Article Text
Statistics from Altmetric.com
- cardiac surgery
- coronary artery disease surgery
- chronic coronary disease
- quality and outcomes of care
- meta-analysis
The art of medicine might well be described as the ability to make the right decisions based on incomplete information. Try as we might to provide ‘evidence–based’ care, the more scientifically candid among us will confess that the ‘definitive study’ on any given topic never has been (nor ever will be) performed. Indeed, a careful 2009 review of the vaunted American College of Cardiology (ACC)/American Heart Association (AHA) guidelines available at that time demonstrated that, in total, less than 10% of recommendations were based on level of evidence (LOE) A (randomised controlled trials (RCTs) or meta-analyses of RCTs), whereas more than half relied on LOE C (studies without controls or expert opinion).1
Despite remarkable and ongoing advances in percutaneous interventions, both meta-analyses of RCTs2 and longitudinal analyses of large registry data3 continue to document the comparative long-term survival benefit of surgical revascularisation for patients with extensive coronary artery disease. Given the now-classic demonstration of the improved survival with the use of the left internal thoracic (or mammary) artery (LITA) rather than saphenous vein grafts (SVGs) for bypassing a diseased left anterior descending coronary artery (LAD), the subsequent demonstration of improved patency of ITA versus SVGs and the clear correlation with improved survival4 would seem to make the rationale for the use of two rather than one ITA somewhat compelling. In fact, meta-analyses addressing this very issue have demonstrated a superior long-term survival benefit for bilateral versus single ITA grafting in patients with extensive coronary artery disease. To these reviews, the meta-analysis presented by Buttar et al herein adds information regarding favourable perioperative and long-term clinical parameters that also appear to favour bilateral internal thoracic artery (BITA) versus single internal thoracic artery (SITA) grafting.5
Almost regardless of this ostensible avalanche of evidence, clinical response has been tepid at best. Both ACC/AHA and European guidelines provide less than enthusiastic Class IIa recommendation for BITA grafting (reasonable; additional studies needed) LOE B (non-randomised studies).6 BITA grafting is used in a meagre 5.4% of coronary artery bypass grafting (CABG) cases reported in the Society of Thoracic Surgeons adult cardiac surgery database, which currently represents 90% of centres performing adult cardiac surgery in the USA. Ironically, the LOE supporting the single LITA graft to the LAD, which is now accepted as a National Quality Forum metric of quality in CABG surgery is no more compelling (level B) than that in favour of BITA versus SITA grafting.
Although there is no shortage of explanations for this ostensible discrepancy between evidence and practice, a deeper dive into the evidence may provide a more sober perspective. Classic understanding would elevate the RCT to the ‘gold standard’ in scientific inquiry. To date, the only adequately powered RCT to address the issue of BITA versus SITA grafting is the Arterial Revascularisation Trial (ART), whose 5-year outcomes were recently reported to show absolutely no difference in any clinical outcome at 5 years, save for an increased risk of sternal wound infection with BITA group.7 Although these findings may be concerning for the BITA enthusiasts, it should be remembered that this trial was powered to demonstrate survival differences at 10 years. Although not a uniform finding, careful 20-year follow-up of BITA versus SITA grafting from the Cleveland Clinic demonstrated nearly overlapping survival curves at 5 years with significant divergence in favour of BITA grafting thereafter.
Even more compelling for the honest seeker of truth is the fact that all that glitters is not gold. RCTs, in order to refine the study question, necessarily tailor their inclusion and exclusion criteria to enhance the ‘signal to noise’ ratio of confounding factors. Furthermore, due to the appropriate rigours of informed consent, it is not uncommon in the cardiac realm for studies to enrol only 7%–10% of potentially eligible patients.
Therefore, even though by controlling for both identified and unforeseen potential confounding variables, the study may have considerable internal validity; there has been an a priori patient selection that may severely limit the applicability of the findings to the broad range of clinical presentations that the authors are seeking to address. Moreover, there are challenges in analysing RCTs specific to surgery. Unlike drugs, which have a defined chemical composition, a measurable pharmacokinetic profile and regulated dosage, surgical procedures can vary considerably from one operator to the next. Were the ITA grafts harvested as a pedicle or in a skeletonised fashion? Were all grafts in situ, or were ‘free’ ITAs used? If the latter, was the graft sewn proximally to the aorta or used as a ‘Y’- or ‘T’-graft arising from another ITA graft? Was the LITA used exclusively for bypassing the LAD, or was the LAD grafted with a right ITA preserving the LITA for the lateral wall? If the LITA was used for the LAD, was the second ITA used for the left or right circulation? How many grafts were performed and how complete was the revascularisation? Was the procedure performed with the aide of cardiopulmonary bypass or ‘off-pump?’ Perhaps surgeons obsess over these things, but the distinctions can sometimes be decisive.
Beyond surgical variability, there is a more subtle but perhaps even more compelling elephant in the room. In order to study any surgical technique, investigators must find those surgeons who are skilled in that approach. The reason why they have become skilled in that approach is their clear belief that it provides the best care for their patients. It is therefore highly unlikely that the most skilled operators share true equipoise regarding the study question. Therefore, either they will not participate in an RCT, or they will enter only those patients that they are least concerned regarding their benefit from the procedure. Therefore, only those patients least likely to benefit will be randomised, and the study will be biased towards the null. Without careful analysis of all the patients operated on but not entered into the study (an exercise rarely ever performed or even accessible), clear understanding of negative results cannot be achieved.
Salvation in the retrospective analysis of large clinical registries? Perhaps as a representation of the ‘real world’ practice of medicine, these data can be more instructive? Here the problem of selective bias is the reverse-all the decisions, including those of interest, have already been made. Selection bias is inevitable. Robust statistical modelling with multivariable analyses and, more recently with propensity score techniques, have attempted to level the playing field. However, such approaches can only control for known variables present in the data. Despite the most sophisticated approaches, the reality remains that given two patients with exact the same ostensible risk profile, the surgeon chose to perform BITA grafting on one and SITA grafting on the other. Was it some gut feeling regarding the patient frailty? Concern regarding ventricular hypertrophy or pulmonary reserve? As we have no way of knowing, we may be left with the conclusion that, despite the statistical gymnastics, the reason why BITA patients tend to live longer is that good surgeons are credible judges of patient longevity and select that technique for those patients whom they perceive are more likely to live longer. Despite the ‘real world’ weight of meta-analysis ostensibly free from publication bias as so admirably performed by Buttar et al,5 there can arise some surprising findings. Virtually every study to date, including the ART trial, has reported equivalent perioperative mortality with SITA versus BITA grafting. Yet recent report from the STS database has shown a significantly increased perioperative mortality with BITA versus SITA grafting.8 Whether this relates to operator experience, patient selection or some other factors will require further investigation. Future research, whether retrospective or prospective, will hopefully help us to better define those patients who may benefit from BITA grafting and those who clearly do not.
‘All we want are the facts.’ Unfortunately, in medicine, the facts cannot be excised from their context, and the honest clinician must keep an open mind to their limitations while, at the same time, seeking to provide the best care for his patients. That said, this surgeon, should he need coronary revascularisation today, would chose a surgeon accomplished in the art of BITA grafting.
Footnotes
Competing interests None declared.
Provenance and peer review Commissioned; internally peer reviewed.