This study was designed as a prospective observational cohort. Consecutive patients with angiographically confirmed coronary artery disease (CAD) were enrolled at Puyang Oilfield General Hospital between February 2023 and June 2024. At baseline, all participants underwent coronary angiography with adjunctive intravascular imaging, including intravascular ultrasound (IVUS) or optical coherence tomography (OCT), to characterize coronary plaque morphology and classify plaque vulnerability. Participants were subsequently followed for 6 months to ascertain the occurrence of major adverse cardiovascular events (MACE).

The study protocol was reviewed and approved by the Ethics Committee of Puyang Oilfield General Hospital, and written informed consent was obtained from all participants in accordance with the Declaration of Helsinki.

Inclusion criteria were as follows:

Exclusion criteria were as follows:

Participants were consecutively enrolled during the predefined study period according to prespecified inclusion and exclusion criteria, without consideration of plaque phenotype. After enrollment, participants were classified into two groups based on baseline intravascular imaging findings obtained by intravascular ultrasound (IVUS) and/or optical coherence tomography (OCT): a stable plaque group and a vulnerable plaque group. Plaque vulnerability was determined after enrollment using prespecified, imaging-based morphological criteria incorporating established IVUS and OCT features of plaque instability. No matching, stratified sampling, or case–control selection was applied.

Follow-up was conducted according to a predefined schedule. At baseline (T0), demographic and clinical characteristics were collected, fasting venous blood samples were obtained for biomarker measurements, and coronary plaque phenotype was assessed using IVUS and/or OCT prior to any revascularization procedures. Participants were then prospectively followed for 6 months. At the end of follow-up (T1), the occurrence of major adverse cardiovascular events (MACE) was ascertained. MACE was defined as a composite endpoint including myocardial infarction, unplanned or urgent percutaneous coronary intervention performed for recurrent ischemia or acute coronary syndromes, cardiovascular death, or rehospitalization for cardiac causes. Planned or elective revascularization procedures based on baseline intravascular imaging findings were not considered outcome events.

The primary exposure variable of interest was the serum concentration of osteopontin (SPP1), which was measured at baseline using a standardized enzyme-linked immunosorbent assay (ELISA). Additional circulating inflammatory biomarkers included high-sensitivity C-reactive protein (hsCRP), interleukin-6 (IL-6), and matrix metalloproteinase-9 (MMP-9). Metabolic and cardiovascular risk parameters assessed at baseline comprised low-density lipoprotein cholesterol (LDL-C), high-density lipoprotein cholesterol (HDL-C), and glycated hemoglobin (HbA1c). Renal function was evaluated using the estimated glomerular filtration rate (eGFR), calculated according to standard equations.

The clinical outcomes of interest were twofold. First, coronary plaque phenotype was classified as stable or vulnerable based on intravascular imaging findings obtained by intravascular ultrasound (IVUS) or optical coherence tomography (OCT). Second, the occurrence of major adverse cardiovascular events (MACE) during the 6-month follow-up period was recorded. MACE was defined as a composite endpoint including myocardial infarction, percutaneous coronary intervention, cardiovascular death, or rehospitalization for cardiac causes.

MACE were defined as a composite endpoint including myocardial infarction, unplanned or urgent percutaneous coronary intervention performed for recurrent ischemia or acute coronary syndromes, cardiovascular death, or rehospitalization for cardiac causes. Planned or elective revascularization procedures based on baseline intravascular imaging findings were not considered outcome events.

Intravascular ultrasound (IVUS) was used to assess plaque burden, vessel remodeling patterns, and the presence of echolucent or echo-attenuated regions suggestive of lipid-rich plaque components. Optical coherence tomography (OCT) was applied to evaluate fibrous cap thickness, lipid-rich plaque characteristics, plaque surface morphology, and the presence of intracoronary thrombus.

Imaging-defined plaque phenotype (stable versus vulnerable) was determined using prespecified, imaging-based morphological criteria derived from established intravascular imaging literature (31, 32). For OCT, vulnerable plaque was defined by the presence of thin-cap fibroatheroma features, characterized by a lipid-rich plaque with a minimum fibrous cap thickness below established thresholds, and/or other high-risk features such as plaque rupture or intracoronary thrombus. For IVUS, surrogate markers of plaque vulnerability included large plaque burden, adverse (positive) remodeling patterns, and echolucent or attenuated regions consistent with lipid-rich plaque. When both modalities were available, OCT was prioritized for cap-based vulnerability assessment given its higher axial resolution; IVUS-based criteria were applied when OCT data were unavailable.

All intravascular imaging data were independently reviewed by two experienced cardiovascular imaging specialists who were blinded to clinical characteristics and biomarker measurements. In cases of disagreement, consensus was achieved through joint review, with adjudication by a third senior reviewer when necessary. Inter-observer reproducibility for plaque classification and key imaging measurements was assessed using appropriate agreement statistics and is reported in the Supplementary Material.

Fasting venous blood samples were collected at baseline upon hospital admission and prior to intravascular imaging and any coronary revascularization procedures, and serum was separated and stored at −80 °C until analysis. In patients presenting with acute coronary syndrome, blood sampling was performed after initial clinical stabilization but before coronary intervention, whereas in patients with stable coronary artery disease, samples were obtained under fasting conditions at admission before coronary angiography.

Serum levels of SPP1, IL-6, and MMP-9 were quantified using standardized commercial enzyme-linked immunosorbent assay (ELISA) kits according to the manufacturers’ instructions. Serum hsCRP was measured using an immunoturbidimetric assay, while other routine biochemical parameters were analyzed in the hospital’s central laboratory using automated analyzers. To ensure analytical reliability, all biomarker measurements were performed in duplicate, and results with a coefficient of variation of less than 10% were considered acceptable.

Continuous variables were summarized as mean ± standard deviation (SD) for normally distributed data and as median with interquartile range (IQR) for non-normally distributed data. Categorical variables were expressed as frequencies and percentages. Between-group comparisons (stable versus vulnerable plaques) were performed using Student’s t test or the Mann–Whitney U test for continuous variables, as appropriate, and the chi-square test for categorical variables.

Associations between serum SPP1 levels and other inflammatory biomarkers (MMP-9, IL-6, and hsCRP) were assessed using Pearson or Spearman correlation coefficients according to data distribution. Logistic regression models were constructed to evaluate the association between serum SPP1 and imaging-defined plaque vulnerability. Receiver operating characteristic (ROC) curve analyses were performed to compare the discriminatory performance of SPP1 with other inflammatory biomarkers, with optimal cutoff values determined using the Youden index.

Covariates for multivariable logistic and Cox regression models were pre-specified a priori based on established clinical relevance to coronary plaque vulnerability and cardiovascular events, with consideration of model parsimony and the risk of overfitting. The core adjustment set included age, sex, and LDL-C for logistic regression models, and age, sex, hypertension, and LDL-C for Cox models. IL-6 was additionally included to account for systemic inflammation. Automated, data-driven variable selection procedures were not applied.

Collinearity among inflammatory biomarkers was formally assessed prior to multivariable modeling using variance inflation factors (VIFs) and tolerance statistics. A VIF value > 5 or a tolerance < 0.20 was considered indicative of potentially problematic collinearity. When SPP1 and MMP-9 were simultaneously entered into multivariable logistic regression models, both variables exceeded these thresholds, indicating substantial shared variance. Accordingly, MMP-9 was excluded from the primary multivariable logistic regression models to avoid model instability and overadjustment, while SPP1 was retained as the primary exposure variable of interest.

For longitudinal analyses, Cox proportional hazards regression models were used to assess the association between baseline serum SPP1 levels and the occurrence of major adverse cardiovascular events (MACE) during the 6-month follow-up period. Kaplan–Meier survival curves were generated to compare MACE-free survival between groups stratified by serum SPP1 levels, and differences were assessed using the log-rank test.

Mediation analyses were conducted using a nonparametric bootstrap approach with 1,000 resamples to explore the potential mediating role of MMP-9 in the association between SPP1 and plaque vulnerability. These analyses were considered exploratory and hypothesis-generating. Prespecified subgroup analyses were performed according to sex, age, diabetes, and hypertension status.

To improve clinical interpretability, serum SPP1 was additionally standardized and analyzed per standard deviation (SD) increase. Categorical analyses based on quartiles of the SPP1 distribution were also performed as sensitivity analyses. Additional sensitivity analyses included alternative modeling strategies such as log-transformation of SPP1, exclusion of outliers, and multiple imputation for missing data to assess the robustness of the findings.

All statistical analyses were performed using appropriate statistical software, and a two-sided P value < 0.05 was considered statistically significant.

Baseline characteristics of the study population are summarized in Table 1. A total of 300 patients were included, with 150 classified as having stable plaques and 150 as having vulnerable plaques based on IVUS/OCT findings. Patients in the vulnerable plaque group were significantly older (65.50 ± 7.15 vs. 61.34 ± 7.54 years, P < 0.001) and had fewer years of formal education (9.08 ± 3.22 vs. 10.18 ± 3.23 years, P = 0.003). Body mass index was higher in the vulnerable group (27.02 ± 3.30 vs. 25.37 ± 3.10 kg/m², P < 0.001), and disease duration was also longer (6.22 ± 2.56 vs. 4.70 ± 2.00 years, P < 0.001).

With respect to laboratory parameters, patients with vulnerable plaques had higher levels of LDL cholesterol (3.17 ± 0.57 vs. 2.80 ± 0.50 mmol/L, P < 0.001), hsCRP (3.03 ± 1.19 vs. 2.11 ± 0.89 mg/L, P < 0.001), MMP-9 (395.55 ± 61.18 vs. 296.69 ± 53.62 ng/mL, P < 0.001), IL-6 (6.91 ± 1.71 vs. 4.84 ± 1.47 pg/mL, P < 0.001), and HbA1c (6.38 ± 0.67 vs. 5.92 ± 0.62%, P < 0.001), while HDL cholesterol was significantly lower (1.01 ± 0.21 vs. 1.11 ± 0.28 mmol/L, P = 0.002). Estimated GFR did not differ significantly between groups. Regarding medication use, there were no significant differences in the proportions of patients receiving statins (56.7% vs. 53.3%) or ACEi/ARB therapy (64.0% vs. 58.7%). The mean SPP1 concentration in the vulnerable group was 55.85 ± 12.26 ng/mL, whereas in the stable group it was 39.18 ± 9.42 ng/mL (Supplementary Figure 1).

During the 6-month follow-up period, a total of 36 major adverse cardiovascular events (MACE) were documented, including 8 myocardial infarctions, 14 unplanned or clinically driven percutaneous coronary interventions, 3 cardiovascular deaths, and 11 cardiac-related rehospitalizations. A detailed breakdown of event types is provided in Supplementary Table S1.

In multivariable linear regression analyses adjusting for age, sex, and LDL-C, serum SPP1 levels were independently associated with MMP-9 (β = 0.71, 95% CI: 0.64–0.78; P < 0.001), IL-6 (β = 0.42, 95% CI: 0.34–0.50; P < 0.001), and, to a lesser extent, hsCRP (β = 0.08, 95% CI: 0.02–0.14; P = 0.009) (Table 2). Correlation analyses demonstrated that serum SPP1 levels were strongly and positively associated with inflammatory mediators (Supplementary Figure 2).

In univariate logistic regression analyses, higher serum levels of SPP1, MMP-9, and IL-6 were each significantly associated with increased odds of imaging-defined plaque vulnerability (all P < 0.001), whereas hsCRP and LDL-C were not significantly associated with plaque vulnerability (P = 0.091 and P = 0.824, respectively) (Table 3).

In the multivariable logistic regression model adjusted for age, sex, LDL-C, and IL-6, serum SPP1 remained independently associated with plaque vulnerability (OR 1.08, 95% CI 1.05–1.11; P < 0.001). LDL-C also showed a modest but statistically significant association (OR 1.36, 95% CI 1.03–1.80; P = 0.032), whereas IL-6 was no longer statistically significant after adjustment (P = 0.122). Age and sex were not independently associated with plaque vulnerability in the adjusted model (both P > 0.05).

In expanded multivariable logistic regression models additionally adjusting for major cardiometabolic risk factors and medication use, including BMI, diabetes mellitus, current smoking status, HDL-C, HbA1c, statin use, and ACEi/ARB use, the association between serum SPP1 and imaging-defined plaque vulnerability remained robust and statistically significant (Supplementary Table S1).

Cox proportional hazards models were used to evaluate the prognostic value of serum biomarkers for 6-month MACE (Table 4). In univariate analyses, higher levels of SPP1, MMP-9, and IL-6 were significantly associated with increased risk of MACE. Specifically, SPP1 showed the strongest association (HR 1.47, 95% CI: 1.23–1.76, P < 0.001), followed by MMP-9 (HR 1.31, 95% CI: 1.11–1.55, P = 0.001) and IL-6 (HR 1.25, 95% CI: 1.05–1.50, P = 0.011). hs-CRP was not predictive of MACE (P = 0.262). In the multivariate model adjusting for age, sex, and hypertension, SPP1 remained an independent predictor of 6-month MACE (HR 1.35, 95% CI: 1.11–1.64, P = 0.002). In contrast, the associations of MMP-9 and IL-6 with MACE were attenuated and no longer statistically significant (both P>0.05).

In expanded multivariable Cox proportional hazards models additionally adjusting for major cardiometabolic risk factors and medication use, including BMI, diabetes mellitus, current smoking status, HDL-C, HbA1c, and statin use, baseline serum SPP1 remained independently associated with an increased risk of 6-month major adverse cardiovascular events (Supplementary Table S2).

When serum SPP1 was re-parameterized to improve clinical interpretability, analyses based on standard deviation (SD) increase and distribution-based quartiles yielded results consistent with the primary per-unit models. In quartile-based analyses, compared with the lowest SPP1 quartile, participants in the highest quartile exhibited markedly increased risks of plaque vulnerability (OR = 2.74, 95% CI 1.78–4.23, P < 0.001) and 6-month MACE (HR = 2.36, 95% CI 1.38–4.04, P = 0.002), with a clear graded increase in effect estimates across higher quartiles. These findings indicate a dose–response relationship between serum SPP1 levels and both plaque instability and short-term cardiovascular risk (Supplementary Table S3).

To address potential confounding by renal function, sensitivity analyses were performed with additional adjustment for baseline estimated glomerular filtration rate (eGFR). In logistic regression models evaluating imaging-defined plaque vulnerability, the association between serum SPP1 levels and plaque vulnerability remained statistically significant after inclusion of eGFR in the multivariable model (OR per unit increase, 1.07; 95% CI, 1.04–1.10; P < 0.001), with minimal attenuation compared with the primary model. Baseline eGFR itself was not independently associated with plaque vulnerability (P = 0.318).

Similarly, in Cox proportional hazards models for 6-month major adverse cardiovascular events, additional adjustment for eGFR did not materially alter the association between SPP1 and incident MACE (HR per unit increase, 1.33; 95% CI, 1.09–1.62; P = 0.004). Renal function was not a significant predictor of MACE in the fully adjusted model (P = 0.176). (Supplementary Table S4).

Receiver operating characteristic (ROC) curve analysis was performed to compare the diagnostic performance of serum biomarkers in discriminating vulnerable from stable plaques (Figure 1; Table 5). Among all tested markers, SPP1 demonstrated the highest diagnostic accuracy, with an area under the curve (AUC) of 0.875 (95% CI: 0.837–0.913). In comparison, IL-6 (AUC = 0.786, 95% CI: 0.736–0.836) and MMP-9 (AUC = 0.729, 95% CI: 0.673–0.786), while hs-CRP demonstrated poor discrimination (AUC = 0.559, 95% CI: 0.491–0.625). Pairwise comparisons of ROC curves using DeLong’s test demonstrated that the AUC of SPP1 was significantly higher than those of IL-6, MMP-9, and hsCRP (Supplementary Table S5).

Kaplan–Meier survival curves demonstrated a significant difference in cumulative MACE-free survival between patients stratified by serum SPP1 levels (Figure 2). Patients were dichotomized into high and low SPP1 groups based on the median serum SPP1 concentration of the study population. For Kaplan–Meier analyses, patients were stratified into high and low SPP1 groups based on the median serum SPP1 concentration of the overall study population (median value: 47.0 ng/mL). During the 6-month follow-up period, patients in the high SPP1 group experienced a higher incidence of MACE compared with those in the low SPP1 group. Separation of event-free survival curves was observed early during follow-up and became more pronounced over time, with MACE-free survival declining to below 70% in the high SPP1 group by the end of follow-up. The difference between groups was statistically significant according to the log-rank test (P = 0.004).

To further explore the mechanistic pathway linking SPP1 to plaque rupture, a mediation analysis was conducted using the bootstrap method (Table 6). The total effect of SPP1 on plaque rupture was significant (estimate = 0.423, 95% CI: 0.271–0.576, P < 0.001). Of this, an indirect effect mediated through MMP-9 accounted for 0.187 (95% CI: 0.072–0.318, P < 0.01), representing 44.2% of the total effect. The direct effect of SPP1 on plaque rupture remained significant (estimate = 0.236, 95% CI: 0.065–0.407).

The association of higher serum SPP1 with plaque rupture and 6-month MACE was consistent across sex, age, diabetes, hypertension, LDL, statin use, smoking, and BMI subgroups. Effect sizes were comparable (ORs 1.55–1.68; HRs 1.38–1.49) with no significant interactions (all P-interaction >0.3) (Supplementary Tables S6, S7). Results remained robust when SPP1 was modeled as log-transformed, dichotomized at the ROC cutoff or median, after trimming outliers or imputing missing data. Adjustment for MMP-9 and IL-6 attenuated but did not abolish the association (OR 1.38, P = 0.006) (Supplementary Table S8). SPP1 consistently predicted 6-month MACE under multiple specifications, including log-transformation, dichotomization, statin-restricted models, IPTW, and multiple imputation. Alternative endpoint definitions yielded similar results, and proportional hazards assumptions were satisfied (Supplementary Table S9).