In Reply:-Beattie, Roizen, and Downing's letter highlights some of the limitations of clinical trials, and emphasizes the major point of our editorial: that further clinical trials of regional versus general anesthesia are unlikely to demonstrate differences in cardiac outcomes between the two anesthetic techniques. One of the constraints of clinical trials is that they require detailed management protocols. Otherwise, it is not possible to determine which particular aspect of an intervention was responsible for any difference in outcome: was it the type of anesthesia, or the regulation of hemodynamic parameters? Another limitation is that patients in trials, because they must meet selection criteria, have fewer comorbidities than patients in routine clinical care. Quality of care and attention to outcome ascertainment may be higher in the context of a study. After a study is completed, nonrandomized analyses often suggest alternative explanations for the results. The question of generalizability in this case, whether a "real world" comparison of regional versus general anesthesia would yield the same results plaguing nearly all trials. A treatment may be better (or worse), cost more (or less), or be impractical in a clinical setting.

We agree that alternative strategies for the prevention of perioperative cardiac morbidity, such as the use of beta blockade or nitrates, may be beneficial, and that they should be studied. Compared with additional studies that compare available regional and general anesthetic agents, these appear to be fruitful lines of inquiry. As to the use of pulmonary artery catheters in selected patients, if Beattie, Roizen, and Downing believe that the risk of postoperative myocardial infarction after lower extremity vascular bypass grafting is truly 12% in their institutions, then high-risk patients have already been identified. A randomized trial of whether such patients benefit from pulmonary artery catheterization would be worthwhile. Such a trial, if it demonstrated benefit, would raise the question of whether hemodynamic management in response to PA abnormalities in the "real world" could match that in the trial. Conversely, a study that showed no difference would be subject to the flip side of the same criticism: that physicians in nonacademic centers may need the information from a PA catheter more.

We have a more difficult time understanding the purpose of clinical trials to study differences in the proxy variable of surgical stress response. Without demonstrated differences in cardiac morbidity and mortality, the meaning of any differences in stress responses will be uncertain.

Warren Browner, M.D., M.P.H.; Alan Go, M.D., General Internal Medicine Section, VA Medical Center, 111A1, San Francisco, California 94121.

(Accepted for publication July 18, 1996.)