Facebook
Twitter
LinkedIn
Email
 |
Checking for direct PDF access through Ovid | |
Excerpt
Dr. Barie and Ms. Hydo have raised numerous points concerning our report [1] of adult respiratory distress syndrome (ARDS) associated with epidural fentanyl infusion. First, Dr. Barie and Ms. Hydo point out that pulmonary edema is known to occur after vascular procedures in which tissue ischemia occurs. Dr. Barie and Ms. Hydo claim that vascular surgical patients are at a risk for ARDS that is similar to the rate of ARDS that we observed. We do not agree. Paterson et al. [2] described noncardiogenic pulmonary edema, a term that generally is used when criteria are insufficient for ARDS, but when a similar pathophysiology is suspected. These authors [2] showed that reperfusion of dysoxic muscle causes pulmonary inflammation during vascular surgery. However, the phenomenon reported by Paterson et al. [2] is clearly of more academic than clinical interest, since pulmonary edema resolved in all patients within 24 hrs. All six of our patients were initially extubated, thereby demonstrating that pulmonary inflammation did not arise from intraoperative muscle ischemia. More importantly, all of our patients satisfied criteria for full-blown ARDS: P(A-a)O sub 2 ranged from 232 to 544 torr (30.9 to 72.5 kPa), and the required level of positive end-expiratory pressure was 10 to 20 cm H2 O. Two of our patients died; all of the patients in the study by Paterson et al. [2] survived.
In our facility, the incidence of postoperative intubation lasting more than 48 hrs for our 422 vascular surgical procedures during the period October 1991 to December 1993 was 5.7%. Of these vascular procedures, 58 were carotid endarterectomies, 77 involved aortic cross-clamping, and 252 were procedures necessitating muscle ischemia. Our local figure is identical to the national Veterans Administration figure of 5.7% during the same period.
Thus, we find ARDS in one of our vascular surgical patients to be unusual, and even more so in the absence of an identifiable, clinically important physiologic insult, such as sepsis or prolonged hypotension. We would find a rate of ARDS (as defined by our criteria) of 12% after elective procedures to be unusual, and cause for concern.
The second broad area of concern raised by Dr. Barie and Ms. Hydo regards the rigor of our statistical analysis. We did not, as they state, use a ``selected control population.'' The control population was, in fact, the entire population of patients admitted to the intensive care unit after major vascular and thoracic surgical procedures. This approach makes selection bias unlikely. Although Dr. Barie and Ms. Hydo state that they ``easily could have reached a conclusion similar to that reported'' by applying subset analysis to their postoperative ARDS data, we are unable to identify such a subset. For example, the difference in incidence of ARDS among their elective surgical patients receiving (4 of 22) and not receiving (3 of 33) fentanyl is not statistically significant (p equals .18). We agree that analysis of multiple subsets to find obscure statistical associations is a dubious practice; thus, we avoided it.
After having warned of the dangers of subset analysis Dr. Barie and Ms. Hydo recommend a multivariate analysis of our patient population to identify other potential risk factors for ARDS. However, construction of a multivariate model requires a minimum of five events per term entered into the model. Since our outbreak involved a total of six patients, construction of a multivariate model would undermine rather than promote statistical rigor.
Finally, Dr. Barie and Ms. Hydo point out that our data represents ``circumstantial evidence.'' We agree. Epidemiologic evidence is inherently circumstantial in nature; cause and effect is difficult to demonstrate by purely epidemiologic means.
