Shuaiwei Guo, Shengyan Cui, Xiao Zhang, Zixuan Xing, Yixin Sun, Mingqian Qin, Xiaoying Ma, Xiangyu Li, Yan Ma, Peng Gao, Yanfei Chen, Yabing Wang, Bin Yang, Jian Chen, Fei Chen, Xia Lu, Jiaqi Jin, Liqun Jiao, Tao Wang
Circulating AGRN and ADAMTS8, measured as serum concentrations, are associated with carotid plaque presence versus healthy controls but do not independently discriminate plaque presence or instability after multivariable adjustment. Clinical covariates alone (age, sex, vascular risk factors) provided near-perfect discrimination of plaque presence, indicating substantial spectrum bias arising from the healthy-control comparator design. These findings underscore the need for age- and risk-matched comparator populations in future biomarker studies, consistent with STARD 2015 recommendations.
BACKGROUND: Carotid atherosclerosis is a major etiologic substrate for ischemic stroke. Although imaging can characterize stenosis and plaque morphology, circulating biomarkers may better support screening and scalable risk stratification, yet reliable serum protein markers for carotid plaque, particularly for plaque instability, remain limited.
METHODS: We enrolled 128 participants, including patients with stable plaques (n = 41), unstable plaques (n = 49), and healthy controls (n = 38). Plaque stability was determined by carotid ultrasonography. We prespecified a serum biomarker panel comprising seven proteins and measured all of them: agrin (AGRN), ADAMTS8, ADAMTS9, collagen type XVIII alpha 1 chain (COL18A1), integrin subunit beta 4 (ITGB4), lysyl oxidase like 4 (LOXL4), and slit guidance ligand 3 (SLIT3). All proteins were quantified by enzyme-linked immunosorbent assay (ELISA). Group comparisons were performed with multiple-comparison adjustment. Receiver operating characteristic (ROC) analyses and the area under the curve (AUC) evaluated the ability of individual biomarkers and the combined prediction probability to discriminate plaque presence and plaque instability.
RESULTS: Baseline characteristics differed substantially between plaque groups and healthy controls. Serum AGRN concentrations were significantly elevated in plaque patients versus controls (mean ± SD, Stable: 2222.42 ± 508.30 pg/mL; Unstable: 2109.59 ± 513.51 pg/mL; Healthy: 1800.36 ± 375.89 pg/mL; overall p < 0.001), while ADAMTS8 levels were comparable across groups (overall p = 0.110). In multivariable Firth penalized logistic regression with log2-transformed biomarkers, neither AGRN nor ADAMTS8 was independently associated with plaque presence or plaque instability after adjustment for age, sex, hypertension, alcohol use, dyslipidemia, smoking, diabetes mellitus, and coronary heart disease (all adjusted p > 0.05). For discriminating plaque presence, the combined AGRN+ADAMTS8 model achieved an apparent AUC of 0.771 (bootstrap 95% CI, 0.678-0.865; optimism-corrected AUC, 0.764; repeated stratified 5-fold cross-validated AUC, 0.749). For plaque instability, the combined model showed no discriminatory capacity (apparent AUC, 0.549; optimism-corrected AUC, 0.498; repeated 5-fold cross-validated AUC, 0.471).
CONCLUSIONS: Circulating AGRN and ADAMTS8, measured as serum concentrations, are associated with carotid plaque presence versus healthy controls but do not independently discriminate plaque presence or instability after multivariable adjustment. Clinical covariates alone (age, sex, vascular risk factors) provided near-perfect discrimination of plaque presence, indicating substantial spectrum bias arising from the healthy-control comparator design. These findings underscore the need for age- and risk-matched comparator populations in future biomarker studies, consistent with STARD 2015 recommendations.