Article Text

Download PDFPDF

Original article
Growth-differentiation factor 15 as predictor of mortality in acute reperfused ST-elevation myocardial infarction: insights from cardiovascular magnetic resonance
  1. Ingo Eitel,
  2. Patrick Blase,
  3. Volker Adams,
  4. Lysann Hildebrand,
  5. Steffen Desch,
  6. Gerhard Schuler,
  7. Holger Thiele
  1. Department of Internal Medicine/Cardiology, University of Leipzig, Heart Center, Leipzig, Germany
  1. Correspondence to Dr Ingo Eitel, Department of Internal Medicine/Cardiology, University of Leipzig, Heart Center, Strümpellstrasse 39, Leipzig 04289, Germany; ingoeitel{at}


Objective Growth-differentiation factor 15 (GDF-15), a stress-responsive member of the transforming growth factor beta cytokine superfamily, has emerged as a biomarker of increased mortality in cardiovascular disease. However, the exact pathophysiological mechanisms of GDF-15 in the cardiovascular system and in acute ST-elevation myocardial infarction (STEMI) are not well defined. The aim of this study was to determine the relation between GDF-15 and myocardial damage assessed by cardiovascular magnetic resonance (CMR) imaging and to evaluate the prognostic value of GDF-15 in a high-risk STEMI population exclusively reperfused by primary angioplasty.

Design, setting, patients GDF-15 concentrations were determined by an ELISA in 238 consecutive patients undergoing primary angioplasty in STEMI less than 12 h after symptom onset. Patients were categorised into two groups defined by the median GDF-15 value on admission. CMR was performed 3 days (IQR 2–4) after infarction for assessment of infarct size, myocardial salvage and microvascular obstruction. The primary clinical endpoint was mortality within 6 months after the index event.

Results Elevated GDF-15 concentrations over and above the median on admission were a strong predictor of mortality (19 vs one death, p<0.001) and major adverse cardiac events (27 vs nine events, p=0.001) at 6 months follow-up. Myocardial salvage was an inverse multivariable predictor of GDF-15 concentrations.

Conclusions GDF-15 on admission is a strong predictor of mortality in patients with STEMI reperfused by primary angioplasty, which is associated with decreased myocardial salvage and subsequent adverse clinical outcome.

Clinical trail registration number NCT00463749.

  • Cardiovascular magnetic resonance
  • growth-differentiation factor-15
  • MRI
  • prognosis
  • ST-elevation myocardial infarction
View Full Text

Statistics from

Despite remarkable advances in the treatment of acute ST-elevation myocardial infarction (STEMI), substantial patient mortality and morbidity remains.1 The role of biomarkers is of increasing interest in the risk stratification of STEMI patients. Biomarkers reflecting different disease pathways may have the potential advantage of improving the assessment of future major adverse cardiovascular events (MACE) as well as for decision-making for intensified treatment after the acute event. Growth-differentiation factor 15 (GDF-15), which is a stress-responsive member of the transforming growth factor beta cytokine superfamily, has recently emerged as a new biomarker of increased mortality and cardiac events in patients diagnosed with non-STEMI and STEMI.2–8 Moreover, a single measurement of GDF-15 on admission provides strong and independent prognostic information in a heterogeneous population of patients with chest pain as well as in patients with heart failure and pulmonary embolism.9–13

GDF-15 may be associated with reperfusion injury, infarct size and/or left ventricular dysfunction, however, the exact pathophysiological relation remains unknown and it is presumed that GDF-15 may provide insight into a distinct pathophysiological process in acute infarction.2–5

Cardiovascular magnetic resonance (CMR), electrocardiographic and angiographic parameters are also of prognostic relevance in STEMI patients. In particular, CMR has emerged as a powerful tool for the assessment of infarct size, global and regional functional recovery as well as prognostication of MACE in STEMI patients.14–16 To date, this imaging technique has not been applied to assess the relationship between GDF-15 and myocardial salvage, infarct size as well as microvascular obstruction, a marker of adverse infarct healing, adverse left ventricular remodelling and no-reflow with subsequent poor prognosis.14 15

To obtain further insight into the pathophysiology of GDF-15 we sought to determine whether GDF-15 relates to markers of myocardial damage (infarct size, myocardial salvage, microvascular obstruction) assessed by CMR as well as other established prognostic parameters of acute reperfused STEMI. A secondary objective was to evaluate the prognostic value of GDF-15 in a consecutive STEMI population exclusively treated by primary percutaneous coronary intervention (PPCI).


Study population

This prospective study was part of the Leipzig Immediate PercutaneouS coronary Intervention Acute myocardial infarction N-ACetylCysteine (LIPSIA-N-ACC) trial, which compared N-acetylcysteine with placebo for reperfusion injury prevention in STEMI patients and that did not show a difference between the treatment groups. The detailed design and main results of the trial have previously been published.17 In brief, patients with STEMI undergoing PPCI were eligible for this trial, if symptoms lasted less than 12 h and if ST-segment elevation of at least 0.1 mV in two or more extremity leads or at least 0.2 mV in two or more precordial leads was present.

Primary angioplasty and subsequent treatment

PPCI was performed as described previously.17 In brief, the use of bare metal or drug-eluting stents was left to the discretion of the interventional cardiologist. The additional use of thrombectomy was recommended depending on relevant thrombus. All patients were treated with aspirin, heparin and clopidogrel. The use of glycoprotein IIb/IIIa-inhibitors, ACE inhibitors, beta-blockers and statins was strongly recommended according to guidelines.18

Angiographic and electrocardiographic analysis

Coronary angiography of the target lesion was performed as described previously.17 In brief, angiographic analysis included initial and final flow of the culprit vessel. Visual assessments were performed offline in the angiographic core laboratory by two blinded observers.

For ECG interpretation cumulative ST-segment resolution approximately 90 min after percutaneous coronary intervention (PCI) expressed as a percentage was calculated by two blinded observers as defined previously.17

Assessment of GDF-15

For GDF-15 concentration measurement blood samples were drawn after vascular puncture before angiography in the catheterisation laboratory and 1 day after PCI. The plasma was separated by centrifugation at 3000g for 10 min at 4°C. Aliquots were stored at −80°C in appropriate cuvettes until assayed. The concentration of GDF-15 was determined by a recently established and validated ELISA (R&D Systems, Abingdon, Oxfordshire, UK),19 which has also been utilised in previous studies.3 5 10 11 20 All biomarker measurements were performed in duplicate by investigators who were unaware of patients' characteristics and outcome.

As previously shown, a GDF-15 value of 1200 ng/l corresponds to the upper limit of normal in healthy, elderly individuals,17 whereas GDF-15 levels of 1200 and 1800 ng/l allowed an identification of patients at low (<1200 ng/l), intermediate (between 1200 and 1800 ng/l), or high risk (>1800 ng/l).3

Other biomarkers

Plasma samples for creatine kinase (CK) and the CK myocardial band fraction were collected on admission and subsequently during the hospitalisation every 6 h for 2 days. Creatinine levels were analysed at admission and every day in the morning for the next 3 days.

Cardiovascular magnetic resonance

Infarct size, myocardial salvage and microvascular obstruction were measured with a 1.5 T CMR scanner (Philips Intera CV, Best, The Netherlands) at days 2–4 after the index event, as described previously.17 For area at risk (AAR) determination short-axis slices covering the whole ventricle using a T2-weighted triple inversion recovery breath-hold pulse sequence were obtained. Late enhancement images covering the whole ventricle were acquired approximately 15 min after intravenous administration of 0.2 mmol/kg bodyweight of gadobutrol (Gadovist, Schering, Germany). A three-dimensional inversion-recovery turbo gradient echo sequence was used for image acquisition.

Off-line image analysis was performed on an independent workstation with dedicated software (ViewForum release 5.2, Phillips Medical Systems, Best, The Netherlands) by two fully blinded operators (IE, HT) at the CMR core laboratory. Infarcted myocardial mass, AAR, microvascular obstruction and salvaged myocardium (AAR minus infarct size) were assessed as previously described (figure 1).17 21 The CMR core laboratory has excellent reproducibility and low interobserver and intraobserver variability for infarct size and myocardial salvage index (MSI) assessment.22 For MSI assessment, the bias and limits of agreement are −0.3±5.0.

Figure 1

Short axis slices of a T2-weighted image (A, B) showing oedema contours (red) and the corresponding late enhancement image (C, D) with endocardial (green), epicardial (yellow), infarcted (red) and microvascular obstruction (dark blue) contours in a patient with a growth-differentiation factor 15 concentration of the median or above on admission.

Clinical endpoints

The primary endpoints of this study were mortality and the occurrence of MACE defined as a composite of death, reinfarction, and new congestive heart failure within 6 months after the index event. Post-hospital follow-up included one outpatient visit at 6 months. The diagnosis of reinfarction and new congestive heart failure were defined as previously described.17 To avoid double counting of patients with more than one event, each patient contributed only once to the composite MACE endpoint (death > recurrent myocarcial infarction > congestive heart failure).

Statistical analysis

Patients were grouped by the median GDF-15 value on admission into a less than the median and a median or above GDF-15 group. Each categorical variable is expressed as the number and percentage of patients. Most continuous variables had non-normal distribution and are therefore presented as medians together with IQR. Differences between groups were assessed by Fisher's exact or the χ2 test for categorical variables and by the Student's t test for continuous data with normal distribution. Otherwise the non-parametric Wilcoxon rank-sum test was used. Correlation analyses were performed by Pearson or Spearman tests, as indicated.

Univariable and multivariable linear logistic regression analyses were performed to identify predictors of GDF-15 values on admission and GDF-15 increase after PCI. Multivariable regression was performed using only variables with a probability value of less than 0.05 in univariable regression analyses.

The Kaplan–Meier method was used to illustrate the timing of events during 6 months follow-up; statistical assessment was performed using the log-rank test. Simple Cox regression analysis was used to identify predictors of death and MACE during 6 months. Significant variables (p<0.05) were then tested in a stepwise multiple Cox regression analysis. Because myocardial salvage has infarct size accounted for by its inclusion in the formula for the calculation of myocardial salvage, we included only infarct size in the Cox regression analysis to avoid collinearity.

For additional comparison of the prognostic values of GDF-15 on admission, infarct size, microvascular obstruction as well as significant parameters in multiple Cox regression analysis with regard to death, receiver operating characteristic (ROC) curves were generated and the areas under the curves (AUC) were calculated.

All statistical tests were performed with SPSS software, version 15.0. A two-tailed p value less than 0.05 was considered statistically significant.


This prospective study included 238 patients (figure 2). In all patients GDF-15 concentrations before and 1 day after reperfusion as well as clinical outcome data 6 months after the index event were available. Results of CMR were available in 207 patients. Reasons for not pursuing CMR are listed in figure 2.

Figure 2

Study flow diagram. CMR, cardiovascular magnetic resonance; GDF-15, growth-differentiation factor 15; STEMI, ST-elevation myocardial infarction.

GDF-15 concentrations and patient characteristics

The concentration of GDF-15 on admission ranged from 545 to 9741 ng/l, with a median of 1319 ng/l (IQR 1023–1977). GDF-15 concentrations on admission and the increase in GDF-15 after PCI did not differ significantly between the patients randomly assigned to N-acetylcysteine or placebo (p=0.59 and 0.58, respectively).

Demographic and clinical characteristics are shown in table 1. Patients with a GDF-15 value of median or above on admission were significantly older and were more likely to have diabetes, a history of cardiovascular disease, or to present with a higher Killip class.

Table 1

Patient characteristics

One day after infarction, there was a significant increase of 11% in GDF-15 plasma concentrations (1460 ng/l, IQR 1101–2186; p<0.001).

Angiographic analysis and ST-segment resolution

Postprocedural thrombolysis in myocardial infarction (TIMI) flow grades were significantly better in the GDF-15 less than median group (table 1). There was no significant correlation between ST resolution and GDF-15 values on admission, whereas the GDF-15 value increase 1 day after PCI correlated inversely with ST resolution (r=−0.214, p=0.001). Creatinine levels on admission were significantly higher in the GDF-15 value of median or above on admission group, but showed only a weak correlation with GDF-15 values on admission (r=0.243, p<0.001).

Cardiovascular magnetic resonance

The median time between the index event and CMR was 3 days (IQR 2–4) for both groups. Patients with a GDF-15 concentration less than the median on admission had a significantly larger AAR, whereas no significant differences in infarct size could be observed between the groups (table 2). Consequently, significantly more myocardium at risk was salvaged in the GDF-15 less than median on admission group and in patients with normal (<1200 ng/l) versus elevated GDF-15 concentrations on admission (figure 3). Microvascular obstruction was identified in 156 (78%) patients with similar occurrence (76 vs 80 patients, p=0.53) and extent in both groups. A major GDF-15 increase in median or above 1 day after successful PCI was significantly associated with larger infarct size, extent of late microvascular obstruction and reduced left ventricular ejection fraction (figure 4A–C).

Table 2

Cardiovascular magnetic resonance results

Figure 3

Growth-differentiation factor 15 (GDF-15) concentrations on admission and amount of myocardial salvage.

Figure 4

Box (25th percentile, median and 75th percentile) and whisker (10th and 90th percentiles) plots of infarct size, microvascular obstruction and left ventricular ejection fraction according to absolute growth-differentiation factor 15 (GDF-15) concentration increase 1 day after reperfusion (A–C). Ten patients who died were not included in the cardiovascular magnetic resonance analysis. LV, left ventricle.

Predictors of GDF-15 on admission and GDF-15 increase after PCI

In a multivariable regression model adjusted for significant variables in univariable regression analysis using the natural logarithm of GDF-15 at admission as the dependent variable, age, serum creatinine on admission, previous infarction, MSI and diabetes were the strongest predictors for GDF-15 concentrations on admission (table 3).

Table 3

Significant predictors of GDF-15 concentrations on admission in univariable and multivariable linear regression analysis

Significant predictors of a GDF-15 increase after PCI are displayed in table 4. In multivariable regression analysis, infarct size, age and anterior infarction emerged as significant predictors of a GDF-15 increase after PCI (table 4).

Table 4

Significant predictors of a GDF-15 increase 1 day after PCI in univariable and multivariable linear regression analysis

Clinical outcome

At 6 months follow-up, there were 19 deaths (16%) in the median or above GDF-15 value on admission group and one (1%) in the GDF-15 less than the median group (p<0.001; figure 5A). Patients who died during follow-up had significantly higher median concentrations of GDF-15 on admission when compared with patients who survived (2545 ng/l (IQR 1869–4799) vs 1228 ng/l (IQR 979–1833), p<0.001). There was also a graded relationship between GDF-15 levels on admission and mortality. Only one death occurred in the subgroup with normal GDF-15 values (<1200 ng/l), compared with three deaths in the intermediate risk group (1200–1800 ng/l) and 16 deaths in the high-risk group (>1800 ng/l; p<0.001).

Figure 5

Kaplan–Meier curve of the incidence of all-cause mortality (A) and of the cumulative incidence of death, reinfarction and new congestive heart failure (B) during the first 6 months after infarction. GDF-15, growth-differentiation factor 15.

The causes of death were as follows: cardiogenic shock (n=6), recurrent infarction (n=6), sudden cardiac death (n=1), stroke (n=4), cancer (n=2), unknown (n=1). In the GDF-15 value less than the median group the only patient died as a result of recurrent infarction (GDF-15 value on admission: 1071 ng/l). Non-fatal reinfarctions and congestive heart failure were similar between groups (eight vs eight events). Consequently, at 6 months follow-up MACE were significantly lower in the GDF-15 value on admission less than the median group (8% vs 23%, p=0.001; figure 5B).

A GDF-15 concentration increase of the median or above (≥110 ng/l) 1 day after reperfusion irrespective of the GDF-15 values on admission was also associated with an increased mortality (15 vs five deaths, p=0.01).

In addition to GDF-15 on admission, several established markers of increased patient risk were associated with an increased mortality at 6-month follow-up by simple Cox regression analysis (table 5). Infarct size (p=0.054) and microvascular obstruction (p=0.051) showed only borderline associations with mortality and were therefore not included in the multivariable analysis. Using stepwise multiple Cox regression analysis, only GDF-15 concentration on admission, ST resolution, Killip class on admission and age emerged as independent predictors of mortality (table 5).

Table 5

Predictors of mortality in univariable and stepwise multivariable cox regression analysis

Significant predictors of MACE using simple Cox regression analysis are shown in table 6. On multivariable analysis, the only significant predictors of MACE were infarct size and left ventricular ejection fraction.

Table 6

Predictors of major adverse cardiac events in univariable and stepwise multivariable Cox regression analysis

Although a GDF-15 concentration increase after PCI had an unadjusted association with mortality (HR 1.82, CI 1.25 to 2.63, p=0.002) and MACE (HR 1.61, CI 1.12 to 2.16, p=0.001), it was unable to predict clinical events independently in multivariable Cox regression analysis.

ROC curve analyses of GDF-15 and other predictors of mortality

ROC curve analysis further illustrated that GDF-15 concentrations on admission are a strong indicator of mortality, with an AUC of 0.74 compared with infarct size (AUC 0.70) and microvascular obstruction (AUC 0.68). GDF-15 was also the strongest indicator of mortality compared with other significant predictors of mortality in Cox regression analysis (Killip class, age, ST resolution; data not presented). The best GDF-15 level for the prediction of 6-month mortality was 1974 ng/l (sensitivity 75%; specificity 79%).


To the best of our knowledge this is the first prospective trial that determines the prognostic value of GDF-15 in a STEMI population exclusively reperfused by PPCI and the first trial that correlates GDF-15 concentrations with CMR parameters of myocardial damage and reperfusion injury (AAR, myocardial salvage, infarct size and microvascular obstruction).

The main findings are: GDF-15 values on admission are a very strong predictor of mortality in STEMI patients exclusively reperfused by PPCI; GDF-15 values on admission were inversely correlated with myocardial salvage; a GDF-15 increase above the median after reperfusion is associated with larger infarcts, impaired microvascular perfusion and left ventricular function.

Role of GDF-15 in the cardiovascular system

GDF-15 is a distinct member of the transforming growth factor beta cytokine superfamily that is only weakly produced under normal conditions in most tissues.23 However, in animal models its expression is strongly induced in the myocardium in response to ischaemia, reperfusion injury, pressure overload and heart failure.20 24 25 In accordance with these findings, median GDF-15 concentrations were elevated in the present STEMI population. After PCI there was an additional increase in GDF-15 levels of 11%. Ischaemia may thus result in a rapid induction of GDF-15, followed by a prolonged expression in the reperfused myocardium, which may be related to increased levels of oxidative stress, inflammation and infarct healing.2 This is supported by the observed upregulation of GDF-15 concentrations in myocardial samples of patients with fatal myocardial infarction.17 Therefore, cardiomyocytes in the ischaemic area seem to contribute to the induction of GDF-15 in the infarcted human heart and even higher GDF-15 concentrations are probably achieved at the tissue level.7

However, it is still unknown whether elevated GDF-15 values may have a potential protective or detrimental effect on the cardiovascular system. Recent experimental studies demonstrated that endogenous GDF-15 may limit myocardial tissue damage,24 25 but failed to provide any pathophysiological basis for the association of GDF-15 with future cardiovascular events. In the present study, elevated GDF-15 concentrations on admission were significantly associated with less myocardial salvage (figure 3). Recent trials have shown that the prognosis in patients with acute reperfused STEMI is directly related to the amount of myocardial salvage.21 26 Therefore, GDF-15 concentrations on admission might carry prognostic information contained on this important outcome marker and may predict the potential benefit of PPCI. Surprisingly, we observed no significant correlation between GDF-15 concentrations on admission and AAR, infarct size as well as reperfusion injury (microvascular obstruction) assessed with CMR. These findings suggest that the prognostic value of GDF-15 values at presentation might be unrelated to the extent of myocardium at risk, the extent of final myocardial necrosis and microvascular injury. This is in line with previous studies, which showed that GDF-15 concentrations on admission do not differ significantly between patients with non-STEMI or STEMI patients, who are considered to have more severe myocardial damage.4 5 There was also no association between GDF-15 on admission and the time from symptom onset to reperfusion, TIMI flow before PCI as well as ST resolution, another important prognostic parameter of successful reperfusion that mirrors tissue and microvascular perfusion.

Higher GDF-15 concentrations at presentation were independently related to age, diabetes, previous infarction, Killip class, serum creatinine and decreased myocardial salvage, suggesting that GDF-15 integrates information from different disease pathways in acute STEMI. Therefore, increased GDF-15 concentrations on admission might reflect, to some extent, premorbid conditions including cardiovascular risk (eg, age, diabetes) associated with endothelial dysfunction, oxidant stress and inflammation as well as the extent of atherosclerotic disease (plaque burden) and other prognostic relevant diseases (eg, renal insufficiency). Such a relation of GDF-15 to chronic cardiac and vascular pathologies is supported by recent epidemiological studies in elderly individuals.19 20

However, we found that a major increase in GDF-15 after reperfusion, irrespective of the GDF-15 concentrations on admission, is related to the AAR, infarct size, left ventricular dysfunction and markers of reperfusion injury (microvascular obstruction, ST resolution). Increased GDF-15 expression with irreversibly damaged myocardium has also been demonstrated in an animal model.24 Consequently, in acute reperfused STEMI an additional GDF-15 increase takes place and appears to be a marker of the extent of myocardial damage including infarct size (similarly to CK, troponin) and microvascular injury.

The findings of our study indicate that elevated GDF-15 concentrations on admission and a major GDF-15 increase after PCI might promote or at least reflect detrimental effects in the clinical setting. Our data thus do not support a cardioprotective effect of GDF-15, as shown in experimental models.24 25 However, the strong prognostic value of GDF-15 independent from established markers of adverse events in acute myocardial infarction may represent a strength of this new biomarker, reflecting several clinical and biochemical indicators of severe disease and poor prognosis. Consequently, research on GDF-15 should be focused on an improved understanding of the underlying pathophysiology of GDF-15 as a potential treatment target.

Prognostic role of GDF-15

GDF-15 has emerged as a new prognostic marker in different cardiovascular diseases.2 3 5 6 8–13 Our trial confirms for the first time the incremental value of GDF-15 as a biomarker of increased mortality in STEMI patients exclusively reperfused by PPCI. Consistent with previous trials in STEMI patients exclusively4 or mainly5 reperfused by thrombolyis, mortality curves showed a very early separation, indicating that the measurement of GDF-15 may enable risk stratification for a time window when important management decisions are made.5 As PPCI has evolved as the preferred reperfusion strategy in STEMI patients our results are more generalisable to a real world STEMI population.18

In addition, we found that a major increase in GDF-15 irrespective of the GDF-15 concentrations on admission might be an additional marker of poor prognosis in STEMI patients. Therefore, increased GDF-15 concentrations on admission as a marker of chronic myocardial/vascular damage and salvageable myocardium together with a significant GDF-15 increase after PCI reflecting ischaemic/reperfusion injury might enhance the clinician's ability to assess risk in infarction rapidly and accurately. Notably, GDF-15 provides prognostic information over and above established clinical factors, biomarkers (eg, brain natriuretic peptide, troponin)3 5 and CMR parameters of prognosis (infarct size, microvascular obstruction) and might have the potential to improve the prognostication of outcome in STEMI. However, further studies are necessary to define whether its plasma values will be useful as a tool of therapeutic decision-making in the management and triage of STEMI patients.


This study represents a single-centre experience with a limited number of patients. Some patients had to be excluded from CMR infarct size, microvascular obstruction and myocardial salvage assessment. Because the number of patients in both groups was identical and the baseline characteristics of patients undergoing and those not undergoing CMR were similar, a potential selection bias is limited. Moreover, these patients were included in the analysis of clinical adverse events. Infarct size and especially microvascular obstruction are dynamic processes after acute infarction. Imaging time after reperfusion is thus an important determinant for the presence and extent of microvascular obstruction as well as infarct size. As in our study both groups underwent CMR after a similar time delay a potential bias is unlikely. Finally, the ideal study design would include a pre and post-intervention CMR study to discern the effects of PPCI from initial myocardial injury and to elucidate further the exact relationship between GDF-15 concentrations on admission and myocardial damage. However, from an ethical point of view such a delay of reperfusion is not acceptable.14 16 19


GDF-15 on admission is a new, powerful predictor of mortality in patients with STEMI reperfused by PPCI, which is associated with decreased myocardial salvage and provides prognostic information beyond established clinical, biochemical and CMR parameters such as infarct size and microvascular obstruction.


View Abstract


  • Funding This work was supported by Rhön Klinikum AG, Germany.

  • Competing interests None.

  • Patient consent Obtained.

  • Ethics approval This study was conducted with the approval of the University of Leipzig.

  • Provenance and peer review Not commissioned; externally peer reviewed.

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.