Reply: Use of Acellular Dermal Matrix in Postmastectomy Breast Reconstruction: Are All Acellular Dermal Matrices Created Equal?

Abstract

Sir: As one can also learn from Jack Schaefer’s Western novel Shane, the protagonist, although outwardly noting the equivalence between a gun, axe, and shovel, ultimately illustrates to the reader the power of the right man with the right tool. In our study, “Use of Acellular Dermal Matrix in Postmastectomy Breast Reconstruction: Are All Acellular Dermal Matrices Created Equal?” we present our analysis of complications among 309 women (521 breasts) who underwent breast reconstruction stratified by the use of either AlloDerm (LifeCell Corp., Branchburg, N.J.) or FlexHD (Ethicon, Inc., Somerville, N.J.). We conclude after multinomial logistic modeling that the use of FlexHD was significantly associated with an increased risk of major infections among patients in our cohort. Although our methodology suffers from shortcomings that are primarily addressed by prospective, randomized, controlled trial study designs, our ability to identify important distinctions between rates of complications using AlloDerm or FlexHD is nonetheless robust and founded on sound statistical principles. Although Dr. Liao accurately argues for the use of propensity matching to minimize the effects of selection bias, we have clearly demonstrated that the cohort that received AlloDerm was not significantly different from the cohort that received FlexHD, which makes it very unlikely that propensity matching would change our findings and ultimate conclusions. The propensity score matching method is most advantageous over regression adjustment when there is not sufficient overlap between the two groups in covariate distribution.1 Besides length of follow-up, we found sufficient overlap between the two groups, which is exemplified by a lack of statistically significant differences between the two groups in Table 1 of the article. In addition, according to basic statistical standards, at least 10 events or patients per covariate are required to be able to accurately detect true differences between study groups. In this sample, because there were 43 patients who developed a major infection, we are able to effectively control for four covariates in our model; adding more covariates, although an attractive option, detracts from the accuracy of the analysis and is statistically incorrect.1 The variables we did include in our model are factors that affect the choice of acellular dermal matrix or the risk of major infection. Importantly, however, we did preliminarily evaluate the impact of comorbidities before formulating our final model; as expected, given the low incidence of complications in our cohort, our conclusion that FlexHD increases the risk for major infection, even after controlling for comorbidities as an added covariate, is maintained. Another important consideration as highlighted by Drs. Liao and Wagner is the impact of surgeon-specific factors on outcomes.2,3 On univariate analyses, there were no significant associations between surgeons and infections with use of AlloDerm or FlexHD (Table 1). Furthermore, each surgeon included in our study used both FlexHD and AlloDerm, making it highly unlikely that a particular surgeon’s complications explain our results. Attempts to control for each surgeon in every regression model as a fixed effect is problematic based on the low event rates for each complication and also based on the quasi-separation of the model that occurs when odds ratios cannot be estimated for surgeons who have zero complications. As an alternative, surgeon can be treated as a random effect in the models, with a focus on the extent to which surgeon accounts for variance in the complication rates. This again is limited by the low event rates. When this analysis was performed by patient or by breast, the models for infection estimated the variance component of the surgeon effect to be 0. As such, the model would not control for surgeon as a random effect and defaults to the original analysis presented in our article. Given this information, we did not include surgeon as a covariate in our regression model to allow for proper control of four other covariates to maintain statistical accuracy in our model.Table 1.: Univariate Analysis of Association between Surgeon and Infection with Use of FlexHD and AllodermIn conclusion, we thoroughly appreciate the questions posed by our colleagues within this forum, as it highlights the yearning for meticulous and rigorous study design within our field. However, not liking our findings does not invalidate our study. More importantly, Dr. Liao’s acknowledgement of his consultant agreements with the Musculoskeletal Transplant Foundation and LifeCell does not change the fact that he has a conflict of interest that can potentially bias his opinion on the subject at hand. Furthermore, our findings, in fact, align with previously published concerns on the risk for complications using FlexHD as published by Liu et al., who note that FlexHD may be a risk factor for implant loss.4 Although statistical significance was not reached in their study with regard to surgical site infection specifically, there was indeed a trend in this direction, indicating the need for a closer look at the question at hand. Additional studies published on this topic were not designed to study the question that we address, and are irrelevant to the accuracy of our findings.5 It is not our intention to overstate our conclusions. In fact, we specifically highlight the need for higher powered, randomized, controlled trials to address the very question we have attempted to study. As surgeons, however, and not cowboys in a work of fiction, we must make clear distinctions in the tools we use and the ways in which we use them, so that we may recognize the significance of the right person with the right tool. DISCLOSURE The authors have no financial interest in any of the products or devices mentioned in this communication. Kavitha Ranganathan, M.D.Adeyiza O. Momoh, M.D.Division of Plastic and Reconstructive SurgeryUniversity of Michigan Health SystemsAnn Arbor, Mich.