The power of randomized controlled trials

The best way to treat postoperative sternal wound infection (SWI) after a median sternotomy is the object of many opinions. Treatment with revision, primary closure and often an irrigation-suction system used to be a common method and could be considered ‘conventional’. However, vacuum-assisted closure, also known as negative wound pressure therapy, has been strongly advocated by many. For several years, it has been the recommended and preferred treatment, at least in complicated cases [1, 2]. However, obviously, not all cardiac surgeons are convinced. A number of variations of the conventional method still exist, indisputably with very acceptable published results [3]. Surely many roads may lead to Rome, and different treatments might work equally well, depending on local experience and patient selection. Nevertheless, it is unsatisfactory that the scientific level of evidence behind all these different methods, with few exceptions, is limited to published single-centre experiences and comparisons with historical controls. In daily clinical practise, I therefore have to select the treatment for each patient based mostly on my personal opinion and experience, which is of course severely biased by local traditions rather than based on scientific evidence. Roemer et al. [4] from St Antonius Hospital deserve all the credit for their attempt to construct a scientifically more solid brick of knowledge by designing and properly conducting a prospective randomized controlled trial (RCT). Local collagen gentamicin has been shown to prevent SWI in several RCTs, and 1 early publication indicated that it might also be effective for the treatment of SWI [5–8]. The hypothesis was therefore well worth testing. In conclusion, the results of this specific study were negative because local gentamicin did not result in a shorter length of stay. Retrospectively, with the results of the study on hand, it is easy to say that the prestudy calculation of sample size may be a bit optimistic, resulting in an apparent risk for lack of power and a statistical type-II error. Also, using hospital stay as the primary outcome measure could be debated because this outcome was poorly correlated with the duration of intravenous antibiotics or the time to removal of the last drain, both of which were shorter in the treatment group, although, importantly, the reductions were not statistically significant. A larger sample size and another end point might have led to another conclusion. Maybe. However, a major quality of a correctly conducted RCT, like the present one, is that a negative result may be as important as a positive one because the head-to-head comparison of treatments adds more reliable information compared to other study designs. This information can be important in clinical practise and can also be a basis for further trials. Large, multicentre RCTs have been the indisputable basis for the successful progress of pharmaceutical treatments in medicine for many decades, although new drugs of course first need to be invented and developed by creative scientists in a laboratory before they are tested in large RCTs. In surgery, this tradition is unfortunately less well established. There are brilliant exceptions of course, like the Arterial Revascularization Trial and the Evaluation of XIENCE Versus Coronary Artery Bypass Surgery for Effectiveness of Left Main Revascularization (EXCEL) trial just to mention 2 [9, 10]. I am the first to admit that the quality of cardiac surgery involves many skills that can never be learned or compared by RCTs. Much of the progress in our field has been made by intelligent, manually skilled, innovative, brave individuals. However, prospectively randomizing patients to alternative treatments remains an unsurpassed method for comparing 2 methods, medications or interventions, especially when there are different opinions among experts. Alternative designs with risk adjustments using multivariable regression, propensity scores or similarly advanced statistical techniques can never, ever reach a similar scientific level of evidence. The problem of achieving large enough studies to reach statistical power can be addressed by a multicentre design. In cardiology, multicentre design in RCTs is axiomatic. In cardiac surgery, we could do this more often! Our speciality, with its centralized surgical units and high volumes of standardized procedures, should lend itself unusually well to a multicentre design. There are obstacles to overcome for sure. Conducting multicentre trials requires much time and work, including statistical and scientific advice when writing the protocol, trust and good will among the centres, the discipline to follow a protocol during the study and also an acceptance of the fact that the study might arrive at a solid conclusion contradictory to the hypothesis and one’s previous beliefs or surgical principles. These circumstances make such a study challenging enough, but definitely not unfeasible. Countless numbers of small or large clinical questions in our daily practise could successfully be the subject of this kind of approach. The results would quickly advance our speciality and benefit our patients. Regarding the treatment of SWI, I hope to see more randomized studies, preferably multicentre studies with large populations, and, if the local application of gentamicin is tested again, probably with an adjusted end point. But why not prospectively compare the conventional treatment and the vacuum-assisted closure treatment for a well-defined, properly selected population? The result would really interest me! Conflict of interest: none declared.