Correspondence  |   January 2006
Lies, Damn Lies, and Statistics
Author Notes
  • Marian Medical Center, Santa Maria, California.
Article Information
Correspondence   |   January 2006
Lies, Damn Lies, and Statistics
Anesthesiology 1 2006, Vol.104, 202. doi:
Anesthesiology 1 2006, Vol.104, 202. doi:
To the Editor:—
Mark Twain may have overstated his distrust of statistics, but the issue of interpretation of statistics comes to the forefront in the study by Arbous et al.  1 and the accompanying editorial by Warner.2 As the results of the study are discussed, Arbous et al.  jump from describing associations  between outcomes and management factors, to cause-and-effect descriptions: “it was found … a checklist decreased the risk,”“the reversal of the effect of opiates and muscle relaxants seems to decrease the risk,” and so on. Warner embraces these ersatz “risks” as showing “anesthetic management processes to dramatically reduce perioperative mortality.”
When one looks at baseline characteristics of the study and control groups, there are, as the authors note, huge differences in the categories of urgent/emergent nature, time of day procedure performed, and American Society of Anesthesiologists physical status. In fact, 40% of the study cases were rated American Society of Anesthesiologists V—not expected to survive for 24 h, with or without surgery (regardless of anesthetic management). If we accept that a large proportion of the study cases carry greater risk by virtue of their physical status and the emergent nature of the injury or disease process, and  that urgent/emergent cases generally account for all the outside working hour cases, differences in anesthetic management processes between the two groups seem more coincidentally associated than causative. Were equipment checks performed less frequently in the study group because of the emergent nature of the cases? Was the lower percentage of two providers at termination of a procedure simply a function of the outside hour the procedure was performed? Was the lower reversal rate of opiates and muscle relaxants due to the fact that the study group was sicker, undergoing more complex procedures, and so remained intubated postoperatively? Did the study groups receive fewer narcotics and local anesthetics for postoperative pain because of their moribund (comatose?) or unstable condition?
This is not to suggest that anesthesia practice factors do not affect morbidity and mortality; some of the anesthetic practice factors in this study may one day be proven to be causative. But let us appreciate the method of this study for what it is: a tool to identify associations. Those associations then need to be further studied to identify them as causative or coincidental. Cars manufactured on Monday do not have more problems than others because that day of the week starts with the letter M  . We need to look for truths, damn truths, and more than associations.
Marian Medical Center, Santa Maria, California.
Arbous MS, Meursing AEE, van Kleef JW, de Lange JJ, Spoormans HHAJM, Touw P, Werner FM, Grobbee DE: Impact of anesthesia management characteristics on severe morbidity and mortality. Anesthesiology 2005; 102:257–68Arbous, MS Meursing, AEE van Kleef, JW de Lange, JJ Spoormans, HHAJM Touw, P Werner, FM Grobbee, DE
Warner MA: Perioperative mortality. Anesthesiology 2005; 102:251–6Warner, MA