Ten-year survival benefit and appropriateness of surgical versus percutaneous revascularization in synergy between percutaneous coronary intervention with Taxus and cardiac surgery randomized trial

Abstract

OBJECTIVES

Average treatment effects from randomized trials do not reflect the heterogeneity of an individual’s response to treatment. This study evaluates the appropriate proportions of patients for coronary artery bypass grafting, or percutaneous intervention based on the predicted/observed ratio of 10-year all-cause mortality in the SYNTAX population.

METHODS

The study included 1800 randomized patients and 1275 patients in the nested percutaneous (n = 198) or surgical (n = 1077) registries. The primary end point was 10-year all-cause mortality. The SYNTAX score II-2020 was validated internally in the randomized cohort and externally in the registry cohort. Proportions of patients with survival benefits from coronary artery bypass grafting or percutaneous intervention were determined using SYNTAX score II-2020.

RESULTS

Ten-year mortality was 23.8% for coronary artery bypass grafting, 28.6% for percutaneous intervention in the randomized cohort, 27.6% for coronary artery bypass grafting and 55.4% for percutaneous intervention in the registries. In the coronary artery bypass grafting registry, the SYNTAX score II-2020 predicted 10-year mortality with helpful calibration and discrimination (C-index: 0.70, intercept: 0.00, slope: 0.76). The proportion of patients deriving a predicted survival benefit from coronary artery bypass grafting over percutaneous intervention was 82.4% (2143/2602) and 17.7% (459/2602) for the entire SYNTAX trial population, translating into a 4.7 to 1 appropriate ratio of treatment allocation to coronary artery bypass grafting and percutaneous intervention.

CONCLUSIONS

Choosing a revascularization modality should depend on an individual’s long-term prognosis rather than average treatment effects. Additionally, patients should be informed about their predicted prognosis.

Trial registration

Registered on clinicaltrial.gov

SYNTAXES

NCT03417050 (https://clinicaltrials.gov/ct2/show/NCT03417050)

SYNTAX

NCT00114972 (https://www.clinicaltrials.gov/ct2/show/NCT00114972)

Open in new tabDownload slide

INTRODUCTION

The Synergy between percutaneous coronary intervention (PCI) with Taxus and Cardiac Surgery Extended (SYNTAXES) trial is the extended 10-year follow-up of the SYNTAX trial [1]. An intention-to-treat analysis of patients in the randomized cohort concluded that coronary artery bypass graft (CABG) did not significantly reduce 10-year mortality compared to PCI [1]; however, this result only provides the average treatment effects, expressed as a risk difference in outcome amongst only those clinically amenable to both treatments and thus eligible for randomization [2].

However, unrestricted inclusion of all patients is critical to establish the generalizability and appropriateness of a treatment selected for an individual in real-world daily practice. The SYNTAX trial’s major strength was including patients with de novo three-vessel disease and/or left main coronary artery disease (CAD) eligible for both CABG and PCI in the randomized cohort. Those with complex CAD expected to have poor PCI outcomes were placed in a CABG registry, while patients ineligible for CABG due to comorbidities and surgical risks were entered into a PCI registry [3].

While average treatment effects from randomized control trials (RCTs) often form the basis of treatment guidelines, they do not apply to all patients with complex CAD treated in everyday practice and do not reflect individual responses to treatment. Patients often favour PCI for its minimal invasiveness, despite objective evidence indicating potentially fatal outcomes. Probabilistic outcome predictions, crucial for informed decisions, are rarely shared with patients, who are the most affected party in the treatment decision process [4].

The SYNTAX score II 2020 (SSII-2020), based on clinical and angiographic characteristics of individual patients, was developed for predicting 10-year all-cause death based on the SYNTAX(ES) randomized cohort, with external validation performed in 4 RCTs with 5–10 year follow-up [5] and in 1 large registry [6].

This study sought to evaluate the appropriateness of the revascularization strategy (PCI or CABG) pursued in the all-comers population of the SYNTAX study based on the predicted and observed ratio of 10-year all-cause mortality.

PATIENTS AND METHODS

Study design and patients

The design, rationale and detailed inclusion and exclusion criteria of the SYNTAX trial have been described previously [3, 7]. The study flow chart is presented in the graphical abstract. Medical Ethical Committee approval for this study was granted at the institution of the principal investigators. The study protocol was consistent with the Declaration of Helsinki. Informed consent to obtain information on 10-year vital status was waived. The primary end point of the SYNTAXES study was all-cause death at 10 years.

Statistical analysis

The analyses were conducted based on intention to treat. Continuous variables were reported as mean ± standard deviation or median and interquartile range, while categorical variables were reported as percentages. Unlike the randomized cohort, serum creatinine was not measured continuously in the registry. To calculate SSII-2020, multiple imputations (20 times) were used based on correlations among predictors, following the missing at-random assumption to efficiently use available data without bias.

The discriminative ability of the SSII-2020 for outcome risk was assessed with Harrell’s C-statistic in the 2 randomized cohorts and the 2 nested registries. Calibration plots were generated to visually evaluate the agreement between predicted and observed rates of 10-year mortality in each treatment arm, with smooth calibration curves based on a Cox model that fitted outcomes to a restricted cubic spline of the predictions [5]. The individual scatter plots of absolute and individual risk differences in predicted all-cause mortality following surgical or percutaneous treatment in the randomized cohorts, or the nested registries, were ranked in descending order according to the predicted PCI minus predicted CABG mortality and connected with the use of locally estimated scatterplot smoothing curves. The observed mortality of treatments in each individual was calculated by the mortality of the nearest 10% population (after ranking) and then fitted by the locally estimated scatterplot smoothing curve. All analyses were performed with R, version 3.5.0 or higher (Foundation for Statistical Computing, Vienna, Austria).

RESULTS

From March 2005 through to April 2007, 4337 consecutive patients were screened for eligibility into the SYNTAX trial [3], with 1800 entering the randomized cohort and 1766 receiving CABG or PCI. Patients not eligible for both PCI and CABG were entered into nested registries [3]. In the CABG registry, 644 of the 1077 patients who underwent CABG were randomly selected by the central allocation service to be clinically followed up for major adverse cardiovascular and cerebrovascular events, whereas all patients in the PCI registry who received PCI (n = 192/198) were followed-up [3]. After 5-year clinical follow-up [8], only vital status was ascertained (Graphical abstract) [1].

Baseline demographics and lesion characteristics in the registry and randomized trial patients

Baseline characteristics are shown in Table 1. Reasons for inclusion in the CABG registry included complex coronary anatomy considered unsuitable for PCI by the interventional cardiologist (70.9%), chronic total occlusions deemed untreatable with PCI (22.0%), inability to take anti-platelet medication (0.9%), refusal to undergo PCI (0.5%) or other reasons (5.7%). Due to prohibitive comorbidities associated with a higher predicted surgical mortality, 198 patients were deemed inoperable and entered into the PCI registry, with 192 treated with PCI. Reasons for inclusion in the PCI registry were as follows: high-risk for CABG (70.7%), no graft material for anastomosis (9.1%), refused CABG (5.6%), small or poor-quality distal vessels (1.5%) or other reasons (13.1%) [3].

Baseline characteristics in the nested CABG and PCI registries have been reported previously [3]. In brief, patients in the CABG registry had similar age and operative risk [Logistic EuroSCORE: 3.9 (2.7) vs 3.8 (2.7)] to the randomized CABG patients. Among patients treated with CABG, lesion complexity, assessed by the anatomical SYNTAX score, was significantly higher [mean difference 8.8 (95% CI 7.5–10.0)] and left ventricular ejection fraction was significantly lower [mean difference −3.5% (95% CI −5.4 to −1.7)] in the registries patients compared with the randomized patients.

Among patients treated with PCI, patients in the registry were significantly older than those randomized to PCI [mean difference 6.0 years (95% CI 4.4–7.6)]. Furthermore, patients treated with PCI in the registry had more comorbidities than patients treated with PCI in the randomized cohort: COPD [mean difference 11.4% (95% CI 6.0–17.8)], peripheral vascular disease [mean difference 7.1% (95% CI 2.1–13.2)], history of congestive heart failure [mean difference 5.7% (95% CI 1.9–10.9)], renal insufficiency [creatinine <200 µmol/l, mean differences 4.6% (95% CI 2.0–8.9)] and active cancer [mean difference 3.4% (95% CI 1.5–7.1)]. The differences in baseline characteristics between the registry and randomized cohort were more remarkable in the PCI group than in the CABG group (Fig. 1, Supplementary Material, Fig. S1).

Observed mortality in the coronary artery bypass graft and percutaneous coronary intervention randomized and registry cohorts

The 10-year all-cause mortality was 28.8% and 24.5% among patients receiving CABG in the nested registry and randomized cohorts, compared to 55.4% and 28.4% for patients in the PCI registry and randomized cohorts, respectively (Fig. 2; Table 2). A sensitivity analysis assuming all lost-to-follow-up patients were deceased showed that the imbalance did not significantly affect all-cause mortality across both randomized and registry cohorts (Supplementary Material, Fig. S2). The 5-year major adverse cardiovascular events (MACE) rates in the nested registry and randomized cohorts have been reported previously (Table 2) [8].

Individual predicted and observed mortality in the overall, randomized and registered population

At cross-validation, the SSII-2020 showed a helpful discriminative ability in both treatment groups for predicting 10-year all-cause deaths as previously reported [5]. In the registry cohort, the SSII-2020 also showed a helpful discriminative ability in both treatment groups for predicting 10-year all-cause deaths [C-index = 0.72 (95% CI 0.64–0.79) for PCI and 0.70 (0.65–0.75) for CABG, Fig. 3] and 5-year MACE [C-index = 0.68 (0.62–0.74) for PCI and C-index = 0.71 (0.67–0.75) for CABG, Supplementary Material, Fig. S3]. In the PCI registry, the SSII-2020 underestimated 5-year MACE and 10-year mortality since specific comorbidities that entail high mortality are not included in the formula of the SSII-2020, whilst in the CABG registry, it predicted mortality with a helpful calibration and discrimination.

The individual predicted and observed 10-year mortalities following CABG and PCI in the randomized and registry cohorts are displayed in Fig. 4. Overall, 14.6% of patients in the PCI registry, 7.5% in the CABG registry and globally 17.7% of the screened population had a potential survival benefit with PCI and could have been safely referred for percutaneous revascularization. The proportions of patients with a predicted survival benefit following CABG and PCI were, respectively, 78.3% (1383/1766) and 21.7% (383/1766) in the randomized cohorts, and 82.4% (2143/2602) and 17.7% (459/2602) in the entire SYNTAX trial population. The latter 2 percentages correspond to a 4.7:1 ratio of appropriate treatment allocation between CABG and PCI across the entire SYNTAX cohort, which includes both the randomized and registry populations. This cohort comprises patients with multivessel disease, including three-vessel disease as well as left main disease, involving 1, 2, or 3 vessels.

Decision curve analysis comparing PCI and CABG across varying thresholds in the randomized and registry cohorts is summarized in Supplementary Material, Fig. S4.

DISCUSSION

The main findings from this 10 year’s results of surgical versus percutaneous revascularization in an all-comers analysis of the SYNTAX trial are:

1. Ten-year mortality was higher in registries for both PCI-treated (registry versus randomized cohort: 55.4% vs 28.4%) and CABG-treated patients (28.8% vs 24.5%).

2. Significant differences in baseline characteristics were observed between the randomized cohort and nested registries, more pronounced in the PCI population than in the CABG population.

3. The SSII-2020 showed reasonable agreement with observed mortality in both randomized and registry cohorts. Predicted survival benefit favoured CABG over PCI for 82.4% (2143/2602) of the SYNTAX trial population; yielding a 4.7 to 1 appropriate ratio for CABG to PCI.

Key differences between randomized and nested registry population

In the intention-to-treat analysis of the randomized cohort, there was no statistically significant average treatment benefit in terms of survival with CABG over PCI [1]. However a recent meta-analysis of 5180 patients from 4 RCTs (BEST, FREEDOM, PRECOMBAT, and SYNTAX) with over 5 years of follow-up found that PCI was linked to higher all-cause mortality over 10 years (HR 1.19; 95% CI 1.04–1.32; P = 0.008) [9]. Within the first 5 years, PCI showed increased mortality risk (HR 1.20; 95% CI 1.06–1.53; P = 0.008), but no significant difference was seen during the 5–10-year period (HR 1.03; 95% CI 0.84–1.26; P = 0.76) [9]. Importantly, out of 3035 SYNTAX trial patients, only 1766 were randomized and treated as assigned, while the rest entered nested registries due to ineligibility for both treatments. This important trial detail is often overlooked. In the registry population, 10-year mortality for PCI was significantly higher than for CABG (55.4% vs 28.8%), strongly influenced by baseline patient characteristic differences between the 2 treatment arms. There were significant differences in the baseline characteristics of patients enrolled in randomized cohort compared with patients entered into nested registries. Regarding CABG, only left ventricular ejection fraction and anatomical SYNTAX score were significantly different between patients included in the randomized cohort and registry cohort, whereas for PCI, eight baseline characteristics (age, current smoking, COPD, peripheral vascular disease, creatinine >200 µmol/l, history of congestive heart failure, active cancer, and anatomical SYNTAX score) were significantly different (Fig. 2).

It is not surprising that patients with extremely complex lesions and extensive disease are often referred to surgeons for CABG by interventional cardiologists. Conversely, those deemed inoperable or very high surgical risk by surgeons are entered into the PCI registry. This small group results from surgeons’ pragmatic refusal to operate due to severe comorbidities.

These differences likely reflect the gap between registries, which are more representative of real-world clinical practice, and RCTs, which are designed to test hypotheses that are aimed at changing and possibly improving current treatment standards [2, 10]. Whilst the results from RCTs are frequently the foundation of treatment guidelines, they focus on average treatment effects, despite there being a potentially wide range of treatment responses among patients.

What is the cornerstone of decision-making?

It is challenging to translate the average treatment effect from RCTs into individuals in real-world clinical practice [10]. Decisions between PCI and CABG should be made by consensus during a heart team consultation (class IC) [11]. However, detailed recommendations for implementing a heart team are not provided in the current guidelines, potentially limiting their utilization in clinical practice and leading to reduced quality of care. Previous reports have provided at best moderate agreement between heart teams for revascularization decisions in patients with complex CAD [12]. Operationalizing the multi-disciplinary heart team (MDHT) model for complex CAD requires adherence to key responsibilities as outlined in societal guidelines: (i) gathering team input; (ii) establishing the final treatment plan; (iii) ensuring communication between the MDHT, patients and providers; (iv) reviewing program metrics (e.g. procedural volumes and clinical outcomes); and (v) keeping team members updated on treatment guidelines and relevant developments. To implement these principles effectively, several operational strategies are essential. Clear criteria for MDHT involvement must be documented in institutional protocols. Regular, structured meetings, ideally held weekly or daily, are necessary for optimal function. Defining each team member’s role is crucial for streamlining meetings. Hospital administration supports the MDHT by managing infrastructure, staffing, coding, resource allocation and marketing. Finally, integrating decision-making tools, such as SYNTAX, STS and EuroSCORE II, ensures consistent, efficient discussions and helps quantify treatment risk trade-offs. Therefore, it is mandatory to use validated models that estimate an individual’s long-term vital prognosis when deciding between modalities of revascularization.

The SSII-2020 has been developed to help decision-making and individualize the heterogeneous treatment responses in patients with complex CAD [5]. To move beyond conventional subgroup analyses, the SSII-2020, a multivariable risk predictive model that simultaneously accounts for multiple patient characteristics affecting treatment effects, enables individualized decision-making for each patient. Notably in the randomized cohorts of the EXCEL trial, deviation from the treatment recommended by the score (i.e. PCI instead of CABG) due to the imposed randomized trial allocation (allocation to PCI) led to an excess of death [13].

The present analysis attempted to segregate the entire enrolled population and predict, with the support of the SSII-2020, the individual survival benefit of patients following PCI or CABG in either the randomized cohort or nested registries. In both randomized and registry cohorts, the SSII-2020 showed helpful discrimination and calibration for predicting 10-year all-cause deaths and 5-year MACE [14]. The patients enrolled in the PCI registry had severe morbid conditions not accounted for in the SSII-2020 (e.g. active cancer or undocumented frailty), in addition to complex coronary anatomy and classical cardiovascular comorbidities well captured by the SSII-2020. Therefore, it is not surprising that the SSII-2020 somewhat loses its predictive value in the PCI registry and the predicted mortality was overall lower than the observed mortality (intercept = 0.38) [15, 16]. Conversely, among CABG patients, the main difference between randomized and registry cohorts was the anatomic SYNTAX score, a key determinant of a high SSII-2020, predicting higher mortality after PCI than CABG. Therefore, the SSII-2020’s accuracy in predicting long-term PCI or surgical mortality was better in the PCI registry cohort. Findings from both cohorts complement each other, advancing knowledge of causation [6].

What was the appropriate ratio between coronary artery bypass graft and percutaneous coronary intervention?

Differences in the allocation to surgical or percutaneous revascularization between randomized clinical trials and registries (‘real world’) are very apparent [17]. In the real world, the APPROACH database (Alberta Provincial Project for Outcome Assessment in Coronary Heart Disease), which includes 4383 patients with three-vessel disease who underwent PCI or CABG between 2017 and 2019 reports a CABG to PCI treatment ratio of 0.35 to 1. However, this ratio varies significantly between the 2 Canadian centres (Centre A: 0.22 to 1, Centre B: 0.57 to 1) [18]. These differing ratios highlight the variability in PCI or CABG assignment within the same country and contrast sharply with the overall appropriate ratio of 4.7 to 1 seen in the SYNTAX trial. This discrepancy becomes evident when analysing observed and predicted all-cause mortality using the SSII-2020, retrospectively.

Only accurate personalized predictions of vital prognosis, validated in both randomized and registry cohorts, will ultimately convince practitioners and scientific societies that personalized treatment recommendations should be rigorously implemented. In the future, a ‘mega-analysis’ by using deep learning based on large randomized or non-randomized data, the so-called ‘big data’, is warranted to predict individual treatment effects not captured by SSII-2020 [19]. Recently, a machine learning-based approach for identifying individuals who benefit from CABG or PCI was feasible and effective [20].

Limitations

In the SYNTAX trial, left ventricular ejection fraction was 98.9% complete as a categorical variable but only 61.9% as continuous. Serum creatinine clearance was 100% complete categorically and 62.1% continuous. To calculate SSII-2020, multiple imputations were performed to account for missing continuous data. This missing data might have impacted the diagnostic performance of SSII-2020, particularly in the registry cohort.

CONCLUSIONS

Selecting the revascularization modality should focus on an individual patient’s long-term prognosis rather than average short-term trial effects, which do not cover all complex CAD patients seen in daily practice. The concordance between individual predicted and observed mortalities (SSII-2020) in both the randomized cohorts and the nested registries of the SYNTAX trial allows us to retrospectively establish the appropriateness of treatment in an all-comers population. With a nearly complete 10-year vital status, the trial shows a realistic CABG to PCI treatment ratio of 4.7 to 1, highlighting surgery’s long-term survival benefits over PCI.

SUPPLEMENTARY MATERIAL

Supplementary material is available at EJCTS online.

FUNDING

German Foundation of Heart Research (SYNTAXES study, 5–10-year follow-up) and Boston Scientific Corporation (SYNTAX study, 0–5-year follow-up).

Conflict of interest: Arie Pieter Kappetein reports to work as an employee of Medtronic, outside the submitted work. Patrick W. Serruys reports personal consultancy fees from Sino Medical Sciences Technology, Philips/Volcano, Xeltis and Heartflow, outside the submitted work. Michael Mack reports non-financial support from Edwards Lifesciences, non-financial support from Medtronic and non-financial support from Abbott, outside the submitted work. All other authors declare no competing interests.

DATA AVAILABILITY

The data underlying this article are available in the article and its online supplementary material.

Author contributions

Patrick W. Serruys: Conceptualization; Formal analysis; Investigation; Methodology; Supervision; Validation; Visualization; Writing—original draft. Kai Ninomiya: Conceptualization; Methodology; Writing—original draft; Writing—review and editing. Pruthvi C. Revaiah: Methodology; Writing—review and editing. Chao Gao: Formal analysis; Methodology; Writing—review and editing. Scot Garg: Writing—review and editing. David van Klaveren: Formal analysis; Writing—review and editing. Yoshinobu Onuma: Conceptualization; Formal analysis; Methodology; Supervision; Writing—original draft; Writing—review and editing. Arie Pieter Kappetein: Investigation; Writing—review and editing. Piroze Davierwala: Investigation; Writing—review and editing. Michael Mack: Investigation; Writing—review and editing. Daniel J.F.M. Thuijs: Investigation; Writing—review and editing. David P. Taggart: Investigation; Writing—review and editing. Milan Milojevic: Writing—review and editing

Reviewer information

European Journal of Cardio-Thoracic Surgery thanks Frank A. Baciewicz Jr., Francesco Formica and Amedeo Anselmi for their contribution to the peer review process of this article.

REFERENCES

ABBREVIATIONS

ABBREVIATIONS
 
• CAD

  Coronary artery disease

 
• MACE

  Major adverse cardiovascular events

 
• MDHT

  Multi-disciplinary heart team

 
• PCI

  Percutaneous coronary intervention

 
• RCT

  Randomized control trial

 
• SYNTAXES

  Synergy between PCI with Taxus and Cardiac Surgery Extended

Author notes