Impact of repeat revascularization within 5 years on 10-year mortality after percutaneous or surgical revascularization

The higher rate of additional revascularization required after percutaneous coronary intervention (PCI) compared to coronary artery bypass graft (CABG) has been one of the Achilles' heels of PCI, and this was first addressed by the introduction and routine use of bare metal stents [1‐3], with further improvements seen in contemporary practice by their replacement with ever improving drug-eluting stents (DES) [4]. Nevertheless, recent trials comparing PCI to CABG continue to underscore the superiority of CABG in the reduction of this adverse event [4‐7].

Repeat revascularization (RR) remains a potential complication in both PCI and CABG patients. Although RR is considered an adverse outcome or failure of the initial treatment, as a strategy it often offers an efficient treatment associated with a reduction in morbidity or mortality [8]. Some authors argue that RR cannot be considered a reliable outcome indicator because of a large number of confounding factors that contribute to the event, ranging from variable indications, the differences in the availability of revascularization targets, to patient preferences [9]. However, the clinical relevance of RR is not negligible, and its negative impact on major adverse events and quality of life has been consistently reported. The SYNTAX trial reported higher rates of the composite endpoint of death, stroke, and myocardial infarction (MI) at 5 years, among patients who were initially randomized to PCI and then underwent secondary revascularization compared to those who did not. A similar, but less impactful, trend was reported among those randomized to CABG [10]. More recently, analyses related to RR in patients with left main (LM) lesions from the EXCEL trial have reaffirmed the association between RR and the risk of 3-year all-cause and cardiovascular mortality after both PCI and CABG [11].

Despite these data, there is currently no evidence as to whether these trends are maintained or amplified long term. Specifically, the impact of RR on all-cause death beyond 5 years has not been fully elucidated. The present study aimed to investigate the impact of RR within 5 years of PCI or CABG on all-cause death among patients with 3VD and/or LMCAD beyond the original 5-year follow-up of SYNTAX trial.

A total of 330 out of 1800 patients (18.3%) underwent ≥ 1 RR within 5 years of their initial procedures, which amounted to 454 RR procedures (390 additional PCI and 64 additional CABG). The 330 patients undergoing RR were made up of 237, 69, 17, and 7 patients having 1, 2, 3, and 4 RR procedures, respectively (Fig. 1). According to the type of RR, patients were categorized as follows: 266 patients in RR-PCI group, 45 in RR-CABG group, and 19 in RR-PCI/CABG group (Fig. 1).

The present study investigated the association between RR within 5 years of the index PCI or CABG and 10-year all-cause death, among patients with 3VD and/or LMCAD in SYNTAX trial. The main findings are:

Numerous studies have demonstrated that RR has a negative impact on short- and mid-term outcomes. In patients who underwent elective primary revascularization with isolated CABG, RR with PCI had a poorer prognosis at a median follow-up of 58 months, compared with those who did not [17]. A pooled patient-level analysis from 21 randomized PCI trials demonstrated that target lesion revascularization (TLR) after PCI increased mortality at a median follow-up of 37 months, which was mainly driven by higher rates of MI occurring after TLR [18]. In another patient-level pooled analysis of 1001 patients, TLR was associated with a higher rate of 5-year mortality compared with those without TLR [19]. Similarly in EXCEL trial, the need for a RR increased the risk of 3-year all-cause death after both PCI and CABG (p_{-interaction =}0.85) [11]. These studies only had medium-term follow-up of 3 to 5 years, the impact of RR on very long-term all-cause death was previously unknown.

In line with the 5-year results of SYNTAX trial [10], we observed a comparable rate of 10-year all-cause death between patients with and without a RR. Whether the impact of RR is related to the mode of primary revascularization is uncertain. In EXCEL trial, among patients having a RR, 3-year all-cause death was numerically higher if the index procedure was with PCI rather than CABG (10.4% vs. 9.1%) [11]. In SYNTAX trial, at 5 years, among patients who underwent RR, those initially randomized to PCI as opposed to CABG had significantly higher rates of the composite of death, MI, or subsequent RR (57.4% vs. 38.4%, p = 0.003), and a trend for higher mortality (20.2% vs. 13.9%, p = 0.095) [10]; with the difference in mortality significant by 10 years (Fig. 3). As the value of SYNTAX scores in predicting clinical outcomes [20], we explored SYNTAX score II 2020 for predicting 10-year mortality in patients with RR. Similarly, we demonstrated that the majority of patients (n = 295) having a RR have a higher rate of 10-year mortality after primary revascularization with PCI than CABG (Fig. 5). SYNTAX score II 2020 clearly identifies those individuals who derive a treatment-specific survival benefit.

There are many possible reasons for these findings. In SYNTAX study, RRs after PCI were mainly target vessel revascularizations, and to a lesser extent treatment of de novo lesions. Disease progression and restenosis appear to be more aggressive after index revascularization with PCI than CABG [10], which may be due to the detrimental effect that coronary stenting has on endothelial function in the distal coronary bed [21, 22]. In addition, bystander enemies, such as diabetes, may contribute to endothelial dysfunction and aggressive disease progression, in particular, in patients who underwent initial PCI, likely influencing their long-term prognosis. As previously reported, indeed, in PCI group, diabetic patients were significantly more frequent among those requiring additional revascularization than those not (34.1% vs. 22.8%, p < 0.001) [10]. In contrast, bypass grafts to the mid-part of coronary vessels not only removes the vulnerability of proximal lesions, but also potentially offers biological protection and prophylaxis against the development of de novo disease [23].

In addition, initial treatment with CABG offers complete revascularization more frequently than PCI, which may result in a greater protective effect on long-term prognosis [24]. Consistently, PCI patients undergoing RR had a significantly higher incidence of incomplete revascularization than those not requiring RR, whereas such a difference was not observed in the index CABG group (Table S2). This suggests that incomplete revascularization during CABG may not have significant impact on RR risk, probably because most vessels that were not revascularized are either chronically occluded or too small, making their treatment irrelevant. However, the vein graft failure over a 10-year period is high, and might attenuate these benefits. Indeed, multi-arterial CABG is associated with lower mortality [25], which may also partially contribute to the higher rate of 10-year all-cause death after primary revascularization with PCI compared to CABG.

The impact of type of RR on mortality remains uncertain, with previous results inconsistent. A patient-level pooled analysis demonstrated that the type of RR did not affect mortality after TLR of LMCAD [19]. In EXCEL trial, RR using CABG was strongly associated with increased 3-year mortality [11]; in our analysis, after adjustment for confounders, RR-CABG was the only independent predictor of 10-year all-cause death. Of note, the high risk of RR with CABG for 10-year all-cause death was mainly contributed by patients whose primary revascularization was with CABG (Table 2). Furthermore, in our study, the risk of mortality was higher following secondary revascularization with CABG, compared to PCI or both PCI and CABG. Similarly, Locker et al. showed that redo CABG increased 30-day mortality compared to RR with PCI in patients with previous CABG [26]. Given that this result was observed after adjustment, our findings indicate that RR with CABG is a high-risk procedure especially for those with prior CABG, and should, therefore, be considered carefully with the patients. However, considering there were very limited patients who underwent RR with CABG in isolation or following PCI, our findings should be considered as exploratory and hypothesis generating. In addition, RR-CABG patients usually present severe 3VD either not amenable with PCI or where PCI already failed with a large area of myocardium at risk, or yet present non-functioning LIMA-LAD bypass that is a predictor of long-term survival, characteristics that generate biases and preclude definite conclusions. Nevertheless, in patients with previous CABG, the ESC Guideline recommends PCI as the first choice for RR if technically feasible, rather than redo CABG [27]. The recently published ACC/AHA Guideline also recommends that in patients with previous CABG with a patent LIMA to the LAD who need RR, if PCI is feasible, it is reasonable to choose PCI over CABG [28].

At 5 years in SYNTAX trial, patients who underwent RR by PCI had a numerically higher rate of all-cause death compared to those who did not, when their initial revascularization was also by PCI (16.6% vs. 13.2%, p = 0.26), whereas the opposite was seen after initial CABG (6.3% vs. 12.1%, p = 0.084) [10]. At 10 years, results were similar, with a trend for increased mortality following primary and secondary revascularization with PCI, and a significantly lower risk when primary CABG was followed by PCI (Table 2). In contrast in EXCEL trial, RR with PCI or CABG were both associated with an increased risk for all-cause death regardless of the initial revascularization approach (PCI or CABG), with no significant interaction observed between the initial revascularization procedure and any type of RR [11]. Ultimately, these inconsistent findings need to be explored in adequately powered clinical studies.

The present study is a post hoc analysis of SYNTAXES trial and may not have adequate statistical power. The main limitation is that RR was no longer recorded after 5 years, we cannot be confident that our findings are reflective of RR in its entirety up to 10 years. Second, the number of patients who underwent RR was limited, especially for those who had repeat CABG or both procedures. Our analysis may lead to the likelihood of spurious findings, all results should be considered exploratory and hypothesis generating. Third, angiographic follow-up was not routinely performed in SYNTAX trial and may underestimate the true rate of RR, especially in patients with silent ischemia. Moreover, not all confounders may have been identified in our multivariable Cox model which assessed the association between RR and all-cause death. Furthermore, the endpoint in SYNTAXES study was all-cause death only; however, death has been considered the most robust and unbiased index for clinical assessment, and is less likely to be affected by ascertainment bias [29]. Finally, the stent-type and peri-interventional therapy used for repeat revascularization were not available. SYNTAX trial enrolled patients with 3VD and/or LMCAD and patients received PCI with first-generation DES; hence, our results should not be extrapolated to general CAD patients in contemporary practice. Further investigations, including latest generation stents and guideline-oriented medical therapy, are warranted.

In patients with 3VD and/or LMCAD undergoing PCI or CABG, RR within 5 years was not associated with the risk of 10-year all-cause death in the whole population. Among patients requiring any repeat procedures, a higher death rate was observed after primary revascularization with PCI than CABG. The SYNTAX score II 2020 can identify individuals who derive a treatment-specific survival benefit. RR with CABG was associated with an increased risk of 10-year all-cause death especially in patients with previous CABG. These exploratory findings should be investigated in larger populations of patients either pooled retrospectively or enrolled in prospective future studies.