Circulating suppression of tumorigenicity 2 (ST2) predicts cardiovascular outcomes and mortality in ischemic heart disease (IHD). ST2 does not correlate with traditional risk indicators as closely as N-terminal pro–brain natriuretic peptide (NT-proBNP) and is only weakly correlated with other biomarkers, indicating distinct pathways for stimulus and release. Although of little diagnostic utility in IHD, ST2 does offer prognostic information. In ST elevation myocardial infarction, ST2 levels increase to peak above the normal reference range (within 6 to 18 hours of symptom onset) in about half of patients. Levels in the upper quartile observed in IHD independently predict cardiovascular death and heart failure with an approximate doubling of risk. Similar but weaker associations have been reported in non–ST elevation myocardial infarction, in which ST2 predicts short-term (30-day) and long-term (>1-year) death and heart failure independent of clinical indicators, but these relations are lost if Global Registry of Acute Coronary Events (GRACE) score and NT-proBNP are added to multivariate models. Early postinfarction levels of ST2 (i.e., <24 hours after admission) have the greatest prognostic utility. Early postinfarction ST2 levels and change over 24 weeks are related to infarct extent and remodeling to a similar extent as NT-proBNP and aldosterone, and ST2 may have a significant pathophysiological role in these postinfarction processes. In long-term follow-up of stable IHD, ST2 is predictive of all-cause and cardiovascular mortality independent of accepted clinical indicators and other biomarkers, including NT-proBNP, high-sensitivity C-reactive protein, interleukin-6, high-sensitivitiy cardiac troponin T, and galectin-3. In conclusion, ST2 in combination with NT-proBNP consistently improves risk stratification compared with either marker alone.
The suppression of tumorigenicity 2 (ST2) protein was identified >20 years ago but remained an orphan receptor until confirmation, in 2005, of interleukin-33 (IL-33) as its endogenous ligand. In vitro expression of ST2 is upregulated with cardiomyocyte stretch. In vivo cardiac loading is associated with an increase in circulating levels of ST2. ST2 is present in circulating and membrane-bound forms. In experimental myocardial infarction (MI), IL-33 prevents apoptosis and improves cardiac function and survival. It has been suggested that circulating “soluble ST2” may act as a “decoy” for IL-33 and in this way reduce access of the interleukin to cardiac membrane–bound receptors, thus reducing the potential cardioprotective effect of IL-33 in acute and chronic heart disease.
This review summarizes findings with respect to plasma ST2 concentrations and their relations to clinical outcomes across the spectrum of ischemic heart disease (IHD) presentations ranging from populations with undifferentiated chest pain through acute coronary syndromes (ACS) to chronic stable IHD ( Table 1 ).
| Article | Study population, disease group | Setting | Primary endpoint | Follow-up | Number of patients who reached endpoint | sST2 assay ∗ | Utilization of sST2 as predictor variable | HR (95% CI) in univariate Cox-regression | HR (95% CI) in multivariate Cox-regression | Other independent variables in multivariate Cox-regression | Cut-point reported for sST2 ∗ | Additional information |
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Brown AM et al 2007 | n=348 chest pain | emergency department patients | composite of all-cause mortality, acute MI, or re-vascularization | exactly 30 days for each study participant | n=23 (7%) | MBL assay (commercially available ready-to-use assay) | continuous | not applied | not applied | not applied | no cut-point provided | area under the curve for prediction of outcome, 0.58 (95% CI, 0.48-0.68) |
| Aldous SJ et al 2012 | n=995 chest pain | emergency department patients, post-hoc analysis of a prospectively conducted multicenter study | composite of all-cause mortality and heart failure | exactly 1.5 years for each study participant | n=110 (11%) | Presage assay (commercially available ready-to-use assay) | dichotomous, >34 ng/mL | not provided | 1.9 (1.2-3.2) | a variety of clinical variables and several biomarkers including BNP and cTnI | 34 ng/mL, cut-point by 95 th percentile | sST2, cTnI and BNP provided independent prognostic information |
| Shimpo M et al 2004 | n=810 STEMI | hospitalized patients, TIMI 14 and ENTIRE-TIMI 23 trial participants | all-cause mortality | exactly 30 days for each study participant | n=28 (3%) | MBL assay (commercially available ready-to-use assay) | incremental, per 1 ng/mL rise | not provided | 1.8 (1.0-3.1) | clinical variables only (age, heart rate, systolic blood pressure, location of MI, Killip class, and time from onset of chest pain) | 0.235 ng/mL, cut-point by median | sST2 lost its independent prognostic value when BNP and cTnI were entered into multivariate analysis |
| Sabatine MS et al 2008 | n=1239 STEMI | hospitalized patients, CLARITY-TIMI 28 substudy | composite of cardiovascular death and heart failure | exactly 30 days for each study participant | n=87 (7%) | MBL assay (commercially available ready-to-use assay) | incremental, log 10 -transformed concentrations per 1 unit | 2.4 (1.7-3.5) (cave: this is an odds ratio because logistic regression was applied) | 1.9 (1.3-3.0) (cave: this is an odds ratio because logistic regression was applied) | clinical variables only (including infarct location, Killip class, time to initiation of fibrinolytic therapy, and type of lytic) | 0.080 ng/mL, cut-point by median | sST2 and NT-proBNP provided independent and additive prognostic information |
| Dhillon OS et al 2013 | n=667 STEMI | hospitalized patients | all-cause mortality | exactly 1.0 year for each study participant | n=58 (9%) | in-house assay (constructed by the respective researchers) | incremental, log 10 -transformed concentrations per 1 unit | 9.1 (5.2-16.0) | 2.6 (1.2-5.4) | a variety of clinical variables including GRACE score and several biomarkers including NT-proBNP (but not cTnI or cTnT) | 0.862 ng/mL, cut-point set arbitrarily for 30-day mortality | sST2 and IL-33 were simultaneously evaluated as outcome predictors |
| Eggers KM et al 2010 | n=403 NSTEMI | hospitalized patients, GUSTO IV substudy on inflammatory markers | all-cause mortality | exactly 1.0 year for each study participant | n=22 (6%) | Presage assay (commercially available ready-to-use assay) | incremental, log e -transformed concentrations per 1 standard deviation | not provided | 2.3 (1.1-4.6) (cave: this is an odds ratio because logistic regression was applied) | age, heart failure, diabetes mellitus, previous MI or stroke, but no biomarkers (such as NT-proBNP, cTnT or CPR) | no cut-point provided | serial measurement of sST2: sST2 concentrations decreased in time course from baseline to 72 hours thereafter |
| Dhillon OS et al 2011 | n=577 NSTEMI | hospitalized patients | composite of all-cause mortality, heart failure hospitalization, and re-infarction | mean of approx. 1.5 years | n=156 (27%) | in-house assay (constructed by the respective researchers) | incremental, log 10 -transformed concentrations per 1 unit | 3.6 (2.3-5.6) | 2.0 (1.2-3.3) | a variety of clinical variables and several biomarkers including cTnI | no cut-point provided | sST2 lost its independent prognostic value when GRACE score and NT-proBNP were entered into multivariate analysis |
| Kohli P et al 2012 | n=4426 NSTEMI | hospitalized patients, substudy of the MERLIN-TIMI 36 trial | composite of cardiovascular mortality and new or worsening heart failure | mean of approx. 1.0 year | not provided | Presage assay (commercially available ready-to-use assay) | incremental, log 10 -transformed concentrations per 1 unit | 1.6 (1.4-1.7) | 1.3 (1.1-1.4) | a variety of clinical variables and several biomarkers including BNP and cTnI | 35 ng/mL, cut-point by highest quartile | sST2, cTnI and BNP provided independent and additive prognostic information |
| Demyanets S et al 2014 | n=373 STEMI, NSTEMI, and stable angina | hospitalized patients | all-cause mortality | mean of approx. 3.5 years | n=37 (10%) | R&D assay (commercially available ready-to-use assay) | dichotomous, >0.538 ng/mL | 2.1 (1.1-4.2) | 2.2 (1.1-4.4) | age, gender, hyper-lipidemia, hyper-tension, smoking, creatinine but no biomarkers (such as NT-proBNP, cTnT or CPR) | 0.538 ng/mL, cut-point by highest quintile | sST2 and IL-33 were simultaneously evaluated as outcome predictors |
| Dieplinger B et al 2014 | n=1345 stable coronary artery disease | hospitalized patients, substudy of the LURIC study | all-cause mortality | median of approx. 9.8 years | n=477 (36%) | Presage assay (commercially available ready-to-use assay) | incremental, log 10 -transformed concentrations per 1 standard deviation | 1.5 (1.3-1.6) | 1.2 (1.0-1.3) | a variety of clinical variables and several biomarkers including cTnT, NT-proBNP, CRP, IL-6, and galectin-3 | 25 ng/mL, cut-point by highest quartile | sST2, cTnT and NT-proBNP provided independent and additive prognostic information |
| Weir RA et al 2010 | n=100 acute MI | hospitalized patients | left ventricular functional recovery after acute MI | exactly 24 weeks for each study participant | not applicable | R&D assay (commercially available ready-to-use assay) | sST2 was correlated with baseline and 24-week measurements of LV function | not applied | not applied | not applied | no cut-point provided | baseline sST2 was associated with baseline and 24-week LVEF and with baseline and 24-week infarct volume index |
∗ Results obtained with different sST2 assays are not comparable; the assays are not standardized. One should be aware with which method the results were produced. Commercially available assays: Presage assay (Presage ST2 kit, Critical Diagnostics, San Diego, CA, USA), the MBL ST2 assay (Human ST2 ELISA kit, Medical&Biological Laboratories, Woburn, MA, USA), and the R&D ST2 assay (ST2/IL-1 R4 DuoSet ELISA or Quantikine ELISA, R&D Systems, Minneapolis, MN, USA).
ST2 in Undifferentiated Chest Pain
Two reports have addressed the diagnostic and prognostic performance of plasma ST2 concentrations in patients presenting to emergency departments with chest pain.
Brown et al enrolled 348 patients, aged ≥25 years, presenting to emergency departments with chest pain triggering the acquisition of an electrocardiogram. Blood for ST2 (measured using a research use–only assay) was obtained at presentation, and levels were subsequently assessed for relations to 30-day event rates for acute MI (AMI), ACS, and adverse cardiac events (defined as all-cause mortality, AMI, or revascularization [angioplasty, stent placement, or coronary artery bypass surgery]). Event rates were low at 4.9% (n = 17) for AMI, 11.2% (n = 39) for ACS, and 6.6% (n = 23) for the composite end point. Areas under the receiver-operating characteristic curves (AUCs) for the detection of AMI, ACS, and the composite end point were low at 0.636, 0.630, and 0.579, respectively. The investigators concluded that ST2 was not useful in the evaluation of emergency department patients with suspected but not proved ACS.
Aldous et al revisited the diagnostic and prognostic utility of ST2 in patients with chest pain in a larger group with longer follow-up and found associations between ST2 and later all-cause mortality and incident heart failure (HF). Nine hundred ninety-five patients reporting ischemic-type pain (including acute chest, gastric, neck, jaw, or arm pain or discomfort or pressure without apparent noncardiac source) were recruited. Blood samples were acquired at 0 and 2 hours after presentation for the measurement of ST2, brain natriuretic peptide (BNP), troponin I (TnI), creatine kinase-MB fraction, and myoglobin. ST2 was weakly correlated with all 4 of the other markers (R = 0.06 to 0.30, p = 0.044 to 0.001, with the strongest correlation with BNP).
The diagnostic performance for the index admission and prognostic performance for events after discharge and up 18 months were assessed for each marker. End points included the composite of all-cause mortality and HF and the individual end points: all-cause mortality, AMI, revascularization, arrhythmia, and stroke. The cut points chosen for each marker were the upper-limit normal values as indicated by the test kit manufacturers. For ST2 this was the 95th percentile of the Critical Diagnostics Presage ST2 assay, 34.3 ng/ml (limit of detection 1.61 ng/ml, coefficient of variation 10% at 16.9 ng/ml, and normal range 1.75 to 34.3 ng/ml). Patients with ST2 >34.3 ng/ml (221 of 995) were more likely to be older; to be female; to have histories of HF, lung disease, diabetes, hypertension, and dyslipidemia; to have lower left ventricular (LV) ejection fractions (LVEFs); to have more frequent regional wall abnormalities; and to have higher plasma creatinine. Events during the index admission included 35 primary end points, 1 death, 34 episodes of HF, and 236 AMIs. ST2 levels were significantly higher in patients with compared with those without these end points.
Relative to the diagnosis of AMI, ST2 clearly underperformed troponin (sensitivity 31.2%, specificity 80.5% compared with TnI at 94.5% and 95.7%, respectively), while for the diagnosis of acute HF, ST2 lacked the sensitivity of BNP (73.5 vs 88.2%) but had higher specificity (79.6 vs 66.2%).
In contrast to performance for diagnosis, the prognostic performance of ST2 for postdischarge events up to 18-month follow-up was notable. Events from discharge till 18-month included 110 primary (i.e., all-cause death and HF) events (11.1%), 73 deaths (7.3%), 52 HF events (5.2%), 86 AMIs (8.6%), 57 revascularizations (5.7%), 49 arrhythmias (4.9%), and 19 strokes (1.9%). Survival (Kaplan-Meier) curve analyses indicated a steep increase in risk in those with admission and/or 2-hour ST2 levels above the third quartile (33.4 ng/ml), a level close to the 95th percentile (34.3 ng/ml) used to define the upper limit of normal. Eighteen-month event rates for the primary end point were 4.9%, 7.6%, 7.2%, and 24.5% in ST2 quartiles 1, 2, 3, and 4, respectively. Similar results were found for each of death, AMI, HF, arrhythmia, and stroke. Prognostic associations between ST2 and important adverse events remained evident after multivariate analyses that adjusted for age; gender; ethnicity; history of IHD, HF, lung disease, revascularization, stroke, hypertension, and dyslipidemia; smoking status; diabetes; body mass index; the LVEF; regional wall abnormality score; and plasma creatinine. ST2, age, BNP, TnI, myoglobin, the LVEF, and previous IHD were all independently prognostic for the primary outcome.
Elevated ST2 predicted the primary outcome (hazard ratio [HR] 1.9; 95% confidence interval [CI] 1.2 to 3.2), which was inferior to BNP (HR 2.8; 95% CI 1.4 to 5.7), equivalent to myoglobin levels, and superior to TnI and creatine kinase-MB, although for all comparisons, overlap of 95% CIs existed. In those with initial diagnoses of AMI, ST2 predicted later primary outcome (HR 2.7; 95% CI 1.2 to 5.7) ST2 also predicted later HF (HR 2.2; 95% CI 1.0 to 4.7) in the population overall. Notably, ST2 was not an independent predictor of later AMI or revascularization.
ST2 combined with BNP improved risk stratification beyond that achieved with either marker alone. Rates for the primary end point over follow-up for those with both markers elevated, BNP alone elevated, ST2 alone elevated, or neither marker raised were 36.8%, 19.4%, 6.8%, and 2.2%, respectively.
This study was the first to assess the relation of early ST2 levels to later mortality and HF events over prolonged follow-up in a large undifferentiated chest pain population and to compare this prognostic performance with that of other markers. The association of ST2 with death and HF was retained after comprehensive adjustment for other predictors. ST2 and BNP combine to provide better risk stratification than obtained with either alone.
ST2 in ST Elevation Myocardial Infarction
Two reports on data derived from 3 randomized clinical trials of experimental therapy in ST elevation MI (STEMI) provide data on the predictive power of plasma ST2 for adverse events up to 30 days after MI, and a further article reported on prognostic performance out to a median follow-up duration of 20 months. ST2 was measured in >2,000 patients participating in Thrombolysis In Myocardial Infarction (TIMI) 14, Enoxaparin and Tenecteplase With or Without Glycoprotein IIb/IIIa Inhibitor as Reperfusion Strategy in ST Elevation MI (ENTIRE)–TIMI 23 and Clopidogrel as Adjunctive Reperfusion Therapy (CLARITY)–TIMI 28 trials.
Shimpo et al measured ST2 in 810 patients with STEMI (362 from TIMI 14 and 448 from ENTIRE–TIMI 23). Patients recruited to these trials had episodes of ischemic discomfort for ≥30 minutes within 6 hours (ENTIRE–TIMI 23) or 12 hours (TIMI 14) and exhibited ≥0.1-mV ST-segment elevation in 2 contiguous electrocardiographic leads. Blood for ST2 measurements was taken at presentation. Using an older generation research use–only ST2 assay (Medical and Biological Laboratories Co., Ltd.) the authors found levels higher (0.379 vs 0.233 ng/ml, p <0.001) in those who died or developed HF over 30 days after STEMI. Ascending quartiles of plasma ST2 significantly corresponded to increasing time from symptom onset, higher heart rates, higher cardiac TnI, higher BNP, higher C-reactive protein (CRP), higher creatinine, and increasing likelihood of an anterior location of the MI. Notably most of the elevation in these covariates occurred between the third and fourth quartiles of plasma ST2. In this STEMI population, ST2 was correlated weakly with TnI (r = 0.26, p <0.0001) but not with BNP. Over 30 days, events included 28 deaths, 29 recurrent MIs, 21 HF events, and 47 death and HF composite events. ST2 levels were higher in subsequent decedents than in survivors and also in those incurring HF but not in those with recurrent MI. This pattern reinforces the repeated finding that ST2 predicts death and HF but not recurrent ischemic events. Despite these relatively small numbers of events, risk for inpatient death increased with quartile of ST2 (0.98%, 1.5%, 3.0%, and 6.4% in quartiles 1, 2, 3, and 4, respectively; p = 0.0008), whereas corresponding percentages for in-hospital recurrent MI (4.4%, 3.5%, 0.5%, and 1.5%, respectively) showed no pattern of association with increasing ST2. Thirty-day event rates also revealed the key ST2-associated end points to be death and HF. Notably, those with inframedian levels of ST2 on admission incurred no further death or HF in the interval between discharge and 30-day follow-up. Again, there was no association with increased risk for recurrent MI at 30 days, with an almost inverse trend between quartile of ST2 and risk for new MI (5.4%, 4.5%, 2.5%, and 2.0% in ST2 quartiles 1, 2, 3, and 4, respectively).
In multivariate analysis ST2 remained predictive of mortality at 30 days, after controlling for age, heart rate, blood pressure, location of infarction, Killip class, and time from symptom onset (HR per 1 ng/ml increase 1.77; 95% CI 1.01 to 3.12, p = 0.047), but the independent association was lost once BNP and TnI were added to the model (this was possible only for n = 448 in the ENTIRE–TIMI 23 subpopulation). In the TIMI 14 subcohort, additional samples were taken at later time points, and ST2 at 12 hours from presentation remained independently predictive of mortality after adjustment for age, heart rate, blood pressure, location of MI, Killip class, and time from symptom onset (p <0.01). The investigators concluded that ST2 is associated with mortality after STEMI independent of established clinical indicators. They did not provide any assessment of the risk stratification potentially offered by marker combinations such as ST2 coupled with either TnI or BNP.
These findings were reinforced and extended by data from the CLARITY–TIMI 28 trial in STEMI, which also provided a compelling analysis of the complementarity of ST2 and the established marker N-terminal pro-BNP (NT-proBNP). ST2 was measured on recruitment to the trial in 1,239 patients. Ascending quartiles of plasma ST2 were associated with diabetes, creatinine clearance (although the difference in estimated glomerular filtration rate (eGFR) between ST2 quartiles 1 to 4 was <6 ml/min) but not related to other clinical characteristics often considered to associate with chronic LV wall stress, including age, hypertension, previous MI, or previous HF. The correlation with NT-proBNP was weak (r = 0.14, p <0.001). NT-proBNP was correlated with age, gender, hypertension, previous MI, previous HF, and lower creatinine clearance (quartile 1 to quartile 4 ST2 interquartile difference in eGFR about 16 ml/min). This striking nonparallelism of the relations of the 2 markers with common covariates supports the concept of different primary drivers for their production and secretion.
The temporal patterns of change in plasma ST2 and NT-proBNP in the post-MI period also differ. Repeat sampling in most patients was undertaken at angiography a mean of about 4 days after MI. ST2 decreased modestly an average of 0.009 ng/ml, while mean NT-proBNP increased nearly sixfold by an average of 386 pg/ml (p <0.001 for the 2 changes from baseline). Notably, plasma ST2 differed between patients with and without angiographic TIMI flow grade 0 and between those with and without TIMI myocardial perfusion grade 0 (both adverse appearances were associated with about threefold higher ST2 levels) than was the case for NT-proBNP (about 40% differences), whereas the natriuretic peptide marker was clearly more closely associated with LVEF than ST2 (r = −0.45 vs r = −0.17). These findings suggest plasma ST2 reflects the quantum of injured tissue and associated necrosis and inflammatory events, whereas the primary driver for NT-proBNP remains cardiac mechanical stress.
For ST2 and NT-proBNP, the mean plasma concentrations observed in those later incurring cardiovascular (CV) death or HF were four- to fivefold those in patients spared these outcomes. In patients later experiencing strokes, marker levels were two- to four-fold those in patients spared stroke. In contrast, ST2 levels were almost identical (0.074 vs 0.079 ng/ml) in those with and without later recurrent MI. Although NT-proBNP levels did differ for this end point (108 vs 70 pg/ml, p = 0.0013), it was to a far lesser degree than for death or HF. Thirty-day CV death rates by quartile of ST2 were 2.5%, 3.2%, 3.2%, and 9% in quartiles 1, 2, 3, and 4, respectively, with corresponding percentages of 1.4%, 1.0%, 5.1%, and 10.5% for successive quartiles of NT-proBNP. Similar patterns were observed for incident HF. The most pronounced increase in risk occurred from the third to the fourth quartile of ST2 values, whereas the increasing risk associated with increasing NT-proBNP was more smoothly graded across successive quartiles. Notably, ST2 levels at baseline were superior to levels 4 days later (i.e., at angiography) for the prediction of death or HF, whereas the opposite was true of NT-proBNP. If corroborated, this observation will inform optimal peri-infarct timing of sampling, which, for prognostic applications, may differ for these 2 markers.
In multivariate analysis with adjustment for age, gender, hypertension, diabetes, previous MI, previous HF, creatinine clearance, infarct location, Killip class, time from symptom onset to lytic therapy, and peak creatine kinase, a 1-SD elevation in log-transformed ST2 was associated with a 1.94-fold (95% CI 1.25-fold to 3.03-fold) increase in risk for CV death or HF over 30 days (p = 0.003). The analogous result for NT-proBNP was 1.46-fold (95% CI 1.22-fold to 1.76-fold, p <0.001). In a multimarker model including ST2 and NT-proBNP plus all the traditional predictors, the 2 markers remained independent predictors of CV death or HF, with odds ratios per 1-SD elevation in log levels of 1.88 (95% CI 1.17 to 3.03, p = 0.009) and 1.41 (95% CI 1.17 to 1.69, p <0.001) for ST2 and NT-proBNP, respectively. Figure 1 illustrates combined risk stratification using TIMI risk score together with ST2 and NT-proBNP.
