Growth Differentiation Factor-15 Levels at Admission Provide Incremental Prognostic Information on All-Cause Long-term Mortality in ST-Segment Elevation Myocardial Infarction Patients Treated with Primary Percutaneous Coronary Intervention

This study demonstrates that higher GDF-15 levels at admission are associated with 10-year all-cause mortality in STEMI patients treated with pPCI. This relation was independent of clinical risk factors and biomarkers. Moreover, GDF-15 levels at admission have additional prognostic value beyond identified risk factors and other cardiac biomarkers such as cTn and NTproBNP as analysed by Chi-square test and C-statistics.

When stratified by median GDF-15 levels, patients with a level below the median show an excellent prognosis with a 10-year all-cause mortality rate of 6.2% compared to 19.3% in the group with a GDF-15 level above the median. NTproBNP levels stratified by the median show the same division of 10-year all-cause mortality rates as GDF-15 levels. However, combining these two biomarkers reveals that they have a complementary relation with 10-year all-cause mortality. So, the combination of these biomarkers seems to identify an interesting group of high-risk patients. In the group of patients with both GDF-15 and NTproBNP levels below the median, only 3 patients (3.8%) died within 10 years compared to 22 (27.8%) in the group with both GDF-15 and NTproBNP levels above the median. Furthermore, GDF-15 shows additional prognostic information when added to clinical information and NTproBNP.

Other studies in a broad spectrum of patients have shown that a high level of GDF-15 is independently associated with mortality and adds extra prognostic value on top of various clinical characteristics and biomarkers [12, 14‐18, 24‐26]. Several of these studies compared the GDF-15 levels of STEMI patients at admission with outcome [14‐16]. Two of these studies used a comparable population with GDF-15 levels at admission of STEMI patients treated with pPCI. Eitel et al. demonstrated that GDF-15 levels at admission are a strong predictor of mortality after 6 years [16]. However, two important biomarkers, cTn and NTproBNP, were not available is this cohort. Recently, Velders et al. showed in a large cohort that GDF-15 is independently associated with cardiovascular death after 1 year after adjusting for these biomarkers, the severity of cardiovascular disease and other clinical information [14]. However, the number of studies that compared GDF-15 and NTproBNP levels at admission with all-cause mortality on the long term in STEMI patients treated with pPCI is virtually absent. To our knowledge the current study is the first that investigated these biomarkers at admission in relation to 10-year all-cause mortality in STEMI patients treated with pPCI.

A relatively low number of patients died in this cohort during the 10 years. In the first year of follow-up after STEMI only 5 patients (1.7%) died, of whom 4 died from a cardiac cause. In the years that followed, 32 more patients died, of whom 6 died from a cardiac cause and 4 patients’ deaths were likely of cardiac origin. In total 19 patients died of a non-cardiac cause, of which 18 were after more than 1 year. By adding biomarkers that reflect a more general state of disease, one might not only capture patients prone to cardiovascular events but also patients prone to non-cardiac events.

Current risk assessment with for example the TIMI or GRACE risk scores is mainly based upon clinical characteristics [14]. Cardiac biomarkers such as cTn and NTproBNP have been shown to improve risk prediction [8, 9]. Using a multimarker strategy that captures a broader spectrum of diseases may have added value since it can reveal novel release mechanisms and therefore potential therapeutic targets. The relation of GDF-15 to cTn and NTproBNP has shown that GDF-15 is involved in cardiac pathologies. GDF-15 is induced in the myocardium after ischemia [11]. Several studies have demonstrated that plasma levels of GDF-15, just as NTproBNP, are associated with mechanical stretch, left ventricular mass, concentric left ventricular hypertrophy, reduced left ventricular ejection fraction, and heart failure [12, 15, 27]. Besides these similarities, the characteristics of GDF-15 release are distinct from those of NTproBNP and cTn. Earlier studies demonstrated that cTn release shows a typical rise and fall pattern, indicating the release from dying cardiomyocytes, whereas GDF-15 values increase within hours after ischemia [11] but remain remarkably stable over time during admission [28] and during 6 months of follow-up [24]. GDF-15 follows several stress pathways that differ from the cardiac-specific biomarkers, like NTproBNP and cTn [11, 29]. Plasma levels of GDF-15 may increase in response to pathological stress associated with vascular inflammation, endothelial activation or tissue damage, and its overexpression has been found in several malignancies [11, 24, 27, 30, 31]. Supported by epidemiological studies, this indicates that GDF-15 seems to be a marker of chronic cardiac and vascular pathologies, and is not per se related to acute injury [27, 32].

Over the last decade, a substantial amount of research has been performed that studied the role of GDF-15 levels for the risk assessment in acute coronary syndrome patients. However, translation of novel biomarkers into clinical practice has been shown to be challenging. For example, after the discovery of NTproBNP as a marker for heart failure it took two decades to implement it into clinical practice [33]. Recently a multidisciplinary working group defined a strategy and checklist to better identify the clinical unmet needs and how novel biomarkers can satisfy them [34]. They provide a practical approach to help assess whether a new biomarker would provide clinical benefit. To give a purely hypothetical example, we designed a map for a clinical pathway using this checklist and clinical approach (supplemental figure) on how we could advance GDF-15 into clinical practice. In this figure we used GDF-15 as an add-on test on top of clinical characteristics and cTn and NTproBNP. We acknowledge that we cannot implement GDF-15 into clinical practice solely on the basis of the results of our study, especially since it is currently unclear how GDF-15 levels can be lowered by medical therapy or interventions and whether lowering these levels results in an improved outcome. However we do hope that we can encourage other stakeholders to follow this example in larger prospective intervention studies to target this unmet clinical need with the development and implementation of GDF-15 in clinical pathway mapping.

Before this can be considered, further research with regard to the pathophysiological mechanisms and the influence of common and novel medical therapies on plasma levels of GDF-15 should be explored. Two potentially ways to do so might be more aggressive lipid-lowering therapies or by anti-inflammatory therapy. The only study so far that investigated the relation between lipid lowering therapy by statin therapy and GDF-15 levels was conducted by Bonaca et al. [35]. They found no interaction of GDF-15 with different kind of statins and, moreover, GDF-15 levels did not decline after 4 months. Whether more aggressive medical therapy by for example PCSK-9 inhibitors has beneficial effects on GDF-15 is worth investigating. Another way to influence GDF-15 levels may be by anti-inflammatory therapy. In the recently published CANTOS (Canakinumab Anti-inflammatory Thrombosis Outcome Study) trial it was shown that targeting the innate immunity pathway with canakinumab, independent of lipid level lowering, led to a significant lower rate of recurrent cardiovascular events than placebo [36]. It would be of interest to explore whether GDF-15 levels may act as biomarker-guided therapy to evaluate the effect of anti-inflammatory therapy.

Several limitations need to be mentioned. First of all, this study was a single-centre randomized trial with a limited number of patients; so, the results are limited to the patients eligible for the trial, although the results apply for a broad range of STEMI patients. All patients were treated according to the institutional MISSION! which provided an integrated approach of MI care to optimize treatment. This yielded a very homogeneous STEMI population. Secondly, limited data is available about the stability of GDF-15 samples after long-term storage at − 80 °C. However, an earlier study conducted by Daniels et al. [37] measured GDF-15 in 1391 serum samples that were stored for 14–18 years. The fact that GDF-15 levels were prognostic for outcome supports that there is sufficient stability to preserve a clinical signal [37]. Moreover, the median of the GDF-15 levels in the present cohort was 37.78 pmol/L which is comparable with the median levels of GDF-15 of 38.63 pmol/L in a similar cohort of STEMI patients [16]. Third, a relatively low number of events were noted during follow-up. To make the results more robust, larger cohorts of STEMI patients treated with pPCI should be followed. An explanation for the low number of events could be the exclusion criteria of patients older than 80 years. This could have led to a relatively young cohort with a mean age of 59 years. The last issue that should be addressed is the low number of events in relation to the multivariate Cox model. This paper has investigated the added value of GDF-15 for the prediction of 10-year all-cause mortality in addition to established risk factors as well as potential confounding variables identified from univariate analysis on our data. As opposed to fitting a prognostic model, a testing procedure was used, which first estimates a multivariate Cox model to account for the joint effect of these risk factors and potential confounders, after which the added value was assessed from likelihood ratio testing for the addition of GDF-15 and NTproBNP to the base model.