Criteria for considering studies for this review
Types of studies
A standard procedure of autologous fat grafting has not been developed. Technique variants make the assessment more complex. In this review, autologous fat grafting is defined as the reconstructive or cosmetic procedure that contains the basic elements of fat grafting, that is fat harvesting, fat processing or preparing, and fat injection or transplantation. The detailed procedure of fat grafting in each study will be recorded.
We will first search for randomised controlled trials (RCTs). The RCTs will involve studies that evaluate the safety and efficacy of autologous fat grafting in operable primary breast cancer. However, for this review, we do not expect any existing RCTs. Therefore, we will extend our criteria to include non-randomised studies (NRSs) that assess the safety and efficacy of autologous fat grafting. In NRSs, women with operable primary breast cancer are allocated into an intervention group or control group based on their own preferences on whether or not to receive autologous fat grafting.
Selection bias, caused by a lack of a randomised allocation mechanism in NRSs, is a key difference between RCTs and NRSs. In our review, we will require that both groups have the same settings, that is, individuals are recruited in the study accept treatments in the same medical centre during the same period. We will only include studies with balanced baseline characteristics. Baseline characteristics mainly include age at diagnosis, menopausal status, tumour stage, hormone receptor status (including oestrogen receptor, progesterone receptor, ki-67%, Her2/neu) and systemic therapy received. In NRSs, only the generation of hypotheses are prospective. Considering these study features, cohort studies may be the most suitable NRSs. No restrictions will be placed on whether the studies are prospective or retrospective. Prospective cohort studies recruit participants before any intervention and follows them into the future. Retrospective cohort studies identify participants from past records describing the interventions received and follows them from the time of those records.
Types of participants
We will include women with pathologically diagnosed operable primary breast cancer. Operable primary breast cancer is defined as the primary breast tumour not fixed to underlying structures (includes TNM classification T1-3 and T4b where there is only minor skin involvement, N0-1, mobile lymph node (UICC 1987)).
No restrictions will be placed on race, age, menopausal status or hormone receptor status.
No restrictions will be placed on the conservative treatment including radiotherapy, neoadjuvant chemotherapy or hormonal therapy.
Participants will be allocated into the intervention (fat grafting) group or control (no fat grafting) group. Criteria for grouping will be explained in detail under the 'Types of interventions' section.
Types of interventions
In RCTs, participants are randomly allocated into the intervention group or control group. In NRSs, women who have undergone breast cancer surgery and autologous fat grafting are allocated into the intervention group. Those who have undergone breast cancer surgery only are allocated into the control group. In this review, autologous fat grafting will be the only option for breast reconstruction.
We will exclude women who have autologous fat grafting combined with other plastic methods (i.e. implant-based breast reconstruction, deep inferior epigastric perforator flap, transverse abdominal musculocutaneous flap) due to the potential interference when assessing safety and efficacy.
Types of outcome measures
Primary outcomes assessing the safety of autologous fat grafting will be the local recurrence rate and post-operative morbidity.
Post-operative morbidity: post-operative morbidity is defined as the rate of surgical complications including infections, formation of necrosis cysts, calcification of transplanted adipose tissues, bruising, temporary swelling and others.
Local recurrence rate: local recurrence is defined as breast cancer recurrence in the chest wall, axillary, supraclavicular, infraclavicular or internal mammary lymph nodes. Local recurrence will be considered as a time-to-event outcome. If time-to-event data are unavailable, 5- and 10-year follow-up data will be collected if available.
Outcomes concerning the safety of autologous fat grafting will also be overall survival, disease-free survival, metastasis rate and metastasis-free survival.
Outcomes concerning the efficacy of autologous fat grafting will be quality of life and satisfaction of the reconstructed breast, cancer surveillance and radiological sequelae, and cost-utility analysis.
Overall survival, defined as the interval from the date of breast cancer diagnosis to the date of death from any cause.
Disease-free survival, defined as the interval from the primary cancer treatment to the date of first cancer recurrence.
Metastasis rate and metastasis-free survival: metastasis rate is defined as the percentage of women with metastasis among all women with breast cancer. Metastasis-free survival is defined as the interval from the date of breast cancer diagnosis to the date of first metastasis. These outcomes will be considered as time-to-event outcomes. If time-to-event data are unavailable, 5- and 10-year follow-up data will be collected if available.
Mortality rates, defined as the percentage of women who died from breast cancer and treatment procedures. Mortality will be considered as a time-to-event outcome. If time-to-event data are unavailable, 5- and 10-year follow-up data will be collected if available.
Quality of life and satisfaction of the reconstructed breast: the European Organisation for Research and Treatment of Cancer (EORTC) Breast Cancer-Specific Quality of Life Questionnaire (QLQ-BR23) and the BREAST-Q will be used to assess the cancer survivor's quality of life and satisfaction of the reconstructed breast. The QLQ-BR23 includes the body image, sexual functioning, arm symptoms, breast symptoms, systematic therapy side effects, sexual enjoyment and hair loss scale. The BREAST-Q includes seven domains and these are satisfaction with the breast, satisfaction with the outcome, psychosocial well-being, sexual well-being, satisfaction with the process of care and additional scales.
Cancer surveillance and radiological sequelae: fat grafting may cause fat necrosis and calcification that may interfere with cancer surveillance. We will assess burden brought by fat grafting during cancer surveillance and reconstruction follow-up by extracting the frequency of visits to clinics, imaging examinations (including mammography, ultrasound, computer tomography, magnetic resonance imaging and others) and biopsies during cancer surveillance.
Cost-utility analysis will be used for an economic evaluation and quality-adjusted life year (QALY) will be used to measure utility.
Search methods for identification of studies
We will search the following databases with key words such as fat transplantation, fat grafting, lipo transfer, breast cancer and breast neoplasm.
The Cochrane Breast Cancer Group's Specialised Register. The Cochrane Breast Cancer Group will search their Specialised Register. Details of the search strategies used by the Group for the identification of studies and the procedure used to code references are outlined in the Group's module (http://www.mrw.interscience.wiley.com/cochrane/clabout/articles/BREASTCA/frame.html). We will extract trials with the key words or text words: 'Early Breast Cancer' 'Breast Cancer Unspecified', 'Surgery', 'Mammoplasty', 'Adipose Tissue', 'Autologous Fat Graft', 'Autologous Fat Grafting', 'Postmastectomy Autologous Fat Graft', 'Postmastectomy Autologous Fat Grafting', 'Autologous Fat Graft for Breast Reconstruction', 'Lipo Transfer' 'Mammaplasty', 'Breast Reconstruction', 'Autogenous Fat Graft', 'Autologous Fat Transplant', 'Autologous Fat Transplantation', 'Postmastectomy Autologous Fat Transplant', 'Postmastectomy Autologous Fat Transplantation', 'Autogenous Fat Transplantation' and 'Lipofilling'.
MEDLINE via OvidSP. Refer to the search strategy in Appendix 1.
EMBASE via EMBASE.com. Refer to the search strategy in Appendix 2.
The WHO International Clinical Trials Registry Platform (ICTRP) search portal (http://apps.who.int/trialsearch/Default.aspx) for all prospectively registered and ongoing trials. Refer to the search strategy in Appendix 3.
ClinicalTrials.gov (http://clinicaltrials.gov/ct2/search). Refer to the search strategy in Appendix 4.
The Cochrane Central Register of Controlled Trials (CENTRAL) (The Cochrane Library, current issue). Refer to the search strategy in Appendix 5.
Searching other resources
We will make attempts to identify further studies from the reference lists of identified relevant studies or reviews. A copy of the full text for each reference reporting potentially eligible studies will be obtained. Where this is not possible, attempts will be made to contact investigators for more information.
We will search conference abstracts from St. Gallen Oncology Conferences, Milan Breast Cancer Conference (Milan, Italy) and San Antonio Breast Cancer Symposium conference proceedings.
We will apply no restrictions on language or publication date during searching.
Data collection and analysis
Selection of studies
Studies evaluating autologous fat grafting in women with operable primary breast cancer may be eligible.
Two authors (JC, QWT) will separately assess each study. For RCTs, the inclusion criteria (details under the 'Types of studies' section) will first be used when screening the titles and abstracts. The full-text article will be obtained and assessed for all potentially eligible studies. We will correspond with investigators to clarify study eligibility, if necessary. We will apply similar procedures described above when screening NRSs. However, for most NRSs, the title and abstracts may not provide us with detailed information on study features and therefore full-text articles may be assessed for study selection.
Multiple publications of the one study will be linked by author information and study design details such as the intervention, number of the participants at baseline data, and follow-up protocol.
Any disagreements about study eligibility will be resolved by discussion among the three authors (JC, QL, QWT).
Excluded studies will be defined as studies that may appear to meet the eligibility criteria but on further inspection do not (Higgins 2011). The reason for excluding a study will be noted in the 'characteristics of excluded studies' table.
Data extraction and management
We will extract relevant data from eligible studies using pre-standardised data extraction forms. The data extraction form will mainly contain the following items: title of the form, revision date of data extraction form, name of author(s) completing the form, study/report identification, assessment of the eligibility of the study, details of the study design (including participants, intervention, comparison and outcome), details of the fat grafting techniques, and results. Pilot testing will be conducted twice to identify data that are missing from the form with sample studies. Consensus between the authors will be achieved before the form is modified. We will use two separate data extraction forms for RCTs and NRSs.
We will extract the following information.
Basic study information: authors and year of publication
Eligibility: For RCTs, we will extract the methods for randomisation and allocation concealment. For NRSs, we will assess the allocation mechanism and the comparability between groups. As NRSs have selection bias, similar baseline characteristics between the two groups will be essential. We will exclude studies without baseline comparability between groups. Items will be recorded as 'Yes', 'No' or 'Can't tell'. Reasons for excluding a study will be recorded in detail under the 'characteristic of the excluded studies' table.
Methods: total study duration, sequence generation, allocation sequence concealment, blinding, measurement techniques for each outcome, and other concerns about bias.
Participants: sample size of the two groups, age at diagnosis, menopausal status, country (race), diagnostic criteria of breast cancer, status of breast cancer at diagnosis, and type of the surgery undertaken.
Intervention: total number of participants in the intervention group and details of the intervention especially the method of autologous fat grafting.
Detailed fat injection method: as there is controversy concerning the fat grafting method, we will extract details from the trial protocol. This information will include: fat harvesting (donor site, tumescent technique, size of the syringe, harvested volume), fat processing (centrifugation or not, the duration and revolution per minute if centrifugation is applied), fat injection (injection instrumentation, injection planes and the injection volume) and others (for studies applying fat grafting in other forms, the detailed information will be recorded).
Outcomes and results: local recurrence rate of breast cancer, post-operative morbidity and mortality, overall survival, disease-free survival, metastasis rate, metastasis survival, quality of life and satisfaction of the breast, cancer surveillance and radiological sequelae, the number of missing participants in each group, summary data for each group, the scoring systems used and other outcomes documented in the study publications.
Others: funding source of the primary study, key conclusion of the study, comments from the investigators and comments from the review authors. Space will be left to extract sufficient details to allow an assessment of the risk of bias. Risk of bias assessment result will be contained in the 'Risk of bias tables'.
Two authors (QL, QWT) will extract data separately. The two data extraction forms will be checked by a third author (JC). Any disagreements will be resolved by discussion or by contacting the investigators, if necessary. For studies with more than one report, we will extract and compare data from all of the reports. If all the reported data are exactly identical, only one copy of data will be kept. If any discrepancies exist among the reported data, we will contact investigators for the original data.
For dichotomous outcomes, we will extract information on the sample size and number of participants who have positive outcomes from the primary study. For continuous outcomes, we will extract data on the number of participants, means and standard deviations (SDs) from the primary study. As we will be including NRSs in our review, the unadjusted effect estimates, adjusted effect estimates and the confounders that were adjusted for will be recorded. If more than one adjusted effect estimate is recorded, we will use the most suitable one based on the recommendation in the Cochrane Handbook for Systemic Review of Interventions (Higgins 2011).
Assessment of risk of bias in included studies
Risk of bias includes selection bias, performance bias, detection bias, attrition bias, reporting bias and other bias.
Selection bias is defined as the systemic difference between the baseline characteristics of the compared groups, and randomisation is the major solution for selection bias. For NRSs, the allocation mechanisms may bring more selection bias. Adequate assessment of selection bias for NRSs will be essential. Performance bias is defined as the systemic difference between groups that receive different care. Detection bias is defined as the systemic difference between groups in which the results are detected and interpreted (Higgins 2011). Blinding is designed for performance bias and detection bias. Attrition bias is defined as the systemic difference between groups that exclude participants or deposit outcome data from the study. In studies with comparisons, the evaluation of performance, detection, attrition and reporting biases are required (Higgins 2011).
Assessment of risk of bias in included RCTs
Two authors (JC, CXT) will assess the risk of bias of the included RCTs in terms of the sequence generation, allocation concealment, blinding of participants, blinding of outcome assessment, incomplete outcome data, selective outcome reporting and reporting bias based on the recommendations in theCochrane Handbook for Systemic Reviews of Interventions (Higgins 2011). Detailed methods for assessing the reporting bias in RCTs will be documented under the "Assessment of reporting biases" section. Each domain will be judged as 'low risk of bias', 'high risk of bias' or 'unclear risk', according to the Cochrane Collaboration's tool for assessing risk of bias (Higgins 2011).
Assessment of risk of bias in included NRSs
Two authors (JC, CXT) will assess the methodological quality of eligible NRSs using the customised Newcastle-Ottawa Scale as described in the Cochrane Handbook for Systemic Reviews of Interventions (Higgins 2011) (Appendix 6). The Newcastle-Ottawa Scale is a convenient tool for quality assessments of NRSs. It is judged on three aspects and these include the selection of the study groups, the comparability between groups and the ascertainment of exposure (Wells 2008). In addition to the customised Newcastle-Ottawa Scale, we will assess selective outcome reporting of each study. Detailed methods for assessing the reporting bias in NRSs will be documented under the "Assessment of reporting biases" section. During the assessment, the review authors will be blinded to the names of the trial authors, institutions, journal and results of the study. Missing information will be collected by contacting the trial investigators. Each bias will be judged using the following categories: "low risk" refers to a low risk of bias, "unclear risk" refers to an uncertain risk of bias and "high risk" refers to a high risk of bias. Independently, two authors (QWT, CXT) will assess the eligible studies. Consensus will be reached by discussion amongst the authors (JC, QWT, CXT).
Measures of treatment effect
We will include dichotomous, continuous and time-to-event data.
We will express dichotomous outcomes (including post-operative morbidity) as risk ratios (RRs) and 95% confidence intervals (CIs). We will interpret the RR as the following: RR > 1: the incidence in the intervention group is greater than in the control group; RR = 1: no statistically significant difference between the intervention group and control group; and RR < 1: the incidence in the intervention group is lower than in the control group.
For continuous outcomes (including quality of life, satisfaction of the breast and cost-effectiveness), we will use the standardised mean difference (SMD). The interpretation of the SMD will involve: SMD > 0, the intervention is beneficial compared to the control; and SMD<0, the intervention is inferior to the control.
We will assess time-to-event outcomes (including local recurrence, overall survival, disease-free survival) in the form of the hazard ratio (HR) and their standard errors.
Unit of analysis issues
We will be including women who have had autologous fat grafting post-surgery in one or both breasts. Therefore the unit of analysis will be the individual participant.
Dealing with missing data
We will contact authors by e-mail or telephone for missing data. For missing data that we are unable to retrieve by e-mail or telephone, we will perform an intention-to-treat analysis and sensitivity analysis.
Assessment of heterogeneity
As we will include NRSs, assessment of the homogeneity of the included studies is essential. We will assess clinical heterogeneity by checking the characteristics of the studies, type of participants, the interventions and outcomes. We will assess the statistical heterogeneity by the Chi-squared (Chi2) test with significance set at P value of < 0.01, and the I2 statistic (Higgins 2011). We will interpret the I2 as follows: 0% to 40%, might not be important; 30% to 60%, may represent moderate heterogeneity; 50% to 90%, may represent substantial heterogeneity; and 75% to 100%, considerable heterogeneity.
If there is no evidence of heterogeneity, we will use the fixed-effect model in the analysis. Otherwise, we will make a cautious attempt to reveal the reason for heterogeneity and apply the random-effects model. If sufficient data are available, we will use forest plots to show the original results of the included studies and the variation in the results of the studies.
Assessment of reporting biases
Reporting bias can be classified into publication bias, time lag bias, multiple publication bias, location bias, citation bias, language bias and outcome bias. If there are sufficient included studies, we will assess the reporting bias of eligible studies by funnel plot asymmetry, following the recommendations of the Cochrane Handbook for Systematic Review of Interventions (Higgins 2011). Funnel plot asymmetry may be caused by selection bias, small-study effect, and true heterogeneity. We will address this issue in the discussion section of the review, if necessary.
Reporting bias can result from selective outcome reporting (SOR) or selective analysis reporting (SAR). The ways to minimise SOR and SAR are to register key details of the study. For RCTs, SOR will be assessed by comparing the study protocol and registration information with all relevant publications to determine if the protocol was finalised before any comparative data analysis, locking of the study database, and access to variables describing how groups were formed. SAR will be assessed by comparing the consistency of the methods and results sections in the protocol and all related publications.
For NRSs, we will identify and document SOR and SAR for each included study. SOR will be assessed by searching for all related publications for each included study. We will also compare the method and results sections of the publications for potential discrepancies in the definitions and measurement techniques for each outcome. SAR will be assessed by checking whether analyses were defined in the methods section of the study publications in sufficient detail to prevent SAR. We will also compare the method and results sections of publications for potential discrepancies in the analytical methods proposed and actually used.
In order to minimise reporting biases, we will place no restrictions on language. We will try to include conference abstracts and other published studies. Data from studies with multiple publications will be included once in the review.
In the event that both RCTs and NRSs exist, we will synthesise results from RCTs and NRSs separately. Two authors (QWT, JC) will analyse the extracted data based on recommendations in the Cochrane Handbook for Systematic Reviews of the Interventions (Higgins 2011). The adjusted effect estimates will be analysed using the inverse-variance method. The fixed-effect meta-analyses or random-effects meta-analyses will be applied based on the assessment of heterogeneity. We will perform all data syntheses and meta-analyses using the Cochrane Collaboration's statistical software, RevMan. We will report the meta-analysis results mainly by forest plots and summary of findings tables.
If the included studies are significantly clinically diverse or most of the included studies have a high risk of bias or serious reporting bias, we will not undertake a meta-analysis. In the case where a meta-analysis cannot be conducted, we will present a narrative synthesis of the included studies.
Subgroup analysis and investigation of heterogeneity
Where adequate studies and data are available, we will perform subgroup analysis. This will include grouping by primary breast cancer stage.
If heterogeneity is present during heterogeneity assessment, we will make an attempt to reveal the reason for heterogeneity and change the analysis method from the fixed-effect model to random-effects model.
If adequate data are available, we will perform sensitivity analysis by comparing results with or without studies of low quality. If we obtain similar results and conclusions during the two analyses, the results of the review will be regarded with a higher degree of certainty. However, if we obtain different results and conclusions, we will make attempts to resolve these uncertainties by checking the extracted data and contacting primary study investigators, if necessary. If consistency still cannot be achieved, we will interpret our result with an appropriate degree of caution.