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.

Graphical Abstract

Open in new tabDownload slide

Introduction

Diagnostic protocols that use high-sensitivity cardiac troponin (hs-cTn) testing are the standard of care for the Emergency Department (ED) assessment of patients presenting with suspected acute coronary syndrome (ACS).^1–4 A significant proportion of these patients in whom myocardial infarction (MI) is excluded (or ‘ruled out’) have underlying coronary artery disease (CAD) that may lead to adverse cardiac events soon after evaluation. Current guidelines recommend objective functional (stress) or anatomical (imaging) testing for individuals in this group to determine individual patient risk; however, uncertainty remains as to deciding which patients truly benefit from such objective testing. Score-based triaging, e.g. the No Objective Testing (NOT) rule,^5–7 and consideration of incremental gradients in sub-threshold hs-cTn levels,^8,9 have been suggested to improve the identification of low-risk individuals. These approaches still leave a sizeable proportion of ED patients (up to one-third) who are not ‘low risk’, have potentially significant CAD, and may still experience post-discharge unstable angina pectoris (UAP).^10,11 This group of patients represents a major unmet need for biomarker support to aid decision-making beyond the current use of clinical scoring and troponin.

In an effort to identify candidate biomarkers indicative of myocardial ischaemia, we utilized a discovery platform comprising ischaemic isolated heart perfusate, coupled with mass spectrometry (MS) proteomic analyses. In doing so, we identified insulin-like growth factor binding protein-3 (IGFBP-3) as a novel candidate marker of myocardial ischaemia. We then sought to determine the potential of IGFBP-3 to identify myocardial ischaemia by the interrogation of circulating levels in (i) stress testing induced ischaemia, (ii) septal ablation procedures (alcohol-induced myocardial necrosis), (iii) trans-organ gradients obtained during clinical catheterization, and (iv) patients presenting to a hospital ED with suspicion of ACS.

Methods

Isolated heart procedure

Isolated heart experiments were carried out in conformity with the ARRIVE guidelines, after animal ethics committee approval (AUP 21-178). Isolated perfused rat heart (IPRH) preparations were carried out as previously described^12,13 (see Supplementary Methods). Briefly, ischaemia was induced by reversible and sham ligation of the left main coronary artery in isolated hearts. Perfusate was collected at −10 min prior to, 0, +10, +20, and +30 min after ligation. Reversible and sham perfusates were subjected to unbiased HPLC/MS to detect differences in protein expression between the two sample types (Figure 1A and Supplementary Methods).

Insulin-like growth factor binding protein-3 and high-sensitivity cardiac troponin assays

Full IGFBP-3 assay details and its analytical validation are given in the Supplementary Methods. Briefly, IGFBP-3 levels were determined using a two-antibody sandwich ELISA on plasma samples diluted 1:100 with phosphate-buffered saline (PBS)/albumin assay buffer (pH 7.2), with samples corrected for dilution in all calculations (i.e. plate values were multiplied by 100×). The IGFBP-3 assay has a limit of detection (LoD) of 0.37 ng/mL, a 10% coefficient of variation (CV), a limit of quantification (LoQ) of 1.56 ng/mL, and an inter-assay CV of 8.9 and 7.9% at 6.5 and 23.9 ng/mL, respectively, over 38 consecutive assays. Incurred sample reanalyses (ISRs) revealed 94.5% of tested samples to be within the required limit of 30% of the mean.¹⁴

Cardiac troponin T (cTn) was measured in all studies using the Roche Cobas e411 assay (hs-cTnT, Roche, Rokruetz, Switzerland) with a limit of blank (LoB) = 3 ng/L, LoQ = 5 ng/L, and a single upper European reference limit (URL) = 14 ng/L. cTnI was measured in APACE and SPACE by the Abbott ARCHITECT (Abbott Park, IL) with a LoB = 0.7 ng/L, LoQ = 3 ng/L, and an overall URL of 26 ng/L. cTnI was measured in the FAST-TRAC study using the Beckman Coulter (Beckman Coulter, Inc, Chaska, MN, USA) assay with a LoB = 0.33 ng/L, LoQ = 3 ng/L, and an overall URL of 18.2 ng/L.

Clinical studies/plasma sample collection

All human studies were conducted in conformity with the Declaration of Helsinki and all participants gave written informed consent. All data were analysed in conformity with STARD guidelines. Human EDTA plasma samples were obtained from multiple studies as follows: individuals undergoing cardiac exercise stress testing [Signal Peptides in Induced Cardiac Ischaemia (SPICI) study, Christchurch, NZ, n = 12], patients undergoing septal ablation treatment^15,16 [SEARCH study, Christchurch, NZ, n = 12], and regional plasma samples from patients undergoing clinically indicated cardiac catheterization from whom matched arterio-venous samples were drawn to obtain trans-organ (venous minus arterial levels of markers) gradients^17,18 [GRADIENT study, Christchurch, NZ, n = 14], and three separate prospective, observational hospital ED studies of patients presenting with chest pain suspicious of ACS [APACE study, international (n = 1056),^11,19 FAST-TRAC, USA (n = 1230),²⁰ and SPACE, Christchurch, New Zealand (n = 766)].^21,22

SPICI study

This study consisted of healthy control individuals (n = 6), and participants with known CAD (n = 6), each undergoing treadmill exercise stress testing as previously described for assessment of possible inducible cardiac ischaemia.²³ The study was approved by the Health and Disability Ethics Committee (HDEC) of New Zealand (URB/10/03/010). Full details of the protocol are given in the Supplementary Methods.

SEARCH study

Twelve patients with hypertrophic cardiomyopathy (HOCM),²⁴ and scheduled for septal alcohol ablation (SAA) from January 2021 to April 2023, were consented and enrolled in this study, approved by the HDEC of New Zealand (14STH73). Details of the protocol are given in the Supplementary Methods. We collected venous blood samples for determination of plasma biomarkers in EDTA-filled tubes 10 min before the procedure and at 15, 30, 45, 60, 120, 240, and 1440 min (24 h) after induction of myocardial necrosis.

GRADIENT study

Patients (n = 14) undergoing cardiac catheterisation for diagnostic coronary angiography or percutaneous intervention were enrolled and gave written consent for additional blood sampling as previously described.^25,26 The study was approved by the HDEC of New Zealand (URB/06/02/010). The right femoral artery and vein were cannulated with 6 Fr sheaths under local anaesthesia. A multipurpose, diagnostic catheter (6 Fr), was advanced via the femoral vein to draw sequential blood samples from femoral, renal, hepatic and jugular veins, and the pulmonary artery. An Amplatz left-1 (6 Fr) catheter was employed to sample from the cardiac coronary sinus. Blood was drawn from the femoral artery before and after the completion of venous sampling to assess the stability of peptide levels during the time required for sampling. Intravenous saline up to 500 mL was administered through the procedure. Blood samples were collected in EDTA and centrifuged within 20 min at 3000 g and 4°C. Plasma was stored at −80°C.

ED studies

To assess the potential diagnostic/prognostic biomarker ability of IGFBP-3 in patients with possible myocardial ischaemia, we interrogated plasma samples obtained from ED patients presenting with chest pain suspicious of ACS from three separate studies (APACE, FAST-TRAC, and SPACE), all with similar recruitment criteria. Patient demographics and heterogeneity are given in Table 1, and details of recruitment criteria are in the Supplementary Methods. In all three ED studies, blood samples were drawn at presentation +1 and +2 h.

Primary endpoints

The primary endpoints for analyses among patients presenting with possible ACS were the diagnosis of UAP, the angiographic finding of obstructive CAD ≥70% in at least one coronary artery, qualification for NOT rule inclusion, and major adverse cardiac events (MACEs; defined as new ACS or CV death) within 365 days of presentation (including any index event).

Adjudication of final diagnoses

Adjudication of the final diagnosis was performed centrally for each study individually by two independent cardiologists, with a third serving as a tiebreaker in the case of diagnostic disagreement. The gold standard diagnosis was determined by a review of all available medical records through locally applied follow-up.

STEMI, NSTEMI, and UAP were defined per the fourth Universal Definition of MI.²⁷ NSTEMI was diagnosed when there was evidence of acute myocardial necrosis (i.e. a rise and/or fall in cTn, with at least one measurement >99th percentile) in a clinical presentation consistent with ACS but without ST-segment elevations on ECG. In cases that included significant new and persistent ST-segment elevations on ECG together with myocardial necrosis, STEMI was diagnosed. The following criteria were used to indicate an increased likelihood of a diagnosis of UA: typical angina pectoris at rest; worsening/deterioration of previously stable angina; cardiac stress test showing myocardial ischaemia; coronary angiography revealing a diameter stenosis of at least 70%; fractional flow reserve documenting the functional significance of a coronary lesion; and sudden cardiac death or MI occurred during follow-up.^17,28

NOT rule calculation and 365-day MACE assessment

Allocation of patients into NOT rule categories was carried out in accordance with previous reports.^5–7 Briefly, patients were considered eligible for NOT rule inclusion if MI was ruled out according to current protocols, defined as hs-cTn levels at 0–2 h below thresholds (<14 ng/L Roche hs-TnT) with no ischaemic changes on the ECG. The NOT rule was only calculated on hs-TnT as that is the only consistent hs-cTn assay across all three studies. Eligible NOT rule patients were then classified into ‘low risk’ (<3 risk factors, age <50 years) and ‘not low risk’ (history of CAD or MI) as previously descibed.²⁸ MACE was defined as ACS (MI or UAP) or cardiovascular death within 365 days of presentation, including index events.

Statistics

Experimental IPRH hemodynamic data are presented as mean ± standard error of the mean. Results are presented as median with interquartile range (IQR). All statistical analysis was carried out using SPSS (IBM/SPSS, v29). Categorical variables were described as percentages. Relational analysis of plasma protein/peptide concentrations was done using Spearman rank order correlation testing and linear regression analysis. The t-test, Mann–Whitney, and χ² tests were used as appropriate to compare patient characteristics. In ED patient studies, binary logistic regression was used to build multivariable clinical and biomarker models that could be incorporated into receiver operating characteristic (ROC) curves to investigate the usefulness of each variable and their combinations for the diagnosis of UAP or the presence of 70% coronary stenosis. Base models were created using clinical characteristics and hs-TnT. IGFBP-3 and ΔIGFBP-3 were then added to the model to evaluate their ability to improve the ROC characteristics [AUC, specificity, sensitivity, positive predictive value (PPV), and NPV]. Pathways and decision curve analyses were used to determine patient movement across risk profiles that were attributable to IGFBP-3.

Results

Identification of IGFBP-3 in ischaemic isolated rat heart perfusate

Coronary ligation resulted in frank changes in hemodynamics in the ischaemic isolated rat heart (IIRH) model, not seen in sham/control hearts (see Supplementary material online, Figure S1A). MS analysis of trypsin-digested isolated heart perfusates identified 148 unique proteins including IGFBP-3 (Figure 1A) as present in samples taken at 10 min post-ligation, but not in sham/control samples. The distribution of these proteins revealed that 35.8% had a cytoplasmic/cytosolic location, 25% were nuclear, 39.2% membranous/mitochondrial, and 6.1% secretory (Figure 1B). Of the secreted proteins, IGFBP-3 was identified via two high-quality, deconvoluted peptide sequences (see Supplementary material online, Figure S1B).

Response of circulating IGFBP-3 to induced cardiac ischaemia in humans

IGFBP-3, as measured by our assay, was confirmed by Western Blot as consistent with the 44–46 kDa molecular weight doublet forms that are well described (see Supplementary material online, Figure S2).^29,30 With the identification of IGFBP-3 as a responsive factor in an experimental model of cardiac ischaemia, we undertook an assessment of the response of circulating IGFBP-3 in human inducible cardiac ischaemia (patient demographics, Supplementary material online, Table S1). In healthy volunteers, with no evidence of CAD and negative stress tests, mean plasma IGFBP-3 levels across the measured time period tended to rise compared with t = 0 values (black triangles, n = 6, Figure 2A). In contrast, in patients with known CAD and positive stress tests, mean plasma IGFBP-3 levels decreased to nadirs at 90 and 150 min post-stress (red squares, n = 6, Figure 2A). Of note, the mean 0–150 min shift in IGFBP-3 (ΔIGFBP-3, defined as the 150 min value minus the 0 min value) in positive stress test cases was −407 ± 392 ng/mL vs. +183 ± 532 ng/mL in negative stress test cases (Figure 2B, Mann–Whitney U test P = 0.03). Pre-test mean hs-cTnT concentrations in positive stress test cases (22.6 ± 39 ng/L) tended to be higher than control (3.7 ± 1.8 ng/L, P = 0.27). The mean 0–150 min delta of hs-TnT (3.8 ± 3.1 ng/L) in positive EST cases was not significant compared with negatives (P = 0.87, data not shown).

Regional trans-organ levels of IGFBP-3 and response to a septal alcohol ablation model of human cardiac necrosis

In order to better define circulating changes in IGFBP-3, its interaction with the heart (and other organs), and its response to more severe cardiac ischaemia, we undertook trans-organ arterio-venous profiling of IGFBP-3. We documented IGFBP-3 temporal response across human localized alcohol-induced myocardial necrosis (demographics of both groups are in Supplementary material online, Table S2). In 14 patients undergoing cardiac catheterization with regional blood sampling, mean coronary sinus (CS) levels of IGFBP-3 were lower than paired arterial levels (1067 ± 449 ng/mL vs. 1314 ± 400 ng/mL, P = 0.01), indicating net cardiac extraction/degradation of IGFBP-3 (Figure 3A). A similar negative gradient was observed across the hepatic circulation (mean hepatic IGFBP-3 = 1194 ± 515 ng/mL vs. mean arterial = 1343 ± 508 ng/mL, P = 0.03). There was no evidence of other trans-organ gradients for IGFBP-3. Mean hs-cTnT concentrations displayed net cardiac release (13.1 ± 12.3 ng/L vs. arterial 10.1 ± 8.2 ng/L, P = 0.02) and renal extraction/degradation (mean renal = 8.6 ± 6.9 ng/L vs. arterial 10.1 ±7.6 ng/L, P = 0.002, Figure 3A).

In 12 patients undergoing SAA (a clinical ‘model’ of MI), IGFBP-3 levels were not significantly altered by the intervention, with no ΔIGFBP-3 observed from 0 to 15 min, nor out to 1440 min (24 h), after injection of ethanol into the selected septal artery (Figure 3B). In contrast, hs-cTnT levels rose markedly from t = 15 min post-ethanol (P = 0.009, t = 15 min vs. 10 min) and continued to rise out to 24 h post-procedure (Figure 3B). We do not have data on the effect of time between balloon stenosis and the introduction of ethanol.

Diagnostic performance of IGFBP-3 in acute chest pain emergency department patients with MI ruled out

Based on the results from experimental and human ischaemia studies, we investigated the potential of plasma t = 0 IGFBP-3 + ΔIGFBP-3 (defined as t = 2 h minus the t = 0 h value) measurement to impact clinical diagnoses/prediction of outcomes in ED patients presenting with chest pain, or equivalent symptoms, triggering assessment for possible ACS, with a focus on those in whom MI had been ruled out. Demographic/heterogeneity details of all patients are contained in Table 1. Neither presenting IGFBP-3 level nor ΔIGFBP-3 predicted the diagnosis of acute MI in the total dataset, including Type 1 or Type 2 MI (AUC for all MI = 0.53, n = 408 cases out of 2901). However, out of 1419 data matched NOT rule eligible cases, the addition of ΔIGFBP-3 identified more low-risk cases, increasing from n = 381 (26.8%) to n = 424 (29.9%) (proportional difference 3.1%; 95% CI −0.4 to 6.4%, Figure 4).

Initial assessment in the NOT rule patients from the SPACE study alone revealed median 0–2 h ΔIGFBP-3 values to be elevated in those presenting with adjudicated UAP [n = 35, 327 (117–719) ng/mL] vs. all other diagnoses [n = 415, −112 (−380–190) ng/mL, P < 0.001]. Further, positive 0–2 h ΔIGFBP-3 values were an independent predictor of a UAP diagnosis in a multivariable regression model that included the history of CAD/MI, cholesterol, sex, and age (OR = 1.6; 1.2–2.3, P = 0.003). Expanding this analysis to NOT rule eligible patients from all three ED studies (n = 1695 total, n = 123 UAP), logistic regression modelling revealed ΔIGFBP-3 to improve the AUC of the base clinical model (age, sex, cholesterol, ECG, history of CAD/MI, and the maximum troponin T concentration) from 0.83 to 0.85 (ΔAUC 0.01, 95% CI 0.002–0.02) (Table 2). At 90% specificity, ΔIGFBP-3 improved the sensitivity of the ROC from 41.5 to 56.1%, PPV 24.6 to 30.6%, and the NPV from 95.1 to 96.3% (Figure 5A, Table 2). Similar results were obtained using the same models when the only criterion was that the maximum 0–2 h hs-cTnT concentration was between 5 and 14 ng/L (PPV at 90% specificity increased from 24.8 to 31.7%, Figure 5B, Table 2, P = 0.05). Overall, the addition of ΔIGFBP-3 to these analyses improved the PPV for UAP at 90% specificity across both assessments by 7%. The logistic regression models for both scenarios are given in Supplementary material online, Table S3 and decision curve analyses in Supplementary material online, Figure S4.

In NOT rule patients who subsequently underwent cardiac stress testing (EST, n = 679), ΔIGFBP-3 values were higher in positive [n = 66, 13.1 (−98.9–112.9) ng/mL] vs. negative [n = 613, −52.5 (−212.5–108.2) ng/mL] tests (P = 0.04). Logistic regression revealed ΔIGFBP-3 to be an independent predictor of a subsequent positive EST [OR = 1.33 (1.02–1.72), P = 0.04], adding to the combined predictive abilities of the maximum hs-TnT value [OR = 3.43 (1.11–10.6), P = 0.03], and history of CAD/MI [OR = 1.86 (1.09–3.21), P = 0.03]. This predictive effect of ΔIGFBP-3 was strongest in those who had no reported history of CAD/MI [n = 41, OR = 1.40 (1.02–1.92), P = 0.04]. In EST patients whose maximum hs-cTnT value was >5 and <14 (n = 470), the predictive ability of ΔIGFBP-3 to detect positive tests [n = 54, OR = 1.37 (0.98–1.92), P = 0.06] was less strong than hs-TnT [OR = 14.4 (1.12–186), P = 0.04] and history of CAD/MI [OR = 1.78 (0.98–3.24), P = 0.06]. In those with no history of CAD/MI in this subgroup (n = 31), neither the maximum hs-cTnT (P = 0.10) or ΔIGFBP-3 (P = 0.17) were predictive of positive tests.

In NOT rule eligible patients and in those with a maximum hs-TnT >5 and <14 (n = 154), IGFBP-3 with ΔIGFBP-3 improved the prediction of 70% coronary artery stenosis over clinical variables (ΔAUC 0.03, 95% CI −0.01 to 0.06, and 0.04, 95% CI 0.002–0.10) (Table 2). Furthermore, in CA patients with a maximum 0–2 h hs-TnT <14 (n = 219), addition of IGFBP-3 improved the ROC AUC to predict 70% stenosis from 0.75 to 0.77 (ΔAUC 0.03, 95% CI −0.007 to 0.06, Figure 6A). In pathway analysis of these same CA patients with 0–2 h hs-TnT <14 ng/L, ΔIGFBP-3 allowed the shifting of 19 intermediate risk cases, primarily to low risk (n = 17 low risk, n = 2 high risk, Figure 6B). Thus, an absolute 9% decrease in intermediate risk cases was achieved, all correctly reallocated to low or high risk.

Finally, in matched ED cases with a maximum hs-TnT < 14 and a finding of 70% stenosis on CA (n = 150), decreased presentation values of IGFBP-3 alone were a predictor of new ACS events within 365 days (n = 16, OR = 0.45; 0.24–0.85), P = 0.01), whereas the maximum value of hs-cTnT was not (OR = 0.63; 0.05–8.3, P = 0.73).

Discussion

The main findings of this work are: (i) IGFBP-3 is a protein that responds to the stimulus of coronary ischaemia without necrosis in an isolated cardiac setting, (ii) circulating IGFBP-3 exhibits dynamic shifts during provocative human cardiac EST such that a ΔIGFBP-3 occurs between 0 and 120/150 min in positive EST cases, (iii) levels of IGFBP-3 do not appear to be responsive to a clinical, acute toxic source of cardiac necrosis (alcohol ablation) nor to naturally occurring MI, (iv) IGFBP-3 undergoes net extraction/degradation by heart and liver, (v) a 0–2 h ΔIGFBP-3 in ED chest pain patients has potential to detect cardiac ischaemia in clinical scored/hs-TnT grey zone (5–14 ng/L) cases experiencing UAP at hospital ED presentation, (vi) ΔIGFBP-3 can improve the identification of patients with sub-threshold troponin values who would most benefit from objective imaging, and (vii) in the ED, ΔIGFBP-3 appears to add to current clinical variables, including hs-cTnT measurement.

The isolated perfused heart has previously been used for the experimental proteomic identification of markers of cardiac ischaemia and infarction.^31,32 However, the vast majority of reports have not taken the further step of attempting to translate findings into human clinical studies. In performing this step, the sheer number of proteins identified herein (n = 148), prompted us to carefully consider which to pursue as potential useful markers. Three factors prompted us to investigate IGFBP-3. First, the early (∼10 min) appearance of IGFBP-3 in cardiac perfusates after coronary ligation suggested it may react to ischaemia rather than MI. Second, Koomen et al.³¹ also identified IGFBP-3 in rat heart perfusates, validating our discovery, and third, IGFBP-7 has found promise as a useful prognostic marker in heart failure, thus linking the IGFBP family to CVD.^25,33 The cardiac stress testing and SAA groups were chosen as trial testing groups to determine temporal and biochemical responses of IGFBP-3 to (i) ischaemia without infarction and (ii) a clinical model of infarction where the onset is precisely known. This is a logical step to take after unbiased proteomic analyses of isolated heart perfusate revealed IGFBP-3 as a potential ischaemia marker. Both cardiac stress testing and SAA groups are not intended to be large robust datasets—they are integrated into our overall discovery process to rapidly identify markers that have the potential to be properly investigated (Figure 1) in the more robust ED datasets which provide the more invaluable information.

Measurement of circulating IGFBP-3 is clinically routine on multiple automated analysers and is an accepted surrogate marker of growth and GH production, due to it being responsible for the carriage of up to 90% of circulating IGF-I/II.³⁴ However, not all immunoassays are equivalent as IGFBP-3 is subject to cleavage by multiple proteases and binds multiple proteins, all of which can affect antibody–antigen binding.³⁵ In our IGFBP-3 assay, we instituted a binding enhancement step to reduce protease and binding protein interference to attempt to minimize these effects. Importantly, although there are well-validated normal range values for IGFBP-3 use in assessing disorders of growth,^34,35 that is not the consideration here; instead, a directional change of IGFBP-3 between 0 and 2 h (i.e. ΔIGFBP-3), regardless of starting value, is what provided potentially useful information.

Preliminary assessment of IGFBP-3 in a small group of cardiac EST cases revealed a marked ΔIGFBP-3 between 0 and 150 min in positive test cases, that was absent in negative normal controls. Further assessment in EST cases in more heterogeneous ED patients also found significant (albeit muted) 0–2 h positive ΔIGFBP-3 in inducible ischaemic cases. It is important to note here that the direction of ΔIGFBP-3 in EST cases is negative, whereas in ED cases, it was positive. This is mostly likely due to the timing of sampling, such that ED cases have their first sample taken some hours post-symptom onset—when IGFBP-3 has reached/is close to the nadir seen at 120–150 min in the experimental EST (Figure 2)—with the second at 2 h reflecting the rise which was seen with later samples in EST. Proper assessment of the ability of ΔIGFBP-3 to improve the diagnostic yield from cardiac EST will require more rigorous assessment in an appropriately designed and recruited study population with sampling at the time of testing.

In contrast with cardiac EST, there was no discernible ΔIGFBP-3 in patients undergoing the non-ischaemic toxic challenge of SAA, out to 24 h post-procedure. A similar result was observed in ED patients suffering acute MI (ROC AUC = 0.52) in the ED cohorts that followed. This suggests that plasma IGFBP-3 may be less responsive to immune system/inflammatory/free radical processes terminal necrotic events such as SAA/MI, possibly due to inhibition of protease activity relevant to IGFBP-3 metabolism, but this requires further studies for confirmation.

Regional arterio-venous assessment of IGFBP-3 indicated net cardiac and hepatic extraction/degradation of IGFBP-3. To our knowledge, our data is the first to show both potential hepatic and cardiac extraction of IGFBP-3 in humans and in this regard, IGFBP-3 has a similar human profile to that of IGFBP-7.²⁵ Reconciliation of the finding of cardiac extraction of IGFBP-3 in humans with our discovery of IGFBP-3 in IIRH perfusates could come from consideration of the IIRH as an isolated organ and the fact that tissue expression of IGFBP-3 is higher in other organs compared with the heart.^36,37 The validity of our arterio-venous results is supported by complementary regional hs-TnT data which showed the expected cardiac release and renal clearance, as we have published previously.¹⁷

In ED patients from three well-published and similar studies, IGFBP-3 did not detect acute MI, but it did improve on the yield of patients satisfying the NOT rule. Further, among patients eligible for NOT rule inclusion, ΔIGFBP-3 provided potential discrimination with respect to three important considerations: the presence of UAP, 70% arterial stenosis on imaging, and whether benefit might accrue from further stress testing. Despite concomitant advances in troponin testing/clinical pathways to improve the identification of high- and low-risk ED cases, a sizeable fraction (25–30%) still have clinically uncertain presentations and identification of those who require further objective testing (either physical or imaging) would benefit from additional biomarker support. Graded troponin levels below mandated cut-offs have been proposed to assist with the identification of significant CAD and subsequent events,^8,9 but do not assist in the detection of UAP. The PPV increments provided for UAP by IGFBP-3 are modest, but for the prediction of 70% stenosis in sub-threshold hs-TnT cases, IGFBP-3 + ΔIGFBP-3 provided a reasonable shift of 9% of intermediate risk cases—primarily to low risk—with no false negatives; if validated by further studies, this latter improvement could be of important clinical benefit to optimize angiography requests. Non-cardiac inflammatory influences on the observed responses of IGFBP-3 cannot be ruled out, but it is noteworthy that in the identification of sole infective disease in our ED cohorts, IGFBP-3 levels did not change between 0 and 2 h sampling times. Also, of interest, the potential for the performance of IGFBP-3 to be enhanced by other factors involved in the IGF-1/IGFBP-3 binding complex remains to be determined, including the determination of 70% stenosis outcome prediction at 365 days, where IGFBP-3 was superior to hs-TnT.

The precise biochemical mechanisms underpinning the performance of IGFBP-3 + ΔIGFBP-3 in chest pain/ACS patients are likely to involve multiple known proteases and plasma binding factors,³⁵ all with cardiovascular localization and relevance. Firstly, IGFBP-3 RNA and protein, as well as IGF-1, are present in smooth muscle cells (SMCs) from atherosclerotic plaques taken from patients who experienced crescendo angina, but not in SMC from normal coronary artery specimens;³⁸ secondly, in a porcine model of embolization induced ischaemia, cardiac myocytes produced elevated levels of IGFBP-3 mRNA at 3–24 h post-onset;³⁹ further, IGFBP-3 RNA and protein are both expressed in human ventricles and are more abundant in failing hearts from patients with IHD compared with HCM.⁴⁰ Thirdly, IGFBP-3 promotes angiogenesis in human endothelial cells via IGF-I receptor signalling and sphingosine kinase and sphingosine-1-phospate formation,⁴¹ and the act of IGFBP-3 binding to endothelial cells inhibits its own proteolysis.⁴² A limited number of studies have addressed plasma measurement of IGFBP-3 as a prognostic marker of future cardiac ischaemic events,^43,44 but in both cases, IGFBP-3 was predictive of future events independent of known clinical and biomarker risk variables. Thus, clear directions are present for studies addressing the mechanistic actions underlying our observations and the potential to enhance/optimize performance further.

Study limitations

First, this report presents observational studies only and we cannot make inferences about the impact on patient outcomes. Second, despite all three ED studies employing guideline definitions for the diagnosis of UAP, we cannot exclude regional differences being present and influencing analyses. Further, we used the presence of 70% stenosis to define significant CAD. We did not perform more detailed analyses on plaque burden and vulnerability. Third, the adjudication of MI across all three studies utilized two different troponin assays. While our analysis used the FAST-TRAC study with Beckman hs-TnI assay for the definition of MI, both SPACE and APACE have been adjudicated with reference to Roche hs-cTnT. To limit the potential effects of this, we performed hs-cTnT measurements in FAST-TRAC to generate biochemical equivalence across all three studies with respect to analyses. Fourth, we cannot rule out statistical over-fitting in some analyses due to smaller case numbers and in that regard, some of the results are hypothesis-generating only. Further studies in appropriate patient groups with adequate sample size and event rates are required to validate the findings of this study.

Conclusions

We provide the first translational report—from basic discovery to clinical assessment—that circulating IGFBP-3 is responsive to acute cardiac ischaemia (without necrosis) and that ΔIGFBP-3 has the potential to aid decision-making with regard to the need for imaging to detect significant coronary stenosis or the need to submit individuals for cardiac stress testing. This applies both to patients eligible for clinical scoring (e.g. NOT rule) and/or those with detectable troponin that does not reach actionable levels with respect to the diagnosis of ACS.

Lead author biography

Chris J. Pemberton is a Research Professor based at the Christchurch Heart Institute, University of Otago, Christchurch, New Zealand, and a co-founder/VP Discovery at Upstream Medical Technologies. His focus is on the identification of new markers and therapies that meet unmet need in cardiac ischaemia and related diseases.

Data availability

The non-identifying data underlying this article will be shared on reasonable request to the corresponding author.

Supplementary material

Supplementary material is available at European Heart Journal Open online.

Acknowledgements

We thank the technical and clinical staff of the CHI (New Zealand), CRIB (Basel, CH), and FAST-TRAC (USA) for assistance with biomarker assays and patient blood sampling.

Funding

The SPACE, SEARCH, and SPICI studies were supported by the Health Research Council of New Zealand, the Canterbury Medical Research Foundation, and the National Heart Foundation of New Zealand. The APACE study was supported by research grants from the Swiss National Science Foundation, the Swiss Heart Foundation, the University Hospital Basel, and the University of Basel. Upstream Medical Technologies provided support for the procurement of study samples, reagents, hs-TnT, and IGFBP-3 tests.

References

Author notes