| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
Eur J Cardiothorac Surg 2004;25:243-245
© 2004 Elsevier Science NL
Review |
Département Cardio-Vasculaire, L'Institut Mutualiste Montsouris, 42, Boulevard Jourdan, 75014 Paris, France
* Tel.: +33-1-56-61-62-63; fax: +33-1-56-61-65-23
In a somewhat provocative article published in the present issue of the European Journal of Cardio-Thoracic Surgery, Myrmel, Lai and Miller question the quality of the publications concerning the treatment of acute aortic dissections through an apparently simple interrogation: can the principles of evidence-based medicine be applied to the treatment of aortic dissections? [1]
During a thorough search of the literature published during the last two decades, the authors have tried to find articles that could answer three questions by using the comparative, prospective, randomized, controlled methodology applied in evidence-based medicine:
Obviously their search was very disappointing and they have found no article following those principles and allowing any unequivocal conclusion concerning those questions. They therefore conclude that, because the great majority of articles report observational studies, most of their recommendations for treatment options are weak and that no evidence exists favouring the use of an open distal anastomosis and the elimination of the distal false lumen during surgery of acute type A dissections, or medical therapy over surgical treatment in non-complicated type B dissections.
This article is rather intriguing not to say puzzling. Not only because it bears the signature of one prestigious surgeon with a large experience of acute aortic dissections and who has participated largely in the relevant literature, but also, more importantly, because it raises fundamental questions about the analyses of our therapeutic methods and the reports of those analyses in that particularly difficult surgical matter.
The principles of evidence-based medicine have been elaborated and designed during the late 1940s and the early 1950s to compare objectively different pharmaceutical treatments (or treatment against refraining) on a large scale and during a sufficiently long period of time. They have become the absolute gold standard in epidemiology and analysis of the effects of drug therapies. They are based on pre-designed prospective, randomized controlled trials with severe inclusion and exclusion criteria and strict end-points. They generally require a great number of patients in each arm of the comparison and often a long period of survey. They overall require that the population under scrutiny be homogeneous.
Obviously those demanding elements make their application to the treatment of acute aortic dissections quite impossible.
Looking at the literature, one may observe that most groups reporting large experiences in acute type A dissection operate an average of 10–15 cases a year. Even in those active and dedicated centres it would probably take several decades to obtain enough patients to significantly compare two modes of treatment. In addition during the long period of study, it is quite probable that many anaesthetic, operative and intensive care management techniques would have changed and would have introduced much bias into the study.
The authors point out accurately that "coronary surgery is one of the best documented treatments in medicine". This is not surprising since coronary patients are numerous, they present under a limited number of clinical patterns, each of which being perfectly characterized. The majority of the randomized, controlled trials are conducted in patients in stable condition or at least in patients whose life is not threatened within hours. Therefore inclusion into or exclusion from the study is easy. Besides, because the compared treatments generally have alternatives, the patients' life is seldom threatened by the randomization.
Conversely patients suffering from acute aortic dissection are few. They present with a broad spectrum of clinical conditions, ranging from the perfectly stable, conscious uncomplicated patient to the highly complicated or even moribund patient suffering from severe tamponnade, cardiogenic shock, stroke, coma, distal malperfusion, etc. or, more often, a combination of those features. Their condition requires immediate active treatment. There is generally no time for selection, randomization, assessment of the inclusion criteria and decision for the corresponding therapeutic solution. How can those patients be included into pre-determined randomized groups? Even if subgroups are defined, they will be obviously small implying low statistical significance and certainly a weak comparison.
All those points are accurately addressed and discussed by Myrmel, Lai and Miller. And they superbly summarize the question by quoting E.S. Crawford stating in 1992 "Ideally, a randomized trial should be conducted to determine the indications for including the arch in all replacement operations for acute De Bakey type I aortic dissections. However, because of the variability that can be expected among patients with acute dissections involving the ascending aorta, a large group of patients would be required, and a very long-term follow-up would be necessary to determine the probability of freedom from unfavourable outcome events. Such a study will probably not be done" [2]. Everything seems to be said. Not quite.
For, beyond the difficulties or impossibility in applying prospective randomized, controlled trials to acute aortic dissection treatment, the article by Myrmel et al. raises other important questions concerning the nature of the treatment itself, its performance, the validity of some statistical methods and the so-called ‘low-rank’ observational studies based on historical controls.
One of the great misleading ideas in assessing ‘scientifically’ our results is that surgery is a science. It is not. It is a technological structure managed, performed and developed daily by ‘technicians’. Of course, like most modern technologies, it is based on scientific grounds and uses science. But this part of the activity belongs mostly to the laboratory whereas the operating theatre gives way only to technological performances. Improvements in technological systems are seldom very spectacular. They come progressively, step by step, and become obvious after a certain delay. But the improvements are then generally evident and they do not need any assessment by prospective randomized, controlled trials to be widely accepted. Who can deny, as an example, that electronically controlled disc brakes in modern automobiles are much safer than the drum brakes used in the automobiles of our childhood? And yet there never was any randomized trial to compare both technologies. The same observation may be made in surgery. To come back to the treatment of aortic dissections, is it really necessary to have prospective randomized, controlled trials to assert with a certain degree of confidence that cannulation of the right axillary artery is a real progress and has reduced dramatically the number of cerebral and visceral malperfusions as well as distension and ruptures of the false channel, as compared to femoral cannulation. But those observations need a certain degree of experience and require that historical controls are carefully documented.
This calls into play an important and yet often ignored issue: the human factor. In epidemiologic or medical prospective, randomized, controlled trials, inclusion and treatment implementation can generally be done by any investigator and the outcome does not depend on the competence or experience of the investigators. The analysis of the results is even generally made in a blind fashion. Conversely, operations for acute aortic dissection are probably among the most difficult, demanding, sometimes disappointing procedures that a cardio-vascular surgeon has to perform. Their result depends on a variety of factors but is undoubtedly experience-dependent. In this regard, we certainly share the opinion of M. De Leval stating that "Surgical performance is dependent on integrated cognitive, physical and affective skills. As in any other skilled handicraft, increased experience leads to better technical performance. Akin to the musician, the surgeon needs practice to maintain peak performance and the more critical the procedure the more important it is that this practice is provided" [3]. Nevertheless, because of the emergent necessity of the treatment, acute dissections are too often operated on in non-optimal conditions (e.g. in the middle of the night or during week-ends, not always with the anaesthetists and nurses usually dedicated to cardiac surgery) and sometimes by the least experienced surgeons who, because of their younger age and hierarchical position, get the larger part of the burden of being on duty. This does not mean that older surgeons are always better than younger ones. As stated by M. De Leval "differences in cognitive performance associated with aging are small relative to the total range of individual differences" [3] and we all know old surgeons who are unskilful and clumsy and young ones who are quite brilliant. But it certainly means that when a surgeon has operated on, say, 100 acute dissections he certainly is more able to appreciate the clinical and pathological features, recognize the threatening elements, chose the most appropriate technical strategy and face unforeseen, sudden complications or technical difficulties. Either from a human or a scientific standpoint, the quality of the operator can hardly be measured and be entered prospectively into a randomized, controlled trial. It can be appreciated only by retrospectively comparing the results of the various operators, assuming that all the other criteria and data are identical. Why should we deny to experienced surgeons the possibility of passing on their experience to the community through observational studies?
In 1997, Sackett et al. stated that "If you find that a study was not randomized, we'd suggest that you stop reading and go to the next article" [4]. This statement seems slightly exaggerated. If we accept the idea that only prospective randomized, controlled trials are able to assess different techniques or modes of treatment and that observational studies are always ‘low-ranked’, most of the surgical literature and obviously all the articles dealing with the treatment of acute dissection (including the excellent ones coming from Stanford) would have to be rejected. It would mean that all those studies are poorly designed and intellectually feeble. In addition, although a great majority of those articles appear in famed international journals, it would mean that the process of selection is weak and that the peer-review system has failed because of the incompetence of the expert reviewers and editors. And that the whole surgical information system has to be reformed. Two articles dealing with this problem appeared in the New England Journal of Medicine in 2000 [5,6]. After a wide scrutiny of the literature the authors of both articles were able to find a certain number of topics that had been studied either through observational studies or randomized, controlled trials. Interestingly enough they could demonstrate that the results in most areas were not very different with both methods. The authors conclude their respective articles as follows: "The fundamental criticism of observational studies is that unrecognized confounding factors may distort the results. According to the conventional wisdom, this distortion is sufficiently common and unpredictable that observational studies are not reliable... Our results suggest that observational studies do provide valid information" [5] and "The popular belief that only randomized, controlled trials produce trustworthy results and that all observational studies are misleading does a disservice to patient care, clinical investigation, and the education of health care professionals." [6]. Indeed. But, too often, the main criticism opposed to many observational studies is that the statistical tools used for comparing treatment methods or identifying risk factors are inappropriate, irrelevant or misleading and that they eventually mix ‘apples and oranges’. This can easily be corrected by scrupulously and carefully designing observational studies, selecting outcome criteria and risk or prognostic factors similar to those included in randomized, controlled trials and, overall, in using appropriate statistical methods. In this regard turning to the help of a specialist in medical statistics is highly advisable.
But, on the other hand, another somewhat erroneous, yet widely accepted notion is that statistical models and analyses always reflect the reality and must be preferred to actual figures. We would certainly not challenge the importance and the absolute necessity of statistics in the analyses of medical and surgical results. We just want to point out that, in some instances, too much statistics may distort the reality and eventually result in erroneous or misleading conclusions. In this regard, and with all due respect to the memory of one of the greatest cardio-vascular surgeons in history, the article by Crawford et al. quoted by Myrmel et al. is a good example. This article was intended to prove, from a limited cohort of 82 acute type A dissections operated between 1968 and 1989, that replacing the arch in addition to the ascending aorta is unnecessary and dangerous. The overall actual mortality was 21%. By applying a sophisticated mathematical model to their results the authors calculated that the predicted 30-days mortality 2 years after the end of the study would be 3% in the ascending aorta and 16% in the arch subgroups. More than a decade later we know for sure that replacing the arch is not a risk-factor of death and may even be beneficial [7]. And we may observe that no surgical group has yet reported hospital mortality rates as low as 3% on a sufficient period of observation, even though the techniques have improved. The truth rests probably in the middle way, as usual: good statistical analysis but not too many mathematical models masking the actual data and figures or making the paper impossible to understand by non-statisticians.
Reality is difficult to apprehend. Each of us has a much reduced experience as compared to the number of patients being treated for acute dissection every year in the world. The only way to know better whether what we do is sound and sensible or not would probably consist in gathering the greatest possible number of patients and the data concerning their clinical condition, the mode of diagnosis, the surgical and perfusion techniques and at least the 30-days mortality and morbidity, in independent national or international data bases. Only this could allow the community to exactly analyse this heterogeneous population and the numerous methods of treatment. But it is only a far dream.
A young man enters a grocery store and says, "Good morning. I would like to buy one litre of potatoes".
"My dear boy", says the grocer, "you cannot buy one litre of potatoes but you can have a kilogram of potatoes".
"All right", says the client, "I would like to buy a kilogram of potatoes"
"Fine", says the grocer. "Do you have a bottle to put them in"?
This small joke illustrates the dilemma concerning most surgical reports. Akin to the grocer we know what is right but we are ignorant about the tools or the means necessary to properly demonstrate our point. Like in the operating room we certainly cannot use and rely on one single instrument (in this case, the randomized, controlled studies). There is no doubt that observational studies must go on and be published as they are used to identify risk factors and prognostic indicators, to promote more appropriate techniques and because trying to replace them systematically by randomized, controlled studies would be unethical in many situations. Considering the wide spectrum of situations and methods, information should be adapted to this spectrum and delivered under many forms provided the quality of the data is ascertained and, as stated recently by Sade and McKneally, "Journals (can) rely on the integrity and honesty of investigators in providing accurate information to their readers, for these character traits are the life blood of the entire corpus of cardio-thoracic research and the clinical practice that depend on it" [8].
 |
References |
|---|
 Top References |
|---|
Related Article
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
| ANN THORAC SURG | ASIAN CARDIOVASC THORAC ANN | EUR J CARDIOTHORAC SURG |
| J THORAC CARDIOVASC SURG | ICVTS | ALL CTSNet JOURNALS |
