Immediate and delayed autologous abdominal microvascular flap breast reconstruction in patients receiving adjuvant, neoadjuvant or no radiotherapy: a meta‐analysis of clinical and quality‐of‐life outcomes

Background Effects of postmastectomy radiotherapy (PMRT) on autologous breast reconstruction (BRR) are controversial regarding surgical complications, cosmetic appearance and quality of life (QOL). This systematic review evaluated these outcomes after abdominal free flap reconstruction in patients undergoing postoperative adjuvant radiotherapy (PMRT), preoperative radiotherapy (neoadjuvant radiotherapy) and no radiotherapy, aiming to establish evidence‐based optimal timings for radiotherapy and BRR to guide contemporary management. Methods The study was registered on PROSPERO (CRD42017077945). Embase, MEDLINE, Google Scholar, CENTRAL, Science Citation Index and ClinicalTrials.gov were searched (January 2000 to August 2018). Study quality and risk of bias were assessed using GRADE and Cochrane's ROBINS‐I respectively. Results Some 12 studies were identified, involving 1756 patients (350 PMRT, 683 no radiotherapy and 723 neoadjuvant radiotherapy), with a mean follow‐up of 27·1 (range 12·0–54·0) months for those having PMRT, 16·8 (1·0–50·3) months for neoadjuvant radiotherapy, and 18·3 (1·0–48·7) months for no radiotherapy. Three prospective and nine retrospective cohorts were included. There were no randomized studies. Five comparative radiotherapy studies evaluated PMRT and four assessed neoadjuvant radiotherapy. Studies were of low quality, with moderate to serious risk of bias. Severe complications were similar between the groups: PMRT versus no radiotherapy (92 versus 141 patients respectively; odds ratio (OR) 2·35, 95 per cent c.i. 0·63 to 8·81, P = 0·200); neoadjuvant radiotherapy versus no radiotherapy (180 versus 392 patients; OR 1·24, 0·76 to 2·04, P = 0·390); and combined PMRT plus neoadjuvant radiotherapy versus no radiotherapy (272 versus 453 patients; OR 1·38, 0·83 to 2·32, P = 0·220). QOL and cosmetic studies used inconsistent methodologies. Conclusion Evidence is conflicting and study quality was poor, limiting recommendations for the timing of autologous BRR and radiotherapy. The impact of PMRT and neoadjuvant radiotherapy appeared to be similar.


Introduction
Breast cancer is the commonest malignancy and leading cause of cancer-related mortality in women 1,2 . free flap and implant-based procedures are the approaches used most frequently in immediate breast reconstruction (BRR) 5 . Autologous BRR has the inherent advantage of using the patient's own tissues, taken from a different part of the body where there is excess fat and skin, to restore breast volume and appearance after mastectomy. Various donor sites can be used, most commonly the abdomen 6 .
Adjuvant locoregional postmastectomy radiotherapy (PMRT) of the chest wall, and potentially of the regional lymph nodes, has been indicated historically for locally advanced disease 7,8 . These indications increased following the Early Breast Cancer Trialists' Collaborative Group 9 meta-analyses, which showed significantly improved disease-free and overall survival after PMRT and regional node irradiation in women at intermediate risk (tumour size 50 mm or less and 1-3 positive lymph nodes) 10 . Newly proposed US guidelines 11 emphasize the need to consider the lower recurrence rates associated with contemporary practice and the benefits of systemic therapy 12 . Current recommendations for PMRT in the intermediate-risk group remain controversial, pending the results of the SUPREMO (Selective Use of Postoperative Radiotherapy aftEr MastectOmy) trial, evaluating chest wall and/or axillary radiotherapy 13,14 .
Adjuvant radiotherapy (PMRT) may have deleterious effects on breast cosmetic outcomes, quality of life (QOL) and surgical complications after immediate BRR 15 . Previous studies evaluating the impact of PMRT on types of immediate BRR showed its potential feasibility in this setting, with lower morbidity rates compared with those of implant-based procedures 5,16 -18 . Surprisingly, the rapid adoption of immediate implant-based reconstruction in about 70 per cent of women, compared with 34 per cent of autologous procedures when PMRT is recommended, may be influenced by surgeon and patient preferences, regardless of current evidence 15,17,19 .
Increasing recommendations for PMRT and immediate BRR have prompted a need to consider their optimal sequence. Previous systematic reviews have not provided clarity concerning the choice between immediate and delayed BRR 9 . Despite this, immediate autologous BRR is commonly recommended in the setting of PMRT, given the potential long-term benefits on patients' QOL and breast cosmetic satisfaction 20,21 . Currently, immediate autologous BRR and PMRT recommendations are variable 22,23 . A systematic review 24 in 2011 showed methodological variations in the definitions of surgical complications, precluding interstudy comparisons.
Complications of autologous breast reconstruction with PMRT include: poor wound-healing, flap-related fat necrosis, fibrosis and contracture, which reduce breast volume 5 . Surgical complications contribute variably to decreased patient satisfaction and impaired cosmetic outcomes 5 . A standardized core set of outcomes for BRR has been proposed 25 involving a range of complications, including flap-related complications and the need for further unplanned surgery. The BRR core outcome set has yet to recommend a standardized measurement tool for evaluating surgical complications. Most surgeons use the Clavien-Dindo classification (CDC) 26 . Patientreported QOL outcomes using validated BRR questionnaires, such as the BREAST-Q and the European Organisation for Research and Treatment of Cancer (EORTC) Quality-of-Life Questionnaire (QLQ)-BRECON23, are recommended to evaluate comparative effectiveness 20,27 -32 . This systematic review aimed to evaluate the quality and strengths of the current evidence regarding surgical complications in autologous abdominal flaps in the context of the receipt and timing of radiotherapy related to PMRT 5,6 and, less commonly, neoadjuvant radiotherapy, generally administered before skin-sparing mastectomy and immediate breast reconstruction 33 , including assessment of QOL 34 .

Methods
The protocol was registered and published on the Prospective Register of Systematic Reviews PROS-PERO (CRD42017077945) 35 . The authors adhered to the PRISMA statement 36 .

Search strategies
A comprehensive search of the MEDLINE (Ovid SP), Embase (Ovid SP), Google Scholar, Cochrane Controlled Register of Trials (CENTRAL), Science citation index databases and ClinicalTrials.gov (January 2000 to August 2018) was conducted, identifying the relevant studies. Combinations of Medical Subject Headings (MeSH) terms and free text were used, including Boolean logical operators for the search strategy. References of included articles were also screened for their relevance. The example of an Embase (Ovid SP) search strategy was adopted for other databases (Appendix S1, supporting information).

Identification and selection of studies
Database-related searches were entered into an EndNote™ X8 library (Clarivate Analytics, Philadelphia, Pennsylvania, USA). Study screening was performed independently in two stages by two investigators using prespecified screening criteria.
In stage 1, two authors independently screened titles and abstracts. Discrepancies were resolved by consensus with the senior author. Remaining doubts regarding an article resulted in a review of the complete publication.
In stage 2, full-text studies from stage 1 were screened independently for their eligibility by two reviewers. Discrepancies were resolved by consensus with a third reviewer. Authors of eligible studies were contacted (via e-mail) to reconcile any methodological issues or to provide more detailed information on data for individual types of autologous flap.

Study design
All primary human studies evaluating surgical complications for autologous free flap (microvascular) abdominal BRR in breast cancer and types of radiotherapy (PMRT, neoadjuvant and no radiotherapy) were included. Outcomes also included patient-reported QOL and cosmetic assessments. Radiotherapy groups were compared with a control or no radiotherapy group in comparative studies, compatible with immediate and delayed BRR. Commonly performed autologous abdominal flaps included: deep inferior epigastric perforator (DIEP), transverse rectus abdominis myocutaneous (TRAM) and the superficial inferior epigastric artery perforator (SIEA) 6 .
Clinical studies that involved at least 50 patients were included (RCTs, prospective and retrospective comparative observational studies, and case series).

Exclusion criteria
Review articles, conference abstracts, simulation studies and clinical studies in non-human subjects were not included, along with studies involving patients who received segmental or partial mastectomy, technical descriptions of operative repair with no outcome measures, BRR unrelated to breast cancer, implant-based reconstructions and other non-abdominal autologous flaps.

Risk of bias and quality of studies
Cochrane's ROBINS-I (Risk Of Bias In Non-randomised Studies -of Interventions) tool was used for comparative studies 37 . This comprises seven domains from which the risk of bias may be ascertained to produce an overall risk-of-bias score 37 . The Grading of Recommendations, Assessment, Development, and Evaluations (GRADE) tool 38 was used to evaluate the methodological quality of individual studies.

Study outcomes
Primary outcomes were surgical complications including: Clavien-Dindo classification (CDC) grades II and III 26 ; partial flap loss; total flap loss; fat necrosis (CDC grades, when reported) 39 ; number(s) of unplanned reoperations for surgical complications (excluding cosmetic revisions); and number(s) of total complications. A surgical complication was defined as an adverse, postoperative, surgery-related event that required additional treatment 16 . If CDC grades were not defined, the complications reported by the included studies were graded retrospectively according to the CDC by two independent authors; any discrepancy was discussed and agreed with the senior author.

Data extraction, collection and management
Two authors independently extracted data from full-text articles using a standard data form. Any discrepancies were resolved by consensus with a third reviewer. Reporting authors of original articles were contacted on up to two occasions relating to missing data or where additional information was required.
Data extraction included: first author, year of publication, study design, study setting, number of centres, duration of follow-up, study population and participant demographics (mean age, BMI, smoking, co-morbidities). Surgical complications were recorded using CDC: grades II-III 26 . Two authors reviewed eligible studies and classified each complication according to the CDC 26 if unreported. QOL and cosmetic outcomes were listed.

Statistical analysis
When two or more studies reported outcome data, these were pooled using Review Manager 5.3 software (The Cochrane Collaboration, The Nordic Cochrane Centre, Copenhagen, Denmark). Odds ratios with 95 per cent confidence intervals were used to evaluate dichotomous outcomes (surgical complications). Standard mean differences (with 95 per cent c.i.) were used for continuous outcomes between treatment groups. Rates of each complication (fat necrosis, partial and total flap loss, infection and wound complications (dehiscence and delayed wound healing)) were compared for PMRT (versus no radiotherapy) and neoadjuvant radiotherapy (versus no radiotherapy). Data were also pooled to provide an overall summary measure of combined radiotherapy (adjuvant and neoadjuvant) compared with no radiotherapy.
Heterogeneity between studies 51 was assessed in Review Manager 5.3 using the Higgins and Thompson I 2 statistic 52 . Levels of heterogeneity were defined as: low (I 2 less than 50 per cent), moderate (I 2 = 50-80 per cent) and high (I 2 above 80 per cent). A random-effects model was used for cohorts with heterogeneity (I 2 above 50 per cent) 53 . As heterogeneity was generally moderate or high, and outcome measures differed between studies, these were combined using the DerSimonian and Laird random-effects model. Results of meta-analyses are shown as forest plots. A sensitivity analysis was performed where possible, to evaluate whether outcomes differed when restricting the analysis exclusively to high-quality studies.
Clinically meaningful differences in QOL items/ questions or domain scores may vary depending on response shift, that is a change in the meaning of QOL scores over time 54 . This is relevant in longitudinal studies and may influence clinical significance, defined as greater than 5-point score differences for EORTC QLQ-C30 and QLQ-BR23 42,43,54 . Clinically meaningful differences are currently being evaluated using a number of methods such as qualitative interviews and using predefined clinical anchors 55 . Clinically meaningful differences in QOL should be differentiated from statistical significance 55 .
BREAST-Q findings have been compared with large population-derived normative data, facilitating clinically meaningful interpretation of data 56,57 .

Clinical outcomes (Tables 2 -5)
No study prospectively graded surgical complications according to an accepted classification such as CDC (fat necrosis, partial or total flap loss, infection and wound complications). One study 64 graded partial flap loss using a novel flap necrosis classification system, adapted from Kwok et al. 70 . Only 30 per cent of all surgical complications (30 of 99) reported across the 12 included studies were defined a priori. Values are *median and †mean. GRADE, Grading of Recommendation, Assessment, Development, and Evaluation (tool for grading the quality of evidence); ROBINS-I, Risk Of Bias In Non-randomised Studies -of Interventions (tool for assessing risk of bias); RT, radiotherapy; n.a., not applicable/available.

Assessment of heterogeneity and meta-analyses
Clinical outcomes within studies of PMRT versus no radiotherapy were homogeneous (I 2 values below 50 per cent). All remaining meta-analyses of outcomes were similar (neoadjuvant radiotherapy versus no radiotherapy, pooled PMRT and neoadjuvant radiotherapy versus no radiotherapy).

Quality of life
There was limited reporting of patient-reported QOL; outcomes were detailed in only two prospective studies 60,62 and one retrospective study 58 , with small patient numbers and short follow-ups for the PMRT groups 58,60,62 . A priori hypothesis-driven selection of QOL domains was absent from methods 58,60,62 , with no reporting of missing data or how this problem was tackled 34 . Three studies 58,60,62 used the BREAST-Q and one 60 used the breast cancer-specific questionnaire (EORTC QLQ-BR23) 42 . One small study 58 reported significantly better 'satisfaction with breast' (P = 0⋅008) after a median follow-up of 27⋅5 months for PMRT compared with 48⋅7 months for no radiotherapy (Table S2, supporting information). The moderate-quality comparative prospective study 60 found a significant adverse impact of PMRT on breast symptoms at 1 year (P < 0⋅001) compared with no radiotherapy (Table S2, supporting information).
The third study 62 evaluated serial QOL outcomes, concluding a significant impact of PMRT on QOL domains (BREAST-Q) at 1 and 2 years, despite the absence of a control group (no radiotherapy). Moreover, clinical significance was defined as P = 0⋅05, which may not account for multiple variables (Table S2, supporting information) 43,62 . Highly significant abdominal adverse effects in a small patient group (108 patients) may be unrelated to PMRT, but rather an indication of donor site morbidity. Interestingly, when evaluating the impact of neoadjuvant radiotherapy in a small non-comparative study 62 , significant time-related improvements in most QOL domains were observed, except lower physical well-being relating to the abdomen at 1 year (Table S3, supporting information).

Cosmetic outcomes
Three studies 58,61,63 evaluated PMRT and the effects on aesthetic outcomes (187 patients). There was no standardized evaluation of cosmetic outcomes, precluding meta-analyses. Studies lacked robust methodology.

Discussion
The mixture of underpowered observational studies included in this review were, in large part, lacking contemporaneous data to reflect current practice. Most were retrospective single-centre cohorts, demonstrating poor levels of clinical evidence (levels 3 and 4) with insufficient follow-up 11 .
A previous study 24 of over 40 000 women undergoing BRR in 134 studies found that only 20 per cent reported a priori surgical complications, as well as inconsistent interstudy definitions 24 . The present review found similar interstudy discrepancies, without uniform adoption of the CDC 26 . The present authors graded all reported surgical complications using the CDC. All surgical interventions were graded as CDC IIIa or IIIb, and surgical reoperations were differentiated according to whether they were for complications or cosmetic revisions. Some complications were not amenable to retrospective grading in three studies 64,66,67 . In one 66 , it was not possible to determine whether fat necrosis required surgical revision for each radiotherapy group (adjuvant or neoadjuvant), compared with no radiotherapy. A second 64 omitted individual abdominal complications relative to timings of radiotherapy, and the third 67 omitted overall numbers of complications. Reviewed studies also failed to define postoperative wound infections according to Centers for Disease Control and Prevention criteria 71 .
The IDEAL (Idea, Development, Exploration, Assessment, Long-term study) Collaboration describes key methodological criteria for robust prospective cohort studies 72 : studies should be powered on the effect size of primary outcomes evaluating interventions of interest. The Mastectomy and Breast Reconstruction Outcomes Collaborative (MROC) is a multicentre prospective cohort study that provides IDEAL level 2b evidence for clinical safety and satisfactory QOL outcomes in the evaluation of surgical complications in immediate autologous reconstructions with PMRT versus no radiotherapy (delayed BRR) in 11 US centres 17,60 . The MROC cohort data were excluded from this systematic review based on its reporting of group-related summative data for all types of autologous reconstruction, as opposed to individual abdominal donor sites.
Eligible studies in the present systematic review were significantly underpowered in comparison with the MROC study, which evaluated irradiated autologous BRR at 1 year (236 patients) and 2 years (199), and non-irradiated procedures at 1 year (1625) and 2 years (332). The MROC data showed no differences between radiotherapy and no radiotherapy groups in the rates of total complications (25⋅6 versus 28⋅3 per cent respectively), major complications (17⋅6 versus 22⋅9 per cent) or flap failure (1⋅0 versus 2⋅4 per cent) at 2 years after immediate autologous reconstruction 17 . Studies in the present review showed significantly lower rates of major complications after radiotherapy compared with the MROC results, suggesting suboptimal overall reporting of surgical complications in the reviewed studies 24 .
The retrospective grading of surgical complications in the two moderate-quality studies reported showed a rate of major complications (CDC grade IIIb) of 9 per cent (6 of 64) at 1 year, and 4⋅6 per cent (5 of 108) at 2 years 60,62 . These rates are also likely to reflect under-reporting compared with the MROC rates of 14⋅8 per cent (35 of 236) at 1 year and 17⋅6 per cent (35 of 199) at 2 years 17 . Despite its strengths, the MROC cohort is based on the review of complications from electronic patient records, potentially also underestimating true complication rates 17 .
One way to measure what matters to patients is to use patient-reported outcome measures (PROMs) to assess the effects of disease or treatment on symptoms, functioning and health-related QOL 34 . In this systematic review, PROMs were poorly reported and underpowered for overall small effect sizes of individual QOL domains 43 . Preliminary conclusions regarding statistical significance were not substantiated by adequate patient numbers, lack of a comparator group or prospectively defined time points for questionnaire collection 58 . Standardized and objective evaluations of cosmetic outcome have also remained elusive with emerging adoption of newer technologies such as the Vectra ® XT 58 . Robust study designs evaluating these innovations should be accompanied by surgery-and disease-specific questionnaires 34 .
Clear recommendations for the optimal timing of radiotherapy in relation to autologous BRR will remain elusive until information from high-quality systematic reviews forms part of shared preoperative decision-making 73 .
Adequately powered prospective studies and ongoing audits, to allow comparisons of postoperative radiotherapy with neoadjuvant radiotherapy, are warranted. Current evidence for irradiating autologous abdominal flaps remains weak, involving only two moderate-quality studies of the 12 included in this report. Future cohort studies should be designed and powered to take advantage of newly evolving study designs, such as multiple-cohort RCTs or trials within cohorts 74 . These designs permit collection of big data within registry or cohort platforms, and allow multiple synchronous randomized trials to be conducted in a cost-effective manner 74 .