A Rebuttal of a Meta-analysis Supporting Breast Implant Pocket Antimicrobial Irrigation

Engels et al1 conclude that triple antibiotic solution is most effective in reducing the risk of capsular contracture and povidone-iodine (Betadine) is most effective in reducing the risk of infection in patients after breast implant surgery. This conclusion contrasts with the findings of a 2017 meta-analysis, also published in Annals of Plastic Surgery,2 which concluded that antimicrobial pocket irrigation significantly increases the risk of capsular contracture after breast augmentation. What could account for this disparity? STATISTICS The authors present their data in forest plots, evaluating the risk of capsular contracture and infection.1 They use the Cochran Q test of heterogeneity to measure the variability of the data and whether the effect is caused solely by the intervention or also by other confounding factors, assigning a random effects model (confounders are in play) or a common effects model (the intervention accounts for the effect). A P value <0.05 indicates significant heterogeneity. Importantly, this P value does not reflect the statistical significance of any comparison between treatments. Surprisingly, the authors do not describe a statistical test comparing the categorical data between treatment groups, such as a χ2 test. This is highly unusual for a scientific study. Usually, forest plots compare an intervention with a control.2–5 A vertical line is set at an odds ratio of 1. Drinane et al2 listed 7 such studies that included comparative groups, finding an odds ratio of 2.6 for the pooled data (indicated by a diamond on the “control” side of the graph), favoring saline irrigation over antimicrobial irrigation. The forest plots presented by Engels et al1 are quite different. They do not represent a comparison between the intervention and a control group. The vertical line does not represent an odds ratio of 1; the diamond does not indicate an overall odds ratio. Instead, the plots simply represent the proportion of events without a comparative group. In fact, the words “odds ratio” does not appear in the article. For example, the results for capsular contracture reported by Walker et al6 are provided separately in the forest plots for saline (0 contractures, 8 patients) and triple antibiotic irrigation (0 contractures, 8 patients). Normally, this data would be presented in the same forest plot, with a marker sitting on the vertical line (no capsular contractures in either the treatment or control group for an odds ratio of 1). Inexplicably, many published studies listed in the plots have zero data listed.1 ELIGIBILITY CRITERIA In comparing the meta-analyses by Engels et al1 and Drinane et al,2 there are obvious differences in eligibility criteria. Engels et al1 included studies of breast reconstruction patients, including a study of tissue expanders,6 instead of limiting the review to breast augmentation patients as done by other investigators.2,7,8 Many of the studies included patients treated with acellular dermal matrix (ADM).1 The inclusion of breast reconstruction patients in a meta-analysis is questionable because these patients add numerous confounders in terms of diagnosis, ancillary treatments such as chemotherapy and radiation, different devices, and of course the introduction of a separate implanted foreign material (ADM). Implants used in breast reconstruction are associated with a greater risk of infection and capsular contracture than cosmetic breast augmentation.9,10 Breast reconstruction is preceded by a mastectomy, changing the microbiological environment by removing breast tissue, which is not sterile and is inhabited by commensal bacteria. The authors include a study that evaluated breast reduction patients,11 which falls outside their inclusion criteria. Similarly, the authors include a study comparing infection rates between reconstruction patients who were treated with absorbable antibiotic beads and a control group who were not treated with these beads.12 There was no mention of pocket irrigation. Even in studies restricted to cosmetic breast augmentation, numerous confounding variables are present, including the surgeon, device, specifics of pocket irrigation, follow-up time, surface characteristics, the insertion plane, the use of drains, and whether the implants are saline-filled or silicone gel. Baker scores are notoriously subjective.2 It is a challenge for any analysis to overcome this degree of data heterogeneity. As the heterogeneity increases, the validity of pooled data is compromised. Meta-analyses need to be selective when including studies.2–5 Fewer studies that are methodologically sound, containing fewer confounding variables, are preferred over more studies that are less robust. SAMPLE SIZES The authors do not mention statistical power or sample size calculation. Some studies included in the meta-analysis feature very small patient numbers. Walker et al6 evaluated 16 patients who underwent tissue expander breast reconstruction, split into 2 groups of 8 patients. One group received saline pocket irrigation. The other group received triple antibiotic irrigation in the form of 1-g cefazolin, 80-mg gentamicin, and 50,000 units of bacitracin (no longer available) in 500 mL of saline. When the incidence of a complication is infrequent, large numbers are needed. Using large studies with well-defined eligibility criteria improves reliability. For example, a 5-year manufacturer core study of 2560 breast augmentation patients provides exponentially greater numbers that improve reliability.13 The antibiotic irrigation cohorts from large core studies with 5-year and 10-year follow-up, containing many patients treated with triple antibiotic irrigation,13,14 are not included in this analysis, leaving very small numbers (only 124 patients) for this popular treatment group. These large studies found a significantly increased (not decreased) risk of capsular contracture among patients who received antibiotic irrigation.13,14 The authors did reference a core study by Calobrace et al,14 but included only one cohort of patients, the group treated with povidone-iodine irrigation. When 2 comparative cohorts are reported from the same study, there is greater likelihood that other variables are similar (eg, surgeon, insertion plane, device). Ideally, the numbers should be of similar magnitude so that an analysis is not overwhelmed by unbalanced data from 1 treatment arm of a single study. Studies with very few patients (eg, 2 groups of 8 patients)6 are essentially irrelevant. ERRORS IN STUDY DATA The analysis for infection after povidone-iodine is almost completely composed of data from a single study by Araco et al (1100 of 1283 patients, 86%).15 Araco et al15 reported 33 infections in a series of 3002 breast augmentation patients. All implants were washed with povidone-iodine solution, and most pockets (1902 patients, 63%) were also irrigated with a single antibiotic solution. Most patients (67%) also were treated with drains, which were found to be associated with a greater risk of infection. Confounders must be considered, including washing the implants rather than irrigating the pockets with povidone-iodine—an important difference in technique—and the use of drains, which is not typical practice in breast augmentation and increases the risk of infection. Engels et al1 assign the large 1902-patient cohort to the triple antibiotic group despite the fact that all the implants were washed with povidone-iodine solution as well. Moreover, these patients did not actually receive triple antibiotic irrigation. The antibiotic solution contained just 1 antibiotic, either cefuroxime or gentamicin (not both), which was used for patients with allergies. Consequently, this data should be removed from the triple antibiotic solution group. This correction reduces the group size from 3243 to 1341 patients and more than doubles the overall infection rate, from 4.1% to 8.7%. For patients treated with saline irrigation, the authors include a cohort reported by Giordano et al.16 In fact, these patients did not receive saline irrigation; they received no irrigation. Giordano et al16 reported 10 Baker III/IV capsular contractures among 165 patients, not 3 events among 167 patients, as listed on the forest plot.1 (Evidently this capsular contracture data was swapped with the infection data). No comparative data were reported from this study, possibly because these authors combined povidone-iodine with antibiotic solution. The data reported by Wiener et al17 are misrepresented. Engels et al1 list 5 capsular contractures among 538 patients treated with povidone-iodine irrigation. In fact, there were 7 affected patients.17 For a brief period in 2004, Wiener et al17 rinsed the pocket with saline after irrigating with Betadine. Fourteen patients (out of 341) developed a capsular contracture. This cohort is included in the “saline irrigation” group despite the fact that these patients were all exposed to Betadine irrigation. Subsequently, Wiener et al17 stopped the saline rinse after Betadine irrigation. Five more patients (out of 154) developed capsular contracture. The data should be correctly listed as 26 affected patients receiving povidone-iodine irrigation among 1033 patients. Ideally, the group treated with a saline rinse after Betadine irrigation would be excluded. This group cannot be considered saline irrigation patients. Making this correction greatly reduces the saline irrigation group size, from 807 patients to 466 patients. For the povidone-iodine irrigation group, Engels et al1 report numbers from 2 original studies by Burkhardt and colleagues.18,19 In their 1986 study, Burkhardt et al18 used 5 different treatments: intraprosthetic cephalothin plus saline irrigation, cephalothin plus povidone-iodine irrigation, povidone-iodine irrigation, antibiotic-steroid foam, and a saline irrigation group. The treatments were randomized by breast (not patient) so that patients received 1 treatment on 1 side and another treatment on the other side. The findings were reported by breast as opposed to by patient, in contrast with other studies. This is not a trivial difference. Burkhardt et al18 caution that for unilateral capsular contractures, the incidence reported per patient (38%) is much higher than the incidence reported per breast (23%). (Mathematically, the risk should double.) In their 1994 study, Burkhardt et al19 used a different implant for each breast—a smooth implant on one side and a textured implant on the other—in the same patient. Each breast pocket received a different irrigation solution, povidone-iodine solution on 1 side and saline solution on the other side. Nineteen women were affected among 45 patients (42% of patients, 21% of breasts). There were no bilateral capsular contractures. The difference in capsular contracture rates was attributed solely to the implant style (P < 0.01), with almost all of the capsular contractures occurring around smooth implants (40%) rather than textured devices (2%). Burkhardt et al19 reported that the antimicrobial irrigation solution did not affect the risk, contrary to the findings of the 1986 study.18 The 2 studies by Burkhardt and colleagues18,19 make any conclusions difficult in view of the different denominators (breasts versus patients) and the important confounder (textured implants) in the second study. All the implants were subglandular, a confounder known to increase the capsular contracture risk.13,14 These considerations make the advisability of pooling this data in a meta-analysis questionable. Because of the different denominators, recording 6 events out of a total of 33 in the povidone-iodine group1 is incorrect. This issue was the shortcoming of a previous review by Awad et al.20 Once the denominator was equalized (converting the denominators to breasts), a significant difference between treatment groups, favoring antimicrobial irrigation, disappeared.21 In their first study, Burkhardt et al18 included a group labeled “antibiotic-steroid foam.” A mixture of 50,000-U bacitracin, 250-mg oxacillin, and 40-mg triamcinolone was foamed in a bulb syringe and instilled into the pocket using a rubber urethral catheter. This is not a triple antibiotic solution. Obviously, triamcinolone is not an antibiotic. The steroid (not used today because of concerns regarding side effects) is a major confounder because of its potent effects on inflammation and wound healing. The 5 events among a total of 36 patients should not be included in this forest plot because this method is not triple antibiotic irrigation and, again, because the authors used a different denominator—breasts rather than patients. When this erroneous data entry is removed, only 1 event remains in the authors' triple antibiotic group, a single capsular contracture. The event total becomes 1 and the group total drops to 88. These numbers are much too small to form a reliable group for comparison. Surprisingly, the triple antibiotic group does not include the cohort of patients reported by Adams et al.22 These authors were the first to describe this irrigation regimen. Instead, this cohort is assigned to “other antibiotics.” This study was not included in the meta-analysis reported by Drinane et al2 because it lacked a control group for comparison. In evaluating infection risk, the authors reference a study by Ahmed et al.12 The forest plot shows 10 infections among 133 breast reconstruction patients who did not receive antibiotic beads. Only the control arm data were considered. In fact, Ahmed et al12 reported 17 infections among 133 control patients. Engels et al1 include a study by Dieterich et al11 that evaluated breast reduction patients, examining a possible role for hydroxyethyl starch in preventing infection (no benefit). Of course, these patients would not have breast implant pockets to irrigate and therefore this data should not be entered. Engels et al1 erroneously entered the number of capsular contractures reported by Giordano et al16 (10/167) instead of the number of infections (3/165) for the saline irrigation group, artificially inflating the number of infections in the saline irrigation group. The inclusion of this cohort is questionable in the first place because the patients did not receive saline irrigation. The incidence of capsular contracture (before correcting for errors) varies between the groups from 3.8% for the povidone-iodine group to 5.9% for the saline group.1 When the values for the 4 groups are compared using a χ2 test,23 the P value is nonsignificant (0.15). It is not possible to compare corrected values (especially after removing the false “triple antibiotic irrigation” data from Burkhardt et al)18 because the sample sizes are inadequate for a reliable comparison, particularly for the triple antibiotic group (n = 1). PUBLISHED REVIEWS Sufficient numbers are needed for a reliable meta-analysis.2,7,8 The meta-analysis by Drinane et al2 evaluated 8 studies and 10,923 patients; 5348 women received antimicrobial irrigation and 5575 patients did not. Counterintuitively, the antibiotic irrigation subgroup, the iodine subgroup, and the combined antimicrobial irrigation subgroup each had a significantly greater risk of capsular contracture than controls (P < 0.0001 for the antibiotic irrigation group and the combined group). A 2018 systematic review by Samargandi et al8 using the Methodological Index for Nonrandomized Studies and Cochrane risk-of-bias tool revealed poor methodologic quality and selection bias among a heterogeneous patient population, predisposing to an erroneous association between antibiotic irrigation and capsular contracture. In 2000, this author discontinued the practice of povidone-iodine pocket irrigation, using only sterile saline solution thereafter. There was no change in the capsular contracture rate, which remained approximately 6%. Only 2 infections were encountered among 522 consecutive breast augmentation patients treated after 2000 with saline pocket irrigation (0.4%).24 MICROBIOME ANALYSIS Interestingly, Walker et al6 determined that triple antibiotic irrigation did not reduce the bacterial abundance or affect microbial diversity compared to saline irrigation at the time of planned exchange of tissue expanders for permanent implants. The prosthesis adopted the microbial composition of the surrounding host tissue. There were no capsular contractures in either group. All patients had unilateral cancer but bilateral mastectomies. On the noncancer side, there was an unexplained increase in bacterial abundance in capsules and ADMs in patients who received triple antibiotic irrigation. Walker et al6 concluded, “these data question the value of antibiotic pocket irrigation in the face of a resilient microbial ecosystem.” These authors,6 and others, using sophisticated microbiological methods, including specialized cultures, polymerase chain reactions, scanning electron microscopy, and next-generation sequencing have failed to identify a specific microbe associated with capsular contracture.6,25–27 Microbiomes tend to reflect the host microbiome as opposed to a bacterial profile associated with an abnormality such as capsular contracture. It is not even clear that capsular contracture is linked to biofilm.25 Microbial DNA was found in only 42% of capsular contracture specimens, and 120 unique bacterial species were identified.27 Without knowledge of a responsible organism, it is difficult to make a rational case for any antimicrobial regimen.21 Interestingly, a recent systematic review and meta-analysis by Villanueva et al4 found no benefit for systemic antibiotic prophylaxis in reducing surgical site infections or capsular contracture rates after breast augmentation. In 1981, Brody and Latts28 observed that the mystery may not be why some women develop capsular contracture, but rather why so many do not develop capsular contracture in response to normal wound contraction, which is so evident in surface wounds. BIAS Engels et al1 dismiss the possibility of publication bias. However, publication bias merits consideration. The Aesthetic Surgery Journal rejected all 4 manuscripts submitted by this discussant challenging the efficacy of implant pocket irrigation. Fortunately, the articles have all been published in other major plastic surgery journals.21,24,29,30 Antibiotic irrigation is a sacred cow in plastic surgery.31 Some authors even believe the supportive evidence is so strong that it would be unethical to conduct a controlled study on this topic.32 Antibiotic irrigation is a prominent part of the (discredited) 14-point plan, promoted to reduce capsular contracture risk.29,31,33 The research arm of the Aesthetic Society has provided funding for clinical research to a proponent of triple antibiotic irrigation over a period of 25 years.34 In an impassioned, but flawed,35 rebuttal, Adams and Calobrace36 defend antibiotic irrigation, insisting that the inside of a Betadine bottle is sterile even if the outside of the bottle is not sterile. In an article co-authored by Adams,34 the authors state that Betadine has an “on-label acceptable use in breast implant pocket/preparation.” This is not true. Because it is a topical antiseptic, 10% Betadine does not require and has not gained Food and Drug Administration approval.30 Engels et al1 are evidently influenced by a bias favoring antibiotic irrigation in the plastic surgery literature. These authors do not reference previous systematic reviews on this topic that do not support this practice.2,7,8 Other recent reviews have also failed to reference the numerous publications2,7,8,21,24,29–31,37 that challenge the evidence for antimicrobial irrigation.32,38,39 Authors avoid addressing the “elephant in the room,”—the fact that Betadine is nonsterile and should never be poured into an open wound.37 The landmark microbiological study by Lineaweaver et al,40 which reported 100% cytotoxicity of 1% povidone-iodine when applied to cultured human fibroblasts, is overlooked. Evidently, Burkhardt et al18 did not read this article before introducing Betadine pocket irrigation to plastic surgery in 1986. From a scientific perspective, it is acceptable to accept or reject conflicting data; it is unacceptable to simply ignore its existence.30 ADDITIONAL RESEARCH The authors believe that antibiotic irrigation is standard practice.1 Engels et al1 recommend a prospective randomized controlled study to determine a standardized protocol. However, in surgery, a randomized controlled trial is seldom possible, practical, or even advisable.41 Patients and surgeons value their freedom to choose their treatment. Equipoise is an important ethical consideration. If a surgeon believes that antibiotic irrigation is truly effective in preventing infection and capsular contracture, that surgeon cannot ethically randomize a patient to a control group. Similarly, if a surgeon believes that antibiotic irrigation is not only unhelpful, but possibly harmful (cytotoxicity, infection risk, lack of manufacturer authorization),21,24,29–31,37 that surgeon cannot ethically allocate a patient to receive this treatment. Fortunately, well-done observational studies can provide evidence that rivals the reliability of randomized controlled trials.41 The reliability of this meta-analysis is by numerous and statistical including Plastic surgeons need to be in evaluating the scientific for before with The case for antibiotic irrigation has been by recent microbiological that to support a bacterial and by systematic reviews that are Of course, using a nonsterile in an the cannot be