This meta-analysis was carried out in accordance with the Cochrane Collaboration Handbook of Interventions Systematic Reviews[31]. The study followed the PRISMA reporting guidelines[32]. This review was prospectively registered in the PROSPERO database (Registration Number CRD42020219183).
Search strategy
We searched the following databases on 10 october 2020: Cochrane Central Register of Controlled Trials (2019, via the Cochrane Library), Central Register of Controlled Trials (CENTRAL, 2017 issue 7, via Wiley), MEDLINE 1966 to 25 July 2019, via Pubmed), and Embase (1988 to July 2019, via Elsevier). The electronic search was conducted with no language, publication year, or publication status restrictions. The selected search terms included (((early breast neoplasm) OR (early breast tumor) OR (early human mammary carcinoma) OR (early human mammary neoplasm) OR (early breast cancer)) AND ((whole-breast irradiation) OR (WBI) OR (APBI) OR (accelerated partial breast irradiation) OR (accelerated partial irradiation) OR (PBI) OR (partial breast irradiation) OR (interstitial brachytherapy) OR (multicatheter interstitial brachytherapy) OR (MIB) OR (balloon catheter brachytherapy) OR (intracavitary brachytherapy) OR (intraoperative radiotherapy) OR (IORT) OR (conformal external beam) OR (three-dimensional conformal radiotherapy) OR (3D-CRT) OR (intensity-modulated radiotherapy) OR (IMRT) OR (hypofraction irradiation) OR (hypofractionated radiotherapy) OR (hypofraction radiotherapy))) AND ((segmental mastectomy) OR (partial mastectomy) OR (limited resection mastectomy) OR (lumpectomy) OR (local excision mastectomy) OR (breast-conserving surgery)).
Study inclusion/exclusion criteria
After the literature search, we included CF-WBI, HF-WBI, and PBI as the interventional postoperative RT arms in patients with early breast cancer. We pooled the results of different PBI techniques, such as external-beam radiation, intraoperative RT using electrons or photons, as well brachytherapy, including single- or multi-catheter-based approaches. The patients had to have undergone breast-conserving surgery and to have histologically confirmed primary breast cancer, either invasive or in situ breast cancer. The boost was assessed by the physicians in each study according to the risk of recurrence. Adjuvant systemic treatments were allowed. The included trials were published after 1 January 2000, to include comparable modern technologies. We excluded trials that compared different HF-WBI radiotherapy. The FAST[8] and FAST FORWARD[33] studies shortened 3–5 weeks of treatment to just 1 week, to explore the efficacy and safety of single-week ultra-hypofractionation. We included the FAST study but excluded the FAST FORWARD study in our analysis, because of the lack of a control arm. The TARGIT group recently updated the long-term results for the post-pathology or delayed-treatment subgroup, and we therefore split the trial into an original pre-pathology group and a delayed group, to allow the most appropriate estimation of the desired comparison22,23. The START study included two studies, Start A and Start B. Data for the longest follow-up period were published jointly in 2013. According to the authors’ intention, we obtained data for the two studies from the joint paper9.
Data extraction
The data were extracted by two independent reviewers (Chen and Yang). A third reviewer resolved any disagreements (Chang). 16 studies with 23,418 patients were included. The outcome indicators in this study were the number of ipsilateral breast tumor recurrences (IBTR), distant metastases, total deaths, and non-breast cancer-related deaths recorded at the endpoints in each included study. IBTR included tumor-bed relapse and new tumors in the ipsilateral breast. Distant metastasis of breast cancer was defined as metastases at all sites outside the quadrant where the tumor was located. Studies evaluating acute and late side effects and cosmetic effects used different evaluation criteria. It was therefore not possible to classify these indicators, which will be examined in another systematic review.
Risk of bias assessment
Two review authors (Chen and Yang) independently evaluated all relevant clinical studies for methodological quality. Each review author performed this assessment using the Cochrane Collaboration’s Risk of Bias tool, which included quality of random allocation concealment, description of dropout and withdrawal, intention-to-treat analysis, and blinding procedures for treatment and outcome assessments[34]. A third reviewer resolved any disagreements by discussion (Chang).
Statistical analysis
Statistical analysis was performed using the GeMTC package in R (version 4.0.2). A node-splitting model[35] for two-category data among all three intervention arms based on a Bayesian approach was established. Odds ratios (OR) with 95% confidence intervals (CI) were calculated to determine the favored arm. Network consistency was analyzed by calculating the ratio of direct and indirect treatment effects within each comparison, with 95% CIs. If the 95% CI was >1.00 or < 1.00, the difference was considered significant (p<0.05). I2 was used to assess the risk of bias in each model. I2 < 50% means that there is a low risk of bias in the model.