LETTERS

Preoperative Skin Preparation

James T Lee, MD, PhD, FACS

Minneapolis, MN

Recent work by Ellenhorn and coworkers1 tested theidea that preoperative scrubbing of skin with povidone-iodine soap adds no incremental protection againstwound infection. Two cohorts (n � 115 and n � 119)were created by randomizing abdominal surgery patientsto “paint-only” or “scrub-plus-paint” categories. If theauthors were fastidious in their experimental execution,if randomization was “perfect” (ie, risk factors balanced),and if infection surveillance accuracy was the same foreach cohort, comparing infection rates for the two ex-perimental cohorts might support a generalization aboutthe value of a povidone-iodine scrubbing step.

The authors observed an incisional infection rate of10% for each of their cohorts and essentially identicalintraabdominal infection rates (3% and 2%) as well. Iworry that certain nuances in their interpretation ofthese findings could unintentionally confuse fatiguedreaders who might skim the article’s abstract after fulldays of elective operating or long nights on emergencycall.

Whenever inferential statistics are used to comparefrequencies of some defined outcomes in two clinicaltrial cohorts, a thought experiment is being performedin hopes of sharpening knowledge of reality. Essentially,such cohorts are viewed as samples taken from two pop-ulations or universes. In the Ellenhorn study, one imag-ined universe was “all abdominal surgery patients onEarth receiving paint and scrub” and the other imagineduniverse was “all abdominal surgery patients on Earthreceiving only paint.” A statistical maneuver estimatesthe range of rate differences between universes that arecompatible with some observed cohort rate difference,which might be zero, as measured in the present case.The authors used a method described by Rodary andcolleagues2 and obtained an nonsignificant p value, con-cluded that the skin preparation routines are equivalent,and called for the elimination of povidone-iodinescrubbing.

The meaning of equivalence is critical in digesting thisarticle. The authors declared before commencing the

trial that an absolute infection rate difference as great as6% would be deemed compatible with equivalence ofthe skin preparation routines, but no clinical rationalewas offered for selecting that value. In addition, hastyreaders can misconstrue the adjective absolute. Unfortu-nately, the authors did not emphasize that their resultscannot rule out a large advantage of adding povidone-iodine scrubbing. For example, their data are compatiblewith an actual infection frequency of 14.5% in the paint-only universe and 7% in the paint-plus-scrub universe. Inclinical practice, this would represent an infection risk dif-ferential of � 50% and few infection prevention protocolscan discount any tactic having that potential impact. Usingthe authors’ raw data, computation of an exact 95% confi-dence interval (with continuity correction) for the differ-ence in infected proportion between universes reveals aconsiderably wide range (�.085 to �.089). We can confi-dently conclude only this: povidone-iodine scrubbingmight add benefit, might have no effect at all, or mightactually increase infection likelihood. These are the samethree possible knowledge states that existed before theauthors’ experiment was performed. Maybe we shoulddiscard preoperative povidone-iodine scrubbing. Maybewe should not.

REFERENCES

1. Ellenhorn JDI, Smith DD, Schwarz RE, et al. Paint-only is equiv-alent to scrub-and-paint in preoperative preparation of abdomi-nal surgery sites. J Am Coll Surg 2005;201:737–741.

2. Rodary C, Con-Nougue C, Tournade M. How to establish equiv-alence between treatments: a one-sided clinical trial in paediatriconcology. Stat Med 1989;8:593–598.

Reply

Joshua Ellenhorn, MD

Duarte, CA

In his letter, Dr Lee points out some subtle statisticalcriticisms in interpreting our findings. Dr Lee is correctin the assumptions of drawing representative samples toestimate population parameters, in this case infectionrate. This is fundamental to all statistical inference. The

853© 2006 by the American College of Surgeons ISSN 1072-7515/06/$32.00Published by Elsevier Inc. doi:10.1016/j.jamcollsurg.2006.01.013

statistical underpinnings of an equivalency or noninfe-riority trial require that an initial difference in outcomesbetween groups be designated as a clinically relevantdifference. Adhering to a strict interpretation of the re-sults of the equivalency trial, one can only make conclu-sions with respect to the difference in outcomes initiallydesignated. For our study, we chose 6% as that desig-nated difference, to avoid a potentially meaningless dif-ference in infection rate between groups, which mightbe statistically different but not clinically relevant. Thiswas balanced with the sample size requirements for asmaller equivalence threshold. In Dr Lee’s example forcomparing 9% infection rate for paint-only versus 4% inpaint-plus-scrub, the sample size requirements would beapproximately 650 patients per group (1,300 patientstotal).

We have contributed to the literature a controlledclinical trial that was designed to test the idea that pre-operative scrubbing of skin with povidone-iodine soapadds no incremental protection against wound infec-tion. We do not agree with Dr Lee that our results havethe same interpretation as a hypothetical trial, where theactual infection frequency was 9% in the paint-only armand 4% in the paint-plus-scrub arm. Although Dr Lee’sworst-case outcomes would be within the tolerance ofour 6% equivalence threshold based on our current re-sults and the results of previous trials, that specific out-come would be unlikely given the data that has beenreported as of this writing. Admittedly, Dr Lee is correctthat the interpretation of our results has not added to thepossible knowledge-space: povidone-iodine scrubbingmight add benefit, might have no effect at all, or mightactually increase infection likelihood. Statistical rea-soning does not deal in perfect knowledge states. Itcan lead to conclusions that speak to the likelihood ofoutcomes. We would argue that our data have shownthat the advantage of povidone-iodine scrubbing overpaint-only is minimal at best because it is the mostlikely interpretation.

Despite the results of even the most rigorously de-signed single clinical trial, the thoughtful clinician willalways weigh the pros and cons of altering his or herclinical practice. Results of our single clinical trial mustbe interpreted in the context of the four earlier clinicaltrials available for review, all of which are cited in ourpublication. Each of these was a negative trial, and eachtrial compared the “gold standard” scrub-plus-paint tosomething different and often less than scrub-plus-

paint. In view of this body of literature and our clearlynegative clinical trial, an objective observer who is notlost in the vagaries of statistical minutiae, can concludethat a reasonable argument can be made for abandoningscrub-and-paint.

Delayed Massive Hemorrhage afterPancreaticoduodenectomy: A NewTherapeutic Approach

Giuseppe Navarra, MD

Marcello Bartolotta, MD

Adalberto Barbera, MD

Messina, Italy

We read with great interest the series of delayed massivehemorrhages after pancreaticoduodenectomy reportedby Tien and colleagues1 in your journal. They have con-firmed that delayed massive hemorrhage after pancreati-coduodenectomy, often associated with septic complica-tions secondary to leakage or intraabdominal abscess, isstill a frequent event carrying a high mortality rate. Afterthe failure of conservative management, final manage-ment can be either radiologic or surgical, as stated byTien and colleagues.1 Although radiologic managementis partly dependent on resuscitation facilities at the de-partment of radiology and the prompt availability ofexperienced interventional radiologists, surgical man-agement is much more invasive and brings with it highmorbidity and mortality, even if it succeeds in stoppingthe bleeding. But there is a pharmacologic agent that canbe used as rescue treatment in case of massive bleeding:recombinant activated factor VII (rFVIIa), which weused to treat a case of delayed massive hemorrhage afterpancreaticoduodenectomy.

Recombinant FVIIa is a major alternative for man-agement of hemophiliac patients with inhibitors.2 Morerecently, it has been used off-label to control bleeding inpatients with trauma or other massive life-threateninghemorrhage, and to reduce blood loss in surgical pa-tients with normal coagulation.3-5 Recombinant FVIIabinds to activated platelets independently of tissue fac-tor. The resulting stimulation of an exaggerated earlythrombin burst at sites of vascular injury makes it anattractive potential treatment for massive, uncontrolledbleeding.

854 Navva J Am Coll Surg