Clinical Science

Research article

Matrix metalloproteinase-2 predicts mortality in patients with acute coronary syndrome

Onkar S. Dhillon, Sohail Q. Khan, Hafid K. Narayan, Kelvin H. Ng, Noor Mohammed, Paulene A. Quinn, Iain B. Squire, Joan E. Davies, Leong L. Ng


The aim of the present study was to investigate the predictive value of MMP (matrix metalloproteinase)-2, MMP-3 and MMP-9 levels in patients with acute coronary syndrome for death, readmission with HF (heart failure) or recurrent MI (myocardial infarction) and to compare them with established markers, NT-proBNP (N-terminal pro-B-type natriuretic peptide) and the GRACE (Global Registry of Acute Coronary Events) score. A single blood test was taken 4 days after admission in 1024 consecutive patients with acute MI with end points observed over 519 (134–1059) days [value is median (range)]. MMP-2 and MMP-3 were increased in patients who died (n=111) compared with survivors (P<0.006 and P=0.01 respectively), but were similar in patients with HF (n=106) or MI (n=138). MMP-9 levels were similar across study end points. Using Cox proportional hazards modelling, MMP-2 demonstrated an independent prediction of death [HR (hazard ratio) 6.60, P=0.001], along with NT-proBNP (HR 4.62, P<0.001) and the GRACE score (HR 1.03, P<0.001), but MMP-3, MMP-9 or log10-troponin I did not. For 1 year mortality, the areas under the receiver operating characteristic curves were 0.60 and 0.58 for MMP-2 and MMP-3 respectively, compared with 0.82 for NT-proBNP and 0.84 for the GRACE score (all P<0.001). Kaplan–Meier analysis revealed that MMP-2 levels in the top quartile were associated with higher mortality rates (log rank 12.49, P=0.006). On univariate analysis, MMP-2 and MMP-3 had a weak association with HF readmission, which was lost after adjustment for clinical factors. None of the MMPs tested predicted MI. In conclusion, this is the first single centre study that identifies MMP2 as an independent predictor of all-cause mortality post-ACS (acute coronary syndrome); however, NT-proBNP and the GRACE score are superior for risk stratification in this cohort.

  • acute coronary syndrome
  • matrix metalloproteinase (MMP)
  • myocardial infarction
  • N-terminal B-type natriuretic peptide (NT-proBNP)
  • prognosis


MMPs (matrix metalloproteinases) are a family of zinc-containing endoproteinases which, along with their inhibitors [TIMPs (tissue inhibitors of metalloproteinases)] are vital in ensuring a balanced turnover of the cardiac ECM (extracellular matrix). MMPs have been classified into four groups based on their substrate specificity. MMP-2 and MMP-9 are known as the gelatinases and are also able to degrade type IV collagen found in basement membranes. MMP-3 is one of a group called the stromelysins and is active against a broad spectrum of ECM components [1]. The discovery of elevated levels of MMP-2, MMP-3 and MMP-9 both in patients with ACS (acute coronary syndrome) [2] and atherosclerosis [3] have stimulated interest in clarifying their precise cardiovascular role.

During the first few days post-AMI [acute MI (myocardial infarction)], the left ventricle dilates with thinning of the infarct zone, which leads to infarct expansion. MMP levels increase early post-AMI and have, therefore, been implicated in this process, which is known as remodelling [4]. These changes in geometry contribute to the development of congestive HF (heart failure) and predict prognosis [5]. Rohde and co-workers [6] were the first to show that administration of an MMP inhibitor attenuated early LV (left ventricular) dilation 4 days after MI in mice [6]. Subsequent studies on genetically modified mice with targeted deletion of the MMP genes have shown improved survival rates over wild-type mice mainly due to reduced cardiac rupture rates [79], with corroboratory results found recently in humans with cardiac rupture [10].

In humans post-AMI, there are multiple reports of MMP levels correlating with echocardiographic parameters of LV function, indicating they possibly contribute towards remodelling [1114]. This is supported by evidence that levels correlate with other neurohormonal markers, such as ANP and BNP (atrial and brain natriuretic peptide respectively) [15,16]. They have also been implicated in atherosclerotic plaque rupture, the underlying cause of ACS [17], with higher levels of MMP-9 found in plasma of the infarct-related artery [18]. MMP levels also contribute to the inter-individual susceptibility to cardiovascular disease and the outcome from ACS, as shown in studies of MMP gene polymorphisms [19]. This central role of MMP-2, MMP-3 and MMP-9 in patients with cardiovascular disease is highlighted further by studies demonstrating the prognostic power of MMP levels in predicting adverse events [2023]. However, their utility as prognostic markers has been hampered by inconsistent results.

BNP and NT-proBNP (N-terminal of its pro-hormone) have a strong association with the extent of LV dysfunction and mortality [24,25]. The aim of the present study was to investigate the prognostic role of MMP-2, MMP-3 and MMP-9 in both STEMI (ST segment elevation MI) and NSTEMI (non-STEMI) using NT-proBNP as a benchmark comparator.


Study population

We studied 1024 consecutive AMI patients admitted to the University Hospitals of Leicester NHS Trust. The study complied with the Declaration of Helsinki, and was approved by the local ethics committee; written informed consent was obtained from patients. MI was diagnosed if a patient had a plasma creatine kinase-MB elevation greater than twice normal or cardiac troponin I (Advia Centaur Troponin-I assay; Siemens) levels >0.1 ng/ml taken up to 30 h after symptom onset with at least one of the following, chest pain lasting >20 min or diagnostic serial ECG changes consisting of new pathological Q waves or ST segment and T wave changes [26]. AMI was subcategorized into STEMI or NSTEMI. Exclusion criteria were known malignancy, renal replacement therapy, recent/pending blood transfusion or surgery within the previous month. Demographical, clinical, biochemical and echocardiographic data were obtained. All patients received standard medical treatment, and revascularization was at the discretion of the attending physician.

Plasma samples were drawn 4 days after the onset of symptoms and after 15 min of bed rest into tubes containing EDTA and aprotinin. Plasma was stored at −80°C until assayed in a blinded fashion in a single batch for determination of plasma MMP-2, MMP-3, MMP-9 and NT-proBNP. eGFR [estimated GFR (glomerular filtration rate)] was calculated from the simplified formula derived from the MDRD (Modification of Diet in Renal Disease) study [27]. Median length of follow-up was 519 days (range, 134–1059 days), with no patients lost to follow-up.


Transthoracic echocardiography was performed in 826 (80.7%) patients during the index admission but after day 3 using either a Sonos 5500 or IE 33 instrument (Philips Medical Systems). A 16-segment LVWMI (LV wall motion index) was performed based on the American Society of Echocardiography method. In suitable patients, LVEF (LV ejection fraction) was calculated using the biplane method of discs formula. Impaired LV systolic function was defined as either an LVEF <40% or an LVWMI >1.8.

GRACE (Global Registry of Acute Coronary Events) scoring

On the basis of an international observational database of ACS patients, GRACE scores can be calculated on initial presentation and on discharge to predict prognosis in these patients. The discharge GRACE score includes age, heart rate, systolic blood pressure, serum creatinine, congestive HF, ST depression, raised cardiac enzymes, a history of previous MI or CABG (coronary artery bypass grafting) and whether in-hospital revascularization was performed. Discharge scores designed to predict death and re-infarction rates up to 6 months [28] are, therefore, used in further analyses. Admission scores are mainly designed to predict in-hospital mortality and do not utilize knowledge of in-hospital revascularization strategies and, consequently, are not as strong a predictor of 6 month mortality [29]. Data were available for GRACE score calculation in 886 (86.5%) patients.

NT-proBNP assay

Our NT-proBNP assay is based on a non-competitive assay, as reported previously [30]. Sheep antibodies were raised to the N-terminal of human NT-proBNP and monoclonal mouse antibodies were raised to the C-terminal. Samples or NT-proBNP standards were incubated in C-terminal IgG-coated wells with the biotinylated N-terminal antibody for 24 h at 4°C. Detection was with MAE (methyl-acridinium ester)-labelled streptavidin on an MLX plate luminometer (Dynex Technologies). The lower limit of detection was 0.3 pmol/l. There was no cross-reactivity with ANP, BNP or CNP (C-type natriuretic peptide). Inter- and intra-assay coefficients of variation were 2.3 and 4.8% respectively. The results from this in-house assay are highly correlated (r=0.90, P<0.0001; n=86) with those obtained on the NT-proBNP assay marketed by Roche Diagnostics.

MMP assays

All MMP-specific antibodies were obtained from R&D Systems. Mouse monoclonal antibodies (200 ng/100 μl) specific for MMP-2, MMP-3 and MMP-9 were coated on to ELISA plates overnight at room temperature (20°C). Plates were then blocked using 10% (v/v) foetal bovine serum. Samples and standards were incubated in the ELISA plates overnight at room temperature. Following washes, the tracer antibody (5 ng/100 μl biotinylated goat antibody specific for MMP-2, MMP-3 or MMP-9) was added to the wells and incubated for 2 h. After washing, detection was with MAE-labelled streptavidin, as for the NT-proBNP assay. There was no cross-reactivity between the different MMP assays. The intra- and inter-assay coefficients of variation for MMP-2, MMP-3 and MMP-9 were as follows: 1.2 and 8.4%, 1.6 and 2.9% and 1.4 and 6.7% respectively. The lowest detectable concentration defined as 2 S.D. above the zero point were 0.26, 0.02 and 0.14 μg/l respectively.

End points

We assessed the value of MMP-2, MMP-3, MMP-9 and NT-proBNP for the prediction of MCEs (major cardiac events). The primary end point was the occurrence of all-cause mortality; HF hospitalization and recurrent MI were considered as secondary end points. Hospitalization for HF was defined as a hospital readmission for which HF was the primary reason requiring treatment with high dose diuretics, inotropes or intravenous nitrate. AMI was diagnosed on established criteria as described above [26]. End points were obtained by reviewing the local hospital databases, the Office of National Statistics Registry and phone calls to patients. We achieved 100% follow-up.

Statistical analysis

Statistical analyses were performed on SPSS Version 14. MMP-2, MMP-3, MMP-9 and NT-proBNP levels were log10-transformed. Proportions between two categorical groups were compared using χ2 tests. Non-parametric tests were employed against non-Gaussian data with a Mann–Whitney U test used to compare continuous variables across two conditions and Spearman (rs) correlations to assess the relationship between two continuous variables. Variation in MMP levels between the several independent end point groups was assessed using the Kruskal–Wallis test. Correction for multiple comparisons was done with Hochberg's GT2 test and all significance levels quoted are two tailed. Assuming an event rate of 10% and that the covariates predict up to 30% of the variance of the biomarker, a sample size of 934 patients would be powered (80% at P<0.05) to detect an HR (hazard ratio) of the biomarker of 2. We recruited more than 1000 cases to cover for loss to follow-up. In order to test the independent predictive power of the peptides for MCEs, survival analysis using Cox proportional hazard modelling was conducted. Owing to log10 transformation of peptide levels, HRs refer to a 10-fold rise in the levels of these markers. Cox models included important clinical variables and any other variables that were statistically significant (at P<0.20) in univariate analyses. These included age, gender, previous history of MI, hypertension, diabetes, type of AMI (STEMI or NSTEMI), Killip class >1 and eGFR, with the addition of log10-troponin I and echocardiographic data in a subset of patients. MMP-2, MMP-3 and MMP-9 were first added individually to this ‘base’ model and then entered simultaneously along with NT-proBNP. A further Cox model was used to assess the relative prognostic power of MMP-2 with the GRACE score (and NT-proBNP), which excluded age and eGFR as both are involved in calculating GRACE scores and may lead to colinearity (rs=0.87 and 0.57 respectively). The areas under the receiver operator characteristic curves were calculated. Kaplan–Meier survival analysis and Mantel–Cox log rank tests were performed. A P value <0.05 was interpreted to be statistically significant.


Patient characteristics

Patient details are recorded in Table 1. We recruited 1024 patients, 478 (46.7%) were STEMI for whom thrombolysis was the primary method of revascularization performed in 310 (64.9%) patients. During follow-up, 111 (10.8%) patients died, 106 (10.3%) were readmitted with HF and 138 (13.5%) suffered a MI.

View this table:
Table 1 Characteristics of the study cohort as a whole and subdivided according to diagnosis of STEMI or NSTEMI

Values are numbers (%) for dichotomous variables or medians±S.D. for continuous variables. *Number of patients with data available for that variable if less than 1024. Troponin I level was taken up to 30 h after symptom onset. ACE, angiotensin-converting enzyme; ARB, angiotensin II receptor blocker; NR, normal range.

Plasma MMP levels (Table 2)


Plasma levels ranged from 0.3 to 371.1 μg/l with a median of 22.8 μg/l. MMP-2 was elevated in patients who died compared with event-free survivors [26.8 (0.7–366.6) compared with 21.8 (0.3–371.1) μg/l respectively; P=0.006; values are medians (range)]. Patients with HF tended to have higher levels, but this was not significant [25.4 (0.6–124.5) μg/l; P=0.55] and similar levels were found in the recurrent MI group when compared with event-free survivors. MMP-2 was also higher in patients with STEMI compared with NSTEMI (P=0.001) and patients with Killip class >1 on admission (P=0.007). There was no significant gender difference in plasma levels or between those with or without a prior history of MI, hypertension, angina or diabetes mellitus. Plasma MMP-2 weakly correlated with eGFR, log10-glucose and NT-proBNP, but not with age, log10-troponin I or the GRACE score.

View this table:
Table 2 MMP-2, MMP-3, MMP-9 and NT-proBNP levels across different patient groups

Median values are all presented in the same order with the value for patients with the variable in question present in the first column and those with variable absent in the second. In (A), the n values indicate the number of patients with that variable present. In (B), the n values indicate the number of patients with data available for that variable.


The median level was 6.0 μg/l in AMI patients and ranged from 0.02 to 43.7 μg/l. MMP-3 was significantly elevated in patients who died compared with event-free survivors [6.9 (1.9–36.8] compared with 5.8 (0.1–43.7] μg/l respectively; P=0.010] with similar levels between event-free survivors and those who had HF [7.6 (0.02–35.5) μg/l; P=0.714] or MI. Men had higher levels (P<0.001) along with patients with a Killip class >1 (P=0.018) and those with prior MI (P=0.036), but was similar across other groups. MMP-3 levels correlated weakly with age, eGFR, log10-glucose, log10-troponin I, discharge GRACE scores and NT-proBNP.


Levels ranged from 0.1 to 893.5 μg/l with a median level of 74.2 μg/l. Levels were higher in patients with NSTEMI compared with STEMI and were lower in patients with Killip class >1 (both P<0.001). MMP-9 levels were similar between patients that suffered an MCE compared with event-free survivors and across most other groups. MMP-9 correlated weakly with the GRACE score, log10 glucose, log10 troponin I and NT-proBNP, but not with eGFR or age.

Relationship between MMPs and echocardiographic parameters

A total of 331 (40.1%) patients were classified as having impaired LV function (see above), and these patients had significantly increased levels of MMP-3 and NT-proBNP, but no significant difference in MMP-2 and MMP-9 levels (Table 2).

MMPs as predictors of MCEs

Primary end point: all-cause mortality

As shown above, plasma levels of MMP-2 and MMP-3 were significantly higher in patients who died and had an association with this end point on univariate analysis (Table 3). For MMP-2, this association remained on multivariate analysis (HR 2.45, P=0.009) along with NT-proBNP (HR 2.87, P<0.001), eGFR, age and prior history of MI. In a separate model, MMP-2 continued to add independent information {HR, 6.60 [95% CI (confidence interval), 2.25–19.36]; P=0.001} after adjustment for GRACE scores, which were also associated with poor outcome [HR, 1.03 (95% CI, 1.01–1.04); P<0.001] along with NT-proBNP [HR, 4.62 (95% CI, 2.06–10.36); P<0.001], but no other variables. Kaplan–Meier analysis revealed that MMP-2 levels in the top quartile were associated with higher mortality rates than those in the lower three quartiles (pooled comparison log-rank test 12.49, P=0.006; Figure 1). MMP-3 levels did not predict death after adjustment for clinical variables

View this table:
Table 3 Association between MMP-2, MMP-3 and MMP-9 and all-cause mortality by univariate and multivariate Cox proportional hazards analysis

Values are HRs (95% CI).

Figure 1 Kaplan–Meier curve showing the difference in mortality in patients with MMP-2 levels in the top quartile compared with first three quartiles (P=0.006)

Of note, we found that a Killip class >1 and impaired LV function were independent predictors of mortality in the base model (P=0.043 and P=0.033 respectively), but were displaced when biomarkers were added (P=0.269 and P=0.137 respectively). Log10-troponin I was not predictive of death in this cohort with probable reasons for this discrepancy being that levels were unlikely to be peak, the cohort are particularly high risk and also possibly not large enough to illustrate this relationship. The type of MI, STEMI or NSTEMI, did not predict mortality.

The areas under the receiver operating characteristic curves for MMP-2 and MMP-3 as predictors of 1 year mortality were 0.60 and 0.58 respectively, compared with 0.84 for NT-proBNP and 0.82 for the GRACE score (all at P<0.001). Models that combined MMP-2 levels with NT-proBNP or the GRACE score did not reveal a significant improvement in the AUC (area under the curve) over either marker alone.

Secondary end points

MMP-2 levels tended to be higher in patients readmitted with HF and were associated with this outcome on univariate Cox analysis (HR 2.18, P=0.015) along with age, gender, previous MI, hypertension, diabetes, Killip class >1, eGFR, impaired LV function, MMP-2 and NT-proBNP. This weak relationship was lost (P=0.29) after adjustment for baseline variables on multivariate analysis with only Killip class >1 (HR 4.01, P<0.001), NT-proBNP (HR 1.62, P=0.027) and age (HR 1.03, P=0.036) remaining. Impaired LV function predicted HF readmission in a base model (P=0.019) until biomarkers were added (P=0.071). Elevated MMP-3 had a weak association with this outcome (HR 1.70, P=0.11) that was lost on multivariate analysis, and MMP-9 levels did not have a relationship with HF.

Variables that predicted re-infarction on univariate analysis included previous history of hypertension, diabetes, Killip class >1, eGFR, impaired LV function, NT-proBNP levels and the GRACE score with none of the MMPs achieving statistical significance. On multivariable analysis, only the GRACE score (HR 1.01, P=0.018) and presence of diabetes (HR 1.59, P=0.019) retained independent associations with re-infarction.


The present single centre study performed in a substantial cohort of patients with AMI revealed that MMP-2 levels predicted the primary end point of all-cause mortality independent of clinical characteristics, NT-proBNP and the GRACE clinical risk score. These are novel results post-AMI with MMP-2, as previous studies have demonstrated prognostic utility in patients with congestive HF [21], but not in a smaller population with stable coronary disease [20]. The positive correlation of MMP2 with NT-proBNP, albeit weak, has been identified previously [15] and provides a degree of internal consistency. However, NT-proBNP or the GRACE score are superior in risk stratification in the present cohort and the results do not support utilizing MMP-2 as a prognostic tool over these markers.

A possible explanation for this interesting link between highly elevated MMP-2 levels (top quartile) and mortality is offered by two studies [8,9] which compared the effect of experimental MI in wild-type (WT) mice and genetically modified MMP-2-KO (knockout) mice. They found significantly improved survival in MMP-2-KO mice that was driven by a marked reduction in myocardial rupture and less adverse remodelling, due to reduced ECM degradation and inflammatory cell infiltration, resulting in the maintenance of the normal collagen framework that is often absent in ruptured myocardium [31]. Similar mortality reductions were observed in mice treated with a selective MMP-2 inhibitor [9]. However, a recent study in humans did not show a significant difference in MMP-2 levels in post-AMI patients with/without cardiac rupture at autopsy [10]. Consequently, one could postulate that it is more likely that the relationship between mortality and MMP-2 levels, as demonstrated in the present study in humans, is secondary to the long term pro-fibrotic effects of MMP-2 on ECM [32] and remodelling or through atherosclerotic plaque instability [17], rather than entirely due to an increase in cardiac rupture rates. This association of MMP-2 levels with mortality raises the question of transient, timely and possibly most importantly selective MMP-2 inhibition aimed at those with the highest MMP-2 levels. MMP-2 inhibition has been effective in dogs [32] and may herald a potential therapy post-AMI requiring further investigation in humans.

In the present study, we also found elevated levels of MMP-2 in patients with STEMI over those with NSTEMI, which is in line with previous findings demonstrating higher levels of MMP-2 in patients with AMI over stable angina over controls [33]. Post-infarction remodelling is greater after STEMI, and higher MMP-2 in these patients and those with elevated Killip class supports the hypothesis that this enzyme may contribute towards this process. MMP-2 levels were predictive of readmission with HF on univariate, but not multivariate, analysis and MMP-2, MMP-3 nor MMP-9 levels were predictive of re-infarction.

With regard to MMP-3, we found elevated levels in patients with echocardiographic evidence of LV dysfunction, a weak correlation with NT-proBNP levels and a univariate association with the primary end point, which are all consistent with a previous report from our centre performed on predominantly STEMI patients [12]. Although a prognostic role for MMP-3 has been shown in patients with stable coronary heart disease [20], the initial association with mortality observed in the present study was lost after adjustment for clinical factors, and the results do not support using MMP-3 levels as a prognostic tool.

In contrast with several previous reports, we did not find an association between MMP-9 and echocardiographic parameters of LV function [11,13,14,34], which may be because our study design was restricted by a single blood sample taken on day 4. MMP-9 is recognized to be highest early after the onset of an AMI, coinciding with the inflammatory response and possible release from neutrophils [4,35]. As a marker of prognosis, MMP-9 is useful in patients with stable coronary disease [21], but its utility post-MI is less clear [11,23,36] and we did not identify a relationship between MMP-9 levels taken on day 4 post-AMI and adverse events. MMP-9 levels were lower in patients with STEMI and it has been reported that a larger fall in MMP-9 relates to more severe LV impairment [11]. Therefore the lower levels observed in patients with STEMI in the present study may be due to MMP-9 falling further by day 4 than in patients with NSTEMI. The correlation between MMP-9 and NT-proBNP levels observed in the present study has been recognized previously [11,16].


Despite the cohort being large, this is a single centre study and the results require replication by other groups. Furthermore, this study has established an association between MMP-2 levels and mortality, but causation cannot be inferred. The cohort did not include patients with troponin-negative ACS. We sampled blood at one time point during the recovery phase and, although post-ACS studies suggest this is appropriate for MMP-2 and MMP-3, as recognized above, MMP-9 peaks early and either admission levels or the change in concentration between serial samples may have shown a stronger relationship with outcome [2,4,1113,35]. We did not perform separate analyses to differentiate between active and latent forms of the MMPs; however, the ELISA method is in keeping with previous studies.


This is the first single centre study that identifies MMP-2 levels as an independent predictor of all-cause mortality in a substantial cohort of post-ACS patients. However, NT-proBNP and the GRACE score retain prognostic superiority. MMP-3 and MMP-9 were not predictive of adverse events.


This work was supported by The British Heart Foundation [Junior Research Fellowship grant numbers FS/05/004 (to O.S.D.) and FS/03/028/15486 (to S.Q.K.)].

Abbreviations: ACS, acute coronary syndrome; ANP, atrial natriuretic peptide; BNP, brain natriuretic peptide; CI, confidence interval; ECM, extracellular matrix; GFR, glomerular filtration rate; eGFR, estimated GFR; GRACE, Global Registry of Acute Coronary Events; HF, heart failure; HR, hazard ratio; KO, knockout; LV, left ventricular; LVEF, LV ejection fraction; LVWMI, LV wall motion index; MAE, methyl-acridinium ester; MCE, major cardiac event; MI, myocardial infarction; AMI, acute MI; MMP, matrix metalloproteinase; NT-proBNP, N-terminal-proBNP; STEMI, ST segment elevation MI; NSTEMI, non-STEMI


View Abstract