Free
Correspondence  |   April 1997
How Can We Demonstrate that New Developments in Anesthesia Are of Real Clinical Importance? 
Author Notes
  • Department of Public Health, University of Western Australia, Nedlands 6907, Western Australia (Rigg, Jamrozik).
  • Clinical Trial Service Unit, Radcliffe Infirmary, Oxford, UK (Clarke).
Article Information
Correspondence
Correspondence   |   April 1997
How Can We Demonstrate that New Developments in Anesthesia Are of Real Clinical Importance? 
Anesthesiology 4 1997, Vol.86, 1008-1110. doi:
Anesthesiology 4 1997, Vol.86, 1008-1110. doi:
To the Editor:-Recently, in their editorial discussion of the paper by Bode et al. [1 ] Go and Browner [2 ] concluded that additional randomized controlled trials comparing regional and general anesthesia are unlikely to be useful. However, even though surgery and anesthesia are so much safer now than 50 years ago, this is no reason to conclude that further significant improvements cannot be made. The problem is how to identify which new developments in anesthesia (and surgery) are of clinical importance. The commentary from Go and Browner does ask two key questions in this regard, namely, the extent to which one can rely on the evidence provided by clinical studies and the extent to which one can generalize from studies performed in one field of surgical and anesthetic practice to other fields.
Individual Trials 
It is not appreciated widely just how large randomized clinical trials need to be to detect reliably moderate, but important, differences between treatments. [3 ] For example, Bode et al. state that their study sought to evaluate whether the type of anesthesia has an important influence on perioperative cardiac outcomes in patients undergoing peripheral vascular surgery. [2 ] However, with an overall mortality rate of just 3.1% in their cohort of 423 patients, they would actually have needed more than 24,000 patients to have an 80% chance of detecting even a 50% reduction in mortality rate. Thus, they rightly acknowledge the inability of their own experiment to answer the question they had posed.
This highlights the major dilemma facing any researcher seeking to evaluate whether changing anesthetic techniques can change patient outcomes. Major surgery for “high risk” patients is now remarkably safe based on the rates of major complications in studies such as that of Bode et al. [1 ] This has important implications for the design of randomized controlled trials to evaluate the impact of any aspect of anesthetic technique on morbidity and mortality rates. As Yusuf et al. [3 ] noted, if a large benefit was associated with a particular clinical maneuver that had been widely used for many years, the evidence for this benefit would be clear, and this would be widely agreed and accepted by clinicians. However, many clinical maneuvers will differ only moderately, and we agree with Go and Browner that the effect of epidural block on surgical outcome is likely to be moderate. Such effects can be important from a clinical and an economic perspective, [3,4 ] but they will not be detected reliably without large randomizd trials.
A second point is a consequence of the low morbidity rate in unselected patients. To have the necessary statistical power to avoid the false conclusion of no benefit, a study should either be extremely large, or it should selectively recruit patients at high risk of complications. Both strategies require a multicenter approach to enroll sufficient patients in a reasonable time, but the selective strategy requires a trial of more achievable size. The Bode et al. study [1 ] can be used to illustrate this. For the sake of simplicity of calculation, we have rounded their observed overall mortality rate from 3.1% to 3% and assumed that each 100 of their patients could be divided into a high-risk group of 10 and a low risk group of 90 with mortality risks as given in Table 1.
Table 1. 
Image not available
Table 1. 
×
We also can estimate, similarly, cardiovascular morbidity from their results, as presented in Table 2.
Table 2. 
Image not available
Table 2. 
×
In the high-risk group, a trial of 945 patients per treatment group would be needed to have an 80% chance of detecting a decrease in mortality rate from 20% to 15%. By concentrating on patients at high risk, trials of modest (1,890) rather than enormous size (> 24,000) are required to have a reasonable chance of identifying modifications to anesthetic and analgesic techniques that could substantially improve outcome for all patients. This assumes that the results in the high-risk group also would be found in the low-risk group. Although we would not expect the benefit to be exactly the same in both groups, there is no good reason to believe that one group would benefit and one would not. We would expect that if there was a benefit, it would exist for both groups, but it may be different in each group.
Such moderate improvements are worthwhile clinically and may be cost-effective if the new or alternative techniques cost no more, or only a little more, than those presently in use. However, to be reasonably sure of identifying benefits of this magnitude in a trial of manageable size and duration, a multicenter study will be required because no single institution is likely to have a sufficient number of high-risk patients.
In practical terms, if 10% of the patients seen by Bode et al. were at high risk and if a new multicenter study could maintain their recruitment fraction of 60%, 11 institutions of the size of the New England Deaconess Hospital could complete the study outlined previously in the same time as Bode et al. took to accrue 423 unselected patients. This goal is realistic and achievable.
Generalizing from Individual Studies 
Bode et al. [1 ] studied cardiac outcomes in patients undergoing peripheral vascular surgery. This represents a narrow focus of outcomes in a specific, but nevertheless important, group of patients. They also studied intraoperative epidural block only and excluded postoperative epidural analgesia. It is recognized that first, postoperative epidural analgesia may be more important than intraoperative epidural block in improving outcome, and secondly, patients having peripheral vascular surgery in the lower limbs do not experience the same intensity of postoperative pain (or surgical traumatic stress) as patients having major abdominal aortic procedures. [1,5–9 ] These observations suggest that the conclusions of Go and Browner may have little validity outside of the field of peripheral vascular surgery. [10,11 ] Their opinion should not be extended uncritically beyond the narrow confines of the investigation by Bode et al. [1 ] to the larger pool of patients at high risk of complications. In such patients, there already is some evidence that in the setting of complex surgery, epidural block may reduce the incidence of adverse events. [4–6,8–11 ] Recent correspondence concerning the report of Bode et al. supports our view of the importance of postoperative epidural block in improving outcome and not extending their conclusions beyond the narrow subset of patients they studied. [10,11 ]
There are three other reasons to question the conclusions of Go and Browner. First, the totality of evidence that they present is small [2 ]; a much larger body of research was needed to provide convincing evidence for the important, but moderate, clinical benefits of Tamoxifen in the management of breast cancer, [12 ] and thrombolytic therapy after myocardial infarction. [13 ] Secondly, it is based solely on published trials, and the addition of any unpublished trials that may exist may lead to different conclusions. [3 ] Thirdly, and most importantly, the evidence cited by Go and Browner [6,14–16 ] actually indicates the existence of genuine uncertainty about the merits or otherwise of epidural block and, therefore, represents a strong argument for an adequately sized, multicenter randomized trial in high-risk patients. This view also is supported by recent correspondence. [10,11 ] The conclusion that no further trials are needed is premature, and this is well illustrated by an evaluation of a series of trials reporting the effects of long-term beta blockade. [3 ] Twenty-one of 24 randomized trials failed to achieve conventional levels of significance because each was too small on its own. Reliable assessment of moderate effects (15–20% improvement in the outcome of myocardial infarction) attributable to beta blockade required randomization of 10,000–20,000 patients with a good prognosis or a somewhat smaller study of patients with a poor prognosis. [3 ] The important consideration in the number of randomized patients needed is not the number of patients, but rather the number and rate of the outcomes to be analyzed.
Conclusions 
Improvements in anesthetic and surgical techniques that already have occurred mean that any improvement from a single future development probably will be moderate rather than substantial. [4 ] However, the aggregate benefit of such a development could well be substantial were it to be applied widely in the course of a procedure that is performed frequently. To detect such moderate improvements reliably requires either large randomized studies of unselected patients or slightly smaller studies that are restricted to patients at highest risk of adverse outcomes. Either approach will require multicenter collaboration.
Real improvements in anesthetic techniques will be most easily identified in trials that involve only patients undergoing major surgery. In addition, a full range of clinically significant complications should be included as endpoints to increase the statistical power of the study and also its statistical reliability by not concentrating on just one subgroup or endpoint.
Failure to appreciate and act on these principles may lead to experimental nihilism and to significant delays in identifying clinically important advances in anesthesiology.
John R.A. Rigg, M.B.B.S., F.A.N.Z.C.A.
Konrad Jamrozik, M.B.B.S., D.Phil., F.A.F.P.H.M.
Department of Public Health; University of Western Australia
Nedlands 6907; Western Australia
Michael Clarke, D.Phil.
Clinical Trial Service Unit; Radcliffe Infirmary
Oxford, UK
(Accepted for publication February 3, 1997.)
References 
References 
Bode RH Jr, Lewis KP, Zarich SW, Pierce ET, Roberts M, Kowalchuk GJ, Satwicz PR, Gibbons GW, Hunter JA, Espanola CC, Nesto RW: Cardiac outcome after peripheral vascular surgery: Comparison of general and regional anesthesia. Anesthesiology 1996; 84:3-13.
Go AS, Browner WS: Cardiac outcomes after general or regional anesthesia (editorial). Anesthesiology 1996; 84:1-2.
Yusuf S, Collins C, Peto R: Why do we need some large simple randomized trials? Stat Med 1984; 3:409-20.
Alpert CC, Conroy JM, Roy RC: Anesthesia and perioperative medicine. A department of anesthesiology changes its name. Anesthesiology 1996; 84:712-5.
Yeager MP, Glass DD, Neff RK: Epidural anesthesia and analgesia in high risk surgical patients. Anesthesiology 1987; 66:729-36.
Tuman KJ, McCarthy RJ, March RJ, DeLaria GA, Patel RV, Ivankovich AD: Effects of epidural anesthesia and analgesia on coagulation and outcome after major vascular surgery. Anesth Analg 1991; 73:696-704.
Baron JF, Bertrand M, Barre E, Godet G, Mundler O, Coriat P, Viars P: Combined epidural and general anesthesia for abdominal aortic surgery. Anesthesiology 1991; 75:611-8.
de Leon-Casasola OA, Lema MJ, Karabella D, Harrison P: Postoperative myocardial ischemia: Epidural versus intravenous patient-controlled analgesia: A pilot project. Reg Anesth 1995; 20:105-12.
Beattie WA, Buckley DN, Forrest JB: Anaesthetic techniques: Epidural morphine reduces the risks of postoperative myocardial ischaemia in patients with cardiac risk factors. Can J Anaesth 1993; 40:532-41.
de Leon-Casasola OA, Lema MJ: General versus regional anesthesia for peripheral vascular surgery (letter). Anesthesiology 1996; 85:224-5.
Bode RH Jr, Lewis KD, Pierce ET: General versus regional anesthesia for peripheral vascular surgery (letter). Anesthesiology 1996; 85:225.
Early Breast Cancer Trialists' Collaborative Group: Systemic treatment of early breast cancer by hormonal, cytotoxic, or immune therapy: 133 randomised trials involving 31,000 recurrences and 24,000 deaths among 75,000 women. Lancet 1992; 339:1-15, 71-85.
Fibrinolytic Therapy Trialists' Collaborative Group: Indications for fibrinolytic therapy in suspected acute myocardial infarction: Collaborative overview of early mortality and major morbidity results from all randomized trials of more than 1,000 patients. Lancet 1994; 343:311-22.
Moller IW, Hjortso E, Krantz T, Wandall E, Kehlet H: The modifying effect of spinal anesthesia on intra- and postoperative adrenocortical and hyperglycaemic response to surgery. Acta Anaesthesiol Scand 1984; 28:266-9.
Raggi R, Dardik H, Mauro AL: Continuous epidural anesthesia and postoperative epidural narcotics in vascular surgery. Am J Surg 1987; 154:192-7.
Kehlet H: Epidural analgesia and the endocrine-metabolic response to surgery: Update and perspectives. Acta Anaesthesiol Scand 1984; 28:125-7.
Table 1. 
Image not available
Table 1. 
×
Table 2. 
Image not available
Table 2. 
×