[Skip to Content]
[Skip to Content Landing]
Citations 0
Invited Critique
October 20, 2008

Management of Delayed Postoperative Hemorrhage After Pancreaticoduodenectomy—Invited Critique

Arch Surg. 2008;143(10):1007. doi:10.1001/archsurg.143.10.1007

Lies, damn lies, and statistics”: This famous phrase could be used to describe recent forays into evidence-based medicine, where the filtering of disparate data from small, underpowered studies using the mechanics of a meta-analysis conveys to the reader a higher degree of precision than is warranted. Take the meta-analysis of DPH after pancreaticoduodenectomy by Limongelli and colleagues published in this issue of the Archives. The authors have followed rigorous steps for identifying the relevant published literature, combining them, and applying appropriate statistical analysis using a random-effects model to examine the overall results. They performed an appropriate, well-done textbook study, even including oversight on their data collection and transfer. What this method fails to convey is that, despite this structure, it remains a composite of small numbers of patients (largest series, 22 patients), treated by a large number of heterogeneous surgeons on several continents who possessed widely disparate expertise and variable access to dissimilar levels of care. It seems inconceivable to me that a P value, no matter how rigorously derived, is capable of accurately adjusting for all of these variables and mathematically accepting one method of treatment over another. It may be time to accept the fact that there are certain disease states encountered not infrequently in biomedical science, particularly the discipline of surgery, that present insurmountable limitations when one attempts to derive a single truth for all patients. Is this the elusive art that we all recognize as a major component of our discipline? As the authors correctly conclude in their discussion, “The management of this life-threatening complication ultimately will be dictated by the clinical status of the patient and the institutional preference.” What they failed to mention is, “despite the P value assigned.”

First Page Preview View Large
First page PDF preview
First page PDF preview