Coronary revascularization strategies in diabetes after FREEDOM – is it already time for another trial?

KEYWORDS:

• Diabetes mellitus
• percutaneous coronary intervention
• coronary artery bypass surgery
• coronary artery disease
• drug eluting stents

Coronary artery bypass grafting (CABG) is considered the gold standard for revascularization in subsets of diabetic, high-risk patients with multivessel disease, high synergy between PCI with taxus and cardiac surgery (SYNTAX) scores, and left main stem involvement. Our recent meta-analysis of 11 randomized control trials showed better outcomes after CABG compared to percutaneous coronary interventions (PCIs) in diabetics with multivessel disease (HR 0.71, confidence intervals 0.61–0.84, p < 0.01, I² = 33%) [1]. A number of further meta-analyses affirmed these findings [2–6].

Diabetics show poorer outcomes and higher revascularization rates after both PCI and CABG. In the bypass angioplasty revascularization investigation trial, survival rates were significantly lower at 10 years among diabetics (PCI 45.5% vs. CABG 57.8%, p < 0.025 in diabetics compared to PCI 77.0% vs. CABG 77.3%, p = 0.59 in nondiabetics) [7]. In the diabetic subgroup, the significant differences in survival between CABG and PCI were confined to only those who received at least a single arterial graft (10-year survival of 64.3% with arterial graft vs. 45.5% with PCI vs. 39.4% with vein grafts). A sub-analysis of diabetics with left main and/or three-vessel coronary artery disease in the SYNTAX trial showed that diabetics were associated with worse outcomes than nondiabetics after either revascularization strategy [8]. Further sub-analyses of Future REvascularization Evaluation in patients with Diabetes Mellitus (FREEDOM) similarly showed worse outcomes for insulin-treated diabetics [9].

1. The FREEDOM trial

FREEDOM was the first and largest randomized control trial of its kind for coronary revascularization in diabetics. Patients with multivessel disease were randomized to either PCI or CABG arms of the treatment group [10,11].

It enrolled 1900 patients from 140 international centers from 2005 through 2010 with 5-year results reported in December 2012. Unlike most of the previous underpowered trials, it was powered to detect a 27% difference in primary end points with 80% power, at a minimum follow-up of 2 years. The therapies were assessed in the background of aggressive optimization of diabetic control (glycated hemoglobin [Hb1Ac]) and risk factors for hypertension and dyslipidemia. A large proportion of patients had triple-vessel disease (83%). Disease complexity was stratified by an objective criterion of SYNTAX scores. The mean SYNTAX score was 26.2 ± 8.6 (mean 5.7 lesions with average 77.6 mm length and high jeopardy scores). The cohorts had significant percentages for hypertension, dyslipidemia, and smoking. Intravenous abciximab was given to PCI patients during procedure and dual antiplatelet agents (aspirin and clopidogrel) were used for a full 1 year with compliance rates >90%. A mean 2.9 ± 0.8 grafts were used in CABG compared to a 3.5 ± 1.4 stents for PCI across all stages. Left internal thoracic artery was used in 94.4%. The cohorts were representatives of the high-risk diabetic populations.

Although the study was sponsored by the National Heart, Lung, and Blood Institute, it was well supported by the medical industry. The stents were provided by Cordis Corporation (Freemont, CA, USA), Johnson & Johnson (Warren, NJ, USA; Taxus – paclitaxel eluting stent), and Boston Scientific Corporation (Natick, MA, USA; sirolimus eluting stent). Eli Lilly (Indianapolis, IN, USA) provided abciximab and an unrestricted research grant; Sanofi-Aventis (Paris, France) and Bristol-Myers Squibb (New York, NY, USA) provided clopidogrel.

2. Limitations of FREEDOM

Historically, major adverse cardiovascular and cerebrovascular event rates (MACCE) have been driven by higher early stroke rates in CABG and higher late revascularization rates in PCI. These factors have remained uncorrected for evaluated therapies in previous trials. The constantly shifting goal posts of improved revascularization and reduced MACCE in PCI have largely been compared to standard on-pump CABG with single internal thoracic artery and vein grafts. This conventional CABG technique has remained unchanged for most surgeons and centers for over three decades now. In the present context in the evaluation of opposing therapies, FREEDOM has the following major limitations:

1. Almost half a decade has already passed since the last FREEDOM patient was recruited. In the meantime, technology for stent designs has relentlessly marched on with the introduction of second-generation drug eluting stents (DESs) with thinner struts and better eluting platforms.

   FREEDOM used paclitaxel and sirolimus DESs, which have been surpassed by second-generation everolimus and zotarolimus stents.

2. At least one internal thoracic artery graft was used in 94.4% in FREEDOM. However, total arterial grafting and bilateral internal thoracic arteries were hardly used. Surgical revascularization has consistently shown superior results with the use of arterial grafts and bilateral internal thoracic arteries in high-risk surgical cohorts.

3. Medical risk factors for progression of coronary artery disease were not adequately controlled. Although a regular communication with treating physicians was maintained for lowering medical risk factors throughout the trial, only 8% actually met prespecified targets.

4. Aspirin and clopidogrel were used for dual antiplatelet therapy. Other combinations with prasugrel and ticagrelor have proven to be superior.

5. Despite the cohorts being representative of high-risk diabetic patients, the trial actually enrolled only 5.8% (1900) of the 32,966 screened patients. The results cannot be extrapolated to ‘real-world’ patients.

6. There was no parallel registry to act as a quality control for the trial.

7. The basis of revascularization was oculo-stenotic anatomical assessment of lesion severities without any physiological assessment of ischemia, even in diffuse multivessel disease with high SYNTAX scores where this is less likely to be reliable.

3. Advances in PCI and CABG

First-generation DES Cypher (sirolimus eluting stent; Cordis Corporation, Johnson & Johnson) and Taxus (paclitaxel eluting stent; Boston Scientific Corporation) were exclusively used in FREEDOM. These stents were associated with high rates of early and late stent thrombosis related to the polymer coating used for drug elusion. Cypher uses polyethylene-co-vinyl acetate and poly n-butyl methacrylate coatings, whereas Taxus uses SIBS (poly styrene-b-isobutylene-b-styrene), copolymer coating. These are non-erodible and their persistent contact with endothelium causes local inflammation, toxicity, and mechanical complications related to stent expansion and mal-apposition. The second-generation stents have superior delivery platforms with thinner struts (Xience V has a strut thickness of 81 µm compared to 132 µm for Taxus and 140 µm for Cypher). These advances together with better balloons reduce the incidence of stent mal-apposition and delayed stent thrombosis.

Both SPIRIT and COMPARE (A Trial of Everolimus Eluting Stents and Paclitaxel Eluting Stents for Coronary Revascularization in Daily Practice) trials showed superiority of second-generation stents over paclitaxel stents [12,13]. In SPIRIT, there were substantial benefits with everolimus eluting stents over paclitaxel stents. Target lesion failure rates were 8.9% vs. 12.5%, p = 0.0002, ischemia-driven target lesion revascularization was 6.0% vs. 8.2%, p = 0.004, stent thrombosis – 0.7% vs. 1.7%, p = 0.003, and major adverse cardiac events were 9.4% vs. 13.0%, p = 0.0002, all in favor of everolimus stents. These benefits even extended to the hard clinical end points of all-cause mortality (3.2% vs. 5.1%, p = 0.003), myocardial infarction (MI; 3.2% vs. 5.1%, p = 0.002), and cardiac death or MI (4.4% vs. 6.3%, p = 0.005). These differences have rarely been seen in stent comparisons. More significantly, these benefits continued to increase in magnitude over a 3-year follow-up in SPIRIT IV, proving the superiority of newer generation stents [10]. Similarly, in COMPARE, everolimus eluting stents were associated with a relative risk reduction of 27% in the primary composite end point of death, MI, and target vessel revascularization (18.4% vs. 25.1%, p = 0.0005) compared with paclitaxel eluting stents. This reduction was mainly due to lower rates of myocardial infarction (7.0% vs. 11.5%, p = 0.001) and target vessel revascularization (7.4% vs. 11.4%, p = 0.003). The rate of stent thrombosis was also lower for everolimus eluting stents at 5 years (3.1% for everolimus stents vs. 5.9% for paclitaxel stents, p = 0.005).

Surgical advances in off-pump, anaortic techniques with complex arterial revascularization have never been evaluated against next-generation everolimus and zotarolimus stents in a randomized control trial.

The pathobiology of diabetic lesions is different and more complex than nondiabetic lesions. Stone et al. showed in SPIRIT IV that improvements in outcomes with newer generation stents in nondiabetic lesions could not be demonstrated in diabetes mellitus [14]. Similarly, the diabetic subgroup study from SORT OUT IV has failed to show any significant differences between everolimus and sirolimus for end points of death and MI despite improvement in revascularization rates [15]. Can these observations be extrapolated to comparisons with surgical revascularization? It would not be long before these advances in PCI catch up with long-term results, for at least conventional CABG.

4. Designs for new trials

For a true comparison of treatment effects between competing modalities, rationalization of trial designs for the CABG arm must now incorporate the use of total arterial grafting techniques, with feasible use of bilateral internal thoracic arteries. This would adjust the concerns for long-term patency of grafts with conventional CABG in older trials. Anaortic off-pump techniques to reduce stroke rates would need to be evaluated separately to adjust for their superior neurological outcomes compared to conventional CABG.

Most historic trials have been underpowered for the treatment effects that they were designed to evaluate. The original FREEDOM protocol was designed to enroll 2400 patients during a 2-year period with a minimum follow-up of 3 years per patient. This was to ensure a power of 85% to detect a relative reduction of 18–23% in 4-year rates of the primary outcome, which were expected to range from 30% to 38% in the less effective study group. In December 2007, the protocol was amended to have a target enrollment of 2058 patients during a 4.25-year period with a minimum of 2.5 years of follow-up to ensure a power of 85% to detect a relative reduction of 24.6% in the rate of the primary outcome, with 1% crossover and loss to follow-up. In April 2009, the protocol was again amended to have a final target enrollment of 1900 patients during a 4.75-year period with the revision of minimum follow-up to 2 years. This was based on an observed aggregate 4-year event rate of 14.85%, with an estimated power of 80% to detect a relative reduction of 27% in the 4-year event rates in the two study groups.

This evolution in the design of FREEDOM represents the perils of running big randomized control trials. There are financial limitations for recruitment of subjects and centers, costs of equipment, data collection and evaluation, follow-ups, and collation of efforts.

To achieve the primary end points in a smaller subset of patients, trials can screen baseline disease characteristics by cardiovascular risk and proteinuria. Cardiovascular risks, for instance, are 6.5-fold (39.9% and 6.3%) for all-cause death and almost 16-fold (18.7% and 1.2%) higher for cardiovascular death in the presence of proteinuria [16]. Any potential differences in benefits accruing from opposing modalities are more likely to be highlighted for these higher cardiovascular risk rates in smaller cohorts.

Concurrent, parallel observational registries can serve as quality controls for the trials. Screened, non-recruited patients from registries can add valuable information to any post hoc analysis. It can explain significant deviations in the study results and overestimation of treatment results that randomized control trials are prone to. It would also allow synthesis of data from the parallel registry for real-world extrapolations of treatment effects.

New trial designs must compare lesion stratified, highest risk subgroups of diabetic patients undergoing advanced surgical techniques (with best graft patency and reduced stroke rates) against newer generation of DESs (with better revascularization rates and probable improvements in survival). This evaluation would have to be in conjunction with intensive medical optimization of insulin-treated diabetes and control of secondary risk factors of hypertension, dyslipidemia, smoking, and obesity.

Coronary lesions are more likely to be diffuse with poor distal coronary runoffs in diabetics. The disease is more likely to be silent with angina equivalent symptoms only. Unlike discreet focal lesions, anatomical oculo-stenotic assessment of diffuse diabetic lesions with poor distal coronary runoff as a basis for revascularization is misleading. Revascularization may still be inadequate despite angiographically functioning grafts after surgery. A physiological basis for revascularization with ischemia assessment is required to target specific territories that are likely to benefit from it. End points, likewise, would need to include completeness of revascularization evaluation and functional ischemia testing.

With increasing harmonization of surgical therapies with percutaneous modalities, comparisons of individual therapies risk becoming increasingly irrelevant in the future. There is likely to be an increasing role for combined revascularization strategies over longer longitudinal follow-ups. With these improvements, hybrid approaches could potentially offer the best of both worlds in complex multivessel disease with surgical revascularization reserved for prognostically important left anterior descending artery lesions [17,18]. Trial designs would further need to be able to identify and evaluate individual lesion-specific rather than modality-specific outcomes.

5. Modifiable risk factors and post-intervention surveillance

Risk score modeling over extended periods from the Framingham Heart study identified modifiable risk factors as significant contributors to comorbidity of cardiovascular disease [19]. FREEDOM protocol recommended guideline-driven targets for lowering medical risk factors. These included smoking cessation, low-density lipoprotein cholesterol < 70 mg/dL, blood pressure < 130/80 mm Hg, and Hb1Ac < 7%. Even in the trial settings, only 8% of FREEDOM patients met all four prespecified treatment targets for systolic blood pressure, cholesterol, smoking cessation, and Hb1Ac at 1 year of follow-up. The role of post-intervention surveillance strategies for control of modifiable risk factors needs to be further evaluated. The FREEDOM trial discussed the Heart team approach consisting of cardiologists and cardiac surgeons. This Heart team must also include diabetologists, nutritionists, cardiac imaging specialists, and noninterventional cardiologists with a well-defined, target-based, post-surveillance strategy, individualized for each patient.

Long-standing insulin-treated diabetes itself remains a major prognostic arbiter of outcomes. FREEDOM answered a lot of unanswered questions from previous trials. The results, however, cannot be extrapolated to present-day modalities. Advancement in therapies has already outpaced these and raised new questions that need to be answered from large, organized, retrospective databases and large meta-analytic studies in addition to carefully designed randomized trials.

Declaration of interest

The author has no relevant affiliations or financial involvement with any organization or entity with a financial interest in or financial conflict with the subject matter or materials discussed in the manuscript. This includes employment, consultancies, honoraria, stock ownership or options, expert testimony, grants or patents received or pending, or royalties.

Additional information

Funding

This paper was not funded.