The association between soluble suppression of tumorigenicity-2 and long-term prognosis in patients with coronary artery disease: A meta-analysis

Objective Findings regarding the prognostic value of soluble suppression of tumorigenecity-2 (sST2) in patients with coronary artery disease (CAD) remain inconsistent. Therefore, we conducted this meta-analysis to investigate the long-term prognostic value of sST2 in patients with CAD. Methods A comprehensive literature search was conducted across the PubMed, Embase, and Cochrane Library databases up to June 3, 2020. The primary outcome was major adverse cardiac events (MACEs). The secondary outcomes were all-cause mortality, cardiovascular (CV) death, heart failure (HF), and myocardial infarction (MI). Pooled estimations and 95% confidence intervals (CIs) were assessed using a random-effects model. Results Twenty-two articles that enrolled a total of 17,432 patients with CAD were included in the final analysis. CAD patients in the highest categories of baseline sST2 had a significantly higher risk of MACEs (HR: 1.42, 95% CI: 1.09–1.76), all-cause mortality (HR: 2.00, 95% CI: 1.54–2.46), and CV death (HR: 1.42, 95% CI: 1.15–1.68), HF (HR: 2.41, 95% CI: 1.87–2.94), but not that of MI (HR: 1.15, 95% CI: -0.73–3.04), than those in the lowest categories. These results were consistent when baseline sST2 was presented as continuous values in one unit increments. Moreover, subgroup analysis showed that elevated baseline sST2 levels increased the long-term risk of MACEs in the acute coronary syndrome (ACS) population (HR: 1.74, 95% CI: 1.39–2.09) but only showed a trend toward higher risk of MACEs in the non-ACS population (HR: 1.09, 95% CI: 0.87–1.30). Conclusions The findings suggest that a higher concentration of baseline sST2 is associated with a higher risk of MACEs, all-cause mortality, CV death, and HF in patients with CAD. Elevated sST2 levels could significantly predict future MACEs in the ACS population but not in the non-ACS population.

including post hoc analyses of randomized clinical trials; 2) studies published in English; 3) studies in which sST2 values related to long-term (>9 months) clinical outcomes in patients with CAD were reported; and 4) studies in which hazard ratio (HR), relative risk (RR), or odds ratio (OR) estimates with the corresponding 95% confidence intervals (CI) for MACEs, allcause mortality, CV death, HF, or MI events were reported or could be calculated. The exclusion criteria were as follows: 1) studies in which sST2 considered only as an element of a prognostic score and 2) studies with a duplicated population.

Data extraction
The following information was extracted from each eligible study by two authors independently: surname name of the first author, publication year, diagnosis of the patients, sample size, age, number of males, cutoff point of sST2 levels, outcomes evaluated, follow-up duration, and variables adjusted in the multivariate model. When sST2 levels were presented as continuous variables, we standardized the group-level exposure estimates to single units, thereby allowing for the integration of the effects of different sST2 values in different studies.

Definitions of outcomes and quality assessments
The primary outcome in this study was MACEs, which was defined as a composite outcome of death and other fatal or non-fatal cardiovascular events. Given the heterogeneity of MACEs definitions in the included studies, the potential cardiovascular events were distinct, such as MI, HF, stroke, rehospitalization, and revascularization, were included. The secondary outcomes were all-cause mortality, CV death, HF, and MI. A dedicated tool [15,16] designed for prognostic study assessment was adopted to evaluate the quality of the included publications. This tool consists of five methodological domains to assess for bias: study participation, prognostic factor measurement, outcome measurement, description and measurement of confounding factors, and statistical analysis and reporting. After assessment, every methodological domain was labeled as good, adequate, or unclear for each study.

Statistical analysis
Statistical analyses were performed using STATA software (version 12.0; StatCorp., College Station, TX, USA), and P < 0.05 was considered statistically significant. Multivariate adjusted HRs/RRs/ORs with 95% CI for each study were extracted and analyzed. For studies using the sST2 level as the categorical variable, the values of the effects for comparing the highest and lowest categories of sST2 level were extracted. The inverse variance-weighted mean of the logarithm of HR/RRs/ORs and 95% CI with a random effect model was used to calculate the association between sST2 levels and different clinical outcomes among CAD patients.
The I 2 and Cochrane's Q tests were adopted to evaluate the heterogeneity among different studies [17,18]. Significant heterogeneity was considered if I 2 > 50% [19]. Meta-regression was performed to explore the study and population characteristics that may contribute to heterogeneity [20]. Publication bias was evaluated by funnel plots using the Egger regression asymmetry test [21].

Literature search and study characteristics
The literature search process is summarized in Fig 1. We identified 159 articles from PubMed, 1,585 from Embase, and 41 from the Cochrane Library. Among them, 22 articles that enrolled a total of 17,432 patients with CAD investigated the relationship between baseline sST2 levels and long-term clinical outcomes in patients with CAD were included [12,13,[22][23][24][25][26][27][28][29][30][31][32][33][34][35][36][37][38][39][40][41]. In these studies, the baseline sST2 level was provided as a categorical variable (18 articles) or a continuous variable (8 articles). The main characteristics of these studies are summarized in Tables 1  and 2. Regarding the types of CAD, 17 studies were conducted on the ACS population and 5 were conducted on the non-ACS population. MACEs as endpoints were reported in 13 articles, followed by all-cause mortality in 12, CV death in 5, HF in 4 articles, and MI in 2. The duration of follow-up in the included studies varied from 9 months to 13 years. Various confounding factors such as age, sex, medical history, comorbidities, biochemical parameters, and medication were adjusted when presenting the HRs/RRs/ORs in the included studies.

Source of heterogeneity and subgroup analysis
There was medium (I 2 = 37.8%) and substantial (I 2 = 75.5%) heterogeneity for the association of sST2 with the primary outcome when sST2 was used as a categorical variable and a continuous variable, respectively. We conducted meta-regression with geographic locations, number of participants, study design, or types of CAD to investigate the potential sources of heterogeneity. The results suggested that the type of CAD might be the main source of heterogeneity (P = 0.009 for CAD patients with sST2 as categorical variables; P = 0.574 for those with sST2 as continuous variables) (S1 Fig). The subgroup analysis stratified by type of CAD is summarized in

Quality assessment
An overview of the methodological quality of all included studies is presented in S2 Table. Nine studies [12,13,23,25,29,30,32,34,38,39] did not report well-defined details of the methods and results in the follow-up, which might have contributed to possible bias in their outcomes. Additionally, one [30] study had a high risk of population selection and another [38] did not provide definitions of the outcomes. All studies provided adequate descriptions of sST2 measurement, control of confounding factors, and statistical analysis.

Sensitivity analysis and publication bias
Sensitivity analysis showed that no single study had a significant effect on the pooled effect, indicating that our results were statistically robust. Based on the results of Egger's test, when    Sst2 was used as a categorization variable, there was no significant publication bias for the primary outcome (P = 0.059). However, a combination of visual assessment of funnel plots (S2 Fig) and Egger's test provided statistical evidence of publication bias for MACEs when sST2 was expressed as a continuous variable (P = 0.015). This may indicate that the positive results in CAD patients with sST2 as a continuous variable might be overrepresented, potentially leading to a relative overestimation of the pooled HR in the meta-analysis.

Discussion
In this study, elevated baseline sST2 levels in patients with CAD were positively associated with increased long-term risk of MACEs, all-cause mortality, CV death, and HF, but not of MI events. Moreover, subgroup analysis showed that the predictive role of sST2 for long-term prognosis in the ACS population was consistent with the main results, while the prognostic value of sST2 was not verified in the non-ACS population. This study is the largest meta-analysis to assess the relationship between sST2 and long-term prognosis in patients with CAD, which allows a much greater possibility of reaching reliable conclusions. sST2 is a novel biomarker in inflammatory conditions and cardiovascular disease [6]. Over the past decade, several meta-analyses have suggested that sST2 is a prognostic indicator of aortic valve replacement [42], acute HF [9], chronic HF [10], and pulmonary hypertension [43]. Gu et al. also demonstrated that a higher circulating baseline sST2 level could predict poor clinical outcomes in patients with ACS in a meta-analysis that included 12 studies [44]. Our updated results with 17 articles involving patients with ACS are consistent with the findings of the above study, which confirmed the positive association between sST2 and increased risk of adverse events in the long-term prognosis of patients with ACS. Nonetheless, the study conducted by Gu et al. may have classified CV death events as all-cause mortality in some studies, which might bias their estimations [23,26,29]. Moreover, our study adds new evidence for the prognostic value of sST2 in the non-ACS population. The elevated baseline sST2 levels showed a trend toward an increased long-term risk of MACEs, all-cause mortality, and CV death in these patients, although without statistical significance.
Heterogeneity is common in meta-analyses. Therefore, it is an essential component of a meta-analysis to explore the sources of heterogeneity. In this study, we used meta-regression to explore the sources of heterogeneity. When sST2 was expressed as a categorical variable, the results showed that the type of CAD (ACS or non-ACS) might be the main source of heterogeneity for the primary outcome. In the subsequent subgroup analysis, the bias risks were similar in the ACS and non-ACS groups. Additionally, significant publication bias was detected in studies using sST2 as a continuous variable. This result might be ascribed to the fact that relatively few studies adopted continuous sST2 values and thus did not report potentially negative results. The biological mechanism underlying the relationship between sST2 and clinical outcomes in patients with CAD is not fully understood. One plausible pathway is that sST2 leads to adverse clinical outcomes by counteracting the protective effect of IL-33. ST2 is expressed by cardiac myocytes in two forms: the soluble subtype (sST2) and the membrane-bound subtype (ST2L) [45]. Its ligand, IL-33, which is released by cardiac cells, normally protects the myocardium under pressure overload [46]. sST2 neutralizes this protective mechanism by acting as a "decoy receptor" for IL-33. When bound with sST2, IL-33 can no longer interact with membrane-ST2L to prevent the activation of downstream signaling pathways, leading to cardiomyocyte hypertrophy and remodeling. The indirect resistance to IL-33 might activate adverse cardiac hypertrophy and remodeling through lymphocyte function [47]. Another possible mechanism linking the ST2/IL-33 axis to CAD outcomes is its correlation with atherosclerosis progression and plaque destabilization [48]. It has been speculated that IL-33 reduces the risk of plaque rupture by inhibiting the immune response involved in atherosclerosis, an effect that can be reversed by sST2. Additionally, the IL-33/ST2 system might be a potential pathophysiological mediator in the inflammatory and fibrotic response to tissue injury and has been identified in inflammatory conditions and fibroproliferative diseases [11]. Accordingly, elevated serum sST2 levels in patients with CAD suggest the presence of active inflammation and fibrosis, factors that have been associated with poor prognosis [49,50]. More importantly, the differential predictive value of sST2 on the occurrence of HF and MI in patients with CAD further suggests that sST2 is predominantly involved in the pathophysiology of HF and independent of CAD progression per se.
Our study has important strengths. First, multivariate-adjusted results were available from all included studies, reducing the likelihood of confounders. Second, all results were derived from cohorts or randomized controlled studies, which would enable us to detect the temporal relationship between sST2 and prognosis in patients with CAD. Nevertheless, several limitations of this study should be acknowledged. First, the heterogeneous definitions of MACEs in different studies might have contributed to potential bias. Second, some degree of heterogeneity was observed, which could be due to different types of CAD. Third, the use of different cutoff values for sST2 may lead to different estimated effects. However, due to the limited information available, we could not address the optimum cutoff value of sST2 for the prognosis of patients with CAD. Further studies with larger sample sizes are warranted to address this issue.

Conclusions
This meta-analysis suggests that elevated baseline sST2 level is associated with a higher risk of MACEs, all-cause mortality, CV death, and HF in patients with CAD. Elevated sST2 levels could significantly predict future MACEs in the ACS population but not in the non-ACS population.
Supporting information S1 Table. Keywords and search strategy in the umbrella review.