The study design included patients with three vessels (71 %) and left main stenosis (29 %) in diabetic and non- diabetic populations [12–14]. SYNTAX was a non-inferiority trial and the primary end point of major adverse cardiac and cerebrovascular events at 1 year was not met due to the higher rate of TVR in the PCI arm. Secondary end points included the same end points at 5 years, rates of each individual component and rates of SET or graft occlusion. Taxus Express stents (Boston Scientific, Natick, MA, USA) were used in all PCI patients; 4.6 stents per patient were used and per protocol, all vessels with a reference diameter > 2.0 mm should be treated.

At 5 years, adverse events were higher in the PCI arm compared to CABG (37.3 and 26.9 %, respectively, p < 0.001). This difference between arms was driven by the higher rate of TVR, MI and cardiac death, although this benefit was more pronounced in patients with 3 vessel CAD (death: 14.6 and 8.2 % with DES and CABG respectively p < 0.001; MI: 10.6 and 6.1 % with DES and CABG respectively p = 0.003). On the contrary, in most patients with left main stenosis, long term outcomes were similar between both revascularization strategies. The authors made a post hoc subgroup analysis using a risk /benefit angiographic score called “Syntax Score”, and they found that patients with low scores (<22) had similar composite end points with both revascularization strategies . In contrast, in patients with an intermediate or high “Syntax score”, the benefit was in favor of bypass surgery. One might argue that, since the primary end point of the study was not met, these subgroup analyses should not be allowed.

Analyzing the subgroup of 452 patients with treated diabetes, CABG had a better survival and lower incidence of MI only in the subgroup with insulin dependent diabetes (IDDM), whereas in non- insulin dependent diabetes (NDDM), the differences between PCI and CABG were only driven by a greater number of revascularization procedures with PCI.

This study had potential limitations:

The FREEDOM trial was a multicenter, open-label prospective randomized superiority trial of PCI/DES versus CABG in 1900 diabetic patients in whom revascularization was indicated with stenoses of more than 70 % in 2 or more major epicardial vessels involving ≥2 separate coronary-artery territories and without left main stenosis. In the PCI arm, DES was used in 100 % of the cases and Taxus and Cypher (Johnson & Jonhson, Cordis, Miami Lakes, FL, USA) were the predominant stents [15, 16].

Recently, the FREEDOM trial results at 5 years have shown that for diabetics and multiple vessel disease, CABG was superior to PCI with DES in that CABG significantly reduced death rates and MI, albeit with a higher rate of non- fatal CVA. Similar to SYNTAX, the primary end point of death, MI and CVA became significant in the CABG arm only after the second year of follow up. The composite end point was 13 % vs 10.9 %, p = ns, at the 2nd year of follow up. It rose to 26.6 % vs 18.7 %, in PCI and CABG respectively at 5 years, making the differences highly significant in favor of CABG, p = 0.005. At 5 years, the advantage of CABG over PCI was seen in both IDDM and NIDM; an increase of 33 and 57 % in the combined end point of mortality and MI respectively with PCI, in spite of the 54 % increase in non- fatal CVA with surgery [15, 16]. The stroke risk was higher with CABG in both groups of diabetic patients-NIDM (1,7 % vs 4.3 % with PCI and CABG respectively) and IDDM (3.7 % vs 7.5 % with PCI and CABG respectively) although the differences were not significant – 5 years stroke HR, PCI vs CABG was 0.51 (0.25–1.06) in NIDM and 0.60 (0.28–1.30) in IDDM. Repeat revascularization procedures were low with CABG [16]. In FREEDOM, the advantages with surgery were independent of the SYNTAX score, underscoring the potential limitations of this score to predict outcome in all groups.

In FREEDOM, there were geographic differences in the results. The primary end point was reached in favor to CABG only in USA and Canada sites; death/MI/CVA was 16 and 28 % with DES and CABG respectively, from a population of 770 patients. Conversely, non-USA and Canada sites, from a population of 1130 patients, the primary end point was 25 % with DES and 21 % with CABG which was not statistically significant, with a p = 0.05 for interaction between North American and outside North American sites. These observed geographic disparities between PCI and CABG in FREEDOM should not be a surprise as regional differences in results between PCI and CABG were also seen in RCTs in the BMS era. Moreover, 2 South American studies -ERACI II and MASS II- showed similar mortality with both revascularization strategies in non- diabetic and diabetic patients treated either with BMS or CABG [RR 1 (0.27–3.72) and 0.95 (0.41–2.22) in ERACI II and MASS II, respectively] [5, 6]. In contrast, another European BMS/CABG trial showed significant survival advantages with CABG at 6 years [8] and the advantages were driven by extremely low in-hospital mortality with surgery.

Therefore, both old (ERACI, MASS and SOS) [5, 6, 8] and new RCTs (FREEDOM and SYNTAX) [12, 15] have shown large geographic inconsistencies in the results between PCI and CABG. These differences may profoundly affect the interpretation of these trials.

In this trial, 510 diabetic patients were randomized to either PCI or CABG. The first year of follow up was published and the 5 years was reported [17, 18]. This was a non-inferiority trial that compared, as the primary end point, a composite of death, MI and CVA. In the PCI arm, DES was used in 69 % and BMS in 31 %.

At 5 years, PCI was non inferior to CABG. Death, MI and CVA were 20.5 % with CABG and 26.6 % with PCI (p = 0.11). However, significant differences in MI, 6.3 % with CABG and 14 % with PCI, p = 0.007, and a repeat revascularization rate of 8.3 % with CABG vs 21.9 % with PCI p < 0.001 was seen. Overall death and non- fatal CVA were similar. A major limitation was the small sample size with lack of power to detect end point differences between groups [18].

This study randomized 198 patients, 97 to surgery and 101 to PCI, and it was terminated due to low recruitment rate, and at that point no significant differences between groups were found [19].

Major limitation: the VA CARDS was severely underpowered for its primary endpoint and therefore no firm conclusions about the comparable effectiveness of CABG and PCI were possible. However, at 2 years, a 76 % reduction of death was noted in CABG group. On the other hand, the incidence of MI, was significantly higher in CABG group. Similarly to the FREEDOM trial [12], SYNTAX score did not predict differences in outcome (Table 14.2).

We have now the opportunity to analyze results from the 8 RCT comparing stents versus CABG; a first meta-analysis showed a significant safety advantage of bypass surgery over PCI only in patients with diabetes with the penalty of a threefold increase stroke risk with CABG (RR 1.84: 1.18–2.53). As we can see in Table 14.3, death in the overall group was significantly lower with CABG (0.79 [0.69–0.90] p < 0.001), although this benefit was only driven by significant differences in the diabetic subgroup of patients. On the contrary, survival was almost identical with PCI or CABG in the non-diabetic group (8.9 and 9.1 % with CABG and PCI respectively, p = 0.80). The same findings were seen with the composite of death/MI/CVA- a significant benefit with CABG only driven by the outcome in diabetics (p < 0.001), Table 14.3.