The following clinical outcome measures were included: 1) operating time, 2) intraoperative blood loss, 3) postoperative drainage volume of the chest wall, 4) postoperative hospital length of stay, 5) total complications, 6) overall seroma, 7) hematoma, 8) wound infection, 9) necrosis, and 10) ecchymosis. The seroma outcome pooled studies reporting overall rates of seroma, as well as studies reporting seroma results specific to the chest and/or axillary regions only. The mean and standard deviation (SD) variance measure was not reported in two studies5,7 for the operating time and intraoperative blood loss outcomes. Study authors were not contacted to retrieve missing data; however, standard methods provided by Cochrane24 and methods outlined by Hozo et al25 were used to impute missing data. Additionally, the SD was not provided in one study26 for operating time; thus, the reported P-value was used to impute this variance measure. Similarly, one study4 did not report the SD variance measure for the postoperative drainage volume from the chest wall outcome, and the P-value was used to impute this missing variance measure. This same study4 did not report the mean and SD variance measure for the postoperative hospital length of stay outcome. Methods provided by Cochrane24 and outlined by Hozo et al25 were used to impute these missing data. Due to missing variance measures, the mean number of lymph nodes dissected was not included as an outcome in this analysis.
The quality of the included studies was assessed using the risk of bias algorithm outlined by the Cochrane guidelines.24 Based on seven prespecified domains (sequence generation, allocation concealment, blinding of participants and personnel, blinding of outcome assessment, incomplete outcome data, selective outcome reporting, and other sources of bias), publications were scored as having low, unclear, or high risk of bias. Final decisions were based on the combination of these factors and individual characteristics of each study. Two authors independently assessed the quality of the study. Differences were resolved through discussion with a third author.
The primary meta-analysis was performed using Review Manager (Version 5.3; The Nordic Cochrane Centre, The Cochrane Collaboration, Copenhagen, Denmark, 2014). Mean differences (MDs) for continuous outcomes (operating time, intraoperative blood loss, postoperative drainage volume of the chest wall, and postoperative hospital length of stay) were calculated using the inverse variance method. The meta-analysis of continuous outcomes used a random-effects model, due to the presence of some study heterogeneity. Risk ratios (RRs) for dichotomous outcomes (total complications, overall seroma, hematoma, wound infection, necrosis, and ecchymosis) were calculated using the Mantel–Haenszel method. A fixed-effects model was used for the meta-analysis of dichotomous outcomes, as several outcome events were rare and displayed low heterogeneity. Forest plots were generated for all outcomes within Review Manager. Heterogeneity of the included studies was assessed using the χ2 test and I2 measure.
The primary analysis compared Harmonic technology to conventional techniques and presented results for four different subgroups of studies, depending on whether studies only included mastectomy, only included BCS, included both mastectomy and BCS, and whether lymph node dissection was performed. A secondary analysis was conducted for dichotomous outcomes, where the odds ratios (ORs) were calculated using the Mantel–Haenszel method. Dichotomous outcomes that were considered rare events (ie, with event rates <1%: hematoma, wound infection, necrosis) were analyzed using the Peto one-step OR method with a fixed-effects model.27 Sensitivity analyses were completed for device-use randomization in included studies, where studies not randomizing device use in breast surgery were excluded28 or studies not randomizing device use in lymph node dissection were excluded.5,29 Further, sensitivity analyses were conducted by excluding any studies from outcomes for which imputed data were required or excluding studies deemed to have a high risk of bias. Forest plots are provided in the main text for results that are statistically significant and when four or more studies informed the analysis.
The systematic search resulted in the identification of 508 citations, and 26 additional studies were identified through manual searches of accepted studies and published systematic reviews. Of the 508 citations identified in the search, 333 were further excluded following abstract screening (Figure 1). Of the 199 full-text articles retrieved and reviewed, 189 were further excluded primarily if studies were non-RCTs, there was no explicit description of a Harmonic branded device, the surgical procedure was not breast surgery or lymph node dissection, and only conference abstracts were available. Overall, 12 studies consisting of 890 patients in total reporting on Harmonic technology use in breast surgery and lymph node dissection procedures were included in the meta-analysis.4–8,26,28–33