Validation of circulating ANXA3 and SOCS3 expression as potential diagnostic biomarkers for acute myocardial infarction: a prospective case-control study

Cardiac troponin is less sensitive within the first few hours of AMI. Both metabolic stress and immune-inflammatory responses have been proposed to contribute to early AMI pathology. This study aimed to validate whether ANXA3 and SOCS3, identified from our earlier bioinformatic work, could serve as early diagnostic blood biomarkers for AMI. 300 AMI patients (symptom onset ≤ 24 h) and 150 matched controls were enrolled in this single-center prospective study. ANXA3 and SOCS3 mRNA were measured by RT-qPCR. ROC analysis was used for individual markers; an ANN model combined them with 5-fold cross-validation. ANXA3 was 2.7-fold higher in AMI patients and SOCS3 2.3-fold higher (both P < 0.001). AUC for ANXA3 was 0.879, for SOCS3 0.839. The primary adjusted logistic regression model had an apparent AUC of 0.952 (95% CI 0.930–0.974), with a 5-fold cross-validated mean AUC of 0.948 (CV 0.7%); this result indicates internal consistency within the current dataset, and generalizability requires validation in independent external multi-center cohorts. As a secondary exploratory sensitivity analysis, the ANN achieved an AUC of 0.938 (95% CI 0.915–0.962). In patients presenting within 4 h, the ANN AUC was 0.938; in those with hypertension or diabetes, AUCs were 0.940 and 0.946, provided for descriptive purposes only. Comparisons with cTnI (AUC 0.917) are limited because cTnI was measured from fasting blood collected the morning after admission, whereas ANXA3 and SOCS3 were measured from blood collected upon emergency department arrival; owing to mismatched sampling time-points, direct head-to-head performance comparisons between these markers cannot be inferred, and all AUC values are presented for descriptive purposes only. ANXA3 and SOCS3 are upregulated in AMI. The primary logistic regression model had an apparent AUC of 0.952 (95% CI 0.930–0.974), with a 5-fold cross-validated mean AUC of 0.948 (CV 0.7%); this result indicates internal consistency within the current dataset, and generalizability requires validation in independent external multi-center cohorts. The exploratory ANN yielded an AUC of 0.938. Subgroup AUC values are descriptive only, and comparisons with cTnI are limited by different sampling time-points. The control group consisted of healthy volunteers, and external validation is lacking. Larger prospective studies in undifferentiated chest pain populations are needed. ChiCTR2600128224. Registered 16 July 2026 (Retrospectively registered).