Incidence and multivariable correlates of long-term mortality in patients treated with surgical or percutaneous revascularization in the Synergy between Percutaneous Coronary Intervention with Taxus and Cardiac Surgery (SYNTAX) trial

Abstract

Aims

The aim of this investigation was to determine the incidence and multivariable correlates of long-term (4-year) mortality in patients treated with surgical or percutaneous revascularization in the synergy between percutaneous coronary intervention (PCI) with TAXUS Express and Cardiac Surgery (SYNTAX) trial.

Methods and results

A total of 1800 patients were randomized to undergo coronary artery bypass graft (CABG) surgery (n = 897) or PCI (n = 903). Prospectively collected baseline and peri- and post-procedural data were used to determine independent correlates of 4-year all-cause death in the CABG and the PCI arms (Cox proportional hazards model). Four-year mortality rates in the CABG and the PCI arms were 9.0% [74 deaths (12 in-hospital)] and 11.8% [104 deaths (16 in-hospital)], respectively (log-rank P-value = 0.063). Censored data comprised 78 patients (8.7%) in the CABG arm, and 24 patients (2.7%) in the PCI arm (log-rank P-value < 0.001). Within the CABG arm, the strongest independent correlates of 4-year mortality were lack of discharge aspirin [hazard ratio (HR) 3.56; 95% CI: 2.04, 6.21; P < 0.001], peripheral vascular disease (PVD) (HR: 2.65; 95% CI: 1.49, 4.72; P = 0.001), chronic obstructive pulmonary disease, age, and serum creatinine. Within the PCI arm, the strongest independent correlate of 4-year mortality was lack of post-procedural anti-platelet therapy (HR: 152.16; 95% CI: 53.57, 432.22; P < 0.001), with 10 reported early (within 45 days) in-hospital deaths secondary to multifactorial causes precluding administration of anti-platelet therapy. Other independent correlates of mortality in the PCI arm included amiodarone therapy on discharge, pre-procedural poor left ventricular ejection fraction, a ‘history of gastrointestinal bleeding or peptic ulcer disease’, PVD (HR: 2.13; 95% CI: 1.26, 3.60; P = 0.005), age, female gender (HR: 1.60; 95% CI: 1.01, 2.56; P = 0.048), and the SYNTAX score (Per increase in 10 points: HR: 1.25; 95% CI: 1.06, 1.47; P = 0.007).

Conclusion

Independent correlates of 4-year mortality in the SYNTAX trial were multifactorial. Lack of discharge aspirin and lack of post-procedural anti-platelet therapy were the strongest independent correlates of mortality in the CABG and the PCI arms, respectively. Peripheral vascular disease is a common independent correlate of 4-year mortality and may be a marker of the severity of baseline coronary disease and risk of future native coronary disease (and extra-cardiac disease) progression.

Introduction

The SYNTAX trial has established a framework in which patients with unprotected left main or 3-vessel coronary artery disease are managed by the consensus of the Heart Team, in which the cardiac surgeon and interventional cardiologist determine the optimal revascularization modality for patients.^1–6 The results from the primary endpoint of the SYNTAX trial at 4 years have recently been reported,^7,8 with follow-up through to 5 years planned. The aim of this study was to analyse the incidence and multivariable correlates of all-cause death at 4 years following surgical or percutaneous revascularization in the randomized SYNTAX trial.

Methods

The SYNTAX trial is a randomized, prospective, multicentre trial that incorporates an ‘All-Comers’ design and in this study, consisted of pre-specified left main [isolated or associated with 1, 2, or 3 vessel disease (3VD)] and 3VD arms.^1,3,5 Patients were randomized on a 1:1 basis by the Heart Team consensus to undergo either coronary artery bypass graft (CABG) or percutaneous coronary intervention (PCI) with TAXUS Express paclitaxel-eluting stents (Boston Scientific Corporation, Natick, MA, USA) (n = 1800). Patients considered unsuitable for randomization were entered in nested registries for CABG- or PCI-ineligible patients. Exclusions were limited only to patients with prior coronary revascularization, the requirement of concomitant cardiac surgery, or on-going acute MI. An independent core laboratory (Cardialysis BV, Rotterdam, the Netherlands), blinded to the treatment assignment, calculated the SYNTAX score (http://www.syntaxscore.com).^3,4,6,9

Pre-procedurally, the study protocol mandated in the CABG arm aspirin (>70 mg per day) at least 12 h pre-CABG, and in the PCI arm aspirin (>70 mg per day) at least 12h pre-PCI and clopidogrel (300 + mg 24h pre-PCI) or ticlopidine (2 × 250 mg daily 48h pre-PCI). Post-procedurally, aspirin (>70 mg/day) was mandated in the CABG and the PCI arms, and clopidogrel (75 mg) or ticlopidine (2 × 250 mg) for at least 6 months in the PCI arm. Within the CABG and the PCI arms, it was recommended that optimal medical treatment be continued long-term according to the ACC/AHA guidelines.^10–12

Baseline and peri- and post-procedural data and blood samples were prospectively collected by the individual participating centres. Post-procedural and discharge medications encompassed specific medications (or lack of) immediately after revascularization (CABG or PCI) and at hospital discharge, respectively. Pre- and post-revascularization per protocol blood sampling with creatine kinase (CK) were undertaken, with the cardiac-specific MB iso-enzyme (CK-MB) measured only if the CK ratio ≥2× upper limit of normal (ULN).⁵ Revascularization with CABG or PCI was permitted only if the cardiac enzymes were <2× ULN. All the samples were evaluated by an independent central chemistry laboratory (Covance, Inc., Geneva, Switzerland, and Indianapolis, IN, USA).

Statistical analysis

In this article, continuous variables are presented as means ± standard deviation (SD). Categorical variables are presented as counts or percentages. Comparisons of 4-year clinical all-cause death (Kaplan–Meier estimates) were performed with the log-rank test. Similar analyses were conducted for censored patients using the time to censored event for all-cause death. Relationships of 4-year all-cause to covariates, utilizing previously published baseline and peri- and post-procedural characteristics,³ were investigated with univariate Cox regression models. The correlates of worse prognosis identified in univariable analyses were then introduced into a multivariable model using the forced enter method, with a variable entry criterion of 0.05; all the variables were retained in the final model irrespective of statistical significance. The proportionality of hazards (PH) assumption was checked using the global PH test based on Schoenfeld residuals.¹³ There was no departure from the PH assumption in the CABG [χ² = 10.29, degrees of freedom (df) = 10, P = 0.42] and the PCI arms (χ² = 13.96, df = 10, P = 0.17). Intercorrelation between variables was undertaken with the Pearson correlation coefficient where appropriate. A two-sided P-value of <0.05 was considered significant for all tests. All analyses were conducted using SPSS 19.0 (SPSS, Inc., Chicago, IL, USA) and STATA 11.0 (Stata Corp., College Station, TX, USA).

Results

Within the randomized SYNTAX trial (n = 1800), baseline demographics, and clinical characteristics for the CABG (n = 897) and the PCI (n = 903) arms were well balanced and have previously been described.³ At 4 years, clinical data were available in 819 of 897 patients in the CABG arm, and 879 of 903 patients in the PCI arm. Censored data comprised 78 patients (8.7%) in the CABG arm, and 24 patients (2.7%) in the PCI arm (log-rank P-value < 0.001). The most common reason for censoring was the withdrawal of consent after randomization [61 patients (CABG n = 50, PCI n = 11)], with the remainder of patients primarily lost to follow-up.

Four-year mortality was 9.0% (74 all-cause deaths) in the CABG arm and 11.8% (104 all-cause Deaths) in the PCI arm (log-rank P-value = 0.063). Univariate and independent correlates of cumulative 4-year mortality in the CABG and the PCI arms are shown in Tables 1–4.

Randomized coronary artery bypass graft arm

Within the randomized CABG arm, incomplete revascularization and on-pump CABG were not shown to be univariate correlates of 4-year mortality (Table 1). Lack of pre-procedural aspirin was not an independent correlate of mortality (Table 2). Lack of discharge aspirin was the strongest independent correlate of 4-year mortality [hazard ratio (HR) 3.56; 95% CI: 2.04, 6.21; P < 0.001]. Only a weak correlation was apparent between lack of pre-procedural aspirin and lack of discharge aspirin (Pearson correlation 0.14, P < 0.001). Twenty-seven out of 96 CABG-treated patients (28.1%) not discharged on aspirin died within 4 years (0 in-hospital deaths). Notably, 15 of the 27 deaths were cardiac (n = 13) or vascular (n = 2) related and 12 deaths non-cardiac related. A ‘history of gastrointestinal (GI) bleeding or peptic ulcer disease’ demonstrated a trend towards significance of being an independent correlate of 4-year mortality (P = 0.085). A ‘history of GI bleeding or peptic ulcer disease’ had no correlation with a lack of discharge aspirin therapy (Pearson correlation 0.03, P = 0.36).

Pre-procedural poor left ventricular ejection fraction (LVEF) was not an independent correlate of 4-year mortality. Amiodarone therapy on discharge demonstrated a trend towards being a significant independent correlate of 4-year mortality (P = 0.070). No correlation of pre-procedural poor LVEF and amiodarone therapy on discharge was evident (Pearson correlation 0.01, P = 0.76).

Other key independent correlates of 4-year mortality in the randomized CABG arm included peripheral vascular disease (PVD) (HR: 2.65; 95% CI: 1.49, 4.72; P = 0.001)—with 21 of 87 patients (24.1%) dying within 4-years (five in-hospital deaths), chronic obstructive pulmonary disease (HR: 2.44; 95% CI: 1.30, 4.60; P = 0.006), age, and renal function (Table 2).

Randomized percutaneous coronary intervention arm

Within the randomized PCI arm, total number of stents, total stent length, and stent length >100 mm were not shown to be correlates of 4-year mortality (Table 3). The strongest independent correlate of mortality was lack of post-procedural anti-platelet therapy (HR: 152.16, 95% CI: 53.57, 432.22, P < 0.001) (Table 4). Further analyses indicated that 11 PCI-treated patients had a lack of post-procedural anti-platelet therapy (neither aspirin nor thienopyridine) with 10 reported early in-hospital deaths (<45 days) and the 11th patient's clinical outcome censored. All 10 deaths were cardiac related and of multifactorial causes which precluded anti-platelet administration. Two deaths were related to GI bleeding, one death secondary to retroperitoneal bleeding resulting from procedure-related femoral vascular access, three deaths secondary to coronary perforation (two cases requiring surgical intervention), and three further cases necessitating CABG after PCI-related complications following stent implantation (two cases secondary to post-procedural acute stent thrombosis and one case for extraction of ruptured balloon catheter after stent implantation), with withdrawal of treatment in one case due to poor prognosis. Detailed patient narratives of these 10 deaths are available in the Supplementary Data.

Being discharged on one anti-platelet agent alone was not an independent correlate of mortality (n = 39, one patient event censored), with no in-hospital deaths reported. There was no correlation between a ‘history of GI bleeding or peptic ulcer disease’ and a lack of post-procedural anti-platelet therapy (Pearson correlation −0.02, P = 0.51) or patients taking one anti-platelet agent at discharge (Pearson correlation 0.04, P = 0.28).

Amiodarone therapy on discharge [5 of 12 patients dying within 4 years (1 death in-hospital)] (HR: 4.49, 95% CI: 1.36, 14.83, P = 0.014) and pre-procedural poor left LVEF were independent correlates of 4-year mortality. Both variables showed minor correlation (Pearson correlation 0.3, P < 0.001).

Other independent correlates of 4-year mortality included a ‘history of GI bleeding or peptic ulcer disease,’ PVD [22 of 80 patients (27.5%) dying within 4 years (4 in-hospital deaths)] (HR = 2.13, 95% CI: 1.26, 3.60, P = 0.005), age and female gender [38 of 207 patients (18.4%) dying within 4 years (8 in-hospital deaths)] (HR: 1.60; 95% CI: 1.01, 2.56; P = 0.048). Female gender had a weak negative correlation with the SYNTAX score (Pearson correlation coefficient: −0.09, P = 0.007) and a weak positive correlation with age (Pearson correlation coefficient 0.18, P < 0.001).

The SYNTAX score was an independent correlate of 4-year mortality in the PCI arm. For every 10-point-increase in the SYNTAX score, this associated with a 1.25 times greater risk of 4-year mortality in the PCI-treated patients.

Discussion

The following are the main findings of this study: (i) within the CABG arm, the lack of discharge aspirin therapy was the strongest independent correlate of 4-year mortality; (ii) within the PCI arm, clinical factors were more strongly associated with 4-year all-cause death compared with anatomical factors (namely the SYNTAX score); in addition procedural related factors, such as stent number or length, were not correlates of 4-year mortality; (iii) female gender was an independent correlate of 4-year mortality in the PCI arm; (iv) PVD was a common independent correlate of 4-year mortality in the CABG and the PCI arms—the presence of which may be a marker of native coronary disease severity and the risk of future coronary disease progression; and (v) the possibility of the outcomes of the SYNTAX trial having been affected by the significant differences in censored patient numbers between the CABG and the PCI arms cannot be excluded.

Mechanism of late (4-year) mortality in coronary artery bypass graft and percutaneous coronary intervention patients

Within the CABG and the PCI arms, PVD was an independent correlate of 4-year Death. The presence and the severity of extra-cardiac disease, defined by parameters such as carotid intima media thickness (CIMT) and ankle brachial pressure index, have previously been associated with more extensive baseline coronary artery disease.^14–21 For example, Ikeda et al.¹⁵ recently associated the severity of the SYNTAX score with the CIMT. More complex baseline coronary artery disease has in turn been associated with adverse long-term prognosis as reported in the Coronary Artery Surgery Study²¹ and Rotterdam¹⁸ registries. Within these registries, more extensive baseline preoperative coronary disease was associated with long-term mortality (>10 years), which were further linked to a higher prevalence and severity of other clinical risk factors. Notably, within the Bypass Angioplasty Revascularization Investigation trial, native coronary disease progression, and not failed revascularization, was the predominant determinant of jeopardized myocardium and recurrence of angina at 5-year follow-up in CABG- or PCI-treated patients.²²

Lack of discharge anti-platelet therapy in the coronary artery bypass graft arm

The findings of lack of discharge aspirin after CABG to be an independent correlate of 4-year mortality are contrasted by previous studies that have potentially related aspirin therapy on the day of surgery to a reduction in in-hospital or at most 30-day mortality.²³ Historically, aspirin therapy on the day of surgery has been associated with a 50–70% reduction in early and late (up to 1-year) aorto-coronary vein-graft occlusion.^24–26 Graft patency at 1-year has, however, not translated into increased mortality. Moreover within the angiographic sub-study of SYNTAX trial,²⁷ despite over a quarter of the CABG patients (27.2%) found to have significantly diseased (≥50 to<100%) or obstructed (100%) bypass grafts at 15 months, this was not significantly associated with early  major adverse cardiovascular and cerebrovascular events. Potential reasons for this latter phenomenon have been related to bypass grafts prematurely failing if anastomosed distal to functionally non-significant lesions, because of competitive flow, or the recurrence of anginal symptoms if bypass grafts to a functionally significant lesion failed.^27–30

Why aspirin therapy in the present study was a strong independent correlate of 4-year mortality in the CABG arm is unclear. As previously discussed, patients with more complex coronary disease have been associated with the presence of extra-cardiac arteriopathy.^14–21 It may, therefore, be hypothesized that aspirin therapy was potentially protective against ischaemic complications arising from cardiac and systemic atherosclerotic disease (cerebral, renal, and GI), which may perhaps be more likely to occur in the study population.^14–21 This hypothesis is supported by the findings of most of the deaths in patients not on aspirin therapy at discharge in the CABG arm to be cardiac or vascular related (15 out of 27 deaths).

History of gastrointestinal bleeding or peptic ulcer disease

A history of GI bleeding or peptic ulcer disease was an independent correlate of 4-year mortality in the PCI arm (P = 0.004), with a trend towards significance in the CABG arm (P = 0.085). Aspirin alone, clopidogrel alone, and their combination have all been associated with an increased risk of GI bleeding.^31–33 Consequently, the association of a ‘history of GI bleeding or peptic ulcer disease’ with mortality is perhaps not unexpected, given the mandatory requirement of dual anti-platelet therapy after PCI to prevent stent thrombosis, and the potential long-term protective effect of aspirin therapy in CABG-treated patients as previously discussed. Notably, Iakovou et al.,³⁴ in a well-designed ‘real-world’ registry of first-generation DES (sirolimus- or paclitaxel-eluting stents) (n = 2229), demonstrated that the premature discontinuation of anti-platelet therapy is strongly associated with acute stent thrombosis at 9-months following DES implantation.

Within PCI patients, the use of concomitant proton pump inhibitors with clopidogrel therapy for gastro-protection has been controversial, because of a potential interaction with the anti-platelet effect of clopidogrel. Until now, no study has convincingly demonstrated any significant impact on adverse clinical outcomes, although this effect cannot be excluded in certain subgroups, such as poor metabolizers of clopidogrel.^35–38 Further studies are required to investigate the cost/risk–benefit ratio of selectively or routinely adopting prophylactic gastro-protective therapy in PCI (and possibly CABG)-treated patients.

Gender

Despite adjustments for risk factors, including age and the SYNTAX score, female gender was an independent correlate of 4-year mortality in the PCI arm of the SYNTAX trial, a finding at odds with other contemporary DES trials showing no gender effect.^39–43 The main difference being that females in the SYNTAX trial had more complex baseline coronary disease (and therefore potentially a greater atherosclerotic burden)—mean SYNTAX score of 26.5 ± 11.9—compared with 12.9 + 8.4 in women from three pooled All-Comers randomized DES trials (SIRTAX, LEADERS, RESOLUTE).³⁹

A recent sub-study of the PROSPECT trial (Providing Regional Observations to Study Predictors of Events in the Coronary Tree)^44,45 has shown that there may be a potential ‘gender effect’ in the pathophysiology and composition of atherosclerotic plaque leading to acute coronary syndrome. Women (compared with men) were shown to have less extensive coronary artery disease by angiographic and intravascular ultrasound assessments, less plaque rupture, less necrotic core and calcium, similar plaque burden, and smaller lumens. Conversely, the presence of thin-cap fibroatheroma was demonstrated to be a stronger marker of plaque vulnerability and risk of future clinical events in women compared with men. Importantly, despite these differences, clinical outcomes were reported to be comparable between the sexes in the PROSPECT trial, although most events occurred in angiographically inapparent lesions. Patients from the SYNTAX trial had substantially more complex coronary disease compared with the PROSPECT trial (which excluded patients with left main coronary artery disease and 3VD requiring PCI). Thus, the possibility of more unfavourable plaque composition in obstructive and non-obstructive lesions of females with advanced atherosclerotic disease being a possible explanation for the potential increased mortality in the SYNTAX trial cannot be dismissed.^14,45,46

Study limitations

This study represents a post hoc analysis of the SYNTAX trial and the results should be considered hypothesis generating. Since the proportion of patients lost to follow-up varied between the two study arms (CABG arm: 8.7%, PCI arm: 2.7%, P< 0.001), this potential confounding factor should be considered when interpreting the results. Although multivariable adjustments were performed for significant confounders (P < 0.05), the possibility of other unmeasured confounders having affected the results cannot be excluded. Within the CABG arm as all correlates that were significantly associated with increased mortality on univariable analyses (P < 0.05) were included in the multivariable model, this may have resulted in over-fitting since there were 78 deaths. As the confidence intervals of these correlates were narrow, this would have limited this risk.⁴⁷ In the PCI population, there were enough events (104 deaths) to include all significant correlates. The lack of CK–MB ratios in all study patients precluded its assessment in the multivariable analyses.

Conclusion

Independent correlates of 4-year mortality in the SYNTAX trial were multifactorial. Lack of discharge aspirin (CABG arm) and female gender (PCI arm) are notable independent correlates of 4-year mortality that require confirmation in further studies. Peripheral vascular disease is a common independent correlate of 4-year mortality and may be a marker of the severity of baseline coronary disease and the risk of future native coronary disease (and extra-cardiac disease) progression.

Supplementary material

Supplementary Data.

Funding

The SYNTAX trial was funded by Boston Scientific Corporation.

Conflict of interest: K.D.D., P.P., and J.H. are all full-time employees in Boston Scientific. K.D.D. holds stock in Boston Scientific. M.M. has served on the Speaker's Bureau of Cordis and Medtronic and reports no financial disclosures. T.F. reported research grants and consulting from Abbott Vascular, Boston Scientific, and Edwards Lifesciences. M.M. reported that her institution received a research grant from Boston Scientific. The other authors report no conflicts of interest.

Acknowledgements

The authors express their gratitude to all the study centres and participants whose work made this study possible. Thanks are due to Peggy Pereda and Jian Huang of Boston Scientific Corporation for their invaluable technical support in accessing the study database.

References