This is how it happens:
- First, RCTs are performed to test a new intervention. Studies are meticulously designed to include only patients who are expected to benefit the most. A list of exclusion criteria avoids patients who might not benefit.
- The RCTs show benefit, leading to use of the intervention. At first, this is limited to patients resembling those studied in the RCTs.
- It remains unclear whether some patients excluded from the initial RCTs might benefit also. However, nobody invests the effort to rigorously evaluate this.
- Years pass. The intervention gains widespread acceptance as the standard of care. Details of the initial RCTs are forgotten. The intervention is applied broadly, including among patients who were initially excluded from the RCTs.
- General consensus forms that it would be unethical to perform an RCT testing the intervention among patients who were initially excluded from RCTs. It becomes impossible to scientifically test broad application of the intervention (which is now dogma).
To some extent, generalization is inevitable. RCTs will always be performed on selected patient groups. It is impossible to perform RCTs including every type of patient. Clinical judgment will always be needed to fill in the gaps. However, some factors exacerbate the problem:
- Investigators are rewarded for “positive” studies showing the benefit of a new intervention. There is less glory involved in going back to parse out whether an intervention works in specific patient subtypes.
- Industry has little motivation to study patients who might not benefit, since this could lead to embarrassing negative studies. Even if such a study were positive, this wouldn't be very profitable because most patients are already using the intervention anyway (an example of investigation bias).
- Specialists are naturally biased to believe that their interventions are beneficial (1).
Overview of percutaneous coronary intervention (PCI) in patients with remote CABG
Most of the grafts in a CABG are constructed from saphenous veins. These veins undergo accelerated atherosclerosis, with ~40% becoming occluded within ten years. Thus, myocardial ischemia remains a major problem in patients after CABG. For example, post-CABG patients account for ~15% of all patients undergoing PCI (Scarsini 2016). Stenting procedures may be performed on native coronary arteries or surgically placed grafts.
There are many reasons that PCI could be less beneficial among patients who have a history of CABG:
- Coronary anatomy is more complex, making the procedure more difficult overall. For example, patients may have numerous stenoses in several vessels without a definite culprit lesion.
- Saphenous vein grafts tend to develop bulky, friable plaques at risk of distal embolization. When intervened upon, this may cause a “no reflow” phenomenon, wherein the occlusion is cleared but embolized material causes severe downstream occlusions (2).
- Saphenous vein grafts are subject to accelerated atherosclerosis. Thus, even if a single stenosis can be opened temporarily, it is unclear whether this improves the long-term function of the entire graft.
- Interventions on native coronary arteries may be difficult for a variety of reasons (e.g. severe calcification, challenging location, or chronic total occlusion). Lack of amenable targets for PCI may have played a role in the original decision to perform CABG surgery instead of PCI (Escaned 2012).
Repeat myocardial revascularization procedures are markedly different from de novo interventions, with increased procedural risk and technical-demanding complexity –Scarsini 2016
Evidence: Foundational studies supporting PCI in acute coronary syndrome.
Below is the classic meta-analysis of RCTs testing early PCI for acute coronary syndrome (3):
- TIMI-IIIB, FRISC II, and RITA 3: Excluded patients with prior CABG
- VANQUISH: Included patients with remote CABG (>3 mos prior)
- TACTICS-TIMI 18: Included patients with remote CABG (>6 mos prior)
- ICTUS: Included CABG patients
Studies in this meta-analysis were heterogeneous (4). The benefit from PCI is driven by studies that exclude CABG patients:
- (#1) RCTs excluding CABG patients show benefit from PCI.
- (#2) RCTs including CABG patients show no benefit from PCI.
Now, RCTs that include CABG patients contain both patients who haven't had a CABG and patients who have had a CABG. If we believe that patients who haven't had a CABG are benefitting from PCI (#1), then the only way that mixed studies could be neutral (#2) is if PCI is hurting patients who have had a CABG:
SAVED trial: Stent placement compared with balloon angioplasty for obstructed coronary bypass grafts (NEJM 1997)
This was a prospective, multicenter RCT comparing angioplasty alone vs. bare metal stent among 220 patients with saphenous vein graft stenosis. Stenting improved graft patency at the end of the procedure, but improvements mostly disappeared during the following six months (figure below). There was no difference in the primary endpoint of restenosis during the first six months (37% within stent group vs. 46% within angioplasty group, p=0.24).
Clinical outcomes were similar among groups, with no differences in death or myocardial infarction (table below). There was an increase in target-lesion revascularization among patients who received angioplasty (likely driven by routine unblinded follow-up catheterization after six months).
Overall this study proves that stenting of saphenous vein grafts is possible and does improve stenotic lesions. However, it is unclear whether this leads to a sustained improvement in graft function or clinical benefit.
More recent evidence about PCI after CABG?
The above studies constitute the foundational evidence upon which PCI is based. However, they are outdated. Would newer techniques would perform better?
Unfortunately, no recent RCTs have tested the role of PCI in patients with prior CABG (compared to medical therapy). Drug-eluting stents don’t appear to be a panacea for intervention on saphenous vein grafts, with one RCT finding increased mortality among patients receiving drug-eluting stents compared to bare metal stents (Vermeersch 2007).
Even if saphenous vein graft PCI is feasible, it is risk-prone in terms of high rates of periprocedural adverse events, immediate-term restenosis, and progression of disease outside the treatment segment. – Dash 2014
What do the guidelines say about PCI for NSTEMI with a history of CABG?
Shown above are the full text of the 2014 NSTEMI guidelines and the 2011 PCI guidelines. These guidelines are complicated. The green text seems to suggest an early-interventional strategy. However, the yellow text seems to suggest a more nuanced approach:
- Patients with extensive anterior ischemia (e.g. occluded LIMA to LAD graft) may benefit from PCI.
- Patients with ischemia in other locations are unlikely to derive a survival benefit from PCI.
Bottom line: How can we apply this information in practice?
Overall, the AHA guidelines may provide a rough blueprint of how to approach these patients:
- Start with an assessment of the patient's baseline anatomy and the myocardium at risk (based on operative reports, prior EKG and echo, current EKG and echo, and clinical presentation).
- Patients with acute anterior ischemia may be most likely to benefit from PCI. Alternatively, patients with ischemia involving other territories are less likely to benefit from PCI.
- Final treatment decisions will involve additional considerations including patient preferences, risk of various complications (e.g. renal failure), and multidisciplinary consultation (with cardiology & cardiothoracic surgery).
Many patients with prior CABG probably benefit from catheterization and repeat revascularization (PCI or a repeat CABG). However, this cannot be assumed to be universally true. In particular, patients with smaller infarcts and advanced renal failure could be harmed.
- Performing PCI in a patient with a remote CABG is technically harder and associated with worse outcomes compared to PCI in a patient without prior CABG.
- The evidence proving benefit of PCI in NSTEMI is driven by studies that excluded patients with prior CABG. When examining large RCTs that included CABG patients, an early-invasive PCI strategy didn't improve outcomes.
- It remains unknown how beneficial modern PCI techniques might be among patients who have had a CABG.
- Decisions regarding PCI should be made carefully, taking into account the patient’s anatomy, the amount of myocardium at risk, procedural risks, and patient preferences.
- Investigation bias: The freakonomics of when industry choses to sponser a clinical trial (PulmCrit)
- Over-diagnosis and over-treatment of MI among critically ill patients (PulmCrit)
- A Case of Identity Part One & Part Two (EM Nerd explores antiplatlet therapy in MI)
- 48-hour heparin infusion is nonbeneficial for noninvasive therapy of NSTEMI (PulmCrit)
- This isn’t intended to demean specialists, but rather it is simply human nature. We all want to think that what we do is the best, indeed we must believe this. For example, in the treatment of prostate cancer, radiation oncologists believe that radiation is superior while urologists believe that surgery is superior (Kim 2014).
- This risk might be reduced by the use of endovascular protection devices, but such devices remain underutilized and it is unclear how well they work (Safian 2016).
- This RCT is cited by the AHA/ACC 2014 guidelines as providing evidence to support the use of PCI in NSTEMI. It was coauthored with Eugene Brauenwald. Classic.
- Ideally, individual studies in a meta-analysis should have generally consistent. The studies in this meta-analysis are heterogeneous and rather inconsistent with each other (e.g. the results of FRISC II are incompatible with VANQUISH). This questions the validity of performing a meta-analysis in the first place.
- This is not a figure from the JAMA 2008 meta-analysis. I created this on powerpoint, with the use of vasserstats to analyze the odds ratio of the combined group. This probably isn't technically kosher, but the results are pretty clear regardless of how you do the statistics.
Image credits: Minyans.