Correspondence

Endovascular vs. Open Repair of Abdominal Aortic Aneurysms

N Engl J Med 2008; 358:2644-2645June 12, 2008

Article

To the Editor:

Schermerhorn and colleagues (Jan. 31 issue)1 examined the relative effectiveness of endovascular versus open repair of abdominal aortic aneurysms. They used a propensity-score approach to match the open-repair and endovascular-repair groups according to demographic and clinical factors. They argued strongly that unmeasured confounding factors were unlikely to bias their results. Unfortunately, propensity scores do not eliminate bias if there are unmeasured confounders.2 The story of the pulmonary-artery catheter provides a cautionary tale. Using a propensity score, investigators found that the use of a pulmonary-artery catheter was associated with a 24% increase in the risk of death as compared with no use of a pulmonary-artery catheter.3 Randomized, controlled trials, however, have not borne out this result.4 The discrepancy arose because there were unmeasured confounders in the observational analysis. In the current study, it is clear that patients in the endovascular-repair group were older and sicker than those in the open-repair group. There is a high likelihood that the propensity score, which is based on administrative data, does not include important confounders and therefore incompletely “controls” for the higher predicted mortality in the endovascular-repair group. Consequently, the analysis probably underestimates the treatment benefit of endovascular repair.

Mark D. Eisner, M.D., M.P.H.
University of California at San Francisco, San Francisco, CA 94143-0111
mark.eisner@ucsf.edu

4 References

1. 1

   Schermerhorn ML, O'Malley AJ, Jhaveri A, Cotterill P, Pomposelli F, Landon BE. Endovascular vs. open repair of abdominal aortic aneurysms in the Medicare population. N Engl J Med 2008;358:464-474
   Full Text | Web of Science | Medline

2. 2

   Weitzen S, Lapane KL, Toledano AY, Hume AL, Mor V. Weaknesses of goodness-of-fit tests for evaluating propensity score models: the case of the omitted confounder. Pharmacoepidemiol Drug Saf 2005;14:227-238
   CrossRef | Web of Science | Medline

3. 3

   Connors AF Jr, Speroff T, Dawson NV, et al. The effectiveness of right heart catheterization in the initial care of critically ill patients. JAMA 1996;276:889-897
   CrossRef | Web of Science | Medline

4. 4

   Wheeler AP, Bernard GR, Thompson BT, et al. Pulmonary-artery versus central venous catheter to guide treatment of acute lung injury. N Engl J Med 2006;354:2213-2224
   Full Text | Web of Science | Medline

Author/Editor Response

With regard to Eisner's comments: we conducted an observational study and used propensity-score methods1 to create evenly matched cohorts of patients. The propensity-score models controlled for an extensive list of potential confounders, but they were limited to variables that could be obtained from administrative data. Although our results could be biased by unmeasured confounders, the magnitude of the differences we observed makes it unlikely that an unmeasured confounder could have altered our results in favor of open repair. Furthermore, we agree with Eisner that unmeasured confounding is more likely to have led to bias against endovascular repair. Thus, in contrast to the example of the pulmonary-artery catheter, the fact that we found an important difference despite the potential for unmeasured confounders further underscores the importance of our results for clinical practice.

Marc L. Schermerhorn, M.D.
Beth Israel Deaconess Medical Center, Boston, MA 02215
mscherm@bidmc.harvard.edu

A. James O'Malley, Ph.D.
Bruce E. Landon, M.D., M.B.A.
Harvard Medical School, Boston, MA 02115

1 References

1. 1

   Rosenbaum PR, Rubin DB. The central role of the propensity score in observational studies for causal effects. Biometrika 1983;70:41-55
   CrossRef | Web of Science