Background

Acute coronary syndrome (ACS), including ST-segment elevation myocardial infarction (STEMI), non-ST-segment elevation myocardial infarction (NSTEMI) and unstable angina pectoris (UAP), result from the rupture or erosion of vulnerable atherosclerotic plaque [1], and death and recurrent myocardial infarction (MI) can occur at any time after the first attack episode. Currently, even with medication and coronary intervention, the prognosis for ACS patients is still not good [2]. Therefore, the early evaluation of ACS and its timely treatment are particularly important.

Growth differentiation factor–15 (GDF-15) is a member of the transforming growth factor β (TGF-β) superfamily. GDF-15 primarily regulates multiple cellular functions, as well as the biological processes underlying the growth of multiple organs and the differentiation and renovation of tissues [3]. Researchers have verified that GDF-15 exerts a suppressive function in the progress, invasiveness, and metastasis of tumors by facilitating the apoptosis of tumor cells [4]. In addition, GDF-15 is also related to cardiovascular diseases, such as heart failure [5], cardiac hypertrophy [6], and coronary heart disease (CHD) [7]. Recent studies have shown that GDF-15 might be a potential predictive cytokine for the prognosis of ACS and, therefore, might assist physicians in evaluating the prognoses of ACS patients [8]. Studies examining the association between GDF-15 and ACS prognosis have included research on patient mortality [7, 9–15] and recurrent MI [9–12, 15]. However, there have been discrepancies in these studies, including the target participants, follow-up period, and patient ethnicity. Hence, this meta-analysis was performed to investigate the role of GDF-15 in predicting the prognosis of ACS patients.

Methods

Inclusion criteria

We performed a meta-analysis to determine whether GDF-15 could predict the prognoses of ACS patients. Studies that had a follow-up duration of least 6 months and that examined the correlation between plasma GDF-15 levels and the prognoses of ACS patients were collected. The quality of each included study has been tested under the Meta-analysis of Observational Studies in Epidemiology (MOOSE) [16]. The MOOSE systemic assessments included the following criteria: 1) a clear explanation of the inclusion and exclusion criteria, 2) the attrition rate of the samples during the follow-up period, 3) a clear explanation of the outcomes and outcome assessments, 4) a sufficient follow-up period, 5) appropriate statistical analyses, and 6) identification of the important confounding and prognostic factors. All items had the following answer options: yes/no/unclear information to answer the question. When a criterion was fulfilled, a score of 1 was given; a score of 0 was given if a criterion was unclear, and −1 was given if a criterion was not achieved. The MOOSE assessment of each study was discussed and carefully evaluated by all authors.

Study selection

The study selection was based on an online search of the PubMed/MEDLINE and EMBASE databases, without any language restrictions; studies published before November 14, 2014 were included. The following medical subject headings were used for the publication search: “ST-segment elevation myocardial infarction, STEMI, Non-ST-segment elevation myocardial infarction, NSTEMI, Unstable angina pectoris, UAP, Acute coronary syndrome, ACS, growth differentiation factor-15, GDF-15, Placental growth factor, PGF, macrophage inhibitory cytokine-1, MIC-1”. The terms were then combined using Boolean the operators “OR, AND”. We extracted the data from individual studies that included the first author’s name, year of publication, ethnicity, participants, outcome data (mortality and recurrent MI), and maximal follow-up years, as well as methods that stratified the GDP-15 levels. The extraction was performed by two independent reviewers (SZ and DD).

Statistical analysis

Review Manager 5.3 and STATA 11.0 statistical software were used for the meta-analysis [17]. Statistical heterogeneity was tested using Cochran’s Q statistic and I2 tests [18]. The random effects model was used for all meta-analyses. The calculated GDF-15 concentration data were extracted from the initial follow-up, and p < 0.05 was considered to be statistically significant.

Results

Study selection

We retrieved 73 studies from PubMed/MEDLINE and 186 studies from EMBASE (Fig. 1). After removing duplicates, 158 of 194 studies were excluded based on title and abstract reviews. In the remaining 36 studies, there were seven reviews, nine studies based on animals, nine studies with different participants, and three studies with inadequate data. Finally, eight articles were included in our meta-analysis, including 4 NSTEMI studies, 2 STEMI studies and 2 other studies that did not clearly provide the types of ACS. No unstable angina pectoris data were found.

Fig. 1
figure 1

Flowchart of selection process in the meta-analysis

Full size image

Characteristics of the individual studies are shown in Table 1. Eight studies, including 8903 participants, investigated the association between GDF-15 and mortality. Five of these studies, including 7193 participants, investigated the association between GDF-15 and recurrent MI. The maximal follow-up duration ranged from 6 months to 6 years.

Table 1 The characteristics of the studies included in the meta-analysis
Full size table

Association between GDF-15 and the ACS prognosis

In addition to comparing the prognoses of the ACS patients with the highest and lowest GDF-15 levels (overall analyses), we also performed subgroup analyses based on the ACS types and a stratification of the GDF-15 values: <1200 ng/L, 1200–1800 ng/L, and >1800 ng/L. Moreover, based on the included studies, we stratified the follow-up duration into “>1 years” and “≤1 years”. Patient mortality and recurrent MI data were analyzed (Table 2, Additional file 1: Figure S1 and Additional file 2: Figure S2).

Table 2 The association between GDF-15 concentration and ACS prognosis
Full size table

For the mortality analyses, a significant association was found in the overall analyses (p < 0.00001, RR = 6.08, 95 % CI = 4.79–7.71) and subgroup analyses of ACS (p < 0.00001, RR = 6.03, 95 % CI =3.66–9.93), NSTEMI (P < 0.00001, RR = 5.94, 95 % CI = 4.38–8.05) and STEMI (p < 0.00001, RR = 8.24, 95 % CI = 3.35–20.25). Positive results were also revealed in the subgroup analyses stratified by GDF-15 Cutoff Point (<1200 ng/L vs. 1200–1800 ng/L:p < 0.00001, RR = 2.23, 95 % CI = 1.66–2.99; 1200–1800 ng/L vs. > 1800 ng/L:p < 0.00001, RR = 2.76, 95 % CI = 2.28–3.34; and >1800 ng/L vs. < 1200 ng/L:p < 0.00001 ng/L, RR = 6.22, 95 % CI = 4.77–8.10). As for subgroup analyses stratified by follow-up duration, positive results were also obtained for the ≤1 year (p < 0.00001, RR =8.83, 95 % CI = 5.32–14.66) and >1 year subgroups (p < 0.00001, RR = 5.47, 95 % CI = 4.18–7.16).

For the recurrent MI analyses, a significant association was observed in the overall analyses (p < 0.00001, RR = 1.76, 95 % CI = 1.49–2.07) and in the subgroup analyses of NSTEMI (p < 0.00001, RR = 1.66, 95 % CI = 1.35–2.03) and ACS (p < 0.00001, RR = 2.02, 95 % CI = 1.56–2.61). A significant association was also found when comparing the following GDF-15 cutoff points: 1200–1800 ng/L vs. >1800 ng/L (p < 0.00001, RR = 1.49, 95 % CI = 1.17–1.89) and >1800 ng/L vs. <1200 ng/L (p < 0.00001, RR = 1.66, 95 % CI = 1.35–2.03). As for the subgroup analyses stratified by the follow-up durations, a positive result was found only in the >1 year follow-up duration (p < 0.00001, RR =1.84, 95 % CI =1.54–2.19); we did not find a significant result for follow-up periods ≤1 year (P = 0.08, RR = 1.42, 95 % CI = 0.96–2.08).

Heterogeneity of the current study

High heterogeneity was found in the comparison of the recurrent MI between the GDF-15 levels of <1200 ng/L and 1200 to 1800 ng/L (I2 = 65 %). Heterogeneity was low for all the above mentioned analyses (I2 <27 %).

The combined prognostic data of hazard ratio

Hazard ratio (HR) was obtained in seven studies [7, 9, 10, 12–15], which was on the association of the GDF-15 and mortality or recurrent of MI. However, after sensitive analysis, we excluded a study [15] that could lead to high heterogeneity to the overall meta-analysis. Our results showed a significant high HR (HR = 1.656, 95 % CI = 1.467–1.871, I2 = 0 %) for GDF-15 to the ACS (Fig. 2).

Fig. 2
figure 2

Funnel plots of GDF-15 with mortality and recurrent MI

Full size image

Discussion

Our meta-analysis identified eight studies, including 8903 participants, and attempted to evaluate the predictive value of plasma GDF-15 levels in the prognosis of ACS patients by calculating the risk of mortality and recurrent MI after the ACS events. We found that higher GDF-15 levels indicated a higher risk for ACS mortality and recurrent MI. This association also remained when the GDF-15 levels were stratified by the cutoff point, which might be valuable in terms of the risk stratification of ACS prognoses because it has been previously reported that 1800 ng/L GDF-15 levels may be an ideal cutoff value for several disease states [19]. Moreover, we also found significant associations between GDF-15 levels and subsequent ACS events. Furthermore, we found high HR for GDF-15 to ACS. These findings might assist with monitoring the prognoses of ACS patients.

GDF-15 has been described as a promising cardio-protective agent [20]. Previous studies have reported that higher expression levels of GDF-15 could predict deteriorating conditions for heart disease patients, particularly ACS patients, independent of troponin or BNP levels [7, 9–15]. It has also been reported that GDF-15 plays a protective role in heart disease, which may be attributed to the anti-apoptotic, anti-inflammatory, or anti-hypertrophic effects demonstrated in animal models [21, 22]. Moreover, GDF-15 might induce angiogenesis, which plays an essential role in the recovery of damaged myocardium [23]. A prolonged increase in GDF-15 levels after episodes of ischemia and reperfusion may be associated with increased levels of oxidative stress, inflammation, and infarct healing, and the increase of GDF-15 requires for enhanced oxidative metabolism and tissue repair [9]. Accordingly, one feasible explanation for the predictive role of GDF-15 is that higher levels of GDF-15 indicate a greater extent of myocardium damage and the risk of adverse remodeling. GDF-15 is also strongly induced in other cardiovascular conditions, such as heart failure [21], Takotsubo cardiomyopathy [24], and atherosclerosis [25].

In this meta-analysis, clear inclusion and exclusion criteria were required for study inclusion. The quality of the current study was well assessed using the MOOSE system. There was no publication bias, according to the funnel plots (Fig. 3). Heterogeneity was high among the “<1200 ng/L vs. 1200–1800 ng/L” subgroup of recurrent MI patients. The following sensitive test showed that this heterogeneity disappeared (I2 = 0 %) when we removed an included study from the Netherlands [11], thus indicating that the heterogeneity might have resulted from ethnicity differences, as the remaining two studies included samples from Sweden. The other comparisons showed low or no heterogeneity (I2 < = 27 %), suggesting that our study results were reliable.

Fig. 3
figure 3

The forest plots of combined hazard rate of GDF-15 with mortality or recurrent of MI

Full size image

There are ethnic differences in GDF-15 levels. In Europeans, the median GDF-15 concentration ranged from 1244 to 1635 ng/L, whereas it was much lower in Asians (650 ng/L) [26]. However, the Asian study cohort was limited in number, and future well-designed studies with different ethnicities are warranted.

Limitations

Although individual patient data from eight eligible studies were included in our meta-analysis, this study had some limitations. First, the definitions of all-cause mortality and cardiovascular mortality were not confirmed. Among the studies included, only two studies provided detailed information [7, 9]. For the recurrent MI patients without causes of death, we only performed subgroup analyses for “recurrent MI”. Second, although the revealed heterogeneity was low for most of the comparisons, GDF-15 levels might be influenced by many factors, including age, male gender, current smoking, symptomatic heart failure, reduced kidney function, and diabetes mellitus, among others [27]. Third, the ACS patients were treated with different treatment strategies, which might alter the GDF-15 levels. Fourth, the obtained blood samples were extracted at different times, therefore the GDF-15 levels may change by different time after ACS events, however, evidence found that GDF-15 levels showed small alterations several months after ACS events [28], hence, we could not give a judgment if the different time of the tests of the GDF-15 would influence our findings. Fifth, the follow-up durations differed in each study. In addition, one study [13] recorded the initial mortality rate 6 months after the ACS occurred, which might slightly alter our study results. Sixthly, the HR of different studies were obtained at different time after the ACS happened, which may slightly influence the results, however, the combined data showed no heterogeneity (I2 = 0 %).

Conclusion

The present study suggests that GDF-15 is a strong predictor of mortality and recurrent MI in ACS prognosis. Thus, it has the potential to become a clinically useful novel biomarker that can provide independent prognostic information and also help to direct optimal treatment strategies. Current ongoing clinical trials are needed to further clarify the benefits of GDF-15 in predicting the prognoses of patients with acute coronary syndrome.

Ethics

This study is not involved in any ethical issues.