The association of young age with local recurrence in women with early-stage breast cancer after breast-conserving therapy: a meta-analysis

The aim of this meta-analysis is to determine the relationship between young age and local recurrence in patients with early-stage breast cancer after breast-conserving therapy. Eligible studies were retrieved from various electronic databases. Among the 19 studies included, 14 studies were analyzed for 5-year local recurrence rate and 8 studies for 10-year local recurrence rate using random effects models. Both results showed that young patients were at higher risk of local recurrence compared to old patients (5-year: RR = 2.64, 95% CI (1.94–3.60); 10-year: RR = 2.37, 95% CI (1.57–3.58)). Harbord’s modified test showed the presence of publication bias in both 5- and 10-year local recurrence rates (P = 0.019 and P = 0.01, respectively). While the Trim and Fill analysis showed that the presence of publication bias did not affect the overall outcome of the 5-year local recurrence rate (RR = 2.21, 95% CI (1.62, 3.02)), it significantly affected the effect size of the 10-year local recurrence rate (RR = 1.47, 95% CI (0.96, 2.27)). Young age is a significant risk factor for local recurrence developed within 5 years of breast-conserving therapy in patients with early-stage breast cancer. Further high-quality studies are needed to elucidate the relationship between young age and the risk of local recurrence developed within 10 years.

lymphatic invasion. However, one report has also shown that local recurrence could still occur despite adequate surgical margin and radiotherapy doses 6 . Other factors, such as young age at diagnosis, could be risk factors for local recurrence in patients after breast-conserving therapy; the significance of these factors are still being debated [5][6][7][8] .
Young age is defined differently based on the epidemiological situation in different regions. For instance, young age is normally defined as age ≤ 35 years; however, in western countries, breast cancer patients younger than 40 are also considered young. The incidence of breast cancer in the United States is reportedly high in patients who are 85 years old, while the rate in patients younger than 35 years old is less than 5%. In contrast, the incidence of breast cancer increases gradually after the age of 30 and peaks at the age of 40 and 50 in the Mainland of China 9 . Similarly, in Taiwan China, patients are also younger when diagnosed with breast cancer; most are between the ages of 45 and 49 10 . As breast cancer patients in China are relatively young when compared with patients in western countries, the study of the impact of young age on the local recurrence of breast cancer after conserving therapy is of clinical significance in China.
Therefore, the present study aims to examine whether the risk of local recurrence is higher in young patients compared to old patients after breast-conserving therapy.

Methods
This study was carried out in accordance with the PRISMA guidelines.
Inclusion criteria. Participants. Subjects were eligible for inclusion if the following criteria were met: (1) subjects had primary breast cancer without both distant metastasis and other severe concurrent diseases diagnosed by clinical symptoms, physical signs, X-ray, pathology and cytology examinations; (2) underwent conserving surgery followed by standard treatment of whole breast radiation; and (3) had a minimum of 5 years follow up. There was no bias related to age, race, ethnicity, or nationality when selecting subjects.
Intervention. The group of younger patients was defined as the experimental group. The group of older patients was defined as control group. Young age was defined as less than 35 or 40 years old (depending on the study cited).
Outcome. The outcome of interest was the rate of local recurrence of breast cancer.
Exclusion criteria. Studies were excluded according to the following criteria: (1) studies without clear diagnosis, inclusion and exclusion criteria of subjects; (2) studies with all subjects under the age of 40; (3) subjects who underwent radio-or chemotherapies before conserving therapy; (4) studies with inaccurate or incomplete data that were unable to provide outcomes; and (5)   Literature selection. Articles were imported into EndNote software to record information such as volume and issue and the completeness of abstract. The accessibility of the articles was also checked. Irrelevant articles were excluded, and the remaining records were assessed. The articles were marked as 'to be included, ' 'pending, ' or 'to be excluded (with reasons)' . For pending articles, full-text was retrieved to select studies that met the selection criteria.
Data extraction. The following data were extracted from all of the included studies: (1) General information: Research title, authors, year of publication and published journal; and (2) Study characteristics: general information of the subjects, intervention methods, and baseline compatibility.
The processes of data selection, evaluation, and extraction were conducted by two investigators. Any discrepancies were solved by discussion or the assistance of a third investigator.
Outcome measurements. The rates of local recurrence at 5 or 10 years after breast-conserving therapy.
Quality assessment. The quality of the included studies was assessed using the Newcastle-Ottawa Scale (NOS). The NOS contains 8 items which are categorized into 3 broad perspectives: Selection (4 items); Comparability (1 item); and Exposure or Outcome (3 items). The NOS applies a "star" system to evaluate the quality of the included studies. A maximum of 1 star can be awarded for each item within the Selection and SCientifiC REPORTS | 7: 11058 | DOI:10.1038/s41598-017-10729-9 Exposure or Outcome categories, while a maximum of 2 stars can be assigned to Comparability. The study with the highest quality is awarded a maximum of nine stars. Statistical analysis. Meta-analysis was performed using Stata software version 13.1. For each included study, individual and pooled risk ratios (RR) and 95% confidence intervals (CI) were calculated. A chi-square (χ 2 ) test was used to test for statistical evidence of heterogeneity between different studies, and the degree of heterogeneity was presented inI 2 statistic. In the case of no heterogeneity (P > 0.1, I 2 ≤ 50%), data were analyzed by a fixed-effects model. If significant heterogeneity was detected between groups (P ≤ 0.1, I 2 > 50%), a random effects model was applied and sources of heterogeneity were evaluated through subgroup analysis. Sensitivity analysis was performed by omitting one study at a time to assess the consistency of the overall effect size.
The existence of publication bias was evaluated by the funnel plot. For the outcome measurement, the pooled effect size was used as the central axis to draw a line which intersected with x-axis vertically. The dots distributed to the left side of the axis represented the effect size that was smaller than the pooled effect size, and the dots on the other side represented the effect size that was larger than the pooled effect size. A symmetrical funnel represents the absence of bias; otherwise, publication bias exists. Asymmetry in the funnel plot was further assessed using Harbord's modified test and Trim and Fill analysis. P value of < 0.05 was considered to be significant.

Results
Study selection. The search strategy yielded a total of 329 articles; of these, 25 were duplicates. The remaining 304 articles were evaluated by 2 investigators. Based on the selection criteria of subjects, intervention methods, type of research, and screening of the titles and abstracts, 272 articles were excluded, 15 were included, and 17 were listed as pending. After a full-text review of the 32 included articles and those in the pending lists, 13 articles were further eliminated. A total of 19 articles were ultimately included in the meta-analysis 7, 11-27 . All articles were in English (Fig. 1). Quality assessment. The included studies were assessed for risk of bias using NOS. The quality of the included studies ranges between 6-8 stars (Table 1). There were 3 studies with a score of 6 stars 15, 22, 25 , 11 with 7 stars 5, 7, 11, 12, 14, 16-18, 20, 21, 24 , and 5 with 8 stars 13,19,23,26,27 . Overall, the qualities of the included studies were high.

Correlation between age and local recurrence in women with early-stage breast cancer after breast-conserving surgery. Comparison of 5-year local recurrence rate between young and old
patients. The experimental group represented young breast cancer patients who had undergone conserving therapy (young group) and the control group comprised of patients who were older than 35, 40, or 45 years (old group). Based on the 14 included studies, there were 1,285 patients in the young group and 11,830 patients in the old group 5, 11-15, 17, 18, 20, 21, 23, 25, 26 . There were 166 patients in the young group and 688 patients in the old group who experienced local occurrence within 5 years of follow-up. Heterogeneity was observed among the studies (P < 0.0001, I 2 = 66.7%) and therefore, a random-effects model was employed to analyze the data. Our results showed a significant difference in the 5-year local recurrence rate between young and old patients (RR = 2.64, 95% CI (1.94-3.60)) ( Fig. 2).
Comparison of 10-year local recurrence rate between the young and old patients. From the 8 included studies 7,16,18,19,22,24,25,27 , there were 812 and 5059 patients in the young and old groups, respectively. Of these, 186 young patients and 642 old patients experienced local occurrence during the 10 years of follow up. As significant heterogeneity was observed among the studies (P < 0.0001, I 2 = 84.8%), a random-effects model was employed to analyze the data. Our results indicated a significant difference in the 10-year local recurrence rate between the young and old patients (RR = 2.37, 95% CI (1.57-3.58) (Fig. 3).

Cohort selection
Comparability Outcome Score Representativeness of the exposed cohort Selection of nonexposed cohort  5).     (Fig. 7).  Sensitivity Analysis. Sensitivity analysis is a method to test the robustness of our findings. Studies are removed one at a time and pooled RR is recalculated to determine if the overall effect size is dependent on one certain study. Our sensitivity analyses for 5-and 10-year local recurrences showed that no individual studies significantly affected the pooled RRs (Figs 8 and 9, respectively), indicating the consistency in our results.

Publication bias. Publication bias was assessed by funnel plot analysis. Our result comparing the 5-year local
recurrence rate between the young and old groups showed an asymmetrical funnel (Fig. 10) and Harbord's modified test showed significant evidence of publication bias among studies (P = 0.019). By trimming and inputting 3 studies using the Trim and Fill method, the recalculated pooled RR for the 5-year local recurrence rate was 2.21, 95% CI (1.62, 3.02) (Fig. 11), which was not significantly changed from the initial estimate (RR = 2.64, 95% CI (1.94, 3.60)). Therefore, the presence of publication bias has no significant effect on the overall finding. Because only 8 of the available studies analyzed the 10-year local recurrence rate, a funnel plot analysis was not suitable for this outcome measure. Nevertheless, Harbord's modified test showed the presence of publication bias (P = 0.01). Using the Trim and Fill method and adjusting for 4 studies, the recalculated pooled RR for the  10-year local recurrence rate was 1.47, 95% CI (0.96, 2.27) (Fig. 12), which was significantly changed from the initial estimate (RR = 2.37, 95% CI (1.57, 3.58)). Our data suggests that there is bias in our overall finding for the 10-year local recurrence rate.

Discussion
Our meta-analysis aims to determine whether young age is a risk factor for local recurrence in patients with early-stage breast cancer after breast-conserving therapy. We searched relevant studies published from 1966 to 2016 in databases including PubMed, EMBASE, Chinese Journal Full-text Database (CNKI), Chinese Biomedical Literature Database (CBM), VIP Chinese scientific and technical journals database (VIP), and Wanfang. Of the   329 studies found, 19 studies met our selection criteria 5,7,11,[13][14][15][16][17][18][19][20][21][22][23][24][25][26][27] . Our meta-analysis showed that young breast cancer patients are at high risk of local recurrence within 5 and 10 years after breast-conserving therapy compared to old patients (RR = 2.64, 95% CI (1.94, 3.60) and RR = 2.37, 95% CI (1.57, 3.58) respectively). Funnel plot and Harbord's modified test showed the presence of publication bias in the analysis of 5-year local recurrence rate. Trim and Fill method showed that the impact of publication bias was within an acceptable range and did not change the overall conclusions. Publication bias also existed in the analysis of 10-year local recurrence rate. The effect size became insignificant after Trim and Fill analysis (RR = 1.47, 95% CI (0.96, 2.27). As there were only 8 studies available for the 10-year local recurrence rate analysis (relatively small sample size), Trim and Fill analysis could be inaccurate and may lead to false negative results. Our result indicates that the meta-analysis for the risk of local recurrence within 10 years of post-breast-conserving therapy is currently inconclusive and further studies are warranted.
Subgroup analyses by age definitions showed that the risk for 5-year local recurrence was significantly increased in all ≤ 35, ≤ 40, and ≤ 45 years old subgroups, while the risk for 10-year local recurrence was only significantly increased in the ≤ 35 and ≤ 40 years old subgroups. The ≤ 45 years old subgroup for the analysis of 10-year local recurrence rate showed no significant difference between the young and old groups (RR = 1.03, 95% CI (0.80, 1.31)). When young was defined as age ≤ 45, the young patients in the study by Elkhuizen et al 18 . were relatively older than those in the other studies, which could lead to the insignificant result when compared to the old group. The insignificant difference observed when the age definition for the young was increased also indicates that age is a determinant for local recurrence risk. In addition, this also shows that the different definitions for young age are a source of heterogeneity in our study.
We also performed subgroup analyses on the regions where the studies were conducted, which could be categorized into the United States, Europe, and others. Our analyses showed that young age is consistently a risk factor for local recurrence in the United States and Europe. However, in the subgroup of "others, " results showed no significant difference in the 5-year local recurrence rate between the young and old patients. The pooled RR for the insignificant 5-year local recurrence result was from two studies: Ohsumi et al. reported a significant difference in the 5-year local recurrence rate between young and old patients in Japan 23 , while Burke et al. found no statistical difference between the young group and the old group in Australia 15 . Notably, in the study by Burke et al., the sample size for the young patients was relatively small with no incidence of local recurrence among all 45 patients (≤35 years old) 15 . Therefore, more studies are needed to validate whether young age is a risk factor for local recurrence developed within 5-year of post-therapy in regions other than the United States and Europe.
Our findings suggest that young age is a risk factor for local recurrence of breast cancer after breast-conserving therapy, which is consistent with the multivariate analysis conducted by De Bock and others in three trials in 2008 28 . Young patients may have some characteristics that promote local recurrence. A review by Anders et al., suggested that there is a higher percentage of ER/PR-negative tumor, HER2-positive tumor, and triple-negative tumor in young breast cancer patients compared to the old patients 29 . There were also studies which demonstrated that breast cancer patients with HER2 type tumor (ER/PR−, HER2+) and basal type tumor (triple negative, ER/PR−, HER2−) displayed increased risk for local recurrence [30][31][32] . Therefore, compared to older patients, the tumor subtypes susceptible to local recurrence are relatively higher in young breast cancer patients. Besides that, breast cancer in young patients is also more likely to be caused by familial mutations 33 . Hereditary familial breast cancer tends to have an early onset, particularly in those with BRCA1/BRCA2 mutations 33 . This type of breast cancer is also more aggressive, and patients tend to relapse after therapies.
This study has a few limitations. Studies included in this meta-analysis are published studies. The results that achieved statistical significance have a greater chance to be reported or published when compared to those results that are invalid or failed to achieve statistical significance. This phenomenon will exaggerate the effect of the group tested. We also did not analyze the effects of other age-related factors, such as the levels of estrogen, progesterone, progesterone receptor, and HER2 receptor, on local recurrence. These factors could be included in future studies for a more convincing result.
With the increasing rate of breast cancer in young patients, especially in Asia, and the significantly higher rate of local recurrences in young patients compared to old patients, young patients should be treated more cautiously. Young patients should be given a variety of auxiliary treatments, which include standardized radiotherapy, chemotherapy, hormonal therapy, and biological therapy after conserving surgery. In addition, more studies should aim to reveal the mechanisms underlying the pathogenesis of early-stage breast cancer in young patients and to improve personalized treatment. Our study found a strong association between young age and 5-year local recurrence rate in patients with early-stage breast cancer after breast-conserving therapy. Future high-quality studies are needed to determine the association of young age with the risk of local recurrence developed within 10 years after breast-conserving surgery.
In the subgroup analysis of regional differences (Fig. 6), we showed that young women in the United States and Europe are more prone to local recurrence within 5 years of breast-conserving therapy than older women when compared to women from a collection of "other" regions, which included Brazil, Japan, and Australia. However, these data do not directly imply that young women from United States and Europe have higher risk of local recurrence when compared to young women from other continents. One reason we obtained statistical significance in the women from United States and Europe but not in the women from other regions is that the present meta-analysis includes more studies from United States and Europe than other regions. When analyzing the local recurrence within 5 years of breast-conserving therapy, the present meta-analysis includes 6 studies from the United States and 6 studies from Europe, but only 2 studies from others regions. Therefore, it is invalid to claim that young women from other continents have lower risk of 5-year local recurrence than women from United States and Europe. The data suggest that more high-quality studies from other countries/regions are needed to confirm whether the higher risk of local recurrence exists in young women from different geographical regions and ethnicities.