiovascular events (MACE) in T2DM is of paramount importance for facilitating prompt clinical intervention in the management of cardiovascular disorders in these patients. Such predictive capabilities are pivotal for enhancing the quality of life, extending life expectancy, and curtailing healthcare expenditure associated with T2DM.

In recent years, the assessment of energy metabolism has played a crucial role in forecasting and managing the risks associated with cardiovascular diseases [3, 4]. The liver occupies a central position in metabolic abnormalities within the body. Excessive hepatic steatosis-driven disruptions in lipid metabolism and inflammation are pivotal not only in the formation of atherosclerotic plaques but also in the instigation of plaque ruptures and the onset of MACE [5,6,7]. Moreover, hepatic steatosis is a common comorbidity in metabolic syndrome and T2DM. Hepatic steatosis can further promote the occurrence and progression of T2DM through the disruption of glycolipid metabolism and insulin resistance, thereby accelerating the onset and progression of its cardiovascular complications [8, 9]. However, clinical research investigating the association between hepatic steatosis and the incidence of MACE in patients with T2DM remains controversial [8, 10], largely due to inconsistent diagnostic criteria for hepatic steatosis. Moreover, traditional non-invasive diagnostic techniques like ultrasound and conventional CT scans lack the precision needed to assess hepatic steatosis accurately [11]. While histopathological assessment can accurately evaluate the severity of hepatic steatosis, its clinical application is constrained by the limitations of standard clinical follow-ups and longitudinal research [12]. Therefore, for T2DM patients, the emergence of innovative imaging technologies for precise assessment of hepatic steatosis may serve as an indicator of metabolic risk, enhancing the predictive value for MACE.

The spectral detector computed tomography (SDCT) capitalizes on the distinct attenuation properties of materials at varying energy levels, offering a more nuanced tissue characterization that surpasses the conventional reliance on Hounsfield units. The Liver fat fraction (LFF), derived from SDCT material decomposition technique, provides a more precise measure of hepatic fat deposition [13, 14]. This research endeavors to investigate the relationship between LFF and the incidence of MACE in a population with suspected coronary artery disease (CAD) among T2DM patients. Additionally, it aims to assess the incremental value of LFF in enhancing risk stratification for MACE prediction, thereby serving as an early warning for potential cardiovascular high-risk patients with metabolic risk factors to prevent adverse outcomes.

Material and method

Study population

We prospectively recruited patients with T2DM who, between August 2021 and August 2022, presented with symptoms of chest distress and pain indicative of suspected CAD and underwent coronary computed tomography angiography (CCTA) at our institution. Eligible patients also received abdominal SDCT within a one-week interval of CCTA. Exclusion criteria were applied to individuals aged 18 years or younger, those with incomplete clinical profiles, and those with a history of various heart diseases (such as dilated cardiomyopathy, hypertrophic obstructive cardiomyopathy, atrial fibrillation, myocarditis, and inflammatory vasculopathies) or prior cardiovascular events (including acute myocardial infarction, unstable angina, chronic stable angina, previous percutaneous or surgical interventions for coronary artery disease, and heart failure). Patients with a history of alcohol use, infectious diseases, or malignancies, as well as those with poor-quality imaging or follow-up periods shorter than one year, were also excluded (Fig. 1). The study protocol was approved by the medical ethics committee of The Shengjing Hospital of China Medical University (2021PS720K). All participants provided written informed consent, and follow-up data were acquired through medical record review and telephone interviews.

Fig. 1

The flowchart of the research. SDCT, spectral detector computed tomography; T2DM, type 2 diabetes mellitus; CCTA, coronary computed tomography angiography; MACE, major adverse cardiovascular; PCI, percutaneous coronary intervention.

Data collection

Blood samples are collected within 24 h of a patient’s admission or outpatient clinic visit. Clinical and demographic characteristics are gathered from the hospital information system (HIS), including age, gender, hypertension, duration of diabetes, smoking status, and medication history. Laboratory indicators are obtained from the laboratory information system (LIS), encompassing white blood cells, neutrophils, lymphocytes, monocytes, total cholesterol, low-density lipoprotein (LDL), high-density lipoprotein (HDL), triglycerides, fasting blood glucose, and glycated hemoglobin levels, and the Triglyceride–glucose (TyG) index = ln [Triglyceride(mg/dl) × Fasting blood glucose(mg/dl)/2]. The systemic inflammatory response index (SIRI) = Neutrophils × Monocyte / Lymphocyte. The framingham risk score (FRS), which includes variables such as age, gender, total cholesterol, high-density lipoprotein, systolic blood pressure, and smoking status, is utilized to categorize patients’ risk for coronary heart disease into three tiers: high risk (≥ 20% within 10 years), moderate risk (10-19% within 10 years), and low risk (< 10% within 10 years).

Measurement of LFF

All scans were conducted using a 64-slice dual-energy detector spectral CT scanner (IQon Spectral CT, Philips Healthcare, Best, the Netherlands). The scan slice thickness was set to 1 mm, and the raw images were reconstructed into spectral-based imaging (SBI) data packages using spectral iterative reconstruction. The fat fraction maps were reconstructed using specialized post-processing software by Philips. Subsequently, two radiologists with 6 and 8 years of experience in abdominal imaging diagnostics, who were blinded to the clinical data, measured the LFF value. They placed three regions of interest (ROI), approximately 300 mm² (with a deviation of less than 10 mm²), at the level of the main portal vein branch in the right lobe of the liver within the left lobe, right anterior lobe, and right posterior lobe [15]. If the left lobe was not visible at this level, the section with the largest cross-sectional area of the left lobe was selected for measurement. The ROIs were carefully drawn to avoid blood vessels, bile ducts, focal liver lesions, areas of non-uniform infiltration, and imaging artifacts within the liver (see later Fig. 2 for details). The mean value of the measurements from the three ROIs was taken as the parameter value. Once consistency between the two sets of measurements was achieved, the average of the values measured by both observers was considered as the final parameter for inclusion in the study.

Fig. 2

Liver SDCT-LFF images of a 51-year-old male patient. The above image illustrates the measurement of LFF based on an SDCT material separation algorithm in a patient. Circular ROIs (in yellow) are placed in the left lobe, right anterior lobe, and right posterior lobe of the liver, with each ROI approximately 300 mm² in size. The delineation of ROIs carefully avoids blood vessels, bile duct structures, focal liver lesions, and imaging artifacts within the liver. The LFF values for the left lobe, right anterior lobe, and right posterior lobe of the liver are 14.50%, 12.00%, and 11.00%, respectively. SDCT, spectral detector computed tomography; LFF, liver fat fraction; ROI, region of Interest

CCTA scan and plaque analysis

All scans were performed using a 64-slice dual-layer spectral detector CT scanner (IQon Spectral CT, Philips Healthcare, Best, the Netherlands). All patients received 0.25 mg sublingual nitroglycerin (Beijing Yimin Pharmaceutical Plant, Beijing) prior to the imaging. For patients with a heart rate exceeding 70 beats per minute, oral beta-blockers (25–50 mg; Metoprolol Succinate sustained-release tablets, AstraZeneca, Sweden) were administered to stabilize the heart rate. The scanning area extended from 1 cm below the tracheal bifurcation to the cardiac apex, with the following scan settings: a tube voltage of 120 kV was set, alongside an automatic tube current modulation technique, prospective gating with a dose right index (DRI) of 13 at a trigger phase of 78% of the R-R interval, with a ± 3% buffer zone for data acquisition; retrospective gating with a DRI of 28, a field of view of 220 mm, a reconstruction matrix of 512 × 512, a tube rotation time of 0.27 s, a detector collimation set to 64 × 0.625 mm, a reconstruction slice thickness of 0.9 mm, a reconstruction interval of 0.45 mm, a total contrast medium volume calculated as body weight×0.8 ml/kg, and an injection rate (ml/s) = the total volume (ml)/the injection duration (12 s), followed by a 20 to 30 ml saline flush at the same injection rate. The ROI was placed within the ascending aorta at the level of the main pulmonary artery window; once the trigger threshold of 150 Hounsfield Units (HU) was reached, patients were instructed to hold their breath for the scan, and the raw data were reconstructed using the IMR-Cardiac Routine-Level 1 algorithm.

Coronary plaque characterization was performed using the IntelliSpace Portal workstation (version 6.5, Philips Healthcare), a sophisticated tool for the detailed analysis of atherosclerotic plaques. The software automatically delineates key plaque metrics, including total plaque volume and total plaque burden, quantified as the total plaque volume multiplied by 100% divided by the vessel volume. It also assesses the percentage of low-attenuation plaques (≤ 30 HU), medium-attenuation plaques (31–130 HU), and high-attenuation plaques (≥ 131 HU). Manual adjustments of the plaque contours can be made when necessary to ensure accuracy. The characteristics of high-risk plaque (HRP) in CCTA images include Low-attenuation plaque with CT values below 30 HU across areas larger than 1 mm²; spotty calcification within non-calcified plaques, marked by lengths less than 3 mm and densities exceeding 130 HU; positive remodeling, signified by a remodeling index above 1.1; and the “napkin-ring” sign, a slightly elevated density rimming low-density plaques [16], a plaque is designated as high-risk when it possesses two or more high-risk characteristics. The coronary artery calcium score (CACS) was meticulously quantified using the Philips IntelliSpace Portal workstation. Obstructive stenosis is characterized by a luminal narrowing exceeding 50% in the left main coronary artery or surpassing 70% in any other coronary artery segments [17]. Two radiologists conducted an independent and blinded analysis of the aforementioned parameters. The quantitative parameters, as ascertained by both observers, were averaged and employed for subsequent analysis.

Sample size estimation

Sample size estimation was performed using PASS software (version 21.0.3, 2021). Based on relevant data reported in the literature [8], the Survival-Cox Regression module was utilized with the following parameters: Power = 0.90, alpha = 0.05, overall event rate = 0.08, B (Log Hazard Ratio) = 0.70, R-squared of X1 with other X’s = 0.3, and S (Standard Deviation of X1) = 1.5. The estimated total sample size was 171. Assuming a 10% dropout rate, the final subjects count was established at 190. However, during the study period, we prospectively enrolled all eligible patients who met the inclusion criteria. This approach allowed us to maximize the available data and enhance the robustness of our findings. As a result, we ended up including a total of 265 patients in our analysis.

Clinical endpoint events

Telephone follow-ups and HIS case inquiries were conducted every six months by physicians who were unaware of the results of the coronary CCTA examinations, ensuring an unbiased recording of the follow-up outcomes. For the purposes of this study, the documented MACE primarily encompassed the following: cardiac death, acute coronary syndrome (ACS), late non-urgent revascularization procedures, and hospital admissions due to heart failure.

Cardiac death was defined as death resulting from acute myocardial infarction, ventricular arrhythmias, or intractable heart failure. ACS was categorized as ST-segment elevation myocardial infarction, unstable angina, and non-ST-segment elevation myocardial infarction. An ACS event was recorded only for cases where the culprit lesion was identified, and an emergency myocardial revascularization was undertaken. Late nonurgent revascularization refers to nonurgent revascularization procedures performed more than six months after coronary CCTA [18].

Statistical analysis

Statistical analysis was performed using SPSS software (version 26.0, SPSS Inc., Armonk, NY, USA) and R software (version 4.3.2, available at https://www.r-project.org). The Shapiro-Wilk test was applied to assess the normality of continuous data. Data adhering to a normal distribution were presented as mean ± standard deviation and compared using the independent samples T-test. Non-normally distributed data were depicted as median (interquartile range) and analyzed using the Mann-Whitney U test for two-group comparisons and the Kruskal-Wallis test for three-group comparisons. Categorical variables were expressed as percentages and compared using the chi-square test or Fisher’s exact test, as appropriate. Univariate cox regression analysis was conducted to identify potential predictors for the occurrence of MACE in patients with T2DM. For continuous independent variables, the optimal cutoff values were determined using R software. All variables were then transformed into binary variables, and those significant in the univariate cox model (P < 0.05) were simultaneously entered into a multivariate cox model. The results of the multivariate cox regression analysis were presented as hazard ratios (HR) with 95% confidence intervals (CI). Based on the results of the multivariate cox regression, risk factors were categorized, and the Kaplan-Meier method was used to calculate the cumulative survival probability, with the log-rank test used for comparisons. The reclassification capability of the updated model was further assessed using the C-index, net reclassification index (NRI), and integrated discrimination improvement (IDI). The clinical net benefit of the model was evaluated using decision curve analysis (DCA).

Results

Patient characteristics

A total of 265 patients with T2DM were enrolled in the study cohort. The follow-up period lasted until November 30, 2024, when all enrolled patients had finished their follow-up assessments. The average follow-up duration was 30 months. Over this period, 51 patients (19.2%) experienced MACE, including 32 cases of ACS, 6 instances of cardiac death and 13 cases of late non-urgent revascularization. No patients were hospitalized due to heart failure.

The study cohort was divided according to the presence or absence of MACE. Clinical and CCTA imaging characteristics of the 265 enrolled patients are detailed in Tables 1 and 2. The patients who experienced MACE demonstrated a significantly longer T2DM time (P = 0.047), and an increased FRS% along with higher risk stratification (P = 0.041 and P = 0.027, respectively). The LFF was significantly higher in the MACE group (P < 0.001). The prevalence of CACS, as measured by both multi-category and binary classifications, the severity of coronary stenosis, the proportion of HRP, and the prevalence of obstructive stenosis were all markedly higher in the MACE group compared to those without MACE (P = 0.001, < 0.001, 0.002, 0.046, and 0.004, respectively). In contrast, the probability of MACE occurrence was lower among patients undergoing early revascularization (P = 0.019). Moreover, the MACE group showed a lower proportion of medium-attenuation plaque burden proportion (P = 0.032).

Table 1 Comparison of baseline characteristics between patients with T2DM with and without MACE

Table 2 Comparison of baseline characteristics between patients with T2DM with and without MACE

Correlation of LFF with plaque characteristics

As depicted in Fig. 2, it illustrates the LFF images and measurement methods of two patients. The measurements of LFF values and the means of three ROIs across different liver lobes demonstrate good consistency (Supplementary Table 1), and the measurement bias is kept within 5% (Fig. 3). Consequently, the average of the two measurements was taken as the final result. Patients were categorized into tertiles based on their admission LFF index levels: Tertile 1 (n = 88, LFF < 4.1), Tertile 2 (n = 88, 4.10 ≤ LFF < 8.30), and Tertile 3 (n = 89, LFF ≥ 8.30).

Fig. 3

Scatter plots of LFF values. A Bland-Altman plot of LL LFF for two viewers. B Bland-Altman plot of RAL LFF for two viewers. C Bland-Altman plot of RPL LFF for two viewers. D Bland-Altman plot of mean LFF for two viewers. SD, standard deviation; LL, Left lobe; RAL, Right Anterior Lobe; RPL, Right Posterior Lobe

As shown in Table 3, the proportion of occurrences of HRP (P = 0.001), HRP feature (positive remodeling, P = 0.003; low-attenuation plaque, P = 0.047; and spotty calcification, P = 0.004) and MACE proportion (P < 0.001) also increases with the elevation of LFF values. Additionally, the prevalence of low-attenuation plaque burden proportion is higher with increasing LFF (P = 0.008), exhibiting a mild linear correlation between LFF and low-attenuation plaque burden proportion (Rs = 0.184, P = 0.003, Supplementary Table 2). However, no correlation was found between LFF and the overall plaque burden, as well as the proportions of high-attenuation plaque burden (P = 0.965, 0.769, respectively).

Table 3 Comparison of coronary plaque characteristics in different LFF groups

Univariate and multivariate Cox regression analysis of MACE

As shown in Table 4, the univariate Cox regression analysis showed that HBA1C > 6.3, FRS > 20%, TYG > 9.68, LFF level, CACS ≥ 100, obstructive stenosis, HRP, early revascularization, total plaque burden, high-attenuation plaque burden proportion were all significantly associated with MACE (All p < 0.05). Given the significant linear correlations between the CACS and total plaque burden, high-attenuation plaque burden proportion, medium-attenuation plaque burden proportion, and low-attenuation plaque burden proportion (with Spearman’s rank correlation coefficients of 0.816, 0.853, -0.717, and − 0.623 respectively, all with P-values < 0.001), and considering that in the univariate regression analysis, the hazard ratio (HR) corresponding to CACS was the highest, we selected CACS for inclusion in the multivariate cox regression analysis to minimize the bias from confounding factors. Furthermore, we developed two multivariate Cox regression models. In Model 1, we adjusted for covariates that were significantly associated with the outcome in univariate Cox regression, excluding treatment modality. The results indicated that CACS ≥ 100 (HR = 2.145, 95% CI: (1.153–3.990, P = 0.016) was an independent predictor of MACE. Using LFF Tertile 1 as the reference, both LFF Tertile 2 (HR = 3.161, 95% CI: 1.163–8.593, P = 0.024) and LFF Tertile 3 (HR = 4.372, 95% CI: 1.591–12.014, P = 0.004) emerged as independent predictors of MACE. In Model 2, we further adjusted for early revascularization based on Model 1. Compared with Model 1, CACS ≥ 100 and LFF tertiles remained robust independent predictors of MACE, underscoring their significant impact on the outcome. Additionally, after adjusting for early revascularization, the true independent risk of obstructive stenosis was revealed (HR = 2.269, 95% CI: 1.433–4.970, p = 0.002).

Table 4 Risk factors associated with MACE in patients with T2DM

Discriminative ability and clinical utility assessment

We constructed four models (Table 5): Model A (FRS), Model B (FRS + CACS), Model C (FRS + CACS + LFF), and Model D (FRS + CACS + LFF + Obstructive stenosis + Early revascularization). Among these models, Model C and Model D, which included CACS and LFF, demonstrated relatively higher C-index values. Furthermore, we employed the NRI and IDI to assess the reclassification capabilities of the models. Compared with Model A, Model C, which incorporated CACS and LFF, exhibited significant NRI of 0.397(95% CI: 0.187–0.528, P < 0.01), and IDI of 0.100 (95% CI: 0.043–0.190, P < 0.01). Similarly, Model D showed significant improvements in both NRI and IDI compared with Model A. However, when compared with Model C, although Model D demonstrated a mild but significant improvement in IDI, the improvement in NRI was not significant (p = 0.348).

The clinical utility of the four models was assessed using DCA (Fig. 4). The DCA demonstrated that Model D, which incorporated FRS, CACS, LFF, obstructive stenosis, and early revascularization, provided a superior clinical net benefit within the threshold probability range of 10–45%. However, when the threshold exceeded 45%, the net benefit of Model D fell below that of Model C (FRS + CACS + LFF).

Table 5 Improvement in MACE risk reclassification

Fig. 4

Decision Curve Analysis of the Models. The y-axis measures the net benefit. The net benefit is determined by calculating the difference between the expected benefit and the expected harm associated with each proposed model Net benefit = true positive rate-(false positive rate x weighting factor), weighting factor = threshold probability/ (1-threshold probability). Model A：FRS；Model B：FRS + CACS；Model C：FRS + CACS + LFF；Model D：FRS + CACS + LFF + Obstructive stenosis + Early revascularization. Abbreviation: CACS, coronary artery calcium score; LFF, liver fat fraction. FRS, Framingham Risk Score

Fig. 5

Kaplan-Meier curve of MACE. A Stratified analysis by LFF tertile. B Stratified analysis by LFF tertile and CACS. C Stratified analysis by LFF tertile and FRS. CACS, coronary artery calcium score; LFF, liver fat fraction. FRS, Framingham Risk Score

Kaplan–Meier survival curve analysis

Individuals with a high tertile LFF had significantly worse MACE-free survival than those with a low tertile LFF (Fig. 5A), both in the presence of CACS ≥ 100 (Fig. 5B) and FRS% ≥ 20 group (Fig. 5C) (All p < 0.05). The cumulative probability of MACE was the highest among individuals with CACS ≥ 100 and LFF tertile 3 group (Fig. 6).

Fig. 6

Total MACE rate in different groups classified by CACS, LFF value, and FRS%. LFF was transformed into categories according to the tertile. MACE, major adverse cardiovascular events; CACS, coronary artery calcium score; LFF, liver fat fraction. FRS, Framingham Risk Score

Discussion

This article evaluates the correlation between LFF and the occurrence of MACE in patients with T2DM. Following a follow-up period of 30 months, the results indicate that LFF is independently associated with the occurrence of MACE and can improve the C-index and reclassification ability of the original model based on conventional clinical risk factors and CACS derived from CCTA. Furthermore, the combination of FRS, CACS, and LFF can enhance the accuracy of MACE prediction in T2DM patients.

To our knowledge, this is the first article to quantitatively assess the severity of hepatic steatosis and its correlation with MACE in T2DM patients. Clinical manifestations of T2DM exhibit a wide range of variability, leading to considerable differences in cardiovascular prognosis. Implementing a thorough risk assessment at an early stage is essential for guiding management strategies that may enhance outcomes for those at higher risk, while also preventing unnecessary financial burdens on lower-risk individuals and minimizing health economic waste [19]. The FRS, a prevalent tool for estimating the 10-year cardiovascular risk in coronary heart disease patients [20], lacks the incorporation of certain critical biomarkers. CACS is strongly associated with an increased risk for cardiovascular events in asymptomatic patients with diabetes and increased lifetime cardiovascular disease risk [21, 22]. Nonetheless, cardiovascular events are not uncommon among those categorized as low risk by CACS, potentially due to the presence of non-calcified and vulnerable plaques [23, 24]. Hepatic steatosis has been identified as an independent risk factor influencing these types of plaques, and in the T2DM population, the disruption of glycolipid metabolism can intensify hepatic steatosis and insulin resistance, consequently elevating the risk of MACE.

Hepatic steatosis has been proposed as a better indicator of lipid metabolism disorders and insulin resistance than visceral adipose tissue (VAT) [25], and it is intricately linked to the underlying pathophysiological mechanisms of diabetes mellitus and comorbidities associated with obesity, thereby emerging as a pivotal maker of cardiometabolic risk [26].In individuals with T2DM, there is an exacerbated susceptibility to the development of hepatic steatosis [9, 27], which in turn intensifies endothelial dysfunction, modulates vascular tone, and augments the formation of atherosclerotic plaques [28, 29]. Consequently, this not only elevates the risk of cardiovascular diseases but may also precipitate a more aggressive and profound progression of cardiovascular events. The fatty liver index (FLI), derived from clinical parameters, has been widely utilized in epidemiological studies and demonstrates robust predictive capability for adverse cardiovascular outcomes [30, 31]. This underscores the broad association between abnormal hepatic lipid metabolism and cardiovascular risk. However, a critical component of FLI, γ-glutamyltransferase (γ-GTP), is susceptible to influences from inflammation, medications, and biliary tract diseases [32, 33], which may compromise the predictive accuracy of FLI in specific populations. Moreover, some studies elucidating the nexus between hepatic steatosis and the incidence of adverse cardiovascular events in T2DM patients remain a contentious domain [8, 10]. The discordance may stem from the imperfect diagnostic precision of hepatic steatosis in prior studies, which has often led to an underestimation or misclassification of the risk stratification of CAD. According to the latest research findings [34, 35], the previous use of an attenuation of less than 40 HU to identify participants with hepatic steatosis at CT would underestimate the presence of hepatic steatosis. Consequently, this research initiates with an accurate quantification of hepatic steatosis and proceeds to assess the prognostic significance of these quantitative metrics for patients with T2DM.

As an imaging biomarker, LFF is computed using a material decomposition algorithm based on the linear equations of Compton scattering and the photoelectric effect. And unlike dual-energy CT (DECT) techniques, SDCT does not require a prospective selection of a dual-energy imaging mode and can retrospectively provide spectral information for every scan at ≥ 120 kV. Research evidence indicates that SDCT-based multiparametric indices can precisely and non-invasively determine the severity of hepatic steatosis, with magnetic resonance imaging proton density fat fraction (MRI-PDFF) serving as the reference standard [36]. Additionally, the LFF derived from the multi-material decomposition (MMD) algorithm based on SDCT has been proven to be an accurate method for assessing hepatic fat deposition in both phantom experiments and small-sample clinical studies [13]. Corroborating previous studies that hepatic steatosis is significantly associated with the presence of HPR [5, 37], our study also yielded similar results. As showed in Table 3, stratified analysis of LFF within the T2DM patient revealed that higher LFF values were associated with a greater prevalence of HRP. This association may be attributed to a complex interplay of factors, including endothelial dysfunction, platelet activation, altered lipid metabolism, insulin resistance, and oxidative stress, which collectively create a proatherogenic environment conducive to plaque formation and progression to HRP under the influence of stressors such as inflammation and shear stress [38]. Additionally, we observed that the burden of low-attenuation plaques increased with rising LFF. These factors are widely recognized as significant risk predictors for the development of MACE [18, 39]. However, in contrast to the previous study [40], we observed that the prevalence of obstructive stenosis tended to increase with higher LFF tertiles, although this association did not reach statistical significance. This discrepancy might be due to the specific characteristics of our study population, which included only individuals with T2DM and had a small sample size. This likely reduced the statistical power to detect a significant relationship between LFF and obstructive stenosis.

Our study diverges from previous study [8] by employing quantitative analysis of hepatic steatosis using SDCT-LFF, which provides a more precision assessment than the binary diagnosis of its presence in previous studies. This approach mitigates the risk of incorrect stratification among the population and allows for a more detailed evaluation of the correlation between the risk of MACE and the progression of hepatic steatosis. In our study, multivariate Cox regression analysis shows that, regardless of whether clinical treatment strategies are adjusted, LFF tertile-group stratification remains an independent predictor of MACE for the T2DM patients enrolled in our study. Upon further evaluation of the incremental predictive value of LFF, it was observed that Model D (FRS + CACS + LFF + Obstructive stenosis + Early revascularization), demonstrated reduced effectiveness in predictive stability when compared to the more streamlined Model C (FRS + CACS + LFF). Moreover, the NRI between the two models did not reach statistical significance (P = 0.348). The underlying reason may be attributed to the low event rate observed in this study. Among the 30 patients who underwent early revascularization, only one experienced MACE during follow-up. This extremely low event rate likely led to decreased model stability, thereby affecting predictive performance and statistical power. It is worth noting that CACS and LFF emerged as potent predictors of MACE. Their clinical accessibility further enhances the practical value of our findings. Research indicates that incorporating CACS and LFF into the traditional FRS model can improve the accuracy of MACE risk prediction by 39.7%, with a corresponding increase of 10.0% in positive predictive probability. The integration of LFF into risk stratification frameworks may refine personalized management for T2DM patients. For instance, individuals with elevated LFF (≥ 8.3%) but low-to-intermediate FRS could be prioritized for intensive metabolic therapy and closer cardiovascular monitoring. Future guidelines may incorporate LFF as a biomarker to guide personalized therapeutic strategies targeting hepatic steatosis and its related metabolic complications. However, prospective trials are needed to validate whether LFF-guided interventions improve clinical outcomes. Additionally, with the increasing application of artificial intelligence (AI) and machine learning technologies in the assessment of CAD [41, 42], future intelligent models integrating LFF parameters are expected to achieve more precise cardiovascular risk stratification, thereby further improving the accuracy of clinical decision-making.

Our study has its limitations. Firstly, the research was conducted as a single-center observational study with a relatively small patient cohort, the generalizability and clinical applicability of LFF as a novel indicator for MACE prediction need to be further established through external validation studies. Secondly, our study population exclusively comprised chinese individuals with T2DM suspected of CAD, and the findings may not be directly applicable to all T2DM patients or other ethnic groups. Thirdly, the nitroglycerin dose (0.25 mg) used in this study aligns with domestic clinical protocols, which differ from international guidelines (400–800 µg). This choice balanced safety and image quality optimization via SDCT’s advanced reconstruction capabilities. Fourthly, the use of medications may potentially influence plaque stability and systemic metabolic status. Additionally, this study did not capture changes in inflammatory markers, diet, physical activity, or the control of risk factors during the follow-up period, which could have influenced the assessment of current cardiovascular metabolic risk. Fifthly, the number of patients undergoing early revascularization was relatively small, which may have affected the model fitting and subsequently impacted the clinical net benefit of the model. Sixthly, the apparent discrepancy between the low short-term MACE rate (within 300 days) and the higher overall incidence may reflect the time-dependent effects of risk factors and medication adherence. Early intensive therapy likely suppresses acute events initially, but long-term non-adherence and cumulative metabolic burden may contribute to delayed adverse outcomes. Future research should encompass large-scale, multicenter cohorts with long-term follow-up, and longitudinal adherence monitoring, to better evaluate the independent predictive capability of LFF for MACE.

Conclusion

Our study demonstrated that LFF assessed by non-enhanced SDCT emerged as an independent prognostic factor for MACE in individuals with T2DM, regardless of whether early revascularization was adjusted for. Incorporating LFF with the FRS and CACS provided incremental prognostic value for MACE in T2DM patients. This research aids in precisely pinpointing high-risk T2DM patients with suspected CAD, enabling timely prevention and intervention to potentially reduce the incidence of cardiovascular events, enhance patient quality of life, and reduce healthcare system strain.