Implant-based Breast Reconstruction after Mastectomy for Breast Cancer: A Systematic Review and Meta-analysis.

Women undergoing implant-based reconstruction (IBR) after mastectomy for breast cancer have numerous options, including timing of IBR relative to radiation and chemotherapy, implant materials, anatomic planes, and use of human acellular dermal matrices. We conducted a systematic review to evaluate these options.
Methods: We searched Medline, Embase, Cochrane CENTRAL, CINAHL, and ClinicalTrials.gov for studies, from inception to March 23, 2021, without language restriction. We assessed risk of bias and strength of evidence (SoE) using standard methods.
Results: We screened 15,936 citations. Thirty-six mostly high or moderate risk of bias studies (48,419 patients) met criteria. Timing of IBR before or after radiation may result in comparable physical, psychosocial, and sexual well-being, and satisfaction with breasts (all low SoE), and probably comparable risks of implant failure/loss or explantation (moderate SoE). No studies addressed timing relative to chemotherapy. Silicone and saline implants may result in clinically comparable satisfaction with breasts (low SoE). Whether the implant is in the prepectoral or total submuscular plane may not impact risk of infections (low SoE). Acellular dermal matrix use probably increases the risk of implant failure/loss or need for explant surgery (moderate SoE) and may increase the risk of infections (low SoE). Risks of seroma and unplanned repeat surgeries for revision are probably comparable (moderate SoE), and risk of necrosis may be comparable with or without human acellular dermal matrices (low SoE). Conclusions: Evidence regarding IBR options is mostly of low SoE. New high-quality research is needed, especially for timing, implant materials, and anatomic planes of implant placement.

Takeaways

Question: What are the comparative benefits and harms of various timing, materials, anatomic planes, and human acellular dermal matrix (ADM) use options for implant-based reconstruction (IBR)?

Findings: In a large systematic review and meta-analysis, 36 studies met criteria. Timing IBR before/after radiation results in comparable physical/psychosocial/sexual well-being, satisfaction with breasts, and risk of implant failure/loss (no studies addressed timing relative to chemotherapy). Silicone/saline implants have comparable satisfaction with breasts. Prepectoral/total submuscular implants have similar risks of infections. ADMs increase risk of implant failure/loss and infections, but risks of seroma, unplanned repeat surgeries, and necrosis are comparable.

Meaning: Evidence regarding IBR options is of low strength.

INTRODUCTION

More than 40% of US women who undergo mastectomy for breast cancer have breast reconstruction,[1] amounting to about 107,000 women in 2019.[2] Most reconstruction procedures in the United States (81%) are implant-based.[2] Considerations for implant-based reconstruction (IBR) include procedure timing relative to chemotherapy and radiation, implant material (eg, silicone, saline, double-lumen), anatomic plane (prepectoral, partial submuscular, or total submuscular), and use of an adjunctive human acellular dermal matrix (ADM). Each consideration can impact aesthetics, complications, and cost.

Objectives

We conducted a systematic review (SR) for the Agency for Healthcare Research and Quality (AHRQ) to support the American Society of Plastic Surgeons in development of a new clinical practice guideline on breast reconstruction after mastectomy.[3] Here, we focus on the research questions concerning IBR. Other articles in this issue focus on autologous reconstruction[4] and the comparison between IBR and autologous reconstruction.[5] All reports focus on women who are undergoing (or who have undergone) mastectomy for breast cancer treatment or prophylaxis. Here, we evaluate the comparative benefits and harms of (1) timing relative to chemotherapy and radiation, (2) implant materials, (3) implant placement planes, and (4) use of human ADMs. We evaluate whether outcomes varied by age, breast cancer stage, occurrence (first/recurrent), chemotherapy/radiation type, timing (immediate/delayed), number of stages (single/multiple), laterality (unilateral/bilateral), implant surface (smooth/textured), implant shape (round/teardrop), and implant size.

METHODS

We used standard SR methodology as outlined in AHRQ’s Methods Guide.[6] We refined the research questions and protocol after discussions with groups of experts. We registered the SR protocol through PROSPERO (CRD42020193183).

Search Strategy

We searched for published studies in Medline (via PubMed), Embase, the Cochrane Central Register of Controlled Trials (CENTRAL), and CINAHL, and for unpublished studies in ClinicalTrials.gov. The searches (for the full SR) included terms related to breast cancer, mastectomy, implants, ADM, and autologous reconstruction. (See table 1, Supplemental Digital Content 1, which displays the search strategies. http://links.lww.com/PRSGO/B950.)

No date or language restrictions were applied. All searches are current as of March 23, 2021. We also scanned the reference lists of available SRs for potentially eligible studies.

Study Selection

Seven investigators independently screened each title and abstract using Abstrackr (http://abstrackr.cebm.brown.edu/). All accepted citations were rescreened in duplicate in full text. At both stages, discrepancies were resolved through discussion and/or consultation with a third investigator.

We included studies of adult women (aged ≥18 years) who had undergone mastectomy for breast cancer or carcinoma in situ (or for cancer prophylaxis) and had IBR. Here, we focus on outcomes prioritized by stakeholder panels. (See table 2, Supplemental Digital Content 2, which displays the outcomes of interest for each research question. http://links.lww.com/PRSGO/B951.)

Additional outcomes are reported in the full report.[3] For benefit outcomes, we included randomized controlled trials (RCTs) with 10 or more patients per group and prospective/retrospective nonrandomized comparative studies (NRCSs) with adequate statistical adjustment analyses and 30 or more patients per group. For surgical complications, we also included single-group studies with 500 patients or more.

Risk of Bias Assessment and Data Extraction

For each study, one investigator assessed risk of bias and extracted data into the Systematic Review Data Repository Plus (http://srdrplus.ahrq.gov/). All extractions were verified by a second investigator. We used questions from the Cochrane Risk of Bias,[7] Risk of Bias in Nonrandomized Studies of Interventions,[8] and National Heart, Lung, and Blood Institute Quality Assessment[2] tools.

Syntheses

For dichotomous outcomes, we preferentially evaluated odds ratios (ORs). For continuous outcomes, we evaluated net mean differences (NMDs) (difference-in-differences) for outcomes measured at both baseline and postintervention, or mean differences (MDs) for outcomes measured only postintervention. When appropriate, we estimated these based on reported data. When feasible, for continuous outcomes, we made conclusions based on published estimates of minimal clinically important differences (MCIDs). For NRCSs, we considered only reported adjusted analyses. Where there were three or more studies reporting results from similar analyses, we conducted pairwise meta-analyses using random-effects models in Stata.

Strength of Evidence (SoE) Assessment

We graded SoE as per the AHRQ Methods Guide.[6] We considered risk of bias, consistency, precision, directness, and sparsity. For each prioritized outcome, we assigned a SoE rating of high, moderate, low, or insufficient. Grades of high, moderate, and low indicate the degree of confidence we have that the estimate lies close to the true effect; an insufficient rating indicates that the evidence precluded estimation of an effect.[6] In accordance with AHRQ guidance,[9,10] we use qualifying language regarding SoE when communicating conclusions: “probably” for moderate SoE and “may” for low SoE.

RESULTS

For the full SR, our searches yielded 15,936 citations (Fig. 1). We screened 1352 full-text articles, of which 36 were eligible for the research questions described in this article.

Fig. 1.

PRISMA diagram depicting identification of studies in this SR.

Characteristics of Included Evidence

Published between 2005 and 2021, the 36 included studies comprised three RCTs[11-13] and 33 NRCSs‚ with adequate statistical adjustment analyses,[14-56] with a total of 48,419 women (Table 1). Twenty-three studies (64%) were from the United States, four (11%) from South Korea, three (8%) from Canada, and three (8%) from the United States and Canada. One study each (3%) was from Italy, Sweden, and Turkey.

Table 1.

Summary of Design, Arm, and Sample Details

Most studies (72%–94%) did not report participant age, race, or body mass index (BMI) for the entire study population. Where reported for the entire population, average patient ages ranged from 46.2 to 51.2 years (12 studies) and average BMIs from 22.3 to 27.0 kg/m2 (nine studies). In two studies,[27,30,57] 79% and 94% of patients were White, and 6.4% and 1.3% were Black. In the one study with data,[28] all patients were treated for their first occurrence of breast cancer. In the two studies with data on reasons for mastectomy, one reported that 90% of mastectomies were therapeutic and 10% prophylactic,[25,51] whereas the other reported that 44% were therapeutic and 56% prophylactic.[43]

Risk of Bias

Two of the three RCTs had a moderate risk of bias and one had a high risk. (See table 3, Supplemental Digital Content 3, which displays the risk of bias assessments. http://links.lww.com/PRSGO/B952.)

The primary concerns about bias in the RCTs were lack of blinding of participants and care providers, evidence of selective outcome reporting, and incompleteness of outcome data. Among the 33 NRCSs, 26 had a high risk of bias, six moderate risk, and one low risk. The primary concerns about bias in the NRCSs were evidence of serious risk of confounding and lack of blinding of outcome assessors.

Timing Relative to Chemotherapy and Radiation

No eligible studies evaluated timing relative to of chemotherapy. Five NRCSs, reported in 10 articles,[21,23,26,44,45,47-49,52,56] and no RCTs evaluated timing relative to radiation in 2834 patients (between 130 and 1143 patients each) (Table 1). Four NRCSs were at a high and one at a moderate risk of bias. Table 2 summarizes the evidence for all comparisons in the review.

Table 2.

Summary of Evidence for all Research Questions

Benefit Outcomes: Two NRCSs (Yoon et al[56] and Cordeiro et al[21]) compared IBR before versus after radiation and reported comparable well-being and satisfaction using subscales of the BREAST-Q (each scored 0–100; higher scores indicate better outcomes). (See table 4, Supplemental Digital Content 4, which displays summary tables. http://links.lww.com/PRSGO/B953.)

Yoon et al[56] reported an adjusted MD (adjMD) of −0.64 [95% confidence interval (CI) −7.19 to 5.90) for physical well-being (MCID = 3[58]), 0.48 (95% CI −7.72 to 8.68) for psychosocial well-being (MCID = 4[58]), −1.00 (95% CI −8.41 to 6.40) for sexual well-being (MCID = 5[58]), and −3.89 (95% CI −11.0 to 3.23) for satisfaction with breasts (MCID = 5[58]). Cordeiro et al[21] did not report adjusted effect sizes but reported no statistically significant between-group differences for physical well-being and satisfaction with breasts. For psychosocial well-being and sexual well-being, Cordeiro et al[21] reported on only statistical significance of MDs (P < 0.01); the unadjusted MDs (−1.2 for psychosocial well-being and −1.4 for sexual well-being) were smaller than their MCIDs. Likewise, Cordeiro et al[21] reported a statistically significant different adjMD for satisfaction with surgical outcome (P = 0.02), but the unadjusted MD (−1.8) was less than the MCID (5[58]).

Surgical complications: Four NRCSs reported on surgical complications, which were generally comparable regardless of timing. Three NRCSs reported on the risk of implant loss/failure or need for explantation at 3.3–3.6 years. Effect sizes ranged from a statistically significant 0.62, favoring before radiation, to a nonsignificant 1.12, yielding a summary effect size of 0.87 (95% CI 0.62–1.24; I2 = 54%) (Fig. 2).

Fig. 2.

Meta-analysis for timing of IBR relative to radiation (Outcome: Implant failure/loss or need for explant surgery). Abbreviations: adj = adjusted, CI = confidence interval, HR = hazard ratio = IBR = implant-based reconstruction, I2 = measure of statistical heterogeneity (% of total variability that is due to between-study variability), NR = not reported, OR = odds ratio, RTX = radiation therapy, y = years.

One NRCS (Yoon et al[56]) reported that 2-year followup data for pain interference (using the Patient-Reported Outcomes Measurement Information System; 100-point scale; higher is better; MCID = 4.5[59]) were comparable irrespective of whether before or after radiation (adjMD = 2.86, 95% CI −1.05 to 6.77) (See table 4-1, Supplemental Digital Content 4, http://links.lww.com/PRSGO/B953). Although no adjusted effect sizes were reported, Yoon et al[56] reported that 2-year risks of five other complications were also comparable between treatment groups: major infections (requiring treatment with intravenous antibiotics) (P = 0.40), minor infections (treated with oral antibiotics) (P = 0.96), wound dehiscence (P = 0.32), seroma (P = 0.46), and capsular contracture (P = 0.80) (See table 4-2, Supplemental Digital Content 4, http://links.lww.com/PRSGO/B953). One NRCS (Eriksson et al[23]) reported comparable risks of unplanned repeat surgeries for revision [adjusted hazard ratio (adjHR) = 0.94, 95% CI 0.63–1.40]. Another NRCS (Hirsch et al[26]) reported comparable risks of necrosis [adjusted OR (adjOR) = 0.96, 95% CI 0.68 to 1.35].

Materials

Five NRCSs,[14,21,30,35,36] but no RCTs, compared implant materials in 2929 patients (between 143 and 1143 patients each) (Table 1). All five NRCSs had a high risk of bias. In Le et al[30] (in USA), the large majority (94%) were White, and in Macadam et al[35] (in Vancouver, Canada), a majority (66%) were Asian; the other studies did not report on race.

Silicone versus Saline Implants

Benefit/Clinical Outcomes: Macadam et al[35] used the European Organization for Research and Treatment of Cancer Quality of Life Questionnaire C30 to report on general quality of life and BREAST-Q to report on physical well-being, psychosocial well-being, sexual well-being, and satisfaction with outcome (See table 4-3, Supplemental Digital Content 4, http://links.lww.com/PRSGO/B953). There was a statistically significant difference in psychosocial well-being (P = 0.03), but not for the other outcomes (P = 0.13 for quality-of-life, P = 0.28 for physical well-being, P = 0.056 for sexual well-being, and P = 0.082 for satisfaction with outcome). No adjusted effect sizes were reported.

Two NRCSs (Macadam et al[35] and McCarthy et al[36]) reported on satisfaction with breasts (using the BREAST-Q; MCID = 5[58]). McCarthy et al[36] reported clinically comparable satisfaction at 2.4–3.3 years (adjMD = 4.1, 95% CI 1.31–6.89). Macadam et al[36] reported a statistically significant between-group difference (P = 0.008), but no adjusted effect size was reported.

One NRCS (Le et al[30]) reported comparable risks of breast cancer mortality (adjHR = 1.01, 95% CI 0.44–2.34) and nonbreast cancer mortality (adjHR = 1.75, 95% CI 0.29–10.34) between silicone and saline groups at 12.4 years of followup (See table 4-4, Supplemental Digital Content 4, http://links.lww.com/PRSGO/B953 ).

Surgical Complications: Two NRCSs (Cordeiro et al[21] and Antony et al[14]) reported on surgical complications (See table 4-4, Supplemental Digital Content 4, http://links.lww.com/PRSGO/B953). Cordeiro et al[21] reported that risk of implant failure/loss was lower among patients with silicone implants (adjOR = 0.61, 95% CI 0.36–1.07). Antony et al[14] reported no statistically significant difference in risks of capsular contracture, but no adjusted effect size or P value was reported.

Double-lumen Implants

One NRCS (Le et al[30]) reported comparable risks of breast cancer mortality at 12.4 years of follow-up between silicone and double-lumen implants (adjHR = 1.49, 95% CI 0.83–2.70) (See table 4-4, Supplemental Digital Content 4, http://links.lww.com/PRSGO/B953.) However, nonbreast cancer mortality was higher among patients with double-lumen implants (adjHR = 3.13, 95% CI 0.91–10.78). No study addressed surgical complications. No study compared saline and double-lumen implants.

Anatomic Planes

We included eight studies (one RCT,[11] and seven NRCSs reported in eight articles[15,16,18,24,28,29,38,39]) that compared prepectoral, partial submuscular, and total submuscular planes of implant placement in 1555 patients. We rated the RCT as having a moderate risk of bias, six of the seven NRCSs at a high risk, and one NRCS at a moderate risk. No study reported on race. Studies followed patients between 6 months and 6.1 years.

Prepectoral versus Total Submuscular Placement of Implants

Benefit Outcomes: One NRCS (Cattelani et al[18]) reported on physical well-being using both the Constant Murley score at 7 days and the Disabilities of the Arm, Shoulder, and Hand instrument at 1 year; psychosocial well-being based on number of days until return to usual work; and satisfaction with breasts using the BREAST-Q (See table 4-5, Supplemental Digital Content 4, http://links.lww.com/PRSGO/B953.) Although no adjusted effect sizes were reported, patients with prepectoral implants fared statistically significantly better than patients with total submuscular implants (adjusted P < 0.001 for each outcome).

Surgical Complications: Two NRCSs (Nealon et al[37] and Kraenzlin et al[29]) reported comparable risks of infections. (See table 4-6, Supplemental Digital Content 4, http://links.lww.com/PRSGO/B953.) Nealon et al[37] reported an imprecise adjOR of 0.31 (95% CI <0.01–8.65) and Kraenzlin et al[29] reported a P value of 0.21 (no adjusted effect size reported).

Although no adjusted effect sizes were reported, two NRCSs (Avila et al[15] and Cattelani et al[18]) reported inconsistent results regarding pain (See table 4-5, Supplemental Digital Content 4, http://links.lww.com/PRSGO/B953). Avila et al[15] used a visual analog scale (VAS) and reported no statistically significantly difference in pain levels. However, Cattelani et al[18] used the Brief Pain Inventory-Short Form (BPI-SF) and reported that patients with total submuscular implants had clinically and statistically significantly lower pain 7 days after surgery (P < 0.001; no adjusted effect size reported).

Other complications were reported by one NRCS each. Avila et al[15] reported lower analgesic use with prepectoral implants (P = 0.03; no adjusted effect size reported). Avila et al[15] also reported comparable risk of unplanned repeat surgeries for revision (P = NS), although no adjusted effect size was reported. Nealon et al[37] reported comparable risks of necrosis (adjOR = 1.01, 95% CI 0.74–5.95), explantation (adjOR = 1.01, 95% CI 0.07 to 14.1), capsular contracture (adjOR = 0.30, 95% CI 0.03 to 1.55), and seroma (adjOR = 1.49, 95% CI 0.37–6.11).

Prepectoral versus Partial Submuscular Placement of Implants

Benefit Outcomes: One RCT (Lee et al[11]) reported that patients with prepectoral or partial submuscular implants had comparable physical well-being measured using the physical component of the SF-36 (0–100; MCID not available) (MD = 0.0, 95% CI −5.0 to 5.0) and comparable psychosocial well-being using the anxiety and depression components of the Hospital Anxiety and Depression Scale (0–21 scale; MCIDs not available) (anxiety MD = 0.0, 95% CI −7.5 to 7.5; depression MD = 1.2, 95% CI −3.2 to 5.6) (See table 4-4, Supplemental Digital Content 4, http://links.lww.com/PRSGO/B953).

Surgical Complications: Specific complications were reported by one study each. One NRCS (Kim and Hong[28]) reported comparable pain using the VAS (0–10; MCID = 2[60]) (adjMD = −0.12; P = 0.12) (See table 4-5, Supplemental Digital Content 4, http://links.lww.com/PRSGO/B953). The RCT (Lee et al[11]) reported comparable risks of seroma (OR = 1.06, 95% 0.15–7.34) and capsular contracture (5% versus 0%; effect size not calculable). (See table 4-6, Supplemental Digital Content 4, http://links.lww.com/PRSGO/B953 .)

Use versus Nonuse of ADMs

We included 22 studies (two RCTs,[12,13] and 20 NRCSs reported in 29 articles[12,13,17-20,22,25-27,31-34,37,40-43,45-48,50-55]) of human ADM use in 43,334 patients (between 36 and 18,977 patients each) (Table 1). Among the 14 of 22 studies that reported funding information, eight explicitly stated that they were not funded, five were funded by nonindustry sources (eg, federal sources, foundations), and one was funded by industry (Lifecell Corporation).

One RCT had a high risk of bias and the other moderate risk. Fifteen NRCSs had a high risk of bias, four moderate risk, and one low risk. Studies followed patients between 2 and 5 years.

Benefit/Clinical Outcomes: One RCT and two NRCSs reported on benefit/clinical outcomes (See table 4-7, Supplemental Digital Content 4, http://links.lww.com/PRSGO/B953). All three studies reported on physical well-being, but the results were inconsistent. Comparable physical well-being (measured by the BREAST-Q; MCID = 3[58]) regardless of ADM use was reported by one RCT (McCarthy et al[12]: NMD = 0.50, 95% CI −5.93 to 6.93) and one NRCS (Ganesh Kumar et al[25]: adjMD = −0.82, 95% CI −3.01 to 1.37). However, the other NRCS (Cattelani et al[18]) reported that patients with ADMs had better unadjusted Constant Murley and Disabilities of the Arm, Shoulder, and Hand (DASH) physical well-being scores (P < 0.001 for both); no adjusted effect sizes were reported.

Two NRCSs reported inconsistent results for psychosocial well-being. Ganesh Kumar et al[25] reported comparable BREAST-Q scores (MCID = 4[58]) regardless of ADM use (adjMD = −0.26, 95% CI −2.97 to 2.45). On the other hand, Cattelani et al[18] reported that patients with ADMs returned to work considerably sooner than who had not (mean 35 versus 57 days, P < 0.001). The same two NRCSs also reported inconsistent results on sexual well-being using the BREAST-Q (MCID 5 points[58]). Ganesh Kumar et al[25] reported comparable scores (adjMD = −1.95, 95% CI −4.96 to 1.06), but Cattelani et al,[18] without mentioning an adjusted effect size, reported that patients with ADMs had considerably higher unadjusted scores (P < 0.001).

One NRCS (Ganesh Kumar et al[25]) reported on sexual well-being using the BREAST-Q (MCID = 5[58]), which was comparable with or without ADM use (adjMD = −2.28, 95% CI −5.63 to 1.06).

Surgical Complications: All 22 studies reported on surgical complications.

Across six NRCSs, the summary adjOR for implant failure/loss or need for explantation was 1.28 (95% CI 0.97–1.70; I2 = 16%) (Fig. 3A). Across seven studies (two RCTs and five NRCSs), the summary adjOR for infections was 1.56 (95% CI 0.96–2.53; I2 = 46%) (Fig. 3B), with similar findings among the RCTs and the NRCSs (P = 0.44, based on a meta-regression). Across four NRCSs, the summary adjOR for necrosis was 0.89 (95% CI 0.63–1.25; I2 = 25%) (Fig. 4A). Across four studies (one RCT and three NRCSs), the summary adjOR for seroma was 1.52 (95% CI 0.62–3.71; I2 = 52%) (Fig. 4B), with no significant difference between the RCT and the NRCSs (P = 0.30, based on a meta-regression). Other studies that reported on these outcomes did not report sufficient data for inclusion in meta-analyses.

Fig. 3.

Meta-analyses for ADM use during IBR: A, Outcome: implant failure/loss or need for explant surgery. B, Outcome: infections. Abbreviations: ADM = acellular dermal matrix, CI = confidence interval, IBR = implant-based reconstruction, I2 = measure of statistical heterogeneity (% of total variability that is due to between-study variability), mo = months, NR = not reported, NRCS = nonrandomized comparative study, OR = odds ratio, RCT = randomized controlled trial, y = years.

Fig. 4.

Meta-analyses for ADM use during IBR: A, Outcome: necrosis. B, Outcome: Seroma. Abbreviations: ADM = acellular dermal matrix, CI = confidence interval, IBR = implant-based reconstruction, I2 = measure of statistical heterogeneity (% of total variability that is due to between-study variability), NR = not reported, OR = odds ratio.

Three NRCSs reported comparable risks of unplanned repeat surgeries for revision of reconstruction (See table 4-8, Supplemental Digital Content 4, http://links.lww.com/PRSGO/B953), but insufficient data were reported to allow meta-analysis. Ibrahim et al[27] reported that, at 6 months, risks were comparable regardless of whether ADMs were used (P = 0.14; no adjusted effect size reported). At approximately 5 years, no significant between-group differences in risk of unplanned repeat surgeries were reported by Nealon et al[38] (adjOR = 0.86, 95% CI 0.69–1.08) and Sobti et al[50] (adjOR = 1.10, 95% CI 0.63–1.92).

Results were inconsistent across studies for various complications. One RCT and one NRCS reported on pain (See table 4-9, Supplemental Digital Content 4, http://links.lww.com/PRSGO/B953). The RCT (McCarthy et al[12]) reported that ADM use was associated with greater pain in the first 24 hours (VAS 0–100 scale: NMD = 6.2, 95% CI −4.9 to 17.3; MCID = 5[61]) and during expansion (NMD = 6.8, 95% CI 1.1–12.5) but not after expansion (NMD = −4.6, 95% CI −9.8 to 0.6). However, the NRCS (Cattelani et al[18]) reported statistically significantly less pain in the ADM group 7 days after surgery on the BPI-SF scale (P < 0.001; no adjusted effect size reported).

Two NRCSs reported inconsistent results on implant malposition (See table 4-8, Supplemental Digital Content 4, http://links.lww.com/PRSGO/B953). Ganesh Kumar et al[25] reported that risks were comparable (P = 0.83; no adjusted effect size reported), but Vardanian et al[53] reported that ADM use was associated with a lower risk (adjOR = 0.23, 95% CI 0.06–0.78).

Four NRCSs reported inconsistent results on capsular contracture (See table 4-8, Supplemental Digital Content 4, http://links.lww.com/PRSGO/B953). Three NRCSs reported comparable risks (Ganesh Kumar et al[25]: P = 0.24; Nealon et al[38]: adjOR = 0.78, 95% CI 0.46–1.36; and Sobti et al[50]: adjOR = 0.57, 95% CI 0.23–1.43). However, Vardanian et al[53] reported that ADM use was associated with a lower risk (adjOR = 0.18, 95% CI 0.08–0.43). We do not report a meta-analysis for this outcome due to substantial statistical heterogeneity (ie, marked between-study variability in results, as suggested by an I2 of 85%).

Four NRCSs reported inconsistent results on wound dehiscence. (See table 4-8, Supplemental Digital Content 4, http://links.lww.com/PRSGO/B953.) ADM use was associated with a greater risk in one NRCS (Ganesh Kumar et al[25]: P = 0.009), a comparable risk in another NRCS (Ibrahim et al[27]: P = 0.26), and a lower risk in a third NRCS (Qureshi et al[42]: adjOR = 0.4; P < 0.05). The fourth NRCS (Craig et al[22]) reported adjusted data only for the subgroup of patients who did not receive postoperative radiation; ADM use was associated with a greater risk (adjOR 2.46, 95% CI 1.23–4.93).

Various complications were reported by one study each. One RCT (McCarthy et al[12]) reported comparable analgesic use within the first 24 hours (MD = −134 mg oral codeine equivalents, 95% CI −440 to 172) (See table 4-9, Supplemental Digital Content 4, http://links.lww.com/PRSGO/B953). One NRCS (Woo et al[55]) reported comparable risk of delayed healing (adjOR = 1.41, 95% CI 0.67–2.96) (See table 4-8, Supplemental Digital Content 4, http://links.lww.com/PRSGO/B953). One NRCS each reported no statistically significant differences in risk of thromboembolic events (Ibrahim et al[27]) or implant rupture (Ganesh Kumar et al[25]), although no adjusted effect sizes were reported. However, Peled et al[41] reported a statistically significant lower risk of unplanned repeat surgeries for complications in patients with ADM use (P < 0.05), but no adjusted effect size was reported.

DISCUSSION

The current evidence does not suggest clearly preferred modalities for IBR after breast-cancer-related mastectomy. We found no evidence regarding timing vis-a-vis chemotherapy. This may be related to the preference of clinicians to base decisions regarding timing of chemotherapy on the severity of the underlying cancer. Limited evidence suggests that timing before or after radiation may not affect physical well-being, psychosocial well-being, sexual well-being, and patient satisfaction with breasts, and probably does not affect implant failure/loss or explantation. Weak evidence suggests clinically comparable satisfaction with breasts after silicone or saline implants, but the evidence is insufficient to make conclusions comparing surgical complications. There is insufficient evidence regarding double-lumen implants. Evidence was also largely insufficient regarding choice of anatomic plane of implant placement. The evidence is also weak for whether the implant should be placed in the prepectoral, total submuscular, or partial submuscular planes. However, prepectoral and total submuscular placements may be associated with comparable risks of infections. Regarding ADMs, there is insufficient evidence whether they impact patient-reported outcomes. However, ADM use may be associated with some surgical complications, such as infections and implant failure, but not others, such as necrosis and seroma. Our findings that ADM use may be associated with infections and implant failure are consistent with a recent U.S. Food and Drug Administration Safety Communication regarding ADM use.[62]

Given the relatively weak evidence addressing some key decisions in clinical practice and the highly patient preference-sensitive nature of the decisions,[63,64] we encourage clinicians to inform patients about the limitations of existing research. Among the limitations is that very little research has focused on patients whose mastectomy was performed for prophylactic (and not therapeutic) purposes. Therefore, the patient’s values and preferences and the clinician’s expertise and experience are highly important.

Strengths and Limitations

We followed contemporary methodological standards for SRs, including multi-stakeholder engagement and use of state-of-the-art methods for searching, screening, assessing risk of bias, extracting and synthesizing data, and assessing SoE.

A few limitations to the evidence base are worth noting. Only three of 36 included studies were RCTs, each small. Most studies were at a moderate or high risk of bias, primarily because participants, care providers, and/or outcome assessors were not blinded, and/or outcome data were incomplete. Studies commonly reported incomplete data regarding adjusted analyses, often reporting only adjusted P values without adjusted effect sizes. Furthermore, comparisons of subgroups were limited in that none of the studies reported statistical analyses of differences between subgroups or, what would have been preferable, evidence of treatment effect heterogeneity. Finally, 80% of studies were conducted in North America (USA or Canada), with some studies from South Korea or Europe. It is unclear to what extent the evidence applies to populations that are not mostly White, middle-aged, nonobese women located in North America. However, the interventions examined in the studies are mostly reflective of available interventions in the USA, such as silicone and saline implants, human ADMs, and prepectoral and total submuscular placements of implants.

Implications for Research

Research is needed to address various questions, especially timing, materials, and anatomic planes. Given the recent increase in prophylactic mastectomies[65-69] and because the risk-benefit tradeoffs may be different from those for women undergoing therapeutic mastectomies, future studies should enroll, and separately report data for, women undergoing prophylactic mastectomies. In addition, studies should enroll more diverse groups of women, particularly by race, ethnicity, age, and socioeconomic position.

It is also important that, when possible, future studies conduct randomization to avoid selection bias. If randomization is not feasible or practical, as may often be the case for surgical topics,[70] studies (such as those using data from the Tracking Operations and Outcomes for Plastic Surgeons registry[71]) should fully report between-group estimates of treatment effect that conduct adequate statistical adjustment analyses to account for important confounders, including at least age, race/ethnicity, weight, and breast cancer stage. Ideally, propensity score analyses (or similar rigorous techniques) should be used to adequately adjust for confounders. Future studies should also evaluate important outcomes that are not sufficiently reported in the identified evidence, including quality of life, number of planned surgeries for reconstruction, incidence and duration of unplanned repeat hospitalizations and surgeries, analgesic use, animation deformity, and complications that may delay other cancer-related treatments.

CONCLUSIONS

The current evidence base allows few conclusions, tempered by the low-to-moderate SoE, for the comparative benefits and harms of IBR-related modalities for women who have undergone mastectomy for breast cancer. IBR before or after radiation may result in comparable benefit outcomes and probably results in a comparable risk of implant failure/loss or explantation. Silicone or saline implants may result in comparable patient satisfaction with breasts, but the evidence for surgical complications is insufficient. Whether the implant is placed in the prepectoral or total submuscular plane may result in comparable risk of infections, but the evidence for beneficial outcomes is insufficient. Regarding human ADM use, the evidence for beneficial outcomes is insufficient, but its use may be associated with greater risks of implant failure/loss or explantation and infections but comparable risks of necrosis and seroma. More research is needed to identify effective and safe surgical options for IBR for women who have undergone mastectomy for treatment or prophylaxis against breast cancer.

ACKNOWLEDGMENTS

We thank Binita Ashar, MD; Katelyn Donnelly, MPH; Phyllis Greenberger, MSW; Michele Manahan, MD; Priscilla McAuliffe, MD, PhD; Steven Nagel, MD; William Sikov, MD; Terence Myckatyn, MD; Anne Taylor, MD; Myelin Torres, MD; Edwin Wilkins, MD; and Sung Yoon, MD, who served as key informants and/or technical experts and helped us refine the research questions and develop the protocol. We are also grateful to the project’s Task Order Officer Jill Huppert, MD, MPH and the Acting EPC Program Director Christine Chang, MD, MPH at the Agency for Healthcare Research and Quality, Rockville, Maryland, and to the full systematic review report’s peer reviewers and Associate Editor Timothy Wilt, MD, MPH, at the Minnesota Evidence-based Practice Center, Minneapolis, Minnesota. We also thank Kristin Konnyu, MSc, PhD and Jonah Popp, PhD for their careful study screening and related contributions to the project. The authors of this study are responsible for its content. Statements in the report should not be construed as endorsement by AHRQ or Department of Health and Human Services. AHRQ retains a license to display, reproduce, and distribute the data and the report from which this article was derived under the terms of the agency’s contract with the author. This systematic review was prospectively registered through PROSPERO (Registration Number: CRD42020193183).