Insulin-like growth factor binding protein-3 (IGFBP-3): a biomarker of coronary artery disease induced myocardial ischaemia

Abstract

Aims: Among individuals presenting to the emergency department (ED) with chest pain, clinical uncertainty surrounds the appropriate identification of non-myocardial infarction (MI) individuals who would most benefit from objective functional/anatomical testing (e.g. imaging). We applied a proteomic biomarker discovery approach to identify novel candidates reflecting coronary artery disease (CAD) induced ischaemia that could translate to measurement in clinical samples.

Methods and results: Mass spectroscopy (MS) of perfusate from an isolated rat heart model of cardiac ischaemia identified >100 novel protein biomarkers. A prominent candidate, insulin-like growth factor binding protein (IGFBP-3), was then interrogated for its ability to identify CAD-related ischaemia (e.g. positive cardiac stress test; unstable angina pectoris, UAP; arterial stenosis >70% on angiography) in multiple patient sample sets [cardiac stress testing, n = 12; septal alcohol ablation (SAA), n = 12; ED chest pain, n = 2977]. In cardiac stress testing, a significant delta IGFBP-3 (ΔIGFBP-3) between 0 and 150 min was seen in positive, but not negative, tests (P = 0.03). In SAA, peripheral IGFBP-3 levels did not change over 24 h (P = 0.57). In ED patients, ΔIGFBP-3 between 0 and 2 h (i) identified more 365-day low-risk major adverse cardiac event cases (27-30%), (ii) provided 7% improvement in positive predictive value over a clinical model for the identification of unstable angina (P = 0.01), and (iii) was a significant, independent predictor of >70% stenosis on angiography, improving indeterminate risk prediction by 9% (95% CI 3-15%).

Conclusion: Our discovery approach has translated IGFBP-3 as a potential biomarker to identify significant CAD/ischaemia in patients who do not meet diagnostic thresholds for MI.

Keywords: Acute coronary syndromes; Biomarker; CAD; IGFBP-3; Imaging; Troponin.

Graphical Abstract

Figure 1

(A) Schematic of the experimental ischaemia marker discovery/test process used herein. Perfusate obtained from ischaemic and control isolated rat hearts was subjected to unbiased tandem MS/MS proteomic interrogation. This process identified 148 protein factors as a novel (i.e. presence in ischaemic hearts but not controls) of which insulin-like growth factor binding protein-3 emerged as a readily testable candidate which was then investigated in human experimental induced ischaemia (cardiac exercise stress testing). (B) Molecular weight ranges and cellular distributions of the 148 proteins identified in trypsin-digested isolated perfused rat heart perfusates at t = 10 min. Only 6% of all identified proteins were predicted to be secreted products, with over half of the proteins between 40 and 100 kDa in size.

Figure 2

(A) Mean (horizontal bars) and individual circulating levels of insulin-like growth factor binding protein-3 in control stress test negative (n = 6, black) and CAD stress test positive (n = 6, red) cases across 240 min. Stress testing was initiated at t = 0 using a standard Bruce protocol with echo imaging to confirm ischaemia. (B) Mean 0–150 min (delta) insulin-like growth factor binding protein-3 (ΔIGFBP-3) changes in CAD stress test positive cases (CAD positive EST) were negative, compared with positive in control negative EST cases at 150 min (P = 0.026).

Figure 3

(A) Mean (blue line) and individual insulin-like growth factor binding protein-3 (upper panel) and hs-TnT (lower panel) concentrations across multiple arterio-venous organ beds (n = 14). FA1, entry arterial; FA2, exit arterial; FV, femoral vein; RV, renal vein; HV, hepatic vein; IVC, inferior vena cava; JUG, jugular; CS, cardiac coronary sinus; PA, pulmonary artery. Due to the time taken for sampling and to account for cannulation effects, arterio-venous comparisons were made with time-comparable sites, i.e. CS vs. FA2 and RV/HV vs. FA1. Extraction was noted across both CS and HV sites for insulin-like growth factor binding protein-3, whereas for hs-TnT, there was production across the heart and extraction across the kidney. Red ** P < 0.05, Ψ = P < 0.01. (B) Mean insulin-like growth factor binding protein-3 (upper panel) and hs-TnT (lower panel) levels in response to septal alcohol ablation (n = 12). Whereas IGFBP-3 levels showed no real change, hs-TnT levels rose significantly from 30 min post-ablation (P = 0.01, red Ψ) and continued to rise out to 1440 min (24 h).

Figure 4

Assessment of the impact of ΔIGFBP-3 in the emergency department upon the No Objective Testing rule calculations in all emergency department studies combined. A total of 1419 patients were eligible for the No Objective Testing rule inclusion, of which the No Objective Testing rule calculations determined n = 381 (26.8%) were low risk with a 365-day major adverse cardiac event rate of 1.0%. The addition of a positive 0–2 h ΔIGFBP-3 increased the absolute number (+43) and proportion of true low-risk cases to n = 424 (29.9%), decreasing the 365-day major adverse cardiac event rate to 0.9%.

Figure 5

Receiver operating characteristic curves generated by logistic regression models for the diagnosis of unstable angina pectoris in (A) all No Objective Testing rule patients (n = 1695 cases, 123 unstable angina pectoris) and (B) where the 0–2 h hs-TnT maximum was between 5 and 14 ng/L (n = 1180 cases, 119 unstable angina pectoris). In both scenarios, a positive ΔIGFBP-3 provided a significant addition to receiver operating characteristic curves (0.012 and P = 0.05, respectively) containing a clinical model comprising age, sex, Hx of Cholesterol, Hx CAD/MI, ECG, and Log10 hs-TnT maximum. ΔIGFBP-3 had the most impact at high specificity through improved sensitivity.

Figure 6

(A) Receiver operating characteristic curves and derived data for the findings of 70% stenosis (n = 112) on coronary angiography in all cases where 0–2 h hs-TnT < 14 ng/L (n = 221). The addition of ΔIGFBP-3 improved the AUC of the clinical + hs-TnT curve by 0.033 points (P = 0.07). (B) In pathways, analyses of cases in which hs-TnT at t = 0 and t = 2 h were both <14 ng/L, inclusion of ΔIGFBP-3 for the detection of 70% coronary stenosis (n = 219 total) improved the clinical + hs-TnT based identification of low-risk cases by an absolute 8% (n = 52, 24% to n = 69, 32%), all of which had no 70% stenosis. This was due to improved specificity of 38–54% at 90% sensitivity.