J Am Coll Surg
is only one way to know what might happen during a larger experimentdit must be performed. In sum, sample size cannot explain a nonsignificant result absent the invocation of multiple, under-the-radar assumptions about the conduct and outcome of an imaginary study not yet performed. Most curiously, Brody and colleagues2 never endorsed or even mentioned confidence intervals. For a nonsignificance randomized trial situation, the “no-differencebetween-treatments” scenario has representation inside a difference confidence interval, but so do contradicting scenarios, any of which plausibly could have generated the raw data. Low-power experiments generate broad confidence intervals. One strategic reason to use the highest practicable power is obvious: should a high-power study show statistical insignificance, the confidence interval for the difference between compared groups will not be broad, and very useful evidence about reality will be forthcoming. Alternatively, it can be eye-opening to examine the confidence interval after a clinical trial has shown that two compared treatments have a significant difference; not a few “positive” trials are actually rather close calls. Using confidence intervals is reader-friendly, visually powerful, nondeceptive, and should be a strict requirement for initial manuscript consideration by all journal editors. REFERENCES 1. Mayo DG. Did Pearson reject the Neyman-Pearson philosophy of statistics? Synthese 1992;90:233e262. 2. Brody BA, Ashton CM, Liu D, et al. Are surgical trials with negative results being interpreted correctly? J Am Coll Surg 2013;216:158e166. 3. Hoenig JM, Heisey DM. The abuse of power: The pervasive fallacy of power calculations for data analysis. Amer Statistician 2001;55:19e24.
Disclosure Information: Nothing to disclose.
Reply Baruch Brody, MD, Nelda Wray, Carol Ashton, MD, MPH Houston, TX
We want to thank Dr Lee for his thoughtful comments, which give us the opportunity to correct one set of mistakes in the paper as well as to discuss some larger issues. Dr Lee said that we have confused beta error and power. We have reviewed every use of both words, and we are confident that we have not. But there were 4 occasions (once in the abstract, once in the introduction, and
twice in the methods) when we wrote (b 0.9) when we should have written power 0.9. We apologize for that error, and we believe that this may explain why he thought that we had confused the two. We also agree with him that confidence intervals would have been very helpful, but we did not discuss them because most of the papers did not present them. We do disagree with one of his comments. We do not believe that potential explanations are an intellectual trap. On the contrary, we believe that they are a crucial portion of any discussion section. If no significant differences have been discovered, it is important for the authors to identify potential explanations (eg, there really is no significant difference or there is one that will be found with a larger sample). This sets the stage for future research to identify the correct explanation. We need to add a final word about post hoc power calculations. As Dr Lee said, for scientific purpose, they are clearly nonsense. We were using them to clarify to the reader with less statistical sophistication the extent of the possibility of a type II error in the case of those trials. Disclosure Information: Nothing to disclose.
Outcomes in Autologous Breast Reconstruction James L Mayo, MD, Charles Dupin, Hugo St Hilaire, MD, DDS, FACS New Orleans, LA
Dr Gart and colleagues should be congratulated on their efforts to more clearly define and compare the outcomes of different autologous breast reconstruction on a large scale.1 Most previously published data have been from a single surgeon or institution that often excels in a specific reconstructive option. By pooling the outcomes data of more than 240 hospitals, the NSQIP database provides an opportunity to significantly improve on the current available data when choosing reconstructive options.2 This information is quite important to the breast surgeon, who represents the “first responder” to a patient undergoing a mastectomy for cancer and considering reconstruction. The breast surgeon participates in both the decision to reconstruct and, ultimately, the decision on type of reconstruction in that a referral may be made to a surgeon who performs predominantly 1 type of autologous reconstruction. This article specifically commented on 30-day morbidity associated with various forms of “autologous” breast reconstruction. The authors concluded that
Vol. 217, No. 1, July 2013
“30-day complication outcomes of pedicled flaps, particularly the latissimus flap are.lower than those for free flaps.”1 The information provided through the NSQIP database is beneficial in its inclusion of data from such a large number of hospitals. However, a few points must be recognized when reviewing these data if they are to be used in the education of patients and the decisionmaking process for choice of breast reconstruction. First, the term autologous for latissimus dorsi flap breast reconstruction is somewhat of a misnomer. The majority of reconstructions using a latissimus flap include placement of a prosthetic, either an expander followed by implant or an immediate implant. This actually creates a mixed reconstruction technique combining the benefits and drawbacks of both autologous and nonautologous reconstruction. Most notably, the rate of capsular contracture poses a risk for future complications and repeated operations. Multiple studies have shown that postmastectomy radiated patients, regardless of the addition of a musculocutaneous flap, have significantly higher contracture rates, which can lead to additional long-term morbidity, including additional surgical procedures.3-6 The addition of radiation therapy to a patient’s treatment, either pre-or postoperatively, can alter the reconstructive plan. Some groups have instituted a delayed-immediate reconstructive timeline in which a prosthetic reconstruction is performed at the time of the mastectomy and followed by an autologous reconstruction once radiation therapy is ruled out or completed.6 The NSQIP database has amassed a population of patients with roughly 1% receiving postmastectomy radiation. This population size seems considerably low given that past publications comparing outcomes in breast reconstruction among radiated and nonradiated patients have shown anywhere from 10% to 49% of postmastectomy patients undergoing radiation therapy.7-9 The lack of representation of the postmastectomy radiated population within the NSQIP database all but eliminates application of the NSQIP data to this group. This specific group is thought to benefit most from a purely autologous reconstruction.10 Also confounding the data is the pooling of all microsurgical reconstruction within 1 group. Numerous free flap reconstructive options exist. Donor sites for these flaps include the abdomen, thigh, or gluteal tissue. Those from the abdomen include the deep inferior epigastric artery perforator (DIEP) flap, the transverse rectus myocutaneous flap (TRAM), and the muscle-sparing TRAM. Each of the flaps and donor sites carries specific pros and cons. The outcomes of the abdominally based flaps differ in the amount of rectus muscle taken with the flap. Although data have often failed to show a difference between outcomes of muscle-sparing TRAM and
DIEP free flaps, data have been presented demonstrating a difference between outcomes when comparing free TRAM with DIEP free flap reconstruction.11-13 Due to the use of a single CPT code for free flap breast reconstruction, the NSQIP database is unable to differentiate between types of free flap reconstruction, but these subgroups do carry differences that do influence reconstruction decision-making. Finally, as mentioned in the article, the data represent only 30-day morbidity. A key difference between various reconstructive options is unrecognized. Although the authors did mention this in the discussion, its importance cannot be understated. Some of the benefits of autologous reconstruction are born out in the reduced long-term morbidity, specifically by avoiding contracture and future resultant operations. Furthermore, within this cohort, both free and pedicled TRAM flaps present a risk for future hernia, a long-term complication. The search for better data regarding various forms of breast reconstruction is clearly daunting. The NSQIP database has allowed analysis of a large cohort of patients from many institutions, offering an opportunity to improve on current data. However, as expressed earlier, these results can be misleading. All free flap reconstructive options should not be grouped as equals. The latissimus flap, although a great option in the right patient, is not a truly autologous reconstruction. Full inclusion of all outcomes and long-term follow-up combined with a regrouping of reconstructive procedures would benefit this analysis. In the meantime, the overall process of selecting a type of breast reconstruction will remain complex and often be guided by experience and physician preference. Surgeonspecific factors such as available resources and technical skill set as well as woman-specific factors such as obesity, postmastectomy radiation, comorbidities, and personal goals/objectives all play a role in choosing a reconstructive option. The “first responder” to a woman in search of breast reconstruction will remain the breast surgeon. Ultimately, knowing the options for reconstruction and the skill sets and resources available to regional reconstructive surgeons will best serve the breast surgeon when discussing reconstruction with a patient. REFERENCES 1. Gart MS, Smetona JT, Hanwright PJ, et al. Autologous options for postmastectomy breast reconstruction: a comparison of outcomes based on the American College of Surgeons National Surgical Quality Improvement Program. J Am Coll Surg 2013;216:229e238. 2. Birkmeyer JD, Shahian DM, Dimick JB, et al. Blueprint for a new American College of Surgeons: National Surgical
6. 7. 8.
9. 10. 11. 12.
Quality Improvement Program. J Am Coll Surg 2008;207: 777e782. Ascherman JA, Hanasono MW, Newman MI, et al. Implant reconstruction in breast cancer patients with radiation therapy. Plast Reconstr Surg 2006;117:359e365. Benediktsson K, Perbeck L. Capsular contracture around saline-filled and textured subcutaneously placed implants in irradiated and non-irradiated breast cancer patients: Five years of monitoring of a prospective trial. J Plast Reconstr Aesthet Surg 2006;59:27e34. Behranwala KA, Dua RS, Ross GM, et al. The influence of radiotherapy on capsule formation and aesthetic outcome after immediate breast reconstruction using biodimensional anatomical expander implants. J Plast Reconstr Aesthet Surg 2006;59:1043e1051. Kronowitz SJ. Delayed-immediate breast reconstruction: technical and timing considerations. Plast Reconstr Surg 2010;125: 463e474. Evans GR, Schusterman MA, Kroll SS, et al. Reconstruction and the radiated breast: Is there a role for implants? Plast Reconstr Surg 1995;96:1111e1115. Carlson GW, Page AL, Peters K. Effects of radiation therapy on pedicled transverse rectus abdominis myocutaneous flap breast reconstruction. Ann Plast Surg 2008;60: 568e572. Spear SL, Ducic I, et al. The effect of radiation on pedicled TRAM flap breast reconstruction: outcomes and implications. Plast Reconstr Surg 2005;115:84e95. Kronowitz SJ, Robb GL. Radiation therapy and breast reconstruction: a critical review of the literature. Plast Reconstr Surg 2009;124:395e408. Man LX, Selber JC, Serletti JM. Abdominal wall following free TRAM or DIEP flap reconstruction: a meta-analysis and critical review. Plast Reconstr Surg 2009;124:752e764. Nahabedian MY, Tsangaris T, Momen B. Breast reconstruction with the DIEP flap or the muscle-sparing (MS-2) free TRAM flap: is there a difference? Plast Reconstr Surg 2005; 115:436e444. Bajaj AK, Chevray PM, Chang DW. Comparison of donorsite complications and functional outcomes in free musclesparing TRAM flap and free DIEP flap breast reconstruction. Plast Reconstr Surg 2006;117:737e746.
Disclosure Information: Nothing to disclose.
Reply Michael S Gart, MD, John T Smetona, BS, BA, Philip J Hanwright, BA, Neil A Fine, MD, FACS, Kevin P Bethke, MD, FACS, Seema A Khan, MD, FACS, John YS Kim, MD Chicago, IL We thank Drs Mayo, Dupin, and Hilaire for their careful reading and thoughtful observations on our paper. The authors correctly pointed out that there are some caveats to labeling latissimus flaps as “autologous” because in the majority of cases, a prosthetic device would
J Am Coll Surg
supplement the reconstruction. Given that our specialty has traditionally categorized reconstructions as autologous or prosthetic, we are following standard nomenclature with reference to autologous reconstructions.1-5 However, the authors cogently pointed out that these categories may be oversimplified. There are some inherent limitations of the NSQIP database as it applies to reconstruction, and the tracking of radiation is one of these. The database only tracks radiation within 90 days in the preoperative setting.6,7 We agree with the respondents that future iterations of the database could be enhanced with a precise elucidation of radiation therapy. The respondents point out that differences in the types of microsurgical reconstruction are not distinguished. Unfortunately, independent recording of outcomes variables with NSQIP depends on CPT codes, and the granularity of the data is dependent on the concomitant nuances in CPT coding. There is some ongoing discussion of the relative outcomes of the varying forms of abdominal free tissue transfer (DIEP vs MS-TRAM vs free TRAM) and pedicled TRAM flaps. Moreover, the respondents correctly pointed out that pooled data such as NSQIP do not factor in individual surgeon or institutional expertise. The literature has been focused almost exclusively on single surgeon and institutional data and we believe that having pooled data as counterpoint to these studies may be helpful to the community at large. We agree with the authors that some key outcomes (hernia, fat necrosis, etc) will extend beyond the 30-day window of captured data in NSQIP. These limitations were highlighted in the discussion of our article. Again, the value proposition of the NSQIP database is that it allows for multi-institutional and independent recording of outcomes. The current conclusions provide an important element of surgeon and patient education; however, we agree with the respondents that as the robustness of the data continues to improve, the conclusions will become more germane. REFERENCES 1. Alderman AK. Discussion: Postmastectomy breast reconstruction in the irradiated breast: a comparative study of DIEP and latissimus dorsi flap outcome. Plast Reconstr Surg 2012; 130:21e22. 2. Chevray PM. Timing of breast reconstruction: immediate versus delayed. Cancer J 2008;14:223e229. 3. Kim Z, Kang SG, Roh JH, et al. Skin-sparing mastectomy and immediate latissimus dorsi flap reconstruction: a retrospective analysis of the surgical and patient-reported outcomes. World J Surg Oncol 2012;10:259. 4. Lindegren A, Halle M, Docherty Skogh AC, Edsandernord A. Post mastectomy breast reconstruction in the irradiated breast: