Prognostic Value of SYNTAX Score II in Patients with Acute Coronary Syndromes Referred for Invasive Management: A Subanalysis from the SPUM and COMFORTABLE AMI Cohorts

Aims To assess the incremental prognostic value of SYNTAX score II (SxSII) as compared to anatomical SYNTAX Score (SxS) and GRACE risk score in patients with acute coronary syndromes who underwent percutaneous coronary intervention. Methods and results SxSII and SxS were determined in 734 ACS patients. Patients were enrolled in the prospective Special Program University Medicine ACS and the COMFORTABLE AMI cohorts and later on stratified according to tertiles of SxSII (SxSIILow ≤21.5 (n=245), SxSIIMid 21.5–30.6 (n=245), and SxSIIHigh ≥30.6 (n=244). The primary endpoint of adjudicated all-cause mortality and secondary endpoints of MACE (cardiac death, repeat revascularization, and myocardial infarction) and MACCE (all-cause mortality, cerebrovascular events, MI, and repeat revascularization) were determined at 1-year follow-up. SxSII provided incremental predictive information for risk stratification when compared to SxS and GRACE risk score (AUC 0.804, 95% CI 0.77–0.84, p < 0.001 versus 0.67, 95% CI 0.63–0.72, p=0.007 versus 0.69, 95% CI 0.6–0.8, p=0.002), respectively. In a multivariable Cox regression analysis, we found that unlike SxS (adjusted HR 1.013, 95% CI (0.96–1.07), p=0.654), SxSII was significantly associated with all-cause mortality (HR = 1.095, 95% CI (1.06–1.11), p < 0.001). This was also true for the prediction of both secondary outcomes MACE (n=60) and MACCE (n=70) with an adjusted HR = 1.055, 95% CI (1.03–1.08), p < 0.001, and HR = 1.065, 95% CI (1.04–1.09), p < 0.001. Conclusion In patients with ACS who underwent PCI, SxSII is an independent predictor of mortality during 1-year follow-up. SxSII shows superiority in discriminating risk compared to conventional SxS and GRACE for all-cause mortality.


Introduction
e anatomical SYNTAX (Synergy between percutaneous coronary intervention with taxus and cardiac surgery) score (SxS) is an angiographic scoring system for assessing the complexity of coronary artery disease (CAD) [1] advocated for decision making in the latest ESC/EACTS guidelines on myocardial revascularization [2]. Originally, the SxS was introduced to predict clinical outcomes in stable patients with 3-vessel and/or left main disease undergoing percutaneous coronary intervention (PCI) or coronary artery bypass grafting (CABG), respectively, based on data from the SYNTAX trial [3,4]. Later on, the SxS was applied to a variety of patient populations with diverse clinical presentations including those with acute coronary syndromes (ACS) undergoing primary PCI [5,6].
However, subanalyses of the SYNTAX trial and results from different studies have implied that the purely anatomybased risk stratification of the SxS score made it prone to misclassification of patient's true risk, particularly for allcause mortality and cardiac death in patients with stable CAD or ACS treated by PCI [7,8]. e addition of clinical variables was a promising step in improving risk stratification by reclassifying patients into more accurate risk categories. erefore, in order to account for the variability of clinical parameters affecting long-term outcomes and hence better classification of patients' risk, the SYNTAX score II (SxSII) was developed by complementing SxS with 7 prognostic variables including age, creatinine clearance, left ventricular ejection fraction (LVEF), presence of unprotected left main coronary artery (ULMCA) disease, peripheral vascular disease (PVD), female gender, and chronic obstructive pulmonary disease (COPD) [9]. So far, the SxSII was validated in patients with left main and multivessel disease showing more accurate patient stratification than SxS [10,11,12].
e aim of the present study was to assess the predictive performance of SxSII in patients presenting with ACS undergoing PCI and to compare it to the previously validated SxS and the commonly used Global Registry of Acute Coronary Events (GRACE) risk score [13].

Study Population.
e prospective multicenter Special Program University Medicine-Acute Coronary Syndromes and Inflammation (SPUM-ACS, ClinicalTrials.gov number, NCT01000701) enrolled consecutive patients who were referred for coronary angiography with a diagnosis of ACS to one of the participating Swiss university hospitals (Zurich, Bern, Lausanne, and Geneva) between December 2009 and October 2012 [14,15]. Inclusion criteria comprised patients of both genders, aged ≥18 years, presenting within 5 days (preferably within 72 hours) after pain onset with a main diagnosis of STEMI, NSTEMI, or unstable angina. Enrolled patients had symptoms compatible with angina pectoris (chest pain and dyspnea) and fulfilled at least one of the following criteria: (a) ECG changes, such as persistent STsegment elevation or depression, T-inversion or dynamic ECG changes, or new left bundle branch block (LBBB); (b) evidence of positive (predominantly conventional) troponin by local laboratory reference values (with a rise and/or fall in serial troponin levels); and (c) known coronary artery disease specified by its status after MI, coronary artery bypass graft (CABG), or PCI or newly documented ≥50% stenosis of an epicardial coronary artery during the initial catheterization. Exclusion criteria for the SxSII study comprised prior CABG, referral to either CABG or medical management after completion of the coronary angiogram, severe physical disability, and inability to comprehend study or less than 1 year of life expectancy for noncardiac reasons. Within this consortium, a centralized electronic database was implemented providing comprehensive information on all patients comprising both clinical and coronary anatomy parameters. A telephone follow-up was performed at 30 days and at 1 year, a clinical visit. Adverse events occurring within 365 days after the index ACS event were adjudicated by an independent adjudication committee consisting of 3 experienced cardiologists (Lukas Kappenberger, MD, Lausanne; Tiziano Moccetti, MD Lugano; and Mathias E. Pfisterer, MD, Basel). An additional 3-year follow-up to assess all-cause mortality was ascertained by telephone for the SxSII study, and only patients recruited in Bern and Zurich were analyzed in the SxSII study. e study was approved by the local ethical committees, and all patients gave written informed consent in compliance with the Declaration of Helsinki. e COMFORTABLE AMI trial included patients aged 18 years or older who had a history of chest pain of more than a 10 min duration and associated ST-segment elevation of >1 mm in ≥2 contiguous leads, new left bundle branch block, or true posterior MI, who underwent primary percutaneous coronary intervention (PCI) within 24 h of symptom onset. In addition, there was angiographic presence of at least one acute infarct-related artery (IRA) with one or multiple coronary artery lesions in a native coronary artery with a diameter between 2.25 and 4.0 mm, which could be treated with one or multiple stents. Exclusion criteria included the use of vitamin K antagonists, mechanical complications of myocardial infarction, acute myocardial infarction secondary to stent thrombosis (ST), planned surgery within 6 months of PCI unless dual antiplatelet therapy could be maintained throughout the perisurgical period, and noncardiac comorbid conditions with life expectancy <1 year. Further study details are described in detail elsewhere [16].
Angiography was digitally recorded and analyzed in a central core laboratory. e MI SxS score was assessed by experienced analysts using the web-based program http: //www.syntaxscore.com as previously described. Angiographic documentation of patients included in the COMFORTABLE AMI trial was scored as described previously.

Clinical
Outcomes. All-cause mortality included cardiac, vascular, and noncardiovascular causes of death. Cerebrovascular events comprised stroke (any, ischemic, hemorrhagic, and unclear etiology) or transient ischemic attack (TIA); repeat revascularization included any repeat coronary revascularization (target and nontarget vessel). Clinically indicated repeat revascularization included any clinically driven repeat coronary revascularization (target and nontarget vessel) [17]. Myocardial infarction was defined based on the universal definition including periprocedural MI in patients with UA [18]. e primary endpoint of our study was adjudicated all-cause mortality at 1-year follow-up. e secondary endpoints were adjudicated major adverse cardiovascular events (MACE) defined as the composite of cardiac death, clinically indicated revascularization, or MI at 1 year and adjudicated major adverse cardiovascular and cerebrovascular events (MACCE) defined as the composite of all-cause mortality, cerebrovascular events, any repeat revascularization, or myocardial infarction (MI) at 1-year, respectively.
2.3. Anatomical SYNTAX Score, SYNTAX Score II, and GRACE Risk Score Calculation. Experienced cardiologists blinded to clinical outcomes assessed the SYNTAX score for each angiogram. e interobserver and intraobserver variabilities of the SxS scoring team were previously reported as moderate (kappa statistic 0.56) and substantial (kappa statistic 0.70, respectively). e intraclass correlation coefficient for calculated SXS in SPUM cohort for absolute agreement was 0.886 (p < 0.001), 95% CI (0.835, 0.919). Before accessing any lesions, all those with ≥50% diameter stenosis in vessels ≥1.5 mm in diameter were scored using the SxS algorithm [1]. e SxSII was then calculated using the PCI SYNTAX score II (http://www.syntaxscore.com) calculator based on the previously published nomogram [9], with scores assigned for the presence and magnitude of each predictor specific for PCI population [9,11]. e GRACE risk score to calculate long-term mortality comprised age, heart rate, systolic blood pressure, initial serum creatinine, history of congestive heart failure, history of myocardial infarction, elevated cardiac markers (conventional troponins as per local laboratories), ST-segment depression, and no in-hospital PCI [13]. e long-term GRACE risk score was calculated using a program written in STATA, and we used the standard scoring of GRACE as mentioned in the reference publications. (http://www. wikidoc.org/index.php/ e_GRACE_risk_score). It was retrospectively calculated for all patients included in our study and was compared to SxSII and SxS as a continuous variable by ROC curve analysis and multivariable Cox regression model.

Statistical Analysis.
Continuous variables are expressed as mean ± SD or medians with interquartile ranges (IQR) and were compared using one-way ANOVA, Student's ttest, Kruskal-Wallis test, or Mann-Whitney test as appropriate. Categorical data are presented as frequency (percentages) and were compared using the Fisher exact or the chi-square test. Cumulative incidences were calculated using Kaplan-Meier curves.
For our analysis, we stratified patients according to tertiles of SxSII [9] (≤21.5, 21.5-30.6, ≥30.6) and SXS (≤12, 12-22, ≥22). e score ranges are referred to as SxSII Low , SxSII Mid , and SxSII High and SxS Low , SxS Mid , and SxS High , respectively. We constructed multivariable Cox proportional hazard models including variables that had a significant association with a p value of <0.05 in univariable analysis. We further added SxS and SxSII separately into the model due to their collinearity. Other variables showed no multicollinearity that exceeding the acceptable threshold of VIF ≥3 or tolerance ≤0.2. From the full model, we selected variables to minimize Akaike's information criterion by backward stepwise methods. Calibration was then determined by the Hosmer-Lemeshow goodness-of-fit test.
To visualize the effect of SxSII on clinical outcomes and confirm its linearity, we constructed an alternative model using penalized splines. Analyses were performed with SPSS version 21.0 software (SPSS Inc., Chicago, Ill).
ROC curves were constructed to assess the ability of the SxSII, SxS, and GRACE risk score to predict events at 1year follow-up. Patients who were lost to follow-up at 1 year were excluded from the analysis. Areas under curves were compared using the DeLong method [19] provided by MedCalc for Windows, version 14.10.2 (MedCalc Software, Acacialaan, Belgium). Category-free net reclassification improvement (NRI) and integrated discrimination improvement (IDI) were both calculated using the "survI-DINRI," R package (R-version 3.3.2), through comparing proportional hazards models, whereas category-based NRI was done using MATLAB version R2015b, all as described by Pencina et al. [20]. A probability value of <0.05 was considered significant, and all tests were two-tailed.

Baseline and Angiographic Characteristics for Tertiles of SYNTAX Score II.
SxS and SxSII could be calculated for all 734 patients with a complete one-year follow-up. e mean anatomical SYNTAX score was 17.56 ± 9.3 with a median of 16 with an interquartile range (IQR) of 13. Patients were categorized into SxS tertiles (SxS Low , n � 251, SxS Mid , n � 244, and SxS High , n � 239). e mean SYNTAX score II was 27.7 ± 10.3 with a median of 25.4. e number of patients stratified according to tertiles of SxSII Low , SxSII Mid , and SxSII High , was 245, 245, and 244, respectively. e tertiles of SxSII and its individual components including the anatomical SYNTAX score are listed in Table 1 as they are compared to tertiles of the conventional SxS.
Compared with patients in the lower tertiles, patients in SxSII High had a higher rate of adverse cardiovascular history and risk factors and fasting glucose, along with hemodynamic instability on admission. e baseline characteristics and risk factors of patients according to SxSII tertiles are listed in Table 2. Angiographic characteristics showed significant differences with a higher rate of MVD with left anterior descending artery (LAD) involvement (74.2%) in the SxSII High tertile compared with patients in the lower tertiles (Table 3).

Clinical Outcomes Stratified by Tertiles of Anatomical
SYNTAX and SYNTAX Score II. In order for better assessment of the impact SxSII has on mortality, we expressed this relationship in terms of penalized splines curves (hazard ratio-based curves) obtained through a Cox proportional hazard regression model (Supplementary Figure 1). As the relationship was curve-linear, our choice of breaking the score into tertiles was justified.

Cardiology Research and Practice
Patients within the SxSII High tertile had a significantly higher incidence of all-cause mortality, MACE, and MACCE compared with patients in lower tertiles (Table 4). ere was also a higher rate of clinically driven revascularization, in SxSII High (8.6% p � 0.002) with a trend for excess cerebrovascular events, p � 0.134. One-year outcomes across the tertiles of SxS are reported in Supplementary Table 1.
Furthermore, Kaplan-Meier curves were plotted to asses all studied outcomes across tertiles of both SxS and SxSII as shown in Figure 1. All-cause mortality (9.4% versus 1.2% versus 0.8%), MACCE (17.6% versus 8.6% versus 2.4%), and MACE (14.3% versus 8.2% versus 2%) occurred at a significantly higher rate among patients in SxSII High compared to SxSII Mid and SxSII Low , respectively (p (log rank) <0.001). Conversely, the anatomical SYNTAX score does not provide consistent risk stratification, specifically when addressing the primary endpoint of all-cause mortality at 1 year.   Depicted are counts, n incidence (%) or mean ± SD. * variables are included in the score; p value reported as N/A. LAD, left anterior descending artery; RCA, right coronary artery; LCX, circumflex artery; TIMI, thrombolysis in myocardial infarction; AHA, American Heart Association.  (Figure 2(a)). is improvement was statistically significant at an AUC difference of 0.145 (95% CI 0.049-0.246, p � 0.0045). Compared with the GRACE risk score (calculated in 500 patients), SxSII showed a persistently higher prognostic accuracy for all-cause mortality (Figure 2 Figure 2). Additionally, we sought to authenticate the benefit in risk assessment, that we have seen so far by SxSII over SXS, through performing a category-free and categorybased net reclassification improvement (NRI) along with integrated discrimination improvement (IDI) for the outcome of all-cause mortality at 1 year. We found a significant categorical net reclassification improvement of 0.344 (p � 0.004) with Z score of 2.714 for all-cause mortality, primarily driven by a significant net gain of 0.321 (p � 0.006) in patients who had the event and a trend (p � 0.39) in patients without the event (Table 5). More importantly, both estimates of category-free net reclassification improvement (NRI) and integrated discrimination improvement (IDI) were significant (p < 0.001) at 0.380 95% CI (0.191-0.554) and 0.08 95% CI (0.035-0.278), respectively. e red shaded area and its extent show a substantial added value of the SxSII model over SxS (Figure 3).

Results from Multivariable Analysis.
For model construction, variables which were found to correlate significantly with the studied outcomes were plotted (Supplementary Figures 3 and 4) as part of hazard assessment in a univariable Cox regression analysis. e anatomical SYNTAX score, age, LVEF, and GFR were predictive of the primary outcome; however, they correlated significantly with SxSII while at the same time showing multicollinearity exceeding the acceptable threshold of VIF<3. Hence, two separate models for SxSII and SxS were constructed, showing the SxSII to be a significant predictor of all-cause mortality (adjusted HR 1.095 95% CI (1.05-1.13), p < 0.001) along with the presence of left main disease, fasting glucose, and resuscitation status ( Table 6). On the other hand, the anatomical SYNTAX score got short of significance for the prediction of our primary outcome (adjusted HR 1.013 95% CI (0.956-1.073), p � 0.656). e complete list of predictors in multivariable analysis for the outcomes of MACE and MACCE is shown in Supplementary Table 2.
Additional analysis regarding the impact of categorizing patients within tertiles of SxSII showed that a classification into SxSII High independently predicted all-cause mortality (

Discussion
To the best of our knowledge, this is the first study to evaluate and compare the novel SxSII score in patients with  ACS undergoing PCI with the currently available SxS and GRACE risk scores. e following key findings were obtained: (i) SxSII independently predicts all-cause mortality, MACE, and MACCE during 1-year follow-up. (ii) SxSII provides superior discrimination of risk for allcause mortality and MACCE than the conventional SxS and GRACE score, respectively.
In the current era of interventional cardiology, the role of the multidisciplinary HeartTeam in choosing the optimal available means of revascularization is emphasized by current ESC guidelines [2] advocating the use of the anatomical SYNTAX score as a fundamental tool to assist in the decision making for surgical versus percutaneous coronary revascularization in stable coronary artery disease patients with left main or multivessel disease [2]. However, risk assessment and the prediction of long-term outcomes in patients presenting with ACS undergoing PCI is still suboptimal with a wide range of old and newly emerging risk scores [10].
e SxS has been extensively studied for a variety of clinical outcomes in different patient populations including all-comers [21,22] as well as patients with NSTEMI [23,24] or STEMI [25,26]. e recently developed SxSII has been complemented with clinically significant prognostic variables, known to be independent predictors of mortality at 4 years in patients with stable CAD enrolled in the SYNTAX trial [9,25,27]. In that patient population, this was translated into better discrimination of risk for long-term mortality for SxSII when compared to SxS [9]. e main findings from our study support the superiority of SxSII also in patients with ACS undergoing PCI who are at particular risk as it was found to be an independent predictor of allcause mortality during 1-year follow-up. In contrast, the anatomical SYNTAX score was not an independent predictor for this endpoint, in line with previous studies [22,28,29]. e discrepant findings between studies evaluating the anatomical SYNTAX score to predict adverse clinical outcome (particularly all-cause mortality) are likely attributable to the heterogeneous patient population analyzed (inclusion or exclusion of patients with STEMI and/or cardiogenic shock). Indeed, the latter patients tend to have a worse outcome which is not predicted by anatomical complexity alone. Furthermore, different cutoffs used in statistical models to calculate SxS may explain the observed differences.
To further substantiate the significant improvement in discrimination or risk for all-cause mortality identified for SxSII, we compared this score to the clinically based, ESCadvocated risk stratification in patients with NSTEMI [2], the GRACE risk score [13]. Despite the fact that patients with cardiogenic shock with high GRACE risk scores were included in our cohort, the SxSII showed greater discrimination of risk for all-cause mortality during 1-year follow- up.
e GRACE risk score lacks some clinical variables which are important predictors of mortality, such as LVEF, as well as classifiers of complex coronary anatomy [30][31][32][33], which may explain the reduced prognostic accuracy to predict all-cause mortality when compared with the SxSII.
In our study, SxSII also showed good prognostic accuracy for the secondary endpoint of MACCE. Conversely, SxSII provided no incremental risk stratification for MACE compared with SxS. is could be explained by the fact that MACE, which were primarily attributable to MI and clinically indicated revascularization in a population with complex coronary anatomy (i.e., a mean SxS of 17.56 ± 9.3, 44% MVD and 63.5% LAD involvement), could be well predicted by pure anatomical complexity [34][35][36] as assessed by SxS. Indeed, in our multivariable analysis, multivessel disease was a predictor of MACE and MACCE when assessed with both models of SxSII. However, it fell short of significance for all-cause mortality along with the anatomical which is concordant with previously published studies [6].
Importantly, SxSII enabled reclassification of risk for allcause mortality compared with SxS.
is was highly significant for 1-year follow-up with a persistent net gain of approximately 33%. Upon exploring the differences in complexity of the coronary anatomy between tertiles of SxSII and SxS, patients would still classify in the middle category of the original anatomical SxS, although the anatomical SYNTAX score in the SxSII High group was higher as compared to that in SxSII Mid and SxSII Low . erefore, the use of SxSII for risk stratification in patients with ACS undergoing PCI identifies a distinctive group of patients who despite having moderately complex coronary artery disease are still at a significantly higher risk of both cardiovascular and noncardiovascular death during follow-up.
ese are patients with multiple risk factors and comorbidities who may benefit from intensive secondary prevention, particularly from some of the most recent lipid-lowering drugs [14,[37][38][39] and risk factor modification with closer follow-up intervals and may derive an advantage of full revascularization when multivessel disease is present.
Surprisingly enough, diabetes was not a predictor of clinical outcomes, which could be explained by the fact that end-organ damage secondary to diabetes-as reflected by scores reflecting coronary anatomy and calcification-rather than metabolic parameters better reflect risk in this patient population [9,13,40]. Interestingly, abnormal fasting glucose levels during hospitalization for the index event were predictive of all-cause mortality irrespective of the diabetes  Figure 3: Category-free IDI and NRI. e red shaded area shows a significant integrated discrimination improvement (IDI), whereas the difference between the two vertical dots represents the improvement in net reclassification by SXSII over SxS.  [41][42][43][44]. e pathophysiological mechanisms are well described elsewhere [45][46][47], including but not limited to the effects on the collateral circulation, infarct size, reperfusion, sympathetic activation with elevated catecholamine levels leading to hepatic glucose release and platelet aggregation.

Study Limitations.
One limitation of our study is the fact that due to the low number of patients undergoing CABG or medical therapy in this ACS population, risk by SxSII in the primary PCI group could not be compared to these different treatment modalities. Furthermore, since the SxSII was developed using prognostic variables in a population where STEMI patients were excluded, there may be additional parameters that could further improve risk stratification of adverse clinical outcome.

Conclusion
In the present study, we demonstrate a clinically relevant superiority of the novel SYNTAX score II when compared to the anatomical SYNTAX and commonly used GRACE risk score, in risk stratification of patients with ACS undergoing PCI. Pending validation in other cohorts, our data suggest that the use of the SYNTAX score II opens a new door for improvement in decision making and management of patients with ACS.

ACS:
Acute coronary syndromes AUC: Area under the curve CABG: Coronary artery bypass graft CI: Confidence intervals ECG: Electrocardiogram ESC: European Society of Cardiology GRACE: Global Registry of Acute Coronary Events LBBB: Left bundle branch block MI: Myocardial infarction NSTEMI: Non-ST-elevation myocardial infarction PCI: Percutaneous coronary intervention ROC: Receiver-operating characteristics.

Data Availability
e data used to support the findings of this study are available from the corresponding author upon request.

Disclosure
CTU Bern, which is part of the University of Bern, has a staff policy of not accepting honoraria or consultancy fees. However, CTU Bern is involved in design, conduct, or analysis of clinical studies funded by several companies and nonprofit organizations. Slayman Obeid, Antonio H. Frangieh, and Lorenz Räber contributed equally.

Conflicts of Interest
LR received speaker fees and research grants to the institution from St. Jude Medical. SW has received research grants to the institution from Abbott, Boston Scientific, Biosensors, Biotronik, the Medicines Company, Medtronic, and St. Jude Medical and honoraria from Abbott, Astra Zeneca, Eli Lilly, Boston Scientific, Biosensors, Biotronik, Medtronic, and Edwards. CMM received research grants to the institution from Eli Lilly, AstraZeneca, Roche, and MSD including speaker or consultant fees. RK received speaker fees from Eli Lilly, Servier, and Bayer Healthcare. TFL received research grants to the institution from AstraZeneca, Bayer Healthcare, Biosensors, Biotronik, Boston Scientific, Eli Lilly, Medtronic, MSD, Merck, Roche, and Servier, including speaker fees by some of them. All other authors have no conflict of interest to declare.